1 /* $NetBSD: linux_misc.c,v 1.147 2005/12/11 12:20:19 christos 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 1.147 2005/12/11 12:20:19 christos 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_machdep.h> 106 #include <compat/linux/common/linux_types.h> 107 #include <compat/linux/common/linux_signal.h> 108 109 #include <compat/linux/linux_syscallargs.h> 110 111 #include <compat/linux/common/linux_fcntl.h> 112 #include <compat/linux/common/linux_mmap.h> 113 #include <compat/linux/common/linux_dirent.h> 114 #include <compat/linux/common/linux_util.h> 115 #include <compat/linux/common/linux_misc.h> 116 #include <compat/linux/common/linux_ptrace.h> 117 #include <compat/linux/common/linux_reboot.h> 118 #include <compat/linux/common/linux_emuldata.h> 119 120 const int linux_ptrace_request_map[] = { 121 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 122 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 123 LINUX_PTRACE_PEEKDATA, PT_READ_D, 124 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 125 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 126 LINUX_PTRACE_CONT, PT_CONTINUE, 127 LINUX_PTRACE_KILL, PT_KILL, 128 LINUX_PTRACE_ATTACH, PT_ATTACH, 129 LINUX_PTRACE_DETACH, PT_DETACH, 130 #ifdef PT_STEP 131 LINUX_PTRACE_SINGLESTEP, PT_STEP, 132 #endif 133 -1 134 }; 135 136 const struct linux_mnttypes linux_fstypes[] = { 137 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 138 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 139 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 140 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 141 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 142 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 143 { MOUNT_PORTAL, LINUX_DEFAULT_SUPER_MAGIC }, 144 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 145 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 146 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 148 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 149 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 150 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 151 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 152 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 153 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 154 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 155 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 156 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 157 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 158 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC }, 159 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC }, 160 { MOUNT_TMPFS, LINUX_DEFAULT_SUPER_MAGIC } 161 }; 162 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 163 164 #ifdef DEBUG_LINUX 165 #define DPRINTF(a) uprintf a 166 #else 167 #define DPRINTF(a) 168 #endif 169 170 /* Local linux_misc.c functions: */ 171 #ifndef __amd64__ 172 static void bsd_to_linux_statfs __P((const struct statvfs *, 173 struct linux_statfs *)); 174 #endif 175 static int linux_to_bsd_limit __P((int)); 176 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *, 177 const struct linux_sys_mmap_args *)); 178 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *, 179 register_t *, off_t)); 180 181 182 /* 183 * The information on a terminated (or stopped) process needs 184 * to be converted in order for Linux binaries to get a valid signal 185 * number out of it. 186 */ 187 void 188 bsd_to_linux_wstat(st) 189 int *st; 190 { 191 192 int sig; 193 194 if (WIFSIGNALED(*st)) { 195 sig = WTERMSIG(*st); 196 if (sig >= 0 && sig < NSIG) 197 *st= (*st& ~0177) | native_to_linux_signo[sig]; 198 } else if (WIFSTOPPED(*st)) { 199 sig = WSTOPSIG(*st); 200 if (sig >= 0 && sig < NSIG) 201 *st = (*st & ~0xff00) | 202 (native_to_linux_signo[sig] << 8); 203 } 204 } 205 206 /* 207 * wait4(2). Passed on to the NetBSD call, surrounded by code to 208 * reserve some space for a NetBSD-style wait status, and converting 209 * it to what Linux wants. 210 */ 211 int 212 linux_sys_wait4(l, v, retval) 213 struct lwp *l; 214 void *v; 215 register_t *retval; 216 { 217 struct linux_sys_wait4_args /* { 218 syscallarg(int) pid; 219 syscallarg(int *) status; 220 syscallarg(int) options; 221 syscallarg(struct rusage *) rusage; 222 } */ *uap = v; 223 struct proc *p = l->l_proc; 224 struct sys_wait4_args w4a; 225 int error, *status, tstat, options, linux_options; 226 caddr_t sg; 227 228 if (SCARG(uap, status) != NULL) { 229 sg = stackgap_init(p, 0); 230 status = (int *) stackgap_alloc(p, &sg, sizeof *status); 231 } else 232 status = NULL; 233 234 linux_options = SCARG(uap, options); 235 options = 0; 236 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS)) 237 return (EINVAL); 238 239 if (linux_options & LINUX_WAIT4_WNOHANG) 240 options |= WNOHANG; 241 if (linux_options & LINUX_WAIT4_WUNTRACED) 242 options |= WUNTRACED; 243 if (linux_options & LINUX_WAIT4_WALL) 244 options |= WALLSIG; 245 if (linux_options & LINUX_WAIT4_WCLONE) 246 options |= WALTSIG; 247 #ifdef DIAGNOSTIC 248 if (linux_options & LINUX_WAIT4_WNOTHREAD) 249 printf("WARNING: %s: linux process %d.%d called " 250 "waitpid with __WNOTHREAD set!", 251 __FILE__, p->p_pid, l->l_lid); 252 253 #endif 254 255 SCARG(&w4a, pid) = SCARG(uap, pid); 256 SCARG(&w4a, status) = status; 257 SCARG(&w4a, options) = options; 258 SCARG(&w4a, rusage) = SCARG(uap, rusage); 259 260 if ((error = sys_wait4(l, &w4a, retval))) 261 return error; 262 263 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD); 264 265 if (status != NULL) { 266 if ((error = copyin(status, &tstat, sizeof tstat))) 267 return error; 268 269 bsd_to_linux_wstat(&tstat); 270 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 271 } 272 273 return 0; 274 } 275 276 /* 277 * Linux brk(2). The check if the new address is >= the old one is 278 * done in the kernel in Linux. NetBSD does it in the library. 