xref: /netbsd-src/sys/compat/linux/common/linux_misc.c (revision 1ffa7b76c40339c17a0fb2a09fac93f287cfc046)
1 /*	$NetBSD: linux_misc.c,v 1.117 2003/03/05 18:44:46 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 1.117 2003/03/05 18:44:46 dsl 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((struct statfs *, struct linux_statfs *));
172 static int linux_to_bsd_limit __P((int));
173 
174 /*
175  * The information on a terminated (or stopped) process needs
176  * to be converted in order for Linux binaries to get a valid signal
177  * number out of it.
178  */
179 void
180 bsd_to_linux_wstat(st)
181 	int *st;
182 {
183 
184 	int sig;
185 
186 	if (WIFSIGNALED(*st)) {
187 		sig = WTERMSIG(*st);
188 		if (sig >= 0 && sig < NSIG)
189 			*st= (*st& ~0177) | native_to_linux_signo[sig];
190 	} else if (WIFSTOPPED(*st)) {
191 		sig = WSTOPSIG(*st);
192 		if (sig >= 0 && sig < NSIG)
193 			*st = (*st & ~0xff00) |
194 			    (native_to_linux_signo[sig] << 8);
195 	}
196 }
197 
198 /*
199  * This is very much the same as waitpid()
200  */
201 int
202 linux_sys_wait4(l, v, retval)
203 	struct lwp *l;
204 	void *v;
205 	register_t *retval;
206 {
207 	struct linux_sys_wait4_args /* {
208 		syscallarg(int) pid;
209 		syscallarg(int *) status;
210 		syscallarg(int) options;
211 		syscallarg(struct rusage *) rusage;
212 	} */ *uap = v;
213 	struct proc *p = l->l_proc;
214 	struct sys_wait4_args w4a;
215 	int error, *status, tstat, options, linux_options;
216 	caddr_t sg;
217 
218 	if (SCARG(uap, status) != NULL) {
219 		sg = stackgap_init(p, 0);
220 		status = (int *) stackgap_alloc(p, &sg, sizeof *status);
221 	} else
222 		status = NULL;
223 
224 	linux_options = SCARG(uap, options);
225 	options = 0;
226 	if (linux_options &
227 	    ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL|
228 	      LINUX_WAIT4_WCLONE))
229 		return (EINVAL);
230 
231 	if (linux_options & LINUX_WAIT4_WNOHANG)
232 		options |= WNOHANG;
233 	if (linux_options & LINUX_WAIT4_WUNTRACED)
234 		options |= WUNTRACED;
235 	if (linux_options & LINUX_WAIT4_WALL)
236 		options |= WALLSIG;
237 	if (linux_options & LINUX_WAIT4_WCLONE)
238 		options |= WALTSIG;
239 
240 	SCARG(&w4a, pid) = SCARG(uap, pid);
241 	SCARG(&w4a, status) = status;
242 	SCARG(&w4a, options) = options;
243 	SCARG(&w4a, rusage) = SCARG(uap, rusage);
244 
245 	if ((error = sys_wait4(l, &w4a, retval)))
246 		return error;
247 
248 	sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD);
249 
250 	if (status != NULL) {
251 		if ((error = copyin(status, &tstat, sizeof tstat)))
252 			return error;
253 
254 		bsd_to_linux_wstat(&tstat);
255 		return copyout(&tstat, SCARG(uap, status), sizeof tstat);
256 	}
257 
258 	return 0;
259 }
260 
261 /*
262  * Linux brk(2). The check if the new address is >= the old one is
263  * done in the kernel in Linux. NetBSD does it in the library.
264  */
265 int
266 linux_sys_brk(l, v, retval)
267 	struct lwp *l;
268 	void *v;
269 	register_t *retval;
270 {
271 	struct linux_sys_brk_args /* {
272 		syscallarg(char *) nsize;
273 	} */ *uap = v;
274 	struct proc *p = l->l_proc;
275 	char *nbrk = SCARG(uap, nsize);
276 	struct sys_obreak_args oba;
277 	struct vmspace *vm = p->p_vmspace;
278 	struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata;
279 
280 	SCARG(&oba, nsize) = nbrk;
281 
282 	if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0)
283 		ed->p_break = (char*)nbrk;
284 	else
285 		nbrk = ed->p_break;
286 
287 	retval[0] = (register_t)nbrk;
288 
289 	return 0;
290 }
291 
292 /*
293  * Convert BSD statfs structure to Linux statfs structure.
294  * The Linux structure has less fields, and it also wants
295  * the length of a name in a dir entry in a field, which
296  * we fake (probably the wrong way).
297  */
298 static void
299 bsd_to_linux_statfs(bsp, lsp)
300 	struct statfs *bsp;
301 	struct linux_statfs *lsp;
302 {
303 	int i;
304 
305 	for (i = 0; i < FSTYPESSIZE; i++)
306 		if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0)
307 			break;
308 
309 	if (i == FSTYPESSIZE) {
310 		DPRINTF(("unhandled fstype in linux emulation: %s\n",
311 		    bsp->f_fstypename));
312 		lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC;
313 	} else {
314 		lsp->l_ftype = fstypes[i].linux;
315 	}
316 
317 	lsp->l_fbsize = bsp->f_bsize;
318 	lsp->l_fblocks = bsp->f_blocks;
319 	lsp->l_fbfree = bsp->f_bfree;
320 	lsp->l_fbavail = bsp->f_bavail;
321 	lsp->l_ffiles = bsp->f_files;
322 	lsp->l_fffree = bsp->f_ffree;
323 	/* Linux sets the fsid to 0..., we don't */
324 	lsp->l_ffsid.val[0] = bsp->f_fsid.val[0];
325 	lsp->l_ffsid.val[1] = bsp->f_fsid.val[1];
326 	lsp->l_fnamelen = MAXNAMLEN;	/* XXX */
327 	(void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare));
328 }
329 
330 /*
331  * Implement the fs stat functions. Straightforward.
