xref: /freebsd/sys/kern/kern_sysctl.c (revision a930f272def8f4c8915467084eb0c2222b39c86c)
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/proc.h>
50 #include <sys/lock.h>
51 #include <sys/mutex.h>
52 #include <sys/sx.h>
53 #include <sys/sysproto.h>
54 
55 #include <security/mac/mac_framework.h>
56 
57 #include <vm/vm.h>
58 #include <vm/vm_extern.h>
59 
60 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
61 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
62 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
63 
64 /*
65  * Locking - this locks the sysctl tree in memory.
66  */
67 static struct sx sysctllock;
68 
69 #define	SYSCTL_LOCK()		sx_xlock(&sysctllock)
70 #define	SYSCTL_UNLOCK()		sx_xunlock(&sysctllock)
71 #define	SYSCTL_INIT()		sx_init(&sysctllock, "sysctl lock")
72 
73 static int sysctl_root(SYSCTL_HANDLER_ARGS);
74 
75 struct sysctl_oid_list sysctl__children; /* root list */
76 
77 static struct sysctl_oid *
78 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
79 {
80 	struct sysctl_oid *oidp;
81 
82 	SLIST_FOREACH(oidp, list, oid_link) {
83 		if (strcmp(oidp->oid_name, name) == 0) {
84 			return (oidp);
85 		}
86 	}
87 	return (NULL);
88 }
89 
90 /*
91  * Initialization of the MIB tree.
92  *
93  * Order by number in each list.
94  */
95 
96 void
97 sysctl_register_oid(struct sysctl_oid *oidp)
98 {
99 	struct sysctl_oid_list *parent = oidp->oid_parent;
100 	struct sysctl_oid *p;
101 	struct sysctl_oid *q;
102 
103 	/*
104 	 * First check if another oid with the same name already
105 	 * exists in the parent's list.
106 	 */
107 	p = sysctl_find_oidname(oidp->oid_name, parent);
108 	if (p != NULL) {
109 		if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
110 			p->oid_refcnt++;
111 			return;
112 		} else {
113 			printf("can't re-use a leaf (%s)!\n", p->oid_name);
114 			return;
115 		}
116 	}
117 	/*
118 	 * If this oid has a number OID_AUTO, give it a number which
119 	 * is greater than any current oid.
120 	 * NOTE: DO NOT change the starting value here, change it in
121 	 * <sys/sysctl.h>, and make sure it is at least 256 to
122 	 * accomodate e.g. net.inet.raw as a static sysctl node.
123 	 */
124 	if (oidp->oid_number == OID_AUTO) {
125 		static int newoid = CTL_AUTO_START;
126 
127 		oidp->oid_number = newoid++;
128 		if (newoid == 0x7fffffff)
129 			panic("out of oids");
130 	}
131 #if 0
132 	else if (oidp->oid_number >= CTL_AUTO_START) {
133 		/* do not panic; this happens when unregistering sysctl sets */
134 		printf("static sysctl oid too high: %d", oidp->oid_number);
135 	}
136 #endif
137 
138 	/*
139 	 * Insert the oid into the parent's list in order.
140 	 */
141 	q = NULL;
142 	SLIST_FOREACH(p, parent, oid_link) {
143 		if (oidp->oid_number < p->oid_number)
144 			break;
145 		q = p;
146 	}
147 	if (q)
148 		SLIST_INSERT_AFTER(q, oidp, oid_link);
149 	else
150 		SLIST_INSERT_HEAD(parent, oidp, oid_link);
151 }
152 
153 void
154 sysctl_unregister_oid(struct sysctl_oid *oidp)
155 {
156 	struct sysctl_oid *p;
157 	int error;
158 
159 	error = ENOENT;
160 	if (oidp->oid_number == OID_AUTO) {
161 		error = EINVAL;
162 	} else {
163 		SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
164 			if (p == oidp) {
165 				SLIST_REMOVE(oidp->oid_parent, oidp,
166 				    sysctl_oid, oid_link);
167 				error = 0;
168 				break;
169 			}
170 		}
171 	}
172 
173 	/*
174 	 * This can happen when a module fails to register and is
175 	 * being unloaded afterwards.  It should not be a panic()
176 	 * for normal use.
177 	 */
178 	if (error)
179 		printf("%s: failed to unregister sysctl\n", __func__);
180 }
181 
182 /* Initialize a new context to keep track of dynamically added sysctls. */
183 int
184 sysctl_ctx_init(struct sysctl_ctx_list *c)
185 {
186 
187 	if (c == NULL) {
188 		return (EINVAL);
189 	}
190 	TAILQ_INIT(c);
191 	return (0);
192 }
193 
194 /* Free the context, and destroy all dynamic oids registered in this context */
195 int
196 sysctl_ctx_free(struct sysctl_ctx_list *clist)
197 {
198 	struct sysctl_ctx_entry *e, *e1;
199 	int error;
200 
201 	error = 0;
202 	/*
203 	 * First perform a "dry run" to check if it's ok to remove oids.
204 	 * XXX FIXME
205 	 * XXX This algorithm is a hack. But I don't know any
206 	 * XXX better solution for now...
207 	 */
208 	TAILQ_FOREACH(e, clist, link) {
209 		error = sysctl_remove_oid(e->entry, 0, 0);
210 		if (error)
211 			break;
212 	}
213 	/*
214 	 * Restore deregistered entries, either from the end,
215 	 * or from the place where error occured.
