1 /* $NetBSD: linux_misc.c,v 1.268 2026/09/20 13:43:51 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 1.268 2026/09/20 13:43:51 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/epoll.h> 68 #include <sys/eventfd.h> 69 #include <sys/file.h> 70 #include <sys/stat.h> 71 #include <sys/filedesc.h> 72 #include <sys/ioctl.h> 73 #include <sys/kernel.h> 74 #include <sys/malloc.h> 75 #include <sys/mbuf.h> 76 #include <sys/mman.h> 77 #include <sys/mount.h> 78 #include <sys/poll.h> 79 #include <sys/prot.h> 80 #include <sys/reboot.h> 81 #include <sys/resource.h> 82 #include <sys/resourcevar.h> 83 #include <sys/select.h> 84 #include <sys/signal.h> 85 #include <sys/signalvar.h> 86 #include <sys/socket.h> 87 #include <sys/time.h> 88 #include <sys/times.h> 89 #include <sys/vnode.h> 90 #include <sys/uio.h> 91 #include <sys/wait.h> 92 #include <sys/utsname.h> 93 #include <sys/unistd.h> 94 #include <sys/vfs_syscalls.h> 95 #include <sys/swap.h> /* for SWAP_ON */ 96 #include <sys/sysctl.h> /* for KERN_DOMAINNAME */ 97 #include <sys/kauth.h> 98 #include <sys/futex.h> 99 100 #include <sys/ptrace.h> 101 #include <machine/ptrace.h> 102 103 #include <sys/syscall.h> 104 #include <sys/syscallargs.h> 105 106 #include <compat/sys/resource.h> 107 108 #include <compat/linux/common/linux_machdep.h> 109 #include <compat/linux/common/linux_types.h> 110 #include <compat/linux/common/linux_signal.h> 111 #include <compat/linux/common/linux_ipc.h> 112 #include <compat/linux/common/linux_sem.h> 113 114 #include <compat/linux/common/linux_fcntl.h> 115 #include <compat/linux/common/linux_mmap.h> 116 #include <compat/linux/common/linux_dirent.h> 117 #include <compat/linux/common/linux_util.h> 118 #include <compat/linux/common/linux_misc.h> 119 #include <compat/linux/common/linux_statfs.h> 120 #include <compat/linux/common/linux_limit.h> 121 #include <compat/linux/common/linux_ptrace.h> 122 #include <compat/linux/common/linux_reboot.h> 123 #include <compat/linux/common/linux_emuldata.h> 124 #include <compat/linux/common/linux_sched.h> 125 126 #include <compat/linux/linux_syscallargs.h> 127 128 const int linux_ptrace_request_map[] = { 129 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 130 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 131 LINUX_PTRACE_PEEKDATA, PT_READ_D, 132 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 133 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 134 LINUX_PTRACE_CONT, PT_CONTINUE, 135 LINUX_PTRACE_KILL, PT_KILL, 136 LINUX_PTRACE_ATTACH, PT_ATTACH, 137 LINUX_PTRACE_DETACH, PT_DETACH, 138 # ifdef PT_STEP 139 LINUX_PTRACE_SINGLESTEP, PT_STEP, 140 # endif 141 LINUX_PTRACE_SYSCALL, PT_SYSCALL, 142 -1 143 }; 144 145 const struct linux_mnttypes linux_fstypes[] = { 146 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 148 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 149 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 150 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 151 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 152 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 153 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 154 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 155 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 156 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 157 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 158 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 159 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 160 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 161 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 162 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 163 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 164 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 165 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 166 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC }, 167 { MOUNT_PTYFS, LINUX_DEVPTS_SUPER_MAGIC }, 168 { MOUNT_TMPFS, LINUX_TMPFS_SUPER_MAGIC } 169 }; 170 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 171 172 #ifdef DEBUG_LINUX 173 #define DPRINTF(a, ...) uprintf(a, __VA_ARGS__) 174 #else 175 #define DPRINTF(a, ...) 176 #endif 177 178 /* Local linux_misc.c functions: */ 179 static void linux_to_bsd_mmap_args(struct sys_mmap_args *, 180 const struct linux_sys_mmap_args *); 181 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *, 182 register_t *, off_t); 183 static int linux_to_native_wait_options(int); 184 185 /* 186 * The information on a terminated (or stopped) process needs 187 * to be converted in order for Linux binaries to get a valid signal 188 * number out of it. 189 */ 190 int 191 bsd_to_linux_wstat(int st) 192 { 193 194 int sig; 195 196 if (WIFSIGNALED(st)) { 197 sig = WTERMSIG(st); 198 if (sig >= 0 && sig < NSIG) 199 st= (st & ~0177) | native_to_linux_signo[sig]; 200 } else if (WIFSTOPPED(st)) { 201 sig = WSTOPSIG(st); 202 if (sig >= 0 && sig < NSIG) 203 st = (st & ~0xff00) | 204 (native_to_linux_signo[sig] << 8); 205 } 206 return st; 207 } 208 209 /* 210 * wait4(2). Passed on to the NetBSD call, surrounded by code to 211 * reserve some space for a NetBSD-style wait status, and converting 212 * it to what Linux wants. 213 */ 214 int 215 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval) 216 { 217 /* { 218 syscallarg(int) pid; 219 syscallarg(int *) status; 220 syscallarg(int) options; 221 syscallarg(struct rusage50 *) rusage; 222 } */ 223 int error, status, options, linux_options, pid = SCARG(uap, pid); 224 struct rusage50 ru50; 225 struct rusage ru; 226 proc_t *p; 227 228 linux_options = SCARG(uap, options); 229 if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS)) 230 return (EINVAL); 231 232 options = linux_to_native_wait_options(linux_options); 233 # ifdef DIAGNOSTIC 234 if (linux_options & LINUX_WNOTHREAD) 235 printf("WARNING: %s: linux process %d.%d called " 236 "waitpid with __WNOTHREAD set!\n", 237 __FILE__, l->l_proc->p_pid, l->l_lid); 238 239 # endif 240 241 error = do_sys_wait(&pid, &status, options, 242 SCARG(uap, rusage) != NULL ? &ru : NULL); 243 244 retval[0] = pid; 245 if (pid == 0) 246 return error; 247 248 p = curproc; 249 mutex_enter(p->p_lock); 250 sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */ 251 mutex_exit(p->p_lock); 252 253 if (SCARG(uap, rusage) != NULL) { 254 rusage_to_rusage50(&ru, &ru50); 255 error = copyout(&ru, SCARG(uap, rusage), sizeof(ru)); 256 } 257 258 if (error == 0 && SCARG(uap, status) != NULL) { 259 status = bsd_to_linux_wstat(status); 260 error = copyout(&status, SCARG(uap, status), sizeof status); 261 } 262 263 return error; 264 } 265 266 /* 267 * waitid(2). Converting arguments to the NetBSD equivalent and 268 * calling it. 269 */ 270 int 271 linux_sys_waitid(struct lwp *l, const struct linux_sys_waitid_args *uap, register_t *retval) 272 { 273 /* { 274 syscallarg(int) idtype; 275 syscallarg(id_t) id; 276 syscallarg(linux_siginfo_t *) infop; 277 syscallarg(int) options; 278 syscallarg(struct rusage50 *) rusage; 279 } */ 280 int error, linux_options, options, linux_idtype, status; 281 pid_t pid; 282 idtype_t idtype; 283 id_t id; 284 siginfo_t info; 285 linux_siginfo_t linux_info; 286 struct wrusage wru; 287 struct rusage50 ru50; 288 289 linux_idtype = SCARG(uap, idtype); 290 switch (linux_idtype) { 291 case LINUX_P_ALL: 292 