1 /* $NetBSD: linux_misc.c,v 1.150 2006/02/09 19:18:56 manu 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 1.150 2006/02/09 19:18:56 manu 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 #ifndef COMPAT_LINUX32 117 #include <compat/linux/common/linux_limit.h> 118 #endif 119 #include <compat/linux/common/linux_ptrace.h> 120 #include <compat/linux/common/linux_reboot.h> 121 #include <compat/linux/common/linux_emuldata.h> 122 123 #ifndef COMPAT_LINUX32 124 const int linux_ptrace_request_map[] = { 125 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 126 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 127 LINUX_PTRACE_PEEKDATA, PT_READ_D, 128 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 129 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 130 LINUX_PTRACE_CONT, PT_CONTINUE, 131 LINUX_PTRACE_KILL, PT_KILL, 132 LINUX_PTRACE_ATTACH, PT_ATTACH, 133 LINUX_PTRACE_DETACH, PT_DETACH, 134 # ifdef PT_STEP 135 LINUX_PTRACE_SINGLESTEP, PT_STEP, 136 # endif 137 -1 138 }; 139 140 const struct linux_mnttypes linux_fstypes[] = { 141 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 142 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 143 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 144 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 145 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 146 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_PORTAL, LINUX_DEFAULT_SUPER_MAGIC }, 148 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 149 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 150 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 151 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 152 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 153 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 154 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 155 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 156 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 157 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 158 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 159 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 160 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 161 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 162 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC }, 163 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC }, 164 { MOUNT_TMPFS, LINUX_DEFAULT_SUPER_MAGIC } 165 }; 166 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 167 168 # ifdef DEBUG_LINUX 169 #define DPRINTF(a) uprintf a 170 # else 171 #define DPRINTF(a) 172 # endif 173 174 /* Local linux_misc.c functions: */ 175 # ifndef __amd64__ 176 static void bsd_to_linux_statfs __P((const struct statvfs *, 177 struct linux_statfs *)); 178 # endif 179 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *, 180 const struct linux_sys_mmap_args *)); 181 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *, 182 register_t *, off_t)); 183 184 185 /* 186 * The information on a terminated (or stopped) process needs 187 * to be converted in order for Linux binaries to get a valid signal 188 * number out of it. 189 */ 190 void 191 bsd_to_linux_wstat(st) 192 int *st; 193 { 194 195 int sig; 196 197 if (WIFSIGNALED(*st)) { 198 sig = WTERMSIG(*st); 199 if (sig >= 0 && sig < NSIG) 200 *st= (*st& ~0177) | native_to_linux_signo[sig]; 201 } else if (WIFSTOPPED(*st)) { 202 sig = WSTOPSIG(*st); 203 if (sig >= 0 && sig < NSIG) 204 *st = (*st & ~0xff00) | 205 (native_to_linux_signo[sig] << 8); 206 } 207 } 208 209 /* 210 * wait4(2). Passed on to the NetBSD call, surrounded by code to 211 * reserve some space for a NetBSD-style wait status, and converting 212 * it to what Linux wants. 213 */ 214 int 215 linux_sys_wait4(l, v, retval) 216 struct lwp *l; 217 void *v; 218 register_t *retval; 219 { 220 struct linux_sys_wait4_args /* { 221 syscallarg(int) pid; 222 syscallarg(int *) status; 223 syscallarg(int) options; 224 syscallarg(struct rusage *) rusage; 225 } */ *uap = v; 226 struct proc *p = l->l_proc; 227 struct sys_wait4_args w4a; 228 int error, *status, tstat, options, linux_options; 229 caddr_t sg; 230 231 if (SCARG(uap, status) != NULL) { 232 sg = stackgap_init(p, 0); 233 status = (int *) stackgap_alloc(p, &sg, sizeof *status); 234 } else 235 status = NULL; 236 237 linux_options = SCARG(uap, options); 238 options = 0; 239 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS)) 240 return (EINVAL); 241 242 if (linux_options & LINUX_WAIT4_WNOHANG) 243 options |= WNOHANG; 244 if (linux_options & LINUX_WAIT4_WUNTRACED) 245 options |= WUNTRACED; 246 if (linux_options & LINUX_WAIT4_WALL) 247 options |= WALLSIG; 248 if (linux_options & LINUX_WAIT4_WCLONE) 249 options |= WALTSIG; 250 # ifdef DIAGNOSTIC 251 if (linux_options & LINUX_WAIT4_WNOTHREAD) 252 printf("WARNING: %s: linux process %d.%d called " 253 "waitpid with __WNOTHREAD set!", 254 __FILE__, p->p_pid, l->l_lid); 255 256 # endif 257 258 SCARG(&w4a, pid) = SCARG(uap, pid); 259 SCARG(&w4a, status) = status; 260 SCARG(&w4a, options) = options; 261 SCARG(&w4a, rusage) = SCARG(uap, rusage); 262 263 if ((error = sys_wait4(l, &w4a, retval))) 264 return error; 265 266 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD); 267 268 if (status != NULL) { 269 if ((error = copyin(status, &tstat, sizeof tstat))) 270 return error; 271 272 bsd_to_linux_wstat(&tstat); 273 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 274 } 275 276 return 0; 277 } 278 279 /* 280 * Linux brk(2). The check if the new address is >= the old one is 281 * done in the kernel in Linux. NetBSD does it in the library. 