xref: /freebsd/sys/kern/kern_sysctl.c (revision 35a04710d7286aa9538917fd7f8e417dbee95b82)
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  * Reparent an existing oid.
419  */
420 int
421 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
422 {
423 	struct sysctl_oid *oidp;
424 
425 	if (oid->oid_parent == parent)
426 		return (0);
427 	oidp = sysctl_find_oidname(oid->oid_name, parent);
428 	if (oidp != NULL)
429 		return (EEXIST);
430 	sysctl_unregister_oid(oid);
431 	oid->oid_parent = parent;
432 	oid->oid_number = OID_AUTO;
433 	sysctl_register_oid(oid);
434 	return (0);
435 }
436 
437 /*
438  * Register the kernel's oids on startup.
439  */
440 SET_DECLARE(sysctl_set, struct sysctl_oid);
441 
442 static void
443 sysctl_register_all(void *arg)
444 {
445 	struct sysctl_oid **oidp;
446 
447 	SYSCTL_INIT();
448 	SET_FOREACH(oidp, sysctl_set)
449 		sysctl_register_oid(*oidp);
450 }
451 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
452 
453 /*
454  * "Staff-functions"
455  *
456  * These functions implement a presently undocumented interface
457  * used by the sysctl program to walk the tree, and get the type
458  * so it can print the value.
459  * This interface is under work and consideration, and should probably
460  * be killed with a big axe by the first person who can find the time.
461  * (be aware though, that the proper interface isn't as obvious as it
462  * may seem, there are various conflicting requirements.
463  *
464  * {0,0}	printf the entire MIB-tree.
465  * {0,1,...}	return the name of the "..." OID.
466  * {0,2,...}	return the next OID.
467  * {0,3}	return the OID of the name in "new"
468  * {0,4,...}	return the kind & format info for the "..." OID.
469  * {0,5,...}	return the description the "..." OID.
470  */
471 
472 #ifdef SYSCTL_DEBUG
473 static void
474 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
475 {
476 	int k;
477 	struct sysctl_oid *oidp;
478 
479 	SLIST_FOREACH(oidp, l, oid_link) {
480 
481 		for (k=0; k<i; k++)
482 			printf(" ");
483 
484 		printf("%d %s ", oidp->oid_number, oidp->oid_name);
485 
486 		printf("%c%c",
487 			oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
488 			oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
489 
490 		if (oidp->oid_handler)
491 			printf(" *Handler");
492 
493 		switch (oidp->oid_kind & CTLTYPE) {
494 			case CTLTYPE_NODE:
495 				printf(" Node\n");
496 				if (!oidp->oid_handler) {
497 					sysctl_sysctl_debug_dump_node(
498 						oidp->oid_arg1, i+2);
499 				}
500 				break;
501 			case CTLTYPE_INT:    printf(" Int\n"); break;
502 			case CTLTYPE_STRING: printf(" String\n"); break;
503 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
504 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
505 			default:	     printf("\n");
506 		}
507 
508 	}
509 }
510 
511 static int
512 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
513 {
514 	int error;
515 
516 	error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
517 	if (error)
518 		return (error);
519 	sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
520 	return (ENOENT);
521 }
522 
523 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
524 	0, 0, sysctl_sysctl_debug, "-", "");
525 #endif
526 
527 static int
528 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
529 {
530 	int *name = (int *) arg1;
531 	u_int namelen = arg2;
532 	int error = 0;
533 	struct sysctl_oid *oid;
534 	struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
535 	char buf[10];
536 
537 	while (namelen) {
538 		if (!lsp) {
539 			snprintf(buf,sizeof(buf),"%d",*name);
540 			if (req->oldidx)
541 				error = SYSCTL_OUT(req, ".", 1);
542 			if (!error)
543 				error = SYSCTL_OUT(req, buf, strlen(buf));
544 			if (error)
545 				return (error);
546 			namelen--;
547 			name++;
548 			continue;
549 		}
550 		lsp2 = 0;
551 		SLIST_FOREACH(oid, lsp, oid_link) {
