xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision deb6f0161a9109e7de9b519dc8dfb9478668dcdd)
1 /*	$NetBSD: linux_misc.c,v 1.240 2018/09/03 16:29:29 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 1.240 2018/09/03 16:29:29 riastradh Exp $");
61 
62 #include <sys/param.h>
63 #include <sys/systm.h>
64 #include <sys/namei.h>
65 #include <sys/proc.h>
66 #include <sys/dirent.h>
67 #include <sys/file.h>
68 #include <sys/stat.h>
69 #include <sys/filedesc.h>
70 #include <sys/ioctl.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/mman.h>
75 #include <sys/mount.h>
76 #include <sys/poll.h>
77 #include <sys/prot.h>
78 #include <sys/reboot.h>
79 #include <sys/resource.h>
80 #include <sys/resourcevar.h>
81 #include <sys/select.h>
82 #include <sys/signal.h>
83 #include <sys/signalvar.h>
84 #include <sys/socket.h>
85 #include <sys/time.h>
86 #include <sys/times.h>
87 #include <sys/vnode.h>
88 #include <sys/uio.h>
89 #include <sys/wait.h>
90 #include <sys/utsname.h>
91 #include <sys/unistd.h>
92 #include <sys/vfs_syscalls.h>
93 #include <sys/swap.h>		/* for SWAP_ON */
94 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
95 #include <sys/kauth.h>
96 
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!",
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 	strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
355 	strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
356 	strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
357 	strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
358 	strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
359 	strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
360 
361 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
362 }
363 
364 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
365 /* Used indirectly on: arm, i386, m68k */
366 
367 /*
368  * New type Linux mmap call.
369  * Only called directly on machines with >= 6 free regs.
370  */
371 int
372 linux_sys_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval)
373 {
374 	/* {
375 		syscallarg(unsigned long) addr;
376 		syscallarg(size_t) len;
377 		syscallarg(int) prot;
378 		syscallarg(int) flags;
379 		syscallarg(int) fd;
380 		syscallarg(linux_off_t) offset;
381 	} */
382 
383 	if (SCARG(uap, offset) & PAGE_MASK)
384 		return EINVAL;
385 
386 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
387 }
388 
389 /*
390  * Guts of most architectures' mmap64() implementations.  This shares
391  * its list of arguments with linux_sys_mmap().
392  *
393  * The difference in linux_sys_mmap2() is that "offset" is actually
394  * (offset / pagesize), not an absolute byte count.  This translation
395  * to pagesize offsets is done inside glibc between the mmap64() call
396  * point, and the actual syscall.
397  */
398 int
399 linux_sys_mmap2(struct lwp *l, const struct linux_sys_mmap2_args *uap, register_t *retval)
400 {
401 	/* {
402 		syscallarg(unsigned long) addr;
403 		syscallarg(size_t) len;
404 		syscallarg(int) prot;
405 		syscallarg(int) flags;
406 		syscallarg(int) fd;
407 		syscallarg(linux_off_t) offset;
408 	} */
409 
410 	return linux_mmap(l, uap, retval,
411 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
412 }
413 
414 /*
415  * Massage arguments and call system mmap(2).
416  */
417 static int
418 linux_mmap(struct lwp *l, const struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
419 {
420 	struct sys_mmap_args cma;
421 	int error;
422 	size_t mmoff=0;
423 
424 	linux_to_bsd_mmap_args(&cma, uap);
425 	SCARG(&cma, pos) = offset;
426 
427 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
428 		/*
429 		 * Request for stack-like memory segment. On linux, this
430 		 * works by mmap()ping (small) segment, which is automatically
431 		 * extended when page fault happens below the currently
432 		 * allocated area. We emulate this by allocating (typically
433 		 * bigger) segment sized at current stack size limit, and
434 		 * offsetting the requested and returned address accordingly.
435 		 * Since physical pages are only allocated on-demand, this
436 		 * is effectively identical.
