xref: /freebsd/sys/kern/kern_sysctl.c (revision 02f2e93b60c2b91feac8f45c4c889a5a8e40d8a2)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Mike Karels at Berkeley Software Design, Inc.
7  *
8  * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
9  * project, to make these variables more userfriendly.
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 University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  *
39  *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
40  * $Id: kern_sysctl.c,v 1.71 1997/10/11 18:31:24 phk Exp $
41  */
42 
43 #include <sys/param.h>
44 #include <sys/buf.h>
45 #include <sys/kernel.h>
46 #include <sys/sysctl.h>
47 #include <sys/malloc.h>
48 #include <sys/proc.h>
49 #include <sys/systm.h>
50 #include <sys/sysproto.h>
51 #include <vm/vm.h>
52 #include <vm/vm_extern.h>
53 
54 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
55 
56 /*
57  * Locking and stats
58  */
59 static struct sysctl_lock {
60 	int	sl_lock;
61 	int	sl_want;
62 	int	sl_locked;
63 } memlock;
64 
65 static int sysctl_root SYSCTL_HANDLER_ARGS;
66 
67 extern struct linker_set sysctl_;
68 
69 /*
70  * Initialization of the MIB tree.
71  *
72  * Order by number in each linker_set.
73  */
74 
75 static int
76 sysctl_order_cmp(const void *a, const void *b)
77 {
78 	struct sysctl_oid const * const *pa;
79 	struct sysctl_oid const * const *pb;
80 
81 	pa = (struct sysctl_oid const * const *)a;
82 	pb = (struct sysctl_oid const * const *)b;
83 	if (*pa == NULL)
84 		return (1);
85 	if (*pb == NULL)
86 		return (-1);
87 	return ((*pa)->oid_number - (*pb)->oid_number);
88 }
89 
90 static void
91 sysctl_order(void *arg)
92 {
93 	int j, k;
94 	struct linker_set *l = (struct linker_set *) arg;
95 	struct sysctl_oid **oidpp;
96 
97 	/* First, find the highest oid we have */
98 	j = l->ls_length;
99 	oidpp = (struct sysctl_oid **) l->ls_items;
100 	for (k = 0; j--; oidpp++) {
101 		if ((*oidpp)->oid_arg1 == arg) {
102 			*oidpp = 0;
103 			continue;
104 		}
105 		if (*oidpp && (*oidpp)->oid_number > k)
106 			k = (*oidpp)->oid_number;
107 	}
108 
109 	/* Next, replace all OID_AUTO oids with new numbers */
110 	j = l->ls_length;
111 	oidpp = (struct sysctl_oid **) l->ls_items;
112 	k += 100;
113 	for (; j--; oidpp++)
114 		if (*oidpp && (*oidpp)->oid_number == OID_AUTO)
115 			(*oidpp)->oid_number = k++;
116 
117 	/* Finally: sort by oid */
118 	j = l->ls_length;
119 	oidpp = (struct sysctl_oid **) l->ls_items;
120 	for (; j--; oidpp++) {
121 		if (!*oidpp)
122 			continue;
123 		if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE)
124 			if (!(*oidpp)->oid_handler)
125 				sysctl_order((*oidpp)->oid_arg1);
126 	}
127 	qsort(l->ls_items, l->ls_length, sizeof l->ls_items[0],
128 		sysctl_order_cmp);
129 }
130 
131 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_order, &sysctl_);
132 
133 /*
134  * "Staff-functions"
135  *
136  * These functions implement a presently undocumented interface
137  * used by the sysctl program to walk the tree, and get the type
138  * so it can print the value.
139  * This interface is under work and consideration, and should probably
140  * be killed with a big axe by the first person who can find the time.
141  * (be aware though, that the proper interface isn't as obvious as it
142  * may seem, there are various conflicting requirements.
143  *
144  * {0,0}	printf the entire MIB-tree.
145  * {0,1,...}	return the name of the "..." OID.
146  * {0,2,...}	return the next OID.
147  * {0,3}	return the OID of the name in "new"
148  * {0,4,...}	return the kind & format info for the "..." OID.