216 	 * e contains the entry that was not unregistered
217 	 */
218 	if (error)
219 		e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
220 	else
221 		e1 = TAILQ_LAST(clist, sysctl_ctx_list);
222 	while (e1 != NULL) {
223 		sysctl_register_oid(e1->entry);
224 		e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
225 	}
226 	if (error)
227 		return(EBUSY);
228 	/* Now really delete the entries */
229 	e = TAILQ_FIRST(clist);
230 	while (e != NULL) {
231 		e1 = TAILQ_NEXT(e, link);
232 		error = sysctl_remove_oid(e->entry, 1, 0);
233 		if (error)
234 			panic("sysctl_remove_oid: corrupt tree, entry: %s",
235 			    e->entry->oid_name);
236 		free(e, M_SYSCTLOID);
237 		e = e1;
238 	}
239 	return (error);
240 }
241 
242 /* Add an entry to the context */
243 struct sysctl_ctx_entry *
244 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
245 {
246 	struct sysctl_ctx_entry *e;
247 
248 	if (clist == NULL || oidp == NULL)
249 		return(NULL);
250 	e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
251 	e->entry = oidp;
252 	TAILQ_INSERT_HEAD(clist, e, link);
253 	return (e);
254 }
255 
256 /* Find an entry in the context */
257 struct sysctl_ctx_entry *
258 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
259 {
260 	struct sysctl_ctx_entry *e;
261 
262 	if (clist == NULL || oidp == NULL)
263 		return(NULL);
264 	TAILQ_FOREACH(e, clist, link) {
265 		if(e->entry == oidp)
266 			return(e);
267 	}
268 	return (e);
269 }
270 
271 /*
272  * Delete an entry from the context.
273  * NOTE: this function doesn't free oidp! You have to remove it
274  * with sysctl_remove_oid().
275  */
276 int
277 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
278 {
279 	struct sysctl_ctx_entry *e;
280 
281 	if (clist == NULL || oidp == NULL)
282 		return (EINVAL);
283 	e = sysctl_ctx_entry_find(clist, oidp);
284 	if (e != NULL) {
285 		TAILQ_REMOVE(clist, e, link);
286 		free(e, M_SYSCTLOID);
287 		return (0);
288 	} else
289 		return (ENOENT);
290 }
291 
292 /*
293  * Remove dynamically created sysctl trees.
294  * oidp - top of the tree to be removed
295  * del - if 0 - just deregister, otherwise free up entries as well
296  * recurse - if != 0 traverse the subtree to be deleted
297  */
298 int
299 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
300 {
301 	struct sysctl_oid *p;
302 	int error;
303 
304 	if (oidp == NULL)
305 		return(EINVAL);
306 	if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
307 		printf("can't remove non-dynamic nodes!\n");
308 		return (EINVAL);
309 	}
310 	/*
311 	 * WARNING: normal method to do this should be through
312 	 * sysctl_ctx_free(). Use recursing as the last resort
313 	 * method to purge your sysctl tree of leftovers...
314 	 * However, if some other code still references these nodes,
315 	 * it will panic.
316 	 */
317 	if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
318 		if (oidp->oid_refcnt == 1) {
319 			SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
320 				if (!recurse)
321 					return (ENOTEMPTY);
322 				error = sysctl_remove_oid(p, del, recurse);
323 				if (error)
324 					return (error);
325 			}
326 			if (del)
327 				free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
328 		}
329 	}
330 	if (oidp->oid_refcnt > 1 ) {
331 		oidp->oid_refcnt--;
332 	} else {
333 		if (oidp->oid_refcnt == 0) {
334 			printf("Warning: bad oid_refcnt=%u (%s)!\n",
335 				oidp->oid_refcnt, oidp->oid_name);
336 			return (EINVAL);
337 		}
338 		sysctl_unregister_oid(oidp);
339 		if (del) {
340 			if (oidp->oid_descr)
341 				free((void *)(uintptr_t)(const void *)oidp->oid_descr, M_SYSCTLOID);
342 			free((void *)(uintptr_t)(const void *)oidp->oid_name,
343 			     M_SYSCTLOID);
344 			free(oidp, M_SYSCTLOID);
345 		}
346 	}
347 	return (0);
348 }
349 
350 /*
351  * Create new sysctls at run time.
352  * clist may point to a valid context initialized with sysctl_ctx_init().
353  */
354 struct sysctl_oid *
355 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
356 	int number, const char *name, int kind, void *arg1, int arg2,
357 	int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
358 {
359 	struct sysctl_oid *oidp;
360 	ssize_t len;
361 	char *newname;
362 
363 	/* You have to hook up somewhere.. */
364 	if (parent == NULL)
365 		return(NULL);
366 	/* Check if the node already exists, otherwise create it */
367 	oidp = sysctl_find_oidname(name, parent);
368 	if (oidp != NULL) {
369 		if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
370 			oidp->oid_refcnt++;
371 			/* Update the context */
372 			if (clist != NULL)
373 				sysctl_ctx_entry_add(clist, oidp);
374 			return (oidp);
375 		} else {
376 			printf("can't re-use a leaf (%s)!\n", name);
377 			return (NULL);
378 		}
379 	}
380 	oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
381 	oidp->oid_parent = parent;
382 	SLIST_NEXT(oidp, oid_link) = NULL;
383 	oidp->oid_number = number;
384 	oidp->oid_refcnt = 1;
385 	len = strlen(name);
386 	newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK);
387 	bcopy(name, newname, len + 1);
388 	newname[len] = '\0';
389 	oidp->oid_name = newname;
390 	oidp->oid_handler = handler;
391 	oidp->oid_kind = CTLFLAG_DYN | kind;
392 	if ((kind & CTLTYPE) == CTLTYPE_NODE) {
393 		/* Allocate space for children */
394 		SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
395 		    M_SYSCTLOID, M_WAITOK));
396 		SLIST_INIT(SYSCTL_CHILDREN(oidp));
397 	} else {
398 		oidp->oid_arg1 = arg1;
399 		oidp->oid_arg2 = arg2;
400 	}
401 	oidp->oid_fmt = fmt;
402 	if (descr) {
403 		int len = strlen(descr) + 1;
404 		oidp->oid_descr = malloc(len, M_SYSCTLOID, M_WAITOK);
405 		if (oidp->oid_descr)
406 			strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
407 	}
408 	/* Update the context, if used */
409 	if (clist != NULL)
410 		sysctl_ctx_entry_add(clist, oidp);
411 	/* Register this oid */
412 	sysctl_register_oid(oidp);
413 	return (oidp);
414 }
415 
416 /*
417  * Reparent an existing oid.