437 		 */
438 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
439 
440 		if (SCARG(&cma, len) < ssl) {
441 			/* Compute the address offset */
442 			mmoff = round_page(ssl) - SCARG(uap, len);
443 
444 			if (SCARG(&cma, addr))
445 				SCARG(&cma, addr) = (char *)SCARG(&cma, addr) - mmoff;
446 
447 			SCARG(&cma, len) = (size_t) ssl;
448 		}
449 	}
450 
451 	error = sys_mmap(l, &cma, retval);
452 	if (error)
453 		return (error);
454 
455 	/* Shift the returned address for stack-like segment if necessary */
456 	retval[0] += mmoff;
457 
458 	return (0);
459 }
460 
461 static void
462 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
463 {
464 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
465 
466 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
467 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
468 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
469 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
470 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_LOCKED, MAP_WIRED);
471 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
472 
473 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
474 	SCARG(cma, len) = SCARG(uap, len);
475 	SCARG(cma, prot) = SCARG(uap, prot);
476 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
477 		SCARG(cma, prot) |= VM_PROT_READ;
478 	SCARG(cma, flags) = flags;
479 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
480 	SCARG(cma, PAD) = 0;
481 }
482 
483 #define	LINUX_MREMAP_MAYMOVE	1
484 #define	LINUX_MREMAP_FIXED	2
485 
486 int
487 linux_sys_mremap(struct lwp *l, const struct linux_sys_mremap_args *uap, register_t *retval)
488 {
489 	/* {
490 		syscallarg(void *) old_address;
491 		syscallarg(size_t) old_size;
492 		syscallarg(size_t) new_size;
493 		syscallarg(u_long) flags;
494 	} */
495 
496 	struct proc *p;
497 	struct vm_map *map;
498 	vaddr_t oldva;
499 	vaddr_t newva;
500 	size_t oldsize;
501 	size_t newsize;
502 	int flags;
503 	int uvmflags;
504 	int error;
505 
506 	flags = SCARG(uap, flags);
507 	oldva = (vaddr_t)SCARG(uap, old_address);
508 	oldsize = round_page(SCARG(uap, old_size));
509 	newsize = round_page(SCARG(uap, new_size));
510 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
511 		error = EINVAL;
512 		goto done;
513 	}
514 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
515 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
516 			error = EINVAL;
517 			goto done;
518 		}
519 #if 0 /* notyet */
520 		newva = SCARG(uap, new_address);
521 		uvmflags = MAP_FIXED;
522 #else /* notyet */
523 		error = EOPNOTSUPP;
524 		goto done;
525 #endif /* notyet */
526 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
527 		uvmflags = 0;
528 	} else {
529 		newva = oldva;
530 		uvmflags = MAP_FIXED;
531 	}
532 	p = l->l_proc;
533 	map = &p->p_vmspace->vm_map;
534 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
535 	    uvmflags);
536 
537 done:
538 	*retval = (error != 0) ? 0 : (register_t)newva;
539 	return error;
540 }
541 
542 #ifdef USRSTACK
543 int
544 linux_sys_mprotect(struct lwp *l, const struct linux_sys_mprotect_args *uap, register_t *retval)
545 {
546 	/* {
547 		syscallarg(const void *) start;
548 		syscallarg(unsigned long) len;
549 		syscallarg(int) prot;
550 	} */
551 	struct vm_map_entry *entry;
552 	struct vm_map *map;
553 	struct proc *p;
554 	vaddr_t end, start, len, stacklim;
555 	int prot, grows;
556 
557 	start = (vaddr_t)SCARG(uap, start);
558 	len = round_page(SCARG(uap, len));
559 	prot = SCARG(uap, prot);
560 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
561 	prot &= ~grows;
562 	end = start + len;
563 
564 	if (start & PAGE_MASK)
565 		return EINVAL;
566 	if (end < start)
567 		return EINVAL;
568 	if (end == start)
569 		return 0;
570 
571 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
572 		return EINVAL;
573 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
574 		return EINVAL;
575 
576 	p = l->l_proc;
577 	map = &p->p_vmspace->vm_map;
578 	vm_map_lock(map);
579 # ifdef notdef
580 	VM_MAP_RANGE_CHECK(map, start, end);
581 # endif
582 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
583 		vm_map_unlock(map);
584 		return ENOMEM;
585 	}
586 
587 	/*
588 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
589 	 */
590 
591 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
592 	if (grows & LINUX_PROT_GROWSDOWN) {
593 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
594 			start = USRSTACK - stacklim;
595 		} else {
596 			start = entry->start;
597 		}
598 	} else if (grows & LINUX_PROT_GROWSUP) {
599 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
600 			end = USRSTACK + stacklim;
601 		} else {
602 			end = entry->end;
603 		}
604 	}
605 	vm_map_unlock(map);
606 	return uvm_map_protect_user(l, start, end, prot);
607 }
608 #endif /* USRSTACK */
609 
610 /*
611  * This code is partly stolen from src/lib/libc/compat-43/times.c
612  */
613 
614 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
615 
616 int
617 linux_sys_times(struct lwp *l, const struct linux_sys_times_args *uap, register_t *retval)
618 {
619 	/* {
620 		syscallarg(struct times *) tms;
621 	} */
622 	struct proc *p = l->l_proc;
623 	struct timeval t;
624 	int error;
625 
626 	if (SCARG(uap, tms)) {
627 		struct linux_tms ltms;
628 		struct rusage ru;
629 
630 		mutex_enter(p->p_lock);
631 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
632 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
633 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
634 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
635 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
636 		mutex_exit(p->p_lock);
637 
638 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
639 			return error;
640 	}
641 
642 	getmicrouptime(&t);
643 
644 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
645 	return 0;
646 }
647 
648 #undef CONVTCK
649 
650 /*
651  * Linux 'readdir' call. This code is mostly taken from the
652  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
653  * an attempt has been made to keep it a little cleaner (failing
654  * miserably, because of the cruft needed if count 1 is passed).
