xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 76c7fc5f6b13ed0b1508e6b313e88e59977ed78e)
1 /*	$NetBSD: linux_misc.c,v 1.245 2019/09/20 15:25:19 kamil 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 1.245 2019/09/20 15:25:19 kamil 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 
97 #include <sys/ptrace.h>
98 #include <machine/ptrace.h>
99 
100 #include <sys/syscall.h>
101 #include <sys/syscallargs.h>
102 
103 #include <compat/sys/resource.h>
104 
105 #include <compat/linux/common/linux_machdep.h>
106 #include <compat/linux/common/linux_types.h>
107 #include <compat/linux/common/linux_signal.h>
108 #include <compat/linux/common/linux_ipc.h>
109 #include <compat/linux/common/linux_sem.h>
110 
111 #include <compat/linux/common/linux_fcntl.h>
112 #include <compat/linux/common/linux_mmap.h>
113 #include <compat/linux/common/linux_dirent.h>
114 #include <compat/linux/common/linux_util.h>
115 #include <compat/linux/common/linux_misc.h>
116 #include <compat/linux/common/linux_statfs.h>
117 #include <compat/linux/common/linux_limit.h>
118 #include <compat/linux/common/linux_ptrace.h>
119 #include <compat/linux/common/linux_reboot.h>
120 #include <compat/linux/common/linux_emuldata.h>
121 #include <compat/linux/common/linux_sched.h>
122 
123 #include <compat/linux/linux_syscallargs.h>
124 
125 const int linux_ptrace_request_map[] = {
126 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
127 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
128 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
129 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
130 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
131 	LINUX_PTRACE_CONT,	PT_CONTINUE,
132 	LINUX_PTRACE_KILL,	PT_KILL,
133 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
134 	LINUX_PTRACE_DETACH,	PT_DETACH,
135 # ifdef PT_STEP
136 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
137 # endif
138 	LINUX_PTRACE_SYSCALL,	PT_SYSCALL,
139 	-1
140 };
141 
142 const struct linux_mnttypes linux_fstypes[] = {
143 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
145 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
147 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
151 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
153 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
154 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
155 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
156 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
158 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
159 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
160 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
161 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
162 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
163 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
164 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
165 	{ MOUNT_TMPFS,		LINUX_TMPFS_SUPER_MAGIC		}
166 };
167 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
168 
169 # ifdef DEBUG_LINUX
170 #define DPRINTF(a)	uprintf a
171 # else
172 #define DPRINTF(a)
173 # endif
174 
175 /* Local linux_misc.c functions: */
176 static void linux_to_bsd_mmap_args(struct sys_mmap_args *,
177     const struct linux_sys_mmap_args *);
178 static int linux_mmap(struct lwp *, const struct linux_sys_mmap_args *,
179     register_t *, off_t);
180 
181 
182 /*
183  * The information on a terminated (or stopped) process needs
184  * to be converted in order for Linux binaries to get a valid signal
185  * number out of it.
186  */
187 int
188 bsd_to_linux_wstat(int st)
189 {
190 
191 	int sig;
192 
193 	if (WIFSIGNALED(st)) {
194 		sig = WTERMSIG(st);
195 		if (sig >= 0 && sig < NSIG)
196 			st= (st & ~0177) | native_to_linux_signo[sig];
197 	} else if (WIFSTOPPED(st)) {
198 		sig = WSTOPSIG(st);
199 		if (sig >= 0 && sig < NSIG)
200 			st = (st & ~0xff00) |
201 			    (native_to_linux_signo[sig] << 8);
202 	}
203 	return st;
204 }
205 
206 /*
207  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
208  * reserve some space for a NetBSD-style wait status, and converting
209  * it to what Linux wants.
210  */
211 int
212 linux_sys_wait4(struct lwp *l, const struct linux_sys_wait4_args *uap, register_t *retval)
213 {
214 	/* {
215 		syscallarg(int) pid;
216 		syscallarg(int *) status;
217 		syscallarg(int) options;
218 		syscallarg(struct rusage50 *) rusage;
219 	} */
220 	int error, status, options, linux_options, pid = SCARG(uap, pid);
221 	struct rusage50 ru50;
222 	struct rusage ru;
223 	proc_t *p;
224 
225 	linux_options = SCARG(uap, options);
226 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
227 		return (EINVAL);
228 
229 	options = 0;
230 	if (linux_options & LINUX_WAIT4_WNOHANG)
231 		options |= WNOHANG;
232 	if (linux_options & LINUX_WAIT4_WUNTRACED)
233 		options |= WUNTRACED;
234 	if (linux_options & LINUX_WAIT4_WCONTINUED)
235 		options |= WCONTINUED;
236 	if (linux_options & LINUX_WAIT4_WALL)
237 		options |= WALLSIG;
238 	if (linux_options & LINUX_WAIT4_WCLONE)
239 		options |= WALTSIG;
240 # ifdef DIAGNOSTIC
241 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
242 		printf("WARNING: %s: linux process %d.%d called "
243 		       "waitpid with __WNOTHREAD set!\n",
244 		       __FILE__, l->l_proc->p_pid, l->l_lid);
245 
246 # endif
247 
248 	error = do_sys_wait(&pid, &status, options,
249 	    SCARG(uap, rusage) != NULL ? &ru : NULL);
250 
251 	retval[0] = pid;
252 	if (pid == 0)
253 		return error;
254 
255 	p = curproc;
256 	mutex_enter(p->p_lock);
257 	sigdelset(&p->p_sigpend.sp_set, SIGCHLD); /* XXXAD ksiginfo leak */
258 	mutex_exit(p->p_lock);
259 
260 	if (SCARG(uap, rusage) != NULL) {
261 		rusage_to_rusage50(&ru, &ru50);
262 		error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
263 	}
264 
265 	if (error == 0 && SCARG(uap, status) != NULL) {
266 		status = bsd_to_linux_wstat(status);
267 		error = copyout(&status, SCARG(uap, status), sizeof status);
268 	}
269 
270 	return error;
271 }
272 
273 /*
274  * Linux brk(2).  Like native, but always return the new break value.