552 			if (oid->oid_number != *name)
553 				continue;
554 
555 			if (req->oldidx)
556 				error = SYSCTL_OUT(req, ".", 1);
557 			if (!error)
558 				error = SYSCTL_OUT(req, oid->oid_name,
559 					strlen(oid->oid_name));
560 			if (error)
561 				return (error);
562 
563 			namelen--;
564 			name++;
565 
566 			if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
567 				break;
568 
569 			if (oid->oid_handler)
570 				break;
571 
572 			lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
573 			break;
574 		}
575 		lsp = lsp2;
576 	}
577 	return (SYSCTL_OUT(req, "", 1));
578 }
579 
580 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
581 
582 static int
583 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
584 	int *next, int *len, int level, struct sysctl_oid **oidpp)
585 {
586 	struct sysctl_oid *oidp;
587 
588 	*len = level;
589 	SLIST_FOREACH(oidp, lsp, oid_link) {
590 		*next = oidp->oid_number;
591 		*oidpp = oidp;
592 
593 		if (oidp->oid_kind & CTLFLAG_SKIP)
594 			continue;
595 
596 		if (!namelen) {
597 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
598 				return (0);
599 			if (oidp->oid_handler)
600 				/* We really should call the handler here...*/
601 				return (0);
602 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
603 			if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
604 				len, level+1, oidpp))
605 				return (0);
606 			goto emptynode;
607 		}
608 
609 		if (oidp->oid_number < *name)
610 			continue;
611 
612 		if (oidp->oid_number > *name) {
613 			if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
614 				return (0);
615 			if (oidp->oid_handler)
616 				return (0);
617 			lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
618 			if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
619 				next+1, len, level+1, oidpp))
620 				return (0);
621 			goto next;
622 		}
623 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
624 			continue;
625 
626 		if (oidp->oid_handler)
627 			continue;
628 
629 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
630 		if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
631 			len, level+1, oidpp))
632 			return (0);
633 	next:
634 		namelen = 1;
635 	emptynode:
636 		*len = level;
637 	}
638 	return (1);
639 }
640 
641 static int
642 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
643 {
644 	int *name = (int *) arg1;
645 	u_int namelen = arg2;
646 	int i, j, error;
647 	struct sysctl_oid *oid;
648 	struct sysctl_oid_list *lsp = &sysctl__children;
649 	int newoid[CTL_MAXNAME];
650 
651 	i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
652 	if (i)
653 		return (ENOENT);
654 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
655 	return (error);
656 }
657 
658 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
659 
660 static int
661 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
662 {
663 	int i;
664 	struct sysctl_oid *oidp;
665 	struct sysctl_oid_list *lsp = &sysctl__children;
666 	char *p;
667 
668 	if (!*name)
669 		return (ENOENT);
670 
671 	p = name + strlen(name) - 1 ;
672 	if (*p == '.')
673 		*p = '\0';
674 
675 	*len = 0;
676 
677 	for (p = name; *p && *p != '.'; p++)
678 		;
679 	i = *p;
680 	if (i == '.')
681 		*p = '\0';
682 
683 	oidp = SLIST_FIRST(lsp);
684 
685 	while (oidp && *len < CTL_MAXNAME) {
686 		if (strcmp(name, oidp->oid_name)) {
687 			oidp = SLIST_NEXT(oidp, oid_link);
688 			continue;
689 		}
690 		*oid++ = oidp->oid_number;
691 		(*len)++;
692 
693 		if (!i) {
694 			if (oidpp)
695 				*oidpp = oidp;
696 			return (0);
697 		}
698 
699 		if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
700 			break;
701 
702 		if (oidp->oid_handler)
703 			break;
704 
705 		lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
706 		oidp = SLIST_FIRST(lsp);
707 		name = p+1;
708 		for (p = name; *p && *p != '.'; p++)
709 				;
710 		i = *p;
711 		if (i == '.')
