linux_misc.c revision 1.140.2.1
1 /* $NetBSD: linux_misc.c,v 1.140.2.1 2005/11/02 11:57:56 yamt 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 1.140.2.1 2005/11/02 11:57:56 yamt Exp $"); 68 69 #include <sys/param.h> 70 #include <sys/systm.h> 71 #include <sys/namei.h> 72 #include <sys/proc.h> 73 #include <sys/dirent.h> 74 #include <sys/file.h> 75 #include <sys/stat.h> 76 #include <sys/filedesc.h> 77 #include <sys/ioctl.h> 78 #include <sys/kernel.h> 79 #include <sys/malloc.h> 80 #include <sys/mbuf.h> 81 #include <sys/mman.h> 82 #include <sys/mount.h> 83 #include <sys/reboot.h> 84 #include <sys/resource.h> 85 #include <sys/resourcevar.h> 86 #include <sys/signal.h> 87 #include <sys/signalvar.h> 88 #include <sys/socket.h> 89 #include <sys/time.h> 90 #include <sys/times.h> 91 #include <sys/vnode.h> 92 #include <sys/uio.h> 93 #include <sys/wait.h> 94 #include <sys/utsname.h> 95 #include <sys/unistd.h> 96 #include <sys/swap.h> /* for SWAP_ON */ 97 #include <sys/sysctl.h> /* for KERN_DOMAINNAME */ 98 99 #include <sys/ptrace.h> 100 #include <machine/ptrace.h> 101 102 #include <sys/sa.h> 103 #include <sys/syscallargs.h> 104 105 #include <compat/linux/common/linux_types.h> 106 #include <compat/linux/common/linux_signal.h> 107 108 #include <compat/linux/linux_syscallargs.h> 109 110 #include <compat/linux/common/linux_fcntl.h> 111 #include <compat/linux/common/linux_mmap.h> 112 #include <compat/linux/common/linux_dirent.h> 113 #include <compat/linux/common/linux_util.h> 114 #include <compat/linux/common/linux_misc.h> 115 #include <compat/linux/common/linux_ptrace.h> 116 #include <compat/linux/common/linux_reboot.h> 117 #include <compat/linux/common/linux_emuldata.h> 118 119 const int linux_ptrace_request_map[] = { 120 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 121 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 122 LINUX_PTRACE_PEEKDATA, PT_READ_D, 123 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 124 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 125 LINUX_PTRACE_CONT, PT_CONTINUE, 126 LINUX_PTRACE_KILL, PT_KILL, 127 LINUX_PTRACE_ATTACH, PT_ATTACH, 128 LINUX_PTRACE_DETACH, PT_DETACH, 129 #ifdef PT_STEP 130 LINUX_PTRACE_SINGLESTEP, PT_STEP, 131 #endif 132 -1 133 }; 134 135 const struct linux_mnttypes linux_fstypes[] = { 136 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 137 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 138 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 139 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 140 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 141 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 142 { MOUNT_PORTAL, LINUX_DEFAULT_SUPER_MAGIC }, 143 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 144 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 145 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 146 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 148 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 149 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 150 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 151 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 152 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 153 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 154 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 155 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 156 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 157 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC }, 158 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC }, 159 { MOUNT_TMPFS, LINUX_DEFAULT_SUPER_MAGIC } 160 }; 161 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 162 163 #ifdef DEBUG_LINUX 164 #define DPRINTF(a) uprintf a 165 #else 166 #define DPRINTF(a) 167 #endif 168 169 /* Local linux_misc.c functions: */ 170 #ifndef __amd64__ 171 static void bsd_to_linux_statfs __P((const struct statvfs *, 172 struct linux_statfs *)); 173 #endif 174 static int linux_to_bsd_limit __P((int)); 175 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *, 176 const struct linux_sys_mmap_args *)); 177 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *, 178 register_t *, off_t)); 179 180 181 /* 182 * The information on a terminated (or stopped) process needs 183 * to be converted in order for Linux binaries to get a valid signal 184 * number out of it. 185 */ 186 void 187 bsd_to_linux_wstat(st) 188 int *st; 189 { 190 191 int sig; 192 193 if (WIFSIGNALED(*st)) { 194 sig = WTERMSIG(*st); 195 if (sig >= 0 && sig < NSIG) 196 *st= (*st& ~0177) | native_to_linux_signo[sig]; 197 } else if (WIFSTOPPED(*st)) { 198 sig = WSTOPSIG(*st); 199 if (sig >= 0 && sig < NSIG) 200 *st = (*st & ~0xff00) | 201 (native_to_linux_signo[sig] << 8); 202 } 203 } 204 205 /* 206 * wait4(2). Passed on to the NetBSD call, surrounded by code to 207 * reserve some space for a NetBSD-style wait status, and converting 208 * it to what Linux wants. 209 */ 210 int 211 linux_sys_wait4(l, v, retval) 212 struct lwp *l; 213 void *v; 214 register_t *retval; 215 { 216 struct linux_sys_wait4_args /* { 217 syscallarg(int) pid; 218 syscallarg(int *) status; 219 syscallarg(int) options; 220 syscallarg(struct rusage *) rusage; 221 } */ *uap = v; 222 struct proc *p = l->l_proc; 223 struct sys_wait4_args w4a; 224 int error, *status, tstat, options, linux_options; 225 caddr_t sg; 226 227 if (SCARG(uap, status) != NULL) { 228 sg = stackgap_init(p, 0); 229 status = (int *) stackgap_alloc(p, &sg, sizeof *status); 230 } else 231 status = NULL; 232 233 linux_options = SCARG(uap, options); 234 options = 0; 235 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS)) 236 return (EINVAL); 237 238 if (linux_options & LINUX_WAIT4_WNOHANG) 239 options |= WNOHANG; 240 if (linux_options & LINUX_WAIT4_WUNTRACED) 241 options |= WUNTRACED; 242 if (linux_options & LINUX_WAIT4_WALL) 243 options |= WALLSIG; 244 if (linux_options & LINUX_WAIT4_WCLONE) 245 options |= WALTSIG; 246 #ifdef DIAGNOSTIC 247 if (linux_options & LINUX_WAIT4_WNOTHREAD) 248 printf("WARNING: %s: linux process %d.