282 */ 283 int 284 linux_sys_brk(l, v, retval) 285 struct lwp *l; 286 void *v; 287 register_t *retval; 288 { 289 struct linux_sys_brk_args /* { 290 syscallarg(char *) nsize; 291 } */ *uap = v; 292 struct proc *p = l->l_proc; 293 char *nbrk = SCARG(uap, nsize); 294 struct sys_obreak_args oba; 295 struct vmspace *vm = p->p_vmspace; 296 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata; 297 298 SCARG(&oba, nsize) = nbrk; 299 300 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0) 301 ed->s->p_break = (char*)nbrk; 302 else 303 nbrk = ed->s->p_break; 304 305 retval[0] = (register_t)nbrk; 306 307 return 0; 308 } 309 310 # ifndef __amd64__ 311 /* 312 * Convert NetBSD statvfs structure to Linux statfs structure. 313 * Linux doesn't have f_flag, and we can't set f_frsize due 314 * to glibc statvfs() bug (see below). 315 */ 316 static void 317 bsd_to_linux_statfs(bsp, lsp) 318 const struct statvfs *bsp; 319 struct linux_statfs *lsp; 320 { 321 int i; 322 323 for (i = 0; i < linux_fstypes_cnt; i++) { 324 if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) { 325 lsp->l_ftype = linux_fstypes[i].linux; 326 break; 327 } 328 } 329 330 if (i == linux_fstypes_cnt) { 331 DPRINTF(("unhandled fstype in linux emulation: %s\n", 332 bsp->f_fstypename)); 333 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC; 334 } 335 336 /* 337 * The sizes are expressed in number of blocks. The block 338 * size used for the size is f_frsize for POSIX-compliant 339 * statvfs. Linux statfs uses f_bsize as the block size 340 * (f_frsize used to not be available in Linux struct statfs). 341 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block 342 * counts for different f_frsize if f_frsize is provided by the kernel. 343 * POSIX conforming apps thus get wrong size if f_frsize 344 * is different to f_bsize. Thus, we just pretend we don't 345 * support f_frsize. 346 */ 347 348 lsp->l_fbsize = bsp->f_frsize; 349 lsp->l_ffrsize = 0; /* compat */ 350 lsp->l_fblocks = bsp->f_blocks; 351 lsp->l_fbfree = bsp->f_bfree; 352 lsp->l_fbavail = bsp->f_bavail; 353 lsp->l_ffiles = bsp->f_files; 354 lsp->l_fffree = bsp->f_ffree; 355 /* Linux sets the fsid to 0..., we don't */ 356 lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0]; 357 lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1]; 358 lsp->l_fnamelen = bsp->f_namemax; 359 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare)); 360 } 361 362 /* 363 * Implement the fs stat functions. Straightforward. 364 */ 365 int 366 linux_sys_statfs(l, v, retval) 367 struct lwp *l; 368 void *v; 369 register_t *retval; 370 { 371 struct linux_sys_statfs_args /* { 372 syscallarg(const char *) path; 373 syscallarg(struct linux_statfs *) sp; 374 } */ *uap = v; 375 struct proc *p = l->l_proc; 376 struct statvfs btmp, *bsp; 377 struct linux_statfs ltmp; 378 struct sys_statvfs1_args bsa; 379 caddr_t sg; 380 int error; 381 382 sg = stackgap_init(p, 0); 383 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 384 385 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path)); 386 387 SCARG(&bsa, path) = SCARG(uap, path); 388 SCARG(&bsa, buf) = bsp; 389 SCARG(&bsa, flags) = ST_WAIT; 390 391 if ((error = sys_statvfs1(l, &bsa, retval))) 392 return error; 393 394 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 395 return error; 396 397 bsd_to_linux_statfs(&btmp, <mp); 398 399 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 400 } 401 402 int 403 linux_sys_fstatfs(l, v, retval) 404 struct lwp *l; 405 void *v; 406 register_t *retval; 407 { 408 struct linux_sys_fstatfs_args /* { 409 syscallarg(int) fd; 410 syscallarg(struct linux_statfs *) sp; 411 } */ *uap = v; 412 struct proc *p = l->l_proc; 413 struct statvfs btmp, *bsp; 414 struct linux_statfs ltmp; 415 struct sys_fstatvfs1_args bsa; 416 caddr_t sg; 417 int error; 418 419 sg = stackgap_init(p, 0); 420 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 421 422 SCARG(&bsa, fd) = SCARG(uap, fd); 423 SCARG(&bsa, buf) = bsp; 424 SCARG(&bsa, flags) = ST_WAIT; 425 426 if ((error = sys_fstatvfs1(l, &bsa, retval))) 427 return error; 428 429 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 430 return error; 431 432 bsd_to_linux_statfs(&btmp, <mp); 433 434 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 435 } 436 # endif /* !__amd64__ */ 437 438 /* 439 * uname(). Just copy the info from the various strings stored in the 440 * kernel, and put it in the Linux utsname structure. That structure 441 * is almost the same as the NetBSD one, only it has fields 65 characters 442 * long, and an extra domainname field. 443 */ 444 int 445 linux_sys_uname(l, v, retval) 446 struct lwp *l; 447 void *v; 448 register_t *retval; 449 { 450 struct linux_sys_uname_args /* { 451 syscallarg(struct linux_utsname *) up; 452 } */ *uap = v; 453 struct linux_utsname luts; 454 455 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 456 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 457 strncpy(luts.l_release, linux_release, sizeof(luts.l_release)); 458 strncpy(luts.l_version, linux_version, sizeof(luts.l_version)); 459 # ifdef LINUX_UNAME_ARCH 460 strncpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine)); 461 # else 462 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 463 # endif 464 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 465 466 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 467 } 468 469 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 470 /* Used indirectly on: arm, i386, m68k */ 471 472 /* 473 * New type Linux mmap call. 474 * Only called directly on machines with >= 6 free regs. 475 */ 476 int 477 linux_sys_mmap(l, v, retval) 478 struct lwp *l; 479 void *v; 480 register_t *retval; 481 { 482 struct linux_sys_mmap_args /* { 483 syscallarg(unsigned long) addr; 484 syscallarg(size_t) len; 485 syscallarg(int) prot; 486 syscallarg(int) flags; 487 syscallarg(int) fd; 488 syscallarg(linux_off_t) offset; 489 } */ *uap = v; 490 491 if (SCARG(uap, offset) & PAGE_MASK) 492 return EINVAL; 493 494 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 495 } 496 497 /* 498 * Guts of most architectures' mmap64() implementations. This shares 499 * its list of arguments with linux_sys_mmap(). 