1 /* $NetBSD: linux_misc.c,v 1.252 2021/09/07 11:43:04 riastradh Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 1998, 1999, 2008 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 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Linux compatibility module. Try to deal with various Linux system calls. 35 */ 36 37 /* 38 * These functions have been moved to multiarch to allow 39 * selection of which machines include them to be 40 * determined by the individual files.linux_<arch> files. 41 * 42 * Function in multiarch: 43 * linux_sys_break : linux_break.c 44 * linux_sys_alarm : linux_misc_notalpha.c 45 * linux_sys_getresgid : linux_misc_notalpha.c 46 * linux_sys_nice : linux_misc_notalpha.c 47 * linux_sys_readdir : linux_misc_notalpha.c 48 * linux_sys_setresgid : linux_misc_notalpha.c 49 * linux_sys_time : linux_misc_notalpha.c 50 * linux_sys_utime : linux_misc_notalpha.c 51 * linux_sys_waitpid : linux_misc_notalpha.c 52 * linux_sys_old_mmap : linux_oldmmap.c 53 * linux_sys_oldolduname : linux_oldolduname.c 54 * linux_sys_oldselect : linux_oldselect.c 55 * linux_sys_olduname : linux_olduname.c 56 * linux_sys_pipe : linux_pipe.c 57 */ 58 59 #include <sys/cdefs.h> 60 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.252 2021/09/07 11:43:04 riastradh Exp $"); 61 62 #include <sys/param.h> 63 #include <sys/systm.h> 64 #include <sys/namei.h> 65 #include <sys/proc.h> 66 #include <sys/dirent.h> 67 #include <sys/file.h> 68 #include <sys/stat.h> 69 #include <sys/filedesc.h> 70 #include <sys/ioctl.h> 71 #include <sys/kernel.h> 72 #include <sys/malloc.h> 73 #include <sys/mbuf.h> 74 #include <sys/mman.h> 75 #include <sys/mount.h> 76 #include <sys/poll.h> 77 #include <sys/prot.h> 78 #include <sys/reboot.h> 79 #include <sys/resource.h> 80 #include <sys/resourcevar.h> 81 #include <sys/select.h> 82 #include <sys/signal.h> 83 #include <sys/signalvar.h> 84 #include <sys/socket.h> 85 #include <sys/time.h> 86 #include <sys/times.h> 87 #include <sys/vnode.h> 88 #include <sys/uio.h> 89 #include <sys/wait.h> 90 #include <sys/utsname.h> 91 #include <sys/unistd.h> 92 #include <sys/vfs_syscalls.h> 93 #include <sys/swap.h> /* for SWAP_ON */ 94 #include <sys/sysctl.h> /* for KERN_DOMAINNAME */ 95 #include <sys/kauth.h> 96 #include <sys/futex.h> 97 98 #include <sys/ptrace.h> 99 #include <machine/ptrace.h> 100 101 #include <sys/syscall.h> 102 #include <sys/syscallargs.h> 103 104 #include <compat/sys/resource.h> 105 106 #include <compat/linux/common/linux_machdep.h> 107 #include <compat/linux/common/linux_types.h> 108 #include <compat/linux/common/linux_signal.h> 109 #include <compat/linux/common/linux_ipc.h> 110 #include <compat/linux/common/linux_sem.h> 111 112 #include <compat/linux/common/linux_fcntl.h> 113 #include <compat/linux/common/linux_mmap.h> 114 #include <compat/linux/common/linux_dirent.h> 115 #include <compat/linux/common/linux_util.h> 116 #include <compat/linux/common/linux_misc.h> 117 #include <compat/linux/common/linux_statfs.h> 118 #include <compat/linux/common/linux_limit.h> 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 #include <compat/linux/common/linux_sched.h> 123 124 #include <compat/linux/linux_syscallargs.h> 125 126 const int linux_ptrace_request_map[] = { 127 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 128 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 129 LINUX_PTRACE_PEEKDATA, PT_READ_D, 130 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 131 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 132 LINUX_PTRACE_CONT, PT_CONTINUE, 133 LINUX_PTRACE_KILL, PT_KILL, 134 LINUX_PTRACE_ATTACH, PT_ATTACH, 135 LINUX_PTRACE_DETACH, PT_DETACH, 136 # ifdef PT_STEP 137 LINUX_PTRACE_SINGLESTEP, PT_STEP, 138 # endif 139 LINUX_PTRACE_SYSCALL, PT_SYSCALL, 140 -1 141 }; 142 143 const struct linux_mnttypes linux_fstypes[] = { 144 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 145 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 146 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 148 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 149 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 150 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 151 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 152 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 153 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 154 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 155 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 156 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 157 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 158 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 159 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 160 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 161 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 162 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 163 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 164 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC }, 165 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC }, 166 { MOUNT_TMPFS, LINUX_TMPFS_SUPER_MAGIC } 167 }; 168 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 169 170 # ifdef DEBUG_LINUX 171 #define DPRINTF(a) uprintf a 172 # else 173 #define DPRINTF(a) 174 # endif 175 176 /* Local linux_misc.c functions: */ 177 static void linux_to_bsd_mmap_args(struct sys_mmap_args *, 178 const struct linux_sys_mmap_args *); 179 