%d called " 249 "waitpid with __WNOTHREAD set!", 250 __FILE__, p->p_pid, l->l_lid); 251 252 #endif 253 254 SCARG(&w4a, pid) = SCARG(uap, pid); 255 SCARG(&w4a, status) = status; 256 SCARG(&w4a, options) = options; 257 SCARG(&w4a, rusage) = SCARG(uap, rusage); 258 259 if ((error = sys_wait4(l, &w4a, retval))) 260 return error; 261 262 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD); 263 264 if (status != NULL) { 265 if ((error = copyin(status, &tstat, sizeof tstat))) 266 return error; 267 268 bsd_to_linux_wstat(&tstat); 269 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 270 } 271 272 return 0; 273 } 274 275 /* 276 * Linux brk(2). The check if the new address is >= the old one is 277 * done in the kernel in Linux. NetBSD does it in the library. 278 */ 279 int 280 linux_sys_brk(l, v, retval) 281 struct lwp *l; 282 void *v; 283 register_t *retval; 284 { 285 struct linux_sys_brk_args /* { 286 syscallarg(char *) nsize; 287 } */ *uap = v; 288 struct proc *p = l->l_proc; 289 char *nbrk = SCARG(uap, nsize); 290 struct sys_obreak_args oba; 291 struct vmspace *vm = p->p_vmspace; 292 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata; 293 294 SCARG(&oba, nsize) = nbrk; 295 296 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0) 297 ed->s->p_break = (char*)nbrk; 298 else 299 nbrk = ed->s->p_break; 300 301 retval[0] = (register_t)nbrk; 302 303 return 0; 304 } 305 306 #ifndef __amd64__ 307 /* 308 * Convert NetBSD statvfs structure to Linux statfs structure. 309 * Linux doesn't have f_flag, and we can't set f_frsize due 310 * to glibc statvfs() bug (see below). 311 */ 312 static void 313 bsd_to_linux_statfs(bsp, lsp) 314 const struct statvfs *bsp; 315 struct linux_statfs *lsp; 316 { 317 int i; 318 319 for (i = 0; i < linux_fstypes_cnt; i++) { 320 if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) { 321 lsp->l_ftype = linux_fstypes[i].linux; 322 break; 323 } 324 } 325 326 if (i == linux_fstypes_cnt) { 327 DPRINTF(("unhandled fstype in linux emulation: %s\n", 328 bsp->f_fstypename)); 329 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC; 330 } 331 332 /* 333 * The sizes are expressed in number of blocks. The block 334 * size used for the size is f_frsize for POSIX-compliant 335 * statvfs. Linux statfs uses f_bsize as the block size 336 * (f_frsize used to not be available in Linux struct statfs). 337 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block 338 * counts for different f_frsize if f_frsize is provided by the kernel. 339 * POSIX conforming apps thus get wrong size if f_frsize 340 * is different to f_bsize. Thus, we just pretend we don't 341 * support f_frsize. 342 */ 343 344 lsp->l_fbsize = bsp->f_frsize; 345 lsp->l_ffrsize = 0; /* compat */ 346 lsp->l_fblocks = bsp->f_blocks; 347 lsp->l_fbfree = bsp->f_bfree; 348 lsp->l_fbavail = bsp->f_bavail; 349 lsp->l_ffiles = bsp->f_files; 350 lsp->l_fffree = bsp->f_ffree; 351 /* Linux sets the fsid to 0..., we don't */ 352 lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0]; 353 lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1]; 354 lsp->l_fnamelen = bsp->f_namemax; 355 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare)); 356 } 357 358 /* 359 * Implement the fs stat functions. Straightforward. 360 */ 361 int 362 linux_sys_statfs(l, v, retval) 363 struct lwp *l; 364 void *v; 365 register_t *retval; 366 { 367 struct linux_sys_statfs_args /* { 368 syscallarg(const char *) path; 369 syscallarg(struct linux_statfs *) sp; 370 } */ *uap = v; 371 struct proc *p = l->l_proc; 372 struct statvfs btmp, *bsp; 373 struct linux_statfs ltmp; 374 struct sys_statvfs1_args bsa; 375 caddr_t sg; 376 int error; 377 378 sg = stackgap_init(p, 0); 379 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 380 381 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 382 383 SCARG(&bsa, path) = SCARG(uap, path); 384 SCARG(&bsa, buf) = bsp; 385 SCARG(&bsa, flags) = ST_WAIT; 386 387 if ((error = sys_statvfs1(l, &bsa, retval))) 388 return error; 389 390 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 391 return error; 392 393 bsd_to_linux_statfs(&btmp, <mp); 394 395 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 396 } 397 398 int 399 linux_sys_fstatfs(l, v, retval) 400 struct lwp *l; 401 void *v; 402 register_t *retval; 403 { 404 struct linux_sys_fstatfs_args /* { 405 syscallarg(int) fd; 406 syscallarg(struct linux_statfs *) sp; 407 } */ *uap = v; 408 struct proc *p = l->l_proc; 409 struct statvfs btmp, *bsp; 410 struct linux_statfs ltmp; 411 struct sys_fstatvfs1_args bsa; 412 caddr_t sg; 413 int error; 414 415 sg = stackgap_init(p, 0); 416 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 417 418 SCARG(&bsa, fd) = SCARG(uap, fd); 419 SCARG(&bsa, buf) = bsp; 420 SCARG(&bsa, flags) = ST_WAIT; 421 422 if ((error = sys_fstatvfs1(l, &bsa, retval))) 423 return error; 424 425 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 426 return error; 427 428 bsd_to_linux_statfs(&btmp, <mp); 429 430 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 431 } 432 #endif /* __amd64__ */ 433 434 /* 435 * uname(). Just copy the info from the various strings stored in the 436 * kernel, and put it in the Linux utsname structure. That structure 437 * is almost the same as the NetBSD one, only it has fields 65 characters 438 * long, and an extra domainname field. 