500 * 501 * The difference in linux_sys_mmap2() is that "offset" is actually 502 * (offset / pagesize), not an absolute byte count. This translation 503 * to pagesize offsets is done inside glibc between the mmap64() call 504 * point, and the actual syscall. 505 */ 506 int 507 linux_sys_mmap2(l, v, retval) 508 struct lwp *l; 509 void *v; 510 register_t *retval; 511 { 512 struct linux_sys_mmap2_args /* { 513 syscallarg(unsigned long) addr; 514 syscallarg(size_t) len; 515 syscallarg(int) prot; 516 syscallarg(int) flags; 517 syscallarg(int) fd; 518 syscallarg(linux_off_t) offset; 519 } */ *uap = v; 520 521 return linux_mmap(l, uap, retval, 522 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 523 } 524 525 /* 526 * Massage arguments and call system mmap(2). 527 */ 528 static int 529 linux_mmap(l, uap, retval, offset) 530 struct lwp *l; 531 struct linux_sys_mmap_args *uap; 532 register_t *retval; 533 off_t offset; 534 { 535 struct sys_mmap_args cma; 536 int error; 537 size_t mmoff=0; 538 539 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 540 /* 541 * Request for stack-like memory segment. On linux, this 542 * works by mmap()ping (small) segment, which is automatically 543 * extended when page fault happens below the currently 544 * allocated area. We emulate this by allocating (typically 545 * bigger) segment sized at current stack size limit, and 546 * offsetting the requested and returned address accordingly. 547 * Since physical pages are only allocated on-demand, this 548 * is effectively identical. 549 */ 550 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 551 552 if (SCARG(uap, len) < ssl) { 553 /* Compute the address offset */ 554 mmoff = round_page(ssl) - SCARG(uap, len); 555 556 if (SCARG(uap, addr)) 557 SCARG(uap, addr) -= mmoff; 558 559 SCARG(uap, len) = (size_t) ssl; 560 } 561 } 562 563 linux_to_bsd_mmap_args(&cma, uap); 564 SCARG(&cma, pos) = offset; 565 566 error = sys_mmap(l, &cma, retval); 567 if (error) 568 return (error); 569 570 /* Shift the returned address for stack-like segment if necessary */ 571 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff) 572 retval[0] += mmoff; 573 574 return (0); 575 } 576 577 static void 578 linux_to_bsd_mmap_args(cma, uap) 579 struct sys_mmap_args *cma; 580 const struct linux_sys_mmap_args *uap; 581 { 582 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 583 584 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 585 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 586 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 587 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 588 /* XXX XAX ERH: Any other flags here? There are more defined... */ 589 590 SCARG(cma, addr) = (void *)SCARG(uap, addr); 591 SCARG(cma, len) = SCARG(uap, len); 592 SCARG(cma, prot) = SCARG(uap, prot); 593 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 594 SCARG(cma, prot) |= VM_PROT_READ; 595 SCARG(cma, flags) = flags; 596 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 597 SCARG(cma, pad) = 0; 598 } 599 600 #define LINUX_MREMAP_MAYMOVE 1 601 #define LINUX_MREMAP_FIXED 2 602 603 int 604 linux_sys_mremap(l, v, retval) 605 struct lwp *l; 606 void *v; 607 register_t *retval; 608 { 609 struct linux_sys_mremap_args /* { 610 syscallarg(void *) old_address; 611 syscallarg(size_t) old_size; 612 syscallarg(size_t) new_size; 613 syscallarg(u_long) flags; 614 } */ *uap = v; 615 616 struct proc *p; 617 struct vm_map *map; 618 vaddr_t oldva; 619 vaddr_t newva; 620 size_t oldsize; 621 size_t newsize; 622 int flags; 623 int uvmflags; 624 int error; 625 626 flags = SCARG(uap, flags); 627 oldva = (vaddr_t)SCARG(uap, old_address); 628 oldsize = round_page(SCARG(uap, old_size)); 629 newsize = round_page(SCARG(uap, new_size)); 630 if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) { 631 error = EINVAL; 632 goto done; 633 } 634 if ((flags & LINUX_MREMAP_FIXED) != 0) { 635 if ((flags & LINUX_MREMAP_MAYMOVE) == 0) { 636 error = EINVAL; 637 goto done; 638 } 639 #if 0 /* notyet */ 640 newva = SCARG(uap, new_address); 641 uvmflags = UVM_MREMAP_FIXED; 642 #else /* notyet */ 643 error = EOPNOTSUPP; 644 goto done; 645 #endif /* notyet */ 646 } else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) { 647 uvmflags = 0; 648 } else { 649 newva = oldva; 650 uvmflags = UVM_MREMAP_FIXED; 651 } 652 p = l->l_proc; 653 map = &p->p_vmspace->vm_map; 654 error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p, 655 uvmflags); 656 657 done: 658 *retval = (error != 0) ? 