xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 21e37cc72a480a47828990a439cde7ac9ffaf0c6)
1 /*	$NetBSD: linux_misc.c,v 1.125 2004/04/21 02:37:20 christos 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 1.125 2004/04/21 02:37:20 christos Exp $");
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/namei.h>
72 #include <sys/proc.h>
73 #include <sys/dirent.h>
74 #include <sys/file.h>
75 #include <sys/stat.h>
76 #include <sys/filedesc.h>
77 #include <sys/ioctl.h>
78 #include <sys/kernel.h>
79 #include <sys/malloc.h>
80 #include <sys/mbuf.h>
81 #include <sys/mman.h>
82 #include <sys/mount.h>
83 #include <sys/reboot.h>
84 #include <sys/resource.h>
85 #include <sys/resourcevar.h>
86 #include <sys/signal.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/times.h>
91 #include <sys/vnode.h>
92 #include <sys/uio.h>
93 #include <sys/wait.h>
94 #include <sys/utsname.h>
95 #include <sys/unistd.h>
96 #include <sys/swap.h>		/* for SWAP_ON */
97 #include <sys/sysctl.h>		/* for KERN_DOMAINNAME */
98 
99 #include <sys/ptrace.h>
100 #include <machine/ptrace.h>
101 
102 #include <sys/sa.h>
103 #include <sys/syscallargs.h>
104 
105 #include <compat/linux/common/linux_types.h>
106 #include <compat/linux/common/linux_signal.h>
107 
108 #include <compat/linux/linux_syscallargs.h>
109 
110 #include <compat/linux/common/linux_fcntl.h>
111 #include <compat/linux/common/linux_mmap.h>
112 #include <compat/linux/common/linux_dirent.h>
113 #include <compat/linux/common/linux_util.h>
114 #include <compat/linux/common/linux_misc.h>
115 #include <compat/linux/common/linux_ptrace.h>
116 #include <compat/linux/common/linux_reboot.h>
117 #include <compat/linux/common/linux_emuldata.h>
118 
119 const int linux_ptrace_request_map[] = {
120 	LINUX_PTRACE_TRACEME,	PT_TRACE_ME,
121 	LINUX_PTRACE_PEEKTEXT,	PT_READ_I,
122 	LINUX_PTRACE_PEEKDATA,	PT_READ_D,
123 	LINUX_PTRACE_POKETEXT,	PT_WRITE_I,
124 	LINUX_PTRACE_POKEDATA,	PT_WRITE_D,
125 	LINUX_PTRACE_CONT,	PT_CONTINUE,
126 	LINUX_PTRACE_KILL,	PT_KILL,
127 	LINUX_PTRACE_ATTACH,	PT_ATTACH,
128 	LINUX_PTRACE_DETACH,	PT_DETACH,
129 #ifdef PT_STEP
130 	LINUX_PTRACE_SINGLESTEP,	PT_STEP,
131 #endif
132 	-1
133 };
134 
135 static const struct mnttypes {
136 	char *bsd;
137 	int linux;
138 } fstypes[] = {
139 	{ MOUNT_FFS,		LINUX_DEFAULT_SUPER_MAGIC	},
140 	{ MOUNT_NFS,		LINUX_NFS_SUPER_MAGIC 		},
141 	{ MOUNT_MFS,		LINUX_DEFAULT_SUPER_MAGIC	},
142 	{ MOUNT_MSDOS,		LINUX_MSDOS_SUPER_MAGIC		},
143 	{ MOUNT_LFS,		LINUX_DEFAULT_SUPER_MAGIC	},
144 	{ MOUNT_FDESC,		LINUX_DEFAULT_SUPER_MAGIC	},
145 	{ MOUNT_PORTAL,		LINUX_DEFAULT_SUPER_MAGIC	},
146 	{ MOUNT_NULL,		LINUX_DEFAULT_SUPER_MAGIC	},
147 	{ MOUNT_OVERLAY,	LINUX_DEFAULT_SUPER_MAGIC	},
148 	{ MOUNT_UMAP,		LINUX_DEFAULT_SUPER_MAGIC	},
149 	{ MOUNT_KERNFS,		LINUX_DEFAULT_SUPER_MAGIC	},
150 	{ MOUNT_PROCFS,		LINUX_PROC_SUPER_MAGIC		},
151 	{ MOUNT_AFS,		LINUX_DEFAULT_SUPER_MAGIC	},
152 	{ MOUNT_CD9660,		LINUX_ISOFS_SUPER_MAGIC		},
153 	{ MOUNT_UNION,		LINUX_DEFAULT_SUPER_MAGIC	},
154 	{ MOUNT_ADOSFS,		LINUX_ADFS_SUPER_MAGIC		},
155 	{ MOUNT_EXT2FS,		LINUX_EXT2_SUPER_MAGIC		},
156 	{ MOUNT_CFS,		LINUX_DEFAULT_SUPER_MAGIC	},
157 	{ MOUNT_CODA,		LINUX_CODA_SUPER_MAGIC		},
158 	{ MOUNT_FILECORE,	LINUX_DEFAULT_SUPER_MAGIC	},
159 	{ MOUNT_NTFS,		LINUX_DEFAULT_SUPER_MAGIC	},
160 	{ MOUNT_SMBFS,		LINUX_SMB_SUPER_MAGIC		}
161 };
162 #define FSTYPESSIZE (sizeof(fstypes) / sizeof(fstypes[0]))
163 
164 #ifdef DEBUG_LINUX
165 #define DPRINTF(a)	uprintf a
166 #else
167 #define DPRINTF(a)
168 #endif
169 
170 /* Local linux_misc.c functions: */
171 static void bsd_to_linux_statfs __P((const struct statvfs *,
172     struct linux_statfs *));
173 static int linux_to_bsd_limit __P((int));
174 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *,
175     const struct linux_sys_mmap_args *));
176 
177 /*
178  * The information on a terminated (or stopped) process needs
179  * to be converted in order for Linux binaries to get a valid signal
180  * number out of it.