439 */ 440 int 441 linux_sys_uname(l, v, retval) 442 struct lwp *l; 443 void *v; 444 register_t *retval; 445 { 446 struct linux_sys_uname_args /* { 447 syscallarg(struct linux_utsname *) up; 448 } */ *uap = v; 449 struct linux_utsname luts; 450 451 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 452 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 453 strncpy(luts.l_release, linux_release, sizeof(luts.l_release)); 454 strncpy(luts.l_version, linux_version, sizeof(luts.l_version)); 455 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 456 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 457 458 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 459 } 460 461 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 462 /* Used indirectly on: arm, i386, m68k */ 463 464 /* 465 * New type Linux mmap call. 466 * Only called directly on machines with >= 6 free regs. 467 */ 468 int 469 linux_sys_mmap(l, v, retval) 470 struct lwp *l; 471 void *v; 472 register_t *retval; 473 { 474 struct linux_sys_mmap_args /* { 475 syscallarg(unsigned long) addr; 476 syscallarg(size_t) len; 477 syscallarg(int) prot; 478 syscallarg(int) flags; 479 syscallarg(int) fd; 480 syscallarg(linux_off_t) offset; 481 } */ *uap = v; 482 483 if (SCARG(uap, offset) & PAGE_MASK) 484 return EINVAL; 485 486 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 487 } 488 489 /* 490 * Guts of most architectures' mmap64() implementations. This shares 491 * its list of arguments with linux_sys_mmap(). 492 * 493 * The difference in linux_sys_mmap2() is that "offset" is actually 494 * (offset / pagesize), not an absolute byte count. This translation 495 * to pagesize offsets is done inside glibc between the mmap64() call 496 * point, and the actual syscall. 497 */ 498 int 499 linux_sys_mmap2(l, v, retval) 500 struct lwp *l; 501 void *v; 502 register_t *retval; 503 { 504 struct linux_sys_mmap2_args /* { 505 syscallarg(unsigned long) addr; 506 syscallarg(size_t) len; 507 syscallarg(int) prot; 508 syscallarg(int) flags; 509 syscallarg(int) fd; 510 syscallarg(linux_off_t) offset; 511 } */ *uap = v; 512 513 return linux_mmap(l, uap, retval, 514 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 515 } 516 517 /* 518 * Massage arguments and call system mmap(2). 519 */ 520 static int 521 linux_mmap(l, uap, retval, offset) 522 struct lwp *l; 523 struct linux_sys_mmap_args *uap; 524 register_t *retval; 525 off_t offset; 526 { 527 struct sys_mmap_args cma; 528 int error; 529 size_t mmoff=0; 530 531 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 532 /* 533 * Request for stack-like memory segment. On linux, this 534 * works by mmap()ping (small) segment, which is automatically 535 * extended when page fault happens below the currently 536 * allocated area. We emulate this by allocating (typically 537 * bigger) segment sized at current stack size limit, and 538 * offsetting the requested and returned address accordingly. 539 * Since physical pages are only allocated on-demand, this 540 * is effectively identical. 541 */ 542 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 543 544 if (SCARG(uap, len) < ssl) { 545 /* Compute the address offset */ 546 mmoff = round_page(ssl) - SCARG(uap, len); 547 548 if (SCARG(uap, addr)) 549 SCARG(uap, addr) -= mmoff; 550 551 SCARG(uap, len) = (size_t) ssl; 552 } 553 } 554 555 linux_to_bsd_mmap_args(&cma, uap); 556 SCARG(&cma, pos) = offset; 557 558 error = sys_mmap(l, &cma, retval); 559 if (error) 560 return (error); 561 562 /* Shift the returned address for stack-like segment if necessary */ 563 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff) 564 retval[0] += mmoff; 565 566 return (0); 567 } 568 569 static void 570 linux_to_bsd_mmap_args(cma, uap) 571 struct sys_mmap_args *cma; 572 const struct linux_sys_mmap_args *uap; 573 { 574 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 575 576 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 577 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 578 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 579 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 580 /* XXX XAX ERH: Any other flags here? There are more defined... */ 581 582 SCARG(cma, addr) = (void *)SCARG(uap, addr); 583 SCARG(cma, len) = SCARG(uap, len); 584 SCARG(cma, prot) = SCARG(uap, prot); 585 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 586 SCARG(cma, prot) |= VM_PROT_READ; 587 SCARG(cma, flags) = flags; 588 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 589 SCARG(cma, pad) = 0; 590 } 591 592 int 593 linux_sys_mremap(l, v, retval) 594 struct lwp *l; 595 void *v; 596 register_t *retval; 597 { 598 struct linux_sys_mremap_args /* { 599 syscallarg(void *) old_address; 600 syscallarg(size_t) old_size; 601 syscallarg(size_t) new_size; 602 syscallarg(u_long) flags; 603 } */ *uap = v; 604 struct sys_munmap_args mua; 605 size_t old_size, new_size; 606 int error; 607 608 old_size = round_page(SCARG(uap, old_size)); 609 new_size = round_page(SCARG(uap, new_size)); 610 611 /* 612 * Growing mapped region. 613 */ 614 if (new_size > old_size) { 615 /* 616 * XXX Implement me. What we probably want to do is 617 * XXX dig out the guts of the old mapping, mmap that 618 * XXX object again with the new size, then munmap 619 * XXX the old mapping. 620 */ 621 *retval = 0; 622 return (ENOMEM); 623 } 624 625 /* 626 * Shrinking mapped region. 627 */ 628 if (new_size < old_size) { 629 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) + 630 new_size; 631 SCARG(&mua, len) = old_size - new_size; 632 error = sys_munmap(l, &mua, retval); 633 *retval = error ? 0 : (register_t)SCARG(uap, old_address); 634 return (error); 635 } 636 637 /* 638 * No change. 