149  */
150 
151 static void
152 sysctl_sysctl_debug_dump_node(struct linker_set *l, int i)
153 {
154 	int j, k;
155 	struct sysctl_oid **oidpp;
156 
157 	j = l->ls_length;
158 	oidpp = (struct sysctl_oid **) l->ls_items;
159 	for (; j--; oidpp++) {
160 
161 		if (!*oidpp)
162 			continue;
163 
164 		for (k=0; k<i; k++)
165 			printf(" ");
166 
167 		printf("%d %s ", (*oidpp)->oid_number, (*oidpp)->oid_name);
168 
169 		printf("%c%c",
170 			(*oidpp)->oid_kind & CTLFLAG_RD ? 'R':' ',
171 			(*oidpp)->oid_kind & CTLFLAG_WR ? 'W':' ');
172 
173 		if ((*oidpp)->oid_handler)
174 			printf(" *Handler");
175 
176 		switch ((*oidpp)->oid_kind & CTLTYPE) {
177 			case CTLTYPE_NODE:
178 				printf(" Node\n");
179 				if (!(*oidpp)->oid_handler) {
180 					sysctl_sysctl_debug_dump_node(
181 						(*oidpp)->oid_arg1, i+2);
182 				}
183 				break;
184 			case CTLTYPE_INT:    printf(" Int\n"); break;
185 			case CTLTYPE_STRING: printf(" String\n"); break;
186 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
187 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
188 			default:	     printf("\n");
189 		}
190 
191 	}
192 }
193 
194 static int
195 sysctl_sysctl_debug SYSCTL_HANDLER_ARGS
196 {
197 	sysctl_sysctl_debug_dump_node(&sysctl_, 0);
198 	return ENOENT;
199 }
200 
201 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
202 	0, 0, sysctl_sysctl_debug, "-", "");
203 
204 static int
205 sysctl_sysctl_name SYSCTL_HANDLER_ARGS
206 {
207 	int *name = (int *) arg1;
208 	u_int namelen = arg2;
209 	int i, j, error = 0;
210 	struct sysctl_oid **oidpp;
211 	struct linker_set *lsp = &sysctl_;
212 	char buf[10];
213 
214 	while (namelen) {
215 		if (!lsp) {
216 			sprintf(buf,"%d",*name);
217 			if (req->oldidx)
218 				error = SYSCTL_OUT(req, ".", 1);
219 			if (!error)
220 				error = SYSCTL_OUT(req, buf, strlen(buf));
221 			if (error)
222 				return (error);
223 			namelen--;
224 			name++;
225 			continue;
226 		}
227 		oidpp = (struct sysctl_oid **) lsp->ls_items;
228 		j = lsp->ls_length;
229 		lsp = 0;
230 		for (i = 0; i < j; i++, oidpp++) {
231 			if (*oidpp && ((*oidpp)->oid_number != *name))
232 				continue;
233 
234 			if (req->oldidx)
235 				error = SYSCTL_OUT(req, ".", 1);
236 			if (!error)
237 				error = SYSCTL_OUT(req, (*oidpp)->oid_name,
238 					strlen((*oidpp)->oid_name));
239 			if (error)
240 				return (error);
241 
242 			namelen--;
243 			name++;
244 
245 			if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
246 				break;
247 
248 			if ((*oidpp)->oid_handler)
249 				break;
250 
251 			lsp = (struct linker_set*)(*oidpp)->oid_arg1;
252 			break;
253 		}
254 	}
255 	return (SYSCTL_OUT(req, "", 1));
256 }
257 
258 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
259 
260 static int
261 sysctl_sysctl_next_ls (struct linker_set *lsp, int *name, u_int namelen,
262 	int *next, int *len, int level, struct sysctl_oid **oidp)
263 {
264 	int i, j;
265 	struct sysctl_oid **oidpp;
266 
267 	oidpp = (struct sysctl_oid **) lsp->ls_items;
268 	j = lsp->ls_length;
269 	*len = level;