655  *
656  * The d_off field should contain the offset of the next valid entry,
657  * but in Linux it has the offset of the entry itself. We emulate
658  * that bug here.
659  *
660  * Read in BSD-style entries, convert them, and copy them out.
661  *
662  * Note that this doesn't handle union-mounted filesystems.
663  */
664 int
665 linux_sys_getdents(struct lwp *l, const struct linux_sys_getdents_args *uap, register_t *retval)
666 {
667 	/* {
668 		syscallarg(int) fd;
669 		syscallarg(struct linux_dirent *) dent;
670 		syscallarg(unsigned int) count;
671 	} */
672 	struct dirent *bdp;
673 	struct vnode *vp;
674 	char *inp, *tbuf;		/* BSD-format */
675 	int len, reclen;		/* BSD-format */
676 	char *outp;			/* Linux-format */
677 	int resid, linux_reclen = 0;	/* Linux-format */
678 	struct file *fp;
679 	struct uio auio;
680 	struct iovec aiov;
681 	struct linux_dirent idb;
682 	off_t off;		/* true file offset */
683 	int buflen, error, eofflag, nbytes, oldcall;
684 	struct vattr va;
685 	off_t *cookiebuf = NULL, *cookie;
686 	int ncookies;
687 
688 	/* fd_getvnode() will use the descriptor for us */
689 	if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
690 		return (error);
691 
692 	if ((fp->f_flag & FREAD) == 0) {
693 		error = EBADF;
694 		goto out1;
695 	}
696 
697 	vp = (struct vnode *)fp->f_data;
698 	if (vp->v_type != VDIR) {
699 		error = ENOTDIR;
700 		goto out1;
701 	}
702 
703 	vn_lock(vp, LK_SHARED | LK_RETRY);
704 	error = VOP_GETATTR(vp, &va, l->l_cred);
705 	VOP_UNLOCK(vp);
706 	if (error)
707 		goto out1;
708 
709 	nbytes = SCARG(uap, count);
710 	if (nbytes == 1) {	/* emulating old, broken behaviour */
711 		nbytes = sizeof (idb);
712 		buflen = uimax(va.va_blocksize, nbytes);
713 		oldcall = 1;
714 	} else {
715 		buflen = uimin(MAXBSIZE, nbytes);
716 		if (buflen < va.va_blocksize)
717 			buflen = va.va_blocksize;
718 		oldcall = 0;
719 	}
720 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
721 
722 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
723 	off = fp->f_offset;
724 again:
725 	aiov.iov_base = tbuf;
726 	aiov.iov_len = buflen;
727 	auio.uio_iov = &aiov;
728 	auio.uio_iovcnt = 1;
729 	auio.uio_rw = UIO_READ;
730 	auio.uio_resid = buflen;
731 	auio.uio_offset = off;
732 	UIO_SETUP_SYSSPACE(&auio);
733 	/*
734          * First we read into the malloc'ed buffer, then
735          * we massage it into user space, one record at a time.
736          */
737 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
738 	    &ncookies);
739 	if (error)
740 		goto out;
741 
742 	inp = tbuf;
743 	outp = (void *)SCARG(uap, dent);
744 	resid = nbytes;
745 	if ((len = buflen - auio.uio_resid) == 0)
746 		goto eof;
747 
748 	for (cookie = cookiebuf; len > 0; len -= reclen) {
749 		bdp = (struct dirent *)inp;
750 		reclen = bdp->d_reclen;
751 		if (reclen & 3) {
752 			error = EIO;
753 			goto out;
754 		}
755 		if (bdp->d_fileno == 0) {
756 			inp += reclen;	/* it is a hole; squish it out */
757 			if (cookie)
758 				off = *cookie++;
759 			else
760 				off += reclen;
761 			continue;
762 		}
763 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
764 		if (reclen > len || resid < linux_reclen) {
765 			/* entry too big for buffer, so just stop */
766 			outp++;
767 			break;
768 		}
769 		/*
770 		 * Massage in place to make a Linux-shaped dirent (otherwise
771 		 * we have to worry about touching user memory outside of
772 		 * the copyout() call).
773 		 */
774 		idb.d_ino = bdp->d_fileno;
775 		/*
776 		 * The old readdir() call misuses the offset and reclen fields.
