linux_misc.c revision 1.134.4.1
1 /* $NetBSD: linux_misc.c,v 1.134.4.1 2005/04/29 11:28:40 kent 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 1.134.4.1 2005/04/29 11:28:40 kent 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 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC } 159 }; 160 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 161 162 #ifdef DEBUG_LINUX 163 #define DPRINTF(a) uprintf a 164 #else 165 #define DPRINTF(a) 166 #endif 167 168 /* Local linux_misc.c functions: */ 169 static void bsd_to_linux_statfs __P((const struct statvfs *, 170 struct linux_statfs *)); 171 static int linux_to_bsd_limit __P((int)); 172 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *, 173 const struct linux_sys_mmap_args *)); 174 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *, 175 register_t *, off_t)); 176 177 178 /* 179 * The information on a terminated (or stopped) process needs 180 * to be converted in order for Linux binaries to get a valid signal 181 * number out of it. 182 */ 183 void 184 bsd_to_linux_wstat(st) 185 int *st; 186 { 187 188 int sig; 189 190 if (WIFSIGNALED(*st)) { 191 sig = WTERMSIG(*st); 192 if (sig >= 0 && sig < NSIG) 193 *st= (*st& ~0177) | native_to_linux_signo[sig]; 194 } else if (WIFSTOPPED(*st)) { 195 sig = WSTOPSIG(*st); 196 if (sig >= 0 && sig < NSIG) 197 *st = (*st & ~0xff00) | 198 (native_to_linux_signo[sig] << 8); 199 } 200 } 201 202 /* 203 * wait4(2). Passed on to the NetBSD call, surrounded by code to 204 * reserve some space for a NetBSD-style wait status, and converting 205 * it to what Linux wants. 206 */ 207 int 208 linux_sys_wait4(l, v, retval) 209 struct lwp *l; 210 void *v; 211 register_t *retval; 212 { 213 struct linux_sys_wait4_args /* { 214 syscallarg(int) pid; 215 syscallarg(int *) status; 216 syscallarg(int) options; 217 syscallarg(struct rusage *) rusage; 218 } */ *uap = v; 219 struct proc *p = l->l_proc; 220 struct sys_wait4_args w4a; 221 int error, *status, tstat, options, linux_options; 222 caddr_t sg; 223 224 if (SCARG(uap, status) != NULL) { 225 sg = stackgap_init(p, 0); 226 status = (int *) stackgap_alloc(p, &sg, sizeof *status); 227 } else 228 status = NULL; 229 230 linux_options = SCARG(uap, options); 231 options = 0; 232 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS)) 233 return (EINVAL); 234 235 if (linux_options & LINUX_WAIT4_WNOHANG) 236 options |= WNOHANG; 237 if (linux_options & LINUX_WAIT4_WUNTRACED) 238 options |= WUNTRACED; 239 if (linux_options & LINUX_WAIT4_WALL) 240 options |= WALLSIG; 241 if (linux_options & LINUX_WAIT4_WCLONE) 242 options |= WALTSIG; 243 #ifdef DIAGNOSTIC 244 if (linux_options & LINUX_WAIT4_WNOTHREAD) 245 printf("WARNING: %s: linux process %d.%d called " 246 "waitpid with __WNOTHREAD set!", 247 __FILE__, p->p_pid, l->l_lid); 248 249 #endif 250 251 SCARG(&w4a, pid) = SCARG(uap, pid); 252 SCARG(&w4a, status) = status; 253 SCARG(&w4a, options) = options; 254 SCARG(&w4a, rusage) = SCARG(uap, rusage); 255 256 if ((error = sys_wait4(l, &w4a, retval))) 257 return error; 258 259 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD); 260 261 if (status != NULL) { 262 if ((error = copyin(status, &tstat, sizeof tstat))) 263 return error; 264 265 bsd_to_linux_wstat(&tstat); 266 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 267 } 268 269 return 0; 270 } 271 272 /* 273 * Linux brk(2). The check if the new address is >= the old one is 274 * done in the kernel in Linux. NetBSD does it in the library. 275 */ 276 int 277 linux_sys_brk(l, v, retval) 278 struct lwp *l; 279 void *v; 280 register_t *retval; 281 { 282 struct linux_sys_brk_args /* { 283 syscallarg(char *) nsize; 284 } */ *uap = v; 285 struct proc *p = l->l_proc; 286 char *nbrk = SCARG(uap, nsize); 287 struct sys_obreak_args oba; 288 struct vmspace *vm = p->p_vmspace; 289 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata; 290 291 SCARG(&oba, nsize) = nbrk; 292 293 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0) 294 ed->s->p_break = (char*)nbrk; 295 else 296 nbrk = ed->s->p_break; 297 298 retval[0] = (register_t)nbrk; 299 300 return 0; 301 } 302 303 /* 304 * Convert NetBSD statvfs structure to Linux statfs structure. 305 * Linux doesn't have f_flag, and we can't set f_frsize due 306 * to glibc statvfs() bug (see below). 307 */ 308 static void 309 bsd_to_linux_statfs(bsp, lsp) 310 const struct statvfs *bsp; 311 struct linux_statfs *lsp; 312 { 313 int i; 314 315 for (i = 0; i < linux_fstypes_cnt; i++) { 316 if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) { 317 lsp->l_ftype = linux_fstypes[i].linux; 318 break; 319 } 320 } 321 322 if (i == linux_fstypes_cnt) { 323 DPRINTF(("unhandled fstype in linux emulation: %s\n", 324 bsp->f_fstypename)); 325 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC; 326 } 327 328 /* 329 * The sizes are expressed in number of blocks. The block 330 * size used for the size is f_frsize for POSIX-compliant 331 * statvfs. Linux statfs uses f_bsize as the block size 332 * (f_frsize used to not be available in Linux struct statfs). 333 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block 334 * counts for different f_frsize if f_frsize is provided by the kernel. 335 * POSIX conforming apps thus get wrong size if f_frsize 336 * is different to f_bsize. Thus, we just pretend we don't 337 * support f_frsize. 