270 	for (i = 0; i < j; i++, oidpp++) {
271 		if (!*oidpp)
272 			continue;
273 
274 		*next = (*oidpp)->oid_number;
275 		*oidp = *oidpp;
276 
277 		if (!namelen) {
278 			if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
279 				return 0;
280 			if ((*oidpp)->oid_handler)
281 				/* We really should call the handler here...*/
282 				return 0;
283 			lsp = (struct linker_set*)(*oidpp)->oid_arg1;
284 			if (!sysctl_sysctl_next_ls (lsp, 0, 0, next+1,
285 				len, level+1, oidp))
286 				return 0;
287 			goto next;
288 		}
289 
290 		if ((*oidpp)->oid_number < *name)
291 			continue;
292 
293 		if ((*oidpp)->oid_number > *name) {
294 			if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
295 				return 0;
296 			if ((*oidpp)->oid_handler)
297 				return 0;
298 			lsp = (struct linker_set*)(*oidpp)->oid_arg1;
299 			if (!sysctl_sysctl_next_ls (lsp, name+1, namelen-1,
300 				next+1, len, level+1, oidp))
301 				return (0);
302 			goto next;
303 		}
304 		if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
305 			continue;
306 
307 		if ((*oidpp)->oid_handler)
308 			continue;
309 
310 		lsp = (struct linker_set*)(*oidpp)->oid_arg1;
311 		if (!sysctl_sysctl_next_ls (lsp, name+1, namelen-1, next+1,
312 			len, level+1, oidp))
313 			return (0);
314 	next:
315 		namelen = 1;
316 		*len = level;
317 	}
318 	return 1;
319 }
320 
321 static int
322 sysctl_sysctl_next SYSCTL_HANDLER_ARGS
323 {
324 	int *name = (int *) arg1;
325 	u_int namelen = arg2;
326 	int i, j, error;
327 	struct sysctl_oid *oid;
328 	struct linker_set *lsp = &sysctl_;
329 	int newoid[CTL_MAXNAME];
330 
331 	i = sysctl_sysctl_next_ls (lsp, name, namelen, newoid, &j, 1, &oid);
332 	if (i)
333 		return ENOENT;
334 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
335 	return (error);
336 }
337 
338 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
339 
340 static int
341 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidp)
342 {
343 	int i, j;
344 	struct sysctl_oid **oidpp;
345 	struct linker_set *lsp = &sysctl_;
346 	char *p;
347 
348 	if (!*name)
349 		return ENOENT;
350 
351 	p = name + strlen(name) - 1 ;
352 	if (*p == '.')
353 		*p = '\0';
354 
355 	*len = 0;
356 
357 	for (p = name; *p && *p != '.'; p++)
358 		;
359 	i = *p;
360 	if (i == '.')
361 		*p = '\0';
362 
363 	j = lsp->ls_length;
364 	oidpp = (struct sysctl_oid **) lsp->ls_items;
365 
366 	while (j-- && *len < CTL_MAXNAME) {
367 		if (!*oidpp)
368 			continue;
369 		if (strcmp(name, (*oidpp)->oid_name)) {
370 			oidpp++;
371 			continue;
372 		}
373 		*oid++ = (*oidpp)->oid_number;
374 		(*len)++;
375 
376 		if (!i) {
377 			if (oidp)
378 				*oidp = *oidpp;
379 			return (0);
380 		}
381 
382 		if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
383 			break;
384 
385 		if ((*oidpp)->oid_handler)
386 			break;
387 
388 		lsp = (struct linker_set*)(*oidpp)->oid_arg1;
389 		j = lsp->ls_length;
390 		oidpp = (struct sysctl_oid **)lsp->ls_items;
391 		name = p+1;
392 		for (p = name; *p && *p != '.'; p++)
393 				;
394 		i = *p;
395 		if (i == '.')
