xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: linux_misc.c,v 1.189 2007/12/08 18:36:08 dsl Exp $	*/
2 
3 /*-
4  * Copyright (c) 1995, 1998, 1999 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Frank van der Linden and Eric Haszlakiewicz; by Jason R. Thorpe
9  * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Linux compatibility module. Try to deal with various Linux system calls.
42  */
43 
44 /*
45  * These functions have been moved to multiarch to allow
46  * selection of which machines include them to be
47  * determined by the individual files.linux_<arch> files.
48  *
49  * Function in multiarch:
50  *	linux_sys_break			: linux_break.c
51  *	linux_sys_alarm			: linux_misc_notalpha.c
52  *	linux_sys_getresgid		: linux_misc_notalpha.c
53  *	linux_sys_nice			: linux_misc_notalpha.c
54  *	linux_sys_readdir		: linux_misc_notalpha.c
55  *	linux_sys_setresgid		: linux_misc_notalpha.c
56  *	linux_sys_time			: linux_misc_notalpha.c
57  *	linux_sys_utime			: linux_misc_notalpha.c
58  *	linux_sys_waitpid		: linux_misc_notalpha.c
59  *	linux_sys_old_mmap		: linux_oldmmap.c
60  *	linux_sys_oldolduname		: linux_oldolduname.c
61  *	linux_sys_oldselect		: linux_oldselect.c
62  *	linux_sys_olduname		: linux_olduname.c
63  *	linux_sys_pipe			: linux_pipe.c
64  */
65 
66 #include <sys/cdefs.h>
67 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.189 2007/12/08 18:36:08 dsl Exp $");
68 
69 #if defined(_KERNEL_OPT)
70 #include "opt_ptrace.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/namei.h>
76 #include <sys/proc.h>
77 #include <sys/dirent.h>
78 #include <sys/file.h>
79 #include <sys/stat.h>
80 #include <sys/filedesc.h>
81 #include <sys/ioctl.h>
82 #include <sys/kernel.h>
83 #include <sys/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/mman.h>
86 #include <sys/mount.h>
87 #include <sys/prot.h>
88 #include <sys/reboot.h>
89 #include <sys/resource.h>
90 #include <sys/resourcevar.h>
91 #include <sys/select.h>
92 #include <sys/signal.h>
93 #include <sys/signalvar.h>
94 #include <sys/socket.h>
95 #include <sys/time.h>
96 #include <sys/times.h>
97 #include <sys/vnode.h>
98 #include <sys/uio.h>
99 #include <sys/wait.h>
100 #include <sys/utsname.h>
101 #include <sys/unistd.h>
102 #include <sys/vfs_syscalls.h>
103 #include <sys/swap.h>		/* for SWAP_ON */
104 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
105 #include <sys/kauth.h>
106 
107 #include <sys/ptrace.h>
108 #include <machine/ptrace.h>
109 
110 #include <sys/syscall.h>
111 #include <sys/syscallargs.h>
112 
113 #include <compat/linux/common/linux_machdep.h>
114 #include <compat/linux/common/linux_types.h>
115 #include <compat/linux/common/linux_signal.h>
116 #include <compat/linux/common/linux_ipc.h>
117 #include <compat/linux/common/linux_sem.h>
118 
119 #include <compat/linux/linux_syscallargs.h>
120 
121 #include <compat/linux/common/linux_fcntl.h>
122 #include <compat/linux/common/linux_mmap.h>
123 #include <compat/linux/common/linux_dirent.h>
124 #include <compat/linux/common/linux_util.h>
125 #include <compat/linux/common/linux_misc.h>
126 #ifndef COMPAT_LINUX32
127 #include <compat/linux/common/linux_statfs.h>
128 #include <compat/linux/common/linux_limit.h>
129 #endif
130 #include <compat/linux/common/linux_ptrace.h>
131 #include <compat/linux/common/linux_reboot.h>
132 #include <compat/linux/common/linux_emuldata.h>
133 
134 #ifndef COMPAT_LINUX32
135 const int linux_ptrace_request_map[] = {
136 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
137 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
138 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
139 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
140 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
141 	LINUX_PTRACE_CONT,	PT_CONTINUE,
142 	LINUX_PTRACE_KILL,	PT_KILL,
143 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
144 	LINUX_PTRACE_DETACH,	PT_DETACH,
145 # ifdef PT_STEP
146 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
147 # endif
148 	LINUX_PTRACE_SYSCALL,	PT_SYSCALL,
149 	-1
150 };
151 
152 const struct linux_mnttypes linux_fstypes[] = {
153 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
155 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
156 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
157 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
158 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
159 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
160 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
161 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
162 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
163 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
164 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
165 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
166 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
167 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
168 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
169 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
170 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
171 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
172 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
173 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
174 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		},
175 	{ MOUNT_PTYFS,		LINUX_DEVPTS_SUPER_MAGIC	},
176 	{ MOUNT_TMPFS,		LINUX_DEFAULT_SUPER_MAGIC	}
177 };
178 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]);
179 
180 # ifdef DEBUG_LINUX
181 #define DPRINTF(a)	uprintf a
182 # else
183 #define DPRINTF(a)
184 # endif
185 
186 /* Local linux_misc.c functions: */
187 static void linux_to_bsd_mmap_args(struct sys_mmap_args *,
188     const struct linux_sys_mmap_args *);
189 static int linux_mmap(struct lwp *, struct linux_sys_mmap_args *,
190     register_t *, off_t);
191 
192 
193 /*
194  * The information on a terminated (or stopped) process needs
195  * to be converted in order for Linux binaries to get a valid signal
196  * number out of it.
197  */
198 int
199 bsd_to_linux_wstat(int st)
200 {
201 
202 	int sig;
203 
204 	if (WIFSIGNALED(st)) {
205 		sig = WTERMSIG(st);
206 		if (sig >= 0 && sig < NSIG)
207 			st= (st & ~0177) | native_to_linux_signo[sig];
208 	} else if (WIFSTOPPED(st)) {
209 		sig = WSTOPSIG(st);
210 		if (sig >= 0 && sig < NSIG)
211 			st = (st & ~0xff00) |
212 			    (native_to_linux_signo[sig] << 8);
213 	}
214 	return st;
215 }
216 
217 /*
218  * wait4(2).  Passed on to the NetBSD call, surrounded by code to
219  * reserve some space for a NetBSD-style wait status, and converting
220  * it to what Linux wants.