idtype = P_ALL; 293 break; 294 case LINUX_P_PID: 295 idtype = P_PID; 296 break; 297 case LINUX_P_PGID: 298 idtype = P_PGID; 299 break; 300 case LINUX_P_PIDFD: 301 return EOPNOTSUPP; 302 default: 303 return EINVAL; 304 } 305 306 linux_options = SCARG(uap, options); 307 if (linux_options & ~(LINUX_WAITID_KNOWNFLAGS)) 308 return EINVAL; 309 310 options = linux_to_native_wait_options(linux_options); 311 id = SCARG(uap, id); 312 313 error = do_sys_waitid(idtype, id, &pid, &status, options, &wru, &info); 314 if (pid == 0 && options & WNOHANG) { 315 info.si_signo = 0; 316 info.si_pid = 0; 317 } 318 319 if (error == 0 && SCARG(uap, infop) != NULL) { 320 /* POSIX says that this NULL check is a bug, but Linux does this. */ 321 native_to_linux_siginfo(&linux_info, &info._info); 322 error = copyout(&linux_info, SCARG(uap, infop), sizeof(linux_info)); 323 } 324 325 if (error == 0 && SCARG(uap, rusage) != NULL) { 326 rusage_to_rusage50(&wru.wru_children, &ru50); 327 error = copyout(&ru50, SCARG(uap, rusage), sizeof(ru50)); 328 } 329 330 return error; 331 } 332 333 /* 334 * Convert the options argument for wait4(2) and waitid(2) from what 335 * Linux wants to what NetBSD wants. 336 */ 337 static int 338 linux_to_native_wait_options(int linux_options) 339 { 340 int options = 0; 341 342 if (linux_options & LINUX_WNOHANG) 343 options |= WNOHANG; 344 if (linux_options & LINUX_WUNTRACED) 345 options |= WUNTRACED; 346 if (linux_options & LINUX_WEXITED) 347 options |= WEXITED; 348 if (linux_options & LINUX_WCONTINUED) 349 options |= WCONTINUED; 350 if (linux_options & LINUX_WNOWAIT) 351 options |= WNOWAIT; 352 if (linux_options & LINUX_WALL) 353 options |= WALLSIG; 354 if (linux_options & LINUX_WCLONE) 355 options |= WALTSIG; 356 357 return options; 358 } 359 360 /* 361 * Linux brk(2). Like native, but always return the new break value. 362 */ 363 int 364 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval) 365 { 366 /* { 367 syscallarg(char *) nsize; 368 } */ 369 struct proc *p = l->l_proc; 370 struct vmspace *vm = p->p_vmspace; 371 struct sys_obreak_args oba; 372 373 memset(&oba, 0, sizeof(oba)); 374 SCARG(&oba, nsize) = SCARG(uap, nsize); 375 376 (void) sys_obreak(l, &oba, retval); 377 retval[0] = (register_t)((char *)vm->vm_daddr + ptoa(vm->vm_dsize)); 378 return 0; 379 } 380 381 /* 382 * Implement the fs stat functions. Straightforward. 383 */ 384 int 385 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval) 386 { 387 /* { 388 syscallarg(const char *) path; 389 syscallarg(struct linux_statfs *) sp; 390 } */ 391 struct statvfs *sb; 392 struct linux_statfs ltmp; 393 int error; 394 395 sb = STATVFSBUF_GET(); 396 error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb); 397 if (error == 0) { 398 bsd_to_linux_statfs(sb, <mp); 399 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp); 400 } 401 STATVFSBUF_PUT(sb); 402 403 return error; 404 } 405 406 int 407 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval) 408 { 409 /* { 410 syscallarg(int) fd; 411 syscallarg(struct linux_statfs *) sp; 412 } */ 413 struct statvfs *sb; 414 struct linux_statfs ltmp; 415 int error; 416 417 sb = STATVFSBUF_GET(); 418 error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb); 419 if (error == 0) { 420 bsd_to_linux_statfs(sb, <mp); 421 error = copyout(<mp, SCARG(uap, sp), sizeof ltmp); 422 } 423 STATVFSBUF_PUT(sb); 424 425 return error; 426 } 427 428 /* 429 * uname(). Just copy the info from the various strings stored in the 430 * kernel, and put it in the Linux utsname structure. That structure 431 * is almost the same as the NetBSD one, only it has fields 65 characters 432 * long, and an extra domainname field. 433 */ 434 int 435 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval) 436 { 437 /* { 438 syscallarg(struct linux_utsname *) up; 439 } */ 440 struct linux_utsname luts; 441 442 memset(&luts, 0, sizeof(luts)); 443 strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 444 strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 445 strlcpy(luts.l_release, linux_release, sizeof(luts.l_release)); 446 strlcpy(luts.l_version, linux_version, sizeof(luts.l_version)); 447 strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine)); 448 strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 449 450 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 451 } 452 453 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 454 /* Used indirectly on: arm, i386, m68k */ 455 456 /* 457 * New type Linux mmap call. 458 * Only called directly on machines with >= 6 free regs. 459 */ 460 int 461 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval) 462 { 463 /* { 464 syscallarg(unsigned long) addr; 465 syscallarg(size_t) len; 466 syscallarg(int) prot; 467 syscallarg(int) flags; 468 syscallarg(int) fd; 469 syscallarg(linux_off_t) offset; 470 } */ 471 472 if (SCARG(uap, offset) & PAGE_MASK) 473 return EINVAL; 474 475 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 476 } 477 478 /* 479 * Guts of most architectures' mmap64() implementations. This shares 480 * its list of arguments with linux_sys_mmap(). 481 * 482 * The difference in linux_sys_mmap2() is that "offset" is actually 483 * (offset / pagesize), not an absolute byte count. This translation 484 * to pagesize offsets is done inside glibc between the mmap64() call 485 * point, and the actual syscall. 486 */ 487 int 488 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval) 489 { 490 /* { 491 syscallarg(unsigned long) addr; 492 syscallarg(size_t) len; 493 syscallarg(int) prot; 494 syscallarg(int) flags; 495 syscallarg(int) fd; 496 syscallarg(linux_off_t) offset; 497 } */ 498 499 return linux_mmap(l, uap, retval, 500 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 501 } 502 503 /* 504 * Massage arguments and call system mmap(2). 505 */ 506 static int 507 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset) 508 { 509 struct sys_mmap_args cma; 510 int error; 511 size_t mmoff=0; 512 513 memset(&cma, 0, sizeof(cma)); 514 linux_to_bsd_mmap_args(&cma, uap); 515 SCARG(&cma, pos) = offset; 516 517 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 518 /* 519 * Request for stack-like memory segment. On linux, this 520 * works by mmap()ping (small) segment, which is automatically 521 * extended when page fault happens below the currently 522 * allocated area. We emulate this by allocating (typically 523 * bigger) segment sized at current stack size limit, and 524 * offsetting the requested and returned address accordingly. 525 * Since physical pages are only allocated on-demand, this 526 * is effectively identical. 