396 			*p = '\0';
397 	}
398 	return ENOENT;
399 }
400 
401 static int
402 sysctl_sysctl_name2oid SYSCTL_HANDLER_ARGS
403 {
404 	char *p;
405 	int error, oid[CTL_MAXNAME], len;
406 	struct sysctl_oid *op = 0;
407 
408 	if (!req->newlen)
409 		return ENOENT;
410 
411 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
412 
413 	error = SYSCTL_IN(req, p, req->newlen);
414 	if (error) {
415 		free(p, M_SYSCTL);
416 		return (error);
417 	}
418 
419 	p [req->newlen] = '\0';
420 
421 	error = name2oid(p, oid, &len, &op);
422 
423 	free(p, M_SYSCTL);
424 
425 	if (error)
426 		return (error);
427 
428 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
429 	return (error);
430 }
431 
432 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
433 	sysctl_sysctl_name2oid, "I", "");
434 
435 static int
436 sysctl_sysctl_oidfmt SYSCTL_HANDLER_ARGS
437 {
438 	int *name = (int *) arg1, error;
439 	u_int namelen = arg2;
440 	int indx, j;
441 	struct sysctl_oid **oidpp;
442 	struct linker_set *lsp = &sysctl_;
443 
444 	j = lsp->ls_length;
445 	oidpp = (struct sysctl_oid **) lsp->ls_items;
446 
447 	indx = 0;
448 	while (j-- && indx < CTL_MAXNAME) {
449 		if (*oidpp && ((*oidpp)->oid_number == name[indx])) {
450 			indx++;
451 			if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
452 				if ((*oidpp)->oid_handler)
453 					goto found;
454 				if (indx == namelen)
455 					goto found;
456 				lsp = (struct linker_set*)(*oidpp)->oid_arg1;
457 				j = lsp->ls_length;
458 				oidpp = (struct sysctl_oid **)lsp->ls_items;
459 			} else {
460 				if (indx != namelen)
461 					return EISDIR;
462 				goto found;
463 			}
464 		} else {
465 			oidpp++;
466 		}
467 	}
468 	return ENOENT;
469 found:
470 	if (!(*oidpp)->oid_fmt)
471 		return ENOENT;
472 	error = SYSCTL_OUT(req,
473 		&(*oidpp)->oid_kind, sizeof((*oidpp)->oid_kind));
474 	if (!error)
475 		error = SYSCTL_OUT(req, (*oidpp)->oid_fmt,
476 			strlen((*oidpp)->oid_fmt)+1);
477 	return (error);
478 }
479 
480 
481 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
482 
483 /*
484  * Default "handler" functions.
485  */
486 
487 /*
488  * Handle an integer, signed or unsigned.
489  * Two cases:
490  *     a variable:  point arg1 at it.
491  *     a constant:  pass it in arg2.
492  */
493 
494 int
495 sysctl_handle_int SYSCTL_HANDLER_ARGS
496 {
497 	int error = 0;
498 
499 	if (arg1)
500 		error = SYSCTL_OUT(req, arg1, sizeof(int));
501 	else
502 		error = SYSCTL_OUT(req, &arg2, sizeof(int));
503 
504 	if (error || !req->newptr)
505 		return (error);
506 
507 	if (!arg1)
508 		error = EPERM;
509 	else
510 		error = SYSCTL_IN(req, arg1, sizeof(int));
511 	return (error);
512 }
513 
514 /*
515  * Handle our generic '\0' terminated 'C' string.
516  * Two cases:
517  * 	a variable string:  point arg1 at it, arg2 is max length.
518  * 	a constant string:  point arg1 at it, arg2 is zero.
519  */
520 
521 int
522 sysctl_handle_string SYSCTL_HANDLER_ARGS
523 {
524 	int error=0;
525 
526 	error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
527 
528 	if (error || !req->newptr || !arg2)
529 		return (error);
530 
531 	if ((req->newlen - req->newidx) > arg2) {
532 		error = E2BIG;
533 	} else {
534 		arg2 = (req->newlen - req->newidx);
535 		error = SYSCTL_IN(req, arg1, arg2);
536 		((char *)arg1)[arg2] = '\0';
537 	}
538 
539 	return (error);
540 }
541 
542 /*
543  * Handle any kind of opaque data.
544  * arg1 points to it, arg2 is the size.