338 */ 339 340 lsp->l_fbsize = bsp->f_frsize; 341 lsp->l_ffrsize = 0; /* compat */ 342 lsp->l_fblocks = bsp->f_blocks; 343 lsp->l_fbfree = bsp->f_bfree; 344 lsp->l_fbavail = bsp->f_bavail; 345 lsp->l_ffiles = bsp->f_files; 346 lsp->l_fffree = bsp->f_ffree; 347 /* Linux sets the fsid to 0..., we don't */ 348 lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0]; 349 lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1]; 350 lsp->l_fnamelen = bsp->f_namemax; 351 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare)); 352 } 353 354 /* 355 * Implement the fs stat functions. Straightforward. 356 */ 357 int 358 linux_sys_statfs(l, v, retval) 359 struct lwp *l; 360 void *v; 361 register_t *retval; 362 { 363 struct linux_sys_statfs_args /* { 364 syscallarg(const char *) path; 365 syscallarg(struct linux_statfs *) sp; 366 } */ *uap = v; 367 struct proc *p = l->l_proc; 368 struct statvfs btmp, *bsp; 369 struct linux_statfs ltmp; 370 struct sys_statvfs1_args bsa; 371 caddr_t sg; 372 int error; 373 374 sg = stackgap_init(p, 0); 375 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 376 377 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 378 379 SCARG(&bsa, path) = SCARG(uap, path); 380 SCARG(&bsa, buf) = bsp; 381 SCARG(&bsa, flags) = ST_WAIT; 382 383 if ((error = sys_statvfs1(l, &bsa, retval))) 384 return error; 385 386 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 387 return error; 388 389 bsd_to_linux_statfs(&btmp, <mp); 390 391 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 392 } 393 394 int 395 linux_sys_fstatfs(l, v, retval) 396 struct lwp *l; 397 void *v; 398 register_t *retval; 399 { 400 struct linux_sys_fstatfs_args /* { 401 syscallarg(int) fd; 402 syscallarg(struct linux_statfs *) sp; 403 } */ *uap = v; 404 struct proc *p = l->l_proc; 405 struct statvfs btmp, *bsp; 406 struct linux_statfs ltmp; 407 struct sys_fstatvfs1_args bsa; 408 caddr_t sg; 409 int error; 410 411 sg = stackgap_init(p, 0); 412 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 413 414 SCARG(&bsa, fd) = SCARG(uap, fd); 415 SCARG(&bsa, buf) = bsp; 416 SCARG(&bsa, flags) = ST_WAIT; 417 418 if ((error = sys_fstatvfs1(l, &bsa, retval))) 419 return error; 420 421 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 422 return error; 423 424 bsd_to_linux_statfs(&btmp, <mp); 425 426 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 427 } 428 429 /* 430 * uname(). Just copy the info from the various strings stored in the 431 * kernel, and put it in the Linux utsname structure. That structure 432 * is almost the same as the NetBSD one, only it has fields 65 characters 433 * long, and an extra domainname field. 434 */ 435 int 436 linux_sys_uname(l, v, retval) 437 struct lwp *l; 438 void *v; 439 register_t *retval; 440 { 441 struct linux_sys_uname_args /* { 442 syscallarg(struct linux_utsname *) up; 443 } */ *uap = v; 444 struct linux_utsname luts; 445 446 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 447 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 448 strncpy(luts.l_release, linux_release, sizeof(luts.l_release)); 449 strncpy(luts.l_version, linux_version, sizeof(luts.l_version)); 450 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 451 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 452 453 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 454 } 455 456 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 457 /* Used indirectly on: arm, i386, m68k */ 458 459 /* 460 * New type Linux mmap call. 461 * Only called directly on machines with >= 6 free regs. 462 */ 463 int 464 linux_sys_mmap(l, v, retval) 465 struct lwp *l; 466 void *v; 467 register_t *retval; 468 { 469 struct linux_sys_mmap_args /* { 470 syscallarg(unsigned long) addr; 471 syscallarg(size_t) len; 472 syscallarg(int) prot; 473 syscallarg(int) flags; 474 syscallarg(int) fd; 475 syscallarg(linux_off_t) offset; 476 } */ *uap = v; 477 478 if (SCARG(uap, offset) & PAGE_MASK) 479 return EINVAL; 480 481 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 482 } 483 484 /* 485 * Guts of most architectures' mmap64() implementations. This shares 486 * its list of arguments with linux_sys_mmap(). 487 * 488 * The difference in linux_sys_mmap2() is that "offset" is actually 489 * (offset / pagesize), not an absolute byte count. This translation 490 * to pagesize offsets is done inside glibc between the mmap64() call 491 * point, and the actual syscall. 492 */ 493 int 494 linux_sys_mmap2(l, v, retval) 495 struct lwp *l; 496 void *v; 497 register_t *retval; 498 { 499 struct linux_sys_mmap2_args /* { 500 syscallarg(unsigned long) addr; 501 syscallarg(size_t) len; 502 syscallarg(int) prot; 503 syscallarg(int) flags; 504 syscallarg(int) fd; 505 syscallarg(linux_off_t) offset; 506 } */ *uap = v; 507 508 return linux_mmap(l, uap, retval, 509 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 510 } 511 512 /* 513 * Massage arguments and call system mmap(2). 