418  */
419 int
420 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
421 {
422 	struct sysctl_oid *oidp;
423 
424 	if (oid->oid_parent == parent)
425 		return (0);
426 	oidp = sysctl_find_oidname(oid->oid_name, parent);
427 	if (oidp != NULL)
428 		return (EEXIST);
429 	sysctl_unregister_oid(oid);
430 	oid->oid_parent = parent;
431 	oid->oid_number = OID_AUTO;
432 	sysctl_register_oid(oid);
433 	return (0);
434 }
435 
436 /*
437  * Register the kernel's oids on startup.
438  */
439 SET_DECLARE(sysctl_set, struct sysctl_oid);
440 
441 static void
442 sysctl_register_all(void *arg)
443 {
444 	struct sysctl_oid **oidp;
445 
446 	SYSCTL_INIT();
447 	SET_FOREACH(oidp, sysctl_set)
448 		sysctl_register_oid(*oidp);
449 }
450 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
451 
452 /*
453  * "Staff-functions"
454  *
455  * These functions implement a presently undocumented interface
456  * used by the sysctl program to walk the tree, and get the type
457  * so it can print the value.
458  * This interface is under work and consideration, and should probably
459  * be killed with a big axe by the first person who can find the time.
460  * (be aware though, that the proper interface isn't as obvious as it
461  * may seem, there are various conflicting requirements.
462  *
463  * {0,0}	printf the entire MIB-tree.
464  * {0,1,...}	return the name of the "..." OID.
465  * {0,2,...}	return the next OID.
466  * {0,3}	return the OID of the name in "new"
467  * {0,4,...}	return the kind & format info for the "..." OID.
468  * {0,5,...}	return the description the "..." OID.
469  */
470 
471 #ifdef SYSCTL_DEBUG
472 static void
473 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
474 {
475 	int k;
476 	struct sysctl_oid *oidp;
477 
478 	SLIST_FOREACH(oidp, l, oid_link) {
479 
480 		for (k=0; k<i; k++)
481 			printf(" ");
482 
483 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
484 
485 		printf("%c%c",
486 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
487 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
488 
489 		if (oidp->oid_handler)
490 			printf(" *Handler");
491 
492 		switch (oidp->oid_kind & CTLTYPE) {
493 			case CTLTYPE_NODE:
494 				printf(" Node\n");
495 				if (!oidp->oid_handler) {
496 					sysctl_sysctl_debug_dump_node(
497 						oidp->oid_arg1, i+2);
498 				}
499 				break;
500 			case CTLTYPE_INT:    printf(" Int\n"); break;
501 			case CTLTYPE_STRING: printf(" String\n"); break;
502 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
503 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
504 			default:	     printf("\n");
505 		}
506 
507 	}
508 }
509 
510 static int
511 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
512 {
513 	int error;
514 
515 	error = suser(req->td);
516 	if (error)
517 		return (error);
518 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
519 	return (ENOENT);
520 }
521 
522 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
523 	0, 0, sysctl_sysctl_debug, "-", "");
524 #endif
525 
526 static int
527 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
528 {
529 	int *name = (int *) arg1;
530 	u_int namelen = arg2;
531 	int error = 0;
532 	struct sysctl_oid *oid;
533 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
534 	char buf[10];
535 
536 	while (namelen) {
537 		if (!lsp) {
538 			snprintf(buf,sizeof(buf),"%d",*name);
539 			if (req->oldidx)
540 				error = SYSCTL_OUT(req, ".", 1);
541 			if (!error)
542 				error = SYSCTL_OUT(req, buf, strlen(buf));
543 			if (error)
544 				return (error);
545 			namelen--;
546 			name++;
547 			continue;
548 		}
549 		lsp2 = 0;
550 		SLIST_FOREACH(oid, lsp, oid_link) {
551 			if (oid->oid_number != *name)
552 				continue;
553 
554 			if (req->oldidx)
555 				error = SYSCTL_OUT(req, ".", 1);
556 			if (!error)
557 				error = SYSCTL_OUT(req, oid->oid_name,
558 					strlen(oid->oid_name));
559 			if (error)
560 				return (error);
561 
562 			namelen--;
563 			name++;
564 
565 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
566 				break;
567 
568 			if (oid->oid_handler)
569 				break;
570 
571 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
572 			break;
573 		}
574 		lsp = lsp2;
575 	}
576 	return (SYSCTL_OUT(req, "", 1));
577 }
578 
579 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
580 
581 static int
582 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
583 	int *next, int *len, int level, struct sysctl_oid **oidpp)
584 {
585 	struct sysctl_oid *oidp;
586 
587 	*len = level;
588 	SLIST_FOREACH(oidp, lsp, oid_link) {
589 		*next = oidp->oid_number;
590 		*oidpp = oidp;
591 
592 		if (oidp->oid_kind & CTLFLAG_SKIP)
593 			continue;
594 
595 		if (!namelen) {
596 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
597 				return (0);
598 			if (oidp->oid_handler)
599 				/* We really should call the handler here...*/
600 				return (0);
601 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
602 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
603 				len, level+1, oidpp))
604 				return (0);
605 			goto emptynode;
606 		}
607 
608 		if (oidp->oid_number < *name)
609 			continue;
610 
611 		if (oidp->oid_number > *name) {
612 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
613 				return (0);
614 			if (oidp->oid_handler)
615 				return (0);
616 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
617 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
618 				next+1, len, level+1, oidpp))
619 				return (0);
620 			goto next;
621 		}
622 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
623 			continue;
624 
625 		if (oidp->oid_handler)
626 			continue;
627 
628 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
629 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
630 			len, level+1, oidpp))
631 			return (0);
632 	next:
633 		namelen = 1;
634 	emptynode:
635 		*len = level;
636 	}
637 	return (1);
638 }
639 
640 static int
641 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
642 {
643 	int *name = (int *) arg1;
644 	u_int namelen = arg2;
645 	int i, j, error;
646 	struct sysctl_oid *oid;
647 	struct sysctl_oid_list *lsp = &sysctl__children;
648 	int newoid[CTL_MAXNAME];
649 
650 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
651 	if (i)
652 		return (ENOENT);
653 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
654 	return (error);
655 }
656 
657 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
658 
659 static int
660 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
661 {
662 	int i;
663 	struct sysctl_oid *oidp;
664 	struct sysctl_oid_list *lsp = &sysctl__children;
665 	char *p;
666 
667 	if (!*name)
668 		return (ENOENT);
669 
670 	p = name + strlen(name) - 1 ;
671 	if (*p == '.')