712 			*p = '\0';
713 	}
714 	return (ENOENT);
715 }
716 
717 static int
718 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
719 {
720 	char *p;
721 	int error, oid[CTL_MAXNAME], len;
722 	struct sysctl_oid *op = 0;
723 
724 	if (!req->newlen)
725 		return (ENOENT);
726 	if (req->newlen >= MAXPATHLEN)	/* XXX arbitrary, undocumented */
727 		return (ENAMETOOLONG);
728 
729 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
730 
731 	error = SYSCTL_IN(req, p, req->newlen);
732 	if (error) {
733 		free(p, M_SYSCTL);
734 		return (error);
735 	}
736 
737 	p [req->newlen] = '\0';
738 
739 	error = name2oid(p, oid, &len, &op);
740 
741 	free(p, M_SYSCTL);
742 
743 	if (error)
744 		return (error);
745 
746 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
747 	return (error);
748 }
749 
750 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
751 	sysctl_sysctl_name2oid, "I", "");
752 
753 static int
754 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
755 {
756 	struct sysctl_oid *oid;
757 	int error;
758 
759 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
760 	if (error)
761 		return (error);
762 
763 	if (!oid->oid_fmt)
764 		return (ENOENT);
765 	error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
766 	if (error)
767 		return (error);
768 	error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
769 	return (error);
770 }
771 
772 
773 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
774 
775 static int
776 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
777 {
778 	struct sysctl_oid *oid;
779 	int error;
780 
781 	error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
782 	if (error)
783 		return (error);
784 
785 	if (!oid->oid_descr)
786 		return (ENOENT);
787 	error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
788 	return (error);
789 }
790 
791 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
792 
793 /*
794  * Default "handler" functions.
795  */
796 
797 /*
798  * Handle an int, signed or unsigned.
799  * Two cases:
800  *     a variable:  point arg1 at it.
801  *     a constant:  pass it in arg2.
802  */
803 
804 int
805 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
806 {
807 	int tmpout, error = 0;
808 
809 	/*
810 	 * Attempt to get a coherent snapshot by making a copy of the data.
811 	 */
812 	if (arg1)
813 		tmpout = *(int *)arg1;
814 	else
815 		tmpout = arg2;
816 	error = SYSCTL_OUT(req, &tmpout, sizeof(int));
817 
818 	if (error || !req->newptr)
819 		return (error);
820 
821 	if (!arg1)
822 		error = EPERM;
823 	else
824 		error = SYSCTL_IN(req, arg1, sizeof(int));
825 	return (error);
826 }
827 
828 
829 /*
830  * Based on on sysctl_handle_int() convert milliseconds into ticks.
831  */
832 
833 int
834 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
835 {
836 	int error, s, tt;
837 
838 	tt = *(int *)oidp->oid_arg1;
839 	s = (int)((int64_t)tt * 1000 / hz);
840 
841 	error = sysctl_handle_int(oidp, &s, 0, req);
842 	if (error || !req->newptr)
843 		return (error);
844 
845 	tt = (int)((int64_t)s * hz / 1000);
846 	if (tt < 1)
847 		return (EINVAL);
848 
849 	*(int *)oidp->oid_arg1 = tt;
850 	return (0);
851 }
852 
853 
854 /*
855  * Handle a long, signed or unsigned.  arg1 points to it.
856  */
857 
858 int
859 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
860 {
861 	int error = 0;
862 	long tmplong;
863 #ifdef SCTL_MASK32
864 	int tmpint;
865 #endif
866 
867 	/*
868 	 * Attempt to get a coherent snapshot by making a copy of the data.