279 */ 280 int 281 linux_sys_brk(l, v, retval) 282 struct lwp *l; 283 void *v; 284 register_t *retval; 285 { 286 struct linux_sys_brk_args /* { 287 syscallarg(char *) nsize; 288 } */ *uap = v; 289 struct proc *p = l->l_proc; 290 char *nbrk = SCARG(uap, nsize); 291 struct sys_obreak_args oba; 292 struct vmspace *vm = p->p_vmspace; 293 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata; 294 295 SCARG(&oba, nsize) = nbrk; 296 297 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0) 298 ed->s->p_break = (char*)nbrk; 299 else 300 nbrk = ed->s->p_break; 301 302 retval[0] = (register_t)nbrk; 303 304 return 0; 305 } 306 307 #ifndef __amd64__ 308 /* 309 * Convert NetBSD statvfs structure to Linux statfs structure. 310 * Linux doesn't have f_flag, and we can't set f_frsize due 311 * to glibc statvfs() bug (see below). 312 */ 313 static void 314 bsd_to_linux_statfs(bsp, lsp) 315 const struct statvfs *bsp; 316 struct linux_statfs *lsp; 317 { 318 int i; 319 320 for (i = 0; i < linux_fstypes_cnt; i++) { 321 if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) { 322 lsp->l_ftype = linux_fstypes[i].linux; 323 break; 324 } 325 } 326 327 if (i == linux_fstypes_cnt) { 328 DPRINTF(("unhandled fstype in linux emulation: %s\n", 329 bsp->f_fstypename)); 330 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC; 331 } 332 333 /* 334 * The sizes are expressed in number of blocks. The block 335 * size used for the size is f_frsize for POSIX-compliant 336 * statvfs. Linux statfs uses f_bsize as the block size 337 * (f_frsize used to not be available in Linux struct statfs). 338 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block 339 * counts for different f_frsize if f_frsize is provided by the kernel. 340 * POSIX conforming apps thus get wrong size if f_frsize 341 * is different to f_bsize. Thus, we just pretend we don't 342 * support f_frsize. 343 */ 344 345 lsp->l_fbsize = bsp->f_frsize; 346 lsp->l_ffrsize = 0; /* compat */ 347 lsp->l_fblocks = bsp->f_blocks; 348 lsp->l_fbfree = bsp->f_bfree; 349 lsp->l_fbavail = bsp->f_bavail; 350 lsp->l_ffiles = bsp->f_files; 351 lsp->l_fffree = bsp->f_ffree; 352 /* Linux sets the fsid to 0..., we don't */ 353 lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0]; 354 lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1]; 355 lsp->l_fnamelen = bsp->f_namemax; 356 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare)); 357 } 358 359 /* 360 * Implement the fs stat functions. Straightforward. 361 */ 362 int 363 linux_sys_statfs(l, v, retval) 364 struct lwp *l; 365 void *v; 366 register_t *retval; 367 { 368 struct linux_sys_statfs_args /* { 369 syscallarg(const char *) path; 370 syscallarg(struct linux_statfs *) sp; 371 } */ *uap = v; 372 struct proc *p = l->l_proc; 373 struct statvfs btmp, *bsp; 374 struct linux_statfs ltmp; 375 struct sys_statvfs1_args bsa; 376 caddr_t sg; 377 int error; 378 379 sg = stackgap_init(p, 0); 380 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 381 382 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path)); 383 384 SCARG(&bsa, path) = SCARG(uap, path); 385 SCARG(&bsa, buf) = bsp; 386 SCARG(&bsa, flags) = ST_WAIT; 387 388 if ((error = sys_statvfs1(l, &bsa, retval))) 389 return error; 390 391 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 392 return error; 393 394 bsd_to_linux_statfs(&btmp, <mp); 395 396 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 397 } 398 399 int 400 linux_sys_fstatfs(l, v, retval) 401 struct lwp *l; 402 void *v; 403 register_t *retval; 404 { 405 struct linux_sys_fstatfs_args /* { 406 syscallarg(int) fd; 407 syscallarg(struct linux_statfs *) sp; 408 } */ *uap = v; 409 struct proc *p = l->l_proc; 410 struct statvfs btmp, *bsp; 411 struct linux_statfs ltmp; 412 struct sys_fstatvfs1_args bsa; 413 caddr_t sg; 414 int error; 415 416 sg = stackgap_init(p, 0); 417 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 418 419 SCARG(&bsa, fd) = SCARG(uap, fd); 420 SCARG(&bsa, buf) = bsp; 421 SCARG(&bsa, flags) = ST_WAIT; 422 423 if ((error = sys_fstatvfs1(l, &bsa, retval))) 424 return error; 425 426 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 427 return error; 428 429 bsd_to_linux_statfs(&btmp, <mp); 430 431 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 432 } 433 #endif /* __amd64__ */ 434 435 /* 436 * uname(). Just copy the info from the various strings stored in the 437 * kernel, and put it in the Linux utsname structure. That structure 438 * is almost the same as the NetBSD one, only it has fields 65 characters 439 * long, and an extra domainname field. 440 */ 441 int 442 linux_sys_uname(l, v, retval) 443 struct lwp *l; 444 void *v; 445 register_t *retval; 446 { 447 struct linux_sys_uname_args /* { 448 syscallarg(struct linux_utsname *) up; 449 } */ *uap = v; 450 struct linux_utsname luts; 451 452 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 453 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 454 strncpy(luts.l_release, linux_release, sizeof(luts.l_release)); 455 strncpy(luts.l_version, linux_version, sizeof(luts.l_version)); 456 #ifdef LINUX_UNAME_ARCH 457 strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine)); 458 #else 459 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 460 #endif 461 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 462 463 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 464 } 465 466 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 467 /* Used indirectly on: arm, i386, m68k */ 468 469 /* 470 * New type Linux mmap call. 471 * Only called directly on machines with >= 6 free regs. 472 */ 473 int 474 linux_sys_mmap(l, v, retval) 475 struct lwp *l; 476 void *v; 477 register_t *retval; 478 { 479 struct linux_sys_mmap_args /* { 480 syscallarg(unsigned long) addr; 481 syscallarg(size_t) len; 482 syscallarg(int) prot; 483 syscallarg(int) flags; 484 syscallarg(int) fd; 485 syscallarg(linux_off_t) offset; 486 } */ *uap = v; 487 488 if (SCARG(uap, offset) & PAGE_MASK) 489 return EINVAL; 490 491 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 492 } 493 494 /* 495 * Guts of most architectures' mmap64() implementations. This shares 496 * its list of arguments with linux_sys_mmap(). 