181  */
182 void
183 bsd_to_linux_wstat(st)
184 	int *st;
185 {
186 
187 	int sig;
188 
189 	if (WIFSIGNALED(*st)) {
190 		sig = WTERMSIG(*st);
191 		if (sig >= 0 && sig < NSIG)
192 			*st= (*st& ~0177) | native_to_linux_signo[sig];
193 	} else if (WIFSTOPPED(*st)) {
194 		sig = WSTOPSIG(*st);
195 		if (sig >= 0 && sig < NSIG)
196 			*st = (*st & ~0xff00) |
197 			    (native_to_linux_signo[sig] << 8);
198 	}
199 }
200 
201 /*
202  * This is very much the same as waitpid()
203  */
204 int
205 linux_sys_wait4(l, v, retval)
206 	struct lwp *l;
207 	void *v;
208 	register_t *retval;
209 {
210 	struct linux_sys_wait4_args /* {
211 		syscallarg(int) pid;
212 		syscallarg(int *) status;
213 		syscallarg(int) options;
214 		syscallarg(struct rusage *) rusage;
215 	} */ *uap = v;
216 	struct proc *p = l->l_proc;
217 	struct sys_wait4_args w4a;
218 	int error, *status, tstat, options, linux_options;
219 	caddr_t sg;
220 
221 	if (SCARG(uap, status) != NULL) {
222 		sg = stackgap_init(p, 0);
223 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
224 	} else
225 		status = NULL;
226 
227 	linux_options = SCARG(uap, options);
228 	options = 0;
229 	if (linux_options &
230 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
231 	      LINUX_WAIT4_WCLONE))
232 		return (EINVAL);
233 
234 	if (linux_options & LINUX_WAIT4_WNOHANG)
235 		options |= WNOHANG;
236 	if (linux_options & LINUX_WAIT4_WUNTRACED)
237 		options |= WUNTRACED;
238 	if (linux_options & LINUX_WAIT4_WALL)
239 		options |= WALLSIG;
240 	if (linux_options & LINUX_WAIT4_WCLONE)
241 		options |= WALTSIG;
242 
243 	SCARG(&w4a, pid) = SCARG(uap, pid);
244 	SCARG(&w4a, status) = status;
245 	SCARG(&w4a, options) = options;
246 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
247 
248 	if ((error = sys_wait4(l, &w4a, retval)))
249 		return error;
250 
251 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
252 
253 	if (status != NULL) {
254 		if ((error = copyin(status, &tstat, sizeof tstat)))
255 			return error;
256 
257 		bsd_to_linux_wstat(&tstat);
258 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
259 	}
260 
261 	return 0;
262 }
263 
264 /*
265  * Linux brk(2). The check if the new address is >= the old one is
266  * done in the kernel in Linux. NetBSD does it in the library.
267  */
268 int
269 linux_sys_brk(l, v, retval)
270 	struct lwp *l;
271 	void *v;
272 	register_t *retval;
273 {
274 	struct linux_sys_brk_args /* {
275 		syscallarg(char *) nsize;
276 	} */ *uap = v;
277 	struct proc *p = l->l_proc;
278 	char *nbrk = SCARG(uap, nsize);
279 	struct sys_obreak_args oba;
280 	struct vmspace *vm = p->p_vmspace;
281 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
282 
283 	SCARG(&oba, nsize) = nbrk;
284 
285 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
286 		ed->p_break = (char*)nbrk;
287 	else
288 		nbrk = ed->p_break;
289 
290 	retval[0] = (register_t)nbrk;
291 
292 	return 0;
293 }
294 
295 /*
296  * Convert BSD statfs structure to Linux statfs structure.
297  * The Linux structure has less fields, and it also wants
298  * the length of a name in a dir entry in a field, which
299  * we fake (probably the wrong way).
300  */
301 static void
302 bsd_to_linux_statfs(bsp, lsp)
303 	const struct statvfs *bsp;
304 	struct linux_statfs *lsp;
305 {
306 	int i;
307 
308 	for (i = 0; i < FSTYPESSIZE; i++)
309 		if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0)
310 			break;
311 
312 	if (i == FSTYPESSIZE) {
313 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
314 		    bsp->f_fstypename));
315 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
316 	} else {
317 		lsp->l_ftype = fstypes[i].linux;
318 	}
319 
320 	lsp->l_fbsize = bsp->f_bsize;
321 	lsp->l_fblocks = bsp->f_blocks;
322 	lsp->l_fbfree = bsp->f_bfree;
323 	lsp->l_fbavail = bsp->f_bavail;
324 	lsp->l_ffiles = bsp->f_files;
325 	lsp->l_fffree = bsp->f_ffree;
326 	/* Linux sets the fsid to 0..., we don't */
327 	lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0];
328 	lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1];
329 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
330 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
331 }
332 
333 /*
334  * Implement the fs stat functions. Straightforward.
335  */
336 int
337 linux_sys_statfs(l, v, retval)
338 	struct lwp *l;
339 	void *v;
340 	register_t *retval;
341 {
342 	struct linux_sys_statfs_args /* {
343 		syscallarg(const char *) path;
344 		syscallarg(struct linux_statfs *) sp;
345 	} */ *uap = v;
346 	struct proc *p = l->l_proc;
347 	struct statvfs btmp, *bsp;
348 	struct linux_statfs ltmp;
349 	struct sys_statvfs1_args bsa;
350 	caddr_t sg;
351 	int error;
352 
353 	sg = stackgap_init(p, 0);
354 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
355 
356 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
357 
358 	SCARG(&bsa, path) = SCARG(uap, path);
359 	SCARG(&bsa, buf) = bsp;
360 	SCARG(&bsa, flags) = ST_WAIT;
361 
362 	if ((error = sys_statvfs1(l, &bsa, retval)))
363 		return error;
364 
365 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
366 		return error;
367 
368 	bsd_to_linux_statfs(&btmp, &ltmp);
369 
370 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
371 }
372 
373 int
374 linux_sys_fstatfs(l, v, retval)
375 	struct lwp *l;
376 	void *v;
377 	register_t *retval;
378 {
379 	struct linux_sys_fstatfs_args /* {
380 		syscallarg(int) fd;
381 		syscallarg(struct linux_statfs *) sp;
382 	} */ *uap = v;
383 	struct proc *p = l->l_proc;
384 	struct statvfs btmp, *bsp;
385 	struct linux_statfs ltmp;
386 	struct sys_fstatvfs1_args bsa;
387 	caddr_t sg;
388 	int error;
389 
390 	sg = stackgap_init(p, 0);
391 	bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs));
392 
393 	SCARG(&bsa, fd) = SCARG(uap, fd);
394 	SCARG(&bsa, buf) = bsp;
395 	SCARG(&bsa, flags) = ST_WAIT;
396 
397 	if ((error = sys_fstatvfs1(l, &bsa, retval)))
398 		return error;
399 
400 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
401 		return error;
402 
403 	bsd_to_linux_statfs(&btmp, &ltmp);
404 
405 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
406 }
407 
408 /*
409  * uname(). Just copy the info from the various strings stored in the
410  * kernel, and put it in the Linux utsname structure. That structure
411  * is almost the same as the NetBSD one, only it has fields 65 characters
412  * long, and an extra domainname field.