332  */
333 int
334 linux_sys_statfs(l, v, retval)
335 	struct lwp *l;
336 	void *v;
337 	register_t *retval;
338 {
339 	struct linux_sys_statfs_args /* {
340 		syscallarg(const char *) path;
341 		syscallarg(struct linux_statfs *) sp;
342 	} */ *uap = v;
343 	struct proc *p = l->l_proc;
344 	struct statfs btmp, *bsp;
345 	struct linux_statfs ltmp;
346 	struct sys_statfs_args bsa;
347 	caddr_t sg;
348 	int error;
349 
350 	sg = stackgap_init(p, 0);
351 	bsp = (struct statfs *) stackgap_alloc(p, &sg, sizeof (struct statfs));
352 
353 	CHECK_ALT_EXIST(p, &sg, SCARG(uap, path));
354 
355 	SCARG(&bsa, path) = SCARG(uap, path);
356 	SCARG(&bsa, buf) = bsp;
357 
358 	if ((error = sys_statfs(l, &bsa, retval)))
359 		return error;
360 
361 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
362 		return error;
363 
364 	bsd_to_linux_statfs(&btmp, &ltmp);
365 
366 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
367 }
368 
369 int
370 linux_sys_fstatfs(l, v, retval)
371 	struct lwp *l;
372 	void *v;
373 	register_t *retval;
374 {
375 	struct linux_sys_fstatfs_args /* {
376 		syscallarg(int) fd;
377 		syscallarg(struct linux_statfs *) sp;
378 	} */ *uap = v;
379 	struct proc *p = l->l_proc;
380 	struct statfs btmp, *bsp;
381 	struct linux_statfs ltmp;
382 	struct sys_fstatfs_args bsa;
383 	caddr_t sg;
384 	int error;
385 
386 	sg = stackgap_init(p, 0);
387 	bsp = (struct statfs *) stackgap_alloc(p, &sg, sizeof (struct statfs));
388 
389 	SCARG(&bsa, fd) = SCARG(uap, fd);
390 	SCARG(&bsa, buf) = bsp;
391 
392 	if ((error = sys_fstatfs(l, &bsa, retval)))
393 		return error;
394 
395 	if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp)))
396 		return error;
397 
398 	bsd_to_linux_statfs(&btmp, &ltmp);
399 
400 	return copyout((caddr_t) &ltmp, (caddr_t) SCARG(uap, sp), sizeof ltmp);
401 }
402 
403 /*
404  * uname(). Just copy the info from the various strings stored in the
405  * kernel, and put it in the Linux utsname structure. That structure
406  * is almost the same as the NetBSD one, only it has fields 65 characters
407  * long, and an extra domainname field.
408  */
409 int
410 linux_sys_uname(l, v, retval)
411 	struct lwp *l;
412 	void *v;
413 	register_t *retval;
414 {
415 	struct linux_sys_uname_args /* {
416 		syscallarg(struct linux_utsname *) up;
417 	} */ *uap = v;
418 	struct linux_utsname luts;
419 
420 	strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname));
421 	strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename));
422 	strncpy(luts.l_release, linux_release, sizeof(luts.l_release));
423 	strncpy(luts.l_version, linux_version, sizeof(luts.l_version));
424 	strncpy(luts.l_machine, machine, sizeof(luts.l_machine));
425 	strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname));
426 
427 	return copyout(&luts, SCARG(uap, up), sizeof(luts));
428 }
429 
430 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */
431 /* Used indirectly on: arm, i386, m68k */
432 
433 /*
434  * New type Linux mmap call.
435  * Only called directly on machines with >= 6 free regs.
436  */
437 int
438 linux_sys_mmap(l, v, retval)
439 	struct lwp *l;
440 	void *v;
441 	register_t *retval;
442 {
443 	struct linux_sys_mmap_args /* {
444 		syscallarg(unsigned long) addr;
445 		syscallarg(size_t) len;
446 		syscallarg(int) prot;
447 		syscallarg(int) flags;
448 		syscallarg(int) fd;
449 		syscallarg(linux_off_t) offset;
450 	} */ *uap = v;
451 	struct sys_mmap_args cma;
452 	int flags = 0, fl = SCARG(uap, flags);
453 
454 	if (SCARG(uap, offset) & PAGE_MASK)
455 		return EINVAL;
456 
457 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED);
458 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE);
459 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED);
460 	flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON);
461 	/* XXX XAX ERH: Any other flags here?  There are more defined... */
462 
463 	SCARG(&cma, addr) = (void *)SCARG(uap, addr);
464 	SCARG(&cma, len) = SCARG(uap, len);
465 	SCARG(&cma, prot) = SCARG(uap, prot);
466 	if (SCARG(&cma, prot) & VM_PROT_WRITE) /* XXX */
467 		SCARG(&cma, prot) |= VM_PROT_READ;
468 	SCARG(&cma, flags) = flags;
469 	SCARG(&cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd);
470 	SCARG(&cma, pad) = 0;
471 	SCARG(&cma, pos) = (off_t)SCARG(uap, offset);
472 
473 	return sys_mmap(l, &cma, retval);
474 }
475 
476 int
477 linux_sys_mremap(l, v, retval)
478 	struct lwp *l;
479 	void *v;
480 	register_t *retval;
481 {
482 	struct linux_sys_mremap_args /* {
483 		syscallarg(void *) old_address;
484 		syscallarg(size_t) old_size;
485 		syscallarg(size_t) new_size;
486 		syscallarg(u_long) flags;
487 	} */ *uap = v;
488 	struct sys_munmap_args mua;
489 	size_t old_size, new_size;
490 	int error;
491 
492 	old_size = round_page(SCARG(uap, old_size));
493 	new_size = round_page(SCARG(uap, new_size));
494 
495 	/*
496 	 * Growing mapped region.
497 	 */
498 	if (new_size > old_size) {
499 		/*
500 		 * XXX Implement me.  What we probably want to do is
501 		 * XXX dig out the guts of the old mapping, mmap that
502 		 * XXX object again with the new size, then munmap
503 		 * XXX the old mapping.
504 		 */
505 		*retval = 0;
506 		return (ENOMEM);
507 	}
508 
509 	/*
510 	 * Shrinking mapped region.
511 	 */
512 	if (new_size < old_size) {
513 		SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) +
514 		    new_size;
515 		SCARG(&mua, len) = old_size - new_size;
516 		error = sys_munmap(l, &mua, retval);
517 		*retval = error ? 0 : (register_t)SCARG(uap, old_address);
518 		return (error);
519 	}
520 
521 	/*
522 	 * No change.