275  */
276 int
277 linux_sys_brk(struct lwp *l, const struct linux_sys_brk_args *uap, register_t *retval)
278 {
279 	/* {
280 		syscallarg(char *) nsize;
281 	} */
282 	struct proc *p = l->l_proc;
283 	struct vmspace *vm = p->p_vmspace;
284 	struct sys_obreak_args oba;
285 
286 	SCARG(&oba, nsize) = SCARG(uap, nsize);
287 
288 	(void) sys_obreak(l, &oba, retval);
289 	retval[0] = (register_t)((char *)vm->vm_daddr + ptoa(vm->vm_dsize));
290 	return 0;
291 }
292 
293 /*
294  * Implement the fs stat functions. Straightforward.
295  */
296 int
297 linux_sys_statfs(struct lwp *l, const struct linux_sys_statfs_args *uap, register_t *retval)
298 {
299 	/* {
300 		syscallarg(const char *) path;
301 		syscallarg(struct linux_statfs *) sp;
302 	} */
303 	struct statvfs *sb;
304 	struct linux_statfs ltmp;
305 	int error;
306 
307 	sb = STATVFSBUF_GET();
308 	error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb);
309 	if (error == 0) {
310 		bsd_to_linux_statfs(sb, &ltmp);
311 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
312 	}
313 	STATVFSBUF_PUT(sb);
314 
315 	return error;
316 }
317 
318 int
319 linux_sys_fstatfs(struct lwp *l, const struct linux_sys_fstatfs_args *uap, register_t *retval)
320 {
321 	/* {
322 		syscallarg(int) fd;
323 		syscallarg(struct linux_statfs *) sp;
324 	} */
325 	struct statvfs *sb;
326 	struct linux_statfs ltmp;
327 	int error;
328 
329 	sb = STATVFSBUF_GET();
330 	error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb);
331 	if (error == 0) {
332 		bsd_to_linux_statfs(sb, &ltmp);
333 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
334 	}
335 	STATVFSBUF_PUT(sb);
336 
337 	return error;
338 }
339 
340 /*
341  * uname(). Just copy the info from the various strings stored in the
342  * kernel, and put it in the Linux utsname structure. That structure
343  * is almost the same as the NetBSD one, only it has fields 65 characters
344  * long, and an extra domainname field.
345  */
346 int
347 linux_sys_uname(struct lwp *l, const struct linux_sys_uname_args *uap, register_t *retval)
348 {
349 	/* {
350 		syscallarg(struct linux_utsname *) up;
351 	} */
352 	struct linux_utsname luts;
353 
354 	memset(&luts, 0, sizeof(luts));
355 	strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
356 	strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
357 	strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
358 	strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
359 	strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
360 	strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
361 
362 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
363 }
364 
365 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
366 /* Used indirectly on: arm, i386, m68k */
367 
368 /*
369  * New type Linux mmap call.
370  * Only called directly on machines with >= 6 free regs.
371  */
372 int
373 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval)
374 {
375 	/* {
376 		syscallarg(unsigned long) addr;
377 		syscallarg(size_t) len;
378 		syscallarg(int) prot;
379 		syscallarg(int) flags;
380 		syscallarg(int) fd;
381 		syscallarg(linux_off_t) offset;
382 	} */
383 
384 	if (SCARG(uap, offset) & PAGE_MASK)
385 		return EINVAL;
386 
387 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
388 }
389 
390 /*
391  * Guts of most architectures' mmap64() implementations.  This shares
392  * its list of arguments with linux_sys_mmap().
393  *
394  * The difference in linux_sys_mmap2() is that "offset" is actually
395  * (offset / pagesize), not an absolute byte count.  This translation
396  * to pagesize offsets is done inside glibc between the mmap64() call
397  * point, and the actual syscall.
398  */
399 int
400 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval)
401 {
402 	/* {
403 		syscallarg(unsigned long) addr;
404 		syscallarg(size_t) len;
405 		syscallarg(int) prot;
406 		syscallarg(int) flags;
407 		syscallarg(int) fd;
408 		syscallarg(linux_off_t) offset;
409 	} */
410 
411 	return linux_mmap(l, uap, retval,
412 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
413 }
414 
415 /*
416  * Massage arguments and call system mmap(2).
417  */
418 static int
419 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
420 {
421 	struct sys_mmap_args cma;
422 	int error;
423 	size_t mmoff=0;
424 
425 	linux_to_bsd_mmap_args(&cma, uap);
426 	SCARG(&cma, pos) = offset;
427 
428 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
429 		/*
430 		 * Request for stack-like memory segment. On linux, this
431 		 * works by mmap()ping (small) segment, which is automatically
432 		 * extended when page fault happens below the currently
433 		 * allocated area. We emulate this by allocating (typically
434 		 * bigger) segment sized at current stack size limit, and
435 		 * offsetting the requested and returned address accordingly.
436 		 * Since physical pages are only allocated on-demand, this
437 		 * is effectively identical.