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *, 180 register_t *, off_t); 181 182 183 /* 184 * The information on a terminated (or stopped) process needs 185 * to be converted in order for Linux binaries to get a valid signal 186 * number out of it. 187 */ 188 int 189 bsd_to_linux_wstat(int st) 190 { 191 192 int sig; 193 194 if (WIFSIGNALED(st)) { 195 sig = WTERMSIG(st); 196 if (sig >= 0 && sig < NSIG) 197 st= (st & ~0177) | native_to_linux_signo[sig]; 198 } else if (WIFSTOPPED(st)) { 199 sig = WSTOPSIG(st); 200 if (sig >= 0 && sig < NSIG) 201 st = (st & ~0xff00) | 202 (native_to_linux_signo[sig] << 8); 203 } 204 return st; 205 } 206 207 /* 208 * wait4(2). Passed on to the NetBSD call, surrounded by code to 209 * reserve some space for a NetBSD-style wait status, and converting 210 * it to what Linux wants. 211 */ 212 int 213 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval) 214 { 215 /* { 216 syscallarg(int) pid; 217 syscallarg(int *) status; 218 syscallarg(int) options; 219 syscallarg(struct rusage50 *) rusage; 220 } */ 221 int error, status, options, linux_options, pid = SCARG(uap, pid); 222 struct rusage50 ru50; 223 struct rusage ru; 224 proc_t *p; 225 226 linux_options = SCARG(uap, options); 227 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS)) 228 return (EINVAL); 229 230 options = 0; 231 if (linux_options & LINUX_WAIT4_WNOHANG) 232 options |= WNOHANG; 233 if (linux_options & LINUX_WAIT4_WUNTRACED) 234 options |= WUNTRACED; 235 if (linux_options & LINUX_WAIT4_WCONTINUED) 236 options |= WCONTINUED; 237 if (linux_options & LINUX_WAIT4_WALL) 238 options |= WALLSIG; 239 if (linux_options & LINUX_WAIT4_WCLONE) 240 options |= WALTSIG; 241 # ifdef DIAGNOSTIC 242 if (linux_options & LINUX_WAIT4_WNOTHREAD) 243 printf("WARNING: %s: linux process %d.%d called " 244 "waitpid with __WNOTHREAD set!\n", 245 __FILE__, l->l_proc->p_pid, l->l_lid); 246 247 # endif 248 249 error = do_sys_wait(&pid, &status, options, 250 SCARG(uap, rusage) != NULL ? &ru : NULL); 251 252 retval[0] = pid; 253 if (pid == 0) 254 return error; 255 256 p = curproc; 257 mutex_enter(p->p_lock); 258 sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */ 259 mutex_exit(p->p_lock); 260 261 if (SCARG(uap, rusage) != NULL) { 262 rusage_to_rusage50(&ru, &ru50); 263 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 264 } 265 266 if (error == 0 && SCARG(uap, status) != NULL) { 267 status = bsd_to_linux_wstat(status); 268 error = copyout(&status, SCARG(uap, status), sizeof status); 269 } 270 271 return error; 272 } 273 274 /* 275 * Linux brk(2). Like native, but always return the new break value. 276 */ 277 int 278 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval) 279 { 280 /* { 281 syscallarg(char *) nsize; 282 } */ 283 struct proc *p = l->l_proc; 284 struct vmspace *vm = p->p_vmspace; 285 struct sys_obreak_args oba; 286 287 SCARG(&oba, nsize) = SCARG(uap, nsize); 288 289 (void) sys_obreak(l, &oba, retval); 290 retval[0] = (register_t)((char *)vm->vm_daddr + ptoa(vm->vm_dsize)); 291 return 0; 292 } 293 294 /* 295 * Implement the fs stat functions. Straightforward. 296 */ 297 int 298 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval) 299 { 300 /* { 301 syscallarg(const char *) path; 302 syscallarg(struct linux_statfs *) sp; 303 } */ 304 struct statvfs *sb; 305 struct linux_statfs ltmp; 306 int error; 307 308 sb = STATVFSBUF_GET(); 309 error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb); 310 if (error == 0) { 311 bsd_to_linux_statfs(sb, <mp); 312 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp); 313 } 314 STATVFSBUF_PUT(sb); 315 316 return error; 317 } 318 319 int 320 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval) 321 { 322 /* { 323 syscallarg(int) fd; 324 syscallarg(struct linux_statfs *) sp; 325 } */ 326 struct statvfs *sb; 327 struct linux_statfs ltmp; 328 int error; 329 330 sb = STATVFSBUF_GET(); 331 error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb); 332 if (error == 0) { 333 bsd_to_linux_statfs(sb, <mp); 334 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp); 335 } 336 STATVFSBUF_PUT(sb); 337 338 return error; 339 } 340 341 /* 342 * uname(). Just copy the info from the various strings stored in the 343 * kernel, and put it in the Linux utsname structure. That structure 344 * is almost the same as the NetBSD one, only it has fields 65 characters 345 * long, and an extra domainname field. 346 */ 347 int 348 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval) 349 { 350 /* { 351 syscallarg(struct linux_utsname *) up; 352 } */ 353 struct linux_utsname luts; 354 355 memset(&luts, 0, sizeof(luts)); 356 strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 357 strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 358 strlcpy(luts.l_release, linux_release, sizeof(luts.l_release)); 359 strlcpy(luts.l_version, linux_version, sizeof(luts.l_version)); 360 strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine)); 361 strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 362 363 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 364 } 365 366 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 367 /* Used indirectly on: arm, i386, m68k */ 368 369 /* 370 * New type Linux mmap call. 371 * Only called directly on machines with >= 6 free regs. 372 */ 373 int 374 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval) 375 { 376 /* { 377 syscallarg(unsigned long) addr; 378 syscallarg(size_t) len; 379 syscallarg(int) prot; 380 syscallarg(int) flags; 381 syscallarg(int) fd; 382 syscallarg(linux_off_t) offset; 383 } */ 384 385 if (SCARG(uap, offset) & PAGE_MASK) 386 return EINVAL; 387 388 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 389 } 390 391 /* 392 * Guts of most architectures' mmap64() implementations. This shares 393 * its list of arguments with linux_sys_mmap(). 