545  */
546 
547 int
548 sysctl_handle_opaque SYSCTL_HANDLER_ARGS
549 {
550 	int error;
551 
552 	error = SYSCTL_OUT(req, arg1, arg2);
553 
554 	if (error || !req->newptr)
555 		return (error);
556 
557 	error = SYSCTL_IN(req, arg1, arg2);
558 
559 	return (error);
560 }
561 
562 /*
563  * Transfer functions to/from kernel space.
564  * XXX: rather untested at this point
565  */
566 static int
567 sysctl_old_kernel(struct sysctl_req *req, const void *p, int l)
568 {
569 	int i = 0;
570 
571 	if (req->oldptr) {
572 		i = min(req->oldlen - req->oldidx, l);
573 		if (i > 0)
574 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
575 	}
576 	req->oldidx += l;
577 	if (req->oldptr && i != l)
578 		return (ENOMEM);
579 	return (0);
580 }
581 
582 static int
583 sysctl_new_kernel(struct sysctl_req *req, void *p, int l)
584 {
585 	if (!req->newptr)
586 		return 0;
587 	if (req->newlen - req->newidx < l)
588 		return (EINVAL);
589 	bcopy((char *)req->newptr + req->newidx, p, l);
590 	req->newidx += l;
591 	return (0);
592 }
593 
594 int
595 kernel_sysctl(struct proc *p, int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, int *retval)
596 {
597 	int error = 0;
598 	struct sysctl_req req;
599 
600 	bzero(&req, sizeof req);
601 
602 	req.p = p;
603 
604 	if (oldlenp) {
605 		req.oldlen = *oldlenp;
606 	}
607 
608 	if (old) {
609 		req.oldptr= old;
610 	}
611 
612 	if (newlen) {
613 		req.newlen = newlen;
614 		req.newptr = new;
615 	}
616 
617 	req.oldfunc = sysctl_old_kernel;
618 	req.newfunc = sysctl_new_kernel;
619 	req.lock = 1;
620 
621 	/* XXX this should probably be done in a general way */
622 	while (memlock.sl_lock) {
623 		memlock.sl_want = 1;
624 		(void) tsleep((caddr_t)&memlock, PRIBIO+1, "sysctl", 0);
625 		memlock.sl_locked++;
626 	}
627 	memlock.sl_lock = 1;
628 
629 	error = sysctl_root(0, name, namelen, &req);
630 
631 	if (req.lock == 2)
632 		vsunlock(req.oldptr, req.oldlen, B_WRITE);
633 
634 	memlock.sl_lock = 0;
635 
636 	if (memlock.sl_want) {
637 		memlock.sl_want = 0;
638 		wakeup((caddr_t)&memlock);
639 	}
640 
641 	if (error && error != ENOMEM)
642 		return (error);
643 
644 	if (retval) {
645 		if (req.oldptr && req.oldidx > req.oldlen)
646 			*retval = req.oldlen;
647 		else
648 			*retval = req.oldidx;
649 	}
650 	return (error);
651 }
652 
653 /*
654  * Transfer function to/from user space.
655  */
656 static int
657 sysctl_old_user(struct sysctl_req *req, const void *p, int l)
658 {
659 	int error = 0, i = 0;
660 
661 	if (req->lock == 1 && req->oldptr) {
662 		vslock(req->oldptr, req->oldlen);
663 		req->lock = 2;
664 	}
665 	if (req->oldptr) {
666 		i = min(req->oldlen - req->oldidx, l);
667 		if (i > 0)
668 			error = copyout(p, (char *)req->oldptr + req->oldidx,
669 					i);
670 	}
671 	req->oldidx += l;
672 	if (error)
673 		return (error);
674 	if (req->oldptr && i < l)
675 		return (ENOMEM);
676 	return (0);
677 }
678 
679 static int
680 sysctl_new_user(struct sysctl_req *req, void *p, int l)
681 {
682 	int error;
683 
684 	if (!req->newptr)
685 		return 0;
686 	if (req->newlen - req->newidx < l)
687 		return (EINVAL);
688 	error = copyin((char *)req->newptr + req->newidx, p, l);
689 	req->newidx += l;
690 	return (error);
691 }
692 
693 /*
694  * Traverse our tree, and find the right node, execute whatever it points
695  * at, and return the resulting error code.