527 */ 528 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 529 530 if (SCARG(&cma, len) < ssl) { 531 /* Compute the address offset */ 532 mmoff = round_page(ssl) - SCARG(uap, len); 533 534 if (SCARG(&cma, addr)) 535 SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff; 536 537 SCARG(&cma, len) = (size_t) ssl; 538 } 539 } 540 541 error = sys_mmap(l, &cma, retval); 542 if (error) 543 return (error); 544 545 /* Shift the returned address for stack-like segment if necessary */ 546 retval[0] += mmoff; 547 548 return (0); 549 } 550 551 static void 552 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap) 553 { 554 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 555 556 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 557 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 558 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 559 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 560 flags |= cvtto_bsd_mask(fl, LINUX_MAP_LOCKED, MAP_WIRED); 561 /* XXX XAX ERH: Any other flags here? There are more defined... */ 562 563 SCARG(cma, addr) = (void *)SCARG(uap, addr); 564 SCARG(cma, len) = SCARG(uap, len); 565 SCARG(cma, prot) = SCARG(uap, prot); 566 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 567 SCARG(cma, prot) |= VM_PROT_READ; 568 SCARG(cma, flags) = flags; 569 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 570 SCARG(cma, PAD) = 0; 571 } 572 573 #define LINUX_MREMAP_MAYMOVE 1 574 #define LINUX_MREMAP_FIXED 2 575 576 int 577 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval) 578 { 579 /* { 580 syscallarg(void *) old_address; 581 syscallarg(size_t) old_size; 582 syscallarg(size_t) new_size; 583 syscallarg(u_long) flags; 584 } */ 585 586 struct proc *p; 587 struct vm_map *map; 588 vaddr_t oldva; 589 vaddr_t newva; 590 size_t oldsize; 591 size_t newsize; 592 int flags; 593 int uvmflags; 594 int error; 595 596 flags = SCARG(uap, flags); 597 oldva = (vaddr_t)SCARG(uap, old_address); 598 oldsize = round_page(SCARG(uap, old_size)); 599 newsize = round_page(SCARG(uap, new_size)); 600 if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) { 601 error = EINVAL; 602 goto done; 603 } 604 if ((flags & LINUX_MREMAP_FIXED) != 0) { 605 if ((flags & LINUX_MREMAP_MAYMOVE) == 0) { 606 error = EINVAL; 607 goto done; 608 } 609 #if 0 /* notyet */ 610 newva = SCARG(uap, new_address); 611 uvmflags = MAP_FIXED; 612 #else /* notyet */ 613 error = EOPNOTSUPP; 614 goto done; 615 #endif /* notyet */ 616 } else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) { 617 uvmflags = 0; 618 } else { 619 newva = oldva; 620 uvmflags = MAP_FIXED; 621 } 622 p = l->l_proc; 623 map = &p->p_vmspace->vm_map; 624 error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p, 625 uvmflags); 626 627 done: 628 *retval = (error != 0) ? 0 : (register_t)newva; 629 return error; 630 } 631 632 #ifdef USRSTACK 633 int 634 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval) 635 { 636 /* { 637 syscallarg(const void *) start; 638 syscallarg(unsigned long) len; 639 syscallarg(int) prot; 640 } */ 641 struct vm_map_entry *entry; 642 struct vm_map *map; 643 struct proc *p; 644 vaddr_t end, start, len, stacklim; 645 int prot, grows; 646 647 start = (vaddr_t)SCARG(uap, start); 648 len = round_page(SCARG(uap, len)); 649 prot = SCARG(uap, prot); 650 grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP); 651 prot &= ~grows; 652 end = start + len; 653 654 if (start & PAGE_MASK) 655 return EINVAL; 656 if (end < start) 657 return EINVAL; 658 if (end == start) 659 return 0; 660 661 if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 662 return EINVAL; 663 if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP)) 664 return EINVAL; 665 666 p = l->l_proc; 667 map = &p->p_vmspace->vm_map; 668 vm_map_lock(map); 669 # ifdef notdef 670 VM_MAP_RANGE_CHECK(map, start, end); 671 # endif 672 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 673 vm_map_unlock(map); 674 return ENOMEM; 675 } 676 677 /* 678 * Approximate the behaviour of PROT_GROWS{DOWN,UP}. 679 */ 680 681 stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur; 682 if (grows & LINUX_PROT_GROWSDOWN) { 683 if (USRSTACK - stacklim <= start && start < USRSTACK) { 684 start = USRSTACK - stacklim; 685 } else { 686 start = entry->start; 687 } 688 } else if (grows & LINUX_PROT_GROWSUP) { 689 if (USRSTACK <= end && end < USRSTACK + stacklim) { 690 end = USRSTACK + stacklim; 691 } else { 692 end = entry->end; 693 } 694 } 695 vm_map_unlock(map); 696 return uvm_map_protect_user(l, start, end, prot); 697 } 698 #endif /* USRSTACK */ 699 700 /* 701 * This code is partly stolen from src/lib/libc/compat-43/times.c 702 */ 703 704 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 705 706 int 707 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval) 708 { 709 /* { 710 syscallarg(struct times *) tms; 711 } */ 712 struct proc *p = l->l_proc; 713 struct timeval t; 714 int error; 715 716 if (SCARG(uap, tms)) { 717 struct linux_tms ltms; 718 struct rusage ru; 719 720 memset(<ms, 0, sizeof(ltms)); 721 722 mutex_enter(p->p_lock); 723 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL); 724 ltms.ltms_utime = CONVTCK(ru.ru_utime); 725 ltms.ltms_stime = CONVTCK(ru.ru_stime); 726 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 727 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 728 mutex_exit(p->p_lock); 729 730 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 731 return error; 732 } 733 734 getmicrouptime(&t); 735 736 retval[0] = ((linux_clock_t)(CONVTCK(t))); 737 return 0; 738 } 739 740 #undef CONVTCK 741 742 #if !defined(__aarch64__) 743 /* 744 * Linux 'readdir' call. This code is mostly taken from the 745 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 746 * an attempt has been made to keep it a little cleaner (failing 747 * miserably, because of the cruft needed if count 1 is passed). 748 * 749 * The d_off field should contain the offset of the next valid entry, 750 * but in Linux it has the offset of the entry itself. We emulate 751 * that bug here. 752 * 753 * Read in BSD-style entries, convert them, and copy them out. 754 * 755 * Note that this doesn't handle union-mounted filesystems. 756 */ 757 int 758 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval) 759 { 760 /* { 761 syscallarg(int) fd; 762 syscallarg(struct linux_dirent *) dent; 763 syscallarg(unsigned int) count; 764 } */ 765 struct dirent *bdp; 766 struct vnode *vp; 767 char *inp, *tbuf; /* BSD-format */ 768 int len, reclen; /* BSD-format */ 769 char *outp; /* Linux-format */ 770 int resid, linux_reclen = 0; /* Linux-format */ 771 struct file *fp; 772 struct uio auio; 773 struct iovec aiov; 774 struct linux_dirent idb; 775 off_t off; /* true file offset */ 776 int buflen, error, eofflag, nbytes, oldcall; 777 struct vattr va; 778 off_t *cookiebuf = NULL, *cookie; 779 int ncookies; 780 781 /* fd_getvnode() will use the descriptor for us */ 782 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 783 return (error); 784 785 if ((fp->f_flag & FREAD) == 0) { 786 error = EBADF; 787 goto out1; 788 } 789 790 vp = (struct vnode *)fp->f_data; 791 if (vp->v_type != VDIR) { 792 error = ENOTDIR; 793 goto out1; 794 } 795 796 vn_lock(vp, LK_SHARED | LK_RETRY); 797 error = VOP_GETATTR(vp, &va, l->l_cred); 798 VOP_UNLOCK(vp); 799 if (error) 800 goto out1; 801 802 nbytes = SCARG(uap, count); 803 if (nbytes == 1) { /* emulating old, broken behaviour */ 804 nbytes = sizeof (idb); 805 buflen = uimax(va.va_blocksize, nbytes); 806 oldcall = 1; 807 } else { 808 buflen = uimin(MAXBSIZE, nbytes); 809 if (buflen < va.va_blocksize) 810 buflen = va.va_blocksize; 811 oldcall = 0; 812 } 813 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 814 815 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 816 off = fp->f_offset; 817 again: 818 aiov.iov_base = tbuf; 819 aiov.iov_len = buflen; 820 auio.uio_iov = &aiov; 821 auio.uio_iovcnt = 1; 822 auio.uio_rw = UIO_READ; 823 auio.uio_resid = buflen; 824 auio.uio_offset = off; 825 UIO_SETUP_SYSSPACE(&auio); 826 /* 827 * First we read into the malloc'ed buffer, then 828 * we massage it into user space, one record at a time. 829 */ 830 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 831 &ncookies); 832 if (error) 833 goto out; 834 835 inp = tbuf; 836 outp = (void *)SCARG(uap, dent); 837 resid = nbytes; 838 if ((len = buflen - auio.uio_resid) == 0) 839 goto eof; 840 841 for (cookie = cookiebuf; len > 0; len -= reclen) { 842 bdp = (struct dirent *)inp; 843 reclen = bdp->d_reclen; 844 if (reclen & 3) { 845 error = EIO; 846 goto out; 847 } 848 if (bdp->d_fileno == 0) { 849 inp += reclen; /* it is a hole; squish it out */ 850 if (cookie) 851 off = *cookie++; 852 else 853 off += reclen; 854 continue; 855 } 856 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 857 if (reclen > len || resid < linux_reclen) { 858 /* entry too big for buffer, so just stop */ 859 outp++; 860 break; 861 } 862 /* 863 * Massage in place to make a Linux-shaped dirent (otherwise 864 * we have to worry about touching user memory outside of 865 * the copyout() call). 