639 */ 640 *retval = (register_t)SCARG(uap, old_address); 641 return (0); 642 } 643 644 int 645 linux_sys_msync(l, v, retval) 646 struct lwp *l; 647 void *v; 648 register_t *retval; 649 { 650 struct linux_sys_msync_args /* { 651 syscallarg(caddr_t) addr; 652 syscallarg(int) len; 653 syscallarg(int) fl; 654 } */ *uap = v; 655 656 struct sys___msync13_args bma; 657 658 /* flags are ignored */ 659 SCARG(&bma, addr) = SCARG(uap, addr); 660 SCARG(&bma, len) = SCARG(uap, len); 661 SCARG(&bma, flags) = SCARG(uap, fl); 662 663 return sys___msync13(l, &bma, retval); 664 } 665 666 int 667 linux_sys_mprotect(l, v, retval) 668 struct lwp *l; 669 void *v; 670 register_t *retval; 671 { 672 struct linux_sys_mprotect_args /* { 673 syscallarg(const void *) start; 674 syscallarg(unsigned long) len; 675 syscallarg(int) prot; 676 } */ *uap = v; 677 struct vm_map_entry *entry; 678 struct vm_map *map; 679 struct proc *p; 680 vaddr_t end, start, len, stacklim; 681 int prot, grows; 682 683 start = (vaddr_t)SCARG(uap, start); 684 len = round_page(SCARG(uap, len)); 685 prot = SCARG(uap, prot); 686 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 687 prot &= ~grows; 688 end = start + len; 689 690 if (start & PAGE_MASK) 691 return EINVAL; 692 if (end < start) 693 return EINVAL; 694 if (end == start) 695 return 0; 696 697 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 698 return EINVAL; 699 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP)) 700 return EINVAL; 701 702 p = l->l_proc; 703 map = &p->p_vmspace->vm_map; 704 vm_map_lock(map); 705 #ifdef notdef 706 VM_MAP_RANGE_CHECK(map, start, end); 707 #endif 708 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 709 vm_map_unlock(map); 710 return ENOMEM; 711 } 712 713 /* 714 * Approximate the behaviour of PROT_GROWS{DOWN,UP}. 715 */ 716 717 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur; 718 if (grows & LINUX_PROT_GROWSDOWN) { 719 if (USRSTACK - stacklim <= start && start < USRSTACK) { 720 start = USRSTACK - stacklim; 721 } else { 722 start = entry->start; 723 } 724 } else if (grows & LINUX_PROT_GROWSUP) { 725 if (USRSTACK <= end && end < USRSTACK + stacklim) { 726 end = USRSTACK + stacklim; 727 } else { 728 end = entry->end; 729 } 730 } 731 vm_map_unlock(map); 732 return uvm_map_protect(map, start, end, prot, FALSE); 733 } 734 735 /* 736 * This code is partly stolen from src/lib/libc/compat-43/times.c 737 */ 738 739 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 740 741 int 742 linux_sys_times(l, v, retval) 743 struct lwp *l; 744 void *v; 745 register_t *retval; 746 { 747 struct linux_sys_times_args /* { 748 syscallarg(struct times *) tms; 749 } */ *uap = v; 750 struct proc *p = l->l_proc; 751 struct timeval t; 752 int error, s; 753 754 if (SCARG(uap, tms)) { 755 struct linux_tms ltms; 756 struct rusage ru; 757 758 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 759 ltms.ltms_utime = CONVTCK(ru.ru_utime); 760 ltms.ltms_stime = CONVTCK(ru.ru_stime); 761 762 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 763 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 764 765 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 766 return error; 767 } 768 769 s = splclock(); 770 timersub(&time, &boottime, &t); 771 splx(s); 772 773 retval[0] = ((linux_clock_t)(CONVTCK(t))); 774 return 0; 775 } 776 777 #undef CONVTCK 778 779 /* 780 * Linux 'readdir' call. This code is mostly taken from the 781 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 782 * an attempt has been made to keep it a little cleaner (failing 783 * miserably, because of the cruft needed if count 1 is passed). 784 * 785 * The d_off field should contain the offset of the next valid entry, 786 * but in Linux it has the offset of the entry itself. We emulate 787 * that bug here. 788 * 789 * Read in BSD-style entries, convert them, and copy them out. 790 * 791 * Note that this doesn't handle union-mounted filesystems. 792 */ 793 int 794 linux_sys_getdents(l, v, retval) 795 struct lwp *l; 796 void *v; 797 register_t *retval; 798 { 799 struct linux_sys_getdents_args /* { 800 syscallarg(int) fd; 801 syscallarg(struct linux_dirent *) dent; 802 syscallarg(unsigned int) count; 803 } */ *uap = v; 804 struct proc *p = l->l_proc; 805 struct dirent *bdp; 806 struct vnode *vp; 807 caddr_t inp, tbuf; /* BSD-format */ 808 int len, reclen; /* BSD-format */ 809 caddr_t outp; /* Linux-format */ 810 int resid, linux_reclen = 0; /* Linux-format */ 811 struct file *fp; 812 struct uio auio; 813 struct iovec aiov; 814 struct linux_dirent idb; 815 off_t off; /* true file offset */ 816 int buflen, error, eofflag, nbytes, oldcall; 817 struct vattr va; 818 off_t *cookiebuf = NULL, *cookie; 819 int ncookies; 820 821 /* getvnode() will use the descriptor for us */ 822 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 823 return (error); 824 825 if ((fp->f_flag & FREAD) == 0) { 826 error = EBADF; 827 goto out1; 828 } 829 830 vp = (struct vnode *)fp->f_data; 831 if (vp->v_type != VDIR) { 832 error = EINVAL; 833 goto out1; 834 } 835 836 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 837 goto out1; 838 839 nbytes = SCARG(uap, count); 840 if (nbytes == 1) { /* emulating old, broken behaviour */ 841 nbytes = sizeof (idb); 842 buflen = max(va.va_blocksize, nbytes); 843 oldcall = 1; 844 } else { 845 buflen = min(MAXBSIZE, nbytes); 846 if (buflen < va.va_blocksize) 847 buflen = va.va_blocksize; 848 oldcall = 0; 849 } 850 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 851 852 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 853 off = fp->f_offset; 854 again: 855 aiov.iov_base = tbuf; 856 aiov.iov_len = buflen; 857 auio.uio_iov = &aiov; 858 auio.uio_iovcnt = 1; 859 auio.uio_rw = UIO_READ; 860 auio.uio_segflg = UIO_SYSSPACE; 861 auio.uio_procp = NULL; 862 auio.uio_resid = buflen; 863 auio.uio_offset = off; 864 /* 865 * First we read into the malloc'ed buffer, then 866 * we massage it into user space, one record at a time. 867 */ 868 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 869 &ncookies); 870 if (error) 871 goto out; 872 873 inp = tbuf; 874 outp = (caddr_t)SCARG(uap, dent); 875 resid = nbytes; 876 if ((len = buflen - auio.uio_resid) == 0) 877 goto eof; 878 879 for (cookie = cookiebuf; len > 0; len -= reclen) { 880 bdp = (struct dirent *)inp; 881 reclen = bdp->d_reclen; 882 if (reclen & 3) 883 panic("linux_readdir"); 884 if (bdp->d_fileno == 0) { 885 inp += reclen; /* it is a hole; squish it out */ 886 if (cookie) 887 off = *cookie++; 888 else 889 off += reclen; 890 continue; 891 } 892 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 893 if (reclen > len || resid < linux_reclen) { 894 /* entry too big for buffer, so just stop */ 895 outp++; 896 break; 897 } 898 /* 899 * Massage in place to make a Linux-shaped dirent (otherwise 900 * we have to worry about touching user memory outside of 901 * the copyout() call). 