777 		 */
778 		if (oldcall) {
779 			idb.d_off = (linux_off_t)linux_reclen;
780 			idb.d_reclen = (u_short)bdp->d_namlen;
781 		} else {
782 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
783 				compat_offseterr(vp, "linux_getdents");
784 				error = EINVAL;
785 				goto out;
786 			}
787 			idb.d_off = (linux_off_t)off;
788 			idb.d_reclen = (u_short)linux_reclen;
789 			/* Linux puts d_type at the end of each record */
790 			*((char *)&idb + idb.d_reclen - 1) = bdp->d_type;
791 		}
792 		memcpy(idb.d_name, bdp->d_name,
793 		    MIN(sizeof(idb.d_name), bdp->d_namlen + 1));
794 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
795 			goto out;
796 		/* advance past this real entry */
797 		inp += reclen;
798 		if (cookie)
799 			off = *cookie++; /* each entry points to itself */
800 		else
801 			off += reclen;
802 		/* advance output past Linux-shaped entry */
803 		outp += linux_reclen;
804 		resid -= linux_reclen;
805 		if (oldcall)
806 			break;
807 	}
808 
809 	/* if we squished out the whole block, try again */
810 	if (outp == (void *)SCARG(uap, dent)) {
811 		if (cookiebuf)
812 			free(cookiebuf, M_TEMP);
813 		cookiebuf = NULL;
814 		goto again;
815 	}
816 	fp->f_offset = off;	/* update the vnode offset */
817 
818 	if (oldcall)
819 		nbytes = resid + linux_reclen;
820 
821 eof:
822 	*retval = nbytes - resid;
823 out:
824 	VOP_UNLOCK(vp);
825 	if (cookiebuf)
826 		free(cookiebuf, M_TEMP);
827 	free(tbuf, M_TEMP);
828 out1:
829 	fd_putfile(SCARG(uap, fd));
830 	return error;
831 }
832 
833 /*
834  * Even when just using registers to pass arguments to syscalls you can
835  * have 5 of them on the i386. So this newer version of select() does
836  * this.
837  */
838 int
839 linux_sys_select(struct lwp *l, const struct linux_sys_select_args *uap, register_t *retval)
840 {
841 	/* {
842 		syscallarg(int) nfds;
843 		syscallarg(fd_set *) readfds;
844 		syscallarg(fd_set *) writefds;
845 		syscallarg(fd_set *) exceptfds;
846 		syscallarg(struct timeval50 *) timeout;
847 	} */
848 
849 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
850 	    SCARG(uap, writefds), SCARG(uap, exceptfds),
851 	    (struct linux_timeval *)SCARG(uap, timeout));
852 }
853 
854 /*
855  * Common code for the old and new versions of select(). A couple of
856  * things are important:
857  * 1) return the amount of time left in the 'timeout' parameter
858  * 2) select never returns ERESTART on Linux, always return EINTR
859  */
860 int
861 linux_select1(struct lwp *l, register_t *retval, int nfds, fd_set *readfds,
862     fd_set *writefds, fd_set *exceptfds, struct linux_timeval *timeout)
863 {
864 	struct timespec ts0, ts1, uts, *ts = NULL;
865 	struct linux_timeval ltv;
866 	int error;
867 
868 	/*
869 	 * Store current time for computation of the amount of
870 	 * time left.
871 	 */
872 	if (timeout) {
873 		if ((error = copyin(timeout, &ltv, sizeof(ltv))))
874 			return error;
875 		uts.tv_sec = ltv.tv_sec;
876 		uts.tv_nsec = ltv.tv_usec * 1000;
877 		if (itimespecfix(&uts)) {
878 			/*
879 			 * The timeval was invalid.  Convert it to something
880 			 * valid that will act as it does under Linux.
881 			 */
882 			uts.tv_sec += uts.tv_nsec / 1000000000;
883 			uts.tv_nsec %= 1000000000;
884 			if (uts.tv_nsec < 0) {
885 				uts.tv_sec -= 1;
886 				uts.tv_nsec += 1000000000;
887 			}
888 			if (uts.tv_sec < 0)
889 				timespecclear(&uts);
890 		}
891 		ts = &uts;
892 		nanotime(&ts0);
893 	}
894 
895 	error = selcommon(retval, nfds, readfds, writefds, exceptfds, ts, NULL);
896 
897 	if (error) {
898 		/*
899 		 * See fs/select.c in the Linux kernel.  Without this,
900 		 * Maelstrom doesn't work.
901 		 */
902 		if (error == ERESTART)
903 			error = EINTR;
904 		return error;
905 	}
906 
907 	if (timeout) {
908 		if (*retval) {
909 			/*
910 			 * Compute how much time was left of the timeout,
911 			 * by subtracting the current time and the time
912 			 * before we started the call, and subtracting
913 			 * that result from the user-supplied value.