413  */
414 int
415 linux_sys_uname(l, v, retval)
416 	struct lwp *l;
417 	void *v;
418 	register_t *retval;
419 {
420 	struct linux_sys_uname_args /* {
421 		syscallarg(struct linux_utsname *) up;
422 	} */ *uap = v;
423 	struct linux_utsname luts;
424 
425 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
426 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
427 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
428 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
429 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
430 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
431 
432 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
433 }
434 
435 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
436 /* Used indirectly on: arm, i386, m68k */
437 
438 /*
439  * New type Linux mmap call.
440  * Only called directly on machines with >= 6 free regs.
441  */
442 int
443 linux_sys_mmap(l, v, retval)
444 	struct lwp *l;
445 	void *v;
446 	register_t *retval;
447 {
448 	struct linux_sys_mmap_args /* {
449 		syscallarg(unsigned long) addr;
450 		syscallarg(size_t) len;
451 		syscallarg(int) prot;
452 		syscallarg(int) flags;
453 		syscallarg(int) fd;
454 		syscallarg(linux_off_t) offset;
455 	} */ *uap = v;
456 	struct sys_mmap_args cma;
457 
458 	if (SCARG(uap, offset) & PAGE_MASK)
459 		return EINVAL;
460 
461 	linux_to_bsd_mmap_args(&cma, uap);
462 	SCARG(&cma, pos) = (off_t)SCARG(uap, offset);
463 
464 	return sys_mmap(l, &cma, retval);
465 }
466 
467 /*
468  * Guts of most architectures' mmap64() implementations.  This shares
469  * its list of arguments with linux_sys_mmap().
470  *
471  * The difference in linux_sys_mmap2() is that "offset" is actually
472  * (offset / pagesize), not an absolute byte count.  This translation
473  * to pagesize offsets is done inside glibc between the mmap64() call
474  * point, and the actual syscall.
475  */
476 int
477 linux_sys_mmap2(l, v, retval)
478 	struct lwp *l;
479 	void *v;
480 	register_t *retval;
481 {
482 	struct linux_sys_mmap2_args /* {
483 		syscallarg(unsigned long) addr;
484 		syscallarg(size_t) len;
485 		syscallarg(int) prot;
486 		syscallarg(int) flags;
487 		syscallarg(int) fd;
488 		syscallarg(linux_off_t) offset;
489 	} */ *uap = v;
490 	struct sys_mmap_args cma;
491 
492 	linux_to_bsd_mmap_args(&cma, uap);
493 	SCARG(&cma, pos) = ((off_t)SCARG(uap, offset)) << PAGE_SHIFT;
494 
495 	return sys_mmap(l, &cma, retval);
496 }
497 
498 static void
499 linux_to_bsd_mmap_args(cma, uap)
500 	struct sys_mmap_args *cma;
501 	const struct linux_sys_mmap_args *uap;
502 {
503 	int flags = MAP_TRYFIXED, fl = SCARG(uap, flags);
504 
505 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
506 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
507 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
508 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
509 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
510 
511 	SCARG(cma, addr) = (void *)SCARG(uap, addr);
512 	SCARG(cma, len) = SCARG(uap, len);
513 	SCARG(cma, prot) = SCARG(uap, prot);
514 	if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */
515 		SCARG(cma, prot) |= VM_PROT_READ;
516 	SCARG(cma, flags) = flags;
517 	SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
518 	SCARG(cma, pad) = 0;
519 }
520 
521 int
522 linux_sys_mremap(l, v, retval)
523 	struct lwp *l;
524 	void *v;
525 	register_t *retval;
526 {
527 	struct linux_sys_mremap_args /* {
528 		syscallarg(void *) old_address;
529 		syscallarg(size_t) old_size;
530 		syscallarg(size_t) new_size;
531 		syscallarg(u_long) flags;
532 	} */ *uap = v;
533 	struct sys_munmap_args mua;
534 	size_t old_size, new_size;
535 	int error;
536 
537 	old_size = round_page(SCARG(uap, old_size));
538 	new_size = round_page(SCARG(uap, new_size));
539 
540 	/*
541 	 * Growing mapped region.
542 	 */
543 	if (new_size > old_size) {
544 		/*
545 		 * XXX Implement me.  What we probably want to do is
546 		 * XXX dig out the guts of the old mapping, mmap that
547 		 * XXX object again with the new size, then munmap
548 		 * XXX the old mapping.
549 		 */
550 		*retval = 0;
551 		return (ENOMEM);
552 	}
553 
554 	/*
555 	 * Shrinking mapped region.
556 	 */
557 	if (new_size < old_size) {
558 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
559 		    new_size;
560 		SCARG(&mua, len) = old_size - new_size;
561 		error = sys_munmap(l, &mua, retval);
562 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
563 		return (error);
564 	}
565 
566 	/*
567 	 * No change.
568 	 */
569 	*retval = (register_t)SCARG(uap, old_address);
570 	return (0);
571 }
572 
573 int
574 linux_sys_msync(l, v, retval)
575 	struct lwp *l;
576 	void *v;
577 	register_t *retval;
578 {
579 	struct linux_sys_msync_args /* {
580 		syscallarg(caddr_t) addr;
581 		syscallarg(int) len;
582 		syscallarg(int) fl;
583 	} */ *uap = v;
584 
585 	struct sys___msync13_args bma;
586 
587 	/* flags are ignored */
588 	SCARG(&bma, addr) = SCARG(uap, addr);
589 	SCARG(&bma, len) = SCARG(uap, len);
590 	SCARG(&bma, flags) = SCARG(uap, fl);
591 
592 	return sys___msync13(l, &bma, retval);
593 }
594 
595 int
596 linux_sys_mprotect(l, v, retval)
597 	struct lwp *l;
598 	void *v;
599 	register_t *retval;
600 {
601 	struct linux_sys_mprotect_args /* {
602 		syscallarg(const void *) start;
603 		syscallarg(unsigned long) len;
604 		syscallarg(int) prot;
605 	} */ *uap = v;
606 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
607 	int prot = SCARG(uap, prot);
608 	struct vm_map_entry *entry;
609 	struct vm_map *map = &l->l_proc->p_vmspace->vm_map;
610 
611 	if (start & PAGE_MASK)
612 		return EINVAL;
613 
614 	len = round_page(SCARG(uap, len));
615 	end = start + len;
616 
617 	if (end < start)
618 		return EINVAL;
619 	else if (end == start)
620 		return 0;
621 
622 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
623 		return EINVAL;
624 
625 	vm_map_lock(map);
626 #ifdef notdef
627 	VM_MAP_RANGE_CHECK(map, start, end);
628 #endif
629 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
630 		vm_map_unlock(map);
631 		return EFAULT;
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(l, v, retval)
645 	struct lwp *l;
646 	void *v;
647 	register_t *retval;
648 {
649 	struct linux_sys_times_args /* {
650 		syscallarg(struct times *) tms;
651 	} */ *uap = v;
652 	struct proc *p = l->l_proc;
653 	struct timeval t;
654 	int error, s;
655 
656 	if (SCARG(uap, tms)) {
657 		struct linux_tms ltms;
658 		struct rusage ru;
659 
660 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
661 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
662 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
663 
664 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
665 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
666 
667 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
668 			return error;
669 	}
670 
671 	s = splclock();
672 	timersub(&time, &boottime, &t);
673 	splx(s);
674 
675 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
676 	return 0;
677 }
678 
679 #undef CONVTCK
680 
681 /*
682  * Linux 'readdir' call. This code is mostly taken from the
683  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
684  * an attempt has been made to keep it a little cleaner (failing
685  * miserably, because of the cruft needed if count 1 is passed).