0 : (register_t)newva; 659 return error; 660 } 661 662 int 663 linux_sys_msync(l, v, retval) 664 struct lwp *l; 665 void *v; 666 register_t *retval; 667 { 668 struct linux_sys_msync_args /* { 669 syscallarg(caddr_t) addr; 670 syscallarg(int) len; 671 syscallarg(int) fl; 672 } */ *uap = v; 673 674 struct sys___msync13_args bma; 675 676 /* flags are ignored */ 677 SCARG(&bma, addr) = SCARG(uap, addr); 678 SCARG(&bma, len) = SCARG(uap, len); 679 SCARG(&bma, flags) = SCARG(uap, fl); 680 681 return sys___msync13(l, &bma, retval); 682 } 683 684 int 685 linux_sys_mprotect(l, v, retval) 686 struct lwp *l; 687 void *v; 688 register_t *retval; 689 { 690 struct linux_sys_mprotect_args /* { 691 syscallarg(const void *) start; 692 syscallarg(unsigned long) len; 693 syscallarg(int) prot; 694 } */ *uap = v; 695 struct vm_map_entry *entry; 696 struct vm_map *map; 697 struct proc *p; 698 vaddr_t end, start, len, stacklim; 699 int prot, grows; 700 701 start = (vaddr_t)SCARG(uap, start); 702 len = round_page(SCARG(uap, len)); 703 prot = SCARG(uap, prot); 704 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 705 prot &= ~grows; 706 end = start + len; 707 708 if (start & PAGE_MASK) 709 return EINVAL; 710 if (end < start) 711 return EINVAL; 712 if (end == start) 713 return 0; 714 715 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 716 return EINVAL; 717 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP)) 718 return EINVAL; 719 720 p = l->l_proc; 721 map = &p->p_vmspace->vm_map; 722 vm_map_lock(map); 723 # ifdef notdef 724 VM_MAP_RANGE_CHECK(map, start, end); 725 # endif 726 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 727 vm_map_unlock(map); 728 return ENOMEM; 729 } 730 731 /* 732 * Approximate the behaviour of PROT_GROWS{DOWN,UP}. 733 */ 734 735 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur; 736 if (grows & LINUX_PROT_GROWSDOWN) { 737 if (USRSTACK - stacklim <= start && start < USRSTACK) { 738 start = USRSTACK - stacklim; 739 } else { 740 start = entry->start; 741 } 742 } else if (grows & LINUX_PROT_GROWSUP) { 743 if (USRSTACK <= end && end < USRSTACK + stacklim) { 744 end = USRSTACK + stacklim; 745 } else { 746 end = entry->end; 747 } 748 } 749 vm_map_unlock(map); 750 return uvm_map_protect(map, start, end, prot, FALSE); 751 } 752 753 /* 754 * This code is partly stolen from src/lib/libc/compat-43/times.c 755 */ 756 757 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 758 759 int 760 linux_sys_times(l, v, retval) 761 struct lwp *l; 762 void *v; 763 register_t *retval; 764 { 765 struct linux_sys_times_args /* { 766 syscallarg(struct times *) tms; 767 } */ *uap = v; 768 struct proc *p = l->l_proc; 769 struct timeval t; 770 int error, s; 771 772 if (SCARG(uap, tms)) { 773 struct linux_tms ltms; 774 struct rusage ru; 775 776 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 777 ltms.ltms_utime = CONVTCK(ru.ru_utime); 778 ltms.ltms_stime = CONVTCK(ru.ru_stime); 779 780 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 781 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 782 783 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 784 return error; 785 } 786 787 s = splclock(); 788 timersub(&time, &boottime, &t); 789 splx(s); 790 791 retval[0] = ((linux_clock_t)(CONVTCK(t))); 792 return 0; 793 } 794 795 #undef CONVTCK 796 797 /* 798 * Linux 'readdir' call. This code is mostly taken from the 799 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 800 * an attempt has been made to keep it a little cleaner (failing 801 * miserably, because of the cruft needed if count 1 is passed). 802 * 803 * The d_off field should contain the offset of the next valid entry, 804 * but in Linux it has the offset of the entry itself. We emulate 805 * that bug here. 806 * 807 * Read in BSD-style entries, convert them, and copy them out. 808 * 809 * Note that this doesn't handle union-mounted filesystems. 810 */ 811 int 812 linux_sys_getdents(l, v, retval) 813 struct lwp *l; 814 void *v; 815 register_t *retval; 816 { 817 struct linux_sys_getdents_args /* { 818 syscallarg(int) fd; 819 syscallarg(struct linux_dirent *) dent; 820 syscallarg(unsigned int) count; 821 } */ *uap = v; 822 struct proc *p = l->l_proc; 823 struct dirent *bdp; 824 struct vnode *vp; 825 caddr_t inp, tbuf; /* BSD-format */ 826 int len, reclen; /* BSD-format */ 827 caddr_t outp; /* Linux-format */ 828 int resid, linux_reclen = 0; /* Linux-format */ 829 struct file *fp; 830 struct uio auio; 831 struct iovec aiov; 832 struct linux_dirent idb; 833 off_t off; /* true file offset */ 834 int buflen, error, eofflag, nbytes, oldcall; 835 struct vattr va; 836 off_t *cookiebuf = NULL, *cookie; 837 int ncookies; 838 839 /* getvnode() will use the descriptor for us */ 840 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 841 return (error); 842 843 if ((fp->f_flag & FREAD) == 0) { 844 error = EBADF; 845 goto out1; 846 } 847 848 vp = (struct vnode *)fp->f_data; 849 if (vp->v_type != VDIR) { 850 error = EINVAL; 851 goto out1; 852 } 853 854 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l))) 855 goto out1; 856 857 nbytes = SCARG(uap, count); 858 if (nbytes == 1) { /* emulating old, broken behaviour */ 859 nbytes = sizeof (idb); 860 buflen = max(va.va_blocksize, nbytes); 861 oldcall = 1; 862 } else { 863 buflen = min(MAXBSIZE, nbytes); 864 if (buflen < va.va_blocksize) 865 buflen = va.va_blocksize; 866 oldcall = 0; 867 } 868 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 869 870 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 871 off = fp->f_offset; 872 again: 873 aiov.iov_base = tbuf; 874 aiov.iov_len = buflen; 875 auio.uio_iov = &aiov; 876 auio.uio_iovcnt = 1; 877 auio.uio_rw = UIO_READ; 878 auio.uio_segflg = UIO_SYSSPACE; 879 auio.uio_lwp = NULL; 880 auio.uio_resid = buflen; 881 auio.uio_offset = off; 882 /* 883 * First we read into