914 			 */
915 			nanotime(&ts1);
916 			timespecsub(&ts1, &ts0, &ts1);
917 			timespecsub(&uts, &ts1, &uts);
918 			if (uts.tv_sec < 0)
919 				timespecclear(&uts);
920 		} else
921 			timespecclear(&uts);
922 		ltv.tv_sec = uts.tv_sec;
923 		ltv.tv_usec = uts.tv_nsec / 1000;
924 		if ((error = copyout(&ltv, timeout, sizeof(ltv))))
925 			return error;
926 	}
927 
928 	return 0;
929 }
930 
931 /*
932  * Derived from FreeBSD's sys/compat/linux/linux_misc.c:linux_pselect6()
933  * which was contributed by Dmitry Chagin
934  * https://svnweb.freebsd.org/base?view=revision&revision=283403
935  */
936 int
937 linux_sys_pselect6(struct lwp *l,
938 	const struct linux_sys_pselect6_args *uap, register_t *retval)
939 {
940 	/* {
941 		syscallarg(int) nfds;
942 		syscallarg(fd_set *) readfds;
943 		syscallarg(fd_set *) writefds;
944 		syscallarg(fd_set *) exceptfds;
945 		syscallarg(struct timespec *) timeout;
946 		syscallarg(linux_sized_sigset_t *) ss;
947 	} */
948 	struct timespec uts, ts0, ts1, *tsp;
949 	linux_sized_sigset_t lsss;
950 	struct linux_timespec lts;
951 	linux_sigset_t lss;
952 	sigset_t *ssp;
953 	sigset_t ss;
954 	int error;
955 
956 	ssp = NULL;
957 	if (SCARG(uap, ss) != NULL) {
958 		if ((error = copyin(SCARG(uap, ss), &lsss, sizeof(lsss))) != 0)
959 			return (error);
960 		if (lsss.ss_len != sizeof(lss))
961 			return (EINVAL);
962 		if (lsss.ss != NULL) {
963 			if ((error = copyin(lsss.ss, &lss, sizeof(lss))) != 0)
964 				return (error);
965 			linux_to_native_sigset(&ss, &lss);
966 			ssp = &ss;
967 		}
968 	}
969 
970 	if (SCARG(uap, timeout) != NULL) {
971 		error = copyin(SCARG(uap, timeout), &lts, sizeof(lts));
972 		if (error != 0)
973 			return (error);
974 		linux_to_native_timespec(&uts, &lts);
975 
976 		if (itimespecfix(&uts))
977 			return (EINVAL);
978 
979 		nanotime(&ts0);
980 		tsp = &uts;
981 	} else {
982 		tsp = NULL;
983 	}
984 
985 	error = selcommon(retval, SCARG(uap, nfds), SCARG(uap, readfds),
986 	    SCARG(uap, writefds), SCARG(uap, exceptfds), tsp, ssp);
987 
988 	if (error == 0 && tsp != NULL) {
989 		if (retval != 0) {
990 			/*
991 			 * Compute how much time was left of the timeout,
992 			 * by subtracting the current time and the time
993 			 * before we started the call, and subtracting
994 			 * that result from the user-supplied value.
995 			 */
996 			nanotime(&ts1);
997 			timespecsub(&ts1, &ts0, &ts1);
998 			timespecsub(&uts, &ts1, &uts);
999 			if (uts.tv_sec < 0)
1000 				timespecclear(&uts);
1001 		} else {
1002 			timespecclear(&uts);
1003 		}
1004 
1005 		native_to_linux_timespec(&lts, &uts);
1006 		error = copyout(&lts, SCARG(uap, timeout), sizeof(lts));
1007 	}
1008 
1009 	return (error);
1010 }
1011 
1012 int
1013 linux_sys_ppoll(struct lwp *l,
1014 	const struct linux_sys_ppoll_args *uap, register_t *retval)
1015 {
1016 	/* {
1017 		syscallarg(struct pollfd *) fds;
1018 		syscallarg(u_int) nfds;
1019 		syscallarg(struct linux_timespec *) timeout;
1020 		syscallarg(linux_sigset_t *) sigset;
1021 	} */
1022 	struct linux_timespec lts0, *lts;
1023 	struct timespec ts0, *ts = NULL;
1024 	linux_sigset_t lsigmask0, *lsigmask;
1025 	sigset_t sigmask0, *sigmask = NULL;
1026 	int error;
1027 
1028 	lts = SCARG(uap, timeout);
1029 	if (lts) {
1030 		if ((error = copyin(lts, &lts0, sizeof(lts0))) != 0)
1031 			return error;
1032 		linux_to_native_timespec(&ts0, &lts0);
1033 		ts = &ts0;
1034 	}
1035 
1036 	lsigmask = SCARG(uap, sigset);
1037 	if (lsigmask) {
1038 		if ((error = copyin(lsigmask, &lsigmask0, sizeof(lsigmask0))))
1039 			return error;
1040 		linux_to_native_sigset(&sigmask0, &lsigmask0);
1041 		sigmask = &sigmask0;
1042 	}
1043 
1044 	return pollcommon(retval, SCARG(uap, fds), SCARG(uap, nfds),
1045 	    ts, sigmask);
1046 }
1047 
1048 /*
1049  * Set the 'personality' (emulation mode) for the current process. Only
1050  * accept the Linux personality here (0). This call is needed because
1051  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
1052  * ELF binaries run in Linux mode, not SVR4 mode.