438 		 */
439 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
440 
441 		if (SCARG(&cma, len) < ssl) {
442 			/* Compute the address offset */
443 			mmoff = round_page(ssl) - SCARG(uap, len);
444 
445 			if (SCARG(&cma, addr))
446 				SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff;
447 
448 			SCARG(&cma, len) = (size_t) ssl;
449 		}
450 	}
451 
452 	error = sys_mmap(l, &cma, retval);
453 	if (error)
454 		return (error);
455 
456 	/* Shift the returned address for stack-like segment if necessary */
457 	retval[0] += mmoff;
458 
459 	return (0);
460 }
461 
462 static void
463 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
464 {
465 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
466 
467 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
468 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
469 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
470 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
471 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_LOCKED, MAP_WIRED);
472 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
473 
474 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
475 	SCARG(cma, len) = SCARG(uap, len);
476 	SCARG(cma, prot) = SCARG(uap, prot);
477 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
478 		SCARG(cma, prot) |= VM_PROT_READ;
479 	SCARG(cma, flags) = flags;
480 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
481 	SCARG(cma, PAD) = 0;
482 }
483 
484 #define	LINUX_MREMAP_MAYMOVE	1
485 #define	LINUX_MREMAP_FIXED	2
486 
487 int
488 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval)
489 {
490 	/* {
491 		syscallarg(void *) old_address;
492 		syscallarg(size_t) old_size;
493 		syscallarg(size_t) new_size;
494 		syscallarg(u_long) flags;
495 	} */
496 
497 	struct proc *p;
498 	struct vm_map *map;
499 	vaddr_t oldva;
500 	vaddr_t newva;
501 	size_t oldsize;
502 	size_t newsize;
503 	int flags;
504 	int uvmflags;
505 	int error;
506 
507 	flags = SCARG(uap, flags);
508 	oldva = (vaddr_t)SCARG(uap, old_address);
509 	oldsize = round_page(SCARG(uap, old_size));
510 	newsize = round_page(SCARG(uap, new_size));
511 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
512 		error = EINVAL;
513 		goto done;
514 	}
515 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
516 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
517 			error = EINVAL;
518 			goto done;
519 		}
520 #if 0 /* notyet */
521 		newva = SCARG(uap, new_address);
522 		uvmflags = MAP_FIXED;
523 #else /* notyet */
524 		error = EOPNOTSUPP;
525 		goto done;
526 #endif /* notyet */
527 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
528 		uvmflags = 0;
529 	} else {
530 		newva = oldva;
531 		uvmflags = MAP_FIXED;
532 	}
533 	p = l->l_proc;
534 	map = &p->p_vmspace->vm_map;
535 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
536 	    uvmflags);
537 
538 done:
539 	*retval = (error != 0) ? 0 : (register_t)newva;
540 	return error;
541 }
542 
543 #ifdef USRSTACK
544 int
545 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval)
546 {
547 	/* {
548 		syscallarg(const void *) start;
549 		syscallarg(unsigned long) len;
550 		syscallarg(int) prot;
551 	} */
552 	struct vm_map_entry *entry;
553 	struct vm_map *map;
554 	struct proc *p;
555 	vaddr_t end, start, len, stacklim;
556 	int prot, grows;
557 
558 	start = (vaddr_t)SCARG(uap, start);
559 	len = round_page(SCARG(uap, len));
560 	prot = SCARG(uap, prot);
561 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
562 	prot &= ~grows;
563 	end = start + len;
564 
565 	if (start & PAGE_MASK)
566 		return EINVAL;
567 	if (end < start)
568 		return EINVAL;
569 	if (end == start)
570 		return 0;
571 
572 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
573 		return EINVAL;
574 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
575 		return EINVAL;
576 
577 	p = l->l_proc;
578 	map = &p->p_vmspace->vm_map;
579 	vm_map_lock(map);
580 # ifdef notdef
581 	VM_MAP_RANGE_CHECK(map, start, end);
582 # endif
583 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
584 		vm_map_unlock(map);
585 		return ENOMEM;
586 	}
587 
588 	/*
589 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
590 	 */
591 
592 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
593 	if (grows & LINUX_PROT_GROWSDOWN) {
594 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
595 			start = USRSTACK - stacklim;
596 		} else {
597 			start = entry->start;
598 		}
599 	} else if (grows & LINUX_PROT_GROWSUP) {
600 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
601 			end = USRSTACK + stacklim;
602 		} else {
603 			end = entry->end;
604 		}
605 	}
606 	vm_map_unlock(map);
607 	return uvm_map_protect_user(l, start, end, prot);
608 }
609 #endif /* USRSTACK */
610 
611 /*
612  * This code is partly stolen from src/lib/libc/compat-43/times.c
613  */
614 
615 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
616 
617 int
618 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval)
619 {
620 	/* {
621 		syscallarg(struct times *) tms;
622 	} */
623 	struct proc *p = l->l_proc;
624 	struct timeval t;
625 	int error;
626 
627 	if (SCARG(uap, tms)) {
628 		struct linux_tms ltms;
629 		struct rusage ru;
630 
631 		mutex_enter(p->p_lock);
632 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
633 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
634 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
635 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
636 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
637 		mutex_exit(p->p_lock);
638 
639 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
640 			return error;
641 	}
642 
643 	getmicrouptime(&t);
644 
645 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
646 	return 0;
647 }
648 
649 #undef CONVTCK
650 
651 /*
652  * Linux 'readdir' call. This code is mostly taken from the
653  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
654  * an attempt has been made to keep it a little cleaner (failing
655  * miserably, because of the cruft needed if count 1 is passed).