869 	 */
870 	if (!arg1)
871 		return (EINVAL);
872 	tmplong = *(long *)arg1;
873 #ifdef SCTL_MASK32
874 	if (req->flags & SCTL_MASK32) {
875 		tmpint = tmplong;
876 		error = SYSCTL_OUT(req, &tmpint, sizeof(int));
877 	} else
878 #endif
879 		error = SYSCTL_OUT(req, &tmplong, sizeof(long));
880 
881 	if (error || !req->newptr)
882 		return (error);
883 
884 #ifdef SCTL_MASK32
885 	if (req->flags & SCTL_MASK32) {
886 		error = SYSCTL_IN(req, &tmpint, sizeof(int));
887 		*(long *)arg1 = (long)tmpint;
888 	} else
889 #endif
890 		error = SYSCTL_IN(req, arg1, sizeof(long));
891 	return (error);
892 }
893 
894 /*
895  * Handle a 64 bit int, signed or unsigned.  arg1 points to it.
896  */
897 
898 int
899 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
900 {
901 	int error = 0;
902 	uint64_t tmpout;
903 
904 	/*
905 	 * Attempt to get a coherent snapshot by making a copy of the data.
906 	 */
907 	if (!arg1)
908 		return (EINVAL);
909 	tmpout = *(uint64_t *)arg1;
910 	error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
911 
912 	if (error || !req->newptr)
913 		return (error);
914 
915 	error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
916 	return (error);
917 }
918 
919 /*
920  * Handle our generic '\0' terminated 'C' string.
921  * Two cases:
922  * 	a variable string:  point arg1 at it, arg2 is max length.
923  * 	a constant string:  point arg1 at it, arg2 is zero.
924  */
925 
926 int
927 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
928 {
929 	int error=0;
930 	char *tmparg;
931 	size_t outlen;
932 
933 	/*
934 	 * Attempt to get a coherent snapshot by copying to a
935 	 * temporary kernel buffer.
936 	 */
937 retry:
938 	outlen = strlen((char *)arg1)+1;
939 	tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
940 
941 	if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
942 		free(tmparg, M_SYSCTLTMP);
943 		goto retry;
944 	}
945 
946 	error = SYSCTL_OUT(req, tmparg, outlen);
947 	free(tmparg, M_SYSCTLTMP);
948 
949 	if (error || !req->newptr)
950 		return (error);
951 
952 	if ((req->newlen - req->newidx) >= arg2) {
953 		error = EINVAL;
954 	} else {
955 		arg2 = (req->newlen - req->newidx);
956 		error = SYSCTL_IN(req, arg1, arg2);
957 		((char *)arg1)[arg2] = '\0';
958 	}
959 
960 	return (error);
961 }
962 
963 /*
964  * Handle any kind of opaque data.
965  * arg1 points to it, arg2 is the size.
966  */
967 
968 int
969 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
970 {
971 	int error, tries;
972 	u_int generation;
973 	struct sysctl_req req2;
974 
975 	/*
976 	 * Attempt to get a coherent snapshot, by using the thread
977 	 * pre-emption counter updated from within mi_switch() to
978 	 * determine if we were pre-empted during a bcopy() or
979 	 * copyout(). Make 3 attempts at doing this before giving up.
980 	 * If we encounter an error, stop immediately.
981 	 */
982 	tries = 0;
983 	req2 = *req;
984 retry:
985 	generation = curthread->td_generation;
986 	error = SYSCTL_OUT(req, arg1, arg2);
987 	if (error)
988 		return (error);
989 	tries++;
990 	if (generation != curthread->td_generation && tries < 3) {
991 		*req = req2;
992 		goto retry;
993 	}
994 
995 	error = SYSCTL_IN(req, arg1, arg2);
996 
997 	return (error);
998 }
999 
1000 /*
1001  * Transfer functions to/from kernel space.