686  *
687  * The d_off field should contain the offset of the next valid entry,
688  * but in Linux it has the offset of the entry itself. We emulate
689  * that bug here.
690  *
691  * Read in BSD-style entries, convert them, and copy them out.
692  *
693  * Note that this doesn't handle union-mounted filesystems.
694  */
695 int
696 linux_sys_getdents(l, v, retval)
697 	struct lwp *l;
698 	void *v;
699 	register_t *retval;
700 {
701 	struct linux_sys_getdents_args /* {
702 		syscallarg(int) fd;
703 		syscallarg(struct linux_dirent *) dent;
704 		syscallarg(unsigned int) count;
705 	} */ *uap = v;
706 	struct proc *p = l->l_proc;
707 	struct dirent *bdp;
708 	struct vnode *vp;
709 	caddr_t	inp, buf;		/* BSD-format */
710 	int len, reclen;		/* BSD-format */
711 	caddr_t outp;			/* Linux-format */
712 	int resid, linux_reclen = 0;	/* Linux-format */
713 	struct file *fp;
714 	struct uio auio;
715 	struct iovec aiov;
716 	struct linux_dirent idb;
717 	off_t off;		/* true file offset */
718 	int buflen, error, eofflag, nbytes, oldcall;
719 	struct vattr va;
720 	off_t *cookiebuf = NULL, *cookie;
721 	int ncookies;
722 
723 	/* getvnode() will use the descriptor for us */
724 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
725 		return (error);
726 
727 	if ((fp->f_flag & FREAD) == 0) {
728 		error = EBADF;
729 		goto out1;
730 	}
731 
732 	vp = (struct vnode *)fp->f_data;
733 	if (vp->v_type != VDIR) {
734 		error = EINVAL;
735 		goto out1;
736 	}
737 
738 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
739 		goto out1;
740 
741 	nbytes = SCARG(uap, count);
742 	if (nbytes == 1) {	/* emulating old, broken behaviour */
743 		nbytes = sizeof (idb);
744 		buflen = max(va.va_blocksize, nbytes);
745 		oldcall = 1;
746 	} else {
747 		buflen = min(MAXBSIZE, nbytes);
748 		if (buflen < va.va_blocksize)
749 			buflen = va.va_blocksize;
750 		oldcall = 0;
751 	}
752 	buf = malloc(buflen, M_TEMP, M_WAITOK);
753 
754 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
755 	off = fp->f_offset;
756 again:
757 	aiov.iov_base = buf;
758 	aiov.iov_len = buflen;
759 	auio.uio_iov = &aiov;
760 	auio.uio_iovcnt = 1;
761 	auio.uio_rw = UIO_READ;
762 	auio.uio_segflg = UIO_SYSSPACE;
763 	auio.uio_procp = p;
764 	auio.uio_resid = buflen;
765 	auio.uio_offset = off;
766 	/*
767          * First we read into the malloc'ed buffer, then
768          * we massage it into user space, one record at a time.
769          */
770 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
771 	    &ncookies);
772 	if (error)
773 		goto out;
774 
775 	inp = buf;
776 	outp = (caddr_t)SCARG(uap, dent);
777 	resid = nbytes;
778 	if ((len = buflen - auio.uio_resid) == 0)
779 		goto eof;
780 
781 	for (cookie = cookiebuf; len > 0; len -= reclen) {
782 		bdp = (struct dirent *)inp;
783 		reclen = bdp->d_reclen;
784 		if (reclen & 3)
785 			panic("linux_readdir");
786 		if (bdp->d_fileno == 0) {
787 			inp += reclen;	/* it is a hole; squish it out */
788 			off = *cookie++;
789 			continue;
790 		}
791 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
792 		if (reclen > len || resid < linux_reclen) {
793 			/* entry too big for buffer, so just stop */
794 			outp++;
795 			break;
796 		}
797 		/*
798 		 * Massage in place to make a Linux-shaped dirent (otherwise
799 		 * we have to worry about touching user memory outside of
800 		 * the copyout() call).
801 		 */
802 		idb.d_ino = bdp->d_fileno;
803 		/*
804 		 * The old readdir() call misuses the offset and reclen fields.
805 		 */
806 		if (oldcall) {
807 			idb.d_off = (linux_off_t)linux_reclen;
808 			idb.d_reclen = (u_short)bdp->d_namlen;
809 		} else {
810 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
811 				compat_offseterr(vp, "linux_getdents");
812 				error = EINVAL;
813 				goto out;
814 			}
815 			idb.d_off = (linux_off_t)off;
816 			idb.d_reclen = (u_short)linux_reclen;
817 		}
818 		strcpy(idb.d_name, bdp->d_name);
819 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
820 			goto out;
821 		/* advance past this real entry */
822 		inp += reclen;
823 		off = *cookie++;	/* each entry points to itself */
824 		/* advance output past Linux-shaped entry */
825 		outp += linux_reclen;
826 		resid -= linux_reclen;
827 		if (oldcall)
828 			break;
829 	}
830 
831 	/* if we squished out the whole block, try again */
832 	if (outp == (caddr_t)SCARG(uap, dent))
833 		goto again;
834 	fp->f_offset = off;	/* update the vnode offset */
835 
836 	if (oldcall)
837 		nbytes = resid + linux_reclen;
838 
839 eof:
840 	*retval = nbytes - resid;
841 out:
842 	VOP_UNLOCK(vp, 0);
843 	if (cookiebuf)
844 		free(cookiebuf, M_TEMP);
845 	free(buf, M_TEMP);
846 out1:
847 	FILE_UNUSE(fp, p);
848 	return error;
849 }
850 
851 /*
852  * Even when just using registers to pass arguments to syscalls you can
853  * have 5 of them on the i386. So this newer version of select() does
854  * this.