497 * 498 * The difference in linux_sys_mmap2() is that "offset" is actually 499 * (offset / pagesize), not an absolute byte count. This translation 500 * to pagesize offsets is done inside glibc between the mmap64() call 501 * point, and the actual syscall. 502 */ 503 int 504 linux_sys_mmap2(l, v, retval) 505 struct lwp *l; 506 void *v; 507 register_t *retval; 508 { 509 struct linux_sys_mmap2_args /* { 510 syscallarg(unsigned long) addr; 511 syscallarg(size_t) len; 512 syscallarg(int) prot; 513 syscallarg(int) flags; 514 syscallarg(int) fd; 515 syscallarg(linux_off_t) offset; 516 } */ *uap = v; 517 518 return linux_mmap(l, uap, retval, 519 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 520 } 521 522 /* 523 * Massage arguments and call system mmap(2). 524 */ 525 static int 526 linux_mmap(l, uap, retval, offset) 527 struct lwp *l; 528 struct linux_sys_mmap_args *uap; 529 register_t *retval; 530 off_t offset; 531 { 532 struct sys_mmap_args cma; 533 int error; 534 size_t mmoff=0; 535 536 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 537 /* 538 * Request for stack-like memory segment. On linux, this 539 * works by mmap()ping (small) segment, which is automatically 540 * extended when page fault happens below the currently 541 * allocated area. We emulate this by allocating (typically 542 * bigger) segment sized at current stack size limit, and 543 * offsetting the requested and returned address accordingly. 544 * Since physical pages are only allocated on-demand, this 545 * is effectively identical. 546 */ 547 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 548 549 if (SCARG(uap, len) < ssl) { 550 /* Compute the address offset */ 551 mmoff = round_page(ssl) - SCARG(uap, len); 552 553 if (SCARG(uap, addr)) 554 SCARG(uap, addr) -= mmoff; 555 556 SCARG(uap, len) = (size_t) ssl; 557 } 558 } 559 560 linux_to_bsd_mmap_args(&cma, uap); 561 SCARG(&cma, pos) = offset; 562 563 error = sys_mmap(l, &cma, retval); 564 if (error) 565 return (error); 566 567 /* Shift the returned address for stack-like segment if necessary */ 568 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff) 569 retval[0] += mmoff; 570 571 return (0); 572 } 573 574 static void 575 linux_to_bsd_mmap_args(cma, uap) 576 struct sys_mmap_args *cma; 577 const struct linux_sys_mmap_args *uap; 578 { 579 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 580 581 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 582 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 583 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 584 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 585 /* XXX XAX ERH: Any other flags here? There are more defined... */ 586 587 SCARG(cma, addr) = (void *)SCARG(uap, addr); 588 SCARG(cma, len) = SCARG(uap, len); 589 SCARG(cma, prot) = SCARG(uap, prot); 590 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 591 SCARG(cma, prot) |= VM_PROT_READ; 592 SCARG(cma, flags) = flags; 593 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 594 SCARG(cma, pad) = 0; 595 } 596 597 int 598 linux_sys_mremap(l, v, retval) 599 struct lwp *l; 600 void *v; 601 register_t *retval; 602 { 603 struct linux_sys_mremap_args /* { 604 syscallarg(void *) old_address; 605 syscallarg(size_t) old_size; 606 syscallarg(size_t) new_size; 607 syscallarg(u_long) flags; 608 } */ *uap = v; 609 struct sys_munmap_args mua; 610 size_t old_size, new_size; 611 int error; 612 613 old_size = round_page(SCARG(uap, old_size)); 614 new_size = round_page(SCARG(uap, new_size)); 615 616 /* 617 * Growing mapped region. 618 */ 619 if (new_size > old_size) { 620 /* 621 * XXX Implement me. What we probably want to do is 622 * XXX dig out the guts of the old mapping, mmap that 623 * XXX object again with the new size, then munmap 624 * XXX the old mapping. 625 */ 626 *retval = 0; 627 return (ENOMEM); 628 } 629 630 /* 631 * Shrinking mapped region. 632 */ 633 if (new_size < old_size) { 634 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) + 635 new_size; 636 SCARG(&mua, len) = old_size - new_size; 637 error = sys_munmap(l, &mua, retval); 638 *retval = error ? 0 : (register_t)SCARG(uap, old_address); 639 return (error); 640 } 641 642 /* 643 * No change. 644 */ 645 *retval = (register_t)SCARG(uap, old_address); 646 return (0); 647 } 648 649 int 650 linux_sys_msync(l, v, retval) 651 struct lwp *l; 652 void *v; 653 register_t *retval; 654 { 655 struct linux_sys_msync_args /* { 656 syscallarg(caddr_t) addr; 657 syscallarg(int) len; 658 syscallarg(int) fl; 659 } */ *uap = v; 660 661 struct sys___msync13_args bma; 662 663 /* flags are ignored */ 664 SCARG(&bma, addr) = SCARG(uap, addr); 665 SCARG(&bma, len) = SCARG(uap, len); 666 SCARG(&bma, flags) = SCARG(uap, fl); 667 668 return sys___msync13(l, &bma, retval); 669 } 670 671 int 672 linux_sys_mprotect(l, v, retval) 673 struct lwp *l; 674 void *v; 675 register_t *retval; 676 { 677 struct linux_sys_mprotect_args /* { 678 syscallarg(const void *) start; 679 syscallarg(unsigned long) len; 680 syscallarg(int) prot; 681 } */ *uap = v; 682 struct vm_map_entry *entry; 683 struct vm_map *map; 684 struct proc *p; 685 vaddr_t end, start, len, stacklim; 686 int prot, grows; 687 688 start = (vaddr_t)SCARG(uap, start); 689 len = round_page(SCARG(uap, len)); 690 prot = SCARG(uap, prot); 691 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 692 prot &= ~grows; 693 end = start + len; 694 695 if (start & PAGE_MASK) 696 return EINVAL; 697 if (end < start) 698 return EINVAL; 699 if (end == start) 700 return 0; 701 702 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 703 return EINVAL; 704 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP)) 705 return EINVAL; 706 707 p = l->l_proc; 708 map = &p->p_vmspace->vm_map; 709 vm_map_lock(map); 710 #ifdef notdef 711 VM_MAP_RANGE_CHECK(map, start, end); 712 #endif 713 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 714 vm_map_unlock(map); 715 return ENOMEM; 716 } 717 718 /* 719 * Approximate the behaviour of PROT_GROWS{DOWN,UP}. 