xref: /freebsd/sys/kern/kern_sysctl.c (revision 1e413cf93298b5b97441a21d9a50fdcd0ee9945e)
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  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_compat.h"
42 #include "opt_mac.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/sysctl.h>
48 #include <sys/malloc.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/lock.h>
52 #include <sys/mutex.h>
53 #include <sys/sx.h>
54 #include <sys/sysproto.h>
55 
56 #include <security/mac/mac_framework.h>
57 
58 #include <vm/vm.h>
59 #include <vm/vm_extern.h>
60 
61 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
62 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
63 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
64 
65 /*
66  * Locking - this locks the sysctl tree in memory.
67  */
68 static struct sx sysctllock;
69 
70 #define	SYSCTL_LOCK()		sx_xlock(&sysctllock)
71 #define	SYSCTL_UNLOCK()		sx_xunlock(&sysctllock)
72 #define	SYSCTL_INIT()		sx_init(&sysctllock, "sysctl lock")
73 
74 static int sysctl_root(SYSCTL_HANDLER_ARGS);
75 
76 struct sysctl_oid_list sysctl__children; /* root list */
77 
78 static struct sysctl_oid *
79 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
80 {
81 	struct sysctl_oid *oidp;
82 
83 	SLIST_FOREACH(oidp, list, oid_link) {
84 		if (strcmp(oidp->oid_name, name) == 0) {
85 			return (oidp);
86 		}
87 	}
88 	return (NULL);
89 }
90 
91 /*
92  * Initialization of the MIB tree.
93  *
94  * Order by number in each list.
95  */
96 
97 void
98 sysctl_register_oid(struct sysctl_oid *oidp)
99 {
100 	struct sysctl_oid_list *parent = oidp->oid_parent;
101 	struct sysctl_oid *p;
102 	struct sysctl_oid *q;
103 
104 	/*
105 	 * First check if another oid with the same name already
106 	 * exists in the parent's list.
107 	 */
108 	p = sysctl_find_oidname(oidp->oid_name, parent);
109 	if (p != NULL) {
110 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
111 			p->oid_refcnt++;
112 			return;
113 		} else {
114 			printf("can't re-use a leaf (%s)!\n", p->oid_name);
115 			return;
116 		}
117 	}
118 	/*
119 	 * If this oid has a number OID_AUTO, give it a number which
120 	 * is greater than any current oid.
121 	 * NOTE: DO NOT change the starting value here, change it in
122 	 * <sys/sysctl.h>, and make sure it is at least 256 to
123 	 * accomodate e.g. net.inet.raw as a static sysctl node.
124 	 */
125 	if (oidp->oid_number == OID_AUTO) {
126 		static int newoid = CTL_AUTO_START;
127 
128 		oidp->oid_number = newoid++;
129 		if (newoid == 0x7fffffff)
130 			panic("out of oids");
131 	}
132 #if 0
133 	else if (oidp->oid_number >= CTL_AUTO_START) {
134 		/* do not panic; this happens when unregistering sysctl sets */
135 		printf("static sysctl oid too high: %d", oidp->oid_number);
136 	}
137 #endif
138 
139 	/*
140 	 * Insert the oid into the parent's list in order.
141 	 */
142 	q = NULL;
143 	SLIST_FOREACH(p, parent, oid_link) {
144 		if (oidp->oid_number < p->oid_number)
145 			break;
146 		q = p;
147 	}
148 	if (q)
149 		SLIST_INSERT_AFTER(q, oidp, oid_link);
150 	else
151 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
152 }
153 
154 void
155 sysctl_unregister_oid(struct sysctl_oid *oidp)
156 {
157 	struct sysctl_oid *p;
158 	int error;
159 
160 	error = ENOENT;
161 	if (oidp->oid_number == OID_AUTO) {
162 		error = EINVAL;
163 	} else {
164 		SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
165 			if (p == oidp) {
166 				SLIST_REMOVE(oidp->oid_parent, oidp,
167 				    sysctl_oid, oid_link);
168 				error = 0;
169 				break;
170 			}
171 		}
172 	}
173 
174 	/*
175 	 * This can happen when a module fails to register and is
176 	 * being unloaded afterwards.  It should not be a panic()
177 	 * for normal use.
178 	 */
179 	if (error)
180 		printf("%s: failed to unregister sysctl\n", __func__);
181 }
182 
183 /* Initialize a new context to keep track of dynamically added sysctls. */
184 int
185 sysctl_ctx_init(struct sysctl_ctx_list *c)
186 {
187 
188 	if (c == NULL) {
189 		return (EINVAL);
190 	}
191 	TAILQ_INIT(c);
192 	return (0);
193 }
194 
195 /* Free the context, and destroy all dynamic oids registered in this context */
196 int
197 sysctl_ctx_free(struct sysctl_ctx_list *clist)
198 {
199 	struct sysctl_ctx_entry *e, *e1;
200 	int error;
201 
202 	error = 0;
203 	/*
204 	 * First perform a "dry run" to check if it's ok to remove oids.
205 	 * XXX FIXME
206 	 * XXX This algorithm is a hack. But I don't know any
207 	 * XXX better solution for now...
208 	 */
209 	TAILQ_FOREACH(e, clist, link) {
210 		error = sysctl_remove_oid(e->entry, 0, 0);
211 		if (error)
212 			break;
213 	}
214 	/*
215 	 * Restore deregistered entries, either from the end,
216 	 * or from the place where error occured.