902 */ 903 idb.d_ino = bdp->d_fileno; 904 /* 905 * The old readdir() call misuses the offset and reclen fields. 906 */ 907 if (oldcall) { 908 idb.d_off = (linux_off_t)linux_reclen; 909 idb.d_reclen = (u_short)bdp->d_namlen; 910 } else { 911 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 912 compat_offseterr(vp, "linux_getdents"); 913 error = EINVAL; 914 goto out; 915 } 916 idb.d_off = (linux_off_t)off; 917 idb.d_reclen = (u_short)linux_reclen; 918 } 919 strcpy(idb.d_name, bdp->d_name); 920 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 921 goto out; 922 /* advance past this real entry */ 923 inp += reclen; 924 if (cookie) 925 off = *cookie++; /* each entry points to itself */ 926 else 927 off += reclen; 928 /* advance output past Linux-shaped entry */ 929 outp += linux_reclen; 930 resid -= linux_reclen; 931 if (oldcall) 932 break; 933 } 934 935 /* if we squished out the whole block, try again */ 936 if (outp == (caddr_t)SCARG(uap, dent)) 937 goto again; 938 fp->f_offset = off; /* update the vnode offset */ 939 940 if (oldcall) 941 nbytes = resid + linux_reclen; 942 943 eof: 944 *retval = nbytes - resid; 945 out: 946 VOP_UNLOCK(vp, 0); 947 if (cookiebuf) 948 free(cookiebuf, M_TEMP); 949 free(tbuf, M_TEMP); 950 out1: 951 FILE_UNUSE(fp, p); 952 return error; 953 } 954 955 /* 956 * Even when just using registers to pass arguments to syscalls you can 957 * have 5 of them on the i386. So this newer version of select() does 958 * this. 959 */ 960 int 961 linux_sys_select(l, v, retval) 962 struct lwp *l; 963 void *v; 964 register_t *retval; 965 { 966 struct linux_sys_select_args /* { 967 syscallarg(int) nfds; 968 syscallarg(fd_set *) readfds; 969 syscallarg(fd_set *) writefds; 970 syscallarg(fd_set *) exceptfds; 971 syscallarg(struct timeval *) timeout; 972 } */ *uap = v; 973 974 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 975 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 976 } 977 978 /* 979 * Common code for the old and new versions of select(). A couple of 980 * things are important: 981 * 1) return the amount of time left in the 'timeout' parameter 982 * 2) select never returns ERESTART on Linux, always return EINTR 983 */ 984 int 985 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout) 986 struct lwp *l; 987 register_t *retval; 988 int nfds; 989 fd_set *readfds, *writefds, *exceptfds; 990 struct timeval *timeout; 991 { 992 struct sys_select_args bsa; 993 struct proc *p = l->l_proc; 994 struct timeval tv0, tv1, utv, *tvp; 995 caddr_t sg; 996 int error; 997 998 SCARG(&bsa, nd) = nfds; 999 SCARG(&bsa, in) = readfds; 1000 SCARG(&bsa, ou) = writefds; 1001 SCARG(&bsa, ex) = exceptfds; 1002 SCARG(&bsa, tv) = timeout; 1003 1004 /* 1005 * Store current time for computation of the amount of 1006 * time left. 1007 */ 1008 if (timeout) { 1009 if ((error = copyin(timeout, &utv, sizeof(utv)))) 1010 return error; 1011 if (itimerfix(&utv)) { 1012 /* 1013 * The timeval was invalid. Convert it to something 1014 * valid that will act as it does under Linux. 1015 */ 1016 sg = stackgap_init(p, 0); 1017 tvp = stackgap_alloc(p, &sg, sizeof(utv)); 1018 utv.tv_sec += utv.tv_usec / 1000000; 1019 utv.tv_usec %= 1000000; 1020 if (utv.tv_usec < 0) { 1021 utv.tv_sec -= 1; 1022 utv.tv_usec += 1000000; 1023 } 1024 if (utv.tv_sec < 0) 1025 timerclear(&utv); 1026 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1027 return error; 1028 SCARG(&bsa, tv) = tvp; 1029 } 1030 microtime(&tv0); 1031 } 1032 1033 error = sys_select(l, &bsa, retval); 1034 if (error) { 1035 /* 1036 * See fs/select.c in the Linux kernel. Without this, 1037 * Maelstrom doesn't work. 1038 */ 1039 if (error == ERESTART) 1040 error = EINTR; 1041 return error; 1042 } 1043 1044 if (timeout) { 1045 if (*retval) { 1046 /* 1047 * Compute how much time was left of the timeout, 1048 * by subtracting the current time and the time 1049 * before we started the call, and subtracting 1050 * that result from the user-supplied value. 1051 */ 1052 microtime(&tv1); 1053 timersub(&tv1, &tv0, &tv1); 1054 timersub(&utv, &tv1, &utv); 1055 if (utv.tv_sec < 0) 1056 timerclear(&utv); 1057 } else 1058 timerclear(&utv); 1059 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1060 return error; 1061 } 1062 1063 return 0; 1064 } 1065 1066 /* 1067 * Get the process group of a certain process. Look it up 1068 * and return the value. 1069 */ 1070 int 1071 linux_sys_getpgid(l, v, retval) 1072 struct lwp *l; 1073 void *v; 1074 register_t *retval; 1075 { 1076 struct linux_sys_getpgid_args /* { 1077 syscallarg(int) pid; 1078 } */ *uap = v; 1079 struct proc *p = l->l_proc; 1080 struct proc *targp; 1081 1082 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1083 if ((targp = pfind(SCARG(uap, pid))) == 0) 1084 return ESRCH; 1085 } 1086 else 1087 targp = p; 1088 1089 retval[0] = targp->p_pgid; 1090 return 0; 1091 } 1092 1093 /* 1094 * Set the 'personality' (emulation mode) for the current process. Only 1095 * accept the Linux personality here (0). This call is needed because 1096 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1097 * ELF binaries run in Linux mode, not SVR4 mode. 1098 */ 1099 int 1100 linux_sys_personality(l, v, retval) 1101 struct lwp *l; 1102 void *v; 1103 register_t *retval; 1104 { 1105 struct linux_sys_personality_args /* { 1106 syscallarg(int) per; 1107 } */ *uap = v; 1108 1109 if (SCARG(uap, per) != 0) 1110 return EINVAL; 1111 retval[0] = 0; 1112 return 0; 1113 } 1114 1115 #if defined(__i386__) || defined(__m68k__) 1116 /* 1117 * The calls are here because of type conversions. 