394 * 395 * The difference in linux_sys_mmap2() is that "offset" is actually 396 * (offset / pagesize), not an absolute byte count. This translation 397 * to pagesize offsets is done inside glibc between the mmap64() call 398 * point, and the actual syscall. 399 */ 400 int 401 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval) 402 { 403 /* { 404 syscallarg(unsigned long) addr; 405 syscallarg(size_t) len; 406 syscallarg(int) prot; 407 syscallarg(int) flags; 408 syscallarg(int) fd; 409 syscallarg(linux_off_t) offset; 410 } */ 411 412 return linux_mmap(l, uap, retval, 413 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 414 } 415 416 /* 417 * Massage arguments and call system mmap(2). 418 */ 419 static int 420 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset) 421 { 422 struct sys_mmap_args cma; 423 int error; 424 size_t mmoff=0; 425 426 linux_to_bsd_mmap_args(&cma, uap); 427 SCARG(&cma, pos) = offset; 428 429 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 430 /* 431 * Request for stack-like memory segment. On linux, this 432 * works by mmap()ping (small) segment, which is automatically 433 * extended when page fault happens below the currently 434 * allocated area. We emulate this by allocating (typically 435 * bigger) segment sized at current stack size limit, and 436 * offsetting the requested and returned address accordingly. 437 * Since physical pages are only allocated on-demand, this 438 * is effectively identical. 439 */ 440 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 441 442 if (SCARG(&cma, len) < ssl) { 443 /* Compute the address offset */ 444 mmoff = round_page(ssl) - SCARG(uap, len); 445 446 if (SCARG(&cma, addr)) 447 SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff; 448 449 SCARG(&cma, len) = (size_t) ssl; 450 } 451 } 452 453 error = sys_mmap(l, &cma, retval); 454 if (error) 455 return (error); 456 457 /* Shift the returned address for stack-like segment if necessary */ 458 retval[0] += mmoff; 459 460 return (0); 461 } 462 463 static void 464 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap) 465 { 466 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 467 468 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 469 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 470 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 471 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 472 flags |= cvtto_bsd_mask(fl, LINUX_MAP_LOCKED, MAP_WIRED); 473 /* XXX XAX ERH: Any other flags here? There are more defined... */ 474 475 SCARG(cma, addr) = (void *)SCARG(uap, addr); 476 SCARG(cma, len) = SCARG(uap, len); 477 SCARG(cma, prot) = SCARG(uap, prot); 478 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 479 SCARG(cma, prot) |= VM_PROT_READ; 480 SCARG(cma, flags) = flags; 481 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 482 SCARG(cma, PAD) = 0; 483 } 484 485 #define LINUX_MREMAP_MAYMOVE 1 486 #define LINUX_MREMAP_FIXED 2 487 488 int 489 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval) 490 { 491 /* { 492 syscallarg(void *) old_address; 493 syscallarg(size_t) old_size; 494 syscallarg(size_t) new_size; 495 syscallarg(u_long) flags; 496 } */ 497 498 struct proc *p; 499 struct vm_map *map; 500 vaddr_t oldva; 501 vaddr_t newva; 502 size_t oldsize; 503 size_t newsize; 504 int flags; 505 int uvmflags; 506 int error; 507 508 flags = SCARG(uap, flags); 509 oldva = (vaddr_t)SCARG(uap, old_address); 510 oldsize = round_page(SCARG(uap, old_size)); 511 newsize = round_page(SCARG(uap, new_size)); 512 if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) { 513 error = EINVAL; 514 goto done; 515 } 516 if ((flags & LINUX_MREMAP_FIXED) != 0) { 517 if ((flags & LINUX_MREMAP_MAYMOVE) == 0) { 518 error = EINVAL; 519 goto done; 520 } 521 #if 0 /* notyet */ 522 newva = SCARG(uap, new_address); 523 uvmflags = MAP_FIXED; 524 #else /* notyet */ 525 error = EOPNOTSUPP; 526 goto done; 527 #endif /* notyet */ 528 } else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) { 529 uvmflags = 0; 530 } else { 531 newva = oldva; 532 uvmflags = MAP_FIXED; 533 } 534 p = l->l_proc; 535 map = &p->p_vmspace->vm_map; 536 error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p, 537 uvmflags); 538 539 done: 540 *retval = (error != 0) ? 0 : (register_t)newva; 541 return error; 542 } 543 544 #ifdef USRSTACK 545 int 546 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval) 547 { 548 /* { 549 syscallarg(const void *) start; 550 syscallarg(unsigned long) len; 551 syscallarg(int) prot; 552 } */ 553 struct vm_map_entry *entry; 554 struct vm_map *map; 555 struct proc *p; 556 vaddr_t end, start, len, stacklim; 557 int prot, grows; 558 559 start = (vaddr_t)SCARG(uap, start); 560 len = round_page(SCARG(uap, len)); 561 prot = SCARG(uap, prot); 562 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 563 prot &= ~grows; 564 end = start + len; 565 566 if (start & PAGE_MASK) 567 return EINVAL; 568 if (end < start) 569 return EINVAL; 570 if (end == start) 571 return 0; 572 573 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 574 return EINVAL; 575 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP)) 576 return EINVAL; 577 578 p = l->l_proc; 579 map = &p->p_vmspace->vm_map; 580 vm_map_lock(map); 581 # ifdef notdef 582 VM_MAP_RANGE_CHECK(map, start, end); 583 # endif 584 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 585 vm_map_unlock(map); 586 return ENOMEM; 587 } 588 589 /* 590 * Approximate the behaviour of PROT_GROWS{DOWN,UP}. 