217 	 * e contains the entry that was not unregistered
218 	 */
219 	if (error)
220 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
221 	else
222 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
223 	while (e1 != NULL) {
224 		sysctl_register_oid(e1->entry);
225 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
226 	}
227 	if (error)
228 		return(EBUSY);
229 	/* Now really delete the entries */
230 	e = TAILQ_FIRST(clist);
231 	while (e != NULL) {
232 		e1 = TAILQ_NEXT(e, link);
233 		error = sysctl_remove_oid(e->entry, 1, 0);
234 		if (error)
235 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
236 			    e->entry->oid_name);
237 		free(e, M_SYSCTLOID);
238 		e = e1;
239 	}
240 	return (error);
241 }
242 
243 /* Add an entry to the context */
244 struct sysctl_ctx_entry *
245 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
246 {
247 	struct sysctl_ctx_entry *e;
248 
249 	if (clist == NULL || oidp == NULL)
250 		return(NULL);
251 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
252 	e->entry = oidp;
253 	TAILQ_INSERT_HEAD(clist, e, link);
254 	return (e);
255 }
256 
257 /* Find an entry in the context */
258 struct sysctl_ctx_entry *
259 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
260 {
261 	struct sysctl_ctx_entry *e;
262 
263 	if (clist == NULL || oidp == NULL)
264 		return(NULL);
265 	TAILQ_FOREACH(e, clist, link) {
266 		if(e->entry == oidp)
267 			return(e);
268 	}
269 	return (e);
270 }
271 
272 /*
273  * Delete an entry from the context.
274  * NOTE: this function doesn't free oidp! You have to remove it
275  * with sysctl_remove_oid().
276  */
277 int
278 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
279 {
280 	struct sysctl_ctx_entry *e;
281 
282 	if (clist == NULL || oidp == NULL)
283 		return (EINVAL);
284 	e = sysctl_ctx_entry_find(clist, oidp);
285 	if (e != NULL) {
286 		TAILQ_REMOVE(clist, e, link);
287 		free(e, M_SYSCTLOID);
288 		return (0);
289 	} else
290 		return (ENOENT);
291 }
292 
293 /*
294  * Remove dynamically created sysctl trees.
295  * oidp - top of the tree to be removed
296  * del - if 0 - just deregister, otherwise free up entries as well
297  * recurse - if != 0 traverse the subtree to be deleted
298  */
299 int
300 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
301 {
302 	struct sysctl_oid *p;
303 	int error;
304 
305 	if (oidp == NULL)
306 		return(EINVAL);
307 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
308 		printf("can't remove non-dynamic nodes!\n");
309 		return (EINVAL);
310 	}
311 	/*
312 	 * WARNING: normal method to do this should be through
313 	 * sysctl_ctx_free(). Use recursing as the last resort
314 	 * method to purge your sysctl tree of leftovers...
315 	 * However, if some other code still references these nodes,
316 	 * it will panic.
317 	 */
318 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
319 		if (oidp->oid_refcnt == 1) {
320 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
321 				if (!recurse)
322 					return (ENOTEMPTY);
323 				error = sysctl_remove_oid(p, del, recurse);
324 				if (error)
325 					return (error);
326 			}
327 			if (del)
328 				free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
329 		}
330 	}
331 	if (oidp->oid_refcnt > 1 ) {
332 		oidp->oid_refcnt--;
333 	} else {
334 		if (oidp->oid_refcnt == 0) {
335 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
336 				oidp->oid_refcnt, oidp->oid_name);
337 			return (EINVAL);
338 		}
339 		sysctl_unregister_oid(oidp);
340 		if (del) {
341 			if (oidp->oid_descr)
342 				free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
343 			free((void *)(uintptr_t)(const void *)oidp->oid_name,
344 			     M_SYSCTLOID);
345 			free(oidp, M_SYSCTLOID);
346 		}
347 	}
348 	return (0);
349 }
350 
351 /*
352  * Create new sysctls at run time.
353  * clist may point to a valid context initialized with sysctl_ctx_init().
354  */
355 struct sysctl_oid *
356 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
357 	int number, const char *name, int kind, void *arg1, int arg2,
358 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
359 {
360 	struct sysctl_oid *oidp;
361 	ssize_t len;
362 	char *newname;
363 
364 	/* You have to hook up somewhere.. */
365 	if (parent == NULL)
366 		return(NULL);
367 	/* Check if the node already exists, otherwise create it */
368 	oidp = sysctl_find_oidname(name, parent);
369 	if (oidp != NULL) {
370 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
371 			oidp->oid_refcnt++;
372 			/* Update the context */
373 			if (clist != NULL)
374 				sysctl_ctx_entry_add(clist, oidp);
375 			return (oidp);
376 		} else {
377 			printf("can't re-use a leaf (%s)!\n", name);
378 			return (NULL);
379 		}
380 	}
381 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
382 	oidp->oid_parent = parent;
383 	SLIST_NEXT(oidp, oid_link) = NULL;
384 	oidp->oid_number = number;
385 	oidp->oid_refcnt = 1;
386 	len = strlen(name);
387 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
388 	bcopy(name, newname, len + 1);
389 	newname[len] = '\0';
390 	oidp->oid_name = newname;
391 	oidp->oid_handler = handler;
392 	oidp->oid_kind = CTLFLAG_DYN | kind;
393 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
394 		/* Allocate space for children */
395 		SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
396 		    M_SYSCTLOID, M_WAITOK));
397 		SLIST_INIT(SYSCTL_CHILDREN(oidp));
398 	} else {
399 		oidp->oid_arg1 = arg1;
400 		oidp->oid_arg2 = arg2;
401 	}
402 	oidp->oid_fmt = fmt;
403 	if (descr) {
404 		int len = strlen(descr) + 1;
405 		oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
406 		if (oidp->oid_descr)
407 			strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
408 	}
409 	/* Update the context, if used */
410 	if (clist != NULL)
411 		sysctl_ctx_entry_add(clist, oidp);
412 	/* Register this oid */
413 	sysctl_register_oid(oidp);
414 	return (oidp);
415 }
416 
417 /*
418  * Rename an existing oid.