1118 */ 1119 int 1120 linux_sys_setreuid16(l, v, retval) 1121 struct lwp *l; 1122 void *v; 1123 register_t *retval; 1124 { 1125 struct linux_sys_setreuid16_args /* { 1126 syscallarg(int) ruid; 1127 syscallarg(int) euid; 1128 } */ *uap = v; 1129 struct sys_setreuid_args bsa; 1130 1131 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1132 (uid_t)-1 : SCARG(uap, ruid); 1133 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1134 (uid_t)-1 : SCARG(uap, euid); 1135 1136 return sys_setreuid(l, &bsa, retval); 1137 } 1138 1139 int 1140 linux_sys_setregid16(l, v, retval) 1141 struct lwp *l; 1142 void *v; 1143 register_t *retval; 1144 { 1145 struct linux_sys_setregid16_args /* { 1146 syscallarg(int) rgid; 1147 syscallarg(int) egid; 1148 } */ *uap = v; 1149 struct sys_setregid_args bsa; 1150 1151 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1152 (uid_t)-1 : SCARG(uap, rgid); 1153 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1154 (uid_t)-1 : SCARG(uap, egid); 1155 1156 return sys_setregid(l, &bsa, retval); 1157 } 1158 1159 int 1160 linux_sys_setresuid16(l, v, retval) 1161 struct lwp *l; 1162 void *v; 1163 register_t *retval; 1164 { 1165 struct linux_sys_setresuid16_args /* { 1166 syscallarg(uid_t) ruid; 1167 syscallarg(uid_t) euid; 1168 syscallarg(uid_t) suid; 1169 } */ *uap = v; 1170 struct linux_sys_setresuid16_args lsa; 1171 1172 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1173 (uid_t)-1 : SCARG(uap, ruid); 1174 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1175 (uid_t)-1 : SCARG(uap, euid); 1176 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ? 1177 (uid_t)-1 : SCARG(uap, suid); 1178 1179 return linux_sys_setresuid(l, &lsa, retval); 1180 } 1181 1182 int 1183 linux_sys_setresgid16(l, v, retval) 1184 struct lwp *l; 1185 void *v; 1186 register_t *retval; 1187 { 1188 struct linux_sys_setresgid16_args /* { 1189 syscallarg(gid_t) rgid; 1190 syscallarg(gid_t) egid; 1191 syscallarg(gid_t) sgid; 1192 } */ *uap = v; 1193 struct linux_sys_setresgid16_args lsa; 1194 1195 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1196 (gid_t)-1 : SCARG(uap, rgid); 1197 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1198 (gid_t)-1 : SCARG(uap, egid); 1199 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ? 1200 (gid_t)-1 : SCARG(uap, sgid); 1201 1202 return linux_sys_setresgid(l, &lsa, retval); 1203 } 1204 1205 int 1206 linux_sys_getgroups16(l, v, retval) 1207 struct lwp *l; 1208 void *v; 1209 register_t *retval; 1210 { 1211 struct linux_sys_getgroups16_args /* { 1212 syscallarg(int) gidsetsize; 1213 syscallarg(linux_gid_t *) gidset; 1214 } */ *uap = v; 1215 struct proc *p = l->l_proc; 1216 caddr_t sg; 1217 int n, error, i; 1218 struct sys_getgroups_args bsa; 1219 gid_t *bset, *kbset; 1220 linux_gid_t *lset; 1221 struct pcred *pc = p->p_cred; 1222 1223 n = SCARG(uap, gidsetsize); 1224 if (n < 0) 1225 return EINVAL; 1226 error = 0; 1227 bset = kbset = NULL; 1228 lset = NULL; 1229 if (n > 0) { 1230 n = min(pc->pc_ucred->cr_ngroups, n); 1231 sg = stackgap_init(p, 0); 1232 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1233 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1234 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1235 if (bset == NULL || kbset == NULL || lset == NULL) 1236 return ENOMEM; 1237 SCARG(&bsa, gidsetsize) = n; 1238 SCARG(&bsa, gidset) = bset; 1239 error = sys_getgroups(l, &bsa, retval); 1240 if (error != 0) 1241 goto out; 1242 error = copyin(bset, kbset, n * sizeof (gid_t)); 1243 if (error != 0) 1244 goto out; 1245 for (i = 0; i < n; i++) 1246 lset[i] = (linux_gid_t)kbset[i]; 1247 error = copyout(lset, SCARG(uap, gidset), 1248 n * sizeof (linux_gid_t)); 1249 } else 1250 *retval = pc->pc_ucred->cr_ngroups; 1251 out: 1252 if (kbset != NULL) 1253 free(kbset, M_TEMP); 1254 if (lset != NULL) 1255 free(lset, M_TEMP); 1256 return error; 1257 } 1258 1259 int 1260 linux_sys_setgroups16(l, v, retval) 1261 struct lwp *l; 1262 void *v; 1263 register_t *retval; 1264 { 1265 struct linux_sys_setgroups16_args /* { 1266 syscallarg(int) gidsetsize; 1267 syscallarg(linux_gid_t *) gidset; 1268 } */ *uap = v; 1269 struct proc *p = l->l_proc; 1270 caddr_t sg; 1271 int n; 1272 int error, i; 1273 struct sys_setgroups_args bsa; 1274 gid_t *bset, *kbset; 1275 linux_gid_t *lset; 1276 1277 n = SCARG(uap, gidsetsize); 1278 if (n < 0 || n > NGROUPS) 1279 return EINVAL; 1280 sg = stackgap_init(p, 0); 1281 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1282 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1283 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1284 if (lset == NULL || bset == NULL) 1285 return ENOMEM; 1286 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t)); 1287 if (error != 0) 1288 goto out; 1289 for (i = 0; i < n; i++) 1290 kbset[i] = (gid_t)lset[i]; 1291 error = copyout(kbset, bset, n * sizeof (gid_t)); 1292 if (error != 0) 1293 goto out; 1294 SCARG(&bsa, gidsetsize) = n; 1295 SCARG(&bsa, gidset) = bset; 1296 error = sys_setgroups(l, &bsa, retval); 1297 1298 out: 1299 if (lset != NULL) 1300 free(lset, M_TEMP); 1301 if (kbset != NULL) 1302 free(kbset, M_TEMP); 1303 1304 return error; 1305 } 1306 1307 #endif /* __i386__ || __m68k__ || __amd64__ */ 1308 1309 /* 1310 * We have nonexistent fsuid equal to uid. 1311 * If modification is requested, refuse. 