523 	 */
524 	*retval = (register_t)SCARG(uap, old_address);
525 	return (0);
526 }
527 
528 int
529 linux_sys_msync(l, v, retval)
530 	struct lwp *l;
531 	void *v;
532 	register_t *retval;
533 {
534 	struct linux_sys_msync_args /* {
535 		syscallarg(caddr_t) addr;
536 		syscallarg(int) len;
537 		syscallarg(int) fl;
538 	} */ *uap = v;
539 
540 	struct sys___msync13_args bma;
541 
542 	/* flags are ignored */
543 	SCARG(&bma, addr) = SCARG(uap, addr);
544 	SCARG(&bma, len) = SCARG(uap, len);
545 	SCARG(&bma, flags) = SCARG(uap, fl);
546 
547 	return sys___msync13(l, &bma, retval);
548 }
549 
550 int
551 linux_sys_mprotect(l, v, retval)
552 	struct lwp *l;
553 	void *v;
554 	register_t *retval;
555 {
556 	struct linux_sys_mprotect_args /* {
557 		syscallarg(const void *) start;
558 		syscallarg(unsigned long) len;
559 		syscallarg(int) prot;
560 	} */ *uap = v;
561 	unsigned long end, start = (unsigned long)SCARG(uap, start), len;
562 	int prot = SCARG(uap, prot);
563 	struct vm_map_entry *entry;
564 	struct vm_map *map = &l->l_proc->p_vmspace->vm_map;
565 
566 	if (start & PAGE_MASK)
567 		return EINVAL;
568 
569 	len = round_page(SCARG(uap, len));
570 	end = start + len;
571 
572 	if (end < start)
573 		return EINVAL;
574 	else if (end == start)
575 		return 0;
576 
577 	if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
578 		return EINVAL;
579 
580 	vm_map_lock(map);
581 #ifdef notdef
582 	VM_MAP_RANGE_CHECK(map, start, end);
583 #endif
584 	if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) {
585 		vm_map_unlock(map);
586 		return EFAULT;
587 	}
588 	vm_map_unlock(map);
589 	return uvm_map_protect(map, start, end, prot, FALSE);
590 }
591 
592 /*
593  * This code is partly stolen from src/lib/libc/compat-43/times.c
594  */
595 
596 #define	CONVTCK(r)	(r.tv_sec * hz + r.tv_usec / (1000000 / hz))
597 
598 int
599 linux_sys_times(l, v, retval)
600 	struct lwp *l;
601 	void *v;
602 	register_t *retval;
603 {
604 	struct linux_sys_times_args /* {
605 		syscallarg(struct times *) tms;
606 	} */ *uap = v;
607 	struct proc *p = l->l_proc;
608 	struct timeval t;
609 	int error, s;
610 
611 	if (SCARG(uap, tms)) {
612 		struct linux_tms ltms;
613 		struct rusage ru;
614 
615 		calcru(p, &ru.ru_utime, &ru.ru_stime, NULL);
616 		ltms.ltms_utime = CONVTCK(ru.ru_utime);
617 		ltms.ltms_stime = CONVTCK(ru.ru_stime);
618 
619 		ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime);
620 		ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime);
621 
622 		if ((error = copyout(&ltms, SCARG(uap, tms), sizeof ltms)))
623 			return error;
624 	}
625 
626 	s = splclock();
627 	timersub(&time, &boottime, &t);
628 	splx(s);
629 
630 	retval[0] = ((linux_clock_t)(CONVTCK(t)));
631 	return 0;
632 }
633 
634 #undef CONVTCK
635 
636 /*
637  * Linux 'readdir' call. This code is mostly taken from the
638  * SunOS getdents call (see compat/sunos/sunos_misc.c), though
639  * an attempt has been made to keep it a little cleaner (failing
640  * miserably, because of the cruft needed if count 1 is passed).
641  *
642  * The d_off field should contain the offset of the next valid entry,
643  * but in Linux it has the offset of the entry itself. We emulate
644  * that bug here.
645  *
646  * Read in BSD-style entries, convert them, and copy them out.
647  *
648  * Note that this doesn't handle union-mounted filesystems.
649  */
650 int
651 linux_sys_getdents(l, v, retval)
652 	struct lwp *l;
653 	void *v;
654 	register_t *retval;
655 {
656 	struct linux_sys_getdents_args /* {
657 		syscallarg(int) fd;
658 		syscallarg(struct linux_dirent *) dent;
659 		syscallarg(unsigned int) count;
660 	} */ *uap = v;
661 	struct proc *p = l->l_proc;
662 	struct dirent *bdp;
663 	struct vnode *vp;
664 	caddr_t	inp, buf;		/* BSD-format */
665 	int len, reclen;		/* BSD-format */
666 	caddr_t outp;			/* Linux-format */
667 	int resid, linux_reclen = 0;	/* Linux-format */
668 	struct file *fp;
669 	struct uio auio;
670 	struct iovec aiov;
671 	struct linux_dirent idb;
672 	off_t off;		/* true file offset */
673 	int buflen, error, eofflag, nbytes, oldcall;
674 	struct vattr va;
675 	off_t *cookiebuf = NULL, *cookie;
676 	int ncookies;
677 
678 	/* getvnode() will use the descriptor for us */
679 	if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0)
680 		return (error);
681 
682 	if ((fp->f_flag & FREAD) == 0) {
683 		error = EBADF;
684 		goto out1;
685 	}
686 
687 	vp = (struct vnode *)fp->f_data;
688 	if (vp->v_type != VDIR) {
689 		error = EINVAL;
690 		goto out1;
691 	}
692 
693 	if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)))
694 		goto out1;
695 
696 	nbytes = SCARG(uap, count);
697 	if (nbytes == 1) {	/* emulating old, broken behaviour */
698 		nbytes = sizeof (idb);
699 		buflen = max(va.va_blocksize, nbytes);
700 		oldcall = 1;
701 	} else {
702 		buflen = min(MAXBSIZE, nbytes);
703 		if (buflen < va.va_blocksize)
704 			buflen = va.va_blocksize;
705 		oldcall = 0;
706 	}
707 	buf = malloc(buflen, M_TEMP, M_WAITOK);
708 
709 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
710 	off = fp->f_offset;
711 again:
712 	aiov.iov_base = buf;
713 	aiov.iov_len = buflen;
714 	auio.uio_iov = &aiov;
715 	auio.uio_iovcnt = 1;
716 	auio.uio_rw = UIO_READ;
717 	auio.uio_segflg = UIO_SYSSPACE;
718 	auio.uio_procp = p;
719 	auio.uio_resid = buflen;
720 	auio.uio_offset = off;
721 	/*
722          * First we read into the malloc'ed buffer, then
723          * we massage it into user space, one record at a time.