419  */
420 void
421 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
422 {
423 	ssize_t len;
424 	char *newname;
425 	void *oldname;
426 
427 	oldname = (void *)(uintptr_t)(const void *)oidp->oid_name;
428 	len = strlen(name);
429 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
430 	bcopy(name, newname, len + 1);
431 	newname[len] = '\0';
432 	oidp->oid_name = newname;
433 	free(oldname, M_SYSCTLOID);
434 }
435 
436 /*
437  * Reparent an existing oid.
438  */
439 int
440 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
441 {
442 	struct sysctl_oid *oidp;
443 
444 	if (oid->oid_parent == parent)
445 		return (0);
446 	oidp = sysctl_find_oidname(oid->oid_name, parent);
447 	if (oidp != NULL)
448 		return (EEXIST);
449 	sysctl_unregister_oid(oid);
450 	oid->oid_parent = parent;
451 	oid->oid_number = OID_AUTO;
452 	sysctl_register_oid(oid);
453 	return (0);
454 }
455 
456 /*
457  * Register the kernel's oids on startup.
458  */
459 SET_DECLARE(sysctl_set, struct sysctl_oid);
460 
461 static void
462 sysctl_register_all(void *arg)
463 {
464 	struct sysctl_oid **oidp;
465 
466 	SYSCTL_INIT();
467 	SET_FOREACH(oidp, sysctl_set)
468 		sysctl_register_oid(*oidp);
469 }
470 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
471 
472 /*
473  * "Staff-functions"
474  *
475  * These functions implement a presently undocumented interface
476  * used by the sysctl program to walk the tree, and get the type
477  * so it can print the value.
478  * This interface is under work and consideration, and should probably
479  * be killed with a big axe by the first person who can find the time.
480  * (be aware though, that the proper interface isn't as obvious as it
481  * may seem, there are various conflicting requirements.
482  *
483  * {0,0}	printf the entire MIB-tree.
484  * {0,1,...}	return the name of the "..." OID.
485  * {0,2,...}	return the next OID.
486  * {0,3}	return the OID of the name in "new"
487  * {0,4,...}	return the kind & format info for the "..." OID.
488  * {0,5,...}	return the description the "..." OID.
489  */
490 
491 #ifdef SYSCTL_DEBUG
492 static void
493 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
494 {
495 	int k;
496 	struct sysctl_oid *oidp;
497 
498 	SLIST_FOREACH(oidp, l, oid_link) {
499 
500 		for (k=0; k<i; k++)
501 			printf(" ");
502 
503 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
504 
505 		printf("%c%c",
506 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
507 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
508 
509 		if (oidp->oid_handler)
510 			printf(" *Handler");
511 
512 		switch (oidp->oid_kind & CTLTYPE) {
513 			case CTLTYPE_NODE:
514 				printf(" Node\n");
515 				if (!oidp->oid_handler) {
516 					sysctl_sysctl_debug_dump_node(
517 						oidp->oid_arg1, i+2);
518 				}
519 				break;
520 			case CTLTYPE_INT:    printf(" Int\n"); break;
521 			case CTLTYPE_STRING: printf(" String\n"); break;
522 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
523 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
524 			default:	     printf("\n");
525 		}
526 
527 	}
528 }
529 
530 static int
531 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
532 {
533 	int error;
534 
535 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
536 	if (error)
537 		return (error);
538 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
539 	return (ENOENT);
540 }
541 
542 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
543 	0, 0, sysctl_sysctl_debug, "-", "");
544 #endif
545 
546 static int
547 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
548 {
549 	int *name = (int *) arg1;
550 	u_int namelen = arg2;
551 	int error = 0;
552 	struct sysctl_oid *oid;
553 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
554 	char buf[10];
555 
556 	while (namelen) {
557 		if (!lsp) {
558 			snprintf(buf,sizeof(buf),"%d",*name);
559 			if (req->oldidx)
560 				error = SYSCTL_OUT(req, ".", 1);
561 			if (!error)
562 				error = SYSCTL_OUT(req, buf, strlen(buf));
563 			if (error)
564 				return (error);
565 			namelen--;
566 			name++;
567 			continue;
568 		}
569 		lsp2 = 0;
570 		SLIST_FOREACH(oid, lsp, oid_link) {
571 			if (oid->oid_number != *name)
572 				continue;
573 
574 			if (req->oldidx)
575 				error = SYSCTL_OUT(req, ".", 1);
576 			if (!error)
577 				error = SYSCTL_OUT(req, oid->oid_name,
578 					strlen(oid->oid_name));
579 			if (error)
580 				return (error);
581 
582 			namelen--;
583 			name++;
584 
585 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
586 				break;
587 
588 			if (oid->oid_handler)
589 				break;
590 
591 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
592 			break;
593 		}
594 		lsp = lsp2;
595 	}
596 	return (SYSCTL_OUT(req, "", 1));
597 }
598 
599 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
600 
601 static int
602 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
603 	int *next, int *len, int level, struct sysctl_oid **oidpp)
604 {
605 	struct sysctl_oid *oidp;
606 
607 	*len = level;
608 	SLIST_FOREACH(oidp, lsp, oid_link) {
609 		*next = oidp->oid_number;
610 		*oidpp = oidp;
611 
612 		if (oidp->oid_kind & CTLFLAG_SKIP)
613 			continue;
614 
615 		if (!namelen) {
616 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
617 				return (0);
618 			if (oidp->oid_handler)
619 				/* We really should call the handler here...*/
620 				return (0);
621 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
622 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
623 				len, level+1, oidpp))
624 				return (0);
625 			goto emptynode;
626 		}
627 
628 		if (oidp->oid_number < *name)
629 			continue;
630 
631 		if (oidp->oid_number > *name) {
632 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
633 				return (0);
634 			if (oidp->oid_handler)
635 				return (0);
636 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
637 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
638 				next+1, len, level+1, oidpp))
639 				return (0);
640 			goto next;
641 		}
642 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
643 			continue;
644 
645 		if (oidp->oid_handler)
646 			continue;
647 
648 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
649 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
650 			len, level+1, oidpp))
651 			return (0);
652 	next:
653 		namelen = 1;
654 	emptynode:
655 		*len = level;
656 	}
657 	return (1);
658 }
659 
660 static int
661 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
662 {
663 	int *name = (int *) arg1;
664 	u_int namelen = arg2;
665 	int i, j, error;
666 	struct sysctl_oid *oid;
667 	struct sysctl_oid_list *lsp = &sysctl__children;
668 	int newoid[CTL_MAXNAME];
669 
670 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
671 	if (i)
672 		return (ENOENT);
673 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
674 	return (error);
675 }
676 
677 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
678 
679 static int
680 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
681 {
682 	int i;
683 	struct sysctl_oid *oidp;
684 	struct sysctl_oid_list *lsp = &sysctl__children;
685 	char *p;
686 
687 	if (!*name)
688 		return (ENOENT);
689 
690 	p = name + strlen(name) - 1 ;
691 	if (*p == '.')