the malloc'ed buffer, then 884 * we massage it into user space, one record at a time. 885 */ 886 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 887 &ncookies); 888 if (error) 889 goto out; 890 891 inp = tbuf; 892 outp = (caddr_t)SCARG(uap, dent); 893 resid = nbytes; 894 if ((len = buflen - auio.uio_resid) == 0) 895 goto eof; 896 897 for (cookie = cookiebuf; len > 0; len -= reclen) { 898 bdp = (struct dirent *)inp; 899 reclen = bdp->d_reclen; 900 if (reclen & 3) 901 panic("linux_readdir"); 902 if (bdp->d_fileno == 0) { 903 inp += reclen; /* it is a hole; squish it out */ 904 if (cookie) 905 off = *cookie++; 906 else 907 off += reclen; 908 continue; 909 } 910 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 911 if (reclen > len || resid < linux_reclen) { 912 /* entry too big for buffer, so just stop */ 913 outp++; 914 break; 915 } 916 /* 917 * Massage in place to make a Linux-shaped dirent (otherwise 918 * we have to worry about touching user memory outside of 919 * the copyout() call). 920 */ 921 idb.d_ino = bdp->d_fileno; 922 /* 923 * The old readdir() call misuses the offset and reclen fields. 924 */ 925 if (oldcall) { 926 idb.d_off = (linux_off_t)linux_reclen; 927 idb.d_reclen = (u_short)bdp->d_namlen; 928 } else { 929 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 930 compat_offseterr(vp, "linux_getdents"); 931 error = EINVAL; 932 goto out; 933 } 934 idb.d_off = (linux_off_t)off; 935 idb.d_reclen = (u_short)linux_reclen; 936 } 937 strcpy(idb.d_name, bdp->d_name); 938 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 939 goto out; 940 /* advance past this real entry */ 941 inp += reclen; 942 if (cookie) 943 off = *cookie++; /* each entry points to itself */ 944 else 945 off += reclen; 946 /* advance output past Linux-shaped entry */ 947 outp += linux_reclen; 948 resid -= linux_reclen; 949 if (oldcall) 950 break; 951 } 952 953 /* if we squished out the whole block, try again */ 954 if (outp == (caddr_t)SCARG(uap, dent)) 955 goto again; 956 fp->f_offset = off; /* update the vnode offset */ 957 958 if (oldcall) 959 nbytes = resid + linux_reclen; 960 961 eof: 962 *retval = nbytes - resid; 963 out: 964 VOP_UNLOCK(vp, 0); 965 if (cookiebuf) 966 free(cookiebuf, M_TEMP); 967 free(tbuf, M_TEMP); 968 out1: 969 FILE_UNUSE(fp, l); 970 return error; 971 } 972 973 /* 974 * Even when just using registers to pass arguments to syscalls you can 975 * have 5 of them on the i386. So this newer version of select() does 976 * this. 977 */ 978 int 979 linux_sys_select(l, v, retval) 980 struct lwp *l; 981 void *v; 982 register_t *retval; 983 { 984 struct linux_sys_select_args /* { 985 syscallarg(int) nfds; 986 syscallarg(fd_set *) readfds; 987 syscallarg(fd_set *) writefds; 988 syscallarg(fd_set *) exceptfds; 989 syscallarg(struct timeval *) timeout; 990 } */ *uap = v; 991 992 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 993 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 994 } 995 996 /* 997 * Common code for the old and new versions of select(). A couple of 998 * things are important: 999 * 1) return the amount of time left in the 'timeout' parameter 1000 * 2) select never returns ERESTART on Linux, always return EINTR 1001 */ 1002 int 1003 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout) 1004 struct lwp *l; 1005 register_t *retval; 1006 int nfds; 1007 fd_set *readfds, *writefds, *exceptfds; 1008 struct timeval *timeout; 1009 { 1010 struct sys_select_args bsa; 1011 struct proc *p = l->l_proc; 1012 struct timeval tv0, tv1, utv, *tvp; 1013 caddr_t sg; 1014 int error; 1015 1016 SCARG(&bsa, nd) = nfds; 1017 SCARG(&bsa, in) = readfds; 1018 SCARG(&bsa, ou) = writefds; 1019 SCARG(&bsa, ex) = exceptfds; 1020 SCARG(&bsa, tv) = timeout; 1021 1022 /* 1023 * Store current time for computation of the amount of 1024 * time left. 1025 */ 1026 if (timeout) { 1027 if ((error = copyin(timeout, &utv, sizeof(utv)))) 1028 return error; 1029 if (itimerfix(&utv)) { 1030 /* 1031 * The timeval was invalid. Convert it to something 1032 * valid that will act as it does under Linux. 1033 */ 1034 sg = stackgap_init(p, 0); 1035 tvp = stackgap_alloc(p, &sg, sizeof(utv)); 1036 utv.tv_sec += utv.tv_usec / 1000000; 1037 utv.tv_usec %= 1000000; 1038 if (utv.tv_usec < 0) { 1039 utv.tv_sec -= 1; 1040 utv.tv_usec += 1000000; 1041 } 1042 if (utv.tv_sec < 0) 1043 timerclear(&utv); 1044 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1045 return error; 1046 SCARG(&bsa, tv) = tvp; 1047 } 1048 microtime(&tv0); 1049 } 1050 1051 error = sys_select(l, &bsa, retval); 1052 if (error) { 1053 /* 1054 * See fs/select.c in the Linux kernel. Without this, 1055 * Maelstrom doesn't work. 1056 */ 1057 if (error == ERESTART) 1058 error = EINTR; 1059 return error; 1060 } 1061 1062 if (timeout) { 1063 if (*retval) { 1064 /* 1065 * Compute how much time was left of the timeout, 1066 * by subtracting the current time and the time 1067 * before we started the call, and subtracting 1068 * that result from the user-supplied value. 1069 */ 1070 microtime(&tv1); 1071 timersub(&tv1, &tv0, &tv1); 1072 timersub(&utv, &tv1, &utv); 1073 if (utv.tv_sec < 0) 1074 timerclear(&utv); 1075 } else 1076 timerclear(&utv); 1077 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1078 return error; 1079 } 1080 1081 return 0; 1082 } 1083 1084 /* 1085 * Get the process group of a certain process. Look it up 1086 * and return the value. 