1053  */
1054 int
1055 linux_sys_personality(struct lwp *l, const struct linux_sys_personality_args *uap, register_t *retval)
1056 {
1057 	/* {
1058 		syscallarg(unsigned long) per;
1059 	} */
1060 	struct linux_emuldata *led;
1061 	int per;
1062 
1063 	per = SCARG(uap, per);
1064 	led = l->l_emuldata;
1065 	if (per == LINUX_PER_QUERY) {
1066 		retval[0] = led->led_personality;
1067 		return 0;
1068 	}
1069 
1070 	switch (per & LINUX_PER_MASK) {
1071 	case LINUX_PER_LINUX:
1072 	case LINUX_PER_LINUX32:
1073 		led->led_personality = per;
1074 		break;
1075 
1076 	default:
1077 		return EINVAL;
1078 	}
1079 
1080 	retval[0] = per;
1081 	return 0;
1082 }
1083 
1084 /*
1085  * We have nonexistent fsuid equal to uid.
1086  * If modification is requested, refuse.
1087  */
1088 int
1089 linux_sys_setfsuid(struct lwp *l, const struct linux_sys_setfsuid_args *uap, register_t *retval)
1090 {
1091 	 /* {
1092 		 syscallarg(uid_t) uid;
1093 	 } */
1094 	 uid_t uid;
1095 
1096 	 uid = SCARG(uap, uid);
1097 	 if (kauth_cred_getuid(l->l_cred) != uid)
1098 		 return sys_nosys(l, uap, retval);
1099 
1100 	 *retval = uid;
1101 	 return 0;
1102 }
1103 
1104 int
1105 linux_sys_setfsgid(struct lwp *l, const struct linux_sys_setfsgid_args *uap, register_t *retval)
1106 {
1107 	/* {
1108 		syscallarg(gid_t) gid;
1109 	} */
1110 	gid_t gid;
1111 
1112 	gid = SCARG(uap, gid);
1113 	if (kauth_cred_getgid(l->l_cred) != gid)
1114 		return sys_nosys(l, uap, retval);
1115 
1116 	*retval = gid;
1117 	return 0;
1118 }
1119 
1120 int
1121 linux_sys_setresuid(struct lwp *l, const struct linux_sys_setresuid_args *uap, register_t *retval)
1122 {
1123 	/* {
1124 		syscallarg(uid_t) ruid;
1125 		syscallarg(uid_t) euid;
1126 		syscallarg(uid_t) suid;
1127 	} */
1128 
1129 	/*
1130 	 * Note: These checks are a little different than the NetBSD
1131 	 * setreuid(2) call performs.  This precisely follows the
1132 	 * behavior of the Linux kernel.
1133 	 */
1134 
1135 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1136 			    SCARG(uap, suid),
1137 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1138 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1139 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1140 }
1141 
1142 int
1143 linux_sys_getresuid(struct lwp *l, const struct linux_sys_getresuid_args *uap, register_t *retval)
1144 {
1145 	/* {
1146 		syscallarg(uid_t *) ruid;
1147 		syscallarg(uid_t *) euid;
1148 		syscallarg(uid_t *) suid;
1149 	} */
1150 	kauth_cred_t pc = l->l_cred;
1151 	int error;
1152 	uid_t uid;
1153 
1154 	/*
1155 	 * Linux copies these values out to userspace like so:
1156 	 *
1157 	 *	1. Copy out ruid.
1158 	 *	2. If that succeeds, copy out euid.
1159 	 *	3. If both of those succeed, copy out suid.
1160 	 */
1161 	uid = kauth_cred_getuid(pc);
1162 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1163 		return (error);
1164 
1165 	uid = kauth_cred_geteuid(pc);
1166 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1167 		return (error);
1168 
1169 	uid = kauth_cred_getsvuid(pc);
1170 
1171 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1172 }
1173 
1174 int
1175 linux_sys_ptrace(struct lwp *l, const struct linux_sys_ptrace_args *uap, register_t *retval)
1176 {
1177 	/* {
1178 		i386, m68k, powerpc: T=int
1179 		alpha, amd64: T=long
1180 		syscallarg(T) request;
1181 		syscallarg(T) pid;
1182 		syscallarg(T) addr;
1183 		syscallarg(T) data;
1184 	} */
1185 	const int *ptr;
1186 	int request;
1187 	int error;
1188 
1189 	ptr = linux_ptrace_request_map;
1190 	request = SCARG(uap, request);
1191 	while (*ptr != -1)
1192 		if (*ptr++ == request) {
1193 			struct sys_ptrace_args pta;
1194 
1195 			SCARG(&pta, req) = *ptr;
1196 			SCARG(&pta, pid) = SCARG(uap, pid);
1197 			SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1198 			SCARG(&pta, data) = SCARG(uap, data);
1199 
1200 			/*
1201 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1202 			 * to continue where the process left off previously.
1203  			 * The same thing is achieved by addr == (void *) 1
1204 			 * on NetBSD, so rewrite 'addr' appropriately.