855  */
856 int
857 linux_sys_select(l, v, retval)
858 	struct lwp *l;
859 	void *v;
860 	register_t *retval;
861 {
862 	struct linux_sys_select_args /* {
863 		syscallarg(int) nfds;
864 		syscallarg(fd_set *) readfds;
865 		syscallarg(fd_set *) writefds;
866 		syscallarg(fd_set *) exceptfds;
867 		syscallarg(struct timeval *) timeout;
868 	} */ *uap = v;
869 
870 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
871 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
872 }
873 
874 /*
875  * Common code for the old and new versions of select(). A couple of
876  * things are important:
877  * 1) return the amount of time left in the 'timeout' parameter
878  * 2) select never returns ERESTART on Linux, always return EINTR
879  */
880 int
881 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
882 	struct lwp *l;
883 	register_t *retval;
884 	int nfds;
885 	fd_set *readfds, *writefds, *exceptfds;
886 	struct timeval *timeout;
887 {
888 	struct sys_select_args bsa;
889 	struct proc *p = l->l_proc;
890 	struct timeval tv0, tv1, utv, *tvp;
891 	caddr_t sg;
892 	int error;
893 
894 	SCARG(&bsa, nd) = nfds;
895 	SCARG(&bsa, in) = readfds;
896 	SCARG(&bsa, ou) = writefds;
897 	SCARG(&bsa, ex) = exceptfds;
898 	SCARG(&bsa, tv) = timeout;
899 
900 	/*
901 	 * Store current time for computation of the amount of
902 	 * time left.
903 	 */
904 	if (timeout) {
905 		if ((error = copyin(timeout, &utv, sizeof(utv))))
906 			return error;
907 		if (itimerfix(&utv)) {
908 			/*
909 			 * The timeval was invalid.  Convert it to something
910 			 * valid that will act as it does under Linux.
911 			 */
912 			sg = stackgap_init(p, 0);
913 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
914 			utv.tv_sec += utv.tv_usec / 1000000;
915 			utv.tv_usec %= 1000000;
916 			if (utv.tv_usec < 0) {
917 				utv.tv_sec -= 1;
918 				utv.tv_usec += 1000000;
919 			}
920 			if (utv.tv_sec < 0)
921 				timerclear(&utv);
922 			if ((error = copyout(&utv, tvp, sizeof(utv))))
923 				return error;
924 			SCARG(&bsa, tv) = tvp;
925 		}
926 		microtime(&tv0);
927 	}
928 
929 	error = sys_select(l, &bsa, retval);
930 	if (error) {
931 		/*
932 		 * See fs/select.c in the Linux kernel.  Without this,
933 		 * Maelstrom doesn't work.
934 		 */
935 		if (error == ERESTART)
936 			error = EINTR;
937 		return error;
938 	}
939 
940 	if (timeout) {
941 		if (*retval) {
942 			/*
943 			 * Compute how much time was left of the timeout,
944 			 * by subtracting the current time and the time
945 			 * before we started the call, and subtracting
946 			 * that result from the user-supplied value.
947 			 */
948 			microtime(&tv1);
949 			timersub(&tv1, &tv0, &tv1);
950 			timersub(&utv, &tv1, &utv);
951 			if (utv.tv_sec < 0)
952 				timerclear(&utv);
953 		} else
954 			timerclear(&utv);
955 		if ((error = copyout(&utv, timeout, sizeof(utv))))
956 			return error;
957 	}
958 
959 	return 0;
960 }
961 
962 /*
963  * Get the process group of a certain process. Look it up
964  * and return the value.
965  */
966 int
967 linux_sys_getpgid(l, v, retval)
968 	struct lwp *l;
969 	void *v;
970 	register_t *retval;
971 {
972 	struct linux_sys_getpgid_args /* {
973 		syscallarg(int) pid;
974 	} */ *uap = v;
975 	struct proc *p = l->l_proc;
976 	struct proc *targp;
977 
978 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
979 		if ((targp = pfind(SCARG(uap, pid))) == 0)
980 			return ESRCH;
981 	}
982 	else
983 		targp = p;
984 
985 	retval[0] = targp->p_pgid;
986 	return 0;
987 }
988 
989 /*
990  * Set the 'personality' (emulation mode) for the current process. Only
991  * accept the Linux personality here (0). This call is needed because
992  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
993  * ELF binaries run in Linux mode, not SVR4 mode.
994  */
995 int
996 linux_sys_personality(l, v, retval)
997 	struct lwp *l;
998 	void *v;
999 	register_t *retval;
1000 {
1001 	struct linux_sys_personality_args /* {
1002 		syscallarg(int) per;
1003 	} */ *uap = v;
1004 
1005 	if (SCARG(uap, per) != 0)
1006 		return EINVAL;
1007 	retval[0] = 0;
1008 	return 0;
1009 }
1010 
1011 #if defined(__i386__) || defined(__m68k__)
1012 /*
1013  * The calls are here because of type conversions.