1087 */ 1088 int 1089 linux_sys_getpgid(l, v, retval) 1090 struct lwp *l; 1091 void *v; 1092 register_t *retval; 1093 { 1094 struct linux_sys_getpgid_args /* { 1095 syscallarg(int) pid; 1096 } */ *uap = v; 1097 struct proc *p = l->l_proc; 1098 struct proc *targp; 1099 1100 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1101 if ((targp = pfind(SCARG(uap, pid))) == 0) 1102 return ESRCH; 1103 } 1104 else 1105 targp = p; 1106 1107 retval[0] = targp->p_pgid; 1108 return 0; 1109 } 1110 1111 /* 1112 * Set the 'personality' (emulation mode) for the current process. Only 1113 * accept the Linux personality here (0). This call is needed because 1114 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1115 * ELF binaries run in Linux mode, not SVR4 mode. 1116 */ 1117 int 1118 linux_sys_personality(l, v, retval) 1119 struct lwp *l; 1120 void *v; 1121 register_t *retval; 1122 { 1123 struct linux_sys_personality_args /* { 1124 syscallarg(int) per; 1125 } */ *uap = v; 1126 1127 if (SCARG(uap, per) != 0) 1128 return EINVAL; 1129 retval[0] = 0; 1130 return 0; 1131 } 1132 #endif /* !COMPAT_LINUX32 */ 1133 1134 #if defined(__i386__) || defined(__m68k__) || defined(COMPAT_LINUX32) 1135 /* 1136 * The calls are here because of type conversions. 1137 */ 1138 int 1139 linux_sys_setreuid16(l, v, retval) 1140 struct lwp *l; 1141 void *v; 1142 register_t *retval; 1143 { 1144 struct linux_sys_setreuid16_args /* { 1145 syscallarg(int) ruid; 1146 syscallarg(int) euid; 1147 } */ *uap = v; 1148 struct sys_setreuid_args bsa; 1149 1150 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1151 (uid_t)-1 : SCARG(uap, ruid); 1152 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1153 (uid_t)-1 : SCARG(uap, euid); 1154 1155 return sys_setreuid(l, &bsa, retval); 1156 } 1157 1158 int 1159 linux_sys_setregid16(l, v, retval) 1160 struct lwp *l; 1161 void *v; 1162 register_t *retval; 1163 { 1164 struct linux_sys_setregid16_args /* { 1165 syscallarg(int) rgid; 1166 syscallarg(int) egid; 1167 } */ *uap = v; 1168 struct sys_setregid_args bsa; 1169 1170 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1171 (uid_t)-1 : SCARG(uap, rgid); 1172 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1173 (uid_t)-1 : SCARG(uap, egid); 1174 1175 return sys_setregid(l, &bsa, retval); 1176 } 1177 1178 int 1179 linux_sys_setresuid16(l, v, retval) 1180 struct lwp *l; 1181 void *v; 1182 register_t *retval; 1183 { 1184 struct linux_sys_setresuid16_args /* { 1185 syscallarg(uid_t) ruid; 1186 syscallarg(uid_t) euid; 1187 syscallarg(uid_t) suid; 1188 } */ *uap = v; 1189 struct linux_sys_setresuid16_args lsa; 1190 1191 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1192 (uid_t)-1 : SCARG(uap, ruid); 1193 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1194 (uid_t)-1 : SCARG(uap, euid); 1195 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ? 1196 (uid_t)-1 : SCARG(uap, suid); 1197 1198 return linux_sys_setresuid(l, &lsa, retval); 1199 } 1200 1201 int 1202 linux_sys_setresgid16(l, v, retval) 1203 struct lwp *l; 1204 void *v; 1205 register_t *retval; 1206 { 1207 struct linux_sys_setresgid16_args /* { 1208 syscallarg(gid_t) rgid; 1209 syscallarg(gid_t) egid; 1210 syscallarg(gid_t) sgid; 1211 } */ *uap = v; 1212 struct linux_sys_setresgid16_args lsa; 1213 1214 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1215 (gid_t)-1 : SCARG(uap, rgid); 1216 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1217 (gid_t)-1 : SCARG(uap, egid); 1218 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ? 1219 (gid_t)-1 : SCARG(uap, sgid); 1220 1221 return linux_sys_setresgid(l, &lsa, retval); 1222 } 1223 1224 int 1225 linux_sys_getgroups16(l, v, retval) 1226 struct lwp *l; 1227 void *v; 1228 register_t *retval; 1229 { 1230 struct linux_sys_getgroups16_args /* { 1231 syscallarg(int) gidsetsize; 1232 syscallarg(linux_gid_t *) gidset; 1233 } */ *uap = v; 1234 struct proc *p = l->l_proc; 1235 caddr_t sg; 1236 int n, error, i; 1237 struct sys_getgroups_args bsa; 1238 gid_t *bset, *kbset; 1239 linux_gid_t *lset; 1240 struct pcred *pc = p->p_cred; 1241 1242 n = SCARG(uap, gidsetsize); 1243 if (n < 0) 1244 return EINVAL; 1245 error = 0; 1246 bset = kbset = NULL; 1247 lset = NULL; 1248 if (n > 0) { 1249 n = min(pc->pc_ucred->cr_ngroups, n); 1250 sg = stackgap_init(p, 0); 1251 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1252 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1253 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1254 if (bset == NULL || kbset == NULL || lset == NULL) 1255 return ENOMEM; 1256 SCARG(&bsa, gidsetsize) = n; 1257 SCARG(&bsa, gidset) = bset; 1258 error = sys_getgroups(l, &bsa, retval); 1259 if (error != 0) 1260 goto out; 1261 error = copyin(bset, kbset, n * sizeof (gid_t)); 1262 if (error != 0) 1263 goto out; 1264 for (i = 0; i < n; i++) 1265 lset[i] = (linux_gid_t)kbset[i]; 1266 error = copyout(lset, SCARG(uap, gidset), 1267 n * sizeof (linux_gid_t)); 1268 } else 1269 *retval = pc->pc_ucred->cr_ngroups; 1270 out: 1271 if (kbset != NULL) 1272 free(kbset, M_TEMP); 1273 if (lset != NULL) 1274 free(lset, M_TEMP); 1275 return error; 1276 } 1277 1278 int 1279 linux_sys_setgroups16(l, v, retval) 1280 struct lwp *l; 1281 void *v; 1282 register_t *retval; 1283 { 1284 struct linux_sys_setgroups16_args /* { 1285 syscallarg(int) gidsetsize; 1286 syscallarg(linux_gid_t *) gidset; 1287 } */ *uap = v; 1288 struct proc *p = l->l_proc; 1289 caddr_t sg; 1290 int n; 1291 int error, i; 1292 struct sys_setgroups_args bsa; 1293 gid_t *bset, *kbset; 1294 linux_gid_t *lset; 1295 1296 n = SCARG(uap, gidsetsize); 1297 if (n < 0 || n > NGROUPS) 1298 return EINVAL; 1299 sg = stackgap_init(p, 0); 1300 