221  */
222 int
223 linux_sys_wait4(struct lwp *l, void *v, register_t *retval)
224 {
225 	struct linux_sys_wait4_args /* {
226 		syscallarg(int) pid;
227 		syscallarg(int *) status;
228 		syscallarg(int) options;
229 		syscallarg(struct rusage *) rusage;
230 	} */ *uap = v;
231 	int error, status, options, linux_options, was_zombie;
232 	struct rusage ru;
233 
234 	linux_options = SCARG(uap, options);
235 	options = WOPTSCHECKED;
236 	if (linux_options & ~(LINUX_WAIT4_KNOWNFLAGS))
237 		return (EINVAL);
238 
239 	if (linux_options & LINUX_WAIT4_WNOHANG)
240 		options |= WNOHANG;
241 	if (linux_options & LINUX_WAIT4_WUNTRACED)
242 		options |= WUNTRACED;
243 	if (linux_options & LINUX_WAIT4_WALL)
244 		options |= WALLSIG;
245 	if (linux_options & LINUX_WAIT4_WCLONE)
246 		options |= WALTSIG;
247 # ifdef DIAGNOSTIC
248 	if (linux_options & LINUX_WAIT4_WNOTHREAD)
249 		printf("WARNING: %s: linux process %d.%d called "
250 		       "waitpid with __WNOTHREAD set!",
251 		       __FILE__, l->l_proc->p_pid, l->l_lid);
252 
253 # endif
254 
255 	error = do_sys_wait(l, &SCARG(uap, pid), &status, options,
256 	    SCARG(uap, rusage) != NULL ? &ru : NULL, &was_zombie);
257 
258 	retval[0] = SCARG(uap, pid);
259 	if (SCARG(uap, pid) == 0)
260 		return error;
261 
262 	sigdelset(&l->l_proc->p_sigpend.sp_set, SIGCHLD);	/* XXXAD ksiginfo leak */
263 
264 	if (SCARG(uap, rusage) != NULL)
265 		error = copyout(&ru, SCARG(uap, rusage), sizeof(ru));
266 
267 	if (error == 0 && SCARG(uap, status) != NULL) {
268 		status = bsd_to_linux_wstat(status);
269 		error = copyout(&status, SCARG(uap, status), sizeof status);
270 	}
271 
272 	return error;
273 }
274 
275 /*
276  * Linux brk(2). The check if the new address is >= the old one is
277  * done in the kernel in Linux. NetBSD does it in the library.
278  */
279 int
280 linux_sys_brk(struct lwp *l, void *v, register_t *retval)
281 {
282 	struct linux_sys_brk_args /* {
283 		syscallarg(char *) nsize;
284 	} */ *uap = v;
285 	struct proc *p = l->l_proc;
286 	char *nbrk = SCARG(uap, nsize);
287 	struct sys_obreak_args oba;
288 	struct vmspace *vm = p->p_vmspace;
289 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
290 
291 	SCARG(&oba, nsize) = nbrk;
292 
293 	if ((void *) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
294 		ed->s->p_break = (char*)nbrk;
295 	else
296 		nbrk = ed->s->p_break;
297 
298 	retval[0] = (register_t)nbrk;
299 
300 	return 0;
301 }
302 
303 /*
304  * Implement the fs stat functions. Straightforward.
305  */
306 int
307 linux_sys_statfs(struct lwp *l, void *v, register_t *retval)
308 {
309 	struct linux_sys_statfs_args /* {
310 		syscallarg(const char *) path;
311 		syscallarg(struct linux_statfs *) sp;
312 	} */ *uap = v;
313 	struct statvfs *sb;
314 	struct linux_statfs ltmp;
315 	int error;
316 
317 	sb = STATVFSBUF_GET();
318 	error = do_sys_pstatvfs(l, SCARG(uap, path), ST_WAIT, sb);
319 	if (error == 0) {
320 		bsd_to_linux_statfs(sb, &ltmp);
321 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
322 	}
323 	STATVFSBUF_PUT(sb);
324 
325 	return error;
326 }
327 
328 int
329 linux_sys_fstatfs(struct lwp *l, void *v, register_t *retval)
330 {
331 	struct linux_sys_fstatfs_args /* {
332 		syscallarg(int) fd;
333 		syscallarg(struct linux_statfs *) sp;
334 	} */ *uap = v;
335 	struct statvfs *sb;
336 	struct linux_statfs ltmp;
337 	int error;
338 
339 	sb = STATVFSBUF_GET();
340 	error = do_sys_fstatvfs(l, SCARG(uap, fd), ST_WAIT, sb);
341 	if (error == 0) {
342 		bsd_to_linux_statfs(sb, &ltmp);
343 		error = copyout(&ltmp, SCARG(uap, sp), sizeof ltmp);
344 	}
345 	STATVFSBUF_PUT(sb);
346 
347 	return error;
348 }
349 
350 /*
351  * uname(). Just copy the info from the various strings stored in the
352  * kernel, and put it in the Linux utsname structure. That structure
353  * is almost the same as the NetBSD one, only it has fields 65 characters
354  * long, and an extra domainname field.
355  */
356 int
357 linux_sys_uname(struct lwp *l, void *v, register_t *retval)
358 {
359 	struct linux_sys_uname_args /* {
360 		syscallarg(struct linux_utsname *) up;
361 	} */ *uap = v;
362 	struct linux_utsname luts;
363 
364 	strlcpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
365 	strlcpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
366 	strlcpy(luts.l_release, linux_release, sizeof(luts.l_release));
367 	strlcpy(luts.l_version, linux_version, sizeof(luts.l_version));
368 	strlcpy(luts.l_machine, LINUX_UNAME_ARCH, sizeof(luts.l_machine));
369 	strlcpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
370 
371 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
372 }
373 
374 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
375 /* Used indirectly on: arm, i386, m68k */
376 
377 /*
378  * New type Linux mmap call.
379  * Only called directly on machines with >= 6 free regs.
380  */
381 int
382 linux_sys_mmap(struct lwp *l, void *v, register_t *retval)
383 {
384 	struct linux_sys_mmap_args /* {
385 		syscallarg(unsigned long) addr;
386 		syscallarg(size_t) len;
387 		syscallarg(int) prot;
388 		syscallarg(int) flags;
389 		syscallarg(int) fd;
390 		syscallarg(linux_off_t) offset;
391 	} */ *uap = v;
392 
393 	if (SCARG(uap, offset) & PAGE_MASK)
394 		return EINVAL;
395 
396 	return linux_mmap(l, uap, retval, SCARG(uap, offset));
397 }
398 
399 /*
400  * Guts of most architectures' mmap64() implementations.  This shares
401  * its list of arguments with linux_sys_mmap().