514 */ 515 static int 516 linux_mmap(l, uap, retval, offset) 517 struct lwp *l; 518 struct linux_sys_mmap_args *uap; 519 register_t *retval; 520 off_t offset; 521 { 522 struct sys_mmap_args cma; 523 int error; 524 size_t mmoff=0; 525 526 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 527 /* 528 * Request for stack-like memory segment. On linux, this 529 * works by mmap()ping (small) segment, which is automatically 530 * extended when page fault happens below the currently 531 * allocated area. We emulate this by allocating (typically 532 * bigger) segment sized at current stack size limit, and 533 * offsetting the requested and returned address accordingly. 534 * Since physical pages are only allocated on-demand, this 535 * is effectively identical. 536 */ 537 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 538 539 if (SCARG(uap, len) < ssl) { 540 /* Compute the address offset */ 541 mmoff = round_page(ssl) - SCARG(uap, len); 542 543 if (SCARG(uap, addr)) 544 SCARG(uap, addr) -= mmoff; 545 546 SCARG(uap, len) = (size_t) ssl; 547 } 548 } 549 550 linux_to_bsd_mmap_args(&cma, uap); 551 SCARG(&cma, pos) = offset; 552 553 error = sys_mmap(l, &cma, retval); 554 if (error) 555 return (error); 556 557 /* Shift the returned address for stack-like segment if necessary */ 558 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff) 559 retval[0] += mmoff; 560 561 return (0); 562 } 563 564 static void 565 linux_to_bsd_mmap_args(cma, uap) 566 struct sys_mmap_args *cma; 567 const struct linux_sys_mmap_args *uap; 568 { 569 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 570 571 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 572 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 573 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 574 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 575 /* XXX XAX ERH: Any other flags here? There are more defined... */ 576 577 SCARG(cma, addr) = (void *)SCARG(uap, addr); 578 SCARG(cma, len) = SCARG(uap, len); 579 SCARG(cma, prot) = SCARG(uap, prot); 580 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 581 SCARG(cma, prot) |= VM_PROT_READ; 582 SCARG(cma, flags) = flags; 583 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 584 SCARG(cma, pad) = 0; 585 } 586 587 int 588 linux_sys_mremap(l, v, retval) 589 struct lwp *l; 590 void *v; 591 register_t *retval; 592 { 593 struct linux_sys_mremap_args /* { 594 syscallarg(void *) old_address; 595 syscallarg(size_t) old_size; 596 syscallarg(size_t) new_size; 597 syscallarg(u_long) flags; 598 } */ *uap = v; 599 struct sys_munmap_args mua; 600 size_t old_size, new_size; 601 int error; 602 603 old_size = round_page(SCARG(uap, old_size)); 604 new_size = round_page(SCARG(uap, new_size)); 605 606 /* 607 * Growing mapped region. 608 */ 609 if (new_size > old_size) { 610 /* 611 * XXX Implement me. What we probably want to do is 612 * XXX dig out the guts of the old mapping, mmap that 613 * XXX object again with the new size, then munmap 614 * XXX the old mapping. 615 */ 616 *retval = 0; 617 return (ENOMEM); 618 } 619 620 /* 621 * Shrinking mapped region. 622 */ 623 if (new_size < old_size) { 624 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) + 625 new_size; 626 SCARG(&mua, len) = old_size - new_size; 627 error = sys_munmap(l, &mua, retval); 628 *retval = error ? 0 : (register_t)SCARG(uap, old_address); 629 return (error); 630 } 631 632 /* 633 * No change. 634 */ 635 *retval = (register_t)SCARG(uap, old_address); 636 return (0); 637 } 638 639 int 640 linux_sys_msync(l, v, retval) 641 struct lwp *l; 642 void *v; 643 register_t *retval; 644 { 645 struct linux_sys_msync_args /* { 646 syscallarg(caddr_t) addr; 647 syscallarg(int) len; 648 syscallarg(int) fl; 649 } */ *uap = v; 650 651 struct sys___msync13_args bma; 652 653 /* flags are ignored */ 654 SCARG(&bma, addr) = SCARG(uap, addr); 655 SCARG(&bma, len) = SCARG(uap, len); 656 SCARG(&bma, flags) = SCARG(uap, fl); 657 658 return sys___msync13(l, &bma, retval); 659 } 660 661 int 662 linux_sys_mprotect(l, v, retval) 663 struct lwp *l; 664 void *v; 665 register_t *retval; 666 { 667 struct linux_sys_mprotect_args /* { 668 syscallarg(const void *) start; 669 syscallarg(unsigned long) len; 670 syscallarg(int) prot; 671 } */ *uap = v; 672 unsigned long end, start = (unsigned long)SCARG(uap, start), len; 673 int prot = SCARG(uap, prot); 674 struct vm_map_entry *entry; 675 struct vm_map *map = &l->l_proc->p_vmspace->vm_map; 676 677 if (start & PAGE_MASK) 678 return EINVAL; 679 680 len = round_page(SCARG(uap, len)); 681 end = start + len; 682 683 if (end < start) 684 return EINVAL; 685 else if (end == start) 686 return 0; 687 688 if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 689 return EINVAL; 690 691 vm_map_lock(map); 692 #ifdef notdef 693 VM_MAP_RANGE_CHECK(map, start, end); 694 #endif 695 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 696 vm_map_unlock(map); 697 return ENOMEM; 698 } 699 vm_map_unlock(map); 700 return uvm_map_protect(map, start, end, prot, FALSE); 701 } 702 703 /* 704 * This code is partly stolen from src/lib/libc/compat-43/times.c 705 */ 706 707 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 708 709 int 710 linux_sys_times(l, v, retval) 711 struct lwp *l; 712 void *v; 713 register_t *retval; 714 { 715 struct linux_sys_times_args /* { 716 syscallarg(struct times *) tms; 717 } */ *uap = v; 718 struct proc *p = l->l_proc; 719 struct timeval t; 720 int error, s; 721 722 if (SCARG(uap, tms)) { 723 struct linux_tms ltms; 724 struct rusage ru; 725 726 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 727 ltms.ltms_utime = CONVTCK(ru.ru_utime); 728 ltms.ltms_stime = CONVTCK(ru.ru_stime); 729 730 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 731 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 732 733 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 734 return error; 735 } 736 737 s = splclock(); 738 timersub(&time, &boottime, &t); 739 splx(s); 740 741 retval[0] = ((linux_clock_t)(CONVTCK(t))); 742 return 0; 743 } 744 745 #undef CONVTCK 746 747 /* 748 * Linux 'readdir' call. This code is mostly taken from the 749 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 750 * an attempt has been made to keep it a little cleaner (failing 751 * miserably, because of the cruft needed if count 1 is passed). 