692 		*p = '\0';
693 
694 	*len = 0;
695 
696 	for (p = name; *p && *p != '.'; p++)
697 		;
698 	i = *p;
699 	if (i == '.')
700 		*p = '\0';
701 
702 	oidp = SLIST_FIRST(lsp);
703 
704 	while (oidp && *len < CTL_MAXNAME) {
705 		if (strcmp(name, oidp->oid_name)) {
706 			oidp = SLIST_NEXT(oidp, oid_link);
707 			continue;
708 		}
709 		*oid++ = oidp->oid_number;
710 		(*len)++;
711 
712 		if (!i) {
713 			if (oidpp)
714 				*oidpp = oidp;
715 			return (0);
716 		}
717 
718 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
719 			break;
720 
721 		if (oidp->oid_handler)
722 			break;
723 
724 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
725 		oidp = SLIST_FIRST(lsp);
726 		name = p+1;
727 		for (p = name; *p && *p != '.'; p++)
728 				;
729 		i = *p;
730 		if (i == '.')
731 			*p = '\0';
732 	}
733 	return (ENOENT);
734 }
735 
736 static int
737 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
738 {
739 	char *p;
740 	int error, oid[CTL_MAXNAME], len;
741 	struct sysctl_oid *op = 0;
742 
743 	if (!req->newlen)
744 		return (ENOENT);
745 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
746 		return (ENAMETOOLONG);
747 
748 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
749 
750 	error = SYSCTL_IN(req, p, req->newlen);
751 	if (error) {
752 		free(p, M_SYSCTL);
753 		return (error);
754 	}
755 
756 	p [req->newlen] = '\0';
757 
758 	error = name2oid(p, oid, &len, &op);
759 
760 	free(p, M_SYSCTL);
761 
762 	if (error)
763 		return (error);
764 
765 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
766 	return (error);
767 }
768 
769 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
770 	sysctl_sysctl_name2oid, "I", "");
771 
772 static int
773 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
774 {
775 	struct sysctl_oid *oid;
776 	int error;
777 
778 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
779 	if (error)
780 		return (error);
781 
782 	if (!oid->oid_fmt)
783 		return (ENOENT);
784 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
785 	if (error)
786 		return (error);
787 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
788 	return (error);
789 }
790 
791 
792 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
793 
794 static int
795 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
796 {
797 	struct sysctl_oid *oid;
798 	int error;
799 
800 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
801 	if (error)
802 		return (error);
803 
804 	if (!oid->oid_descr)
805 		return (ENOENT);
806 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
807 	return (error);
808 }
809 
810 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
811 
812 /*
813  * Default "handler" functions.
814  */
815 
816 /*
817  * Handle an int, signed or unsigned.
818  * Two cases:
819  *     a variable:  point arg1 at it.
820  *     a constant:  pass it in arg2.
821  */
822 
823 int
824 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
825 {
826 	int tmpout, error = 0;
827 
828 	/*
829 	 * Attempt to get a coherent snapshot by making a copy of the data.
830 	 */
831 	if (arg1)
832 		tmpout = *(int *)arg1;
833 	else
834 		tmpout = arg2;
835 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
836 
837 	if (error || !req->newptr)
838 		return (error);
839 
840 	if (!arg1)
841 		error = EPERM;
842 	else
843 		error = SYSCTL_IN(req, arg1, sizeof(int));
844 	return (error);
845 }
846 
847 
848 /*
849  * Based on on sysctl_handle_int() convert milliseconds into ticks.
850  */
851 
852 int
853 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
854 {
855 	int error, s, tt;
856 
857 	tt = *(int *)oidp->oid_arg1;
858 	s = (int)((int64_t)tt * 1000 / hz);
859 
860 	error = sysctl_handle_int(oidp, &s, 0, req);
861 	if (error || !req->newptr)
862 		return (error);
863 
864 	tt = (int)((int64_t)s * hz / 1000);
865 	if (tt < 1)
866 		return (EINVAL);
867 
868 	*(int *)oidp->oid_arg1 = tt;
869 	return (0);
870 }
871 
872 
873 /*
874  * Handle a long, signed or unsigned.  arg1 points to it.
875  */
876 
877 int
878 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
879 {
880 	int error = 0;
881 	long tmplong;
882 #ifdef SCTL_MASK32
883 	int tmpint;
884 #endif
885 
886 	/*
887 	 * Attempt to get a coherent snapshot by making a copy of the data.