720 */ 721 722 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur; 723 if (grows & LINUX_PROT_GROWSDOWN) { 724 if (USRSTACK - stacklim <= start && start < USRSTACK) { 725 start = USRSTACK - stacklim; 726 } else { 727 start = entry->start; 728 } 729 } else if (grows & LINUX_PROT_GROWSUP) { 730 if (USRSTACK <= end && end < USRSTACK + stacklim) { 731 end = USRSTACK + stacklim; 732 } else { 733 end = entry->end; 734 } 735 } 736 vm_map_unlock(map); 737 return uvm_map_protect(map, start, end, prot, FALSE); 738 } 739 740 /* 741 * This code is partly stolen from src/lib/libc/compat-43/times.c 742 */ 743 744 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 745 746 int 747 linux_sys_times(l, v, retval) 748 struct lwp *l; 749 void *v; 750 register_t *retval; 751 { 752 struct linux_sys_times_args /* { 753 syscallarg(struct times *) tms; 754 } */ *uap = v; 755 struct proc *p = l->l_proc; 756 struct timeval t; 757 int error, s; 758 759 if (SCARG(uap, tms)) { 760 struct linux_tms ltms; 761 struct rusage ru; 762 763 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 764 ltms.ltms_utime = CONVTCK(ru.ru_utime); 765 ltms.ltms_stime = CONVTCK(ru.ru_stime); 766 767 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 768 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 769 770 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 771 return error; 772 } 773 774 s = splclock(); 775 timersub(&time, &boottime, &t); 776 splx(s); 777 778 retval[0] = ((linux_clock_t)(CONVTCK(t))); 779 return 0; 780 } 781 782 #undef CONVTCK 783 784 /* 785 * Linux 'readdir' call. This code is mostly taken from the 786 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 787 * an attempt has been made to keep it a little cleaner (failing 788 * miserably, because of the cruft needed if count 1 is passed). 789 * 790 * The d_off field should contain the offset of the next valid entry, 791 * but in Linux it has the offset of the entry itself. We emulate 792 * that bug here. 793 * 794 * Read in BSD-style entries, convert them, and copy them out. 795 * 796 * Note that this doesn't handle union-mounted filesystems. 797 */ 798 int 799 linux_sys_getdents(l, v, retval) 800 struct lwp *l; 801 void *v; 802 register_t *retval; 803 { 804 struct linux_sys_getdents_args /* { 805 syscallarg(int) fd; 806 syscallarg(struct linux_dirent *) dent; 807 syscallarg(unsigned int) count; 808 } */ *uap = v; 809 struct proc *p = l->l_proc; 810 struct dirent *bdp; 811 struct vnode *vp; 812 caddr_t inp, tbuf; /* BSD-format */ 813 int len, reclen; /* BSD-format */ 814 caddr_t outp; /* Linux-format */ 815 int resid, linux_reclen = 0; /* Linux-format */ 816 struct file *fp; 817 struct uio auio; 818 struct iovec aiov; 819 struct linux_dirent idb; 820 off_t off; /* true file offset */ 821 int buflen, error, eofflag, nbytes, oldcall; 822 struct vattr va; 823 off_t *cookiebuf = NULL, *cookie; 824 int ncookies; 825 826 /* getvnode() will use the descriptor for us */ 827 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 828 return (error); 829 830 if ((fp->f_flag & FREAD) == 0) { 831 error = EBADF; 832 goto out1; 833 } 834 835 vp = (struct vnode *)fp->f_data; 836 if (vp->v_type != VDIR) { 837 error = EINVAL; 838 goto out1; 839 } 840 841 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l))) 842 goto out1; 843 844 nbytes = SCARG(uap, count); 845 if (nbytes == 1) { /* emulating old, broken behaviour */ 846 nbytes = sizeof (idb); 847 buflen = max(va.va_blocksize, nbytes); 848 oldcall = 1; 849 } else { 850 buflen = min(MAXBSIZE, nbytes); 851 if (buflen < va.va_blocksize) 852 buflen = va.va_blocksize; 853 oldcall = 0; 854 } 855 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 856 857 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 858 off = fp->f_offset; 859 again: 860 aiov.iov_base = tbuf; 861 aiov.iov_len = buflen; 862 auio.uio_iov = &aiov; 863 auio.uio_iovcnt = 1; 864 auio.uio_rw = UIO_READ; 865 auio.uio_segflg = UIO_SYSSPACE; 866 auio.uio_lwp = NULL; 867 auio.uio_resid = buflen; 868 auio.uio_offset = off; 869 /* 870 * First we read into the malloc'ed buffer, then 871 * we massage it into user space, one record at a time. 872 */ 873 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 874 &ncookies); 875 if (error) 876 goto out; 877 878 inp = tbuf; 879 outp = (caddr_t)SCARG(uap, dent); 880 resid = nbytes; 881 if ((len = buflen - auio.uio_resid) == 0) 882 goto eof; 883 884 for (cookie = cookiebuf; len > 0; len -= reclen) { 885 bdp = (struct dirent *)inp; 886 reclen = bdp->d_reclen; 887 if (reclen & 3) 888 panic("linux_readdir"); 889 if (bdp->d_fileno == 0) { 890 inp += reclen; /* it is a hole; squish it out */ 891 if (cookie) 892 off = *cookie++; 893 else 894 off += reclen; 895 continue; 896 } 897 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 898 if (reclen > len || resid < linux_reclen) { 899 /* entry too big for buffer, so just stop */ 900 outp++; 901 break; 902 } 903 /* 904 * Massage in place to make a Linux-shaped dirent (otherwise 905 * we have to worry about touching user memory outside of 906 * the copyout() call). 