752 * 753 * The d_off field should contain the offset of the next valid entry, 754 * but in Linux it has the offset of the entry itself. We emulate 755 * that bug here. 756 * 757 * Read in BSD-style entries, convert them, and copy them out. 758 * 759 * Note that this doesn't handle union-mounted filesystems. 760 */ 761 int 762 linux_sys_getdents(l, v, retval) 763 struct lwp *l; 764 void *v; 765 register_t *retval; 766 { 767 struct linux_sys_getdents_args /* { 768 syscallarg(int) fd; 769 syscallarg(struct linux_dirent *) dent; 770 syscallarg(unsigned int) count; 771 } */ *uap = v; 772 struct proc *p = l->l_proc; 773 struct dirent *bdp; 774 struct vnode *vp; 775 caddr_t inp, buf; /* BSD-format */ 776 int len, reclen; /* BSD-format */ 777 caddr_t outp; /* Linux-format */ 778 int resid, linux_reclen = 0; /* Linux-format */ 779 struct file *fp; 780 struct uio auio; 781 struct iovec aiov; 782 struct linux_dirent idb; 783 off_t off; /* true file offset */ 784 int buflen, error, eofflag, nbytes, oldcall; 785 struct vattr va; 786 off_t *cookiebuf = NULL, *cookie; 787 int ncookies; 788 789 /* getvnode() will use the descriptor for us */ 790 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 791 return (error); 792 793 if ((fp->f_flag & FREAD) == 0) { 794 error = EBADF; 795 goto out1; 796 } 797 798 vp = (struct vnode *)fp->f_data; 799 if (vp->v_type != VDIR) { 800 error = EINVAL; 801 goto out1; 802 } 803 804 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 805 goto out1; 806 807 nbytes = SCARG(uap, count); 808 if (nbytes == 1) { /* emulating old, broken behaviour */ 809 nbytes = sizeof (idb); 810 buflen = max(va.va_blocksize, nbytes); 811 oldcall = 1; 812 } else { 813 buflen = min(MAXBSIZE, nbytes); 814 if (buflen < va.va_blocksize) 815 buflen = va.va_blocksize; 816 oldcall = 0; 817 } 818 buf = malloc(buflen, M_TEMP, M_WAITOK); 819 820 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 821 off = fp->f_offset; 822 again: 823 aiov.iov_base = buf; 824 aiov.iov_len = buflen; 825 auio.uio_iov = &aiov; 826 auio.uio_iovcnt = 1; 827 auio.uio_rw = UIO_READ; 828 auio.uio_segflg = UIO_SYSSPACE; 829 auio.uio_procp = NULL; 830 auio.uio_resid = buflen; 831 auio.uio_offset = off; 832 /* 833 * First we read into the malloc'ed buffer, then 834 * we massage it into user space, one record at a time. 835 */ 836 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 837 &ncookies); 838 if (error) 839 goto out; 840 841 inp = buf; 842 outp = (caddr_t)SCARG(uap, dent); 843 resid = nbytes; 844 if ((len = buflen - auio.uio_resid) == 0) 845 goto eof; 846 847 for (cookie = cookiebuf; len > 0; len -= reclen) { 848 bdp = (struct dirent *)inp; 849 reclen = bdp->d_reclen; 850 if (reclen & 3) 851 panic("linux_readdir"); 852 if (bdp->d_fileno == 0) { 853 inp += reclen; /* it is a hole; squish it out */ 854 if (cookie) 855 off = *cookie++; 856 else 857 off += reclen; 858 continue; 859 } 860 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 861 if (reclen > len || resid < linux_reclen) { 862 /* entry too big for buffer, so just stop */ 863 outp++; 864 break; 865 } 866 /* 867 * Massage in place to make a Linux-shaped dirent (otherwise 868 * we have to worry about touching user memory outside of 869 * the copyout() call). 870 */ 871 idb.d_ino = bdp->d_fileno; 872 /* 873 * The old readdir() call misuses the offset and reclen fields. 874 */ 875 if (oldcall) { 876 idb.d_off = (linux_off_t)linux_reclen; 877 idb.d_reclen = (u_short)bdp->d_namlen; 878 } else { 879 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 880 compat_offseterr(vp, "linux_getdents"); 881 error = EINVAL; 882 goto out; 883 } 884 idb.d_off = (linux_off_t)off; 885 idb.d_reclen = (u_short)linux_reclen; 886 } 887 strcpy(idb.d_name, bdp->d_name); 888 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 889 goto out; 890 /* advance past this real entry */ 891 inp += reclen; 892 if (cookie) 893 off = *cookie++; /* each entry points to itself */ 894 else 895 off += reclen; 896 /* advance output past Linux-shaped entry */ 897 outp += linux_reclen; 898 resid -= linux_reclen; 899 if (oldcall) 900 break; 901 } 902 903 /* if we squished out the whole block, try again */ 904 if (outp == (caddr_t)SCARG(uap, dent)) 905 goto again; 906 fp->f_offset = off; /* update the vnode offset */ 907 908 if (oldcall) 909 nbytes = resid + linux_reclen; 910 911 eof: 912 *retval = nbytes - resid; 913 out: 914 VOP_UNLOCK(vp, 0); 915 if (cookiebuf) 916 free(cookiebuf, M_TEMP); 917 free(buf, M_TEMP); 