888 	 */
889 	if (!arg1)
890 		return (EINVAL);
891 	tmplong = *(long *)arg1;
892 #ifdef SCTL_MASK32
893 	if (req->flags & SCTL_MASK32) {
894 		tmpint = tmplong;
895 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
896 	} else
897 #endif
898 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
899 
900 	if (error || !req->newptr)
901 		return (error);
902 
903 #ifdef SCTL_MASK32
904 	if (req->flags & SCTL_MASK32) {
905 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
906 		*(long *)arg1 = (long)tmpint;
907 	} else
908 #endif
909 		error = SYSCTL_IN(req, arg1, sizeof(long));
910 	return (error);
911 }
912 
913 /*
914  * Handle a 64 bit int, signed or unsigned.  arg1 points to it.
915  */
916 
917 int
918 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
919 {
920 	int error = 0;
921 	uint64_t tmpout;
922 
923 	/*
924 	 * Attempt to get a coherent snapshot by making a copy of the data.
925 	 */
926 	if (!arg1)
927 		return (EINVAL);
928 	tmpout = *(uint64_t *)arg1;
929 	error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
930 
931 	if (error || !req->newptr)
932 		return (error);
933 
934 	error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
935 	return (error);
936 }
937 
938 /*
939  * Handle our generic '\0' terminated 'C' string.
940  * Two cases:
941  * 	a variable string:  point arg1 at it, arg2 is max length.
942  * 	a constant string:  point arg1 at it, arg2 is zero.
943  */
944 
945 int
946 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
947 {
948 	int error=0;
949 	char *tmparg;
950 	size_t outlen;
951 
952 	/*
953 	 * Attempt to get a coherent snapshot by copying to a
954 	 * temporary kernel buffer.
955 	 */
956 retry:
957 	outlen = strlen((char *)arg1)+1;
958 	tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
959 
960 	if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
961 		free(tmparg, M_SYSCTLTMP);
962 		goto retry;
963 	}
964 
965 	error = SYSCTL_OUT(req, tmparg, outlen);
966 	free(tmparg, M_SYSCTLTMP);
967 
968 	if (error || !req->newptr)
969 		return (error);
970 
971 	if ((req->newlen - req->newidx) >= arg2) {
972 		error = EINVAL;
973 	} else {
974 		arg2 = (req->newlen - req->newidx);
975 		error = SYSCTL_IN(req, arg1, arg2);
976 		((char *)arg1)[arg2] = '\0';
977 	}
978 
979 	return (error);
980 }
981 
982 /*
983  * Handle any kind of opaque data.
984  * arg1 points to it, arg2 is the size.
985  */
986 
987 int
988 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
989 {
990 	int error, tries;
991 	u_int generation;
992 	struct sysctl_req req2;
993 
994 	/*
995 	 * Attempt to get a coherent snapshot, by using the thread
996 	 * pre-emption counter updated from within mi_switch() to
997 	 * determine if we were pre-empted during a bcopy() or
998 	 * copyout(). Make 3 attempts at doing this before giving up.
999 	 * If we encounter an error, stop immediately.
1000 	 */
1001 	tries = 0;
1002 	req2 = *req;
1003 retry:
1004 	generation = curthread->td_generation;
1005 	error = SYSCTL_OUT(req, arg1, arg2);
1006 	if (error)
1007 		return (error);
1008 	tries++;
1009 	if (generation != curthread->td_generation && tries < 3) {
1010 		*req = req2;
1011 		goto retry;
1012 	}
1013 
1014 	error = SYSCTL_IN(req, arg1, arg2);
1015 
1016 	return (error);
1017 }
1018 
1019 /*
1020  * Transfer functions to/from kernel space.
1021  * XXX: rather untested at this point
1022  */
1023 static int
1024 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1025 {
1026 	size_t i = 0;
1027 
1028 	if (req->oldptr) {
1029 		i = l;
1030 		if (req->oldlen <= req->oldidx)
1031 			i = 0;
1032 		else
1033 			if (i > req->oldlen - req->oldidx)
1034 				i = req->oldlen - req->oldidx;
1035 		if (i > 0)
1036 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
1037 	}
1038 	req->oldidx += l;
1039 	if (req->oldptr && i != l)
1040 		return (ENOMEM);
1041 	return (0);
1042 }
1043 
1044 static int
1045 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1046 {
1047 	if (!req->newptr)
1048 		return (0);
1049 	if (req->newlen - req->newidx < l)
1050 		return (EINVAL);
1051 	bcopy((char *)req->newptr + req->newidx, p, l);
1052 	req->newidx += l;
1053 	return (0);
1054 }
1055 
1056 int
1057 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1058     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1059 {
1060 	int error = 0;
1061 	struct sysctl_req req;
1062 
1063 	bzero(&req, sizeof req);
1064 
1065 	req.td = td;
1066 	req.flags = flags;
1067 
1068 	if (oldlenp) {
1069 		req.oldlen = *oldlenp;
1070 	}
1071 	req.validlen = req.oldlen;
1072 
1073 	if (old) {
1074 		req.oldptr= old;
1075 	}
1076 
1077 	if (new != NULL) {
1078 		req.newlen = newlen;
1079 		req.newptr = new;
1080 	}
1081 
1082 	req.oldfunc = sysctl_old_kernel;
1083 	req.newfunc = sysctl_new_kernel;
1084 	req.lock = REQ_LOCKED;
1085 
1086 	SYSCTL_LOCK();
1087 
1088 	error = sysctl_root(0, name, namelen, &req);
1089 
1090 	if (req.lock == REQ_WIRED && req.validlen > 0)
1091 		vsunlock(req.oldptr, req.validlen);
1092 
1093 	SYSCTL_UNLOCK();
1094 
1095 	if (error && error != ENOMEM)
1096 		return (error);
1097 
1098 	if (retval) {
1099 		if (req.oldptr && req.oldidx > req.validlen)
1100 			*retval = req.validlen;
1101 		else
1102 			*retval = req.oldidx;
1103 	}
1104 	return (error);
1105 }
1106 
1107 int
1108 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1109     void *new, size_t newlen, size_t *retval, int flags)
1110 {
1111         int oid[CTL_MAXNAME];
1112         size_t oidlen, plen;
1113 	int error;
1114 
1115 	oid[0] = 0;		/* sysctl internal magic */
1116 	oid[1] = 3;		/* name2oid */
1117 	oidlen = sizeof(oid);
1118 
1119 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1120 	    (void *)name, strlen(name), &plen, flags);
1121 	if (error)
1122 		return (error);
1123 
1124 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1125 	    new, newlen, retval, flags);
1126 	return (error);
1127 }
1128 
1129 /*
1130  * Transfer function to/from user space.