402  *
403  * The difference in linux_sys_mmap2() is that "offset" is actually
404  * (offset / pagesize), not an absolute byte count.  This translation
405  * to pagesize offsets is done inside glibc between the mmap64() call
406  * point, and the actual syscall.
407  */
408 int
409 linux_sys_mmap2(struct lwp *l, void *v, register_t *retval)
410 {
411 	struct linux_sys_mmap2_args /* {
412 		syscallarg(unsigned long) addr;
413 		syscallarg(size_t) len;
414 		syscallarg(int) prot;
415 		syscallarg(int) flags;
416 		syscallarg(int) fd;
417 		syscallarg(linux_off_t) offset;
418 	} */ *uap = v;
419 
420 	return linux_mmap(l, uap, retval,
421 	    ((off_t)SCARG(uap, offset)) << PAGE_SHIFT);
422 }
423 
424 /*
425  * Massage arguments and call system mmap(2).
426  */
427 static int
428 linux_mmap(struct lwp *l, struct linux_sys_mmap_args *uap, register_t *retval, off_t offset)
429 {
430 	struct sys_mmap_args cma;
431 	int error;
432 	size_t mmoff=0;
433 
434 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) {
435 		/*
436 		 * Request for stack-like memory segment. On linux, this
437 		 * works by mmap()ping (small) segment, which is automatically
438 		 * extended when page fault happens below the currently
439 		 * allocated area. We emulate this by allocating (typically
440 		 * bigger) segment sized at current stack size limit, and
441 		 * offsetting the requested and returned address accordingly.
442 		 * Since physical pages are only allocated on-demand, this
443 		 * is effectively identical.
444 		 */
445 		rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur;
446 
447 		if (SCARG(uap, len) < ssl) {
448 			/* Compute the address offset */
449 			mmoff = round_page(ssl) - SCARG(uap, len);
450 
451 			if (SCARG(uap, addr))
452 				SCARG(uap, addr) -= mmoff;
453 
454 			SCARG(uap, len) = (size_t) ssl;
455 		}
456 	}
457 
458 	linux_to_bsd_mmap_args(&cma, uap);
459 	SCARG(&cma, pos) = offset;
460 
461 	error = sys_mmap(l, &cma, retval);
462 	if (error)
463 		return (error);
464 
465 	/* Shift the returned address for stack-like segment if necessary */
466 	if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff)
467 		retval[0] += mmoff;
468 
469 	return (0);
470 }
471 
472 static void
473 linux_to_bsd_mmap_args(struct sys_mmap_args *cma, const struct linux_sys_mmap_args *uap)
474 {
475 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
476 
477 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
478 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
479 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
480 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
481 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
482 
483 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
484 	SCARG(cma, len) = SCARG(uap, len);
485 	SCARG(cma, prot) = SCARG(uap, prot);
486 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
487 		SCARG(cma, prot) |= VM_PROT_READ;
488 	SCARG(cma, flags) = flags;
489 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
490 	SCARG(cma, pad) = 0;
491 }
492 
493 #define	LINUX_MREMAP_MAYMOVE	1
494 #define	LINUX_MREMAP_FIXED	2
495 
496 int
497 linux_sys_mremap(struct lwp *l, void *v, register_t *retval)
498 {
499 	struct linux_sys_mremap_args /* {
500 		syscallarg(void *) old_address;
501 		syscallarg(size_t) old_size;
502 		syscallarg(size_t) new_size;
503 		syscallarg(u_long) flags;
504 	} */ *uap = v;
505 
506 	struct proc *p;
507 	struct vm_map *map;
508 	vaddr_t oldva;
509 	vaddr_t newva;
510 	size_t oldsize;
511 	size_t newsize;
512 	int flags;
513 	int uvmflags;
514 	int error;
515 
516 	flags = SCARG(uap, flags);
517 	oldva = (vaddr_t)SCARG(uap, old_address);
518 	oldsize = round_page(SCARG(uap, old_size));
519 	newsize = round_page(SCARG(uap, new_size));
520 	if ((flags & ~(LINUX_MREMAP_FIXED|LINUX_MREMAP_MAYMOVE)) != 0) {
521 		error = EINVAL;
522 		goto done;
523 	}
524 	if ((flags & LINUX_MREMAP_FIXED) != 0) {
525 		if ((flags & LINUX_MREMAP_MAYMOVE) == 0) {
526 			error = EINVAL;
527 			goto done;
528 		}
529 #if 0 /* notyet */
530 		newva = SCARG(uap, new_address);
531 		uvmflags = MAP_FIXED;
532 #else /* notyet */
533 		error = EOPNOTSUPP;
534 		goto done;
535 #endif /* notyet */
536 	} else if ((flags & LINUX_MREMAP_MAYMOVE) != 0) {
537 		uvmflags = 0;
538 	} else {
539 		newva = oldva;
540 		uvmflags = MAP_FIXED;
541 	}
542 	p = l->l_proc;
543 	map = &p->p_vmspace->vm_map;
544 	error = uvm_mremap(map, oldva, oldsize, map, &newva, newsize, p,
545 	    uvmflags);
546 
547 done:
548 	*retval = (error != 0) ? 0 : (register_t)newva;
549 	return error;
550 }
551 
552 int
553 linux_sys_msync(struct lwp *l, void *v, register_t *retval)
554 {
555 	struct linux_sys_msync_args /* {
556 		syscallarg(void *) addr;
557 		syscallarg(int) len;
558 		syscallarg(int) fl;
559 	} */ *uap = v;
560 
561 	struct sys___msync13_args bma;
562 
563 	/* flags are ignored */
564 	SCARG(&bma, addr) = SCARG(uap, addr);
565 	SCARG(&bma, len) = SCARG(uap, len);
566 	SCARG(&bma, flags) = SCARG(uap, fl);
567 
568 	return sys___msync13(l, &bma, retval);
569 }
570 
571 int
572 linux_sys_mprotect(struct lwp *l, void *v, register_t *retval)
573 {
574 	struct linux_sys_mprotect_args /* {
575 		syscallarg(const void *) start;
576 		syscallarg(unsigned long) len;
577 		syscallarg(int) prot;
578 	} */ *uap = v;
579 	struct vm_map_entry *entry;
580 	struct vm_map *map;
581 	struct proc *p;
582 	vaddr_t end, start, len, stacklim;
583 	int prot, grows;
584 
585 	start = (vaddr_t)SCARG(uap, start);
586 	len = round_page(SCARG(uap, len));
587 	prot = SCARG(uap, prot);
588 	grows = prot & (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP);
589 	prot &= ~grows;
590 	end = start + len;
591 
592 	if (start & PAGE_MASK)
593 		return EINVAL;
594 	if (end < start)
595 		return EINVAL;
596 	if (end == start)
597 		return 0;
598 
599 	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
600 		return EINVAL;
601 	if (grows == (LINUX_PROT_GROWSDOWN | LINUX_PROT_GROWSUP))
602 		return EINVAL;
603 
604 	p = l->l_proc;
605 	map = &p->p_vmspace->vm_map;
606 	vm_map_lock(map);
607 # ifdef notdef
608 	VM_MAP_RANGE_CHECK(map, start, end);
609 # endif
610 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
611 		vm_map_unlock(map);
612 		return ENOMEM;
613 	}
614 
615 	/*
616 	 * Approximate the behaviour of PROT_GROWS{DOWN,UP}.