656  *
657  * The d_off field should contain the offset of the next valid entry,
658  * but in Linux it has the offset of the entry itself. We emulate
659  * that bug here.
660  *
661  * Read in BSD-style entries, convert them, and copy them out.
662  *
663  * Note that this doesn't handle union-mounted filesystems.
664  */
665 int
666 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval)
667 {
668 	/* {
669 		syscallarg(int) fd;
670 		syscallarg(struct linux_dirent *) dent;
671 		syscallarg(unsigned int) count;
672 	} */
673 	struct dirent *bdp;
674 	struct vnode *vp;
675 	char *inp, *tbuf;		/* BSD-format */
676 	int len, reclen;		/* BSD-format */
677 	char *outp;			/* Linux-format */
678 	int resid, linux_reclen = 0;	/* Linux-format */
679 	struct file *fp;
680 	struct uio auio;
681 	struct iovec aiov;
682 	struct linux_dirent idb;
683 	off_t off;		/* true file offset */
684 	int buflen, error, eofflag, nbytes, oldcall;
685 	struct vattr va;
686 	off_t *cookiebuf = NULL, *cookie;
687 	int ncookies;
688 
689 	/* fd_getvnode() will use the descriptor for us */
690 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
691 		return (error);
692 
693 	if ((fp->f_flag & FREAD) == 0) {
694 		error = EBADF;
695 		goto out1;
696 	}
697 
698 	vp = (struct vnode *)fp->f_data;
699 	if (vp->v_type != VDIR) {
700 		error = ENOTDIR;
701 		goto out1;
702 	}
703 
704 	vn_lock(vp, LK_SHARED | LK_RETRY);
705 	error = VOP_GETATTR(vp, &va, l->l_cred);
706 	VOP_UNLOCK(vp);
707 	if (error)
708 		goto out1;
709 
710 	nbytes = SCARG(uap, count);
711 	if (nbytes == 1) {	/* emulating old, broken behaviour */
712 		nbytes = sizeof (idb);
713 		buflen = uimax(va.va_blocksize, nbytes);
714 		oldcall = 1;
715 	} else {
716 		buflen = uimin(MAXBSIZE, nbytes);
717 		if (buflen < va.va_blocksize)
718 			buflen = va.va_blocksize;
719 		oldcall = 0;
720 	}
721 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
722 
723 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
724 	off = fp->f_offset;
725 again:
726 	aiov.iov_base = tbuf;
727 	aiov.iov_len = buflen;
728 	auio.uio_iov = &aiov;
729 	auio.uio_iovcnt = 1;
730 	auio.uio_rw = UIO_READ;
731 	auio.uio_resid = buflen;
732 	auio.uio_offset = off;
733 	UIO_SETUP_SYSSPACE(&auio);
734 	/*
735          * First we read into the malloc'ed buffer, then
736          * we massage it into user space, one record at a time.
737          */
738 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
739 	    &ncookies);
740 	if (error)
741 		goto out;
742 
743 	inp = tbuf;
744 	outp = (void *)SCARG(uap, dent);
745 	resid = nbytes;
746 	if ((len = buflen - auio.uio_resid) == 0)
747 		goto eof;
748 
749 	for (cookie = cookiebuf; len > 0; len -= reclen) {
750 		bdp = (struct dirent *)inp;
751 		reclen = bdp->d_reclen;
752 		if (reclen & 3) {
753 			error = EIO;
754 			goto out;
755 		}
756 		if (bdp->d_fileno == 0) {
757 			inp += reclen;	/* it is a hole; squish it out */
758 			if (cookie)
759 				off = *cookie++;
760 			else
761 				off += reclen;
762 			continue;
763 		}
764 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
765 		if (reclen > len || resid < linux_reclen) {
766 			/* entry too big for buffer, so just stop */
767 			outp++;
768 			break;
769 		}
770 		/*
771 		 * Massage in place to make a Linux-shaped dirent (otherwise
772 		 * we have to worry about touching user memory outside of
773 		 * the copyout() call).
774 		 */
775 		memset(&idb, 0, sizeof(idb));
776 		idb.d_ino = bdp->d_fileno;
777 		/*
778 		 * The old readdir() call misuses the offset and reclen fields.
779 		 */
780 		if (oldcall) {
781 			idb.d_off = (linux_off_t)linux_reclen;
782 			idb.d_reclen = (u_short)bdp->d_namlen;
783 		} else {
784 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
785 				compat_offseterr(vp, "linux_getdents");
786 				error = EINVAL;
787 				goto out;
788 			}
789 			idb.d_off = (linux_off_t)off;
790 			idb.d_reclen = (u_short)linux_reclen;
791 			/* Linux puts d_type at the end of each record */
792 			*((char *)&idb + idb.d_reclen - 1) = bdp->d_type;
793 		}
794 		memcpy(idb.d_name, bdp->d_name,
795 		    MIN(sizeof(idb.d_name), bdp->d_namlen + 1));
796 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
797 			goto out;
798 		/* advance past this real entry */
799 		inp += reclen;
800 		if (cookie)
801 			off = *cookie++; /* each entry points to itself */
802 		else
803 			off += reclen;
804 		/* advance output past Linux-shaped entry */
805 		outp += linux_reclen;
806 		resid -= linux_reclen;
807 		if (oldcall)
808 			break;
809 	}
810 
811 	/* if we squished out the whole block, try again */
812 	if (outp == (void *)SCARG(uap, dent)) {
813 		if (cookiebuf)
814 			free(cookiebuf, M_TEMP);
815 		cookiebuf = NULL;
816 		goto again;
817 	}
818 	fp->f_offset = off;	/* update the vnode offset */
819 
820 	if (oldcall)
821 		nbytes = resid + linux_reclen;
822 
823 eof:
824 	*retval = nbytes - resid;
825 out:
826 	VOP_UNLOCK(vp);
827 	if (cookiebuf)
828 		free(cookiebuf, M_TEMP);
829 	free(tbuf, M_TEMP);
830 out1:
831 	fd_putfile(SCARG(uap, fd));
832 	return error;
833 }
834 
835 /*
836  * Even when just using registers to pass arguments to syscalls you can
837  * have 5 of them on the i386. So this newer version of select() does
838  * this.