907 */ 908 idb.d_ino = bdp->d_fileno; 909 /* 910 * The old readdir() call misuses the offset and reclen fields. 911 */ 912 if (oldcall) { 913 idb.d_off = (linux_off_t)linux_reclen; 914 idb.d_reclen = (u_short)bdp->d_namlen; 915 } else { 916 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 917 compat_offseterr(vp, "linux_getdents"); 918 error = EINVAL; 919 goto out; 920 } 921 idb.d_off = (linux_off_t)off; 922 idb.d_reclen = (u_short)linux_reclen; 923 } 924 strcpy(idb.d_name, bdp->d_name); 925 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 926 goto out; 927 /* advance past this real entry */ 928 inp += reclen; 929 if (cookie) 930 off = *cookie++; /* each entry points to itself */ 931 else 932 off += reclen; 933 /* advance output past Linux-shaped entry */ 934 outp += linux_reclen; 935 resid -= linux_reclen; 936 if (oldcall) 937 break; 938 } 939 940 /* if we squished out the whole block, try again */ 941 if (outp == (caddr_t)SCARG(uap, dent)) 942 goto again; 943 fp->f_offset = off; /* update the vnode offset */ 944 945 if (oldcall) 946 nbytes = resid + linux_reclen; 947 948 eof: 949 *retval = nbytes - resid; 950 out: 951 VOP_UNLOCK(vp, 0); 952 if (cookiebuf) 953 free(cookiebuf, M_TEMP); 954 free(tbuf, M_TEMP); 955 out1: 956 FILE_UNUSE(fp, l); 957 return error; 958 } 959 960 /* 961 * Even when just using registers to pass arguments to syscalls you can 962 * have 5 of them on the i386. So this newer version of select() does 963 * this. 964 */ 965 int 966 linux_sys_select(l, v, retval) 967 struct lwp *l; 968 void *v; 969 register_t *retval; 970 { 971 struct linux_sys_select_args /* { 972 syscallarg(int) nfds; 973 syscallarg(fd_set *) readfds; 974 syscallarg(fd_set *) writefds; 975 syscallarg(fd_set *) exceptfds; 976 syscallarg(struct timeval *) timeout; 977 } */ *uap = v; 978 979 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 980 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 981 } 982 983 /* 984 * Common code for the old and new versions of select(). A couple of 985 * things are important: 986 * 1) return the amount of time left in the 'timeout' parameter 987 * 2) select never returns ERESTART on Linux, always return EINTR 988 */ 989 int 990 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout) 991 struct lwp *l; 992 register_t *retval; 993 int nfds; 994 fd_set *readfds, *writefds, *exceptfds; 995 struct timeval *timeout; 996 { 997 struct sys_select_args bsa; 998 struct proc *p = l->l_proc; 999 struct timeval tv0, tv1, utv, *tvp; 1000 caddr_t sg; 1001 int error; 1002 1003 SCARG(&bsa, nd) = nfds; 1004 SCARG(&bsa, in) = readfds; 1005 SCARG(&bsa, ou) = writefds; 1006 SCARG(&bsa, ex) = exceptfds; 1007 SCARG(&bsa, tv) = timeout; 1008 1009 /* 1010 * Store current time for computation of the amount of 1011 * time left. 1012 */ 1013 if (timeout) { 1014 if ((error = copyin(timeout, &utv, sizeof(utv)))) 1015 return error; 1016 if (itimerfix(&utv)) { 1017 /* 1018 * The timeval was invalid. Convert it to something 1019 * valid that will act as it does under Linux. 1020 */ 1021 sg = stackgap_init(p, 0); 1022 tvp = stackgap_alloc(p, &sg, sizeof(utv)); 1023 utv.tv_sec += utv.tv_usec / 1000000; 1024 utv.tv_usec %= 1000000; 1025 if (utv.tv_usec < 0) { 1026 utv.tv_sec -= 1; 1027 utv.tv_usec += 1000000; 1028 } 1029 if (utv.tv_sec < 0) 1030 timerclear(&utv); 1031 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1032 return error; 1033 SCARG(&bsa, tv) = tvp; 1034 } 1035 microtime(&tv0); 1036 } 1037 1038 error = sys_select(l, &bsa, retval); 1039 if (error) { 1040 /* 1041 * See fs/select.c in the Linux kernel. Without this, 1042 * Maelstrom doesn't work. 1043 */ 1044 if (error == ERESTART) 1045 error = EINTR; 1046 return error; 1047 } 1048 1049 if (timeout) { 1050 if (*retval) { 1051 /* 1052 * Compute how much time was left of the timeout, 1053 * by subtracting the current time and the time 1054 * before we started the call, and subtracting 1055 * that result from the user-supplied value. 1056 */ 1057 microtime(&tv1); 1058 timersub(&tv1, &tv0, &tv1); 1059 timersub(&utv, &tv1, &utv); 1060 if (utv.tv_sec < 0) 1061 timerclear(&utv); 1062 } else 1063 timerclear(&utv); 1064 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1065 return error; 1066 } 1067 1068 return 0; 1069 } 1070 1071 /* 1072 * Get the process group of a certain process. Look it up 1073 * and return the value. 1074 */ 1075 int 1076 linux_sys_getpgid(l, v, retval) 1077 struct lwp *l; 1078 void *v; 1079 register_t *retval; 1080 { 1081 struct linux_sys_getpgid_args /* { 1082 syscallarg(int) pid; 1083 } */ *uap = v; 1084 struct proc *p = l->l_proc; 1085 struct proc *targp; 1086 1087 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1088 if ((targp = pfind(SCARG(uap, pid))) == 0) 1089 return ESRCH; 1090 } 1091 else 1092 targp = p; 1093 1094 retval[0] = targp->p_pgid; 1095 return 0; 1096 } 1097 1098 /* 1099 * Set the 'personality' (emulation mode) for the current process. Only 1100 * accept the Linux personality here (0). This call is needed because 1101 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1102 * ELF binaries run in Linux mode, not SVR4 mode. 1103 */ 1104 int 1105 linux_sys_personality(l, v, retval) 1106 struct lwp *l; 1107 void *v; 1108 register_t *retval; 1109 { 1110 struct linux_sys_personality_args /* { 1111 syscallarg(int) per; 1112 } */ *uap = v; 1113 1114 if (SCARG(uap, per) != 0) 1115 return EINVAL; 1116 retval[0] = 0; 1117 return 0; 1118 } 1119 1120 #if defined(__i386__) || defined(__m68k__) 1121 /* 1122 * The calls are here because of type conversions. 