1131  */
1132 static int
1133 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1134 {
1135 	int error = 0;
1136 	size_t i, len, origidx;
1137 
1138 	origidx = req->oldidx;
1139 	req->oldidx += l;
1140 	if (req->oldptr == NULL)
1141 		return (0);
1142 	/*
1143 	 * If we have not wired the user supplied buffer and we are currently
1144 	 * holding locks, drop a witness warning, as it's possible that
1145 	 * write operations to the user page can sleep.
1146 	 */
1147 	if (req->lock != REQ_WIRED)
1148 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1149 		    "sysctl_old_user()");
1150 	i = l;
1151 	len = req->validlen;
1152 	if (len <= origidx)
1153 		i = 0;
1154 	else {
1155 		if (i > len - origidx)
1156 			i = len - origidx;
1157 		error = copyout(p, (char *)req->oldptr + origidx, i);
1158 	}
1159 	if (error)
1160 		return (error);
1161 	if (i < l)
1162 		return (ENOMEM);
1163 	return (0);
1164 }
1165 
1166 static int
1167 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1168 {
1169 	int error;
1170 
1171 	if (!req->newptr)
1172 		return (0);
1173 	if (req->newlen - req->newidx < l)
1174 		return (EINVAL);
1175 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1176 	    "sysctl_new_user()");
1177 	error = copyin((char *)req->newptr + req->newidx, p, l);
1178 	req->newidx += l;
1179 	return (error);
1180 }
1181 
1182 /*
1183  * Wire the user space destination buffer.  If set to a value greater than
1184  * zero, the len parameter limits the maximum amount of wired memory.
1185  */
1186 int
1187 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1188 {
1189 	int ret;
1190 	size_t i, wiredlen;
1191 	char *cp, dummy;
1192 
1193 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1194 	ret = 0;
1195 	if (req->lock == REQ_LOCKED && req->oldptr &&
1196 	    req->oldfunc == sysctl_old_user) {
1197 		if (wiredlen != 0) {
1198 			ret = vslock(req->oldptr, wiredlen);
1199 			if (ret != 0) {
1200 				if (ret != ENOMEM)
1201 					return (ret);
1202 				wiredlen = 0;
1203 			}
1204 			/*
1205 			 * Touch all the wired pages to avoid PTE modified
1206 			 * bit emulation traps on Alpha while holding locks
1207 			 * in the sysctl handler.
1208 			 */
1209 			for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1210 			    cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1211 				copyin(cp, &dummy, 1);
1212 				copyout(&dummy, cp, 1);
1213 			}
1214 		}
1215 		req->lock = REQ_WIRED;
1216 		req->validlen = wiredlen;
1217 	}
1218 	return (0);
1219 }
1220 
1221 int
1222 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1223     int *nindx, struct sysctl_req *req)
1224 {
1225 	struct sysctl_oid *oid;
1226 	int indx;
1227 
1228 	oid = SLIST_FIRST(&sysctl__children);
1229 	indx = 0;
1230 	while (oid && indx < CTL_MAXNAME) {
1231 		if (oid->oid_number == name[indx]) {
1232 			indx++;
1233 			if (oid->oid_kind & CTLFLAG_NOLOCK)
1234 				req->lock = REQ_UNLOCKED;
1235 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1236 				if (oid->oid_handler != NULL ||
1237 				    indx == namelen) {
1238 					*noid = oid;
1239 					if (nindx != NULL)
1240 						*nindx = indx;
1241 					return (0);
1242 				}
1243 				oid = SLIST_FIRST(
1244 				    (struct sysctl_oid_list *)oid->oid_arg1);
1245 			} else if (indx == namelen) {
1246 				*noid = oid;
1247 				if (nindx != NULL)
1248 					*nindx = indx;
1249 				return (0);
1250 			} else {
1251 				return (ENOTDIR);
1252 			}
1253 		} else {
1254 			oid = SLIST_NEXT(oid, oid_link);
1255 		}
1256 	}
1257 	return (ENOENT);
1258 }
1259 
1260 /*
1261  * Traverse our tree, and find the right node, execute whatever it points
1262  * to, and return the resulting error code.
1263  */
1264 
1265 static int
1266 sysctl_root(SYSCTL_HANDLER_ARGS)
1267 {
1268 	struct sysctl_oid *oid;
1269 	int error, indx, lvl;
1270 
1271 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1272 	if (error)
1273 		return (error);
1274 
1275 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1276 		/*
1277 		 * You can't call a sysctl when it's a node, but has
1278 		 * no handler.  Inform the user that it's a node.
1279 		 * The indx may or may not be the same as namelen.