918 out1: 919 FILE_UNUSE(fp, p); 920 return error; 921 } 922 923 /* 924 * Even when just using registers to pass arguments to syscalls you can 925 * have 5 of them on the i386. So this newer version of select() does 926 * this. 927 */ 928 int 929 linux_sys_select(l, v, retval) 930 struct lwp *l; 931 void *v; 932 register_t *retval; 933 { 934 struct linux_sys_select_args /* { 935 syscallarg(int) nfds; 936 syscallarg(fd_set *) readfds; 937 syscallarg(fd_set *) writefds; 938 syscallarg(fd_set *) exceptfds; 939 syscallarg(struct timeval *) timeout; 940 } */ *uap = v; 941 942 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 943 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 944 } 945 946 /* 947 * Common code for the old and new versions of select(). A couple of 948 * things are important: 949 * 1) return the amount of time left in the 'timeout' parameter 950 * 2) select never returns ERESTART on Linux, always return EINTR 951 */ 952 int 953 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout) 954 struct lwp *l; 955 register_t *retval; 956 int nfds; 957 fd_set *readfds, *writefds, *exceptfds; 958 struct timeval *timeout; 959 { 960 struct sys_select_args bsa; 961 struct proc *p = l->l_proc; 962 struct timeval tv0, tv1, utv, *tvp; 963 caddr_t sg; 964 int error; 965 966 SCARG(&bsa, nd) = nfds; 967 SCARG(&bsa, in) = readfds; 968 SCARG(&bsa, ou) = writefds; 969 SCARG(&bsa, ex) = exceptfds; 970 SCARG(&bsa, tv) = timeout; 971 972 /* 973 * Store current time for computation of the amount of 974 * time left. 975 */ 976 if (timeout) { 977 if ((error = copyin(timeout, &utv, sizeof(utv)))) 978 return error; 979 if (itimerfix(&utv)) { 980 /* 981 * The timeval was invalid. Convert it to something 982 * valid that will act as it does under Linux. 983 */ 984 sg = stackgap_init(p, 0); 985 tvp = stackgap_alloc(p, &sg, sizeof(utv)); 986 utv.tv_sec += utv.tv_usec / 1000000; 987 utv.tv_usec %= 1000000; 988 if (utv.tv_usec < 0) { 989 utv.tv_sec -= 1; 990 utv.tv_usec += 1000000; 991 } 992 if (utv.tv_sec < 0) 993 timerclear(&utv); 994 if ((error = copyout(&utv, tvp, sizeof(utv)))) 995 return error; 996 SCARG(&bsa, tv) = tvp; 997 } 998 microtime(&tv0); 999 } 1000 1001 error = sys_select(l, &bsa, retval); 1002 if (error) { 1003 /* 1004 * See fs/select.c in the Linux kernel. Without this, 1005 * Maelstrom doesn't work. 1006 */ 1007 if (error == ERESTART) 1008 error = EINTR; 1009 return error; 1010 } 1011 1012 if (timeout) { 1013 if (*retval) { 1014 /* 1015 * Compute how much time was left of the timeout, 1016 * by subtracting the current time and the time 1017 * before we started the call, and subtracting 1018 * that result from the user-supplied value. 1019 */ 1020 microtime(&tv1); 1021 timersub(&tv1, &tv0, &tv1); 1022 timersub(&utv, &tv1, &utv); 1023 if (utv.tv_sec < 0) 1024 timerclear(&utv); 1025 } else 1026 timerclear(&utv); 1027 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1028 return error; 1029 } 1030 1031 return 0; 1032 } 1033 1034 /* 1035 * Get the process group of a certain process. Look it up 1036 * and return the value. 1037 */ 1038 int 1039 linux_sys_getpgid(l, v, retval) 1040 struct lwp *l; 1041 void *v; 1042 register_t *retval; 1043 { 1044 struct linux_sys_getpgid_args /* { 1045 syscallarg(int) pid; 1046 } */ *uap = v; 1047 struct proc *p = l->l_proc; 1048 struct proc *targp; 1049 1050 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1051 if ((targp = pfind(SCARG(uap, pid))) == 0) 1052 return ESRCH; 1053 } 1054 else 1055 targp = p; 1056 1057 retval[0] = targp->p_pgid; 1058 return 0; 1059 } 1060 1061 /* 1062 * Set the 'personality' (emulation mode) for the current process. Only 1063 * accept the Linux personality here (0). This call is needed because 1064 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1065 * ELF binaries run in Linux mode, not SVR4 mode. 1066 */ 1067 int 1068 linux_sys_personality(l, v, retval) 1069 struct lwp *l; 1070 void *v; 1071 register_t *retval; 1072 { 1073 struct linux_sys_personality_args /* { 1074 syscallarg(int) per; 1075 } */ *uap = v; 1076 1077 if (SCARG(uap, per) != 0) 1078 return EINVAL; 1079 retval[0] = 0; 1080 return 0; 1081 } 1082 1083 #if defined(__i386__) || defined(__m68k__) 1084 /* 1085 * The calls are here because of type conversions. 1086 */ 1087 int 1088 linux_sys_setreuid16(l, v, retval) 1089 struct lwp *l; 1090 void *v; 1091 register_t *retval; 1092 { 1093 struct linux_sys_setreuid16_args /* { 1094 syscallarg(int) ruid; 1095 syscallarg(int) euid; 1096 } */ *uap = v; 1097 struct sys_setreuid_args bsa; 1098 1099 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1100 (uid_t)-1 : SCARG(uap, ruid); 1101 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1102 (uid_t)-1 : SCARG(uap, euid); 1103 1104 return sys_setreuid(l, &bsa, retval); 1105 } 1106 1107 int 1108 linux_sys_setregid16(l, v, retval) 1109 struct lwp *l; 1110 void *v; 1111 register_t *retval; 1112 { 1113 struct linux_sys_setregid16_args /* { 1114 syscallarg(int) rgid; 1115 syscallarg(int) egid; 1116 } */ *uap = v; 1117 struct sys_setregid_args bsa; 1118 1119 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1120 (uid_t)-1 : SCARG(uap, rgid); 1121 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1122 (uid_t)-1 : SCARG(uap, egid); 1123 1124 return sys_setregid(l, &bsa, retval); 1125 } 1126 1127 int 1128 linux_sys_setresuid16(l, v, retval) 1129 struct lwp *l; 1130 void *v; 1131 register_t *retval; 1132 { 1133 struct linux_sys_setresuid16_args /* { 1134 syscallarg(uid_t) ruid; 1135 syscallarg(uid_t) euid; 1136 syscallarg(uid_t) suid; 1137 } */ *uap = v; 1138 struct linux_sys_setresuid16_args lsa; 1139 1140 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1141 (uid_t)-1 : SCARG(uap, ruid); 1142 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1143 (uid_t)-1 : SCARG(uap, euid); 1144 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ? 