1002  * XXX: rather untested at this point
1003  */
1004 static int
1005 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
1006 {
1007 	size_t i = 0;
1008 
1009 	if (req->oldptr) {
1010 		i = l;
1011 		if (req->oldlen <= req->oldidx)
1012 			i = 0;
1013 		else
1014 			if (i > req->oldlen - req->oldidx)
1015 				i = req->oldlen - req->oldidx;
1016 		if (i > 0)
1017 			bcopy(p, (char *)req->oldptr + req->oldidx, i);
1018 	}
1019 	req->oldidx += l;
1020 	if (req->oldptr && i != l)
1021 		return (ENOMEM);
1022 	return (0);
1023 }
1024 
1025 static int
1026 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
1027 {
1028 	if (!req->newptr)
1029 		return (0);
1030 	if (req->newlen - req->newidx < l)
1031 		return (EINVAL);
1032 	bcopy((char *)req->newptr + req->newidx, p, l);
1033 	req->newidx += l;
1034 	return (0);
1035 }
1036 
1037 int
1038 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1039     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
1040 {
1041 	int error = 0;
1042 	struct sysctl_req req;
1043 
1044 	bzero(&req, sizeof req);
1045 
1046 	req.td = td;
1047 	req.flags = flags;
1048 
1049 	if (oldlenp) {
1050 		req.oldlen = *oldlenp;
1051 	}
1052 	req.validlen = req.oldlen;
1053 
1054 	if (old) {
1055 		req.oldptr= old;
1056 	}
1057 
1058 	if (new != NULL) {
1059 		req.newlen = newlen;
1060 		req.newptr = new;
1061 	}
1062 
1063 	req.oldfunc = sysctl_old_kernel;
1064 	req.newfunc = sysctl_new_kernel;
1065 	req.lock = REQ_LOCKED;
1066 
1067 	SYSCTL_LOCK();
1068 
1069 	error = sysctl_root(0, name, namelen, &req);
1070 
1071 	if (req.lock == REQ_WIRED && req.validlen > 0)
1072 		vsunlock(req.oldptr, req.validlen);
1073 
1074 	SYSCTL_UNLOCK();
1075 
1076 	if (error && error != ENOMEM)
1077 		return (error);
1078 
1079 	if (retval) {
1080 		if (req.oldptr && req.oldidx > req.validlen)
1081 			*retval = req.validlen;
1082 		else
1083 			*retval = req.oldidx;
1084 	}
1085 	return (error);
1086 }
1087 
1088 int
1089 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
1090     void *new, size_t newlen, size_t *retval, int flags)
1091 {
1092         int oid[CTL_MAXNAME];
1093         size_t oidlen, plen;
1094 	int error;
1095 
1096 	oid[0] = 0;		/* sysctl internal magic */
1097 	oid[1] = 3;		/* name2oid */
1098 	oidlen = sizeof(oid);
1099 
1100 	error = kernel_sysctl(td, oid, 2, oid, &oidlen,
1101 	    (void *)name, strlen(name), &plen, flags);
1102 	if (error)
1103 		return (error);
1104 
1105 	error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
1106 	    new, newlen, retval, flags);
1107 	return (error);
1108 }
1109 
1110 /*
1111  * Transfer function to/from user space.
1112  */
1113 static int
1114 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1115 {
1116 	int error = 0;
1117 	size_t i, len, origidx;
1118 
1119 	origidx = req->oldidx;
1120 	req->oldidx += l;
1121 	if (req->oldptr == NULL)
1122 		return (0);
1123 	/*
1124 	 * If we have not wired the user supplied buffer and we are currently
1125 	 * holding locks, drop a witness warning, as it's possible that
1126 	 * write operations to the user page can sleep.