1280 		 */
1281 		if (oid->oid_handler == NULL)
1282 			return (EISDIR);
1283 	}
1284 
1285 	/* Is this sysctl writable? */
1286 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1287 		return (EPERM);
1288 
1289 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1290 
1291 	/* Is this sysctl sensitive to securelevels? */
1292 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1293 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1294 		error = securelevel_gt(req->td->td_ucred, lvl);
1295 		if (error)
1296 			return (error);
1297 	}
1298 
1299 	/* Is this sysctl writable by only privileged users? */
1300 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1301 		if (oid->oid_kind & CTLFLAG_PRISON)
1302 			error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1303 		else
1304 			error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1305 		if (error)
1306 			return (error);
1307 	}
1308 
1309 	if (!oid->oid_handler)
1310 		return (EINVAL);
1311 
1312 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1313 		arg1 = (int *)arg1 + indx;
1314 		arg2 -= indx;
1315 	} else {
1316 		arg1 = oid->oid_arg1;
1317 		arg2 = oid->oid_arg2;
1318 	}
1319 #ifdef MAC
1320 	error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1321 	    req);
1322 	if (error != 0)
1323 		return (error);
1324 #endif
1325 	error = oid->oid_handler(oid, arg1, arg2, req);
1326 
1327 	return (error);
1328 }
1329 
1330 #ifndef _SYS_SYSPROTO_H_
1331 struct sysctl_args {
1332 	int	*name;
1333 	u_int	namelen;
1334 	void	*old;
1335 	size_t	*oldlenp;
1336 	void	*new;
1337 	size_t	newlen;
1338 };
1339 #endif
1340 int
1341 __sysctl(struct thread *td, struct sysctl_args *uap)
1342 {
1343 	int error, name[CTL_MAXNAME];
1344 	size_t j;
1345 
1346 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1347 		return (EINVAL);
1348 
1349  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1350  	if (error)
1351 		return (error);
1352 
1353 	mtx_lock(&Giant);
1354 
1355 	error = userland_sysctl(td, name, uap->namelen,
1356 		uap->old, uap->oldlenp, 0,
1357 		uap->new, uap->newlen, &j, 0);
1358 	if (error && error != ENOMEM)
1359 		goto done2;
1360 	if (uap->oldlenp) {
1361 		int i = copyout(&j, uap->oldlenp, sizeof(j));
1362 		if (i)
1363 			error = i;
1364 	}
1365 done2:
1366 	mtx_unlock(&Giant);
1367 	return (error);
1368 }
1369 
1370 /*
1371  * This is used from various compatibility syscalls too.  That's why name
1372  * must be in kernel space.
1373  */
1374 int
1375 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1376     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1377     int flags)
1378 {
1379 	int error = 0;
1380 	struct sysctl_req req;
1381 
1382 	bzero(&req, sizeof req);
1383 
1384 	req.td = td;
1385 	req.flags = flags;
1386 
1387 	if (oldlenp) {
1388 		if (inkernel) {
1389 			req.oldlen = *oldlenp;
1390 		} else {
1391 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1392 			if (error)
1393 				return (error);
1394 		}
1395 	}
1396 	req.validlen = req.oldlen;
1397 
1398 	if (old) {
1399 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1400 			return (EFAULT);
1401 		req.oldptr= old;
1402 	}
1403 
1404 	if (new != NULL) {
1405 		if (!useracc(new, newlen, VM_PROT_READ))
1406 			return (EFAULT);
1407 		req.newlen = newlen;
1408 		req.newptr = new;
1409 	}
1410 
1411 	req.oldfunc = sysctl_old_user;
1412 	req.newfunc = sysctl_new_user;
1413 	req.lock = REQ_LOCKED;
1414 
1415 	SYSCTL_LOCK();
1416 
1417 	do {
1418 		req.oldidx = 0;
1419 		req.newidx = 0;
1420 		error = sysctl_root(0, name, namelen, &req);
1421 	} while (error == EAGAIN);
1422 
1423 	if (req.lock == REQ_WIRED && req.validlen > 0)
1424 		vsunlock(req.oldptr, req.validlen);
1425 
1426 	SYSCTL_UNLOCK();
1427 
1428 	if (error && error != ENOMEM)
1429 		return (error);
1430 
1431 	if (retval) {
1432 		if (req.oldptr && req.oldidx > req.validlen)
1433 			*retval = req.validlen;
1434 		else
1435 			*retval = req.oldidx;
1436 	}
1437 	return (error);
1438 }
1439 
1440 #ifdef COMPAT_43
1441 #include <sys/socket.h>
1442 #include <vm/vm_param.h>
1443 
1444 #define	KINFO_PROC		(0<<8)
1445 #define	KINFO_RT		(1<<8)
1446 #define	KINFO_VNODE		(2<<8)
1447 #define	KINFO_FILE		(3<<8)
1448 #define	KINFO_METER		(4<<8)
1449 #define	KINFO_LOADAVG		(5<<8)
1450 #define	KINFO_CLOCKRATE		(6<<8)
1451 
1452 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
1453 #define	KINFO_BSDI_SYSINFO	(101<<8)
1454 
1455 /*
1456  * XXX this is bloat, but I hope it's better here than on the potentially
1457  * limited kernel stack...  -Peter
1458  */
1459 
1460 static struct {
1461 	int	bsdi_machine;		/* "i386" on BSD/386 */
1462 /*      ^^^ this is an offset to the string, relative to the struct start */
1463 	char	*pad0;
1464 	long	pad1;
1465 	long	pad2;
1466 	long	pad3;
1467 	u_long	pad4;
1468 	u_long	pad5;
1469 	u_long	pad6;
1470 
1471 	int	bsdi_ostype;		/* "BSD/386" on BSD/386 */
1472 	int	bsdi_osrelease;		/* "1.1" on BSD/386 */
1473 	long	pad7;
1474 	long	pad8;
1475 	char	*pad9;
1476 
1477 	long	pad10;
1478 	long	pad11;
1479 	int	pad12;
1480 	long	pad13;
1481 	quad_t	pad14;
1482 	long	pad15;
1483 
1484 	struct	timeval pad16;
1485 	/* we dont set this, because BSDI's uname used gethostname() instead */
1486 	int	bsdi_hostname;		/* hostname on BSD/386 */
1487 
1488 	/* the actual string data is appended here */
1489 
1490 } bsdi_si;
1491 
1492 /*
1493  * this data is appended to the end of the bsdi_si structure during copyout.