591 */ 592 593 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur; 594 if (grows & LINUX_PROT_GROWSDOWN) { 595 if (USRSTACK - stacklim <= start && start < USRSTACK) { 596 start = USRSTACK - stacklim; 597 } else { 598 start = entry->start; 599 } 600 } else if (grows & LINUX_PROT_GROWSUP) { 601 if (USRSTACK <= end && end < USRSTACK + stacklim) { 602 end = USRSTACK + stacklim; 603 } else { 604 end = entry->end; 605 } 606 } 607 vm_map_unlock(map); 608 return uvm_map_protect_user(l, start, end, prot); 609 } 610 #endif /* USRSTACK */ 611 612 /* 613 * This code is partly stolen from src/lib/libc/compat-43/times.c 614 */ 615 616 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 617 618 int 619 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval) 620 { 621 /* { 622 syscallarg(struct times *) tms; 623 } */ 624 struct proc *p = l->l_proc; 625 struct timeval t; 626 int error; 627 628 if (SCARG(uap, tms)) { 629 struct linux_tms ltms; 630 struct rusage ru; 631 632 memset(<ms, 0, sizeof(ltms)); 633 634 mutex_enter(p->p_lock); 635 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 636 ltms.ltms_utime = CONVTCK(ru.ru_utime); 637 ltms.ltms_stime = CONVTCK(ru.ru_stime); 638 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 639 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 640 mutex_exit(p->p_lock); 641 642 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 643 return error; 644 } 645 646 getmicrouptime(&t); 647 648 retval[0] = ((linux_clock_t)(CONVTCK(t))); 649 return 0; 650 } 651 652 #undef CONVTCK 653 654 /* 655 * Linux 'readdir' call. This code is mostly taken from the 656 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 657 * an attempt has been made to keep it a little cleaner (failing 658 * miserably, because of the cruft needed if count 1 is passed). 659 * 660 * The d_off field should contain the offset of the next valid entry, 661 * but in Linux it has the offset of the entry itself. We emulate 662 * that bug here. 663 * 664 * Read in BSD-style entries, convert them, and copy them out. 665 * 666 * Note that this doesn't handle union-mounted filesystems. 667 */ 668 int 669 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval) 670 { 671 /* { 672 syscallarg(int) fd; 673 syscallarg(struct linux_dirent *) dent; 674 syscallarg(unsigned int) count; 675 } */ 676 struct dirent *bdp; 677 struct vnode *vp; 678 char *inp, *tbuf; /* BSD-format */ 679 int len, reclen; /* BSD-format */ 680 char *outp; /* Linux-format */ 681 int resid, linux_reclen = 0; /* Linux-format */ 682 struct file *fp; 683 struct uio auio; 684 struct iovec aiov; 685 struct linux_dirent idb; 686 off_t off; /* true file offset */ 687 int buflen, error, eofflag, nbytes, oldcall; 688 struct vattr va; 689 off_t *cookiebuf = NULL, *cookie; 690 int ncookies; 691 692 /* fd_getvnode() will use the descriptor for us */ 693 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 694 return (error); 695 696 if ((fp->f_flag & FREAD) == 0) { 697 error = EBADF; 698 goto out1; 699 } 700 701 vp = (struct vnode *)fp->f_data; 702 if (vp->v_type != VDIR) { 703 error = ENOTDIR; 704 goto out1; 705 } 706 707 vn_lock(vp, LK_SHARED | LK_RETRY); 708 error = VOP_GETATTR(vp, &va, l->l_cred); 709 VOP_UNLOCK(vp); 710 if (error) 711 goto out1; 712 713 nbytes = SCARG(uap, count); 714 if (nbytes == 1) { /* emulating old, broken behaviour */ 715 nbytes = sizeof (idb); 716 buflen = uimax(va.va_blocksize, nbytes); 717 oldcall = 1; 718 } else { 719 buflen = uimin(MAXBSIZE, nbytes); 720 if (buflen < va.va_blocksize) 721 buflen = va.va_blocksize; 722 oldcall = 0; 723 } 724 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 725 726 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 727 off = fp->f_offset; 728 again: 729 aiov.iov_base = tbuf; 730 aiov.iov_len = buflen; 731 auio.uio_iov = &aiov; 732 auio.uio_iovcnt = 1; 733 auio.uio_rw = UIO_READ; 734 auio.uio_resid = buflen; 735 auio.uio_offset = off; 736 UIO_SETUP_SYSSPACE(&auio); 737 /* 738 * First we read into the malloc'ed buffer, then 739 * we massage it into user space, one record at a time. 740 */ 741 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 742 &ncookies); 743 if (error) 744 goto out; 745 746 inp = tbuf; 747 outp = (void *)SCARG(uap, dent); 748 resid = nbytes; 749 if ((len = buflen - auio.uio_resid) == 0) 750 goto eof; 751 752 for (cookie = cookiebuf; len > 0; len -= reclen) { 753 bdp = (struct dirent *)inp; 754 reclen = bdp->d_reclen; 755 if (reclen & 3) { 756 error = EIO; 757 goto out; 758 } 759 if (bdp->d_fileno == 0) { 760 inp += reclen; /* it is a hole; squish it out */ 761 if (cookie) 762 off = *cookie++; 763 else 764 off += reclen; 765 continue; 766 } 767 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 768 if (reclen > len || resid < linux_reclen) { 769 /* entry too big for buffer, so just stop */ 770 outp++; 771 break; 772 } 773 /* 774 * Massage in place to make a Linux-shaped dirent (otherwise 775 * we have to worry about touching user memory outside of 776 * the copyout() call). 