672 		*p = '\0';
673 
674 	*len = 0;
675 
676 	for (p = name; *p && *p != '.'; p++)
677 		;
678 	i = *p;
679 	if (i == '.')
680 		*p = '\0';
681 
682 	oidp = SLIST_FIRST(lsp);
683 
684 	while (oidp && *len < CTL_MAXNAME) {
685 		if (strcmp(name, oidp->oid_name)) {
686 			oidp = SLIST_NEXT(oidp, oid_link);
687 			continue;
688 		}
689 		*oid++ = oidp->oid_number;
690 		(*len)++;
691 
692 		if (!i) {
693 			if (oidpp)
694 				*oidpp = oidp;
695 			return (0);
696 		}
697 
698 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
699 			break;
700 
701 		if (oidp->oid_handler)
702 			break;
703 
704 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
705 		oidp = SLIST_FIRST(lsp);
706 		name = p+1;
707 		for (p = name; *p && *p != '.'; p++)
708 				;
709 		i = *p;
710 		if (i == '.')
711 			*p = '\0';
712 	}
713 	return (ENOENT);
714 }
715 
716 static int
717 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
718 {
719 	char *p;
720 	int error, oid[CTL_MAXNAME], len;
721 	struct sysctl_oid *op = 0;
722 
723 	if (!req->newlen)
724 		return (ENOENT);
725 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
726 		return (ENAMETOOLONG);
727 
728 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
729 
730 	error = SYSCTL_IN(req, p, req->newlen);
731 	if (error) {
732 		free(p, M_SYSCTL);
733 		return (error);
734 	}
735 
736 	p [req->newlen] = '\0';
737 
738 	error = name2oid(p, oid, &len, &op);
739 
740 	free(p, M_SYSCTL);
741 
742 	if (error)
743 		return (error);
744 
745 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
746 	return (error);
747 }
748 
749 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
750 	sysctl_sysctl_name2oid, "I", "");
751 
752 static int
753 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
754 {
755 	struct sysctl_oid *oid;
756 	int error;
757 
758 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
759 	if (error)
760 		return (error);
761 
762 	if (!oid->oid_fmt)
763 		return (ENOENT);
764 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
765 	if (error)
766 		return (error);
767 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
768 	return (error);
769 }
770 
771 
772 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
773 
774 static int
775 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
776 {
777 	struct sysctl_oid *oid;
778 	int error;
779 
780 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
781 	if (error)
782 		return (error);
783 
784 	if (!oid->oid_descr)
785 		return (ENOENT);
786 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
787 	return (error);
788 }
789 
790 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
791 
792 /*
793  * Default "handler" functions.
794  */
795 
796 /*
797  * Handle an int, signed or unsigned.
798  * Two cases:
799  *     a variable:  point arg1 at it.
800  *     a constant:  pass it in arg2.
801  */
802 
803 int
804 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
805 {
806 	int tmpout, error = 0;
807 
808 	/*
809 	 * Attempt to get a coherent snapshot by making a copy of the data.
810 	 */
811 	if (arg1)
812 		tmpout = *(int *)arg1;
813 	else
814 		tmpout = arg2;
815 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
816 
817 	if (error || !req->newptr)
818 		return (error);
819 
820 	if (!arg1)
821 		error = EPERM;
822 	else
823 		error = SYSCTL_IN(req, arg1, sizeof(int));
824 	return (error);
825 }
826 
827 
828 /*
829  * Based on on sysctl_handle_int() convert milliseconds into ticks.
830  */
831 
832 int
833 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
834 {
835 	int error, s, tt;
836 
837 	tt = *(int *)oidp->oid_arg1;
838 	s = (int)((int64_t)tt * 1000 / hz);
839 
840 	error = sysctl_handle_int(oidp, &s, 0, req);
841 	if (error || !req->newptr)
842 		return (error);
843 
844 	tt = (int)((int64_t)s * hz / 1000);
845 	if (tt < 1)
846 		return (EINVAL);
847 
848 	*(int *)oidp->oid_arg1 = tt;
849 	return (0);
850 }
851 
852 
853 /*
854  * Handle a long, signed or unsigned.  arg1 points to it.
855  */
856 
857 int
858 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
859 {
860 	int error = 0;
861 	long tmplong;
862 #ifdef SCTL_MASK32
863 	int tmpint;
864 #endif
865 
866 	/*
867 	 * Attempt to get a coherent snapshot by making a copy of the data.