1205 			 */
1206 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1207 				SCARG(&pta, addr) = (void *) 1;
1208 
1209 			error = sysent[SYS_ptrace].sy_call(l, &pta, retval);
1210 			if (error)
1211 				return error;
1212 			switch (request) {
1213 			case LINUX_PTRACE_PEEKTEXT:
1214 			case LINUX_PTRACE_PEEKDATA:
1215 				error = copyout (retval,
1216 				    (void *)SCARG(uap, data),
1217 				    sizeof *retval);
1218 				*retval = SCARG(uap, data);
1219 				break;
1220 			default:
1221 				break;
1222 			}
1223 			return error;
1224 		}
1225 		else
1226 			ptr++;
1227 
1228 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1229 }
1230 
1231 int
1232 linux_sys_reboot(struct lwp *l, const struct linux_sys_reboot_args *uap, register_t *retval)
1233 {
1234 	/* {
1235 		syscallarg(int) magic1;
1236 		syscallarg(int) magic2;
1237 		syscallarg(int) cmd;
1238 		syscallarg(void *) arg;
1239 	} */
1240 	struct sys_reboot_args /* {
1241 		syscallarg(int) opt;
1242 		syscallarg(char *) bootstr;
1243 	} */ sra;
1244 	int error;
1245 
1246 	if ((error = kauth_authorize_system(l->l_cred,
1247 	    KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1248 		return(error);
1249 
1250 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1251 		return(EINVAL);
1252 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1253 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1254 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1255 		return(EINVAL);
1256 
1257 	switch ((unsigned long)SCARG(uap, cmd)) {
1258 	case LINUX_REBOOT_CMD_RESTART:
1259 		SCARG(&sra, opt) = RB_AUTOBOOT;
1260 		break;
1261 	case LINUX_REBOOT_CMD_HALT:
1262 		SCARG(&sra, opt) = RB_HALT;
1263 		break;
1264 	case LINUX_REBOOT_CMD_POWER_OFF:
1265 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1266 		break;
1267 	case LINUX_REBOOT_CMD_RESTART2:
1268 		/* Reboot with an argument. */
1269 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1270 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1271 		break;
1272 	case LINUX_REBOOT_CMD_CAD_ON:
1273 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1274 	case LINUX_REBOOT_CMD_CAD_OFF:
1275 		return(0);
1276 	default:
1277 		return(EINVAL);
1278 	}
1279 
1280 	return(sys_reboot(l, &sra, retval));
1281 }
1282 
1283 /*
1284  * Copy of compat_12_sys_swapon().
1285  */
1286 int
1287 linux_sys_swapon(struct lwp *l, const struct linux_sys_swapon_args *uap, register_t *retval)
1288 {
1289 	/* {
1290 		syscallarg(const char *) name;
1291 	} */
1292 	struct sys_swapctl_args ua;
1293 
1294 	SCARG(&ua, cmd) = SWAP_ON;
1295 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1296 	SCARG(&ua, misc) = 0;	/* priority */
1297 	return (sys_swapctl(l, &ua, retval));
1298 }
1299 
1300 /*
1301  * Stop swapping to the file or block device specified by path.
1302  */
1303 int
1304 linux_sys_swapoff(struct lwp *l, const struct linux_sys_swapoff_args *uap, register_t *retval)
1305 {
1306 	/* {
1307 		syscallarg(const char *) path;
1308 	} */
1309 	struct sys_swapctl_args ua;
1310 
1311 	SCARG(&ua, cmd) = SWAP_OFF;
1312 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1313 	return (sys_swapctl(l, &ua, retval));
1314 }
1315 
1316 /*
1317  * Copy of compat_09_sys_setdomainname()
1318  */
1319 /* ARGSUSED */
1320 int
1321 linux_sys_setdomainname(struct lwp *l, const struct linux_sys_setdomainname_args *uap, register_t *retval)
1322 {
1323 	/* {
1324 		syscallarg(char *) domainname;
1325 		syscallarg(int) len;
1326 	} */
1327 	int name[2];
1328 
1329 	name[0] = CTL_KERN;
1330 	name[1] = KERN_DOMAINNAME;
1331 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1332 			    SCARG(uap, len), l));
1333 }
1334 
1335 /*
1336  * sysinfo()
1337  */
1338 /* ARGSUSED */
1339 int
1340 linux_sys_sysinfo(struct lwp *l, const struct linux_sys_sysinfo_args *uap, register_t *retval)
1341 {
1342 	/* {
1343 		syscallarg(struct linux_sysinfo *) arg;
1344 	} */
1345 	struct linux_sysinfo si;
1346 	struct loadavg *la;
1347 
1348 	si.uptime = time_uptime;
1349 	la = &averunnable;
1350 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1351 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1352 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1353 	si.totalram = ctob((u_long)physmem);
1354 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1355 	si.sharedram = 0;	/* XXX */
1356 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1357 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1358 	si.freeswap =