777 */ 778 memset(&idb, 0, sizeof(idb)); 779 idb.d_ino = bdp->d_fileno; 780 /* 781 * The old readdir() call misuses the offset and reclen fields. 782 */ 783 if (oldcall) { 784 idb.d_off = (linux_off_t)linux_reclen; 785 idb.d_reclen = (u_short)bdp->d_namlen; 786 } else { 787 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 788 compat_offseterr(vp, "linux_getdents"); 789 error = EINVAL; 790 goto out; 791 } 792 idb.d_off = (linux_off_t)off; 793 idb.d_reclen = (u_short)linux_reclen; 794 /* Linux puts d_type at the end of each record */ 795 *((char *)&idb + idb.d_reclen - 1) = bdp->d_type; 796 } 797 memcpy(idb.d_name, bdp->d_name, 798 MIN(sizeof(idb.d_name), bdp->d_namlen + 1)); 799 if ((error = copyout((void *)&idb, outp, linux_reclen))) 800 goto out; 801 /* advance past this real entry */ 802 inp += reclen; 803 if (cookie) 804 off = *cookie++; /* each entry points to itself */ 805 else 806 off += reclen; 807 /* advance output past Linux-shaped entry */ 808 outp += linux_reclen; 809 resid -= linux_reclen; 810 if (oldcall) 811 break; 812 } 813 814 /* if we squished out the whole block, try again */ 815 if (outp == (void *)SCARG(uap, dent)) { 816 if (cookiebuf) 817 free(cookiebuf, M_TEMP); 818 cookiebuf = NULL; 819 goto again; 820 } 821 fp->f_offset = off; /* update the vnode offset */ 822 823 if (oldcall) 824 nbytes = resid + linux_reclen; 825 826 eof: 827 *retval = nbytes - resid; 828 out: 829 VOP_UNLOCK(vp); 830 if (cookiebuf) 831 free(cookiebuf, M_TEMP); 832 free(tbuf, M_TEMP); 833 out1: 834 fd_putfile(SCARG(uap, fd)); 835 return error; 836 } 837 838 /* 839 * Even when just using registers to pass arguments to syscalls you can 840 * have 5 of them on the i386. So this newer version of select() does 841 * this. 842 */ 843 int 844 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval) 845 { 846 /* { 847 syscallarg(int) nfds; 848 syscallarg(fd_set *) readfds; 849 syscallarg(fd_set *) writefds; 850 syscallarg(fd_set *) exceptfds; 851 syscallarg(struct timeval50 *) timeout; 852 } */ 853 854 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 855 SCARG(uap, writefds), SCARG(uap, exceptfds), 856 (struct linux_timeval *)SCARG(uap, timeout)); 857 } 858 859 /* 860 * Common code for the old and new versions of select(). A couple of 861 * things are important: 862 * 1) return the amount of time left in the 'timeout' parameter 863 * 2) select never returns ERESTART on Linux, always return EINTR 864 */ 865 int 866 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds, 867 fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout) 868 { 869 struct timespec ts0, ts1, uts, *ts = NULL; 870 struct linux_timeval ltv; 871 int error; 872 873 /* 874 * Store current time for computation of the amount of 875 * time left. 876 */ 877 if (timeout) { 878 if ((error = copyin(timeout, <v, sizeof(ltv)))) 879 return error; 880 uts.tv_sec = ltv.tv_sec; 881 uts.tv_nsec = (long)((unsigned long)ltv.tv_usec * 1000); 882 if (itimespecfix(&uts)) { 883 /* 884 * The timeval was invalid. Convert it to something 885 * valid that will act as it does under Linux. 886 */ 887 uts.tv_sec += uts.tv_nsec / 1000000000; 888 uts.tv_nsec %= 1000000000; 889 if (uts.tv_nsec < 0) { 890 uts.tv_sec -= 1; 891 uts.tv_nsec += 1000000000; 892 } 893 if (uts.tv_sec < 0) 894 timespecclear(&uts); 895 } 896 ts = &uts; 897 nanotime(&ts0); 898 } 899 900 error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL); 901 902 if (error) { 903 /* 904 * See fs/select.c in the Linux kernel. Without this, 905 * Maelstrom doesn't work. 906 */ 907 if (error == ERESTART) 908 error = EINTR; 909 return error; 910 } 911 912 if (timeout) { 913 if (*retval) { 914 /* 915 * Compute how much time was left of the timeout, 916 * by subtracting the current time and the time 917 * before we started the call, and subtracting 918 * that result from the user-supplied value. 919 */ 920 nanotime(&ts1); 921 timespecsub(&ts1, &ts0, &ts1); 922 timespecsub(&uts, &ts1, &uts); 923 if (uts.tv_sec < 0) 924 timespecclear(&uts); 925 } else 926 timespecclear(&uts); 927 ltv.tv_sec = uts.tv_sec; 928 ltv.tv_usec = uts.tv_nsec / 1000; 929 if ((error = copyout(<v, timeout, sizeof(ltv)))) 930 return error; 931 } 932 933 return 0; 934 } 935 936 /* 937 * Derived from FreeBSD's sys/compat/linux/linux_misc.c:linux_pselect6() 938 * which was contributed by Dmitry Chagin 939 * https://svnweb.freebsd.org/base?view=revision&revision=283403 940 */ 941 int 942 linux_sys_pselect6(struct lwp *l, 943 const struct linux_sys_pselect6_args *uap, register_t *retval) 944 { 945 /* { 946 syscallarg(int) nfds; 947 syscallarg(fd_set *) readfds; 948 syscallarg(fd_set *) writefds; 949 syscallarg(fd_set *) exceptfds; 950 syscallarg(struct timespec *) timeout; 951 syscallarg(linux_sized_sigset_t *) ss; 952 } */ 953 struct timespec uts, ts0, ts1, *tsp; 954 linux_sized_sigset_t lsss; 955 struct linux_timespec lts; 956 linux_sigset_t lss; 957 sigset_t *ssp; 958 sigset_t ss; 959 int error; 960 961 ssp = NULL; 962 if (SCARG(uap, ss) != NULL) { 963 if ((error = copyin(SCARG(uap, ss), &lsss, sizeof(lsss))) != 0) 964 return (error); 965 if (lsss.ss_len != sizeof(lss)) 966 return (EINVAL); 967 if (lsss.ss != NULL) { 968 if ((error = copyin(lsss.ss, &lss, sizeof(lss))) != 0) 969 return (error); 970 linux_to_native_sigset(&ss, &lss); 971 ssp = &ss; 972 } 973 } 974 975 if (SCARG(uap, timeout) != NULL) { 976 error = copyin(SCARG(uap, timeout), <s, sizeof(lts)); 977 if (error != 0) 978 return (error); 979 linux_to_native_timespec(&uts, <s); 980 981 if (itimespecfix(&uts)) 982 return (EINVAL); 983 984 nanotime(&ts0); 985 tsp = &uts; 986 } else { 987 tsp = NULL; 988 } 989 990 error = selcommon(retval, SCARG(uap, nfds), SCARG(uap, readfds), 991 SCARG(uap, writefds), SCARG(uap, exceptfds), tsp, ssp); 992 993 if (error == 0 && tsp != NULL) { 994 if (retval != 0) { 995 /* 996 * Compute how much time was left of the timeout, 997 * by subtracting the current time and the time 998 * before we started the call, and subtracting 999 * that result from the user-supplied value. 