868 	 */
869 	if (!arg1)
870 		return (EINVAL);
871 	tmplong = *(long *)arg1;
872 #ifdef SCTL_MASK32
873 	if (req->flags & SCTL_MASK32) {
874 		tmpint = tmplong;
875 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
876 	} else
877 #endif
878 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
879 
880 	if (error || !req->newptr)
881 		return (error);
882 
883 #ifdef SCTL_MASK32
884 	if (req->flags & SCTL_MASK32) {
885 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
886 		*(long *)arg1 = (long)tmpint;
887 	} else
888 #endif
889 		error = SYSCTL_IN(req, arg1, sizeof(long));
890 	return (error);
891 }
892 
893 /*
894  * Handle our generic '\0' terminated 'C' string.
895  * Two cases:
896  * 	a variable string:  point arg1 at it, arg2 is max length.
897  * 	a constant string:  point arg1 at it, arg2 is zero.
898  */
899 
900 int
901 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
902 {
903 	int error=0;
904 	char *tmparg;
905 	size_t outlen;
906 
907 	/*
908 	 * Attempt to get a coherent snapshot by copying to a
909 	 * temporary kernel buffer.
910 	 */
911 retry:
912 	outlen = strlen((char *)arg1)+1;
913 	tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
914 
915 	if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
916 		free(tmparg, M_SYSCTLTMP);
917 		goto retry;
918 	}
919 
920 	error = SYSCTL_OUT(req, tmparg, outlen);
921 	free(tmparg, M_SYSCTLTMP);
922 
923 	if (error || !req->newptr)
924 		return (error);
925 
926 	if ((req->newlen - req->newidx) >= arg2) {
927 		error = EINVAL;
928 	} else {
929 		arg2 = (req->newlen - req->newidx);
930 		error = SYSCTL_IN(req, arg1, arg2);
931 		((char *)arg1)[arg2] = '\0';
932 	}
933 
934 	return (error);
935 }
936 
937 /*
938  * Handle any kind of opaque data.
939  * arg1 points to it, arg2 is the size.
940  */
941 
942 int
943 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
944 {
945 	int error, tries;
946 	u_int generation;
947 	struct sysctl_req req2;
948 
949 	/*
950 	 * Attempt to get a coherent snapshot, by using the thread
951 	 * pre-emption counter updated from within mi_switch() to
952 	 * determine if we were pre-empted during a bcopy() or
953 	 * copyout(). Make 3 attempts at doing this before giving up.
954 	 * If we encounter an error, stop immediately.
955 	 */
956 	tries = 0;
957 	req2 = *req;
958 retry:
959 	generation = curthread->td_generation;
960 	error = SYSCTL_OUT(req, arg1, arg2);
961 	if (error)
962 		return (error);
963 	tries++;
964 	if (generation != curthread->td_generation && tries < 3) {
965 		*req = req2;
966 		goto retry;
967 	}
968 
969 	error = SYSCTL_IN(req, arg1, arg2);
970 
971 	return (error);
972 }
973 
974 /*
975  * Transfer functions to/from kernel space.
976  * XXX: rather untested at this point
977  */
978 static int
979 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
980 {
981 	size_t i = 0;
982 
983 	if (req->oldptr) {
984 		i = l;
985 		if (req->oldlen <= req->oldidx)
986 			i = 0;
987 		else
988 			if (i > req->oldlen - req->oldidx)
989 				i = req->oldlen - req->oldidx;
990 		if (i > 0)
991 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
992 	}
993 	req->oldidx += l;
994 	if (req->oldptr && i != l)
995 		return (ENOMEM);
996 	return (0);
997 }
998 
999 static int
1000 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1001 {
1002 	if (!req->newptr)
1003 		return (0);
1004 	if (req->newlen - req->newidx < l)
1005 		return (EINVAL);
1006 	bcopy((char *)req->newptr + req->newidx, p, l);
1007 	req->newidx += l;
1008 	return (0);
1009 }
1010 
1011 int
1012 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1013     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1014 {
1015 	int error = 0;
1016 	struct sysctl_req req;
1017 
1018 	bzero(&req, sizeof req);
1019 
1020 	req.td = td;
1021 	req.flags = flags;
1022 
1023 	if (oldlenp) {
1024 		req.oldlen = *oldlenp;
1025 	}
1026 	req.validlen = req.oldlen;
1027 
1028 	if (old) {
1029 		req.oldptr= old;
1030 	}
1031 
1032 	if (new != NULL) {
1033 		req.newlen = newlen;
1034 		req.newptr = new;
1035 	}
1036 
1037 	req.oldfunc = sysctl_old_kernel;
1038 	req.newfunc = sysctl_new_kernel;
1039 	req.lock = REQ_LOCKED;
1040 
1041 	SYSCTL_LOCK();
1042 
1043 	error = sysctl_root(0, name, namelen, &req);
1044 
1045 	if (req.lock == REQ_WIRED && req.validlen > 0)
1046 		vsunlock(req.oldptr, req.validlen);
1047 
1048 	SYSCTL_UNLOCK();
1049 
1050 	if (error && error != ENOMEM)
1051 		return (error);
1052 
1053 	if (retval) {
1054 		if (req.oldptr && req.oldidx > req.validlen)
1055 			*retval = req.validlen;
1056 		else
1057 			*retval = req.oldidx;
1058 	}
1059 	return (error);
1060 }
1061 
1062 int
1063 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1064     void *new, size_t newlen, size_t *retval, int flags)
1065 {
1066         int oid[CTL_MAXNAME];
1067         size_t oidlen, plen;
1068 	int error;
1069 
1070 	oid[0] = 0;		/* sysctl internal magic */
1071 	oid[1] = 3;		/* name2oid */
1072 	oidlen = sizeof(oid);
1073 
1074 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1075 	    (void *)name, strlen(name), &plen, flags);
1076 	if (error)
1077 		return (error);
1078 
1079 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1080 	    new, newlen, retval, flags);
1081 	return (error);
1082 }
1083 
1084 /*
1085  * Transfer function to/from user space.