617 	 */
618 
619 	stacklim = (vaddr_t)p->p_limit->pl_rlimit[RLIMIT_STACK].rlim_cur;
620 	if (grows & LINUX_PROT_GROWSDOWN) {
621 		if (USRSTACK - stacklim <= start && start < USRSTACK) {
622 			start = USRSTACK - stacklim;
623 		} else {
624 			start = entry->start;
625 		}
626 	} else if (grows & LINUX_PROT_GROWSUP) {
627 		if (USRSTACK <= end && end < USRSTACK + stacklim) {
628 			end = USRSTACK + stacklim;
629 		} else {
630 			end = entry->end;
631 		}
632 	}
633 	vm_map_unlock(map);
634 	return uvm_map_protect(map, start, end, prot, FALSE);
635 }
636 
637 /*
638  * This code is partly stolen from src/lib/libc/compat-43/times.c
639  */
640 
641 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
642 
643 int
644 linux_sys_times(struct lwp *l, void *v, register_t *retval)
645 {
646 	struct linux_sys_times_args /* {
647 		syscallarg(struct times *) tms;
648 	} */ *uap = v;
649 	struct proc *p = l->l_proc;
650 	struct timeval t;
651 	int error;
652 
653 	if (SCARG(uap, tms)) {
654 		struct linux_tms ltms;
655 		struct rusage ru;
656 
657 		mutex_enter(&p->p_smutex);
658 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL, NULL);
659 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
660 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
661 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
662 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
663 		mutex_exit(&p->p_smutex);
664 
665 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
666 			return error;
667 	}
668 
669 	getmicrouptime(&t);
670 
671 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
672 	return 0;
673 }
674 
675 #undef CONVTCK
676 
677 /*
678  * Linux 'readdir' call. This code is mostly taken from the
679  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
680  * an attempt has been made to keep it a little cleaner (failing
681  * miserably, because of the cruft needed if count 1 is passed).
682  *
683  * The d_off field should contain the offset of the next valid entry,
684  * but in Linux it has the offset of the entry itself. We emulate
685  * that bug here.
686  *
687  * Read in BSD-style entries, convert them, and copy them out.
688  *
689  * Note that this doesn't handle union-mounted filesystems.
690  */
691 int
692 linux_sys_getdents(struct lwp *l, void *v, register_t *retval)
693 {
694 	struct linux_sys_getdents_args /* {
695 		syscallarg(int) fd;
696 		syscallarg(struct linux_dirent *) dent;
697 		syscallarg(unsigned int) count;
698 	} */ *uap = v;
699 	struct dirent *bdp;
700 	struct vnode *vp;
701 	char *inp, *tbuf;		/* BSD-format */
702 	int len, reclen;		/* BSD-format */
703 	char *outp;			/* Linux-format */
704 	int resid, linux_reclen = 0;	/* Linux-format */
705 	struct file *fp;
706 	struct uio auio;
707 	struct iovec aiov;
708 	struct linux_dirent idb;
709 	off_t off;		/* true file offset */
710 	int buflen, error, eofflag, nbytes, oldcall;
711 	struct vattr va;
712 	off_t *cookiebuf = NULL, *cookie;
713 	int ncookies;
714 
715 	/* getvnode() will use the descriptor for us */
716 	if ((error = getvnode(l->l_proc->p_fd, SCARG(uap, fd), &fp)) != 0)
717 		return (error);
718 
719 	if ((fp->f_flag & FREAD) == 0) {
720 		error = EBADF;
721 		goto out1;
722 	}
723 
724 	vp = (struct vnode *)fp->f_data;
725 	if (vp->v_type != VDIR) {
726 		error = EINVAL;
727 		goto out1;
728 	}
729 
730 	if ((error = VOP_GETATTR(vp, &va, l->l_cred)))
731 		goto out1;
732 
733 	nbytes = SCARG(uap, count);
734 	if (nbytes == 1) {	/* emulating old, broken behaviour */
735 		nbytes = sizeof (idb);
736 		buflen = max(va.va_blocksize, nbytes);
737 		oldcall = 1;
738 	} else {
739 		buflen = min(MAXBSIZE, nbytes);
740 		if (buflen < va.va_blocksize)
741 			buflen = va.va_blocksize;
742 		oldcall = 0;
743 	}
744 	tbuf = malloc(buflen, M_TEMP, M_WAITOK);
745 
746 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
747 	off = fp->f_offset;
748 again:
749 	aiov.iov_base = tbuf;
750 	aiov.iov_len = buflen;
751 	auio.uio_iov = &aiov;
752 	auio.uio_iovcnt = 1;
753 	auio.uio_rw = UIO_READ;
754 	auio.uio_resid = buflen;
755 	auio.uio_offset = off;
756 	UIO_SETUP_SYSSPACE(&auio);
757 	/*
758          * First we read into the malloc'ed buffer, then
759          * we massage it into user space, one record at a time.
760          */
761 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
762 	    &ncookies);
763 	if (error)
764 		goto out;
765 
766 	inp = tbuf;
767 	outp = (void *)SCARG(uap, dent);
768 	resid = nbytes;
769 	if ((len = buflen - auio.uio_resid) == 0)
770 		goto eof;
771 
772 	for (cookie = cookiebuf; len > 0; len -= reclen) {
773 		bdp = (struct dirent *)inp;
774 		reclen = bdp->d_reclen;
775 		if (reclen & 3)
776 			panic("linux_readdir");
777 		if (bdp->d_fileno == 0) {
778 			inp += reclen;	/* it is a hole; squish it out */
779 			if (cookie)
780 				off = *cookie++;
781 			else
782 				off += reclen;
783 			continue;
784 		}
785 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
786 		if (reclen > len || resid < linux_reclen) {
787 			/* entry too big for buffer, so just stop */
788 			outp++;
789 			break;
790 		}
791 		/*
792 		 * Massage in place to make a Linux-shaped dirent (otherwise
793 		 * we have to worry about touching user memory outside of
794 		 * the copyout() call).