724          */
725 	error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf,
726 	    &ncookies);
727 	if (error)
728 		goto out;
729 
730 	inp = buf;
731 	outp = (caddr_t)SCARG(uap, dent);
732 	resid = nbytes;
733 	if ((len = buflen - auio.uio_resid) == 0)
734 		goto eof;
735 
736 	for (cookie = cookiebuf; len > 0; len -= reclen) {
737 		bdp = (struct dirent *)inp;
738 		reclen = bdp->d_reclen;
739 		if (reclen & 3)
740 			panic("linux_readdir");
741 		if (bdp->d_fileno == 0) {
742 			inp += reclen;	/* it is a hole; squish it out */
743 			off = *cookie++;
744 			continue;
745 		}
746 		linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen);
747 		if (reclen > len || resid < linux_reclen) {
748 			/* entry too big for buffer, so just stop */
749 			outp++;
750 			break;
751 		}
752 		/*
753 		 * Massage in place to make a Linux-shaped dirent (otherwise
754 		 * we have to worry about touching user memory outside of
755 		 * the copyout() call).
756 		 */
757 		idb.d_ino = bdp->d_fileno;
758 		/*
759 		 * The old readdir() call misuses the offset and reclen fields.
760 		 */
761 		if (oldcall) {
762 			idb.d_off = (linux_off_t)linux_reclen;
763 			idb.d_reclen = (u_short)bdp->d_namlen;
764 		} else {
765 			if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) {
766 				compat_offseterr(vp, "linux_getdents");
767 				error = EINVAL;
768 				goto out;
769 			}
770 			idb.d_off = (linux_off_t)off;
771 			idb.d_reclen = (u_short)linux_reclen;
772 		}
773 		strcpy(idb.d_name, bdp->d_name);
774 		if ((error = copyout((caddr_t)&idb, outp, linux_reclen)))
775 			goto out;
776 		/* advance past this real entry */
777 		inp += reclen;
778 		off = *cookie++;	/* each entry points to itself */
779 		/* advance output past Linux-shaped entry */
780 		outp += linux_reclen;
781 		resid -= linux_reclen;
782 		if (oldcall)
783 			break;
784 	}
785 
786 	/* if we squished out the whole block, try again */
787 	if (outp == (caddr_t)SCARG(uap, dent))
788 		goto again;
789 	fp->f_offset = off;	/* update the vnode offset */
790 
791 	if (oldcall)
792 		nbytes = resid + linux_reclen;
793 
794 eof:
795 	*retval = nbytes - resid;
796 out:
797 	VOP_UNLOCK(vp, 0);
798 	if (cookiebuf)
799 		free(cookiebuf, M_TEMP);
800 	free(buf, M_TEMP);
801 out1:
802 	FILE_UNUSE(fp, p);
803 	return error;
804 }
805 
806 /*
807  * Even when just using registers to pass arguments to syscalls you can
808  * have 5 of them on the i386. So this newer version of select() does
809  * this.
810  */
811 int
812 linux_sys_select(l, v, retval)
813 	struct lwp *l;
814 	void *v;
815 	register_t *retval;
816 {
817 	struct linux_sys_select_args /* {
818 		syscallarg(int) nfds;
819 		syscallarg(fd_set *) readfds;
820 		syscallarg(fd_set *) writefds;
821 		syscallarg(fd_set *) exceptfds;
822 		syscallarg(struct timeval *) timeout;
823 	} */ *uap = v;
824 
825 	return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds),
826 	    SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout));
827 }
828 
829 /*
830  * Common code for the old and new versions of select(). A couple of
831  * things are important:
832  * 1) return the amount of time left in the 'timeout' parameter
833  * 2) select never returns ERESTART on Linux, always return EINTR
834  */
835 int
836 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout)
837 	struct lwp *l;
838 	register_t *retval;
839 	int nfds;
840 	fd_set *readfds, *writefds, *exceptfds;
841 	struct timeval *timeout;
842 {
843 	struct sys_select_args bsa;
844 	struct proc *p = l->l_proc;
845 	struct timeval tv0, tv1, utv, *tvp;
846 	caddr_t sg;
847 	int error;
848 
849 	SCARG(&bsa, nd) = nfds;
850 	SCARG(&bsa, in) = readfds;
851 	SCARG(&bsa, ou) = writefds;
852 	SCARG(&bsa, ex) = exceptfds;
853 	SCARG(&bsa, tv) = timeout;
854 
855 	/*
856 	 * Store current time for computation of the amount of
857 	 * time left.
858 	 */
859 	if (timeout) {
860 		if ((error = copyin(timeout, &utv, sizeof(utv))))
861 			return error;
862 		if (itimerfix(&utv)) {
863 			/*
864 			 * The timeval was invalid.  Convert it to something
865 			 * valid that will act as it does under Linux.
866 			 */
867 			sg = stackgap_init(p, 0);
868 			tvp = stackgap_alloc(p, &sg, sizeof(utv));
869 			utv.tv_sec += utv.tv_usec / 1000000;
870 			utv.tv_usec %= 1000000;
871 			if (utv.tv_usec < 0) {
872 				utv.tv_sec -= 1;
873 				utv.tv_usec += 1000000;
874 			}
875 			if (utv.tv_sec < 0)
876 				timerclear(&utv);
877 			if ((error = copyout(&utv, tvp, sizeof(utv))))
878 				return error;
879 			SCARG(&bsa, tv) = tvp;
880 		}
881 		microtime(&tv0);
882 	}
883 
884 	error = sys_select(l, &bsa, retval);
885 	if (error) {
886 		/*
887 		 * See fs/select.c in the Linux kernel.  Without this,
888 		 * Maelstrom doesn't work.