1086  */
1087 static int
1088 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1089 {
1090 	int error = 0;
1091 	size_t i, len, origidx;
1092 
1093 	origidx = req->oldidx;
1094 	req->oldidx += l;
1095 	if (req->oldptr == NULL)
1096 		return (0);
1097 	/*
1098 	 * If we have not wired the user supplied buffer and we are currently
1099 	 * holding locks, drop a witness warning, as it's possible that
1100 	 * write operations to the user page can sleep.
1101 	 */
1102 	if (req->lock != REQ_WIRED)
1103 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1104 		    "sysctl_old_user()");
1105 	i = l;
1106 	len = req->validlen;
1107 	if (len <= origidx)
1108 		i = 0;
1109 	else {
1110 		if (i > len - origidx)
1111 			i = len - origidx;
1112 		error = copyout(p, (char *)req->oldptr + origidx, i);
1113 	}
1114 	if (error)
1115 		return (error);
1116 	if (i < l)
1117 		return (ENOMEM);
1118 	return (0);
1119 }
1120 
1121 static int
1122 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1123 {
1124 	int error;
1125 
1126 	if (!req->newptr)
1127 		return (0);
1128 	if (req->newlen - req->newidx < l)
1129 		return (EINVAL);
1130 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1131 	    "sysctl_new_user()");
1132 	error = copyin((char *)req->newptr + req->newidx, p, l);
1133 	req->newidx += l;
1134 	return (error);
1135 }
1136 
1137 /*
1138  * Wire the user space destination buffer.  If set to a value greater than
1139  * zero, the len parameter limits the maximum amount of wired memory.
1140  */
1141 int
1142 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1143 {
1144 	int ret;
1145 	size_t i, wiredlen;
1146 	char *cp, dummy;
1147 
1148 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1149 	ret = 0;
1150 	if (req->lock == REQ_LOCKED && req->oldptr &&
1151 	    req->oldfunc == sysctl_old_user) {
1152 		if (wiredlen != 0) {
1153 			ret = vslock(req->oldptr, wiredlen);
1154 			if (ret != 0) {
1155 				if (ret != ENOMEM)
1156 					return (ret);
1157 				wiredlen = 0;
1158 			}
1159 			/*
1160 			 * Touch all the wired pages to avoid PTE modified
1161 			 * bit emulation traps on Alpha while holding locks
1162 			 * in the sysctl handler.
1163 			 */
1164 			for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1165 			    cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1166 				copyin(cp, &dummy, 1);
1167 				copyout(&dummy, cp, 1);
1168 			}
1169 		}
1170 		req->lock = REQ_WIRED;
1171 		req->validlen = wiredlen;
1172 	}
1173 	return (0);
1174 }
1175 
1176 int
1177 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1178     int *nindx, struct sysctl_req *req)
1179 {
1180 	struct sysctl_oid *oid;
1181 	int indx;
1182 
1183 	oid = SLIST_FIRST(&sysctl__children);
1184 	indx = 0;
1185 	while (oid && indx < CTL_MAXNAME) {
1186 		if (oid->oid_number == name[indx]) {
1187 			indx++;
1188 			if (oid->oid_kind & CTLFLAG_NOLOCK)
1189 				req->lock = REQ_UNLOCKED;
1190 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1191 				if (oid->oid_handler != NULL ||
1192 				    indx == namelen) {
1193 					*noid = oid;
1194 					if (nindx != NULL)
1195 						*nindx = indx;
1196 					return (0);
1197 				}
1198 				oid = SLIST_FIRST(
1199 				    (struct sysctl_oid_list *)oid->oid_arg1);
1200 			} else if (indx == namelen) {
1201 				*noid = oid;
1202 				if (nindx != NULL)
1203 					*nindx = indx;
1204 				return (0);
1205 			} else {
1206 				return (ENOTDIR);
1207 			}
1208 		} else {
1209 			oid = SLIST_NEXT(oid, oid_link);
1210 		}
1211 	}
1212 	return (ENOENT);
1213 }
1214 
1215 /*
1216  * Traverse our tree, and find the right node, execute whatever it points
1217  * to, and return the resulting error code.
1218  */
1219 
1220 static int
1221 sysctl_root(SYSCTL_HANDLER_ARGS)
1222 {
1223 	struct sysctl_oid *oid;
1224 	int error, indx, lvl;
1225 
1226 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1227 	if (error)
1228 		return (error);
1229 
1230 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1231 		/*
1232 		 * You can't call a sysctl when it's a node, but has
1233 		 * no handler.  Inform the user that it's a node.
1234 		 * The indx may or may not be the same as namelen.