795 		 */
796 		idb.d_ino = bdp->d_fileno;
797 		/*
798 		 * The old readdir() call misuses the offset and reclen fields.
799 		 */
800 		if (oldcall) {
801 			idb.d_off = (linux_off_t)linux_reclen;
802 			idb.d_reclen = (u_short)bdp->d_namlen;
803 		} else {
804 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
805 				compat_offseterr(vp, "linux_getdents");
806 				error = EINVAL;
807 				goto out;
808 			}
809 			idb.d_off = (linux_off_t)off;
810 			idb.d_reclen = (u_short)linux_reclen;
811 		}
812 		strcpy(idb.d_name, bdp->d_name);
813 		if ((error = copyout((void *)&idb, outp, linux_reclen)))
814 			goto out;
815 		/* advance past this real entry */
816 		inp += reclen;
817 		if (cookie)
818 			off = *cookie++; /* each entry points to itself */
819 		else
820 			off += reclen;
821 		/* advance output past Linux-shaped entry */
822 		outp += linux_reclen;
823 		resid -= linux_reclen;
824 		if (oldcall)
825 			break;
826 	}
827 
828 	/* if we squished out the whole block, try again */
829 	if (outp == (void *)SCARG(uap, dent))
830 		goto again;
831 	fp->f_offset = off;	/* update the vnode offset */
832 
833 	if (oldcall)
834 		nbytes = resid + linux_reclen;
835 
836 eof:
837 	*retval = nbytes - resid;
838 out:
839 	VOP_UNLOCK(vp, 0);
840 	if (cookiebuf)
841 		free(cookiebuf, M_TEMP);
842 	free(tbuf, M_TEMP);
843 out1:
844 	FILE_UNUSE(fp, l);
845 	return error;
846 }
847 
848 /*
849  * Even when just using registers to pass arguments to syscalls you can
850  * have 5 of them on the i386. So this newer version of select() does
851  * this.
852  */
853 int
854 linux_sys_select(struct lwp *l, void *v, register_t *retval)
855 {
856 	struct linux_sys_select_args /* {
857 		syscallarg(int) nfds;
858 		syscallarg(fd_set *) readfds;
859 		syscallarg(fd_set *) writefds;
860 		syscallarg(fd_set *) exceptfds;
861 		syscallarg(struct timeval *) timeout;
862 	} */ *uap = v;
863 
864 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
865 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
866 }
867 
868 /*
869  * Common code for the old and new versions of select(). A couple of
870  * things are important:
871  * 1) return the amount of time left in the 'timeout' parameter
872  * 2) select never returns ERESTART on Linux, always return EINTR
873  */
874 int
875 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
876 	struct lwp *l;
877 	register_t *retval;
878 	int nfds;
879 	fd_set *readfds, *writefds, *exceptfds;
880 	struct timeval *timeout;
881 {
882 	struct timeval tv0, tv1, utv, *tv = NULL;
883 	int error;
884 
885 	/*
886 	 * Store current time for computation of the amount of
887 	 * time left.
888 	 */
889 	if (timeout) {
890 		if ((error = copyin(timeout, &utv, sizeof(utv))))
891 			return error;
892 		if (itimerfix(&utv)) {
893 			/*
894 			 * The timeval was invalid.  Convert it to something
895 			 * valid that will act as it does under Linux.
896 			 */
897 			utv.tv_sec += utv.tv_usec / 1000000;
898 			utv.tv_usec %= 1000000;
899 			if (utv.tv_usec < 0) {
900 				utv.tv_sec -= 1;
901 				utv.tv_usec += 1000000;
902 			}
903 			if (utv.tv_sec < 0)
904 				timerclear(&utv);
905 		}
906 		tv = &utv;
907 		microtime(&tv0);
908 	}
909 
910 	error = selcommon(l, retval, nfds, readfds, writefds, exceptfds,
911 	    tv, NULL);
912 
913 	if (error) {
914 		/*
915 		 * See fs/select.c in the Linux kernel.  Without this,
916 		 * Maelstrom doesn't work.
917 		 */
918 		if (error == ERESTART)
919 			error = EINTR;
920 		return error;
921 	}
922 
923 	if (timeout) {
924 		if (*retval) {
925 			/*
926 			 * Compute how much time was left of the timeout,
927 			 * by subtracting the current time and the time
928 			 * before we started the call, and subtracting
929 			 * that result from the user-supplied value.
930 			 */
931 			microtime(&tv1);
932 			timersub(&tv1, &tv0, &tv1);
933 			timersub(&utv, &tv1, &utv);
934 			if (utv.tv_sec < 0)
935 				timerclear(&utv);
936 		} else
937 			timerclear(&utv);
938 		if ((error = copyout(&utv, timeout, sizeof(utv))))
939 			return error;
940 	}
941 
942 	return 0;
943 }
944 
945 /*
946  * Get the process group of a certain process. Look it up
947  * and return the value.
948  */
949 int
950 linux_sys_getpgid(struct lwp *l, void *v, register_t *retval)
951 {
952 	struct linux_sys_getpgid_args /* {
953 		syscallarg(int) pid;
954 	} */ *uap = v;
955 	struct proc *p = l->l_proc;
956 	struct proc *targp;
957 
958 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
959 		if ((targp = pfind(SCARG(uap, pid))) == 0)
960 			return ESRCH;
961 	}
962 	else
963 		targp = p;
964 
965 	retval[0] = targp->p_pgid;
966 	return 0;
967 }
968 
969 /*
970  * Set the 'personality' (emulation mode) for the current process. Only
971  * accept the Linux personality here (0). This call is needed because
972  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
973  * ELF binaries run in Linux mode, not SVR4 mode.
974  */
975 int
976 linux_sys_personality(struct lwp *l, void *v, register_t *retval)
977 {
978 	struct linux_sys_personality_args /* {
979 		syscallarg(int) per;
980 	} */ *uap = v;
981 
982 	if (SCARG(uap, per) != 0)
983 		return EINVAL;
984 	retval[0] = 0;
985 	return 0;
986 }
987 #endif /* !COMPAT_LINUX32 */
988 
989 #if defined(__i386__) || defined(__m68k__) || defined(COMPAT_LINUX32)
990 /*
991  * The calls are here because of type conversions.