1494  * The "char *" offsets are relative to the base of the bsdi_si struct.
1495  * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
1496  * should not exceed the length of the buffer here... (or else!! :-)
1497  */
1498 static char bsdi_strings[80];	/* It had better be less than this! */
1499 
1500 #ifndef _SYS_SYSPROTO_H_
1501 struct getkerninfo_args {
1502 	int	op;
1503 	char	*where;
1504 	size_t	*size;
1505 	int	arg;
1506 };
1507 #endif
1508 int
1509 ogetkerninfo(struct thread *td, struct getkerninfo_args *uap)
1510 {
1511 	int error, name[6];
1512 	size_t size;
1513 	u_int needed = 0;
1514 
1515 	mtx_lock(&Giant);
1516 
1517 	switch (uap->op & 0xff00) {
1518 
1519 	case KINFO_RT:
1520 		name[0] = CTL_NET;
1521 		name[1] = PF_ROUTE;
1522 		name[2] = 0;
1523 		name[3] = (uap->op & 0xff0000) >> 16;
1524 		name[4] = uap->op & 0xff;
1525 		name[5] = uap->arg;
1526 		error = userland_sysctl(td, name, 6, uap->where, uap->size,
1527 			0, 0, 0, &size, 0);
1528 		break;
1529 
1530 	case KINFO_VNODE:
1531 		name[0] = CTL_KERN;
1532 		name[1] = KERN_VNODE;
1533 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1534 			0, 0, 0, &size, 0);
1535 		break;
1536 
1537 	case KINFO_PROC:
1538 		name[0] = CTL_KERN;
1539 		name[1] = KERN_PROC;
1540 		name[2] = uap->op & 0xff;
1541 		name[3] = uap->arg;
1542 		error = userland_sysctl(td, name, 4, uap->where, uap->size,
1543 			0, 0, 0, &size, 0);
1544 		break;
1545 
1546 	case KINFO_FILE:
1547 		name[0] = CTL_KERN;
1548 		name[1] = KERN_FILE;
1549 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1550 			0, 0, 0, &size, 0);
1551 		break;
1552 
1553 	case KINFO_METER:
1554 		name[0] = CTL_VM;
1555 		name[1] = VM_TOTAL;
1556 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1557 			0, 0, 0, &size, 0);
1558 		break;
1559 
1560 	case KINFO_LOADAVG:
1561 		name[0] = CTL_VM;
1562 		name[1] = VM_LOADAVG;
1563 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1564 			0, 0, 0, &size, 0);
1565 		break;
1566 
1567 	case KINFO_CLOCKRATE:
1568 		name[0] = CTL_KERN;
1569 		name[1] = KERN_CLOCKRATE;
1570 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1571 			0, 0, 0, &size, 0);
1572 		break;
1573 
1574 	case KINFO_BSDI_SYSINFO: {
1575 		/*
1576 		 * this is pretty crude, but it's just enough for uname()
1577 		 * from BSDI's 1.x libc to work.
1578 		 *
1579 		 * *size gives the size of the buffer before the call, and
1580 		 * the amount of data copied after a successful call.
1581 		 * If successful, the return value is the amount of data
1582 		 * available, which can be larger than *size.
1583 		 *
1584 		 * BSDI's 2.x product apparently fails with ENOMEM if *size
1585 		 * is too small.
1586 		 */
1587 
1588 		u_int left;
1589 		char *s;
1590 
1591 		bzero((char *)&bsdi_si, sizeof(bsdi_si));
1592 		bzero(bsdi_strings, sizeof(bsdi_strings));
1593 
1594 		s = bsdi_strings;
1595 
1596 		bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1597 		strcpy(s, ostype);
1598 		s += strlen(s) + 1;
1599 
1600 		bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1601 		strcpy(s, osrelease);
1602 		s += strlen(s) + 1;
1603 
1604 		bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1605 		strcpy(s, machine);
1606 		s += strlen(s) + 1;
1607 
1608 		needed = sizeof(bsdi_si) + (s - bsdi_strings);
1609 
1610 		if ((uap->where == NULL) || (uap->size == NULL)) {
1611 			/* process is asking how much buffer to supply.. */
1612 			size = needed;
1613 			error = 0;
1614 			break;
1615 		}
1616 
1617 		if ((error = copyin(uap->size, &size, sizeof(size))) != 0)
1618 			break;
1619 
1620 		/* if too much buffer supplied, trim it down */
1621 		if (size > needed)
1622 			size = needed;
1623 
1624 		/* how much of the buffer is remaining */
1625 		left = size;
1626 
1627 		if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1628 			break;
1629 
1630 		/* is there any point in continuing? */
1631 		if (left > sizeof(bsdi_si)) {
1632 			left -= sizeof(bsdi_si);
1633 			error = copyout(&bsdi_strings,
1634 					uap->where + sizeof(bsdi_si), left);
1635 		}
1636 		break;
1637 	}
1638 
1639 	default:
1640 		error = EOPNOTSUPP;
1641 		break;
1642 	}
1643 	if (error == 0) {
1644 		td->td_retval[0] = needed ? needed : size;
1645 		if (uap->size) {
1646 			error = copyout(&size, uap->size, sizeof(size));
1647 		}
1648 	}
1649 	mtx_unlock(&Giant);
1650 	return (error);
1651 }
1652 #endif /* COMPAT_43 */
1653