1235 		 */
1236 		if (oid->oid_handler == NULL)
1237 			return (EISDIR);
1238 	}
1239 
1240 	/* Is this sysctl writable? */
1241 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1242 		return (EPERM);
1243 
1244 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1245 
1246 	/* Is this sysctl sensitive to securelevels? */
1247 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1248 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1249 		error = securelevel_gt(req->td->td_ucred, lvl);
1250 		if (error)
1251 			return (error);
1252 	}
1253 
1254 	/* Is this sysctl writable by only privileged users? */
1255 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1256 		int flags;
1257 
1258 		if (oid->oid_kind & CTLFLAG_PRISON)
1259 			flags = SUSER_ALLOWJAIL;
1260 		else
1261 			flags = 0;
1262 		error = suser_cred(req->td->td_ucred, flags);
1263 		if (error)
1264 			return (error);
1265 	}
1266 
1267 	if (!oid->oid_handler)
1268 		return (EINVAL);
1269 
1270 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1271 		arg1 = (int *)arg1 + indx;
1272 		arg2 -= indx;
1273 	} else {
1274 		arg1 = oid->oid_arg1;
1275 		arg2 = oid->oid_arg2;
1276 	}
1277 #ifdef MAC
1278 	error = mac_check_system_sysctl(req->td->td_ucred, oid, arg1, arg2,
1279 	    req);
1280 	if (error != 0)
1281 		return (error);
1282 #endif
1283 	error = oid->oid_handler(oid, arg1, arg2, req);
1284 
1285 	return (error);
1286 }
1287 
1288 #ifndef _SYS_SYSPROTO_H_
1289 struct sysctl_args {
1290 	int	*name;
1291 	u_int	namelen;
1292 	void	*old;
1293 	size_t	*oldlenp;
1294 	void	*new;
1295 	size_t	newlen;
1296 };
1297 #endif
1298 
1299 /*
1300  * MPSAFE
1301  */
1302 int
1303 __sysctl(struct thread *td, struct sysctl_args *uap)
1304 {
1305 	int error, name[CTL_MAXNAME];
1306 	size_t j;
1307 
1308 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1309 		return (EINVAL);
1310 
1311  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1312  	if (error)
1313 		return (error);
1314 
1315 	mtx_lock(&Giant);
1316 
1317 	error = userland_sysctl(td, name, uap->namelen,
1318 		uap->old, uap->oldlenp, 0,
1319 		uap->new, uap->newlen, &j, 0);
1320 	if (error && error != ENOMEM)
1321 		goto done2;
1322 	if (uap->oldlenp) {
1323 		int i = copyout(&j, uap->oldlenp, sizeof(j));
1324 		if (i)
1325 			error = i;
1326 	}
1327 done2:
1328 	mtx_unlock(&Giant);
1329 	return (error);
1330 }
1331 
1332 /*
1333  * This is used from various compatibility syscalls too.  That's why name
1334  * must be in kernel space.
1335  */
1336 int
1337 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1338     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1339     int flags)
1340 {
1341 	int error = 0;
1342 	struct sysctl_req req;
1343 
1344 	bzero(&req, sizeof req);
1345 
1346 	req.td = td;
1347 	req.flags = flags;
1348 
1349 	if (oldlenp) {
1350 		if (inkernel) {
1351 			req.oldlen = *oldlenp;
1352 		} else {
1353 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1354 			if (error)
1355 				return (error);
1356 		}
1357 	}
1358 	req.validlen = req.oldlen;
1359 
1360 	if (old) {
1361 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1362 			return (EFAULT);
1363 		req.oldptr= old;
1364 	}
1365 
1366 	if (new != NULL) {
1367 		if (!useracc(new, req.newlen, VM_PROT_READ))
1368 			return (EFAULT);
1369 		req.newlen = newlen;
1370 		req.newptr = new;
1371 	}
1372 
1373 	req.oldfunc = sysctl_old_user;
1374 	req.newfunc = sysctl_new_user;
1375 	req.lock = REQ_LOCKED;
1376 
1377 	SYSCTL_LOCK();
1378 
1379 	do {
1380 		req.oldidx = 0;
1381 		req.newidx = 0;
1382 		error = sysctl_root(0, name, namelen, &req);
1383 	} while (error == EAGAIN);
1384 
1385 	if (req.lock == REQ_WIRED && req.validlen > 0)
1386 		vsunlock(req.oldptr, req.validlen);
1387 
1388 	SYSCTL_UNLOCK();
1389 
1390 	if (error && error != ENOMEM)
1391 		return (error);
1392 
1393 	if (retval) {
1394 		if (req.oldptr && req.oldidx > req.validlen)
1395 			*retval = req.validlen;
1396 		else
1397 			*retval = req.oldidx;
1398 	}
1399 	return (error);
1400 }
1401 
1402 #ifdef COMPAT_43
1403 #include <sys/socket.h>
1404 #include <vm/vm_param.h>
1405 
1406 #define	KINFO_PROC		(0<<8)
1407 #define	KINFO_RT		(1<<8)
1408 #define	KINFO_VNODE		(2<<8)
1409 #define	KINFO_FILE		(3<<8)
1410 #define	KINFO_METER		(4<<8)
1411 #define	KINFO_LOADAVG		(5<<8)
1412 #define	KINFO_CLOCKRATE		(6<<8)
1413 
1414 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
1415 #define	KINFO_BSDI_SYSINFO	(101<<8)
1416 
1417 /*
1418  * XXX this is bloat, but I hope it's better here than on the potentially
1419  * limited kernel stack...  -Peter
1420  */
1421 
1422 static struct {
1423 	int	bsdi_machine;		/* "i386" on BSD/386 */
1424 /*      ^^^ this is an offset to the string, relative to the struct start */
1425 	char	*pad0;
1426 	long	pad1;
1427 	long	pad2;
1428 	long	pad3;
1429 	u_long	pad4;
1430 	u_long	pad5;
1431 	u_long	pad6;
1432 
1433 	int	bsdi_ostype;		/* "BSD/386" on BSD/386 */
1434 	int	bsdi_osrelease;		/* "1.1" on BSD/386 */
1435 	long	pad7;
1436 	long	pad8;
1437 	char	*pad9;
1438 
1439 	long	pad10;
1440 	long	pad11;
1441 	int	pad12;
1442 	long	pad13;
1443 	quad_t	pad14;
1444 	long	pad15;
1445 
1446 	struct	timeval pad16;
1447 	/* we dont set this, because BSDI's uname used gethostname() instead */
1448 	int	bsdi_hostname;		/* hostname on BSD/386 */
1449 
1450 	/* the actual string data is appended here */
1451 
1452 } bsdi_si;
1453 /*
1454  * this data is appended to the end of the bsdi_si structure during copyout.