992  */
993 #ifndef COMPAT_LINUX32
994 int
995 linux_sys_setreuid16(struct lwp *l, void *v, register_t *retval)
996 {
997 	struct linux_sys_setreuid16_args /* {
998 		syscallarg(int) ruid;
999 		syscallarg(int) euid;
1000 	} */ *uap = v;
1001 	struct sys_setreuid_args bsa;
1002 
1003 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1004 		(uid_t)-1 : SCARG(uap, ruid);
1005 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1006 		(uid_t)-1 : SCARG(uap, euid);
1007 
1008 	return sys_setreuid(l, &bsa, retval);
1009 }
1010 
1011 int
1012 linux_sys_setregid16(struct lwp *l, void *v, register_t *retval)
1013 {
1014 	struct linux_sys_setregid16_args /* {
1015 		syscallarg(int) rgid;
1016 		syscallarg(int) egid;
1017 	} */ *uap = v;
1018 	struct sys_setregid_args bsa;
1019 
1020 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1021 		(uid_t)-1 : SCARG(uap, rgid);
1022 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1023 		(uid_t)-1 : SCARG(uap, egid);
1024 
1025 	return sys_setregid(l, &bsa, retval);
1026 }
1027 
1028 int
1029 linux_sys_setresuid16(struct lwp *l, void *v, register_t *retval)
1030 {
1031 	struct linux_sys_setresuid16_args /* {
1032 		syscallarg(uid_t) ruid;
1033 		syscallarg(uid_t) euid;
1034 		syscallarg(uid_t) suid;
1035 	} */ *uap = v;
1036 	struct linux_sys_setresuid16_args lsa;
1037 
1038 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1039 		(uid_t)-1 : SCARG(uap, ruid);
1040 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1041 		(uid_t)-1 : SCARG(uap, euid);
1042 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1043 		(uid_t)-1 : SCARG(uap, suid);
1044 
1045 	return linux_sys_setresuid(l, &lsa, retval);
1046 }
1047 
1048 int
1049 linux_sys_setresgid16(struct lwp *l, void *v, register_t *retval)
1050 {
1051 	struct linux_sys_setresgid16_args /* {
1052 		syscallarg(gid_t) rgid;
1053 		syscallarg(gid_t) egid;
1054 		syscallarg(gid_t) sgid;
1055 	} */ *uap = v;
1056 	struct linux_sys_setresgid16_args lsa;
1057 
1058 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1059 		(gid_t)-1 : SCARG(uap, rgid);
1060 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1061 		(gid_t)-1 : SCARG(uap, egid);
1062 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1063 		(gid_t)-1 : SCARG(uap, sgid);
1064 
1065 	return linux_sys_setresgid(l, &lsa, retval);
1066 }
1067 #endif /* COMPAT_LINUX32 */
1068 
1069 int
1070 linux_sys_getgroups16(struct lwp *l, void *v, register_t *retval)
1071 {
1072 	struct linux_sys_getgroups16_args /* {
1073 		syscallarg(int) gidsetsize;
1074 		syscallarg(linux_gid_t *) gidset;
1075 	} */ *uap = v;
1076 	linux_gid_t lset[16];
1077 	linux_gid_t *gidset;
1078 	unsigned int ngrps;
1079 	int i, n, j;
1080 	int error;
1081 
1082 	ngrps = kauth_cred_ngroups(l->l_cred);
1083 	*retval = ngrps;
1084 	if (SCARG(uap, gidsetsize) == 0)
1085 		return 0;
1086 	if (SCARG(uap, gidsetsize) < ngrps)
1087 		return EINVAL;
1088 
1089 	gidset = SCARG(uap, gidset);
1090 	for (i = 0; i < (n = ngrps); i += n, gidset += n) {
1091 		n -= i;
1092 		if (n > __arraycount(lset))
1093 			n = __arraycount(lset);
1094 		for (j = 0; j < n; j++)
1095 			lset[j] = kauth_cred_group(l->l_cred, i + j);
1096 		error = copyout(lset, gidset, n * sizeof(lset[0]));
1097 		if (error != 0)
1098 			return error;
1099 	}
1100 
1101 	return 0;
1102 }
1103 
1104 /*
1105  * It is very unlikly that any problem using 16bit groups is written
1106  * to allow for more than 16 of them, so don't bother trying to
1107  * support that.
1108  */
1109 #define COMPAT_NGROUPS16 16
1110 
1111 int
1112 linux_sys_setgroups16(struct lwp *l, void *v, register_t *retval)
1113 {
1114 	struct linux_sys_setgroups16_args /* {
1115 		syscallarg(int) gidsetsize;
1116 		syscallarg(linux_gid_t *) gidset;
1117 	} */ *uap = v;
1118 	linux_gid_t lset[COMPAT_NGROUPS16];
1119 	kauth_cred_t ncred;
1120 	int error;
1121 	gid_t grbuf[COMPAT_NGROUPS16];
1122 	unsigned int i, ngroups = SCARG(uap, gidsetsize);
1123 
1124 	if (ngroups > COMPAT_NGROUPS16)
1125 		return EINVAL;
1126 	error = copyin(SCARG(uap, gidset), lset, ngroups);
1127 	if (error != 0)
1128 		return error;
1129 
1130 	for (i = 0; i < ngroups; i++)
1131 		grbuf[i] = lset[i];
1132 
1133 	ncred = kauth_cred_alloc();
1134 	error = kauth_cred_setgroups(ncred, grbuf, SCARG(uap, gidsetsize),
1135 	    -1, UIO_SYSSPACE);
1136 	if (error != 0) {
1137 		kauth_cred_free(ncred);
1138 		return error;
1139 	}
1140 
1141 	return kauth_proc_setgroups(l, ncred);
1142 }
1143 
1144 #endif /* __i386__ || __m68k__ || COMPAT_LINUX32 */
1145 
1146 #ifndef COMPAT_LINUX32
1147 /*
1148  * We have nonexistent fsuid equal to uid.
1149  * If modification is requested, refuse.