866 */ 867 memset(&idb, 0, sizeof(idb)); 868 idb.d_ino = bdp->d_fileno; 869 /* 870 * The old readdir() call misuses the offset and reclen fields. 871 */ 872 if (oldcall) { 873 idb.d_off = (linux_off_t)linux_reclen; 874 idb.d_reclen = (u_short)bdp->d_namlen; 875 } else { 876 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 877 compat_offseterr(vp, "linux_getdents"); 878 error = EINVAL; 879 goto out; 880 } 881 idb.d_off = (linux_off_t)off; 882 idb.d_reclen = (u_short)linux_reclen; 883 /* Linux puts d_type at the end of each record */ 884 *((char *)&idb + idb.d_reclen - 1) = bdp->d_type; 885 } 886 memcpy(idb.d_name, bdp->d_name, 887 MIN(sizeof(idb.d_name), bdp->d_namlen + 1)); 888 if ((error = copyout((void *)&idb, outp, linux_reclen))) 889 goto out; 890 /* advance past this real entry */ 891 inp += reclen; 892 if (cookie) 893 off = *cookie++; /* each entry points to itself */ 894 else 895 off += reclen; 896 /* advance output past Linux-shaped entry */ 897 outp += linux_reclen; 898 resid -= linux_reclen; 899 if (oldcall) 900 break; 901 } 902 903 /* if we squished out the whole block, try again */ 904 if (outp == (void *)SCARG(uap, dent)) { 905 if (cookiebuf) 906 free(cookiebuf, M_TEMP); 907 cookiebuf = NULL; 908 goto again; 909 } 910 fp->f_offset = off; /* update the vnode offset */ 911 912 if (oldcall) 913 nbytes = resid + linux_reclen; 914 915 eof: 916 *retval = nbytes - resid; 917 out: 918 VOP_UNLOCK(vp); 919 if (cookiebuf) 920 free(cookiebuf, M_TEMP); 921 free(tbuf, M_TEMP); 922 out1: 923 fd_putfile(SCARG(uap, fd)); 924 return error; 925 } 926 #endif 927 928 #if !defined(__aarch64__) 929 /* 930 * Even when just using registers to pass arguments to syscalls you can 931 * have 5 of them on the i386. So this newer version of select() does 932 * this. 933 */ 934 int 935 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval) 936 { 937 /* { 938 syscallarg(int) nfds; 939 syscallarg(fd_set *) readfds; 940 syscallarg(fd_set *) writefds; 941 syscallarg(fd_set *) exceptfds; 942 syscallarg(struct timeval50 *) timeout; 943 } */ 944 945 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 946 SCARG(uap, writefds), SCARG(uap, exceptfds), 947 (struct linux_timeval *)SCARG(uap, timeout)); 948 } 949 950 /* 951 * Common code for the old and new versions of select(). A couple of 952 * things are important: 953 * 1) return the amount of time left in the 'timeout' parameter 954 * 2) select never returns ERESTART on Linux, always return EINTR 955 */ 956 int 957 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds, 958 fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout) 959 { 960 struct timespec ts0, ts1, uts, *ts = NULL; 961 struct linux_timeval ltv; 962 int error; 963 964 /* 965 * Store current time for computation of the amount of 966 * time left. 967 */ 968 if (timeout) { 969 if ((error = copyin(timeout, <v, sizeof(ltv)))) 970 return error; 971 uts.tv_sec = ltv.tv_sec; 972 uts.tv_nsec = (long)((unsigned long)ltv.tv_usec * 1000); 973 if (itimespecfix(&uts)) { 974 /* 975 * The timeval was invalid. Convert it to something 976 * valid that will act as it does under Linux. 977 */ 978 uts.tv_sec += uts.tv_nsec / 1000000000; 979 uts.tv_nsec %= 1000000000; 980 if (uts.tv_nsec < 0) { 981 uts.tv_sec -= 1; 982 uts.tv_nsec += 1000000000; 983 } 984 if (uts.tv_sec < 0) 985 timespecclear(&uts); 986 } 987 ts = &uts; 988 nanotime(&ts0); 989 } 990 991 error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL); 992 993 if (error) { 994 /* 995 * See fs/select.c in the Linux kernel. Without this, 996 * Maelstrom doesn't work. 997 */ 998 if (error == ERESTART) 999 error = EINTR; 1000 return error; 1001 } 1002 1003 if (timeout) { 1004 if (*retval) { 1005 /* 1006 * Compute how much time was left of the timeout, 1007 * by subtracting the current time and the time 1008 * before we started the call, and subtracting 1009 * that result from the user-supplied value. 1010 */ 1011 nanotime(&ts1); 1012 timespecsub(&ts1, &ts0, &ts1); 1013 timespecsub(&uts, &ts1, &uts); 1014 if (uts.tv_sec < 0) 1015 timespecclear(&uts); 1016 } else 1017 timespecclear(&uts); 1018 ltv.tv_sec = uts.tv_sec; 1019 ltv.tv_usec = uts.tv_nsec / 1000; 1020 if ((error = copyout(<v, timeout, sizeof(ltv)))) 1021 return error; 1022 } 1023 1024 return 0; 1025 } 1026 #endif 1027 1028 /* 1029 * Derived from FreeBSD's sys/compat/linux/linux_misc.c:linux_pselect6() 1030 * which was contributed by Dmitry Chagin 1031 * https://svnweb.freebsd.org/base?view=revision&revision=283403 1032 */ 1033 int 1034 linux_sys_pselect6(struct lwp *l, 1035 const struct linux_sys_pselect6_args *uap, register_t *retval) 1036 { 1037 /* { 1038 syscallarg(int) nfds; 1039 syscallarg(fd_set *) readfds; 1040 syscallarg(fd_set *) writefds; 1041 syscallarg(fd_set *) exceptfds; 1042 syscallarg(struct timespec *) timeout; 1043 syscallarg(linux_sized_sigset_t *) ss; 1044 } */ 1045 struct timespec uts, ts0, ts1, *tsp; 1046 linux_sized_sigset_t lsss; 1047 struct linux_timespec lts; 1048 linux_sigset_t lss; 1049 sigset_t *ssp; 1050 sigset_t ss; 1051 int error; 1052 1053 ssp = NULL; 1054 if (SCARG(uap, ss) != NULL) { 1055 if ((error = copyin(SCARG(uap, ss), &lsss, sizeof(lsss))) != 0) 1056 return (error); 1057 if (lsss.ss_len != sizeof(lss)) 1058 return (EINVAL); 1059 if (lsss.ss != NULL) { 1060 if ((error = copyin(lsss.ss, &lss, sizeof(lss))) != 0) 1061 return (error); 1062 linux_to_native_sigset(&ss, &lss); 1063 ssp = &ss; 1064 } 1065 } 1066 1067 if (SCARG(uap, timeout) != NULL) { 1068 error = copyin(SCARG(uap, timeout), <s, sizeof(lts)); 1069 if (error != 0) 1070 return (error); 1071 linux_to_native_timespec(&uts, <s); 1072 1073 if (itimespecfix(&uts)) 1074 return (EINVAL); 1075 1076 nanotime(&ts0); 1077 tsp = &uts; 1078 } else { 1079 tsp = NULL; 1080 } 1081 1082 error = selcommon(retval, SCARG(uap, nfds), SCARG(uap, readfds), 1083 SCARG(uap, writefds), SCARG(uap, exceptfds), tsp, ssp); 1084 1085 if (error == 0 && tsp != NULL) { 1086 if (retval != 0) { 1087 /* 1088 * Compute how much time was left of the timeout, 1089 * by subtracting the current time and the time 1090 * before we started the call, and subtracting 1091 * that result from the user-supplied value. 