1123 */ 1124 int 1125 linux_sys_setreuid16(l, v, retval) 1126 struct lwp *l; 1127 void *v; 1128 register_t *retval; 1129 { 1130 struct linux_sys_setreuid16_args /* { 1131 syscallarg(int) ruid; 1132 syscallarg(int) euid; 1133 } */ *uap = v; 1134 struct sys_setreuid_args bsa; 1135 1136 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1137 (uid_t)-1 : SCARG(uap, ruid); 1138 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1139 (uid_t)-1 : SCARG(uap, euid); 1140 1141 return sys_setreuid(l, &bsa, retval); 1142 } 1143 1144 int 1145 linux_sys_setregid16(l, v, retval) 1146 struct lwp *l; 1147 void *v; 1148 register_t *retval; 1149 { 1150 struct linux_sys_setregid16_args /* { 1151 syscallarg(int) rgid; 1152 syscallarg(int) egid; 1153 } */ *uap = v; 1154 struct sys_setregid_args bsa; 1155 1156 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1157 (uid_t)-1 : SCARG(uap, rgid); 1158 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1159 (uid_t)-1 : SCARG(uap, egid); 1160 1161 return sys_setregid(l, &bsa, retval); 1162 } 1163 1164 int 1165 linux_sys_setresuid16(l, v, retval) 1166 struct lwp *l; 1167 void *v; 1168 register_t *retval; 1169 { 1170 struct linux_sys_setresuid16_args /* { 1171 syscallarg(uid_t) ruid; 1172 syscallarg(uid_t) euid; 1173 syscallarg(uid_t) suid; 1174 } */ *uap = v; 1175 struct linux_sys_setresuid16_args lsa; 1176 1177 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1178 (uid_t)-1 : SCARG(uap, ruid); 1179 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1180 (uid_t)-1 : SCARG(uap, euid); 1181 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ? 1182 (uid_t)-1 : SCARG(uap, suid); 1183 1184 return linux_sys_setresuid(l, &lsa, retval); 1185 } 1186 1187 int 1188 linux_sys_setresgid16(l, v, retval) 1189 struct lwp *l; 1190 void *v; 1191 register_t *retval; 1192 { 1193 struct linux_sys_setresgid16_args /* { 1194 syscallarg(gid_t) rgid; 1195 syscallarg(gid_t) egid; 1196 syscallarg(gid_t) sgid; 1197 } */ *uap = v; 1198 struct linux_sys_setresgid16_args lsa; 1199 1200 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1201 (gid_t)-1 : SCARG(uap, rgid); 1202 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1203 (gid_t)-1 : SCARG(uap, egid); 1204 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ? 1205 (gid_t)-1 : SCARG(uap, sgid); 1206 1207 return linux_sys_setresgid(l, &lsa, retval); 1208 } 1209 1210 int 1211 linux_sys_getgroups16(l, v, retval) 1212 struct lwp *l; 1213 void *v; 1214 register_t *retval; 1215 { 1216 struct linux_sys_getgroups16_args /* { 1217 syscallarg(int) gidsetsize; 1218 syscallarg(linux_gid_t *) gidset; 1219 } */ *uap = v; 1220 struct proc *p = l->l_proc; 1221 caddr_t sg; 1222 int n, error, i; 1223 struct sys_getgroups_args bsa; 1224 gid_t *bset, *kbset; 1225 linux_gid_t *lset; 1226 struct pcred *pc = p->p_cred; 1227 1228 n = SCARG(uap, gidsetsize); 1229 if (n < 0) 1230 return EINVAL; 1231 error = 0; 1232 bset = kbset = NULL; 1233 lset = NULL; 1234 if (n > 0) { 1235 n = min(pc->pc_ucred->cr_ngroups, n); 1236 sg = stackgap_init(p, 0); 1237 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1238 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1239 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1240 if (bset == NULL || kbset == NULL || lset == NULL) 1241 return ENOMEM; 1242 SCARG(&bsa, gidsetsize) = n; 1243 SCARG(&bsa, gidset) = bset; 1244 error = sys_getgroups(l, &bsa, retval); 1245 if (error != 0) 1246 goto out; 1247 error = copyin(bset, kbset, n * sizeof (gid_t)); 1248 if (error != 0) 1249 goto out; 1250 for (i = 0; i < n; i++) 1251 lset[i] = (linux_gid_t)kbset[i]; 1252 error = copyout(lset, SCARG(uap, gidset), 1253 n * sizeof (linux_gid_t)); 1254 } else 1255 *retval = pc->pc_ucred->cr_ngroups; 1256 out: 1257 if (kbset != NULL) 1258 free(kbset, M_TEMP); 1259 if (lset != NULL) 1260 free(lset, M_TEMP); 1261 return error; 1262 } 1263 1264 int 1265 linux_sys_setgroups16(l, v, retval) 1266 struct lwp *l; 1267 void *v; 1268 register_t *retval; 1269 { 1270 struct linux_sys_setgroups16_args /* { 1271 syscallarg(int) gidsetsize; 1272 syscallarg(linux_gid_t *) gidset; 1273 } */ *uap = v; 1274 struct proc *p = l->l_proc; 1275 caddr_t sg; 1276 int n; 1277 int error, i; 1278 struct sys_setgroups_args bsa; 1279 gid_t *bset, *kbset; 1280 linux_gid_t *lset; 1281 1282 n = SCARG(uap, gidsetsize); 1283 if (n < 0 || n > NGROUPS) 1284 return EINVAL; 1285 sg = stackgap_init(p, 0); 1286 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1287 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1288 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1289 if (lset == NULL || bset == NULL) 1290 return ENOMEM; 1291 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t)); 1292 if (error != 0) 1293 goto out; 1294 for (i = 0; i < n; i++) 1295 kbset[i] = (gid_t)lset[i]; 1296 error = copyout(kbset, bset, n * sizeof (gid_t)); 1297 if (error != 0) 1298 goto out; 1299 SCARG(&bsa, gidsetsize) = n; 1300 SCARG(&bsa, gidset) = bset; 1301 error = sys_setgroups(l, &bsa, retval); 1302 1303 out: 1304 if (lset != NULL) 1305 free(lset, M_TEMP); 1306 if (kbset != NULL) 1307 free(kbset, M_TEMP); 1308 1309 return error; 1310 } 1311 1312 #endif /* __i386__ || __m68k__ || __amd64__ */ 1313 1314 /* 1315 * We have nonexistent fsuid equal to uid. 1316 * If modification is requested, refuse. 