839  */
840 int
841 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval)
842 {
843 	/* {
844 		syscallarg(int) nfds;
845 		syscallarg(fd_set *) readfds;
846 		syscallarg(fd_set *) writefds;
847 		syscallarg(fd_set *) exceptfds;
848 		syscallarg(struct timeval50 *) timeout;
849 	} */
850 
851 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
852 	    SCARG(uap, writefds), SCARG(uap, exceptfds),
853 	    (struct linux_timeval *)SCARG(uap, timeout));
854 }
855 
856 /*
857  * Common code for the old and new versions of select(). A couple of
858  * things are important:
859  * 1) return the amount of time left in the 'timeout' parameter
860  * 2) select never returns ERESTART on Linux, always return EINTR
861  */
862 int
863 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds,
864     fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout)
865 {
866 	struct timespec ts0, ts1, uts, *ts = NULL;
867 	struct linux_timeval ltv;
868 	int error;
869 
870 	/*
871 	 * Store current time for computation of the amount of
872 	 * time left.
873 	 */
874 	if (timeout) {
875 		if ((error = copyin(timeout, &ltv, sizeof(ltv))))
876 			return error;
877 		uts.tv_sec = ltv.tv_sec;
878 		uts.tv_nsec = (long)((unsigned long)ltv.tv_usec * 1000);
879 		if (itimespecfix(&uts)) {
880 			/*
881 			 * The timeval was invalid.  Convert it to something
882 			 * valid that will act as it does under Linux.
883 			 */
884 			uts.tv_sec += uts.tv_nsec / 1000000000;
885 			uts.tv_nsec %= 1000000000;
886 			if (uts.tv_nsec < 0) {
887 				uts.tv_sec -= 1;
888 				uts.tv_nsec += 1000000000;
889 			}
890 			if (uts.tv_sec < 0)
891 				timespecclear(&uts);
892 		}
893 		ts = &uts;
894 		nanotime(&ts0);
895 	}
896 
897 	error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL);
898 
899 	if (error) {
900 		/*
901 		 * See fs/select.c in the Linux kernel.  Without this,
902 		 * Maelstrom doesn't work.
903 		 */
904 		if (error == ERESTART)
905 			error = EINTR;
906 		return error;
907 	}
908 
909 	if (timeout) {
910 		if (*retval) {
911 			/*
912 			 * Compute how much time was left of the timeout,
913 			 * by subtracting the current time and the time
914 			 * before we started the call, and subtracting
915 			 * that result from the user-supplied value.
916 			 */
917 			nanotime(&ts1);
918 			timespecsub(&ts1, &ts0, &ts1);
919 			timespecsub(&uts, &ts1, &uts);
920 			if (uts.tv_sec < 0)
921 				timespecclear(&uts);
922 		} else
923 			timespecclear(&uts);
924 		ltv.tv_sec = uts.tv_sec;
925 		ltv.tv_usec = uts.tv_nsec / 1000;
926 		if ((error = copyout(&ltv, timeout, sizeof(ltv))))
927 			return error;
928 	}
929 
930 	return 0;
931 }
932 
933 /*
934  * Derived from FreeBSD's sys/compat/linux/linux_misc.c:linux_pselect6()
935  * which was contributed by Dmitry Chagin
936  * https://svnweb.freebsd.org/base?view=revision&revision=283403
937  */
938 int
939 linux_sys_pselect6(struct lwp *l,
940 	const struct linux_sys_pselect6_args *uap, register_t *retval)
941 {
942 	/* {
943 		syscallarg(int) nfds;
944 		syscallarg(fd_set *) readfds;
945 		syscallarg(fd_set *) writefds;
946 		syscallarg(fd_set *) exceptfds;
947 		syscallarg(struct timespec *) timeout;
948 		syscallarg(linux_sized_sigset_t *) ss;
949 	} */
950 	struct timespec uts, ts0, ts1, *tsp;
951 	linux_sized_sigset_t lsss;
952 	struct linux_timespec lts;
953 	linux_sigset_t lss;
954 	sigset_t *ssp;
955 	sigset_t ss;
956 	int error;
957 
958 	ssp = NULL;
959 	if (SCARG(uap, ss) != NULL) {
960 		if ((error = copyin(SCARG(uap, ss), &lsss, sizeof(lsss))) != 0)
961 			return (error);
962 		if (lsss.ss_len != sizeof(lss))
963 			return (EINVAL);
964 		if (lsss.ss != NULL) {
965 			if ((error = copyin(lsss.ss, &lss, sizeof(lss))) != 0)
966 				return (error);
967 			linux_to_native_sigset(&ss, &lss);
968 			ssp = &ss;
969 		}
970 	}
971 
972 	if (SCARG(uap, timeout) != NULL) {
973 		error = copyin(SCARG(uap, timeout), &lts, sizeof(lts));
974 		if (error != 0)
975 			return (error);
976 		linux_to_native_timespec(&uts, &lts);
977 
978 		if (itimespecfix(&uts))
979 			return (EINVAL);
980 
981 		nanotime(&ts0);
982 		tsp = &uts;
983 	} else {
984 		tsp = NULL;
985 	}
986 
987 	error = selcommon(retval, SCARG(uap, nfds), SCARG(uap, readfds),
988 	    SCARG(uap, writefds), SCARG(uap, exceptfds), tsp, ssp);
989 
990 	if (error == 0 && tsp != NULL) {
991 		if (retval != 0) {
992 			/*
993 			 * Compute how much time was left of the timeout,
994 			 * by subtracting the current time and the time
995 			 * before we started the call, and subtracting
996 			 * that result from the user-supplied value.