1092 */ 1093 nanotime(&ts1); 1094 timespecsub(&ts1, &ts0, &ts1); 1095 timespecsub(&uts, &ts1, &uts); 1096 if (uts.tv_sec < 0) 1097 timespecclear(&uts); 1098 } else { 1099 timespecclear(&uts); 1100 } 1101 1102 native_to_linux_timespec(<s, &uts); 1103 error = copyout(<s, SCARG(uap, timeout), sizeof(lts)); 1104 } 1105 1106 return (error); 1107 } 1108 1109 int 1110 linux_sys_ppoll(struct lwp *l, 1111 const struct linux_sys_ppoll_args *uap, register_t *retval) 1112 { 1113 /* { 1114 syscallarg(struct pollfd *) fds; 1115 syscallarg(u_int) nfds; 1116 syscallarg(struct linux_timespec *) timeout; 1117 syscallarg(linux_sigset_t *) sigset; 1118 } */ 1119 struct linux_timespec lts0, *lts; 1120 struct timespec ts0, *ts = NULL; 1121 linux_sigset_t lsigmask0, *lsigmask; 1122 sigset_t sigmask0, *sigmask = NULL; 1123 int error; 1124 1125 lts = SCARG(uap, timeout); 1126 if (lts) { 1127 if ((error = copyin(lts, <s0, sizeof(lts0))) != 0) 1128 return error; 1129 linux_to_native_timespec(&ts0, <s0); 1130 ts = &ts0; 1131 } 1132 1133 lsigmask = SCARG(uap, sigset); 1134 if (lsigmask) { 1135 if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0)))) 1136 return error; 1137 linux_to_native_sigset(&sigmask0, &lsigmask0); 1138 sigmask = &sigmask0; 1139 } 1140 1141 return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds), 1142 ts, sigmask); 1143 } 1144 1145 /* 1146 * Set the 'personality' (emulation mode) for the current process. Only 1147 * accept the Linux personality here (0). This call is needed because 1148 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1149 * ELF binaries run in Linux mode, not SVR4 mode. 1150 */ 1151 int 1152 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval) 1153 { 1154 /* { 1155 syscallarg(unsigned long) per; 1156 } */ 1157 struct linux_emuldata *led; 1158 int per; 1159 1160 per = SCARG(uap, per); 1161 led = l->l_emuldata; 1162 if (per == LINUX_PER_QUERY) { 1163 retval[0] = led->led_personality; 1164 return 0; 1165 } 1166 1167 switch (per & LINUX_PER_MASK) { 1168 case LINUX_PER_LINUX: 1169 case LINUX_PER_LINUX32: 1170 led->led_personality = per; 1171 break; 1172 1173 default: 1174 return EINVAL; 1175 } 1176 1177 retval[0] = per; 1178 return 0; 1179 } 1180 1181 /* 1182 * We have nonexistent fsuid equal to uid. 1183 * If modification is requested, refuse. 1184 */ 1185 int 1186 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval) 1187 { 1188 /* { 1189 syscallarg(uid_t) uid; 1190 } */ 1191 uid_t uid; 1192 1193 uid = SCARG(uap, uid); 1194 if (kauth_cred_getuid(l->l_cred) != uid) 1195 return sys_nosys(l, uap, retval); 1196 1197 *retval = uid; 1198 return 0; 1199 } 1200 1201 int 1202 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval) 1203 { 1204 /* { 1205 syscallarg(gid_t) gid; 1206 } */ 1207 gid_t gid; 1208 1209 gid = SCARG(uap, gid); 1210 if (kauth_cred_getgid(l->l_cred) != gid) 1211 return sys_nosys(l, uap, retval); 1212 1213 *retval = gid; 1214 return 0; 1215 } 1216 1217 int 1218 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval) 1219 { 1220 /* { 1221 syscallarg(uid_t) ruid; 1222 syscallarg(uid_t) euid; 1223 syscallarg(uid_t) suid; 1224 } */ 1225 1226 /* 1227 * Note: These checks are a little different than the NetBSD 1228 * setreuid(2) call performs. This precisely follows the 1229 * behavior of the Linux kernel. 1230 */ 1231 1232 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1233 SCARG(uap, suid), 1234 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1235 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1236 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1237 } 1238 1239 int 1240 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval) 1241 { 1242 /* { 1243 syscallarg(uid_t *) ruid; 1244 syscallarg(uid_t *) euid; 1245 syscallarg(uid_t *) suid; 1246 } */ 1247 kauth_cred_t pc = l->l_cred; 1248 int error; 1249 uid_t uid; 1250 1251 /* 1252 * Linux copies these values out to userspace like so: 1253 * 1254 * 1. Copy out ruid. 1255 * 2. If that succeeds, copy out euid. 1256 * 3. If both of those succeed, copy out suid. 1257 */ 1258 uid = kauth_cred_getuid(pc); 1259 if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0) 1260 return (error); 1261 1262 uid = kauth_cred_geteuid(pc); 1263 if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0) 1264 return (error); 1265 1266 uid = kauth_cred_getsvuid(pc); 1267 1268 return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t))); 1269 } 1270 1271 int 1272 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval) 1273 { 1274 /* { 1275 i386, m68k, powerpc: T=int 1276 alpha, amd64: T=long 1277 syscallarg(T) request; 1278 syscallarg(T) pid; 1279 syscallarg(T) addr; 1280 syscallarg(T) data; 1281 } */ 1282 const int *ptr; 1283 int request; 1284 int error; 1285 1286 ptr = linux_ptrace_request_map; 1287 request = SCARG(uap, request); 1288 while (*ptr != -1) 1289 if (*ptr++ == request) { 1290 struct sys_ptrace_args pta; 1291 1292 memset(&pta, 0, sizeof(pta)); 1293 SCARG(&pta, req) = *ptr; 1294 SCARG(&pta, pid) = SCARG(uap, pid); 1295 SCARG(&pta, addr) = (void *)SCARG(uap, addr); 1296 SCARG(&pta, data) = SCARG(uap, data); 1297 1298 /* 1299 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1300 * to continue where the process left off previously. 1301 * The same thing is achieved by addr == (void *) 1 1302 * on NetBSD, so rewrite 'addr' appropriately. 1303 */ 1304 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1305 SCARG(&pta, addr) = (void *) 1; 1306 1307 error = sysent[SYS_ptrace].sy_call(l, &pta, retval); 1308 if (error) 1309 return error; 1310 switch (request) { 1311 case LINUX_PTRACE_PEEKTEXT: 1312 case LINUX_PTRACE_PEEKDATA: 1313 error = copyout (retval, 1314 (void *)SCARG(uap, data), 1315 sizeof *retval); 1316 *retval = SCARG(uap, data); 1317 break; 1318 default: 1319 break; 1320 } 1321 return error; 1322 } 1323 else 1324 ptr++; 1325 1326 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1327 } 1328 1329 int 1330 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval) 1331 { 1332 /* { 1333 syscallarg(int) magic1; 1334 syscallarg(int) magic2; 1335 syscallarg(int) cmd; 1336 syscallarg(void *) arg; 1337 } */ 1338 struct sys_reboot_args /* { 1339 syscallarg(int) opt; 1340 syscallarg(char *) bootstr; 1341 } */ sra; 1342 int error; 1343 1344 if ((error = kauth_authorize_system(l->l_cred, 1345 KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0) 1346 return(error); 1347 1348 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1349 return(EINVAL); 1350 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1351 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1352 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1353 return(EINVAL); 1354 1355 memset(&sra, 0, sizeof(sra)); 1356 1357 switch ((unsigned long)SCARG(uap, cmd)) { 1358 case LINUX_REBOOT_CMD_RESTART: 1359 SCARG(&sra, opt) = RB_AUTOBOOT; 1360 break; 1361 case LINUX_REBOOT_CMD_HALT: 1362 SCARG(&sra, opt) = RB_HALT; 1363 break; 1364 case LINUX_REBOOT_CMD_POWER_OFF: 1365 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1366 break; 1367 case LINUX_REBOOT_CMD_RESTART2: 1368 /* Reboot with an argument. */ 1369 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1370 SCARG(&sra, bootstr) = SCARG(uap, arg); 1371 break; 1372 case LINUX_REBOOT_CMD_CAD_ON: 1373 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1374 case LINUX_REBOOT_CMD_CAD_OFF: 1375 return(0); 1376 default: 1377 return(EINVAL); 1378 } 1379 1380 return(sys_reboot(l, &sra, retval)); 1381 } 1382 1383 /* 1384 * Copy of compat_12_sys_swapon(). 