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1301 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1302 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1303 if (lset == NULL || bset == NULL) 1304 return ENOMEM; 1305 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t)); 1306 if (error != 0) 1307 goto out; 1308 for (i = 0; i < n; i++) 1309 kbset[i] = (gid_t)lset[i]; 1310 error = copyout(kbset, bset, n * sizeof (gid_t)); 1311 if (error != 0) 1312 goto out; 1313 SCARG(&bsa, gidsetsize) = n; 1314 SCARG(&bsa, gidset) = bset; 1315 error = sys_setgroups(l, &bsa, retval); 1316 1317 out: 1318 if (lset != NULL) 1319 free(lset, M_TEMP); 1320 if (kbset != NULL) 1321 free(kbset, M_TEMP); 1322 1323 return error; 1324 } 1325 1326 #endif /* __i386__ || __m68k__ || COMPAT_LINUX32 */ 1327 1328 #ifndef COMPAT_LINUX32 1329 /* 1330 * We have nonexistent fsuid equal to uid. 1331 * If modification is requested, refuse. 1332 */ 1333 int 1334 linux_sys_setfsuid(l, v, retval) 1335 struct lwp *l; 1336 void *v; 1337 register_t *retval; 1338 { 1339 struct linux_sys_setfsuid_args /* { 1340 syscallarg(uid_t) uid; 1341 } */ *uap = v; 1342 struct proc *p = l->l_proc; 1343 uid_t uid; 1344 1345 uid = SCARG(uap, uid); 1346 if (p->p_cred->p_ruid != uid) 1347 return sys_nosys(l, v, retval); 1348 else 1349 return (0); 1350 } 1351 1352 /* XXX XXX XXX */ 1353 # ifndef alpha 1354 int 1355 linux_sys_getfsuid(l, v, retval) 1356 struct lwp *l; 1357 void *v; 1358 register_t *retval; 1359 { 1360 return sys_getuid(l, v, retval); 1361 } 1362 # endif 1363 1364 int 1365 linux_sys_setresuid(l, v, retval) 1366 struct lwp *l; 1367 void *v; 1368 register_t *retval; 1369 { 1370 struct linux_sys_setresuid_args /* { 1371 syscallarg(uid_t) ruid; 1372 syscallarg(uid_t) euid; 1373 syscallarg(uid_t) suid; 1374 } */ *uap = v; 1375 1376 /* 1377 * Note: These checks are a little different than the NetBSD 1378 * setreuid(2) call performs. This precisely follows the 1379 * behavior of the Linux kernel. 1380 */ 1381 1382 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1383 SCARG(uap, suid), 1384 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1385 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1386 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1387 } 1388 1389 int 1390 linux_sys_getresuid(l, v, retval) 1391 struct lwp *l; 1392 void *v; 1393 register_t *retval; 1394 { 1395 struct linux_sys_getresuid_args /* { 1396 syscallarg(uid_t *) ruid; 1397 syscallarg(uid_t *) euid; 1398 syscallarg(uid_t *) suid; 1399 } */ *uap = v; 1400 struct proc *p = l->l_proc; 1401 struct pcred *pc = p->p_cred; 1402 int error; 1403 1404 /* 1405 * Linux copies these values out to userspace like so: 1406 * 1407 * 1. Copy out ruid. 1408 * 2. If that succeeds, copy out euid. 1409 * 3. If both of those succeed, copy out suid. 1410 */ 1411 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid), 1412 sizeof(uid_t))) != 0) 1413 return (error); 1414 1415 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid), 1416 sizeof(uid_t))) != 0) 1417 return (error); 1418 1419 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t))); 1420 } 1421 1422 int 1423 linux_sys_ptrace(l, v, retval) 1424 struct lwp *l; 1425 void *v; 1426 register_t *retval; 1427 { 1428 struct linux_sys_ptrace_args /* { 1429 i386, m68k, powerpc: T=int 1430 alpha, amd64: T=long 1431 syscallarg(T) request; 1432 syscallarg(T) pid; 1433 syscallarg(T) addr; 1434 syscallarg(T) data; 1435 } */ *uap = v; 1436 const int *ptr; 1437 int request; 1438 int error; 1439 1440 ptr = linux_ptrace_request_map; 1441 request = SCARG(uap, request); 1442 while (*ptr != -1) 1443 if (*ptr++ == request) { 1444 struct sys_ptrace_args pta; 1445 1446 SCARG(&pta, req) = *ptr; 1447 SCARG(&pta, pid) = SCARG(uap, pid); 1448 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr); 1449 SCARG(&pta, data) = SCARG(uap, data); 1450 1451 /* 1452 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1453 * to continue where the process left off previously. 1454 * The same thing is achieved by addr == (caddr_t) 1 1455 * on NetBSD, so rewrite 'addr' appropriately. 1456 */ 1457 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1458 SCARG(&pta, addr) = (caddr_t) 1; 1459 1460 error = sys_ptrace(l, &pta, retval); 1461 if (error) 1462 return error; 1463 switch (request) { 1464 case LINUX_PTRACE_PEEKTEXT: 1465 case LINUX_PTRACE_PEEKDATA: 1466 error = copyout (retval, 1467 (caddr_t)SCARG(uap, data), 1468 sizeof *retval); 1469 *retval = SCARG(uap, data); 1470 break; 1471 default: 1472 break; 1473 } 1474 return error; 1475 } 1476 else 1477 ptr++; 1478 1479 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1480 } 1481 1482 int 1483 linux_sys_reboot(struct lwp *l, void *v, register_t *retval) 1484 { 1485 struct linux_sys_reboot_args /* { 1486 syscallarg(int) magic1; 1487 syscallarg(int) magic2; 1488 syscallarg(int) cmd; 1489 syscallarg(void *) arg; 1490 } */ *uap = v; 1491 struct sys_reboot_args /* { 1492 syscallarg(int) opt; 1493 syscallarg(char *) bootstr; 1494 } */ sra; 1495 struct proc *p = l->l_proc; 1496 int error; 1497 1498 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1499 return(error); 1500 1501 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1502 return(EINVAL); 1503 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1504 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1505 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1506 return(EINVAL); 1507 1508 switch (SCARG(uap, cmd)) { 1509 case LINUX_REBOOT_CMD_RESTART: 1510 SCARG(&sra, opt) = RB_AUTOBOOT; 1511 break; 1512 case LINUX_REBOOT_CMD_HALT: 1513 SCARG(&sra, opt) = RB_HALT; 1514 break; 1515 case LINUX_REBOOT_CMD_POWER_OFF: 1516 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1517 break; 1518 case LINUX_REBOOT_CMD_RESTART2: 1519 /* Reboot with an argument. */ 1520 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1521 SCARG(&sra, bootstr) = SCARG(uap, arg); 1522 break; 1523 case LINUX_REBOOT_CMD_CAD_ON: 1524 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1525 case LINUX_REBOOT_CMD_CAD_OFF: 1526 return(0); 1527 default: 1528 return(EINVAL); 1529 } 1530 1531 return(sys_reboot(l, &sra, retval)); 1532 } 1533 1534 /* 1535 * Copy of compat_12_sys_swapon(). 