997 			 */
998 			nanotime(&ts1);
999 			timespecsub(&ts1, &ts0, &ts1);
1000 			timespecsub(&uts, &ts1, &uts);
1001 			if (uts.tv_sec < 0)
1002 				timespecclear(&uts);
1003 		} else {
1004 			timespecclear(&uts);
1005 		}
1006 
1007 		native_to_linux_timespec(&lts, &uts);
1008 		error = copyout(&lts, SCARG(uap, timeout), sizeof(lts));
1009 	}
1010 
1011 	return (error);
1012 }
1013 
1014 int
1015 linux_sys_ppoll(struct lwp *l,
1016 	const struct linux_sys_ppoll_args *uap, register_t *retval)
1017 {
1018 	/* {
1019 		syscallarg(struct pollfd *) fds;
1020 		syscallarg(u_int) nfds;
1021 		syscallarg(struct linux_timespec *) timeout;
1022 		syscallarg(linux_sigset_t *) sigset;
1023 	} */
1024 	struct linux_timespec lts0, *lts;
1025 	struct timespec ts0, *ts = NULL;
1026 	linux_sigset_t lsigmask0, *lsigmask;
1027 	sigset_t sigmask0, *sigmask = NULL;
1028 	int error;
1029 
1030 	lts = SCARG(uap, timeout);
1031 	if (lts) {
1032 		if ((error = copyin(lts, &lts0, sizeof(lts0))) != 0)
1033 			return error;
1034 		linux_to_native_timespec(&ts0, &lts0);
1035 		ts = &ts0;
1036 	}
1037 
1038 	lsigmask = SCARG(uap, sigset);
1039 	if (lsigmask) {
1040 		if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0))))
1041 			return error;
1042 		linux_to_native_sigset(&sigmask0, &lsigmask0);
1043 		sigmask = &sigmask0;
1044 	}
1045 
1046 	return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds),
1047 	    ts, sigmask);
1048 }
1049 
1050 /*
1051  * Set the 'personality' (emulation mode) for the current process. Only
1052  * accept the Linux personality here (0). This call is needed because
1053  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1054  * ELF binaries run in Linux mode, not SVR4 mode.
1055  */
1056 int
1057 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval)
1058 {
1059 	/* {
1060 		syscallarg(unsigned long) per;
1061 	} */
1062 	struct linux_emuldata *led;
1063 	int per;
1064 
1065 	per = SCARG(uap, per);
1066 	led = l->l_emuldata;
1067 	if (per == LINUX_PER_QUERY) {
1068 		retval[0] = led->led_personality;
1069 		return 0;
1070 	}
1071 
1072 	switch (per & LINUX_PER_MASK) {
1073 	case LINUX_PER_LINUX:
1074 	case LINUX_PER_LINUX32:
1075 		led->led_personality = per;
1076 		break;
1077 
1078 	default:
1079 		return EINVAL;
1080 	}
1081 
1082 	retval[0] = per;
1083 	return 0;
1084 }
1085 
1086 /*
1087  * We have nonexistent fsuid equal to uid.
1088  * If modification is requested, refuse.
1089  */
1090 int
1091 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval)
1092 {
1093 	 /* {
1094 		 syscallarg(uid_t) uid;
1095 	 } */
1096 	 uid_t uid;
1097 
1098 	 uid = SCARG(uap, uid);
1099 	 if (kauth_cred_getuid(l->l_cred) != uid)
1100 		 return sys_nosys(l, uap, retval);
1101 
1102 	 *retval = uid;
1103 	 return 0;
1104 }
1105 
1106 int
1107 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval)
1108 {
1109 	/* {
1110 		syscallarg(gid_t) gid;
1111 	} */
1112 	gid_t gid;
1113 
1114 	gid = SCARG(uap, gid);
1115 	if (kauth_cred_getgid(l->l_cred) != gid)
1116 		return sys_nosys(l, uap, retval);
1117 
1118 	*retval = gid;
1119 	return 0;
1120 }
1121 
1122 int
1123 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval)
1124 {
1125 	/* {
1126 		syscallarg(uid_t) ruid;
1127 		syscallarg(uid_t) euid;
1128 		syscallarg(uid_t) suid;
1129 	} */
1130 
1131 	/*
1132 	 * Note: These checks are a little different than the NetBSD
1133 	 * setreuid(2) call performs.  This precisely follows the
1134 	 * behavior of the Linux kernel.
1135 	 */
1136 
1137 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1138 			    SCARG(uap, suid),
1139 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1140 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1141 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1142 }
1143 
1144 int
1145 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval)
1146 {
1147 	/* {
1148 		syscallarg(uid_t *) ruid;
1149 		syscallarg(uid_t *) euid;
1150 		syscallarg(uid_t *) suid;
1151 	} */
1152 	kauth_cred_t pc = l->l_cred;
1153 	int error;
1154 	uid_t uid;
1155 
1156 	/*
1157 	 * Linux copies these values out to userspace like so:
1158 	 *
1159 	 *	1. Copy out ruid.