1145 (uid_t)-1 : SCARG(uap, suid); 1146 1147 return linux_sys_setresuid(l, &lsa, retval); 1148 } 1149 1150 int 1151 linux_sys_setresgid16(l, v, retval) 1152 struct lwp *l; 1153 void *v; 1154 register_t *retval; 1155 { 1156 struct linux_sys_setresgid16_args /* { 1157 syscallarg(gid_t) rgid; 1158 syscallarg(gid_t) egid; 1159 syscallarg(gid_t) sgid; 1160 } */ *uap = v; 1161 struct linux_sys_setresgid16_args lsa; 1162 1163 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1164 (gid_t)-1 : SCARG(uap, rgid); 1165 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1166 (gid_t)-1 : SCARG(uap, egid); 1167 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ? 1168 (gid_t)-1 : SCARG(uap, sgid); 1169 1170 return linux_sys_setresgid(l, &lsa, retval); 1171 } 1172 1173 int 1174 linux_sys_getgroups16(l, v, retval) 1175 struct lwp *l; 1176 void *v; 1177 register_t *retval; 1178 { 1179 struct linux_sys_getgroups16_args /* { 1180 syscallarg(int) gidsetsize; 1181 syscallarg(linux_gid_t *) gidset; 1182 } */ *uap = v; 1183 struct proc *p = l->l_proc; 1184 caddr_t sg; 1185 int n, error, i; 1186 struct sys_getgroups_args bsa; 1187 gid_t *bset, *kbset; 1188 linux_gid_t *lset; 1189 struct pcred *pc = p->p_cred; 1190 1191 n = SCARG(uap, gidsetsize); 1192 if (n < 0) 1193 return EINVAL; 1194 error = 0; 1195 bset = kbset = NULL; 1196 lset = NULL; 1197 if (n > 0) { 1198 n = min(pc->pc_ucred->cr_ngroups, n); 1199 sg = stackgap_init(p, 0); 1200 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1201 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1202 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1203 if (bset == NULL || kbset == NULL || lset == NULL) 1204 return ENOMEM; 1205 SCARG(&bsa, gidsetsize) = n; 1206 SCARG(&bsa, gidset) = bset; 1207 error = sys_getgroups(l, &bsa, retval); 1208 if (error != 0) 1209 goto out; 1210 error = copyin(bset, kbset, n * sizeof (gid_t)); 1211 if (error != 0) 1212 goto out; 1213 for (i = 0; i < n; i++) 1214 lset[i] = (linux_gid_t)kbset[i]; 1215 error = copyout(lset, SCARG(uap, gidset), 1216 n * sizeof (linux_gid_t)); 1217 } else 1218 *retval = pc->pc_ucred->cr_ngroups; 1219 out: 1220 if (kbset != NULL) 1221 free(kbset, M_TEMP); 1222 if (lset != NULL) 1223 free(lset, M_TEMP); 1224 return error; 1225 } 1226 1227 int 1228 linux_sys_setgroups16(l, v, retval) 1229 struct lwp *l; 1230 void *v; 1231 register_t *retval; 1232 { 1233 struct linux_sys_setgroups16_args /* { 1234 syscallarg(int) gidsetsize; 1235 syscallarg(linux_gid_t *) gidset; 1236 } */ *uap = v; 1237 struct proc *p = l->l_proc; 1238 caddr_t sg; 1239 int n; 1240 int error, i; 1241 struct sys_setgroups_args bsa; 1242 gid_t *bset, *kbset; 1243 linux_gid_t *lset; 1244 1245 n = SCARG(uap, gidsetsize); 1246 if (n < 0 || n > NGROUPS) 1247 return EINVAL; 1248 sg = stackgap_init(p, 0); 1249 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1250 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1251 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1252 if (lset == NULL || bset == NULL) 1253 return ENOMEM; 1254 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t)); 1255 if (error != 0) 1256 goto out; 1257 for (i = 0; i < n; i++) 1258 kbset[i] = (gid_t)lset[i]; 1259 error = copyout(kbset, bset, n * sizeof (gid_t)); 1260 if (error != 0) 1261 goto out; 1262 SCARG(&bsa, gidsetsize) = n; 1263 SCARG(&bsa, gidset) = bset; 1264 error = sys_setgroups(l, &bsa, retval); 1265 1266 out: 1267 if (lset != NULL) 1268 free(lset, M_TEMP); 1269 if (kbset != NULL) 1270 free(kbset, M_TEMP); 1271 1272 return error; 1273 } 1274 1275 #endif /* __i386__ || __m68k__ */ 1276 1277 /* 1278 * We have nonexistent fsuid equal to uid. 1279 * If modification is requested, refuse. 1280 */ 1281 int 1282 linux_sys_setfsuid(l, v, retval) 1283 struct lwp *l; 1284 void *v; 1285 register_t *retval; 1286 { 1287 struct linux_sys_setfsuid_args /* { 1288 syscallarg(uid_t) uid; 1289 } */ *uap = v; 1290 struct proc *p = l->l_proc; 1291 uid_t uid; 1292 1293 uid = SCARG(uap, uid); 1294 if (p->p_cred->p_ruid != uid) 1295 return sys_nosys(l, v, retval); 1296 else 1297 return (0); 1298 } 1299 1300 /* XXX XXX XXX */ 1301 #ifndef alpha 1302 int 1303 linux_sys_getfsuid(l, v, retval) 1304 struct lwp *l; 1305 void *v; 1306 register_t *retval; 1307 { 1308 return sys_getuid(l, v, retval); 1309 } 1310 #endif 1311 1312 int 1313 linux_sys_setresuid(l, v, retval) 1314 struct lwp *l; 1315 void *v; 1316 register_t *retval; 1317 { 1318 struct linux_sys_setresuid_args /* { 1319 syscallarg(uid_t) ruid; 1320 syscallarg(uid_t) euid; 1321 syscallarg(uid_t) suid; 1322 } */ *uap = v; 1323 1324 /* 1325 * Note: These checks are a little different than the NetBSD 1326 * setreuid(2) call performs. This precisely follows the 1327 * behavior of the Linux kernel. 