1317 */ 1318 int 1319 linux_sys_setfsuid(l, v, retval) 1320 struct lwp *l; 1321 void *v; 1322 register_t *retval; 1323 { 1324 struct linux_sys_setfsuid_args /* { 1325 syscallarg(uid_t) uid; 1326 } */ *uap = v; 1327 struct proc *p = l->l_proc; 1328 uid_t uid; 1329 1330 uid = SCARG(uap, uid); 1331 if (p->p_cred->p_ruid != uid) 1332 return sys_nosys(l, v, retval); 1333 else 1334 return (0); 1335 } 1336 1337 /* XXX XXX XXX */ 1338 #ifndef alpha 1339 int 1340 linux_sys_getfsuid(l, v, retval) 1341 struct lwp *l; 1342 void *v; 1343 register_t *retval; 1344 { 1345 return sys_getuid(l, v, retval); 1346 } 1347 #endif 1348 1349 int 1350 linux_sys_setresuid(l, v, retval) 1351 struct lwp *l; 1352 void *v; 1353 register_t *retval; 1354 { 1355 struct linux_sys_setresuid_args /* { 1356 syscallarg(uid_t) ruid; 1357 syscallarg(uid_t) euid; 1358 syscallarg(uid_t) suid; 1359 } */ *uap = v; 1360 1361 /* 1362 * Note: These checks are a little different than the NetBSD 1363 * setreuid(2) call performs. This precisely follows the 1364 * behavior of the Linux kernel. 1365 */ 1366 1367 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1368 SCARG(uap, suid), 1369 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1370 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1371 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1372 } 1373 1374 int 1375 linux_sys_getresuid(l, v, retval) 1376 struct lwp *l; 1377 void *v; 1378 register_t *retval; 1379 { 1380 struct linux_sys_getresuid_args /* { 1381 syscallarg(uid_t *) ruid; 1382 syscallarg(uid_t *) euid; 1383 syscallarg(uid_t *) suid; 1384 } */ *uap = v; 1385 struct proc *p = l->l_proc; 1386 struct pcred *pc = p->p_cred; 1387 int error; 1388 1389 /* 1390 * Linux copies these values out to userspace like so: 1391 * 1392 * 1. Copy out ruid. 1393 * 2. If that succeeds, copy out euid. 1394 * 3. If both of those succeed, copy out suid. 1395 */ 1396 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid), 1397 sizeof(uid_t))) != 0) 1398 return (error); 1399 1400 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid), 1401 sizeof(uid_t))) != 0) 1402 return (error); 1403 1404 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t))); 1405 } 1406 1407 int 1408 linux_sys_ptrace(l, v, retval) 1409 struct lwp *l; 1410 void *v; 1411 register_t *retval; 1412 { 1413 struct linux_sys_ptrace_args /* { 1414 i386, m68k, powerpc: T=int 1415 alpha, amd64: T=long 1416 syscallarg(T) request; 1417 syscallarg(T) pid; 1418 syscallarg(T) addr; 1419 syscallarg(T) data; 1420 } */ *uap = v; 1421 const int *ptr; 1422 int request; 1423 int error; 1424 1425 ptr = linux_ptrace_request_map; 1426 request = SCARG(uap, request); 1427 while (*ptr != -1) 1428 if (*ptr++ == request) { 1429 struct sys_ptrace_args pta; 1430 1431 SCARG(&pta, req) = *ptr; 1432 SCARG(&pta, pid) = SCARG(uap, pid); 1433 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr); 1434 SCARG(&pta, data) = SCARG(uap, data); 1435 1436 /* 1437 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1438 * to continue where the process left off previously. 1439 * The same thing is achieved by addr == (caddr_t) 1 1440 * on NetBSD, so rewrite 'addr' appropriately. 1441 */ 1442 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1443 SCARG(&pta, addr) = (caddr_t) 1; 1444 1445 error = sys_ptrace(l, &pta, retval); 1446 if (error) 1447 return error; 1448 switch (request) { 1449 case LINUX_PTRACE_PEEKTEXT: 1450 case LINUX_PTRACE_PEEKDATA: 1451 error = copyout (retval, 1452 (caddr_t)SCARG(uap, data), 1453 sizeof *retval); 1454 *retval = SCARG(uap, data); 1455 break; 1456 default: 1457 break; 1458 } 1459 return error; 1460 } 1461 else 1462 ptr++; 1463 1464 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1465 } 1466 1467 int 1468 linux_sys_reboot(struct lwp *l, void *v, register_t *retval) 1469 { 1470 struct linux_sys_reboot_args /* { 1471 syscallarg(int) magic1; 1472 syscallarg(int) magic2; 1473 syscallarg(int) cmd; 1474 syscallarg(void *) arg; 1475 } */ *uap = v; 1476 struct sys_reboot_args /* { 1477 syscallarg(int) opt; 1478 syscallarg(char *) bootstr; 1479 } */ sra; 1480 struct proc *p = l->l_proc; 1481 int error; 1482 1483 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1484 return(error); 1485 1486 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1487 return(EINVAL); 1488 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1489 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1490 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1491 return(EINVAL); 1492 1493 switch (SCARG(uap, cmd)) { 1494 case LINUX_REBOOT_CMD_RESTART: 1495 SCARG(&sra, opt) = RB_AUTOBOOT; 1496 break; 1497 case LINUX_REBOOT_CMD_HALT: 1498 SCARG(&sra, opt) = RB_HALT; 1499 break; 1500 case LINUX_REBOOT_CMD_POWER_OFF: 1501 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1502 break; 1503 case LINUX_REBOOT_CMD_RESTART2: 1504 /* Reboot with an argument. */ 1505 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1506 SCARG(&sra, bootstr) = SCARG(uap, arg); 1507 break; 1508 case LINUX_REBOOT_CMD_CAD_ON: 1509 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1510 case LINUX_REBOOT_CMD_CAD_OFF: 1511 return(0); 1512 default: 1513 return(EINVAL); 1514 } 1515 1516 return(sys_reboot(l, &sra, retval)); 1517 } 1518 1519 /* 1520 * Copy of compat_12_sys_swapon(). 1521 */ 1522 int 1523 linux_sys_swapon(l, v, retval) 1524 struct lwp *l; 1525 void *v; 1526 register_t *retval; 1527 { 1528 struct sys_swapctl_args ua; 1529 struct linux_sys_swapon_args /* { 1530 syscallarg(const char *) name; 1531 } */ *uap = v; 1532 1533 SCARG(&ua, cmd) = SWAP_ON; 1534 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name)); 1535 SCARG(&ua, misc) = 0; /* priority */ 1536 return (sys_swapctl(l, &ua, retval)); 1537 } 1538 1539 /* 1540 * Stop swapping to the file or block device specified by path. 1541 */ 1542 int 1543 linux_sys_swapoff(l, v, retval) 1544 struct lwp *l; 1545 void *v; 1546 register_t *retval; 1547 { 1548 struct sys_swapctl_args ua; 1549 struct linux_sys_swapoff_args /* { 1550 syscallarg(const char *) path; 1551 } */ *uap = v; 1552 1553 SCARG(&ua, cmd) = SWAP_OFF; 1554 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/ 1555 return (sys_swapctl(l, &ua, retval)); 1556 } 1557 1558 /* 1559 * Copy of compat_09_sys_setdomainname() 1560 */ 1561 /* ARGSUSED */ 1562 int 1563 linux_sys_setdomainname(l, v, retval) 1564 struct lwp *l; 1565 void *v; 1566 register_t *retval; 1567 { 1568 struct linux_sys_setdomainname_args /* { 1569 syscallarg(char *) domainname; 1570 syscallarg(int) len; 1571 } */ *uap = v; 1572 int name[2]; 1573 1574 name[0] = CTL_KERN; 1575 name[1] = KERN_DOMAINNAME; 1576 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1577 SCARG(uap, len), l)); 1578 } 1579 1580 /* 1581 * sysinfo() 1582 */ 1583 /* ARGSUSED */ 1584 int 1585 linux_sys_sysinfo(l, v, retval) 1586 struct lwp *l; 1587 void *v; 1588 register_t *retval; 1589 { 1590 struct linux_sys_sysinfo_args /* { 1591 syscallarg(struct linux_sysinfo *) arg; 1592 } */ *uap = v; 1593 struct linux_sysinfo si; 1594 struct loadavg *la; 1595 1596 si.uptime = time.tv_sec - boottime.tv_sec; 1597 la = &averunnable; 1598 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1599 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1600 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1601 si.totalram = ctob(physmem); 1602 si.freeram = uvmexp.free * uvmexp.pagesize; 1603 si.sharedram = 0; /* XXX */ 1604 si.bufferram = uvmexp.filepages * uvmexp.pagesize; 1605 si.totalswap = uvmexp.swpages * uvmexp.pagesize; 1606 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1607 si.procs = nprocs; 1608 1609 /* The following are only present in newer Linux kernels. */ 1610 si.totalbig = 0; 1611 si.freebig = 0; 1612 si.mem_unit = 1; 1613 1614 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1615 } 1616 1617 #ifdef LINUX_LARGEFILE64 1618 #define bsd_to_linux_rlimit1(l, b, f) \ 1619 (l)->f = ((b)->f == RLIM_INFINITY || \ 1620 ((b)->f & 0x8000000000000000UL) != 0) ? \ 1621 LINUX_RLIM_INFINITY : (b)->f 1622 #else 1623 #define bsd_to_linux_rlimit1(l, b, f) \ 1624 (l)->f = ((b)->f == RLIM_INFINITY || \ 1625 ((b)->f & 0xffffffff00000000ULL) != 0) ? \ 1626 LINUX_RLIM_INFINITY : (int32_t)(b)->f 1627 #endif 1628 #define bsd_to_linux_rlimit(l, b) \ 1629 bsd_to_linux_rlimit1(l, b, rlim_cur); \ 1630 bsd_to_linux_rlimit1(l, b, rlim_max) 1631 1632 #define linux_to_bsd_rlimit1(b, l, f) \ 1633 (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f 1634 #define linux_to_bsd_rlimit(b, l) \ 1635 linux_to_bsd_rlimit1(b, l, rlim_cur); \ 1636 linux_to_bsd_rlimit1(b, l, rlim_max) 1637 1638 static int 1639 linux_to_bsd_limit(lim) 1640 int lim; 1641 { 1642 switch (lim) { 1643 case LINUX_RLIMIT_CPU: 1644 return RLIMIT_CPU; 1645 case LINUX_RLIMIT_FSIZE: 1646 return RLIMIT_FSIZE; 1647 case LINUX_RLIMIT_DATA: 1648 return RLIMIT_DATA; 1649 case LINUX_RLIMIT_STACK: 1650 return RLIMIT_STACK; 1651 case LINUX_RLIMIT_CORE: 1652 return RLIMIT_CORE; 1653 case LINUX_RLIMIT_RSS: 1654 return RLIMIT_RSS; 1655 case LINUX_RLIMIT_NPROC: 1656 return RLIMIT_NPROC; 1657 case LINUX_RLIMIT_NOFILE: 1658 return RLIMIT_NOFILE; 1659 case LINUX_RLIMIT_MEMLOCK: 1660 return RLIMIT_MEMLOCK; 1661 case LINUX_RLIMIT_AS: 1662 case LINUX_RLIMIT_LOCKS: 1663 return -EOPNOTSUPP; 1664 default: 1665 return -EINVAL; 1666 } 1667 } 1668 1669 1670 int 1671 linux_sys_getrlimit(l, v, retval) 1672 struct lwp *l; 1673 void *v; 1674 register_t *retval; 1675 { 1676 struct linux_sys_getrlimit_args /* { 1677 syscallarg(int) which; 1678 #ifdef LINUX_LARGEFILE64 1679 syscallarg(struct rlimit *) rlp; 1680 #else 1681 syscallarg(struct orlimit *) rlp; 1682 #endif 1683 } */ *uap = v; 1684 struct proc *p = l->l_proc; 1685 caddr_t sg = stackgap_init(p, 0); 1686 struct sys_getrlimit_args ap; 1687 struct rlimit rl; 1688 #ifdef LINUX_LARGEFILE64 1689 struct rlimit orl; 1690 #else 1691 struct orlimit orl; 1692 #endif 1693 int error; 1694 1695 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1696 if ((error = SCARG(&ap, which)) < 0) 1697 return -error; 1698 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1699 if ((error = sys_getrlimit(l, &ap, retval)) != 0) 1700 return error; 1701 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0) 1702 return error; 1703 bsd_to_linux_rlimit(&orl, &rl); 1704 1705 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1706 } 1707 1708 int 1709 linux_sys_setrlimit(l, v, retval) 1710 struct lwp *l; 1711 void *v; 1712 register_t *retval; 1713 { 1714 struct linux_sys_setrlimit_args /* { 1715 syscallarg(int) which; 1716 #ifdef LINUX_LARGEFILE64 1717 syscallarg(struct rlimit *) rlp; 1718 #else 1719 syscallarg(struct orlimit *) rlp; 1720 #endif 1721 } */ *uap = v; 1722 struct proc *p = l->l_proc; 1723 caddr_t sg = stackgap_init(p, 0); 1724 struct sys_getrlimit_args ap; 1725 struct rlimit rl; 1726 #ifdef LINUX_LARGEFILE64 1727 struct rlimit orl; 1728 #else 1729 struct orlimit orl; 1730 #endif 1731 int error; 1732 1733 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1734 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1735 if ((error = SCARG(&ap, which)) < 0) 1736 return -error; 1737 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1738 return error; 1739 linux_to_bsd_rlimit(&rl, &orl); 1740 if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0) 1741 return error; 1742 return sys_setrlimit(l, &ap, retval); 1743 } 1744 1745 #if !defined(__mips__) && !defined(__amd64__) 1746 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1747 int 1748 linux_sys_ugetrlimit(l, v, retval) 1749 struct lwp *l; 1750 void *v; 1751 register_t *retval; 1752 { 1753 return linux_sys_getrlimit(l, v, retval); 1754 } 1755 #endif 1756 1757 /* 1758 * This gets called for unsupported syscalls. The difference to sys_nosys() 1759 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1760 * This is the way Linux does it and glibc depends on this behaviour. 1761 */ 1762 int 1763 linux_sys_nosys(l, v, retval) 1764 struct lwp *l; 1765 void *v; 1766 register_t *retval; 1767 { 1768 return (ENOSYS); 1769 } 1770