889 		 */
890 		if (error == ERESTART)
891 			error = EINTR;
892 		return error;
893 	}
894 
895 	if (timeout) {
896 		if (*retval) {
897 			/*
898 			 * Compute how much time was left of the timeout,
899 			 * by subtracting the current time and the time
900 			 * before we started the call, and subtracting
901 			 * that result from the user-supplied value.
902 			 */
903 			microtime(&tv1);
904 			timersub(&tv1, &tv0, &tv1);
905 			timersub(&utv, &tv1, &utv);
906 			if (utv.tv_sec < 0)
907 				timerclear(&utv);
908 		} else
909 			timerclear(&utv);
910 		if ((error = copyout(&utv, timeout, sizeof(utv))))
911 			return error;
912 	}
913 
914 	return 0;
915 }
916 
917 /*
918  * Get the process group of a certain process. Look it up
919  * and return the value.
920  */
921 int
922 linux_sys_getpgid(l, v, retval)
923 	struct lwp *l;
924 	void *v;
925 	register_t *retval;
926 {
927 	struct linux_sys_getpgid_args /* {
928 		syscallarg(int) pid;
929 	} */ *uap = v;
930 	struct proc *p = l->l_proc;
931 	struct proc *targp;
932 
933 	if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) {
934 		if ((targp = pfind(SCARG(uap, pid))) == 0)
935 			return ESRCH;
936 	}
937 	else
938 		targp = p;
939 
940 	retval[0] = targp->p_pgid;
941 	return 0;
942 }
943 
944 /*
945  * Set the 'personality' (emulation mode) for the current process. Only
946  * accept the Linux personality here (0). This call is needed because
947  * the Linux ELF crt0 issues it in an ugly kludge to make sure that
948  * ELF binaries run in Linux mode, not SVR4 mode.
949  */
950 int
951 linux_sys_personality(l, v, retval)
952 	struct lwp *l;
953 	void *v;
954 	register_t *retval;
955 {
956 	struct linux_sys_personality_args /* {
957 		syscallarg(int) per;
958 	} */ *uap = v;
959 
960 	if (SCARG(uap, per) != 0)
961 		return EINVAL;
962 	retval[0] = 0;
963 	return 0;
964 }
965 
966 #if defined(__i386__) || defined(__m68k__)
967 /*
968  * The calls are here because of type conversions.
969  */
970 int
971 linux_sys_setreuid16(l, v, retval)
972 	struct lwp *l;
973 	void *v;
974 	register_t *retval;
975 {
976 	struct linux_sys_setreuid16_args /* {
977 		syscallarg(int) ruid;
978 		syscallarg(int) euid;
979 	} */ *uap = v;
980 	struct sys_setreuid_args bsa;
981 
982 	SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
983 		(uid_t)-1 : SCARG(uap, ruid);
984 	SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
985 		(uid_t)-1 : SCARG(uap, euid);
986 
987 	return sys_setreuid(l, &bsa, retval);
988 }
989 
990 int
991 linux_sys_setregid16(l, v, retval)
992 	struct lwp *l;
993 	void *v;
994 	register_t *retval;
995 {
996 	struct linux_sys_setregid16_args /* {
997 		syscallarg(int) rgid;
998 		syscallarg(int) egid;
999 	} */ *uap = v;
1000 	struct sys_setregid_args bsa;
1001 
1002 	SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1003 		(uid_t)-1 : SCARG(uap, rgid);
1004 	SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1005 		(uid_t)-1 : SCARG(uap, egid);
1006 
1007 	return sys_setregid(l, &bsa, retval);
1008 }
1009 
1010 int
1011 linux_sys_setresuid16(l, v, retval)
1012 	struct lwp *l;
1013 	void *v;
1014 	register_t *retval;
1015 {
1016 	struct linux_sys_setresuid16_args /* {
1017 		syscallarg(uid_t) ruid;
1018 		syscallarg(uid_t) euid;
1019 		syscallarg(uid_t) suid;
1020 	} */ *uap = v;
1021 	struct linux_sys_setresuid16_args lsa;
1022 
1023 	SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ?
1024 		(uid_t)-1 : SCARG(uap, ruid);
1025 	SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ?
1026 		(uid_t)-1 : SCARG(uap, euid);
1027 	SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ?
1028 		(uid_t)-1 : SCARG(uap, suid);
1029 
1030 	return linux_sys_setresuid(l, &lsa, retval);
1031 }
1032 
1033 int
1034 linux_sys_setresgid16(l, v, retval)
1035 	struct lwp *l;
1036 	void *v;
1037 	register_t *retval;
1038 {
1039 	struct linux_sys_setresgid16_args /* {
1040 		syscallarg(gid_t) rgid;
1041 		syscallarg(gid_t) egid;
1042 		syscallarg(gid_t) sgid;
1043 	} */ *uap = v;
1044 	struct linux_sys_setresgid16_args lsa;
1045 
1046 	SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ?
1047 		(gid_t)-1 : SCARG(uap, rgid);
1048 	SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ?
1049 		(gid_t)-1 : SCARG(uap, egid);