1312 */ 1313 int 1314 linux_sys_setfsuid(l, v, retval) 1315 struct lwp *l; 1316 void *v; 1317 register_t *retval; 1318 { 1319 struct linux_sys_setfsuid_args /* { 1320 syscallarg(uid_t) uid; 1321 } */ *uap = v; 1322 struct proc *p = l->l_proc; 1323 uid_t uid; 1324 1325 uid = SCARG(uap, uid); 1326 if (p->p_cred->p_ruid != uid) 1327 return sys_nosys(l, v, retval); 1328 else 1329 return (0); 1330 } 1331 1332 /* XXX XXX XXX */ 1333 #ifndef alpha 1334 int 1335 linux_sys_getfsuid(l, v, retval) 1336 struct lwp *l; 1337 void *v; 1338 register_t *retval; 1339 { 1340 return sys_getuid(l, v, retval); 1341 } 1342 #endif 1343 1344 int 1345 linux_sys_setresuid(l, v, retval) 1346 struct lwp *l; 1347 void *v; 1348 register_t *retval; 1349 { 1350 struct linux_sys_setresuid_args /* { 1351 syscallarg(uid_t) ruid; 1352 syscallarg(uid_t) euid; 1353 syscallarg(uid_t) suid; 1354 } */ *uap = v; 1355 1356 /* 1357 * Note: These checks are a little different than the NetBSD 1358 * setreuid(2) call performs. This precisely follows the 1359 * behavior of the Linux kernel. 1360 */ 1361 1362 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1363 SCARG(uap, suid), 1364 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1365 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1366 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1367 } 1368 1369 int 1370 linux_sys_getresuid(l, v, retval) 1371 struct lwp *l; 1372 void *v; 1373 register_t *retval; 1374 { 1375 struct linux_sys_getresuid_args /* { 1376 syscallarg(uid_t *) ruid; 1377 syscallarg(uid_t *) euid; 1378 syscallarg(uid_t *) suid; 1379 } */ *uap = v; 1380 struct proc *p = l->l_proc; 1381 struct pcred *pc = p->p_cred; 1382 int error; 1383 1384 /* 1385 * Linux copies these values out to userspace like so: 1386 * 1387 * 1. Copy out ruid. 1388 * 2. If that succeeds, copy out euid. 1389 * 3. If both of those succeed, copy out suid. 1390 */ 1391 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid), 1392 sizeof(uid_t))) != 0) 1393 return (error); 1394 1395 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid), 1396 sizeof(uid_t))) != 0) 1397 return (error); 1398 1399 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t))); 1400 } 1401 1402 int 1403 linux_sys_ptrace(l, v, retval) 1404 struct lwp *l; 1405 void *v; 1406 register_t *retval; 1407 { 1408 struct linux_sys_ptrace_args /* { 1409 i386, m68k, powerpc: T=int 1410 alpha, amd64: T=long 1411 syscallarg(T) request; 1412 syscallarg(T) pid; 1413 syscallarg(T) addr; 1414 syscallarg(T) data; 1415 } */ *uap = v; 1416 const int *ptr; 1417 int request; 1418 int error; 1419 1420 ptr = linux_ptrace_request_map; 1421 request = SCARG(uap, request); 1422 while (*ptr != -1) 1423 if (*ptr++ == request) { 1424 struct sys_ptrace_args pta; 1425 1426 SCARG(&pta, req) = *ptr; 1427 SCARG(&pta, pid) = SCARG(uap, pid); 1428 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr); 1429 SCARG(&pta, data) = SCARG(uap, data); 1430 1431 /* 1432 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1433 * to continue where the process left off previously. 1434 * The same thing is achieved by addr == (caddr_t) 1 1435 * on NetBSD, so rewrite 'addr' appropriately. 1436 */ 1437 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1438 SCARG(&pta, addr) = (caddr_t) 1; 1439 1440 error = sys_ptrace(l, &pta, retval); 1441 if (error) 1442 return error; 1443 switch (request) { 1444 case LINUX_PTRACE_PEEKTEXT: 1445 case LINUX_PTRACE_PEEKDATA: 1446 error = copyout (retval, 1447 (caddr_t)SCARG(uap, data), 1448 sizeof *retval); 1449 *retval = SCARG(uap, data); 1450 break; 1451 default: 1452 break; 1453 } 1454 return error; 1455 } 1456 else 1457 ptr++; 1458 1459 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1460 } 1461 1462 int 1463 linux_sys_reboot(struct lwp *l, void *v, register_t *retval) 1464 { 1465 struct linux_sys_reboot_args /* { 1466 syscallarg(int) magic1; 1467 syscallarg(int) magic2; 1468 syscallarg(int) cmd; 1469 syscallarg(void *) arg; 1470 } */ *uap = v; 1471 struct sys_reboot_args /* { 1472 syscallarg(int) opt; 1473 syscallarg(char *) bootstr; 1474 } */ sra; 1475 struct proc *p = l->l_proc; 1476 int error; 1477 1478 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1479 return(error); 1480 1481 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1482 return(EINVAL); 1483 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1484 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1485 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1486 return(EINVAL); 1487 1488 switch (SCARG(uap, cmd)) { 1489 case LINUX_REBOOT_CMD_RESTART: 1490 SCARG(&sra, opt) = RB_AUTOBOOT; 1491 break; 1492 case LINUX_REBOOT_CMD_HALT: 1493 SCARG(&sra, opt) = RB_HALT; 1494 break; 1495 case LINUX_REBOOT_CMD_POWER_OFF: 1496 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1497 break; 1498 case LINUX_REBOOT_CMD_RESTART2: 1499 /* Reboot with an argument. */ 1500 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1501 SCARG(&sra, bootstr) = SCARG(uap, arg); 1502 break; 1503 case LINUX_REBOOT_CMD_CAD_ON: 1504 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1505 case LINUX_REBOOT_CMD_CAD_OFF: 1506 return(0); 1507 default: 1508 return(EINVAL); 1509 } 1510 1511 return(sys_reboot(l, &sra, retval)); 1512 } 1513 1514 /* 1515 * Copy of compat_12_sys_swapon(). 