1385 */ 1386 int 1387 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval) 1388 { 1389 /* { 1390 syscallarg(const char *) name; 1391 } */ 1392 struct sys_swapctl_args ua; 1393 1394 memset(&ua, 0, sizeof(ua)); 1395 SCARG(&ua, cmd) = SWAP_ON; 1396 SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name)); 1397 SCARG(&ua, misc) = 0; /* priority */ 1398 return (sys_swapctl(l, &ua, retval)); 1399 } 1400 1401 /* 1402 * Stop swapping to the file or block device specified by path. 1403 */ 1404 int 1405 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval) 1406 { 1407 /* { 1408 syscallarg(const char *) path; 1409 } */ 1410 struct sys_swapctl_args ua; 1411 1412 memset(&ua, 0, sizeof(ua)); 1413 SCARG(&ua, cmd) = SWAP_OFF; 1414 SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/ 1415 return (sys_swapctl(l, &ua, retval)); 1416 } 1417 1418 /* 1419 * Copy of compat_09_sys_setdomainname() 1420 */ 1421 /* ARGSUSED */ 1422 int 1423 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval) 1424 { 1425 /* { 1426 syscallarg(char *) domainname; 1427 syscallarg(int) len; 1428 } */ 1429 int name[2]; 1430 1431 name[0] = CTL_KERN; 1432 name[1] = KERN_DOMAINNAME; 1433 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1434 SCARG(uap, len), l)); 1435 } 1436 1437 /* 1438 * sysinfo() 1439 */ 1440 /* ARGSUSED */ 1441 int 1442 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval) 1443 { 1444 /* { 1445 syscallarg(struct linux_sysinfo *) arg; 1446 } */ 1447 struct linux_sysinfo si; 1448 struct loadavg *la; 1449 int64_t filepg; 1450 1451 memset(&si, 0, sizeof(si)); 1452 si.uptime = time_uptime; 1453 la = &averunnable; 1454 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1455 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1456 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1457 si.totalram = ctob((u_long)physmem); 1458 /* uvm_availmem() may sync the counters. */ 1459 si.freeram = (u_long)uvm_availmem(true) * uvmexp.pagesize; 1460 filepg = cpu_count_get(CPU_COUNT_FILECLEAN) + 1461 cpu_count_get(CPU_COUNT_FILEDIRTY) + 1462 cpu_count_get(CPU_COUNT_FILEUNKNOWN) - 1463 cpu_count_get(CPU_COUNT_EXECPAGES); 1464 si.sharedram = 0; /* XXX */ 1465 si.bufferram = (u_long)(filepg * uvmexp.pagesize); 1466 si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize; 1467 si.freeswap = 1468 (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1469 si.procs = atomic_load_relaxed(&nprocs); 1470 1471 /* The following are only present in newer Linux kernels. */ 1472 si.totalbig = 0; 1473 si.freebig = 0; 1474 si.mem_unit = 1; 1475 1476 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1477 } 1478 1479 int 1480 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval) 1481 { 1482 /* { 1483 syscallarg(int) which; 1484 # ifdef LINUX_LARGEFILE64 1485 syscallarg(struct rlimit *) rlp; 1486 # else 1487 syscallarg(struct orlimit *) rlp; 1488 # endif 1489 } */ 1490 # ifdef LINUX_LARGEFILE64 1491 struct rlimit orl; 1492 # else 1493 struct orlimit orl; 1494 # endif 1495 int which; 1496 1497 which = linux_to_bsd_limit(SCARG(uap, which)); 1498 if (which < 0) 1499 return -which; 1500 1501 memset(&orl, 0, sizeof(orl)); 1502 bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]); 1503 1504 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1505 } 1506 1507 int 1508 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval) 1509 { 1510 /* { 1511 syscallarg(int) which; 1512 # ifdef LINUX_LARGEFILE64 1513 syscallarg(struct rlimit *) rlp; 1514 # else 1515 syscallarg(struct orlimit *) rlp; 1516 # endif 1517 } */ 1518 struct rlimit rl; 1519 # ifdef LINUX_LARGEFILE64 1520 struct rlimit orl; 1521 # else 1522 struct orlimit orl; 1523 # endif 1524 int error; 1525 int which; 1526 1527 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1528 return error; 1529 1530 which = linux_to_bsd_limit(SCARG(uap, which)); 1531 if (which < 0) 1532 return -which; 1533 1534 linux_to_bsd_rlimit(&rl, &orl); 1535 return dosetrlimit(l, l->l_proc, which, &rl); 1536 } 1537 1538 # if !defined(__aarch64__) && !defined(__mips__) && !defined(__amd64__) 1539 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1540 int 1541 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval) 1542 { 1543 return linux_sys_getrlimit(l, (const void *)uap, retval); 1544 } 1545 # endif 1546 1547 int 1548 linux_sys_prlimit64(struct lwp *l, const struct linux_sys_prlimit64_args *uap, register_t *retval) 1549 { 1550 /* { 1551 syscallarg(pid_t) pid; 1552 syscallarg(int) witch; 1553 syscallarg(struct rlimit *) new_rlp; 1554 syscallarg(struct rlimit *) old_rlp; 1555 }; */ 1556 struct rlimit rl, nrl, orl; 1557 struct rlimit *p; 1558 int which; 1559 int error; 1560 1561 /* XXX: Cannot operate any process other than its own */ 1562 if (SCARG(uap, pid) != 0) 1563 return EPERM; 1564 1565 which = linux_to_bsd_limit(SCARG(uap, which)); 1566 if (which < 0) 1567 return -which; 1568 1569 p = SCARG(uap, old_rlp); 1570 if (p != NULL) { 1571 memset(&orl, 0, sizeof(orl)); 1572 bsd_to_linux_rlimit64(&orl, &l->l_proc->p_rlimit[which]); 1573 if ((error = copyout(&orl, p, sizeof(orl))) != 0) 1574 return error; 1575 } 1576 1577 p = SCARG(uap, new_rlp); 1578 if (p != NULL) { 1579 if ((error = copyin(p, &nrl, sizeof(nrl))) != 0) 1580 return error; 1581 1582 linux_to_bsd_rlimit(&rl, &nrl); 1583 return dosetrlimit(l, l->l_proc, which, &rl); 1584 } 1585 1586 return 0; 1587 } 1588 1589 /* 1590 * This gets called for unsupported syscalls. The difference to sys_nosys() 1591 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1592 * This is the way Linux does it and glibc depends on this behaviour. 1593 */ 1594 int 1595 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval) 1596 { 1597 return (ENOSYS); 1598 } 1599 1600 int 1601 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval) 1602 { 1603 /* { 1604 syscallarg(int) which; 1605 syscallarg(int) who; 1606 } */ 1607 struct sys_getpriority_args bsa; 1608 int error; 1609 1610 memset(&bsa, 0, sizeof(bsa)); 1611 SCARG(&bsa, which) = SCARG(uap, which); 1612 SCARG(&bsa, who) = SCARG(uap, who); 1613 1614 if ((error = sys_getpriority(l, &bsa, retval))) 1615 return error; 1616 1617 *retval = NZERO - *retval; 1618 1619 return 0; 1620 } 1621 1622 int 1623 linux_do_sys_utimensat(struct lwp *l, int fd, const char *path, struct timespec *tsp, int flags, register_t *retval) 1624 { 1625 int follow, error; 1626 1627 follow = (flags & LINUX_AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW; 1628 1629 if (path == NULL && fd != AT_FDCWD) { 1630 file_t *fp; 1631 1632 /* fd_getvnode() will use the descriptor for us */ 1633 if ((error = fd_getvnode(fd, &fp)) != 0) 1634 return error; 1635 error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0, 1636 tsp, UIO_SYSSPACE); 1637 fd_putfile(fd); 1638 return error; 1639 } 1640 1641 return do_sys_utimensat(l, fd, NULL, path, follow, tsp, UIO_SYSSPACE); 1642 } 1643 1644 int 1645 linux_sys_utimensat(struct lwp *l, const struct linux_sys_utimensat_args *uap, 1646 register_t *retval) 1647 { 1648 /* { 1649 syscallarg(int) fd; 1650 syscallarg(const char *) path; 1651 syscallarg(const struct linux_timespec *) times; 1652 syscallarg(int) flag; 1653 } */ 1654 int error; 1655 struct linux_timespec lts[2]; 1656 struct timespec *tsp = NULL, ts[2]; 1657 1658 if (SCARG(uap, times)) { 1659 error = copyin(SCARG(uap, times), <s, sizeof(lts)); 1660 if (error != 0) 1661 return error; 1662 linux_to_native_timespec(&ts[0], <s[0]); 1663 linux_to_native_timespec(&ts[1], <s[1]); 1664 tsp = ts; 1665 } 1666 1667 return linux_do_sys_utimensat(l, SCARG(uap, fd), SCARG(uap, path), 1668 tsp, SCARG(uap, flag), retval); 1669 } 1670 1671 int 1672 linux_sys_futex(struct lwp *l, const struct linux_sys_futex_args *uap, 1673 register_t *retval) 1674 { 1675 /* { 1676 syscallarg(int *) uaddr; 1677 syscallarg(int) op; 1678 syscallarg(int) val; 1679 syscallarg(const struct linux_timespec *) timeout; 1680 syscallarg(int *) uaddr2; 1681 syscallarg(int) val3; 1682 } */ 1683 struct linux_timespec lts; 1684 struct timespec ts, *tsp = NULL; 1685 int val2 = 0; 1686 int error; 1687 1688 /* 1689 * Linux overlays the "timeout" field and the "val2" field. 