696  */
697 
698 int
699 sysctl_root SYSCTL_HANDLER_ARGS
700 {
701 	int *name = (int *) arg1;
702 	u_int namelen = arg2;
703 	int indx, i, j;
704 	struct sysctl_oid **oidpp;
705 	struct linker_set *lsp = &sysctl_;
706 
707 	j = lsp->ls_length;
708 	oidpp = (struct sysctl_oid **) lsp->ls_items;
709 
710 	indx = 0;
711 	while (j-- && indx < CTL_MAXNAME) {
712 		if (*oidpp && ((*oidpp)->oid_number == name[indx])) {
713 			indx++;
714 			if ((*oidpp)->oid_kind & CTLFLAG_NOLOCK)
715 				req->lock = 0;
716 			if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
717 				if ((*oidpp)->oid_handler)
718 					goto found;
719 				if (indx == namelen)
720 					return ENOENT;
721 				lsp = (struct linker_set*)(*oidpp)->oid_arg1;
722 				j = lsp->ls_length;
723 				oidpp = (struct sysctl_oid **)lsp->ls_items;
724 			} else {
725 				if (indx != namelen)
726 					return EISDIR;
727 				goto found;
728 			}
729 		} else {
730 			oidpp++;
731 		}
732 	}
733 	return ENOENT;
734 found:
735 	/* If writing isn't allowed */
736 	if (req->newptr && !((*oidpp)->oid_kind & CTLFLAG_WR))
737 		return (EPERM);
738 
739 	/* Most likely only root can write */
740 	if (!((*oidpp)->oid_kind & CTLFLAG_ANYBODY) &&
741 	    req->newptr && req->p &&
742 	    (i = suser(req->p->p_ucred, &req->p->p_acflag)))
743 		return (i);
744 
745 	if (!(*oidpp)->oid_handler)
746 		return EINVAL;
747 
748 	if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
749 		i = ((*oidpp)->oid_handler) (*oidpp,
750 					name + indx, namelen - indx,
751 					req);
752 	} else {
753 		i = ((*oidpp)->oid_handler) (*oidpp,
754 					(*oidpp)->oid_arg1, (*oidpp)->oid_arg2,
755 					req);
756 	}
757 	return (i);
758 }
759 
760 #ifndef _SYS_SYSPROTO_H_
761 struct sysctl_args {
762 	int	*name;
763 	u_int	namelen;
764 	void	*old;
765 	size_t	*oldlenp;
766 	void	*new;
767 	size_t	newlen;
768 };
769 #endif
770 
771 int
772 __sysctl(struct proc *p, struct sysctl_args *uap, int *retval)
773 {
774 	int error, i, j, name[CTL_MAXNAME];
775 
776 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
777 		return (EINVAL);
778 
779  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
780  	if (error)
781 		return (error);
782 
783 	error = userland_sysctl(p, name, uap->namelen,
784 		uap->old, uap->oldlenp, 0,
785 		uap->new, uap->newlen, &j);
786 	if (error && error != ENOMEM)
787 		return (error);
788 	if (uap->oldlenp) {
789 		i = copyout(&j, uap->oldlenp, sizeof(j));
790 		if (i)
791 			return (i);
792 	}
793 	return (error);
794 }
795 
796 /*
797  * This is used from various compatibility syscalls too.  That's why name
798  * must be in kernel space.