1516 */ 1517 int 1518 linux_sys_swapon(l, v, retval) 1519 struct lwp *l; 1520 void *v; 1521 register_t *retval; 1522 { 1523 struct sys_swapctl_args ua; 1524 struct linux_sys_swapon_args /* { 1525 syscallarg(const char *) name; 1526 } */ *uap = v; 1527 1528 SCARG(&ua, cmd) = SWAP_ON; 1529 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name)); 1530 SCARG(&ua, misc) = 0; /* priority */ 1531 return (sys_swapctl(l, &ua, retval)); 1532 } 1533 1534 /* 1535 * Stop swapping to the file or block device specified by path. 1536 */ 1537 int 1538 linux_sys_swapoff(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_swapoff_args /* { 1545 syscallarg(const char *) path; 1546 } */ *uap = v; 1547 1548 SCARG(&ua, cmd) = SWAP_OFF; 1549 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/ 1550 return (sys_swapctl(l, &ua, retval)); 1551 } 1552 1553 /* 1554 * Copy of compat_09_sys_setdomainname() 1555 */ 1556 /* ARGSUSED */ 1557 int 1558 linux_sys_setdomainname(l, v, retval) 1559 struct lwp *l; 1560 void *v; 1561 register_t *retval; 1562 { 1563 struct linux_sys_setdomainname_args /* { 1564 syscallarg(char *) domainname; 1565 syscallarg(int) len; 1566 } */ *uap = v; 1567 int name[2]; 1568 1569 name[0] = CTL_KERN; 1570 name[1] = KERN_DOMAINNAME; 1571 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1572 SCARG(uap, len), l)); 1573 } 1574 1575 /* 1576 * sysinfo() 1577 */ 1578 /* ARGSUSED */ 1579 int 1580 linux_sys_sysinfo(l, v, retval) 1581 struct lwp *l; 1582 void *v; 1583 register_t *retval; 1584 { 1585 struct linux_sys_sysinfo_args /* { 1586 syscallarg(struct linux_sysinfo *) arg; 1587 } */ *uap = v; 1588 struct linux_sysinfo si; 1589 struct loadavg *la; 1590 1591 si.uptime = time.tv_sec - boottime.tv_sec; 1592 la = &averunnable; 1593 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1594 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1595 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1596 si.totalram = ctob(physmem); 1597 si.freeram = uvmexp.free * uvmexp.pagesize; 1598 si.sharedram = 0; /* XXX */ 1599 si.bufferram = uvmexp.filepages * uvmexp.pagesize; 1600 si.totalswap = uvmexp.swpages * uvmexp.pagesize; 1601 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1602 si.procs = nprocs; 1603 1604 /* The following are only present in newer Linux kernels. */ 1605 si.totalbig = 0; 1606 si.freebig = 0; 1607 si.mem_unit = 1; 1608 1609 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1610 } 1611 1612 #define bsd_to_linux_rlimit1(l, b, f) \ 1613 (l)->f = ((b)->f == RLIM_INFINITY || \ 1614 ((b)->f & 0xffffffff00000000ULL) != 0) ? \ 1615 LINUX_RLIM_INFINITY : (int32_t)(b)->f 1616 #define bsd_to_linux_rlimit(l, b) \ 1617 bsd_to_linux_rlimit1(l, b, rlim_cur); \ 1618 bsd_to_linux_rlimit1(l, b, rlim_max) 1619 1620 #define linux_to_bsd_rlimit1(b, l, f) \ 1621 (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f 1622 #define linux_to_bsd_rlimit(b, l) \ 1623 linux_to_bsd_rlimit1(b, l, rlim_cur); \ 1624 linux_to_bsd_rlimit1(b, l, rlim_max) 1625 1626 static int 1627 linux_to_bsd_limit(lim) 1628 int lim; 1629 { 1630 switch (lim) { 1631 case LINUX_RLIMIT_CPU: 1632 return RLIMIT_CPU; 1633 case LINUX_RLIMIT_FSIZE: 1634 return RLIMIT_FSIZE; 1635 case LINUX_RLIMIT_DATA: 1636 return RLIMIT_DATA; 1637 case LINUX_RLIMIT_STACK: 1638 return RLIMIT_STACK; 1639 case LINUX_RLIMIT_CORE: 1640 return RLIMIT_CORE; 1641 case LINUX_RLIMIT_RSS: 1642 return RLIMIT_RSS; 1643 case LINUX_RLIMIT_NPROC: 1644 return RLIMIT_NPROC; 1645 case LINUX_RLIMIT_NOFILE: 1646 return RLIMIT_NOFILE; 1647 case LINUX_RLIMIT_MEMLOCK: 1648 return RLIMIT_MEMLOCK; 1649 case LINUX_RLIMIT_AS: 1650 case LINUX_RLIMIT_LOCKS: 1651 return -EOPNOTSUPP; 1652 default: 1653 return -EINVAL; 1654 } 1655 } 1656 1657 1658 int 1659 linux_sys_getrlimit(l, v, retval) 1660 struct lwp *l; 1661 void *v; 1662 register_t *retval; 1663 { 1664 struct linux_sys_getrlimit_args /* { 1665 syscallarg(int) which; 1666 syscallarg(struct orlimit *) rlp; 1667 } */ *uap = v; 1668 struct proc *p = l->l_proc; 1669 caddr_t sg = stackgap_init(p, 0); 1670 struct sys_getrlimit_args ap; 1671 struct rlimit rl; 1672 struct orlimit orl; 1673 int error; 1674 1675 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1676 if ((error = SCARG(&ap, which)) < 0) 1677 return -error; 1678 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1679 if ((error = sys_getrlimit(l, &ap, retval)) != 0) 1680 return error; 1681 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0) 1682 return error; 1683 bsd_to_linux_rlimit(&orl, &rl); 1684 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1685 } 1686 1687 int 1688 linux_sys_setrlimit(l, v, retval) 1689 struct lwp *l; 1690 void *v; 1691 register_t *retval; 1692 { 1693 struct linux_sys_setrlimit_args /* { 1694 syscallarg(int) which; 1695 syscallarg(struct orlimit *) rlp; 1696 } */ *uap = v; 1697 struct proc *p = l->l_proc; 1698 caddr_t sg = stackgap_init(p, 0); 1699 struct sys_getrlimit_args ap; 1700 struct rlimit rl; 1701 struct orlimit orl; 1702 int error; 1703 1704 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1705 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1706 if ((error = SCARG(&ap, which)) < 0) 1707 return -error; 1708 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1709 return error; 1710 linux_to_bsd_rlimit(&rl, &orl); 1711 if ((error = copyout(&rl, SCARG(&ap, rlp), sizeof(rl))) != 0) 1712 return error; 1713 return sys_setrlimit(l, &ap, retval); 1714 } 1715 1716 #if !defined(__mips__) && !defined(__amd64__) 1717 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1718 int 1719 linux_sys_ugetrlimit(l, v, retval) 1720 struct lwp *l; 1721 void *v; 1722 register_t *retval; 1723 { 1724 return linux_sys_getrlimit(l, v, retval); 1725 } 1726 #endif 1727 1728 /* 1729 * This gets called for unsupported syscalls. The difference to sys_nosys() 1730 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1731 * This is the way Linux does it and glibc depends on this behaviour. 1732 */ 1733 int 1734 linux_sys_nosys(l, v, retval) 1735 struct lwp *l; 1736 void *v; 1737 register_t *retval; 1738 { 1739 return (ENOSYS); 1740 } 1741
Indexes created Mon Sep 22 21:09:42 GMT 2025