1160 	 *	2. If that succeeds, copy out euid.
1161 	 *	3. If both of those succeed, copy out suid.
1162 	 */
1163 	uid = kauth_cred_getuid(pc);
1164 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1165 		return (error);
1166 
1167 	uid = kauth_cred_geteuid(pc);
1168 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1169 		return (error);
1170 
1171 	uid = kauth_cred_getsvuid(pc);
1172 
1173 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1174 }
1175 
1176 int
1177 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval)
1178 {
1179 	/* {
1180 		i386, m68k, powerpc: T=int
1181 		alpha, amd64: T=long
1182 		syscallarg(T) request;
1183 		syscallarg(T) pid;
1184 		syscallarg(T) addr;
1185 		syscallarg(T) data;
1186 	} */
1187 	const int *ptr;
1188 	int request;
1189 	int error;
1190 
1191 	ptr = linux_ptrace_request_map;
1192 	request = SCARG(uap, request);
1193 	while (*ptr != -1)
1194 		if (*ptr++ == request) {
1195 			struct sys_ptrace_args pta;
1196 
1197 			SCARG(&pta, req) = *ptr;
1198 			SCARG(&pta, pid) = SCARG(uap, pid);
1199 			SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1200 			SCARG(&pta, data) = SCARG(uap, data);
1201 
1202 			/*
1203 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1204 			 * to continue where the process left off previously.
1205  			 * The same thing is achieved by addr == (void *) 1
1206 			 * on NetBSD, so rewrite 'addr' appropriately.
1207 			 */
1208 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1209 				SCARG(&pta, addr) = (void *) 1;
1210 
1211 			error = sysent[SYS_ptrace].sy_call(l, &pta, retval);
1212 			if (error)
1213 				return error;
1214 			switch (request) {
1215 			case LINUX_PTRACE_PEEKTEXT:
1216 			case LINUX_PTRACE_PEEKDATA:
1217 				error = copyout (retval,
1218 				    (void *)SCARG(uap, data),
1219 				    sizeof *retval);
1220 				*retval = SCARG(uap, data);
1221 				break;
1222 			default:
1223 				break;
1224 			}
1225 			return error;
1226 		}
1227 		else
1228 			ptr++;
1229 
1230 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1231 }
1232 
1233 int
1234 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval)
1235 {
1236 	/* {
1237 		syscallarg(int) magic1;
1238 		syscallarg(int) magic2;
1239 		syscallarg(int) cmd;
1240 		syscallarg(void *) arg;
1241 	} */
1242 	struct sys_reboot_args /* {
1243 		syscallarg(int) opt;
1244 		syscallarg(char *) bootstr;
1245 	} */ sra;
1246 	int error;
1247 
1248 	if ((error = kauth_authorize_system(l->l_cred,
1249 	    KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1250 		return(error);
1251 
1252 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1253 		return(EINVAL);
1254 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1255 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1256 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1257 		return(EINVAL);
1258 
1259 	switch ((unsigned long)SCARG(uap, cmd)) {
1260 	case LINUX_REBOOT_CMD_RESTART:
1261 		SCARG(&sra, opt) = RB_AUTOBOOT;
1262 		break;
1263 	case LINUX_REBOOT_CMD_HALT:
1264 		SCARG(&sra, opt) = RB_HALT;
1265 		break;
1266 	case LINUX_REBOOT_CMD_POWER_OFF:
1267 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1268 		break;
1269 	case LINUX_REBOOT_CMD_RESTART2:
1270 		/* Reboot with an argument. */
1271 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1272 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1273 		break;
1274 	case LINUX_REBOOT_CMD_CAD_ON:
1275 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1276 	case LINUX_REBOOT_CMD_CAD_OFF:
1277 		return(0);
1278 	default:
1279 		return(EINVAL);
1280 	}
1281 
1282 	return(sys_reboot(l, &sra, retval));
1283 }
1284 
1285 /*
1286  * Copy of compat_12_sys_swapon().
1287  */
1288 int
1289 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval)
1290 {
1291 	/* {
1292 		syscallarg(const char *) name;
1293 	} */
1294 	struct sys_swapctl_args ua;
1295 
1296 	SCARG(&ua, cmd) = SWAP_ON;
1297 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1298 	SCARG(&ua, misc) = 0;	/* priority */
1299 	return (sys_swapctl(l, &ua, retval));
1300 }
1301 
1302 /*
1303  * Stop swapping to the file or block device specified by path.