1014  */
1015 int
1016 linux_sys_setreuid16(l, v, retval)
1017 	struct lwp *l;
1018 	void *v;
1019 	register_t *retval;
1020 {
1021 	struct linux_sys_setreuid16_args /* {
1022 		syscallarg(int) ruid;
1023 		syscallarg(int) euid;
1024 	} */ *uap = v;
1025 	struct sys_setreuid_args bsa;
1026 
1027 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1028 		(uid_t)-1 : SCARG(uap, ruid);
1029 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1030 		(uid_t)-1 : SCARG(uap, euid);
1031 
1032 	return sys_setreuid(l, &bsa, retval);
1033 }
1034 
1035 int
1036 linux_sys_setregid16(l, v, retval)
1037 	struct lwp *l;
1038 	void *v;
1039 	register_t *retval;
1040 {
1041 	struct linux_sys_setregid16_args /* {
1042 		syscallarg(int) rgid;
1043 		syscallarg(int) egid;
1044 	} */ *uap = v;
1045 	struct sys_setregid_args bsa;
1046 
1047 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1048 		(uid_t)-1 : SCARG(uap, rgid);
1049 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1050 		(uid_t)-1 : SCARG(uap, egid);
1051 
1052 	return sys_setregid(l, &bsa, retval);
1053 }
1054 
1055 int
1056 linux_sys_setresuid16(l, v, retval)
1057 	struct lwp *l;
1058 	void *v;
1059 	register_t *retval;
1060 {
1061 	struct linux_sys_setresuid16_args /* {
1062 		syscallarg(uid_t) ruid;
1063 		syscallarg(uid_t) euid;
1064 		syscallarg(uid_t) suid;
1065 	} */ *uap = v;
1066 	struct linux_sys_setresuid16_args lsa;
1067 
1068 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1069 		(uid_t)-1 : SCARG(uap, ruid);
1070 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1071 		(uid_t)-1 : SCARG(uap, euid);
1072 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1073 		(uid_t)-1 : SCARG(uap, suid);
1074 
1075 	return linux_sys_setresuid(l, &lsa, retval);
1076 }
1077 
1078 int
1079 linux_sys_setresgid16(l, v, retval)
1080 	struct lwp *l;
1081 	void *v;
1082 	register_t *retval;
1083 {
1084 	struct linux_sys_setresgid16_args /* {
1085 		syscallarg(gid_t) rgid;
1086 		syscallarg(gid_t) egid;
1087 		syscallarg(gid_t) sgid;
1088 	} */ *uap = v;
1089 	struct linux_sys_setresgid16_args lsa;
1090 
1091 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1092 		(gid_t)-1 : SCARG(uap, rgid);
1093 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1094 		(gid_t)-1 : SCARG(uap, egid);
1095 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1096 		(gid_t)-1 : SCARG(uap, sgid);
1097 
1098 	return linux_sys_setresgid(l, &lsa, retval);
1099 }
1100 
1101 int
1102 linux_sys_getgroups16(l, v, retval)
1103 	struct lwp *l;
1104 	void *v;
1105 	register_t *retval;
1106 {
1107 	struct linux_sys_getgroups16_args /* {
1108 		syscallarg(int) gidsetsize;
1109 		syscallarg(linux_gid_t *) gidset;
1110 	} */ *uap = v;
1111 	struct proc *p = l->l_proc;
1112 	caddr_t sg;
1113 	int n, error, i;
1114 	struct sys_getgroups_args bsa;
1115 	gid_t *bset, *kbset;
1116 	linux_gid_t *lset;
1117 	struct pcred *pc = p->p_cred;
1118 
1119 	n = SCARG(uap, gidsetsize);
1120 	if (n < 0)
1121 		return EINVAL;
1122 	error = 0;
1123 	bset = kbset = NULL;
1124 	lset = NULL;
1125 	if (n > 0) {
1126 		n = min(pc->pc_ucred->cr_ngroups, n);
1127 		sg = stackgap_init(p, 0);
1128 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1129 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1130 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1131 		if (bset == NULL || kbset == NULL || lset == NULL)
1132 			return ENOMEM;
1133 		SCARG(&bsa, gidsetsize) = n;
1134 		SCARG(&bsa, gidset) = bset;
1135 		error = sys_getgroups(l, &bsa, retval);
1136 		if (error != 0)
1137 			goto out;
1138 		error = copyin(bset, kbset, n * sizeof (gid_t));
1139 		if (error != 0)
1140 			goto out;
1141 		for (i = 0; i < n; i++)
1142 			lset[i] = (linux_gid_t)kbset[i];
1143 		error = copyout(lset, SCARG(uap, gidset),
1144 		    n * sizeof (linux_gid_t));
1145 	} else
1146 		*retval = pc->pc_ucred->cr_ngroups;
1147 out:
1148 	if (kbset != NULL)
1149 		free(kbset, M_TEMP);
1150 	if (lset != NULL)
1151 		free(lset, M_TEMP);
1152 	return error;
1153 }
1154 
1155 int
1156 linux_sys_setgroups16(l, v, retval)
1157 	struct lwp *l;
1158 	void *v;
1159 	register_t *retval;
1160 {
1161 	struct linux_sys_setgroups16_args /* {
1162 		syscallarg(int) gidsetsize;
1163 		syscallarg(linux_gid_t *) gidset;
1164 	} */ *uap = v;
1165 	struct proc *p = l->l_proc;
1166 	caddr_t sg;
1167 	int n;
1168 	int error, i;
1169 	struct sys_setgroups_args bsa;
1170 	gid_t *bset, *kbset;
1171 	linux_gid_t *lset;
1172 
1173 	n = SCARG(uap, gidsetsize);
1174 	if (n < 0 || n > NGROUPS)
1175 		return EINVAL;
1176 	sg = stackgap_init(p, 0);
1177 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1178 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1179 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1180 	if (lset == NULL || bset == NULL)
1181 		return ENOMEM;
1182 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1183 	if (error != 0)
1184 		goto out;
1185 	for (i = 0; i < n; i++)
1186 		kbset[i] = (gid_t)lset[i];
1187 	error = copyout(kbset, bset, n * sizeof (gid_t));
1188 	if (error != 0)
1189 		goto out;
1190 	SCARG(&bsa, gidsetsize) = n;
1191 	SCARG(&bsa, gidset) = bset;
1192 	error = sys_setgroups(l, &bsa, retval);
1193 
1194 out:
1195 	if (lset != NULL)
1196 		free(lset, M_TEMP);
1197 	if (kbset != NULL)
1198 		free(kbset, M_TEMP);
1199 
1200 	return error;
1201 }
1202 
1203 #endif /* __i386__ || __m68k__ */
1204 
1205 /*
1206  * We have nonexistent fsuid equal to uid.
1207  * If modification is requested, refuse.