799  */
800 int
801 userland_sysctl(struct proc *p, int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, int *retval)
802 {
803 	int error = 0;
804 	struct sysctl_req req, req2;
805 
806 	bzero(&req, sizeof req);
807 
808 	req.p = p;
809 
810 	if (oldlenp) {
811 		if (inkernel) {
812 			req.oldlen = *oldlenp;
813 		} else {
814 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
815 			if (error)
816 				return (error);
817 		}
818 	}
819 
820 	if (old) {
821 		if (!useracc(old, req.oldlen, B_WRITE))
822 			return (EFAULT);
823 		req.oldptr= old;
824 	}
825 
826 	if (newlen) {
827 		if (!useracc(new, req.newlen, B_READ))
828 			return (EFAULT);
829 		req.newlen = newlen;
830 		req.newptr = new;
831 	}
832 
833 	req.oldfunc = sysctl_old_user;
834 	req.newfunc = sysctl_new_user;
835 	req.lock = 1;
836 
837 	/* XXX this should probably be done in a general way */
838 	while (memlock.sl_lock) {
839 		memlock.sl_want = 1;
840 		(void) tsleep((caddr_t)&memlock, PRIBIO+1, "sysctl", 0);
841 		memlock.sl_locked++;
842 	}
843 	memlock.sl_lock = 1;
844 
845 	do {
846 	    req2 = req;
847 	    error = sysctl_root(0, name, namelen, &req2);
848 	} while (error == EAGAIN);
849 
850 	req = req2;
851 	if (req.lock == 2)
852 		vsunlock(req.oldptr, req.oldlen, B_WRITE);
853 
854 	memlock.sl_lock = 0;
855 
856 	if (memlock.sl_want) {
857 		memlock.sl_want = 0;
858 		wakeup((caddr_t)&memlock);
859 	}
860 
861 	if (error && error != ENOMEM)
862 		return (error);
863 
864 	if (retval) {
865 		if (req.oldptr && req.oldidx > req.oldlen)
866 			*retval = req.oldlen;
867 		else
868 			*retval = req.oldidx;
869 	}
870 	return (error);
871 }
872 
873 #ifdef COMPAT_43
874 #include <sys/socket.h>
875 #include <vm/vm_param.h>
876 
877 #define	KINFO_PROC		(0<<8)
878 #define	KINFO_RT		(1<<8)
879 #define	KINFO_VNODE		(2<<8)
880 #define	KINFO_FILE		(3<<8)
881 #define	KINFO_METER		(4<<8)
882 #define	KINFO_LOADAVG		(5<<8)
883 #define	KINFO_CLOCKRATE		(6<<8)
884 
885 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
886 #define	KINFO_BSDI_SYSINFO	(101<<8)
887 
888 /*
889  * XXX this is bloat, but I hope it's better here than on the potentially
890  * limited kernel stack...  -Peter
891  */
892 
893 static struct {
894 	int	bsdi_machine;		/* "i386" on BSD/386 */
895 /*      ^^^ this is an offset to the string, relative to the struct start */
896 	char	*pad0;
897 	long	pad1;
898 	long	pad2;
899 	long	pad3;
900 	u_long	pad4;
901 	u_long	pad5;
902 	u_long	pad6;
903 
904 	int	bsdi_ostype;		/* "BSD/386" on BSD/386 */
905 	int	bsdi_osrelease;		/* "1.1" on BSD/386 */
906 	long	pad7;
907 	long	pad8;
908 	char	*pad9;
909 
910 	long	pad10;
911 	long	pad11;
912 	int	pad12;
913 	long	pad13;
914 	quad_t	pad14;
915 	long	pad15;
916 
917 	struct	timeval pad16;
918 	/* we dont set this, because BSDI's uname used gethostname() instead */
919 	int	bsdi_hostname;		/* hostname on BSD/386 */
920 
921 	/* the actual string data is appended here */
922 
923 } bsdi_si;
924 /*
925  * this data is appended to the end of the bsdi_si structure during copyout.
926  * The "char *" offsets are relative to the base of the bsdi_si struct.