1127 	 */
1128 	if (req->lock != REQ_WIRED)
1129 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1130 		    "sysctl_old_user()");
1131 	i = l;
1132 	len = req->validlen;
1133 	if (len <= origidx)
1134 		i = 0;
1135 	else {
1136 		if (i > len - origidx)
1137 			i = len - origidx;
1138 		error = copyout(p, (char *)req->oldptr + origidx, i);
1139 	}
1140 	if (error)
1141 		return (error);
1142 	if (i < l)
1143 		return (ENOMEM);
1144 	return (0);
1145 }
1146 
1147 static int
1148 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1149 {
1150 	int error;
1151 
1152 	if (!req->newptr)
1153 		return (0);
1154 	if (req->newlen - req->newidx < l)
1155 		return (EINVAL);
1156 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1157 	    "sysctl_new_user()");
1158 	error = copyin((char *)req->newptr + req->newidx, p, l);
1159 	req->newidx += l;
1160 	return (error);
1161 }
1162 
1163 /*
1164  * Wire the user space destination buffer.  If set to a value greater than
1165  * zero, the len parameter limits the maximum amount of wired memory.
1166  */
1167 int
1168 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
1169 {
1170 	int ret;
1171 	size_t i, wiredlen;
1172 	char *cp, dummy;
1173 
1174 	wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
1175 	ret = 0;
1176 	if (req->lock == REQ_LOCKED && req->oldptr &&
1177 	    req->oldfunc == sysctl_old_user) {
1178 		if (wiredlen != 0) {
1179 			ret = vslock(req->oldptr, wiredlen);
1180 			if (ret != 0) {
1181 				if (ret != ENOMEM)
1182 					return (ret);
1183 				wiredlen = 0;
1184 			}
1185 			/*
1186 			 * Touch all the wired pages to avoid PTE modified
1187 			 * bit emulation traps on Alpha while holding locks
1188 			 * in the sysctl handler.
1189 			 */
1190 			for (i = (wiredlen + PAGE_SIZE - 1) / PAGE_SIZE,
1191 			    cp = req->oldptr; i > 0; i--, cp += PAGE_SIZE) {
1192 				copyin(cp, &dummy, 1);
1193 				copyout(&dummy, cp, 1);
1194 			}
1195 		}
1196 		req->lock = REQ_WIRED;
1197 		req->validlen = wiredlen;
1198 	}
1199 	return (0);
1200 }
1201 
1202 int
1203 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1204     int *nindx, struct sysctl_req *req)
1205 {
1206 	struct sysctl_oid *oid;
1207 	int indx;
1208 
1209 	oid = SLIST_FIRST(&sysctl__children);
1210 	indx = 0;
1211 	while (oid && indx < CTL_MAXNAME) {
1212 		if (oid->oid_number == name[indx]) {
1213 			indx++;
1214 			if (oid->oid_kind & CTLFLAG_NOLOCK)
1215 				req->lock = REQ_UNLOCKED;
1216 			if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1217 				if (oid->oid_handler != NULL ||
1218 				    indx == namelen) {
1219 					*noid = oid;
1220 					if (nindx != NULL)
1221 						*nindx = indx;
1222 					return (0);
1223 				}
1224 				oid = SLIST_FIRST(
1225 				    (struct sysctl_oid_list *)oid->oid_arg1);
1226 			} else if (indx == namelen) {
1227 				*noid = oid;
1228 				if (nindx != NULL)
1229 					*nindx = indx;
1230 				return (0);
1231 			} else {
1232 				return (ENOTDIR);
1233 			}
1234 		} else {
1235 			oid = SLIST_NEXT(oid, oid_link);
1236 		}
1237 	}
1238 	return (ENOENT);
1239 }
1240 
1241 /*
1242  * Traverse our tree, and find the right node, execute whatever it points
1243  * to, and return the resulting error code.
1244  */
1245 
1246 static int
1247 sysctl_root(SYSCTL_HANDLER_ARGS)
1248 {
1249 	struct sysctl_oid *oid;
1250 	int error, indx, lvl;
1251 
1252 	error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1253 	if (error)
1254 		return (error);
1255 
1256 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1257 		/*
1258 		 * You can't call a sysctl when it's a node, but has
1259 		 * no handler.  Inform the user that it's a node.
1260 		 * The indx may or may not be the same as namelen.