1359 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1360 	si.procs = nprocs;
1361 
1362 	/* The following are only present in newer Linux kernels. */
1363 	si.totalbig = 0;
1364 	si.freebig = 0;
1365 	si.mem_unit = 1;
1366 
1367 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1368 }
1369 
1370 int
1371 linux_sys_getrlimit(struct lwp *l, const struct linux_sys_getrlimit_args *uap, register_t *retval)
1372 {
1373 	/* {
1374 		syscallarg(int) which;
1375 # ifdef LINUX_LARGEFILE64
1376 		syscallarg(struct rlimit *) rlp;
1377 # else
1378 		syscallarg(struct orlimit *) rlp;
1379 # endif
1380 	} */
1381 # ifdef LINUX_LARGEFILE64
1382 	struct rlimit orl;
1383 # else
1384 	struct orlimit orl;
1385 # endif
1386 	int which;
1387 
1388 	which = linux_to_bsd_limit(SCARG(uap, which));
1389 	if (which < 0)
1390 		return -which;
1391 
1392 	bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1393 
1394 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1395 }
1396 
1397 int
1398 linux_sys_setrlimit(struct lwp *l, const struct linux_sys_setrlimit_args *uap, register_t *retval)
1399 {
1400 	/* {
1401 		syscallarg(int) which;
1402 # ifdef LINUX_LARGEFILE64
1403 		syscallarg(struct rlimit *) rlp;
1404 # else
1405 		syscallarg(struct orlimit *) rlp;
1406 # endif
1407 	} */
1408 	struct rlimit rl;
1409 # ifdef LINUX_LARGEFILE64
1410 	struct rlimit orl;
1411 # else
1412 	struct orlimit orl;
1413 # endif
1414 	int error;
1415 	int which;
1416 
1417 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1418 		return error;
1419 
1420 	which = linux_to_bsd_limit(SCARG(uap, which));
1421 	if (which < 0)
1422 		return -which;
1423 
1424 	linux_to_bsd_rlimit(&rl, &orl);
1425 	return dosetrlimit(l, l->l_proc, which, &rl);
1426 }
1427 
1428 # if !defined(__mips__) && !defined(__amd64__)
1429 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1430 int
1431 linux_sys_ugetrlimit(struct lwp *l, const struct linux_sys_ugetrlimit_args *uap, register_t *retval)
1432 {
1433 	return linux_sys_getrlimit(l, (const void *)uap, retval);
1434 }
1435 # endif
1436 
1437 /*
1438  * This gets called for unsupported syscalls. The difference to sys_nosys()
1439  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1440  * This is the way Linux does it and glibc depends on this behaviour.
1441  */
1442 int
1443 linux_sys_nosys(struct lwp *l, const void *v, register_t *retval)
1444 {
1445 	return (ENOSYS);
1446 }
1447 
1448 int
1449 linux_sys_getpriority(struct lwp *l, const struct linux_sys_getpriority_args *uap, register_t *retval)
1450 {
1451         /* {
1452                 syscallarg(int) which;
1453                 syscallarg(int) who;
1454         } */
1455         struct sys_getpriority_args bsa;
1456         int error;
1457 
1458         SCARG(&bsa, which) = SCARG(uap, which);
1459         SCARG(&bsa, who) = SCARG(uap, who);
1460 
1461         if ((error = sys_getpriority(l, &bsa, retval)))
1462                 return error;
1463 
1464         *retval = NZERO - *retval;
1465 
1466         return 0;
1467 }
1468 
1469 int
1470 linux_do_sys_utimensat(struct lwp *l, int fd, const char *path, struct timespec *tsp, int flags, register_t *retval)
1471 {
1472 	int follow, error;
1473 
1474 	follow = (flags & LINUX_AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
1475 
1476 	if (path == NULL && fd != AT_FDCWD) {
1477 		file_t *fp;
1478 
1479 		/* fd_getvnode() will use the descriptor for us */
1480 		if ((error = fd_getvnode(fd, &fp)) != 0)
1481 			return error;
1482 		error = do_sys_utimensat(l, AT_FDCWD, fp->f_data, NULL, 0,
1483 		    tsp, UIO_SYSSPACE);
1484 		fd_putfile(fd);
1485 		return error;
1486 	}
1487 
1488 	return do_sys_utimensat(l, fd, NULL, path, follow, tsp, UIO_SYSSPACE);
1489 }
1490 
1491 int
1492 linux_sys_utimensat(struct lwp *l, const struct linux_sys_utimensat_args *uap,
1493 	register_t *retval)
1494 {
1495 	/* {
1496 		syscallarg(int) fd;
1497 		syscallarg(const char *) path;
1498 		syscallarg(const struct linux_timespec *) times;
1499 		syscallarg(int) flag;
1500 	} */
1501 	int error;
1502 	struct linux_timespec lts[2];
1503 	struct timespec *tsp = NULL, ts[2];
1504 
1505 	if (SCARG(uap, times)) {
1506 		error = copyin(SCARG(uap, times), &lts, sizeof(lts));
1507 		if (error != 0)
1508 			return error;
1509 		linux_to_native_timespec(&ts[0], &lts[0]);
1510 		linux_to_native_timespec(&ts[1], &lts[1]);
1511 		tsp = ts;
1512 	}
1513 
1514 	return linux_do_sys_utimensat(l, SCARG(uap, fd), SCARG(uap, path),
1515 	    tsp, SCARG(uap, flag), retval);
1516 }
1517