1050 	SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ?
1051 		(gid_t)-1 : SCARG(uap, sgid);
1052 
1053 	return linux_sys_setresgid(l, &lsa, retval);
1054 }
1055 
1056 int
1057 linux_sys_getgroups16(l, v, retval)
1058 	struct lwp *l;
1059 	void *v;
1060 	register_t *retval;
1061 {
1062 	struct linux_sys_getgroups16_args /* {
1063 		syscallarg(int) gidsetsize;
1064 		syscallarg(linux_gid_t *) gidset;
1065 	} */ *uap = v;
1066 	struct proc *p = l->l_proc;
1067 	caddr_t sg;
1068 	int n, error, i;
1069 	struct sys_getgroups_args bsa;
1070 	gid_t *bset, *kbset;
1071 	linux_gid_t *lset;
1072 	struct pcred *pc = p->p_cred;
1073 
1074 	n = SCARG(uap, gidsetsize);
1075 	if (n < 0)
1076 		return EINVAL;
1077 	error = 0;
1078 	bset = kbset = NULL;
1079 	lset = NULL;
1080 	if (n > 0) {
1081 		n = min(pc->pc_ucred->cr_ngroups, n);
1082 		sg = stackgap_init(p, 0);
1083 		bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1084 		kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1085 		lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1086 		if (bset == NULL || kbset == NULL || lset == NULL)
1087 			return ENOMEM;
1088 		SCARG(&bsa, gidsetsize) = n;
1089 		SCARG(&bsa, gidset) = bset;
1090 		error = sys_getgroups(l, &bsa, retval);
1091 		if (error != 0)
1092 			goto out;
1093 		error = copyin(bset, kbset, n * sizeof (gid_t));
1094 		if (error != 0)
1095 			goto out;
1096 		for (i = 0; i < n; i++)
1097 			lset[i] = (linux_gid_t)kbset[i];
1098 		error = copyout(lset, SCARG(uap, gidset),
1099 		    n * sizeof (linux_gid_t));
1100 	} else
1101 		*retval = pc->pc_ucred->cr_ngroups;
1102 out:
1103 	if (kbset != NULL)
1104 		free(kbset, M_TEMP);
1105 	if (lset != NULL)
1106 		free(lset, M_TEMP);
1107 	return error;
1108 }
1109 
1110 int
1111 linux_sys_setgroups16(l, v, retval)
1112 	struct lwp *l;
1113 	void *v;
1114 	register_t *retval;
1115 {
1116 	struct linux_sys_setgroups16_args /* {
1117 		syscallarg(int) gidsetsize;
1118 		syscallarg(linux_gid_t *) gidset;
1119 	} */ *uap = v;
1120 	struct proc *p = l->l_proc;
1121 	caddr_t sg;
1122 	int n;
1123 	int error, i;
1124 	struct sys_setgroups_args bsa;
1125 	gid_t *bset, *kbset;
1126 	linux_gid_t *lset;
1127 
1128 	n = SCARG(uap, gidsetsize);
1129 	if (n < 0 || n > NGROUPS)
1130 		return EINVAL;
1131 	sg = stackgap_init(p, 0);
1132 	bset = stackgap_alloc(p, &sg, n * sizeof (gid_t));
1133 	lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK);
1134 	kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK);
1135 	if (lset == NULL || bset == NULL)
1136 		return ENOMEM;
1137 	error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t));
1138 	if (error != 0)
1139 		goto out;
1140 	for (i = 0; i < n; i++)
1141 		kbset[i] = (gid_t)lset[i];
1142 	error = copyout(kbset, bset, n * sizeof (gid_t));
1143 	if (error != 0)
1144 		goto out;
1145 	SCARG(&bsa, gidsetsize) = n;
1146 	SCARG(&bsa, gidset) = bset;
1147 	error = sys_setgroups(l, &bsa, retval);
1148 
1149 out:
1150 	if (lset != NULL)
1151 		free(lset, M_TEMP);
1152 	if (kbset != NULL)
1153 		free(kbset, M_TEMP);
1154 
1155 	return error;
1156 }
1157 
1158 #endif /* __i386__ || __m68k__ */
1159 
1160 /*
1161  * We have nonexistent fsuid equal to uid.
1162  * If modification is requested, refuse.
1163  */
1164 int
1165 linux_sys_setfsuid(l, v, retval)
1166 	 struct lwp *l;
1167 	 void *v;
1168 	 register_t *retval;
1169 {
1170 	 struct linux_sys_setfsuid_args /* {
1171 		 syscallarg(uid_t) uid;
1172 	 } */ *uap = v;
1173 	 struct proc *p = l->l_proc;
1174 	 uid_t uid;
1175 
1176 	 uid = SCARG(uap, uid);
1177 	 if (p->p_cred->p_ruid != uid)
1178 		 return sys_nosys(l, v, retval);
1179 	 else
1180 		 return (0);
1181 }
1182 
1183 /* XXX XXX XXX */
1184 #ifndef alpha
1185 int
1186 linux_sys_getfsuid(l, v, retval)
1187 	struct lwp *l;
1188 	void *v;
1189 	register_t *retval;
1190 {
1191 	return sys_getuid(l, v, retval);
1192 }
1193 #endif
1194 
1195 int
1196 linux_sys_setresuid(l, v, retval)
1197 	struct lwp *l;
1198 	void *v;
1199 	register_t *retval;
1200 {
1201 	struct linux_sys_setresuid_args /* {
1202 		syscallarg(uid_t) ruid;
1203 		syscallarg(uid_t) euid;
1204 		syscallarg(uid_t) suid;
1205 	} */ *uap = v;
1206 
1207 	/*
1208 	 * Note: These checks are a little different than the NetBSD
1209 	 * setreuid(2) call performs.  This precisely follows the
1210 	 * behavior of the Linux kernel.
1211 	 */
1212 
1213 	return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid),
1214 			    SCARG(uap, suid),
1215 			    ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S |
1216 			    ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S |
1217 			    ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S );
1218 }
1219 
1220 int
1221 linux_sys_getresuid(l, v, retval)
1222 	struct lwp *l;
1223 	void *v;
1224 	register_t *retval;
1225 {
1226 	struct linux_sys_getresuid_args /* {
1227 		syscallarg(uid_t *) ruid;
1228 		syscallarg(uid_t *) euid;
1229 		syscallarg(uid_t *) suid;
1230 	} */ *uap = v;
1231 	struct proc *p = l->l_proc;
1232 	struct pcred *pc = p->p_cred;
1233 	int error;
1234 
1235 	/*
1236 	 * Linux copies these values out to userspace like so:
1237 	 *
1238 	 *	1. Copy out ruid.
1239 	 *	2. If that succeeds, copy out euid.
1240 	 *	3. If both of those succeed, copy out suid.