1208  */
1209 int
1210 linux_sys_setfsuid(l, v, retval)
1211 	 struct lwp *l;
1212 	 void *v;
1213 	 register_t *retval;
1214 {
1215 	 struct linux_sys_setfsuid_args /* {
1216 		 syscallarg(uid_t) uid;
1217 	 } */ *uap = v;
1218 	 struct proc *p = l->l_proc;
1219 	 uid_t uid;
1220 
1221 	 uid = SCARG(uap, uid);
1222 	 if (p->p_cred->p_ruid != uid)
1223 		 return sys_nosys(l, v, retval);
1224 	 else
1225 		 return (0);
1226 }
1227 
1228 /* XXX XXX XXX */
1229 #ifndef alpha
1230 int
1231 linux_sys_getfsuid(l, v, retval)
1232 	struct lwp *l;
1233 	void *v;
1234 	register_t *retval;
1235 {
1236 	return sys_getuid(l, v, retval);
1237 }
1238 #endif
1239 
1240 int
1241 linux_sys_setresuid(l, v, retval)
1242 	struct lwp *l;
1243 	void *v;
1244 	register_t *retval;
1245 {
1246 	struct linux_sys_setresuid_args /* {
1247 		syscallarg(uid_t) ruid;
1248 		syscallarg(uid_t) euid;
1249 		syscallarg(uid_t) suid;
1250 	} */ *uap = v;
1251 
1252 	/*
1253 	 * Note: These checks are a little different than the NetBSD
1254 	 * setreuid(2) call performs.  This precisely follows the
1255 	 * behavior of the Linux kernel.
1256 	 */
1257 
1258 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1259 			    SCARG(uap, suid),
1260 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1261 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1262 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1263 }
1264 
1265 int
1266 linux_sys_getresuid(l, v, retval)
1267 	struct lwp *l;
1268 	void *v;
1269 	register_t *retval;
1270 {
1271 	struct linux_sys_getresuid_args /* {
1272 		syscallarg(uid_t *) ruid;
1273 		syscallarg(uid_t *) euid;
1274 		syscallarg(uid_t *) suid;
1275 	} */ *uap = v;
1276 	struct proc *p = l->l_proc;
1277 	struct pcred *pc = p->p_cred;
1278 	int error;
1279 
1280 	/*
1281 	 * Linux copies these values out to userspace like so:
1282 	 *
1283 	 *	1. Copy out ruid.
1284 	 *	2. If that succeeds, copy out euid.
1285 	 *	3. If both of those succeed, copy out suid.
1286 	 */
1287 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1288 			     sizeof(uid_t))) != 0)
1289 		return (error);
1290 
1291 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1292 			     sizeof(uid_t))) != 0)
1293 		return (error);
1294 
1295 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1296 }
1297 
1298 int
1299 linux_sys_ptrace(l, v, retval)
1300 	struct lwp *l;
1301 	void *v;
1302 	register_t *retval;
1303 {
1304 	struct linux_sys_ptrace_args /* {
1305 		i386, m68k, powerpc: T=int
1306 		alpha: T=long
1307 		syscallarg(T) request;
1308 		syscallarg(T) pid;
1309 		syscallarg(T) addr;
1310 		syscallarg(T) data;
1311 	} */ *uap = v;
1312 	const int *ptr;
1313 	int request;
1314 	int error;
1315 
1316 	ptr = linux_ptrace_request_map;
1317 	request = SCARG(uap, request);
1318 	while (*ptr != -1)
1319 		if (*ptr++ == request) {
1320 			struct sys_ptrace_args pta;
1321 
1322 			SCARG(&pta, req) = *ptr;
1323 			SCARG(&pta, pid) = SCARG(uap, pid);
1324 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1325 			SCARG(&pta, data) = SCARG(uap, data);
1326 
1327 			/*
1328 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1329 			 * to continue where the process left off previously.
1330 			 * The same thing is achieved by addr == (caddr_t) 1
1331 			 * on NetBSD, so rewrite 'addr' appropriately.
1332 			 */
1333 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1334 				SCARG(&pta, addr) = (caddr_t) 1;
1335 
1336 			error = sys_ptrace(l, &pta, retval);
1337 			if (error)
1338 				return error;
1339 			switch (request) {
1340 			case LINUX_PTRACE_PEEKTEXT:
1341 			case LINUX_PTRACE_PEEKDATA:
1342 				error = copyout (retval,
1343 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1344 				*retval = SCARG(uap, data);
1345 				break;
1346 			default:
1347 				break;
1348 			}
1349 			return error;
1350 		}
1351 		else
1352 			ptr++;
1353 
1354 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1355 }
1356 
1357 int
1358 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1359 {
1360 	struct linux_sys_reboot_args /* {
1361 		syscallarg(int) magic1;
1362 		syscallarg(int) magic2;
1363 		syscallarg(int) cmd;
1364 		syscallarg(void *) arg;
1365 	} */ *uap = v;
1366 	struct sys_reboot_args /* {
1367 		syscallarg(int) opt;
1368 		syscallarg(char *) bootstr;
1369 	} */ sra;
1370 	struct proc *p = l->l_proc;
1371 	int error;
1372 
1373 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1374 		return(error);
1375 
1376 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1377 		return(EINVAL);
1378 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1379 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1380 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1381 		return(EINVAL);
1382 
1383 	switch (SCARG(uap, cmd)) {
1384 	case LINUX_REBOOT_CMD_RESTART:
1385 		SCARG(&sra, opt) = RB_AUTOBOOT;
1386 		break;
1387 	case LINUX_REBOOT_CMD_HALT:
1388 		SCARG(&sra, opt) = RB_HALT;
1389 		break;
1390 	case LINUX_REBOOT_CMD_POWER_OFF:
1391 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1392 		break;
1393 	case LINUX_REBOOT_CMD_RESTART2:
1394 		/* Reboot with an argument. */
1395 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1396 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1397 		break;
1398 	case LINUX_REBOOT_CMD_CAD_ON:
1399 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1400 	case LINUX_REBOOT_CMD_CAD_OFF:
1401 		return(0);
1402 	default:
1403 		return(EINVAL);
1404 	}
1405 
1406 	return(sys_reboot(l, &sra, retval));
1407 }
1408 
1409 /*
1410  * Copy of compat_12_sys_swapon().
1411  */
1412 int
1413 linux_sys_swapon(l, v, retval)
1414 	struct lwp *l;
1415 	void *v;
1416 	register_t *retval;
1417 {
1418 	struct sys_swapctl_args ua;
1419 	struct linux_sys_swapon_args /* {
1420 		syscallarg(const char *) name;
1421 	} */ *uap = v;
1422 
1423 	SCARG(&ua, cmd) = SWAP_ON;
1424 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1425 	SCARG(&ua, misc) = 0;	/* priority */
1426 	return (sys_swapctl(l, &ua, retval));
1427 }
1428 
1429 /*
1430  * Stop swapping to the file or block device specified by path.