1328 */ 1329 1330 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1331 SCARG(uap, suid), 1332 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1333 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1334 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1335 } 1336 1337 int 1338 linux_sys_getresuid(l, v, retval) 1339 struct lwp *l; 1340 void *v; 1341 register_t *retval; 1342 { 1343 struct linux_sys_getresuid_args /* { 1344 syscallarg(uid_t *) ruid; 1345 syscallarg(uid_t *) euid; 1346 syscallarg(uid_t *) suid; 1347 } */ *uap = v; 1348 struct proc *p = l->l_proc; 1349 struct pcred *pc = p->p_cred; 1350 int error; 1351 1352 /* 1353 * Linux copies these values out to userspace like so: 1354 * 1355 * 1. Copy out ruid. 1356 * 2. If that succeeds, copy out euid. 1357 * 3. If both of those succeed, copy out suid. 1358 */ 1359 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid), 1360 sizeof(uid_t))) != 0) 1361 return (error); 1362 1363 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid), 1364 sizeof(uid_t))) != 0) 1365 return (error); 1366 1367 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t))); 1368 } 1369 1370 int 1371 linux_sys_ptrace(l, v, retval) 1372 struct lwp *l; 1373 void *v; 1374 register_t *retval; 1375 { 1376 struct linux_sys_ptrace_args /* { 1377 i386, m68k, powerpc: T=int 1378 alpha: T=long 1379 syscallarg(T) request; 1380 syscallarg(T) pid; 1381 syscallarg(T) addr; 1382 syscallarg(T) data; 1383 } */ *uap = v; 1384 const int *ptr; 1385 int request; 1386 int error; 1387 1388 ptr = linux_ptrace_request_map; 1389 request = SCARG(uap, request); 1390 while (*ptr != -1) 1391 if (*ptr++ == request) { 1392 struct sys_ptrace_args pta; 1393 1394 SCARG(&pta, req) = *ptr; 1395 SCARG(&pta, pid) = SCARG(uap, pid); 1396 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr); 1397 SCARG(&pta, data) = SCARG(uap, data); 1398 1399 /* 1400 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1401 * to continue where the process left off previously. 1402 * The same thing is achieved by addr == (caddr_t) 1 1403 * on NetBSD, so rewrite 'addr' appropriately. 1404 */ 1405 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1406 SCARG(&pta, addr) = (caddr_t) 1; 1407 1408 error = sys_ptrace(l, &pta, retval); 1409 if (error) 1410 return error; 1411 switch (request) { 1412 case LINUX_PTRACE_PEEKTEXT: 1413 case LINUX_PTRACE_PEEKDATA: 1414 error = copyout (retval, 1415 (caddr_t)SCARG(uap, data), sizeof *retval); 1416 *retval = SCARG(uap, data); 1417 break; 1418 default: 1419 break; 1420 } 1421 return error; 1422 } 1423 else 1424 ptr++; 1425 1426 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1427 } 1428 1429 int 1430 linux_sys_reboot(struct lwp *l, void *v, register_t *retval) 1431 { 1432 struct linux_sys_reboot_args /* { 1433 syscallarg(int) magic1; 1434 syscallarg(int) magic2; 1435 syscallarg(int) cmd; 1436 syscallarg(void *) arg; 1437 } */ *uap = v; 1438 struct sys_reboot_args /* { 1439 syscallarg(int) opt; 1440 syscallarg(char *) bootstr; 1441 } */ sra; 1442 struct proc *p = l->l_proc; 1443 int error; 1444 1445 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1446 return(error); 1447 1448 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1449 return(EINVAL); 1450 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1451 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1452 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1453 return(EINVAL); 1454 1455 switch (SCARG(uap, cmd)) { 1456 case LINUX_REBOOT_CMD_RESTART: 1457 SCARG(&sra, opt) = RB_AUTOBOOT; 1458 break; 1459 case LINUX_REBOOT_CMD_HALT: 1460 SCARG(&sra, opt) = RB_HALT; 1461 break; 1462 case LINUX_REBOOT_CMD_POWER_OFF: 1463 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1464 break; 1465 case LINUX_REBOOT_CMD_RESTART2: 1466 /* Reboot with an argument. */ 1467 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1468 SCARG(&sra, bootstr) = SCARG(uap, arg); 1469 break; 1470 case LINUX_REBOOT_CMD_CAD_ON: 1471 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1472 case LINUX_REBOOT_CMD_CAD_OFF: 1473 return(0); 1474 default: 1475 return(EINVAL); 1476 } 1477 1478 return(sys_reboot(l, &sra, retval)); 1479 } 1480 1481 /* 1482 * Copy of compat_12_sys_swapon(). 1483 */ 1484 int 1485 linux_sys_swapon(l, v, retval) 1486 struct lwp *l; 1487 void *v; 1488 register_t *retval; 1489 { 1490 struct sys_swapctl_args ua; 1491 struct linux_sys_swapon_args /* { 1492 syscallarg(const char *) name; 1493 } */ *uap = v; 1494 1495 SCARG(&ua, cmd) = SWAP_ON; 1496 SCARG(&ua, arg) = (void *)SCARG(uap, name); 1497 SCARG(&ua, misc) = 0; /* priority */ 1498 return (sys_swapctl(l, &ua, retval)); 1499 } 1500 1501 /* 1502 * Stop swapping to the file or block device specified by path. 1503 */ 1504 int 1505 linux_sys_swapoff(l, v, retval) 1506 struct lwp *l; 1507 void *v; 1508 register_t *retval; 1509 { 1510 struct sys_swapctl_args ua; 1511 struct linux_sys_swapoff_args /* { 1512 syscallarg(const char *) path; 1513 } */ *uap = v; 1514 1515 SCARG(&ua, cmd) = SWAP_OFF; 1516 SCARG(&ua, arg) = (void *)SCARG(uap, path); 1517 return (sys_swapctl(l, &ua, retval)); 1518 } 1519 1520 /* 1521 * Copy of compat_09_sys_setdomainname() 