1241 	 */
1242 	if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid),
1243 			     sizeof(uid_t))) != 0)
1244 		return (error);
1245 
1246 	if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid),
1247 			     sizeof(uid_t))) != 0)
1248 		return (error);
1249 
1250 	return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t)));
1251 }
1252 
1253 int
1254 linux_sys_ptrace(l, v, retval)
1255 	struct lwp *l;
1256 	void *v;
1257 	register_t *retval;
1258 {
1259 	struct linux_sys_ptrace_args /* {
1260 		i386, m68k, powerpc: T=int
1261 		alpha: T=long
1262 		syscallarg(T) request;
1263 		syscallarg(T) pid;
1264 		syscallarg(T) addr;
1265 		syscallarg(T) data;
1266 	} */ *uap = v;
1267 	const int *ptr;
1268 	int request;
1269 	int error;
1270 
1271 	ptr = linux_ptrace_request_map;
1272 	request = SCARG(uap, request);
1273 	while (*ptr != -1)
1274 		if (*ptr++ == request) {
1275 			struct sys_ptrace_args pta;
1276 
1277 			SCARG(&pta, req) = *ptr;
1278 			SCARG(&pta, pid) = SCARG(uap, pid);
1279 			SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr);
1280 			SCARG(&pta, data) = SCARG(uap, data);
1281 
1282 			/*
1283 			 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually
1284 			 * to continue where the process left off previously.
1285 			 * The same thing is achieved by addr == (caddr_t) 1
1286 			 * on NetBSD, so rewrite 'addr' appropriately.
1287 			 */
1288 			if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0)
1289 				SCARG(&pta, addr) = (caddr_t) 1;
1290 
1291 			error = sys_ptrace(l, &pta, retval);
1292 			if (error)
1293 				return error;
1294 			switch (request) {
1295 			case LINUX_PTRACE_PEEKTEXT:
1296 			case LINUX_PTRACE_PEEKDATA:
1297 				error = copyout (retval,
1298 				    (caddr_t)SCARG(uap, data), sizeof *retval);
1299 				*retval = SCARG(uap, data);
1300 				break;
1301 			default:
1302 				break;
1303 			}
1304 			return error;
1305 		}
1306 		else
1307 			ptr++;
1308 
1309 	return LINUX_SYS_PTRACE_ARCH(l, uap, retval);
1310 }
1311 
1312 int
1313 linux_sys_reboot(struct lwp *l, void *v, register_t *retval)
1314 {
1315 	struct linux_sys_reboot_args /* {
1316 		syscallarg(int) magic1;
1317 		syscallarg(int) magic2;
1318 		syscallarg(int) cmd;
1319 		syscallarg(void *) arg;
1320 	} */ *uap = v;
1321 	struct sys_reboot_args /* {
1322 		syscallarg(int) opt;
1323 		syscallarg(char *) bootstr;
1324 	} */ sra;
1325 	struct proc *p = l->l_proc;
1326 	int error;
1327 
1328 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1329 		return(error);
1330 
1331 	if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1)
1332 		return(EINVAL);
1333 	if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 &&
1334 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A &&
1335 	    SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B)
1336 		return(EINVAL);
1337 
1338 	switch (SCARG(uap, cmd)) {
1339 	case LINUX_REBOOT_CMD_RESTART:
1340 		SCARG(&sra, opt) = RB_AUTOBOOT;
1341 		break;
1342 	case LINUX_REBOOT_CMD_HALT:
1343 		SCARG(&sra, opt) = RB_HALT;
1344 		break;
1345 	case LINUX_REBOOT_CMD_POWER_OFF:
1346 		SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN;
1347 		break;
1348 	case LINUX_REBOOT_CMD_RESTART2:
1349 		/* Reboot with an argument. */
1350 		SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING;
1351 		SCARG(&sra, bootstr) = SCARG(uap, arg);
1352 		break;
1353 	case LINUX_REBOOT_CMD_CAD_ON:
1354 		return(EINVAL);	/* We don't implement ctrl-alt-delete */
1355 	case LINUX_REBOOT_CMD_CAD_OFF:
1356 		return(0);
1357 	default:
1358 		return(EINVAL);
1359 	}
1360 
1361 	return(sys_reboot(l, &sra, retval));
1362 }
1363 
1364 /*
1365  * Copy of compat_12_sys_swapon().
1366  */
1367 int
1368 linux_sys_swapon(l, v, retval)
1369 	struct lwp *l;
1370 	void *v;
1371 	register_t *retval;
1372 {
1373 	struct sys_swapctl_args ua;
1374 	struct linux_sys_swapon_args /* {
1375 		syscallarg(const char *) name;
1376 	} */ *uap = v;
1377 
1378 	SCARG(&ua, cmd) = SWAP_ON;
1379 	SCARG(&ua, arg) = (void *)SCARG(uap, name);
1380 	SCARG(&ua, misc) = 0;	/* priority */
1381 	return (sys_swapctl(l, &ua, retval));
1382 }
1383 
1384 /*
1385  * Stop swapping to the file or block device specified by path.