1150  */
1151 int
1152 linux_sys_setfsuid(struct lwp *l, void *v, register_t *retval)
1153 {
1154 	 struct linux_sys_setfsuid_args /* {
1155 		 syscallarg(uid_t) uid;
1156 	 } */ *uap = v;
1157 	 uid_t uid;
1158 
1159 	 uid = SCARG(uap, uid);
1160 	 if (kauth_cred_getuid(l->l_cred) != uid)
1161 		 return sys_nosys(l, v, retval);
1162 	 else
1163 		 return (0);
1164 }
1165 
1166 /* XXX XXX XXX */
1167 # ifndef alpha
1168 int
1169 linux_sys_getfsuid(struct lwp *l, void *v, register_t *retval)
1170 {
1171 	return sys_getuid(l, v, retval);
1172 }
1173 # endif
1174 
1175 int
1176 linux_sys_setresuid(struct lwp *l, void *v, register_t *retval)
1177 {
1178 	struct linux_sys_setresuid_args /* {
1179 		syscallarg(uid_t) ruid;
1180 		syscallarg(uid_t) euid;
1181 		syscallarg(uid_t) suid;
1182 	} */ *uap = v;
1183 
1184 	/*
1185 	 * Note: These checks are a little different than the NetBSD
1186 	 * setreuid(2) call performs.  This precisely follows the
1187 	 * behavior of the Linux kernel.
1188 	 */
1189 
1190 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1191 			    SCARG(uap, suid),
1192 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1193 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1194 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1195 }
1196 
1197 int
1198 linux_sys_getresuid(struct lwp *l, void *v, register_t *retval)
1199 {
1200 	struct linux_sys_getresuid_args /* {
1201 		syscallarg(uid_t *) ruid;
1202 		syscallarg(uid_t *) euid;
1203 		syscallarg(uid_t *) suid;
1204 	} */ *uap = v;
1205 	kauth_cred_t pc = l->l_cred;
1206 	int error;
1207 	uid_t uid;
1208 
1209 	/*
1210 	 * Linux copies these values out to userspace like so:
1211 	 *
1212 	 *	1. Copy out ruid.
1213 	 *	2. If that succeeds, copy out euid.
1214 	 *	3. If both of those succeed, copy out suid.
1215 	 */
1216 	uid = kauth_cred_getuid(pc);
1217 	if ((error = copyout(&uid, SCARG(uap, ruid), sizeof(uid_t))) != 0)
1218 		return (error);
1219 
1220 	uid = kauth_cred_geteuid(pc);
1221 	if ((error = copyout(&uid, SCARG(uap, euid), sizeof(uid_t))) != 0)
1222 		return (error);
1223 
1224 	uid = kauth_cred_getsvuid(pc);
1225 
1226 	return (copyout(&uid, SCARG(uap, suid), sizeof(uid_t)));
1227 }
1228 
1229 int
1230 linux_sys_ptrace(struct lwp *l, void *v, register_t *retval)
1231 {
1232 #if defined(PTRACE) || defined(_LKM)
1233 	struct linux_sys_ptrace_args /* {
1234 		i386, m68k, powerpc: T=int
1235 		alpha, amd64: T=long
1236 		syscallarg(T) request;
1237 		syscallarg(T) pid;
1238 		syscallarg(T) addr;
1239 		syscallarg(T) data;
1240 	} */ *uap = v;
1241 	const int *ptr;
1242 	int request;
1243 	int error;
1244 #ifdef _LKM
1245 #define sys_ptrace (*sysent[SYS_ptrace].sy_call)
1246 #endif
1247 
1248 	ptr = linux_ptrace_request_map;
1249 	request = SCARG(uap, request);
1250 	while (*ptr != -1)
1251 		if (*ptr++ == request) {
1252 			struct sys_ptrace_args pta;
1253 
1254 			SCARG(&pta, req) = *ptr;
1255 			SCARG(&pta, pid) = SCARG(uap, pid);
1256 			SCARG(&pta, addr) = (void *)SCARG(uap, addr);
1257 			SCARG(&pta, data) = SCARG(uap, data);
1258 
1259 			/*
1260 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1261 			 * to continue where the process left off previously.
1262 			 * The same thing is achieved by addr == (void *) 1
1263 			 * on NetBSD, so rewrite 'addr' appropriately.
1264 			 */
1265 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1266 				SCARG(&pta, addr) = (void *) 1;
1267 
1268 			error = sys_ptrace(l, &pta, retval);
1269 			if (error)
1270 				return error;
1271 			switch (request) {
1272 			case LINUX_PTRACE_PEEKTEXT:
1273 			case LINUX_PTRACE_PEEKDATA:
1274 				error = copyout (retval,
1275 				    (void *)SCARG(uap, data),
1276 				    sizeof *retval);
1277 				*retval = SCARG(uap, data);
1278 				break;
1279 			default:
1280 				break;
1281 			}
1282 			return error;
1283 		}
1284 		else
1285 			ptr++;
1286 
1287 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1288 #else
1289 	return ENOSYS;
1290 #endif /* PTRACE || _LKM */
1291 }
1292 
1293 int
1294 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1295 {
1296 	struct linux_sys_reboot_args /* {
1297 		syscallarg(int) magic1;
1298 		syscallarg(int) magic2;
1299 		syscallarg(int) cmd;
1300 		syscallarg(void *) arg;
1301 	} */ *uap = v;
1302 	struct sys_reboot_args /* {
1303 		syscallarg(int) opt;
1304 		syscallarg(char *) bootstr;
1305 	} */ sra;
1306 	int error;
1307 
1308 	if ((error = kauth_authorize_system(l->l_cred,
1309 	    KAUTH_SYSTEM_REBOOT, 0, NULL, NULL, NULL)) != 0)
1310 		return(error);
1311 
1312 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1313 		return(EINVAL);
1314 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1315 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1316 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1317 		return(EINVAL);
1318 
1319 	switch (SCARG(uap, cmd)) {
1320 	case LINUX_REBOOT_CMD_RESTART:
1321 		SCARG(&sra, opt) = RB_AUTOBOOT;
1322 		break;
1323 	case LINUX_REBOOT_CMD_HALT:
1324 		SCARG(&sra, opt) = RB_HALT;
1325 		break;
1326 	case LINUX_REBOOT_CMD_POWER_OFF:
1327 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1328 		break;
1329 	case LINUX_REBOOT_CMD_RESTART2:
1330 		/* Reboot with an argument. */
1331 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1332 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1333 		break;
1334 	case LINUX_REBOOT_CMD_CAD_ON:
1335 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1336 	case LINUX_REBOOT_CMD_CAD_OFF:
1337 		return(0);
1338 	default:
1339 		return(EINVAL);
1340 	}
1341 
1342 	return(sys_reboot(l, &sra, retval));
1343 }
1344 
1345 /*
1346  * Copy of compat_12_sys_swapon().