1522 */ 1523 /* ARGSUSED */ 1524 int 1525 linux_sys_setdomainname(l, v, retval) 1526 struct lwp *l; 1527 void *v; 1528 register_t *retval; 1529 { 1530 struct linux_sys_setdomainname_args /* { 1531 syscallarg(char *) domainname; 1532 syscallarg(int) len; 1533 } */ *uap = v; 1534 int name[2]; 1535 1536 name[0] = CTL_KERN; 1537 name[1] = KERN_DOMAINNAME; 1538 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1539 SCARG(uap, len), l)); 1540 } 1541 1542 /* 1543 * sysinfo() 1544 */ 1545 /* ARGSUSED */ 1546 int 1547 linux_sys_sysinfo(l, v, retval) 1548 struct lwp *l; 1549 void *v; 1550 register_t *retval; 1551 { 1552 struct linux_sys_sysinfo_args /* { 1553 syscallarg(struct linux_sysinfo *) arg; 1554 } */ *uap = v; 1555 struct linux_sysinfo si; 1556 struct loadavg *la; 1557 1558 si.uptime = time.tv_sec - boottime.tv_sec; 1559 la = &averunnable; 1560 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1561 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1562 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1563 si.totalram = ctob(physmem); 1564 si.freeram = uvmexp.free * uvmexp.pagesize; 1565 si.sharedram = 0; /* XXX */ 1566 si.bufferram = uvmexp.filepages * uvmexp.pagesize; 1567 si.totalswap = uvmexp.swpages * uvmexp.pagesize; 1568 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1569 si.procs = nprocs; 1570 1571 /* The following are only present in newer Linux kernels. */ 1572 si.totalbig = 0; 1573 si.freebig = 0; 1574 si.mem_unit = 1; 1575 1576 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1577 } 1578 1579 #define bsd_to_linux_rlimit1(l, b, f) \ 1580 (l)->f = ((b)->f == RLIM_INFINITY || \ 1581 ((b)->f & 0xffffffff00000000ULL) != 0) ? \ 1582 LINUX_RLIM_INFINITY : (int32_t)(b)->f 1583 #define bsd_to_linux_rlimit(l, b) \ 1584 bsd_to_linux_rlimit1(l, b, rlim_cur); \ 1585 bsd_to_linux_rlimit1(l, b, rlim_max) 1586 1587 #define linux_to_bsd_rlimit1(b, l, f) \ 1588 (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f 1589 #define linux_to_bsd_rlimit(b, l) \ 1590 linux_to_bsd_rlimit1(b, l, rlim_cur); \ 1591 linux_to_bsd_rlimit1(b, l, rlim_max) 1592 1593 static int 1594 linux_to_bsd_limit(lim) 1595 int lim; 1596 { 1597 switch (lim) { 1598 case LINUX_RLIMIT_CPU: 1599 return RLIMIT_CPU; 1600 case LINUX_RLIMIT_FSIZE: 1601 return RLIMIT_FSIZE; 1602 case LINUX_RLIMIT_DATA: 1603 return RLIMIT_DATA; 1604 case LINUX_RLIMIT_STACK: 1605 return RLIMIT_STACK; 1606 case LINUX_RLIMIT_CORE: 1607 return RLIMIT_CORE; 1608 case LINUX_RLIMIT_RSS: 1609 return RLIMIT_RSS; 1610 case LINUX_RLIMIT_NPROC: 1611 return RLIMIT_NPROC; 1612 case LINUX_RLIMIT_NOFILE: 1613 return RLIMIT_NOFILE; 1614 case LINUX_RLIMIT_MEMLOCK: 1615 return RLIMIT_MEMLOCK; 1616 case LINUX_RLIMIT_AS: 1617 case LINUX_RLIMIT_LOCKS: 1618 return -EOPNOTSUPP; 1619 default: 1620 return -EINVAL; 1621 } 1622 } 1623 1624 1625 int 1626 linux_sys_getrlimit(l, v, retval) 1627 struct lwp *l; 1628 void *v; 1629 register_t *retval; 1630 { 1631 struct linux_sys_getrlimit_args /* { 1632 syscallarg(int) which; 1633 syscallarg(struct orlimit *) rlp; 1634 } */ *uap = v; 1635 struct proc *p = l->l_proc; 1636 caddr_t sg = stackgap_init(p, 0); 1637 struct sys_getrlimit_args ap; 1638 struct rlimit rl; 1639 struct orlimit orl; 1640 int error; 1641 1642 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1643 if ((error = SCARG(&ap, which)) < 0) 1644 return -error; 1645 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1646 if ((error = sys_getrlimit(l, &ap, retval)) != 0) 1647 return error; 1648 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0) 1649 return error; 1650 bsd_to_linux_rlimit(&orl, &rl); 1651 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1652 } 1653 1654 int 1655 linux_sys_setrlimit(l, v, retval) 1656 struct lwp *l; 1657 void *v; 1658 register_t *retval; 1659 { 1660 struct linux_sys_setrlimit_args /* { 1661 syscallarg(int) which; 1662 syscallarg(struct orlimit *) rlp; 1663 } */ *uap = v; 1664 struct proc *p = l->l_proc; 1665 caddr_t sg = stackgap_init(p, 0); 1666 struct sys_setrlimit_args ap; 1667 struct rlimit rl; 1668 struct orlimit orl; 1669 int error; 1670 1671 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1672 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1673 if ((error = SCARG(&ap, which)) < 0) 1674 return -error; 1675 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1676 return error; 1677 linux_to_bsd_rlimit(&rl, &orl); 1678 /* XXX: alpha complains about this */ 1679 if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0) 1680 return error; 1681 return sys_setrlimit(l, &ap, retval); 1682 } 1683 1684 #ifndef __mips__ 1685 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1686 int 1687 linux_sys_ugetrlimit(l, v, retval) 1688 struct lwp *l; 1689 void *v; 1690 register_t *retval; 1691 { 1692 return linux_sys_getrlimit(l, v, retval); 1693 } 1694 #endif 1695 1696 /* 1697 * This gets called for unsupported syscalls. The difference to sys_nosys() 1698 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1699 * This is the way Linux does it and glibc depends on this behaviour. 1700 */ 1701 int 1702 linux_sys_nosys(l, v, retval) 1703 struct lwp *l; 1704 void *v; 1705 register_t *retval; 1706 { 1707 return (ENOSYS); 1708 } 1709
Indexes created Mon Sep 22 21:09:42 GMT 2025