1386  */
1387 int
1388 linux_sys_swapoff(l, v, retval)
1389 	struct lwp *l;
1390 	void *v;
1391 	register_t *retval;
1392 {
1393 	struct sys_swapctl_args ua;
1394 	struct linux_sys_swapoff_args /* {
1395 		syscallarg(const char *) path;
1396 	} */ *uap = v;
1397 
1398 	SCARG(&ua, cmd) = SWAP_OFF;
1399 	SCARG(&ua, arg) = (void *)SCARG(uap, path);
1400 	return (sys_swapctl(l, &ua, retval));
1401 }
1402 
1403 /*
1404  * Copy of compat_09_sys_setdomainname()
1405  */
1406 /* ARGSUSED */
1407 int
1408 linux_sys_setdomainname(l, v, retval)
1409 	struct lwp *l;
1410 	void *v;
1411 	register_t *retval;
1412 {
1413 	struct linux_sys_setdomainname_args /* {
1414 		syscallarg(char *) domainname;
1415 		syscallarg(int) len;
1416 	} */ *uap = v;
1417 	struct proc *p = l->l_proc;
1418 	int name;
1419 	int error;
1420 
1421 	if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
1422 		return (error);
1423 	name = KERN_DOMAINNAME;
1424 	return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname),
1425 			    SCARG(uap, len), p));
1426 }
1427 
1428 /*
1429  * sysinfo()
1430  */
1431 /* ARGSUSED */
1432 int
1433 linux_sys_sysinfo(l, v, retval)
1434 	struct lwp *l;
1435 	void *v;
1436 	register_t *retval;
1437 {
1438 	struct linux_sys_sysinfo_args /* {
1439 		syscallarg(struct linux_sysinfo *) arg;
1440 	} */ *uap = v;
1441 	struct linux_sysinfo si;
1442 	struct loadavg *la;
1443 
1444 	si.uptime = time.tv_sec - boottime.tv_sec;
1445 	la = &averunnable;
1446 	si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1447 	si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1448 	si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale;
1449 	si.totalram = ctob(physmem);
1450 	si.freeram = uvmexp.free * uvmexp.pagesize;
1451 	si.sharedram = 0;	/* XXX */
1452 	si.bufferram = uvmexp.filepages * uvmexp.pagesize;
1453 	si.totalswap = uvmexp.swpages * uvmexp.pagesize;
1454 	si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize;
1455 	si.procs = nprocs;
1456 
1457 	/* The following are only present in newer Linux kernels. */
1458 	si.totalbig = 0;
1459 	si.freebig = 0;
1460 	si.mem_unit = 1;
1461 
1462 	return (copyout(&si, SCARG(uap, arg), sizeof si));
1463 }
1464 
1465 #define bsd_to_linux_rlimit1(l, b, f) \
1466     (l)->f = ((b)->f == RLIM_INFINITY || \
1467 	     ((b)->f & 0xffffffff00000000ULL) != 0) ? \
1468     LINUX_RLIM_INFINITY : (int32_t)(b)->f
1469 #define bsd_to_linux_rlimit(l, b) \
1470     bsd_to_linux_rlimit1(l, b, rlim_cur); \
1471     bsd_to_linux_rlimit1(l, b, rlim_max)
1472 
1473 #define linux_to_bsd_rlimit1(b, l, f) \
1474     (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f
1475 #define linux_to_bsd_rlimit(b, l) \
1476     linux_to_bsd_rlimit1(b, l, rlim_cur); \
1477     linux_to_bsd_rlimit1(b, l, rlim_max)
1478 
1479 static int
1480 linux_to_bsd_limit(lim)
1481 	int lim;
1482 {
1483 	switch (lim) {
1484 	case LINUX_RLIMIT_CPU:
1485 		return RLIMIT_CPU;
1486 	case LINUX_RLIMIT_FSIZE:
1487 		return RLIMIT_FSIZE;
1488 	case LINUX_RLIMIT_DATA:
1489 		return RLIMIT_DATA;
1490 	case LINUX_RLIMIT_STACK:
1491 		return RLIMIT_STACK;
1492 	case LINUX_RLIMIT_CORE:
1493 		return RLIMIT_CORE;
1494 	case LINUX_RLIMIT_RSS:
1495 		return RLIMIT_RSS;
1496 	case LINUX_RLIMIT_NPROC:
1497 		return RLIMIT_NPROC;
1498 	case LINUX_RLIMIT_NOFILE:
1499 		return RLIMIT_NOFILE;
1500 	case LINUX_RLIMIT_MEMLOCK:
1501 		return RLIMIT_MEMLOCK;
1502 	case LINUX_RLIMIT_AS:
1503 	case LINUX_RLIMIT_LOCKS:
1504 		return -EOPNOTSUPP;
1505 	default:
1506 		return -EINVAL;
1507 	}
1508 }
1509 
1510 
1511 int
1512 linux_sys_getrlimit(l, v, retval)
1513 	struct lwp *l;
1514 	void *v;
1515 	register_t *retval;
1516 {
1517 	struct linux_sys_getrlimit_args /* {
1518 		syscallarg(int) which;
1519 		syscallarg(struct orlimit *) rlp;
1520 	} */ *uap = v;
1521 	struct proc *p = l->l_proc;
1522 	caddr_t sg = stackgap_init(p, 0);
1523 	struct sys_getrlimit_args ap;
1524 	struct rlimit rl;
1525 	struct orlimit orl;
1526 	int error;
1527 
1528 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1529 	if ((error = SCARG(&ap, which)) < 0)
1530 		return -error;
1531 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1532 	if ((error = sys_getrlimit(l, &ap, retval)) != 0)
1533 		return error;
1534 	if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0)
1535 		return error;
1536 	bsd_to_linux_rlimit(&orl, &rl);
1537 	return copyout(&orl, SCARG(uap, rlp), sizeof(orl));
1538 }
1539 
1540 int
1541 linux_sys_setrlimit(l, v, retval)
1542 	struct lwp *l;
1543 	void *v;
1544 	register_t *retval;
1545 {
1546 	struct linux_sys_setrlimit_args /* {
1547 		syscallarg(int) which;
1548 		syscallarg(struct orlimit *) rlp;
1549 	} */ *uap = v;
1550 	struct proc *p = l->l_proc;
1551 	caddr_t sg = stackgap_init(p, 0);
1552 	struct sys_setrlimit_args ap;
1553 	struct rlimit rl;
1554 	struct orlimit orl;
1555 	int error;
1556 
1557 	SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which));
1558 	SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl);
1559 	if ((error = SCARG(&ap, which)) < 0)
1560 		return -error;
1561 	if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0)
1562 		return error;
1563 	linux_to_bsd_rlimit(&rl, &orl);
1564 	/* XXX: alpha complains about this */
1565 	if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0)
1566 		return error;
1567 	return sys_setrlimit(l, &ap, retval);
1568 }
1569 
1570 #ifndef __mips__
1571 /* XXX: this doesn't look 100% common, at least mips doesn't have it */
1572 int
1573 linux_sys_ugetrlimit(l, v, retval)
1574 	struct lwp *l;
1575 	void *v;
1576 	register_t *retval;
1577 {
1578 	return linux_sys_getrlimit(l, v, retval);
1579 }
1580 #endif
1581 
1582 /*
1583  * This gets called for unsupported syscalls. The difference to sys_nosys()
1584  * is that process does not get SIGSYS, the call just returns with ENOSYS.
1585  * This is the way Linux does it and glibc depends on this behaviour.
1586  */
1587 int
1588 linux_sys_nosys(l, v, retval)
1589 	struct lwp *l;
1590 	void *v;
1591 	register_t *retval;
1592 {
1593 	return (ENOSYS);
1594 }
1595