1536 */ 1537 int 1538 linux_sys_swapon(l, v, retval) 1539 struct lwp *l; 1540 void *v; 1541 register_t *retval; 1542 { 1543 struct sys_swapctl_args ua; 1544 struct linux_sys_swapon_args /* { 1545 syscallarg(const char *) name; 1546 } */ *uap = v; 1547 1548 SCARG(&ua, cmd) = SWAP_ON; 1549 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name)); 1550 SCARG(&ua, misc) = 0; /* priority */ 1551 return (sys_swapctl(l, &ua, retval)); 1552 } 1553 1554 /* 1555 * Stop swapping to the file or block device specified by path. 1556 */ 1557 int 1558 linux_sys_swapoff(l, v, retval) 1559 struct lwp *l; 1560 void *v; 1561 register_t *retval; 1562 { 1563 struct sys_swapctl_args ua; 1564 struct linux_sys_swapoff_args /* { 1565 syscallarg(const char *) path; 1566 } */ *uap = v; 1567 1568 SCARG(&ua, cmd) = SWAP_OFF; 1569 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/ 1570 return (sys_swapctl(l, &ua, retval)); 1571 } 1572 1573 /* 1574 * Copy of compat_09_sys_setdomainname() 1575 */ 1576 /* ARGSUSED */ 1577 int 1578 linux_sys_setdomainname(l, v, retval) 1579 struct lwp *l; 1580 void *v; 1581 register_t *retval; 1582 { 1583 struct linux_sys_setdomainname_args /* { 1584 syscallarg(char *) domainname; 1585 syscallarg(int) len; 1586 } */ *uap = v; 1587 int name[2]; 1588 1589 name[0] = CTL_KERN; 1590 name[1] = KERN_DOMAINNAME; 1591 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1592 SCARG(uap, len), l)); 1593 } 1594 1595 /* 1596 * sysinfo() 1597 */ 1598 /* ARGSUSED */ 1599 int 1600 linux_sys_sysinfo(l, v, retval) 1601 struct lwp *l; 1602 void *v; 1603 register_t *retval; 1604 { 1605 struct linux_sys_sysinfo_args /* { 1606 syscallarg(struct linux_sysinfo *) arg; 1607 } */ *uap = v; 1608 struct linux_sysinfo si; 1609 struct loadavg *la; 1610 1611 si.uptime = time.tv_sec - boottime.tv_sec; 1612 la = &averunnable; 1613 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1614 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1615 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1616 si.totalram = ctob(physmem); 1617 si.freeram = uvmexp.free * uvmexp.pagesize; 1618 si.sharedram = 0; /* XXX */ 1619 si.bufferram = uvmexp.filepages * uvmexp.pagesize; 1620 si.totalswap = uvmexp.swpages * uvmexp.pagesize; 1621 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1622 si.procs = nprocs; 1623 1624 /* The following are only present in newer Linux kernels. */ 1625 si.totalbig = 0; 1626 si.freebig = 0; 1627 si.mem_unit = 1; 1628 1629 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1630 } 1631 1632 int 1633 linux_sys_getrlimit(l, v, retval) 1634 struct lwp *l; 1635 void *v; 1636 register_t *retval; 1637 { 1638 struct linux_sys_getrlimit_args /* { 1639 syscallarg(int) which; 1640 # ifdef LINUX_LARGEFILE64 1641 syscallarg(struct rlimit *) rlp; 1642 # else 1643 syscallarg(struct orlimit *) rlp; 1644 # endif 1645 } */ *uap = v; 1646 struct proc *p = l->l_proc; 1647 caddr_t sg = stackgap_init(p, 0); 1648 struct sys_getrlimit_args ap; 1649 struct rlimit rl; 1650 # ifdef LINUX_LARGEFILE64 1651 struct rlimit orl; 1652 # else 1653 struct orlimit orl; 1654 # endif 1655 int error; 1656 1657 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1658 if ((error = SCARG(&ap, which)) < 0) 1659 return -error; 1660 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1661 if ((error = sys_getrlimit(l, &ap, retval)) != 0) 1662 return error; 1663 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0) 1664 return error; 1665 bsd_to_linux_rlimit(&orl, &rl); 1666 1667 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1668 } 1669 1670 int 1671 linux_sys_setrlimit(l, v, retval) 1672 struct lwp *l; 1673 void *v; 1674 register_t *retval; 1675 { 1676 struct linux_sys_setrlimit_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 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1697 if ((error = SCARG(&ap, which)) < 0) 1698 return -error; 1699 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1700 return error; 1701 linux_to_bsd_rlimit(&rl, &orl); 1702 if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0) 1703 return error; 1704 return sys_setrlimit(l, &ap, retval); 1705 } 1706 1707 # if !defined(__mips__) && !defined(__amd64__) 1708 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1709 int 1710 linux_sys_ugetrlimit(l, v, retval) 1711 struct lwp *l; 1712 void *v; 1713 register_t *retval; 1714 { 1715 return linux_sys_getrlimit(l, v, retval); 1716 } 1717 # endif 1718 1719 /* 1720 * This gets called for unsupported syscalls. The difference to sys_nosys() 1721 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1722 * This is the way Linux does it and glibc depends on this behaviour. 1723 */ 1724 int 1725 linux_sys_nosys(l, v, retval) 1726 struct lwp *l; 1727 void *v; 1728 register_t *retval; 1729 { 1730 return (ENOSYS); 1731 } 1732 1733 #endif /* !COMPAT_LINUX32 */ 1734