1347  */
1348 int
1349 linux_sys_swapon(struct lwp *l, void *v, register_t *retval)
1350 {
1351 	struct sys_swapctl_args ua;
1352 	struct linux_sys_swapon_args /* {
1353 		syscallarg(const char *) name;
1354 	} */ *uap = v;
1355 
1356 	SCARG(&ua, cmd) = SWAP_ON;
1357 	SCARG(&ua, arg) = (void *)__UNCONST(SCARG(uap, name));
1358 	SCARG(&ua, misc) = 0;	/* priority */
1359 	return (sys_swapctl(l, &ua, retval));
1360 }
1361 
1362 /*
1363  * Stop swapping to the file or block device specified by path.
1364  */
1365 int
1366 linux_sys_swapoff(struct lwp *l, void *v, register_t *retval)
1367 {
1368 	struct sys_swapctl_args ua;
1369 	struct linux_sys_swapoff_args /* {
1370 		syscallarg(const char *) path;
1371 	} */ *uap = v;
1372 
1373 	SCARG(&ua, cmd) = SWAP_OFF;
1374 	SCARG(&ua, arg) = __UNCONST(SCARG(uap, path)); /*XXXUNCONST*/
1375 	return (sys_swapctl(l, &ua, retval));
1376 }
1377 
1378 /*
1379  * Copy of compat_09_sys_setdomainname()
1380  */
1381 /* ARGSUSED */
1382 int
1383 linux_sys_setdomainname(struct lwp *l, void *v, register_t *retval)
1384 {
1385 	struct linux_sys_setdomainname_args /* {
1386 		syscallarg(char *) domainname;
1387 		syscallarg(int) len;
1388 	} */ *uap = v;
1389 	int name[2];
1390 
1391 	name[0] = CTL_KERN;
1392 	name[1] = KERN_DOMAINNAME;
1393 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1394 			    SCARG(uap, len), l));
1395 }
1396 
1397 /*
1398  * sysinfo()
1399  */
1400 /* ARGSUSED */
1401 int
1402 linux_sys_sysinfo(struct lwp *l, void *v, register_t *retval)
1403 {
1404 	struct linux_sys_sysinfo_args /* {
1405 		syscallarg(struct linux_sysinfo *) arg;
1406 	} */ *uap = v;
1407 	struct linux_sysinfo si;
1408 	struct loadavg *la;
1409 
1410 	si.uptime = time_uptime;
1411 	la = &averunnable;
1412 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1413 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1414 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1415 	si.totalram = ctob((u_long)physmem);
1416 	si.freeram = (u_long)uvmexp.free * uvmexp.pagesize;
1417 	si.sharedram = 0;	/* XXX */
1418 	si.bufferram = (u_long)uvmexp.filepages * uvmexp.pagesize;
1419 	si.totalswap = (u_long)uvmexp.swpages * uvmexp.pagesize;
1420 	si.freeswap =
1421 	    (u_long)(uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1422 	si.procs = nprocs;
1423 
1424 	/* The following are only present in newer Linux kernels. */
1425 	si.totalbig = 0;
1426 	si.freebig = 0;
1427 	si.mem_unit = 1;
1428 
1429 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1430 }
1431 
1432 int
1433 linux_sys_getrlimit(struct lwp *l, void *v, register_t *retval)
1434 {
1435 	struct linux_sys_getrlimit_args /* {
1436 		syscallarg(int) which;
1437 # ifdef LINUX_LARGEFILE64
1438 		syscallarg(struct rlimit *) rlp;
1439 # else
1440 		syscallarg(struct orlimit *) rlp;
1441 # endif
1442 	} */ *uap = v;
1443 # ifdef LINUX_LARGEFILE64
1444 	struct rlimit orl;
1445 # else
1446 	struct orlimit orl;
1447 # endif
1448 	int which;
1449 
1450 	which = linux_to_bsd_limit(SCARG(uap, which));
1451 	if (which < 0)
1452 		return -which;
1453 
1454 	bsd_to_linux_rlimit(&orl, &l->l_proc->p_rlimit[which]);
1455 
1456 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1457 }
1458 
1459 int
1460 linux_sys_setrlimit(struct lwp *l, void *v, register_t *retval)
1461 {
1462 	struct linux_sys_setrlimit_args /* {
1463 		syscallarg(int) which;
1464 # ifdef LINUX_LARGEFILE64
1465 		syscallarg(struct rlimit *) rlp;
1466 # else
1467 		syscallarg(struct orlimit *) rlp;
1468 # endif
1469 	} */ *uap = v;
1470 	struct rlimit rl;
1471 # ifdef LINUX_LARGEFILE64
1472 	struct rlimit orl;
1473 # else
1474 	struct orlimit orl;
1475 # endif
1476 	int error;
1477 	int which;
1478 
1479 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1480 		return error;
1481 
1482 	which = linux_to_bsd_limit(SCARG(uap, which));
1483 	if (which < 0)
1484 		return -which;
1485 
1486 	linux_to_bsd_rlimit(&rl, &orl);
1487 	return dosetrlimit(l, l->l_proc, which, &rl);
1488 }
1489 
1490 # if !defined(__mips__) && !defined(__amd64__)
1491 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1492 int
1493 linux_sys_ugetrlimit(struct lwp *l, void *v, register_t *retval)
1494 {
1495 	return linux_sys_getrlimit(l, v, retval);
1496 }
1497 # endif
1498 
1499 /*
1500  * This gets called for unsupported syscalls. The difference to sys_nosys()
1501  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1502  * This is the way Linux does it and glibc depends on this behaviour.
1503  */
1504 int
1505 linux_sys_nosys(struct lwp *l, void *v,
1506     register_t *retval)
1507 {
1508 	return (ENOSYS);
1509 }
1510 
1511 int
1512 linux_sys_getpriority(struct lwp *l, void *v, register_t *retval)
1513 {
1514         struct linux_sys_getpriority_args /* {
1515                 syscallarg(int) which;
1516                 syscallarg(int) who;
1517         } */ *uap = v;
1518         struct sys_getpriority_args bsa;
1519         int error;
1520 
1521         SCARG(&bsa, which) = SCARG(uap, which);
1522         SCARG(&bsa, who) = SCARG(uap, who);
1523 
1524         if ((error = sys_getpriority(l, &bsa, retval)))
1525                 return error;
1526 
1527         *retval = NZERO - *retval;
1528 
1529         return 0;
1530 }
1531 
1532 #endif /* !COMPAT_LINUX32 */
1533