1000 */ 1001 nanotime(&ts1); 1002 timespecsub(&ts1, &ts0, &ts1); 1003 timespecsub(&uts, &ts1, &uts); 1004 if (uts.tv_sec < 0) 1005 timespecclear(&uts); 1006 } else { 1007 timespecclear(&uts); 1008 } 1009 1010 native_to_linux_timespec(<s, &uts); 1011 error = copyout(<s, SCARG(uap, timeout), sizeof(lts)); 1012 } 1013 1014 return (error); 1015 } 1016 1017 int 1018 linux_sys_ppoll(struct lwp *l, 1019 const struct linux_sys_ppoll_args *uap, register_t *retval) 1020 { 1021 /* { 1022 syscallarg(struct pollfd *) fds; 1023 syscallarg(u_int) nfds; 1024 syscallarg(struct linux_timespec *) timeout; 1025 syscallarg(linux_sigset_t *) sigset; 1026 } */ 1027 struct linux_timespec lts0, *lts; 1028 struct timespec ts0, *ts = NULL; 1029 linux_sigset_t lsigmask0, *lsigmask; 1030 sigset_t sigmask0, *sigmask = NULL; 1031 int error; 1032 1033 lts = SCARG(uap, timeout); 1034 if (lts) { 1035 if ((error = copyin(lts, <s0, sizeof(lts0))) != 0) 1036 return error; 1037 linux_to_native_timespec(&ts0, <s0); 1038 ts = &ts0; 1039 } 1040 1041 lsigmask = SCARG(uap, sigset); 1042 if (lsigmask) { 1043 if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0)))) 1044 return error; 1045 linux_to_native_sigset(&sigmask0, &lsigmask0); 1046 sigmask = &sigmask0; 1047 } 1048 1049 return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds), 1050 ts, sigmask); 1051 } 1052 1053 /* 1054 * Set the 'personality' (emulation mode) for the current process. Only 1055 * accept the Linux personality here (0). This call is needed because 1056 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1057 * ELF binaries run in Linux mode, not SVR4 mode. 1058 */ 1059 int 1060 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval) 1061 { 1062 /* { 1063 syscallarg(unsigned long) per; 1064 } */ 1065 struct linux_emuldata *led; 1066 int per; 1067 1068 per = SCARG(uap, per); 1069 led = l->l_emuldata; 1070 if (per == LINUX_PER_QUERY) { 1071 retval[0] = led->led_personality; 1072 return 0; 1073 } 1074 1075 switch (per & LINUX_PER_MASK) { 1076 case LINUX_PER_LINUX: 1077 case LINUX_PER_LINUX32: 1078 led->led_personality = per; 1079 break; 1080 1081 default: 1082 return EINVAL; 1083 } 1084 1085 retval[0] = per; 1086 return 0; 1087 } 1088 1089 /* 1090 * We have nonexistent fsuid equal to uid. 1091 * If modification is requested, refuse. 1092 */ 1093 int 1094 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval) 1095 { 1096 /* { 1097 syscallarg(uid_t) uid; 1098 } */ 1099 uid_t uid; 1100 1101 uid = SCARG(uap, uid); 1102 if (kauth_cred_getuid(l->l_cred) != uid) 1103 return sys_nosys(l, uap, retval); 1104 1105 *retval = uid; 1106 return 0; 1107 } 1108 1109 int 1110 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval) 1111 { 1112 /* { 1113 syscallarg(gid_t) gid; 1114 } */ 1115 gid_t gid; 1116 1117 gid = SCARG(uap, gid); 1118 if (kauth_cred_getgid(l->l_cred) != gid) 1119 return sys_nosys(l, uap, retval); 1120 1121 *retval = gid; 1122 return 0; 1123 } 1124 1125 int 1126 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval) 1127 { 1128 /* { 1129 syscallarg(uid_t) ruid; 1130 syscallarg(uid_t) euid; 1131 syscallarg(uid_t) suid; 1132 } */ 1133 1134 /* 1135 * Note: These checks are a little different than the NetBSD 1136 * setreuid(2) call performs. This precisely follows the 1137 * behavior of the Linux kernel. 1138 */ 1139 1140 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1141 SCARG(uap, suid), 1142 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1143 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1144 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1145 } 1146 1147 int 1148 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval) 1149 { 1150 /* { 1151 syscallarg(uid_t *) ruid; 1152 syscallarg(uid_t *) euid; 1153 syscallarg(uid_t *) suid; 1154 } */ 1155 kauth_cred_t pc = l->l_cred; 1156 int error; 1157 uid_t uid; 1158 1159 /* 1160 * Linux copies these values out to userspace like so: 1161 * 1162 * 1. Copy out ruid. 1163 * 2. If that succeeds, copy out euid. 1164 * 3. If both of those succeed, copy out suid. 1165 */ 1166 uid = kauth_cred_getuid(pc); 1167 if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0) 1168 return (error); 1169 1170 uid = kauth_cred_geteuid(pc); 1171 if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0) 1172 return (error); 1173 1174 uid = kauth_cred_getsvuid(pc); 1175 1176 return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t))); 1177 } 1178 1179 int 1180 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval) 1181 { 1182 /* { 1183 i386, m68k, powerpc: T=int 1184 alpha, amd64: T=long 1185 syscallarg(T) request; 1186 syscallarg(T) pid; 1187 syscallarg(T) addr; 1188 syscallarg(T) data; 1189 } */ 1190 const int *ptr; 1191 int request; 1192 int error; 1193 1194 ptr = linux_ptrace_request_map; 1195 request = SCARG(uap, request); 1196 while (*ptr != -1) 1197 if (*ptr++ == request) { 1198 struct sys_ptrace_args pta; 1199 1200 SCARG(&pta, req) = *ptr; 1201 SCARG(&pta, pid) = SCARG(uap, pid); 1202 SCARG(&pta, addr) = (void *)SCARG(uap, addr); 1203 SCARG(&pta, data) = SCARG(uap, data); 1204 1205 /* 1206 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1207 * to continue where the process left off previously. 