1690 * "timeout" is only valid for FUTEX_WAIT and FUTEX_WAIT_BITSET 1691 * on Linux. 1692 */ 1693 const int op = (SCARG(uap, op) & FUTEX_CMD_MASK); 1694 if ((op == FUTEX_WAIT || op == FUTEX_WAIT_BITSET) && 1695 SCARG(uap, timeout) != NULL) { 1696 if ((error = copyin(SCARG(uap, timeout), 1697 <s, sizeof(lts))) != 0) { 1698 return error; 1699 } 1700 linux_to_native_timespec(&ts, <s); 1701 tsp = &ts; 1702 } else { 1703 val2 = (int)(uintptr_t)SCARG(uap, timeout); 1704 } 1705 1706 return linux_do_futex(SCARG(uap, uaddr), SCARG(uap, op), 1707 SCARG(uap, val), tsp, SCARG(uap, uaddr2), val2, 1708 SCARG(uap, val3), retval); 1709 } 1710 1711 int 1712 linux_do_futex(int *uaddr, int op, int val, struct timespec *timeout, 1713 int *uaddr2, int val2, int val3, register_t *retval) 1714 { 1715 /* 1716 * Always clear FUTEX_PRIVATE_FLAG for Linux processes. 1717 * NetBSD-native futexes exist in different namespace 1718 * depending on FUTEX_PRIVATE_FLAG. This appears not 1719 * to be the case in Linux, and some futex users will 1720 * mix private and non-private ops on the same futex 1721 * object. 1722 */ 1723 return do_futex(uaddr, op & ~FUTEX_PRIVATE_FLAG, 1724 val, timeout, uaddr2, val2, val3, retval); 1725 } 1726 1727 #define LINUX_EFD_SEMAPHORE 0x0001 1728 #define LINUX_EFD_CLOEXEC LINUX_O_CLOEXEC 1729 #define LINUX_EFD_NONBLOCK LINUX_O_NONBLOCK 1730 1731 static int 1732 linux_do_eventfd2(struct lwp *l, unsigned int initval, int flags, 1733 register_t *retval) 1734 { 1735 int nflags = 0; 1736 1737 if (flags & ~(LINUX_EFD_SEMAPHORE | LINUX_EFD_CLOEXEC | 1738 LINUX_EFD_NONBLOCK)) { 1739 return EINVAL; 1740 } 1741 if (flags & LINUX_EFD_SEMAPHORE) { 1742 nflags |= EFD_SEMAPHORE; 1743 } 1744 if (flags & LINUX_EFD_CLOEXEC) { 1745 nflags |= EFD_CLOEXEC; 1746 } 1747 if (flags & LINUX_EFD_NONBLOCK) { 1748 nflags |= EFD_NONBLOCK; 1749 } 1750 1751 return do_eventfd(l, initval, nflags, retval); 1752 } 1753 1754 int 1755 linux_sys_eventfd(struct lwp *l, const struct linux_sys_eventfd_args *uap, 1756 register_t *retval) 1757 { 1758 /* { 1759 syscallarg(unsigned int) initval; 1760 } */ 1761 1762 return linux_do_eventfd2(l, SCARG(uap, initval), 0, retval); 1763 } 1764 1765 int 1766 linux_sys_eventfd2(struct lwp *l, const struct linux_sys_eventfd2_args *uap, 1767 register_t *retval) 1768 { 1769 /* { 1770 syscallarg(unsigned int) initval; 1771 syscallarg(int) flags; 1772 } */ 1773 1774 return linux_do_eventfd2(l, SCARG(uap, initval), SCARG(uap, flags), 1775 retval); 1776 } 1777 1778 #ifndef __aarch64__ 1779 /* 1780 * epoll_create(2). Check size and call sys_epoll_create1. 1781 */ 1782 int 1783 linux_sys_epoll_create(struct lwp *l, 1784 const struct linux_sys_epoll_create_args *uap, register_t *retval) 1785 { 1786 /* { 1787 syscallarg(int) size; 1788 } */ 1789 struct sys_epoll_create1_args ca; 1790 1791 /* 1792 * SCARG(uap, size) is unused. Linux just tests it and then 1793 * forgets it as well. 1794 */ 1795 if (SCARG(uap, size) <= 0) 1796 return EINVAL; 1797 1798 memset(&ca, 0, sizeof(ca)); 1799 SCARG(&ca, flags) = 0; 1800 return sys_epoll_create1(l, &ca, retval); 1801 } 1802 #endif /* !__aarch64__ */ 1803 1804 /* 1805 * epoll_create1(2). Translate the flags and call sys_epoll_create1. 1806 */ 1807 int 1808 linux_sys_epoll_create1(struct lwp *l, 1809 const struct linux_sys_epoll_create1_args *uap, register_t *retval) 1810 { 1811 /* { 1812 syscallarg(int) flags; 1813 } */ 1814 struct sys_epoll_create1_args ca; 1815 1816 if ((SCARG(uap, flags) & ~(LINUX_O_CLOEXEC)) != 0) 1817 return EINVAL; 1818 1819 memset(&ca, 0, sizeof(ca)); 1820 SCARG(&ca, flags) = 0; 1821 if ((SCARG(uap, flags) & LINUX_O_CLOEXEC) != 0) 1822 SCARG(&ca, flags) |= EPOLL_CLOEXEC; 1823 1824 return sys_epoll_create1(l, &ca, retval); 1825 } 1826 1827 /* 1828 * epoll_ctl(2). Copyin event and translate it if necessary and then 1829 * call epoll_ctl_common(). 1830 */ 1831 int 1832 linux_sys_epoll_ctl(struct lwp *l, const struct linux_sys_epoll_ctl_args *uap, 1833 register_t *retval) 1834 { 1835 /* { 1836 syscallarg(int) epfd; 1837 syscallarg(int) op; 1838 syscallarg(int) fd; 1839 syscallarg(struct linux_epoll_event *) event; 1840 } */ 1841 struct linux_epoll_event lee; 1842 struct epoll_event ee; 1843 struct epoll_event *eep; 1844 int error; 1845 1846 if (SCARG(uap, op) != EPOLL_CTL_DEL) { 1847 error = copyin(SCARG(uap, event), &lee, sizeof(lee)); 1848 if (error != 0) 1849 return error; 1850 1851 /* 1852 * On some architectures, struct linux_epoll_event and 1853 * struct epoll_event are packed differently... but otherwise 1854 * the contents are the same. 1855 */ 1856 ee.events = lee.events; 1857 ee.data = lee.data; 1858 1859 eep = ⅇ 1860 } else 1861 eep = NULL; 1862 1863 return epoll_ctl_common(l, retval, SCARG(uap, epfd), SCARG(uap, op), 1864 SCARG(uap, fd), eep); 1865 } 1866 1867 #ifndef __aarch64__ 1868 /* 1869 * epoll_wait(2). Call sys_epoll_pwait(). 1870 */ 1871 int 1872 linux_sys_epoll_wait(struct lwp *l, 1873 const struct linux_sys_epoll_wait_args *uap, register_t *retval) 1874 { 1875 /* { 1876 syscallarg(int) epfd; 1877 syscallarg(struct linux_epoll_event *) events; 1878 syscallarg(int) maxevents; 1879 syscallarg(int) timeout; 1880 } */ 1881 struct linux_sys_epoll_pwait_args ea; 1882 1883 memset(&ea, 0, sizeof(ea)); 1884 SCARG(&ea, epfd) = SCARG(uap, epfd); 1885 SCARG(&ea, events) = SCARG(uap, events); 1886 SCARG(&ea, maxevents) = SCARG(uap, maxevents); 1887 SCARG(&ea, timeout) = SCARG(uap, timeout); 1888 SCARG(&ea, sigmask) = NULL; 1889 1890 return linux_sys_epoll_pwait(l, &ea, retval); 1891 } 1892 #endif /* !__aarch64__ */ 1893 1894 /* 1895 * Main body of epoll_pwait2(2). Translate timeout and sigmask and 1896 * call epoll_wait_common. 