927  * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
928  * should not exceed the length of the buffer here... (or else!! :-)
929  */
930 static char bsdi_strings[80];	/* It had better be less than this! */
931 
932 #ifndef _SYS_SYSPROTO_H_
933 struct getkerninfo_args {
934 	int	op;
935 	char	*where;
936 	int	*size;
937 	int	arg;
938 };
939 #endif
940 
941 int
942 ogetkerninfo(struct proc *p, struct getkerninfo_args *uap, int *retval)
943 {
944 	int error, name[6];
945 	u_int size;
946 
947 	switch (uap->op & 0xff00) {
948 
949 	case KINFO_RT:
950 		name[0] = CTL_NET;
951 		name[1] = PF_ROUTE;
952 		name[2] = 0;
953 		name[3] = (uap->op & 0xff0000) >> 16;
954 		name[4] = uap->op & 0xff;
955 		name[5] = uap->arg;
956 		error = userland_sysctl(p, name, 6, uap->where, uap->size,
957 			0, 0, 0, &size);
958 		break;
959 
960 	case KINFO_VNODE:
961 		name[0] = CTL_KERN;
962 		name[1] = KERN_VNODE;
963 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
964 			0, 0, 0, &size);
965 		break;
966 
967 	case KINFO_PROC:
968 		name[0] = CTL_KERN;
969 		name[1] = KERN_PROC;
970 		name[2] = uap->op & 0xff;
971 		name[3] = uap->arg;
972 		error = userland_sysctl(p, name, 4, uap->where, uap->size,
973 			0, 0, 0, &size);
974 		break;
975 
976 	case KINFO_FILE:
977 		name[0] = CTL_KERN;
978 		name[1] = KERN_FILE;
979 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
980 			0, 0, 0, &size);
981 		break;
982 
983 	case KINFO_METER:
984 		name[0] = CTL_VM;
985 		name[1] = VM_METER;
986 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
987 			0, 0, 0, &size);
988 		break;
989 
990 	case KINFO_LOADAVG:
991 		name[0] = CTL_VM;
992 		name[1] = VM_LOADAVG;
993 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
994 			0, 0, 0, &size);
995 		break;
996 
997 	case KINFO_CLOCKRATE:
998 		name[0] = CTL_KERN;
999 		name[1] = KERN_CLOCKRATE;
1000 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1001 			0, 0, 0, &size);
1002 		break;
1003 
1004 	case KINFO_BSDI_SYSINFO: {
1005 		/*
1006 		 * this is pretty crude, but it's just enough for uname()
1007 		 * from BSDI's 1.x libc to work.
1008 		 *
1009 		 * In particular, it doesn't return the same results when
1010 		 * the supplied buffer is too small.  BSDI's version apparently
1011 		 * will return the amount copied, and set the *size to how
1012 		 * much was needed.  The emulation framework here isn't capable
1013 		 * of that, so we just set both to the amount copied.
1014 		 * BSDI's 2.x product apparently fails with ENOMEM in this
1015 		 * scenario.
1016 		 */
1017 
1018 		u_int needed;
1019 		u_int left;
1020 		char *s;
1021 
1022 		bzero((char *)&bsdi_si, sizeof(bsdi_si));
1023 		bzero(bsdi_strings, sizeof(bsdi_strings));
1024 
1025 		s = bsdi_strings;
1026 
1027 		bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1028 		strcpy(s, ostype);
1029 		s += strlen(s) + 1;
1030 
1031 		bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1032 		strcpy(s, osrelease);
1033 		s += strlen(s) + 1;
1034 
1035 		bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1036 		strcpy(s, machine);
1037 		s += strlen(s) + 1;
1038 
1039 		needed = sizeof(bsdi_si) + (s - bsdi_strings);
1040 
1041 		if (uap->where == NULL) {
1042 			/* process is asking how much buffer to supply.. */
1043 			size = needed;
1044 			error = 0;
1045 			break;
1046 		}
1047 
1048 
1049 		/* if too much buffer supplied, trim it down */
1050 		if (size > needed)
1051 			size = needed;
1052 
1053 		/* how much of the buffer is remaining */
1054 		left = size;
1055 
1056 		if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1057 			break;
1058 
1059 		/* is there any point in continuing? */
1060 		if (left > sizeof(bsdi_si)) {
1061 			left -= sizeof(bsdi_si);
1062 			error = copyout(&bsdi_strings,
1063 					uap->where + sizeof(bsdi_si), left);
1064 		}
1065 		break;
1066 	}
1067 
1068 	default:
1069 		return (EOPNOTSUPP);
1070 	}
1071 	if (error)
1072 		return (error);
1073 	*retval = size;
1074 	if (uap->size)
1075 		error = copyout((caddr_t)&size, (caddr_t)uap->size,
1076 		    sizeof(size));
1077 	return (error);
1078 }
1079 #endif /* COMPAT_43 */
1080