1208 * The same thing is achieved by addr == (void *) 1 1209 * on NetBSD, so rewrite 'addr' appropriately. 1210 */ 1211 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1212 SCARG(&pta, addr) = (void *) 1; 1213 1214 error = sysent[SYS_ptrace].sy_call(l, &pta, retval); 1215 if (error) 1216 return error; 1217 switch (request) { 1218 case LINUX_PTRACE_PEEKTEXT: 1219 case LINUX_PTRACE_PEEKDATA: 1220 error = copyout (retval, 1221 (void *)SCARG(uap, data), 1222 sizeof *retval); 1223 *retval = SCARG(uap, data); 1224 break; 1225 default: 1226 break; 1227 } 1228 return error; 1229 } 1230 else 1231 ptr++; 1232 1233 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1234 } 1235 1236 int 1237 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval) 1238 { 1239 /* { 1240 syscallarg(int) magic1; 1241 syscallarg(int) magic2; 1242 syscallarg(int) cmd; 1243 syscallarg(void *) arg; 1244 } */ 1245 struct sys_reboot_args /* { 1246 syscallarg(int) opt; 1247 syscallarg(char *) bootstr; 1248 } */ sra; 1249 int error; 1250 1251 if ((error = kauth_authorize_system(l->l_cred, 1252 KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0) 1253 return(error); 1254 1255 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1256 return(EINVAL); 1257 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1258 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1259 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1260 return(EINVAL); 1261 1262 switch ((unsigned long)SCARG(uap, cmd)) { 1263 case LINUX_REBOOT_CMD_RESTART: 1264 SCARG(&sra, opt) = RB_AUTOBOOT; 1265 break; 1266 case LINUX_REBOOT_CMD_HALT: 1267 SCARG(&sra, opt) = RB_HALT; 1268 break; 1269 case LINUX_REBOOT_CMD_POWER_OFF: 1270 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1271 break; 1272 case LINUX_REBOOT_CMD_RESTART2: 1273 /* Reboot with an argument. */ 1274 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1275 SCARG(&sra, bootstr) = SCARG(uap, arg); 1276 break; 1277 case LINUX_REBOOT_CMD_CAD_ON: 1278 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1279 case LINUX_REBOOT_CMD_CAD_OFF: 1280 return(0); 1281 default: 1282 return(EINVAL); 1283 } 1284 1285 return(sys_reboot(l, &sra, retval)); 1286 } 1287 1288 /* 1289 * Copy of compat_12_sys_swapon(). 1290 */ 1291 int 1292 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval) 1293 { 1294 /* { 1295 syscallarg(const char *) name; 1296 } */ 1297 struct sys_swapctl_args ua; 1298 1299 SCARG(&ua, cmd) = SWAP_ON; 1300 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name)); 1301 SCARG(&ua, misc) = 0; /* priority */ 1302 return (sys_swapctl(l, &ua, retval)); 1303 } 1304 1305 /* 1306 * Stop swapping to the file or block device specified by path. 1307 */ 1308 int 1309 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval) 1310 { 1311 /* { 1312 syscallarg(const char *) path; 1313 } */ 1314 struct sys_swapctl_args ua; 1315 1316 SCARG(&ua, cmd) = SWAP_OFF; 1317 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/ 1318 return (sys_swapctl(l, &ua, retval)); 1319 } 1320 1321 /* 1322 * Copy of compat_09_sys_setdomainname() 1323 */ 1324 /* ARGSUSED */ 1325 int 1326 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval) 1327 { 1328 /* { 1329 syscallarg(char *) domainname; 1330 syscallarg(int) len; 1331 } */ 1332 int name[2]; 1333 1334 name[0] = CTL_KERN; 1335 name[1] = KERN_DOMAINNAME; 1336 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1337 SCARG(uap, len), l)); 1338 } 1339 1340 /* 1341 * sysinfo() 1342 */ 1343 /* ARGSUSED */ 1344 int 1345 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval) 1346 { 1347 /* { 1348 syscallarg(struct linux_sysinfo *) arg; 1349 } */ 1350 struct linux_sysinfo si; 1351 struct loadavg *la; 1352 int64_t filepg; 1353 1354 memset(&si, 0, sizeof(si)); 1355 si.uptime = time_uptime; 1356 la = &averunnable; 1357 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1358 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1359 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1360 si.totalram = ctob((u_long)physmem); 1361 /* uvm_availmem() may sync the counters. */ 1362 si.freeram = (u_long)uvm_availmem(true) * uvmexp.pagesize; 1363 filepg = cpu_count_get(CPU_COUNT_FILECLEAN) + 1364 cpu_count_get(CPU_COUNT_FILEDIRTY) + 1365 cpu_count_get(CPU_COUNT_FILEUNKNOWN) - 1366 cpu_count_get(CPU_COUNT_EXECPAGES); 1367 si.sharedram = 0; /* XXX */ 1368 si.bufferram = (u_long)(filepg * uvmexp.pagesize); 1369 si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize; 1370 si.freeswap = 1371 (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1372 si.procs = atomic_load_relaxed(&nprocs); 1373 1374 /* The following are only present in newer Linux kernels. */ 1375 si.totalbig = 0; 1376 si.freebig = 0; 1377 si.mem_unit = 1; 1378 1379 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1380 } 1381 1382 int 1383 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval) 1384 { 1385 /* { 1386 syscallarg(int) which; 1387 # ifdef LINUX_LARGEFILE64 1388 