1455  * The "char *" offsets are relative to the base of the bsdi_si struct.
1456  * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
1457  * should not exceed the length of the buffer here... (or else!! :-)
1458  */
1459 static char bsdi_strings[80];	/* It had better be less than this! */
1460 
1461 #ifndef _SYS_SYSPROTO_H_
1462 struct getkerninfo_args {
1463 	int	op;
1464 	char	*where;
1465 	size_t	*size;
1466 	int	arg;
1467 };
1468 #endif
1469 
1470 /*
1471  * MPSAFE
1472  */
1473 int
1474 ogetkerninfo(struct thread *td, struct getkerninfo_args *uap)
1475 {
1476 	int error, name[6];
1477 	size_t size;
1478 	u_int needed = 0;
1479 
1480 	mtx_lock(&Giant);
1481 
1482 	switch (uap->op & 0xff00) {
1483 
1484 	case KINFO_RT:
1485 		name[0] = CTL_NET;
1486 		name[1] = PF_ROUTE;
1487 		name[2] = 0;
1488 		name[3] = (uap->op & 0xff0000) >> 16;
1489 		name[4] = uap->op & 0xff;
1490 		name[5] = uap->arg;
1491 		error = userland_sysctl(td, name, 6, uap->where, uap->size,
1492 			0, 0, 0, &size, 0);
1493 		break;
1494 
1495 	case KINFO_VNODE:
1496 		name[0] = CTL_KERN;
1497 		name[1] = KERN_VNODE;
1498 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1499 			0, 0, 0, &size, 0);
1500 		break;
1501 
1502 	case KINFO_PROC:
1503 		name[0] = CTL_KERN;
1504 		name[1] = KERN_PROC;
1505 		name[2] = uap->op & 0xff;
1506 		name[3] = uap->arg;
1507 		error = userland_sysctl(td, name, 4, uap->where, uap->size,
1508 			0, 0, 0, &size, 0);
1509 		break;
1510 
1511 	case KINFO_FILE:
1512 		name[0] = CTL_KERN;
1513 		name[1] = KERN_FILE;
1514 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1515 			0, 0, 0, &size, 0);
1516 		break;
1517 
1518 	case KINFO_METER:
1519 		name[0] = CTL_VM;
1520 		name[1] = VM_TOTAL;
1521 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1522 			0, 0, 0, &size, 0);
1523 		break;
1524 
1525 	case KINFO_LOADAVG:
1526 		name[0] = CTL_VM;
1527 		name[1] = VM_LOADAVG;
1528 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1529 			0, 0, 0, &size, 0);
1530 		break;
1531 
1532 	case KINFO_CLOCKRATE:
1533 		name[0] = CTL_KERN;
1534 		name[1] = KERN_CLOCKRATE;
1535 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1536 			0, 0, 0, &size, 0);
1537 		break;
1538 
1539 	case KINFO_BSDI_SYSINFO: {
1540 		/*
1541 		 * this is pretty crude, but it's just enough for uname()
1542 		 * from BSDI's 1.x libc to work.
1543 		 *
1544 		 * *size gives the size of the buffer before the call, and
1545 		 * the amount of data copied after a successful call.
1546 		 * If successful, the return value is the amount of data
1547 		 * available, which can be larger than *size.
1548 		 *
1549 		 * BSDI's 2.x product apparently fails with ENOMEM if *size
1550 		 * is too small.
1551 		 */
1552 
1553 		u_int left;
1554 		char *s;
1555 
1556 		bzero((char *)&bsdi_si, sizeof(bsdi_si));
1557 		bzero(bsdi_strings, sizeof(bsdi_strings));
1558 
1559 		s = bsdi_strings;
1560 
1561 		bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1562 		strcpy(s, ostype);
1563 		s += strlen(s) + 1;
1564 
1565 		bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1566 		strcpy(s, osrelease);
1567 		s += strlen(s) + 1;
1568 
1569 		bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1570 		strcpy(s, machine);
1571 		s += strlen(s) + 1;
1572 
1573 		needed = sizeof(bsdi_si) + (s - bsdi_strings);
1574 
1575 		if ((uap->where == NULL) || (uap->size == NULL)) {
1576 			/* process is asking how much buffer to supply.. */
1577 			size = needed;
1578 			error = 0;
1579 			break;
1580 		}
1581 
1582 		if ((error = copyin(uap->size, &size, sizeof(size))) != 0)
1583 			break;
1584 
1585 		/* if too much buffer supplied, trim it down */
1586 		if (size > needed)
1587 			size = needed;
1588 
1589 		/* how much of the buffer is remaining */
1590 		left = size;
1591 
1592 		if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1593 			break;
1594 
1595 		/* is there any point in continuing? */
1596 		if (left > sizeof(bsdi_si)) {
1597 			left -= sizeof(bsdi_si);
1598 			error = copyout(&bsdi_strings,
1599 					uap->where + sizeof(bsdi_si), left);
1600 		}
1601 		break;
1602 	}
1603 
1604 	default:
1605 		error = EOPNOTSUPP;
1606 		break;
1607 	}
1608 	if (error == 0) {
1609 		td->td_retval[0] = needed ? needed : size;
1610 		if (uap->size) {
1611 			error = copyout(&size, uap->size, sizeof(size));
1612 		}
1613 	}
1614 	mtx_unlock(&Giant);
1615 	return (error);
1616 }
1617 #endif /* COMPAT_43 */
1618