1897 */ 1898 static int 1899 linux_epoll_pwait2_common(struct lwp *l, register_t *retval, int epfd, 1900 struct linux_epoll_event *events, int maxevents, 1901 struct linux_timespec *timeout, const linux_sigset_t *sigmask) 1902 { 1903 struct timespec ts, *tsp; 1904 linux_sigset_t lss; 1905 sigset_t ss, *ssp; 1906 struct epoll_event *eep; 1907 struct linux_epoll_event *leep; 1908 int i, error; 1909 1910 if (maxevents <= 0 || maxevents > EPOLL_MAX_EVENTS) 1911 return EINVAL; 1912 1913 if (timeout != NULL) { 1914 linux_to_native_timespec(&ts, timeout); 1915 tsp = &ts; 1916 } else 1917 tsp = NULL; 1918 1919 if (sigmask != NULL) { 1920 error = copyin(sigmask, &lss, sizeof(lss)); 1921 if (error != 0) 1922 return error; 1923 1924 linux_to_native_sigset(&ss, &lss); 1925 ssp = &ss; 1926 } else 1927 ssp = NULL; 1928 1929 eep = kmem_alloc(maxevents * sizeof(*eep), KM_SLEEP); 1930 1931 error = epoll_wait_common(l, retval, epfd, eep, maxevents, tsp, 1932 ssp); 1933 if (error == 0 && *retval > 0) { 1934 leep = kmem_alloc((*retval) * sizeof(*leep), KM_SLEEP); 1935 1936 /* Translate the events (because of packing). */ 1937 for (i = 0; i < *retval; i++) { 1938 leep[i].events = eep[i].events; 1939 leep[i].data = eep[i].data; 1940 } 1941 1942 error = copyout(leep, events, (*retval) * sizeof(*leep)); 1943 kmem_free(leep, (*retval) * sizeof(*leep)); 1944 } 1945 1946 kmem_free(eep, maxevents * sizeof(*eep)); 1947 return error; 1948 } 1949 1950 /* 1951 * epoll_pwait(2). Translate timeout and call sys_epoll_pwait2. 1952 */ 1953 int 1954 linux_sys_epoll_pwait(struct lwp *l, 1955 const struct linux_sys_epoll_pwait_args *uap, register_t *retval) 1956 { 1957 /* { 1958 syscallarg(int) epfd; 1959 syscallarg(struct linux_epoll_event *) events; 1960 syscallarg(int) maxevents; 1961 syscallarg(int) timeout; 1962 syscallarg(linux_sigset_t *) sigmask; 1963 } */ 1964 struct linux_timespec lts, *ltsp; 1965 const int timeout = SCARG(uap, timeout); 1966 1967 if (timeout >= 0) { 1968 /* Convert from milliseconds to timespec. */ 1969 lts.tv_sec = timeout / 1000; 1970 lts.tv_nsec = (timeout % 1000) * 1000000; 1971 1972 ltsp = <s; 1973 } else 1974 ltsp = NULL; 1975 1976 return linux_epoll_pwait2_common(l, retval, SCARG(uap, epfd), 1977 SCARG(uap, events), SCARG(uap, maxevents), ltsp, 1978 SCARG(uap, sigmask)); 1979 } 1980 1981 1982 /* 1983 * epoll_pwait2(2). Copyin timeout and call linux_epoll_pwait2_common(). 1984 */ 1985 int 1986 linux_sys_epoll_pwait2(struct lwp *l, 1987 const struct linux_sys_epoll_pwait2_args *uap, register_t *retval) 1988 { 1989 /* { 1990 syscallarg(int) epfd; 1991 syscallarg(struct linux_epoll_event *) events; 1992 syscallarg(int) maxevents; 1993 syscallarg(struct linux_timespec *) timeout; 1994 syscallarg(linux_sigset_t *) sigmask; 1995 } */ 1996 struct linux_timespec lts, *ltsp; 1997 int error; 1998 1999 if (SCARG(uap, timeout) != NULL) { 2000 error = copyin(SCARG(uap, timeout), <s, sizeof(lts)); 2001 if (error != 0) 2002 return error; 2003 2004 ltsp = <s; 2005 } else 2006 ltsp = NULL; 2007 2008 return linux_epoll_pwait2_common(l, retval, SCARG(uap, epfd), 2009 SCARG(uap, events), SCARG(uap, maxevents), ltsp, 2010 SCARG(uap, sigmask)); 2011 } 2012 2013 #define LINUX_MFD_CLOEXEC 0x0001U 2014 #define LINUX_MFD_ALLOW_SEALING 0x0002U 2015 #define LINUX_MFD_HUGETLB 0x0004U 2016 #define LINUX_MFD_NOEXEC_SEAL 0x0008U 2017 #define LINUX_MFD_EXEC 0x0010U 2018 #define LINUX_MFD_HUGE_FLAGS (0x3f << 26) 2019 2020 #define LINUX_MFD_ALL_FLAGS (LINUX_MFD_CLOEXEC|LINUX_MFD_ALLOW_SEALING \ 2021 |LINUX_MFD_HUGETLB|LINUX_MFD_NOEXEC_SEAL \ 2022 |LINUX_MFD_EXEC|LINUX_MFD_HUGE_FLAGS) 2023 #define LINUX_MFD_KNOWN_FLAGS (LINUX_MFD_CLOEXEC|LINUX_MFD_ALLOW_SEALING) 2024 2025 #define LINUX_MFD_NAME_MAX 249 2026 2027 /* 2028 * memfd_create(2). Do some error checking and then call NetBSD's 2029 * version. 2030 */ 2031 int 2032 linux_sys_memfd_create(struct lwp *l, 2033 const struct linux_sys_memfd_create_args *uap, register_t *retval) 2034 { 2035 /* { 2036 syscallarg(const char *) name; 2037 syscallarg(unsigned int) flags; 2038 } */ 2039 int error; 2040 char *pbuf; 2041 struct sys_memfd_create_args muap; 2042 const unsigned int lflags = SCARG(uap, flags); 2043 2044 KASSERT(LINUX_MFD_NAME_MAX < NAME_MAX); /* sanity check */ 2045 2046 if (lflags & ~LINUX_MFD_ALL_FLAGS) 2047 return EINVAL; 2048 if ((lflags & LINUX_MFD_HUGE_FLAGS) != 0 && 2049 (lflags & LINUX_MFD_HUGETLB) == 0) 2050 return EINVAL; 2051 if ((lflags & LINUX_MFD_HUGETLB) && (lflags & LINUX_MFD_ALLOW_SEALING)) 2052 return EINVAL; 2053 2054 /* Linux has a stricter limit for name size */ 2055 pbuf = PNBUF_GET(); 2056 error = copyinstr(SCARG(uap, name), pbuf, LINUX_MFD_NAME_MAX+1, NULL); 2057 PNBUF_PUT(pbuf); 2058 pbuf = NULL; 2059 if (error != 0) { 2060 if (error == ENAMETOOLONG) 2061 error = EINVAL; 2062 return error; 2063 } 2064 2065 if (lflags & ~LINUX_MFD_KNOWN_FLAGS) { 2066 DPRINTF("%s: ignored flags %#x\n", __func__, 2067 lflags & ~LINUX_MFD_KNOWN_FLAGS); 2068 } 2069 2070 memset(&muap, 0, sizeof(muap)); 2071 SCARG(&muap, name) = SCARG(uap, name); 2072 SCARG(&muap, flags) = lflags & LINUX_MFD_KNOWN_FLAGS; 2073 2074 return sys_memfd_create(l, &muap, retval); 2075 } 2076 2077 #define LINUX_CLOSE_RANGE_UNSHARE 0x02U 2078 #define LINUX_CLOSE_RANGE_CLOEXEC 0x04U 2079 2080 /* 2081 * close_range(2). 2082 */ 2083 int 2084 linux_sys_close_range(struct lwp *l, 2085 const struct linux_sys_close_range_args *uap, register_t *retval) 2086 { 2087 /* { 2088 syscallarg(unsigned int) first; 2089 syscallarg(unsigned int) last; 2090 syscallarg(unsigned int) flags; 2091 } */ 2092 unsigned int fd, last; 2093 file_t *fp; 2094 filedesc_t *fdp; 2095 const unsigned int flags = SCARG(uap, flags); 2096 2097 if (flags & ~(LINUX_CLOSE_RANGE_CLOEXEC|LINUX_CLOSE_RANGE_UNSHARE)) 2098 return EINVAL; 2099 if (SCARG(uap, first) > SCARG(uap, last)) 2100 return EINVAL; 2101 2102 if (flags & LINUX_CLOSE_RANGE_UNSHARE) { 2103 fdp = fd_copy(); 2104 fd_free(); 2105 l->l_proc->p_fd = fdp; 2106 l->l_fd = fdp; 2107 } 2108 2109 last = MIN(SCARG(uap, last), l->l_proc->p_fd->fd_lastfile); 2110 for (fd = SCARG(uap, first); fd <= last; fd++) { 2111 fp = fd_getfile(fd); 2112 if (fp == NULL) 2113 continue; 2114 2115 if (flags & LINUX_CLOSE_RANGE_CLOEXEC) { 2116 fd_set_exclose(l, fd, true); 2117 fd_putfile(fd); 2118 } else 2119 fd_close(fd); 2120 } 2121 2122 return 0; 2123 } 2124 2125 /* 2126 * readahead(2). Call posix_fadvise with POSIX_FADV_WILLNEED with some extra 2127 * error checking. 2128 */ 2129 int 2130 linux_sys_readahead(struct lwp *l, const struct linux_sys_readahead_args *uap, 2131 register_t *retval) 2132 { 2133 /* { 2134 syscallarg(int) fd; 2135 syscallarg(off_t) offset; 2136 syscallarg(size_t) count; 2137 } */ 2138 file_t *fp; 2139 int error = 0; 2140 const int fd = SCARG(uap, fd); 2141 2142 fp = fd_getfile(fd); 2143 if (fp == NULL) 2144 return EBADF; 2145 if ((fp->f_flag & FREAD) == 0) 2146 error = EBADF; 2147 else if (fp->f_type != DTYPE_VNODE || fp->f_vnode->v_type != VREG) 2148 error = EINVAL; 2149 fd_putfile(fd); 2150 if (error != 0) 2151 return error; 2152 2153 return do_posix_fadvise(fd, SCARG(uap, offset), SCARG(uap, count), 2154 POSIX_FADV_WILLNEED); 2155 } 2156 2157 int 2158 linux_sys_getcpu(lwp_t *l, const struct linux_sys_getcpu_args *uap, 2159 register_t *retval) 2160 { 2161 /* { 2162 syscallarg(unsigned int *) cpu; 2163 syscallarg(unsigned int *) node; 2164 syscallarg(struct linux_getcpu_cache *) tcache; 2165 }*/ 2166 int error; 2167 2168 if (SCARG(uap, cpu)) { 2169 u_int cpu_id = l->l_cpu->ci_data.cpu_index; 2170 error = copyout(&cpu_id, SCARG(uap, cpu), sizeof(cpu_id)); 2171 if (error) 2172 return error; 2173 2174 } 2175 2176 // TO-DO: Test on a NUMA machine if the node_id returned is correct 2177 if (SCARG(uap, node)) { 2178 u_int node_id = l->l_cpu->ci_data.cpu_numa_id; 2179 error = copyout(&node_id, SCARG(uap, node), sizeof(node_id)); 2180 if (error) 2181 return error; 2182 } 2183 2184 return 0; 2185 } 2186
Indexes created Sun Oct 04 00:26:11 UTC 2026