1261 		 */
1262 		if (oid->oid_handler == NULL)
1263 			return (EISDIR);
1264 	}
1265 
1266 	/* Is this sysctl writable? */
1267 	if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
1268 		return (EPERM);
1269 
1270 	KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
1271 
1272 	/* Is this sysctl sensitive to securelevels? */
1273 	if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
1274 		lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
1275 		error = securelevel_gt(req->td->td_ucred, lvl);
1276 		if (error)
1277 			return (error);
1278 	}
1279 
1280 	/* Is this sysctl writable by only privileged users? */
1281 	if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
1282 		if (oid->oid_kind & CTLFLAG_PRISON)
1283 			error = priv_check(req->td, PRIV_SYSCTL_WRITEJAIL);
1284 		else
1285 			error = priv_check(req->td, PRIV_SYSCTL_WRITE);
1286 		if (error)
1287 			return (error);
1288 	}
1289 
1290 	if (!oid->oid_handler)
1291 		return (EINVAL);
1292 
1293 	if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1294 		arg1 = (int *)arg1 + indx;
1295 		arg2 -= indx;
1296 	} else {
1297 		arg1 = oid->oid_arg1;
1298 		arg2 = oid->oid_arg2;
1299 	}
1300 #ifdef MAC
1301 	error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
1302 	    req);
1303 	if (error != 0)
1304 		return (error);
1305 #endif
1306 	error = oid->oid_handler(oid, arg1, arg2, req);
1307 
1308 	return (error);
1309 }
1310 
1311 #ifndef _SYS_SYSPROTO_H_
1312 struct sysctl_args {
1313 	int	*name;
1314 	u_int	namelen;
1315 	void	*old;
1316 	size_t	*oldlenp;
1317 	void	*new;
1318 	size_t	newlen;
1319 };
1320 #endif
1321 int
1322 __sysctl(struct thread *td, struct sysctl_args *uap)
1323 {
1324 	int error, name[CTL_MAXNAME];
1325 	size_t j;
1326 
1327 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1328 		return (EINVAL);
1329 
1330  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1331  	if (error)
1332 		return (error);
1333 
1334 	mtx_lock(&Giant);
1335 
1336 	error = userland_sysctl(td, name, uap->namelen,
1337 		uap->old, uap->oldlenp, 0,
1338 		uap->new, uap->newlen, &j, 0);
1339 	if (error && error != ENOMEM)
1340 		goto done2;
1341 	if (uap->oldlenp) {
1342 		int i = copyout(&j, uap->oldlenp, sizeof(j));
1343 		if (i)
1344 			error = i;
1345 	}
1346 done2:
1347 	mtx_unlock(&Giant);
1348 	return (error);
1349 }
1350 
1351 /*
1352  * This is used from various compatibility syscalls too.  That's why name
1353  * must be in kernel space.