1431  */
1432 int
1433 linux_sys_swapoff(l, v, retval)
1434 	struct lwp *l;
1435 	void *v;
1436 	register_t *retval;
1437 {
1438 	struct sys_swapctl_args ua;
1439 	struct linux_sys_swapoff_args /* {
1440 		syscallarg(const char *) path;
1441 	} */ *uap = v;
1442 
1443 	SCARG(&ua, cmd) = SWAP_OFF;
1444 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1445 	return (sys_swapctl(l, &ua, retval));
1446 }
1447 
1448 /*
1449  * Copy of compat_09_sys_setdomainname()
1450  */
1451 /* ARGSUSED */
1452 int
1453 linux_sys_setdomainname(l, v, retval)
1454 	struct lwp *l;
1455 	void *v;
1456 	register_t *retval;
1457 {
1458 	struct linux_sys_setdomainname_args /* {
1459 		syscallarg(char *) domainname;
1460 		syscallarg(int) len;
1461 	} */ *uap = v;
1462 	int name[2];
1463 
1464 	name[0] = CTL_KERN;
1465 	name[1] = KERN_DOMAINNAME;
1466 	return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname),
1467 			    SCARG(uap, len), l));
1468 }
1469 
1470 /*
1471  * sysinfo()
1472  */
1473 /* ARGSUSED */
1474 int
1475 linux_sys_sysinfo(l, v, retval)
1476 	struct lwp *l;
1477 	void *v;
1478 	register_t *retval;
1479 {
1480 	struct linux_sys_sysinfo_args /* {
1481 		syscallarg(struct linux_sysinfo *) arg;
1482 	} */ *uap = v;
1483 	struct linux_sysinfo si;
1484 	struct loadavg *la;
1485 
1486 	si.uptime = time.tv_sec - boottime.tv_sec;
1487 	la = &averunnable;
1488 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1489 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1490 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1491 	si.totalram = ctob(physmem);
1492 	si.freeram = uvmexp.free * uvmexp.pagesize;
1493 	si.sharedram = 0;	/* XXX */
1494 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1495 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1496 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1497 	si.procs = nprocs;
1498 
1499 	/* The following are only present in newer Linux kernels. */
1500 	si.totalbig = 0;
1501 	si.freebig = 0;
1502 	si.mem_unit = 1;
1503 
1504 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1505 }
1506 
1507 #define bsd_to_linux_rlimit1(l, b, f) \
1508     (l)->f = ((b)->f == RLIM_INFINITY || \
1509 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1510     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1511 #define bsd_to_linux_rlimit(l, b) \
1512     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1513     bsd_to_linux_rlimit1(l, b, rlim_max)
1514 
1515 #define linux_to_bsd_rlimit1(b, l, f) \
1516     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1517 #define linux_to_bsd_rlimit(b, l) \
1518     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1519     linux_to_bsd_rlimit1(b, l, rlim_max)
1520 
1521 static int
1522 linux_to_bsd_limit(lim)
1523 	int lim;
1524 {
1525 	switch (lim) {
1526 	case LINUX_RLIMIT_CPU:
1527 		return RLIMIT_CPU;
1528 	case LINUX_RLIMIT_FSIZE:
1529 		return RLIMIT_FSIZE;
1530 	case LINUX_RLIMIT_DATA:
1531 		return RLIMIT_DATA;
1532 	case LINUX_RLIMIT_STACK:
1533 		return RLIMIT_STACK;
1534 	case LINUX_RLIMIT_CORE:
1535 		return RLIMIT_CORE;
1536 	case LINUX_RLIMIT_RSS:
1537 		return RLIMIT_RSS;
1538 	case LINUX_RLIMIT_NPROC:
1539 		return RLIMIT_NPROC;
1540 	case LINUX_RLIMIT_NOFILE:
1541 		return RLIMIT_NOFILE;
1542 	case LINUX_RLIMIT_MEMLOCK:
1543 		return RLIMIT_MEMLOCK;
1544 	case LINUX_RLIMIT_AS:
1545 	case LINUX_RLIMIT_LOCKS:
1546 		return -EOPNOTSUPP;
1547 	default:
1548 		return -EINVAL;
1549 	}
1550 }
1551 
1552 
1553 int
1554 linux_sys_getrlimit(l, v, retval)
1555 	struct lwp *l;
1556 	void *v;
1557 	register_t *retval;
1558 {
1559 	struct linux_sys_getrlimit_args /* {
1560 		syscallarg(int) which;
1561 		syscallarg(struct orlimit *) rlp;
1562 	} */ *uap = v;
1563 	struct proc *p = l->l_proc;
1564 	caddr_t sg = stackgap_init(p, 0);
1565 	struct sys_getrlimit_args ap;
1566 	struct rlimit rl;
1567 	struct orlimit orl;
1568 	int error;
1569 
1570 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1571 	if ((error = SCARG(&ap, which)) < 0)
1572 		return -error;
1573 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1574 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1575 		return error;
1576 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1577 		return error;
1578 	bsd_to_linux_rlimit(&orl, &rl);
1579 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1580 }
1581 
1582 int
1583 linux_sys_setrlimit(l, v, retval)
1584 	struct lwp *l;
1585 	void *v;
1586 	register_t *retval;
1587 {
1588 	struct linux_sys_setrlimit_args /* {
1589 		syscallarg(int) which;
1590 		syscallarg(struct orlimit *) rlp;
1591 	} */ *uap = v;
1592 	struct proc *p = l->l_proc;
1593 	caddr_t sg = stackgap_init(p, 0);
1594 	struct sys_setrlimit_args ap;
1595 	struct rlimit rl;
1596 	struct orlimit orl;
1597 	int error;
1598 
1599 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1600 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1601 	if ((error = SCARG(&ap, which)) < 0)
1602 		return -error;
1603 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1604 		return error;
1605 	linux_to_bsd_rlimit(&rl, &orl);
1606 	/* XXX: alpha complains about this */
1607 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1608 		return error;
1609 	return sys_setrlimit(l, &ap, retval);
1610 }
1611 
1612 #ifndef __mips__
1613 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1614 int
1615 linux_sys_ugetrlimit(l, v, retval)
1616 	struct lwp *l;
1617 	void *v;
1618 	register_t *retval;
1619 {
1620 	return linux_sys_getrlimit(l, v, retval);
1621 }
1622 #endif
1623 
1624 /*
1625  * This gets called for unsupported syscalls. The difference to sys_nosys()
1626  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1627  * This is the way Linux does it and glibc depends on this behaviour.
1628  */
1629 int
1630 linux_sys_nosys(l, v, retval)
1631 	struct lwp *l;
1632 	void *v;
1633 	register_t *retval;
1634 {
1635 	return (ENOSYS);
1636 }
1637