syscallarg(struct rlimit *) rlp; 1389 # else 1390 syscallarg(struct orlimit *) rlp; 1391 # endif 1392 } */ 1393 # ifdef LINUX_LARGEFILE64 1394 struct rlimit orl; 1395 # else 1396 struct orlimit orl; 1397 # endif 1398 int which; 1399 1400 which = linux_to_bsd_limit(SCARG(uap, which)); 1401 if (which < 0) 1402 return -which; 1403 1404 memset(&orl, 0, sizeof(orl)); 1405 bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]); 1406 1407 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1408 } 1409 1410 int 1411 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval) 1412 { 1413 /* { 1414 syscallarg(int) which; 1415 # ifdef LINUX_LARGEFILE64 1416 syscallarg(struct rlimit *) rlp; 1417 # else 1418 syscallarg(struct orlimit *) rlp; 1419 # endif 1420 } */ 1421 struct rlimit rl; 1422 # ifdef LINUX_LARGEFILE64 1423 struct rlimit orl; 1424 # else 1425 struct orlimit orl; 1426 # endif 1427 int error; 1428 int which; 1429 1430 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1431 return error; 1432 1433 which = linux_to_bsd_limit(SCARG(uap, which)); 1434 if (which < 0) 1435 return -which; 1436 1437 linux_to_bsd_rlimit(&rl, &orl); 1438 return dosetrlimit(l, l->l_proc, which, &rl); 1439 } 1440 1441 # if !defined(__mips__) && !defined(__amd64__) 1442 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1443 int 1444 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval) 1445 { 1446 return linux_sys_getrlimit(l, (const void *)uap, retval); 1447 } 1448 # endif 1449 1450 /* 1451 * This gets called for unsupported syscalls. The difference to sys_nosys() 1452 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1453 * This is the way Linux does it and glibc depends on this behaviour. 1454 */ 1455 int 1456 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval) 1457 { 1458 return (ENOSYS); 1459 } 1460 1461 int 1462 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval) 1463 { 1464 /* { 1465 syscallarg(int) which; 1466 syscallarg(int) who; 1467 } */ 1468 struct sys_getpriority_args bsa; 1469 int error; 1470 1471 SCARG(&bsa, which) = SCARG(uap, which); 1472 SCARG(&bsa, who) = SCARG(uap, who); 1473 1474 if ((error = sys_getpriority(l, &bsa, retval))) 1475 return error; 1476 1477 *retval = NZERO - *retval; 1478 1479 return 0; 1480 } 1481 1482 int 1483 linux_do_sys_utimensat(struct lwp *l, int fd, const char *path, struct timespec *tsp, int flags, register_t *retval) 1484 { 1485 int follow, error; 1486 1487 follow = (flags & LINUX_AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 1488 1489 if (path == NULL && fd != AT_FDCWD) { 1490 file_t *fp; 1491 1492 /* fd_getvnode() will use the descriptor for us */ 1493 if ((error = fd_getvnode(fd, &fp)) != 0) 1494 return error; 1495 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 1496 tsp, UIO_SYSSPACE); 1497 fd_putfile(fd); 1498 return error; 1499 } 1500 1501 return do_sys_utimensat(l, fd, NULL, path, follow, tsp, UIO_SYSSPACE); 1502 } 1503 1504 int 1505 linux_sys_utimensat(struct lwp *l, const struct linux_sys_utimensat_args *uap, 1506 register_t *retval) 1507 { 1508 /* { 1509 syscallarg(int) fd; 1510 syscallarg(const char *) path; 1511 syscallarg(const struct linux_timespec *) times; 1512 syscallarg(int) flag; 1513 } */ 1514 int error; 1515 struct linux_timespec lts[2]; 1516 struct timespec *tsp = NULL, ts[2]; 1517 1518 if (SCARG(uap, times)) { 1519 error = copyin(SCARG(uap, times), <s, sizeof(lts)); 1520 if (error != 0) 1521 return error; 1522 linux_to_native_timespec(&ts[0], <s[0]); 1523 linux_to_native_timespec(&ts[1], <s[1]); 1524 tsp = ts; 1525 } 1526 1527 return linux_do_sys_utimensat(l, SCARG(uap, fd), SCARG(uap, path), 1528 tsp, SCARG(uap, flag), retval); 1529 } 1530 1531 int 1532 linux_sys_futex(struct lwp *l, const struct linux_sys_futex_args *uap, 1533 register_t *retval) 1534 { 1535 /* { 1536 syscallarg(int *) uaddr; 1537 syscallarg(int) op; 1538 syscallarg(int) val; 1539 syscallarg(const struct linux_timespec *) timeout; 1540 syscallarg(int *) uaddr2; 1541 syscallarg(int) val3; 1542 } */ 1543 struct linux_timespec lts; 1544 struct timespec ts, *tsp = NULL; 1545 int val2 = 0; 1546 int error; 1547 1548 /* 1549 * Linux overlays the "timeout" field and the "val2" field. 1550 * "timeout" is only valid for FUTEX_WAIT and FUTEX_WAIT_BITSET 1551 * on Linux. 1552 */ 1553 const int op = (SCARG(uap, op) & FUTEX_CMD_MASK); 1554 if ((op == FUTEX_WAIT || op == FUTEX_WAIT_BITSET) && 1555 SCARG(uap, timeout) != NULL) { 1556 if ((error = copyin(SCARG(uap, timeout), 1557 <s, sizeof(lts))) != 0) { 1558 return error; 1559 } 1560 linux_to_native_timespec(&ts, <s); 1561 tsp = &ts; 1562 } else { 1563 val2 = (int)(uintptr_t)SCARG(uap, timeout); 1564 } 1565 1566 return linux_do_futex(SCARG(uap, uaddr), SCARG(uap, op), 1567 SCARG(uap, val), tsp, SCARG(uap, uaddr2), val2, 1568 SCARG(uap, val3), retval); 1569 } 1570 1571 int 1572 linux_do_futex(int *uaddr, int op, int val, struct timespec *timeout, 1573 int *uaddr2, int val2, int val3, register_t *retval) 1574 { 1575 /* 1576 * Always clear FUTEX_PRIVATE_FLAG for Linux processes. 1577 * NetBSD-native futexes exist in different namespace 1578 * depending on FUTEX_PRIVATE_FLAG. This appears not 1579 * to be the case in Linux, and some futex users will 1580 * mix private and non-private ops on the same futex 1581 * object. 1582 */ 1583 return do_futex(uaddr, op & ~FUTEX_PRIVATE_FLAG, 1584 val, timeout, uaddr2, val2, val3, retval); 1585 } 1586