1304  */
1305 int
1306 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval)
1307 {
1308 	/* {
1309 		syscallarg(const char *) path;
1310 	} */
1311 	struct sys_swapctl_args ua;
1312 
1313 	SCARG(&ua, cmd) = SWAP_OFF;
1314 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1315 	return (sys_swapctl(l, &ua, retval));
1316 }
1317 
1318 /*
1319  * Copy of compat_09_sys_setdomainname()
1320  */
1321 /* ARGSUSED */
1322 int
1323 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval)
1324 {
1325 	/* {
1326 		syscallarg(char *) domainname;
1327 		syscallarg(int) len;
1328 	} */
1329 	int name[2];
1330 
1331 	name[0] = CTL_KERN;
1332 	name[1] = KERN_DOMAINNAME;
1333 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1334 			    SCARG(uap, len), l));
1335 }
1336 
1337 /*
1338  * sysinfo()
1339  */
1340 /* ARGSUSED */
1341 int
1342 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval)
1343 {
1344 	/* {
1345 		syscallarg(struct linux_sysinfo *) arg;
1346 	} */
1347 	struct linux_sysinfo si;
1348 	struct loadavg *la;
1349 
1350 	memset(&si, 0, sizeof(si));
1351 	si.uptime = time_uptime;
1352 	la = &averunnable;
1353 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1354 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1355 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1356 	si.totalram = ctob((u_long)physmem);
1357 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1358 	si.sharedram = 0;	/* XXX */
1359 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1360 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1361 	si.freeswap =
1362 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1363 	si.procs = nprocs;
1364 
1365 	/* The following are only present in newer Linux kernels. */
1366 	si.totalbig = 0;
1367 	si.freebig = 0;
1368 	si.mem_unit = 1;
1369 
1370 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1371 }
1372 
1373 int
1374 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval)
1375 {
1376 	/* {
1377 		syscallarg(int) which;
1378 # ifdef LINUX_LARGEFILE64
1379 		syscallarg(struct rlimit *) rlp;
1380 # else
1381 		syscallarg(struct orlimit *) rlp;
1382 # endif
1383 	} */
1384 # ifdef LINUX_LARGEFILE64
1385 	struct rlimit orl;
1386 # else
1387 	struct orlimit orl;
1388 # endif
1389 	int which;
1390 
1391 	which = linux_to_bsd_limit(SCARG(uap, which));
1392 	if (which < 0)
1393 		return -which;
1394 
1395 	bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1396 
1397 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1398 }
1399 
1400 int
1401 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval)
1402 {
1403 	/* {
1404 		syscallarg(int) which;
1405 # ifdef LINUX_LARGEFILE64
1406 		syscallarg(struct rlimit *) rlp;
1407 # else
1408 		syscallarg(struct orlimit *) rlp;
1409 # endif
1410 	} */
1411 	struct rlimit rl;
1412 # ifdef LINUX_LARGEFILE64
1413 	struct rlimit orl;
1414 # else
1415 	struct orlimit orl;
1416 # endif
1417 	int error;
1418 	int which;
1419 
1420 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1421 		return error;
1422 
1423 	which = linux_to_bsd_limit(SCARG(uap, which));
1424 	if (which < 0)
1425 		return -which;
1426 
1427 	linux_to_bsd_rlimit(&rl, &orl);
1428 	return dosetrlimit(l, l->l_proc, which, &rl);
1429 }
1430 
1431 # if !defined(__mips__) && !defined(__amd64__)
1432 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1433 int
1434 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval)
1435 {
1436 	return linux_sys_getrlimit(l, (const void *)uap, retval);
1437 }
1438 # endif
1439 
1440 /*
1441  * This gets called for unsupported syscalls. The difference to sys_nosys()
1442  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1443  * This is the way Linux does it and glibc depends on this behaviour.
1444  */
1445 int
1446 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval)
1447 {
1448 	return (ENOSYS);
1449 }
1450 
1451 int
1452 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval)
1453 {
1454         /* {
1455                 syscallarg(int) which;
1456                 syscallarg(int) who;
1457         } */
1458         struct sys_getpriority_args bsa;
1459         int error;
1460 
1461         SCARG(&bsa, which) = SCARG(uap, which);
1462         SCARG(&bsa, who) = SCARG(uap, who);
1463 
1464         if ((error = sys_getpriority(l, &bsa, retval)))
1465                 return error;
1466 
1467         *retval = NZERO - *retval;
1468 
1469         return 0;
1470 }
1471 
1472 int
1473 linux_do_sys_utimensat(struct lwp *l, int fd, const char *path, struct timespec *tsp, int flags, register_t *retval)
1474 {
1475 	int follow, error;
1476 
1477 	follow = (flags & LINUX_AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
1478 
1479 	if (path == NULL && fd != AT_FDCWD) {
1480 		file_t *fp;
1481 
1482 		/* fd_getvnode() will use the descriptor for us */
1483 		if ((error = fd_getvnode(fd, &fp)) != 0)
1484 			return error;
1485 		error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0,
1486 		    tsp, UIO_SYSSPACE);
1487 		fd_putfile(fd);
1488 		return error;
1489 	}
1490 
1491 	return do_sys_utimensat(l, fd, NULL, path, follow, tsp, UIO_SYSSPACE);
1492 }
1493 
1494 int
1495 linux_sys_utimensat(struct lwp *l, const struct linux_sys_utimensat_args *uap,
1496 	register_t *retval)
1497 {
1498 	/* {
1499 		syscallarg(int) fd;
1500 		syscallarg(const char *) path;
1501 		syscallarg(const struct linux_timespec *) times;
1502 		syscallarg(int) flag;
1503 	} */
1504 	int error;
1505 	struct linux_timespec lts[2];
1506 	struct timespec *tsp = NULL, ts[2];
1507 
1508 	if (SCARG(uap, times)) {
1509 		error = copyin(SCARG(uap, times), &lts, sizeof(lts));
1510 		if (error != 0)
1511 			return error;
1512 		linux_to_native_timespec(&ts[0], &lts[0]);
1513 		linux_to_native_timespec(&ts[1], &lts[1]);
1514 		tsp = ts;
1515 	}
1516 
1517 	return linux_do_sys_utimensat(l, SCARG(uap, fd), SCARG(uap, path),
1518 	    tsp, SCARG(uap, flag), retval);
1519 }
1520