1354  */
1355 int
1356 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
1357     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
1358     int flags)
1359 {
1360 	int error = 0;
1361 	struct sysctl_req req;
1362 
1363 	bzero(&req, sizeof req);
1364 
1365 	req.td = td;
1366 	req.flags = flags;
1367 
1368 	if (oldlenp) {
1369 		if (inkernel) {
1370 			req.oldlen = *oldlenp;
1371 		} else {
1372 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1373 			if (error)
1374 				return (error);
1375 		}
1376 	}
1377 	req.validlen = req.oldlen;
1378 
1379 	if (old) {
1380 		if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1381 			return (EFAULT);
1382 		req.oldptr= old;
1383 	}
1384 
1385 	if (new != NULL) {
1386 		if (!useracc(new, newlen, VM_PROT_READ))
1387 			return (EFAULT);
1388 		req.newlen = newlen;
1389 		req.newptr = new;
1390 	}
1391 
1392 	req.oldfunc = sysctl_old_user;
1393 	req.newfunc = sysctl_new_user;
1394 	req.lock = REQ_LOCKED;
1395 
1396 	SYSCTL_LOCK();
1397 
1398 	do {
1399 		req.oldidx = 0;
1400 		req.newidx = 0;
1401 		error = sysctl_root(0, name, namelen, &req);
1402 	} while (error == EAGAIN);
1403 
1404 	if (req.lock == REQ_WIRED && req.validlen > 0)
1405 		vsunlock(req.oldptr, req.validlen);
1406 
1407 	SYSCTL_UNLOCK();
1408 
1409 	if (error && error != ENOMEM)
1410 		return (error);
1411 
1412 	if (retval) {
1413 		if (req.oldptr && req.oldidx > req.validlen)
1414 			*retval = req.validlen;
1415 		else
1416 			*retval = req.oldidx;
1417 	}
1418 	return (error);
1419 }
1420 
1421 #ifdef COMPAT_43
1422 #include <sys/socket.h>
1423 #include <vm/vm_param.h>
1424 
1425 #define	KINFO_PROC		(0<<8)
1426 #define	KINFO_RT		(1<<8)
1427 #define	KINFO_VNODE		(2<<8)
1428 #define	KINFO_FILE		(3<<8)
1429 #define	KINFO_METER		(4<<8)
1430 #define	KINFO_LOADAVG		(5<<8)
1431 #define	KINFO_CLOCKRATE		(6<<8)
1432 
1433 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
1434 #define	KINFO_BSDI_SYSINFO	(101<<8)
1435 
1436 /*
1437  * XXX this is bloat, but I hope it's better here than on the potentially
1438  * limited kernel stack...  -Peter
1439  */
1440 
1441 static struct {
1442 	int	bsdi_machine;		/* "i386" on BSD/386 */
1443 /*      ^^^ this is an offset to the string, relative to the struct start */
1444 	char	*pad0;
1445 	long	pad1;
1446 	long	pad2;
1447 	long	pad3;
1448 	u_long	pad4;
1449 	u_long	pad5;
1450 	u_long	pad6;
1451 
1452 	int	bsdi_ostype;		/* "BSD/386" on BSD/386 */
1453 	int	bsdi_osrelease;		/* "1.1" on BSD/386 */
1454 	long	pad7;
1455 	long	pad8;
1456 	char	*pad9;
1457 
1458 	long	pad10;
1459 	long	pad11;
1460 	int	pad12;
1461 	long	pad13;
1462 	quad_t	pad14;
1463 	long	pad15;
1464 
1465 	struct	timeval pad16;
1466 	/* we dont set this, because BSDI's uname used gethostname() instead */
1467 	int	bsdi_hostname;		/* hostname on BSD/386 */
1468 
1469 	/* the actual string data is appended here */
1470 
1471 } bsdi_si;
1472 
1473 /*
1474  * this data is appended to the end of the bsdi_si structure during copyout.
1475  * The "char *" offsets are relative to the base of the bsdi_si struct.
1476  * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
1477  * should not exceed the length of the buffer here... (or else!! :-)
1478  */
1479 static char bsdi_strings[80];	/* It had better be less than this! */
1480 
1481 #ifndef _SYS_SYSPROTO_H_
1482 struct getkerninfo_args {
1483 	int	op;
1484 	char	*where;
1485 	size_t	*size;
1486 	int	arg;
1487 };
1488 #endif
1489 int
1490 ogetkerninfo(struct thread *td, struct getkerninfo_args *uap)
1491 {
1492 	int error, name[6];
1493 	size_t size;
1494 	u_int needed = 0;
1495 
1496 	mtx_lock(&Giant);
1497 
1498 	switch (uap->op & 0xff00) {
1499 
1500 	case KINFO_RT:
1501 		name[0] = CTL_NET;
1502 		name[1] = PF_ROUTE;
1503 		name[2] = 0;
1504 		name[3] = (uap->op & 0xff0000) >> 16;
1505 		name[4] = uap->op & 0xff;
1506 		name[5] = uap->arg;
1507 		error = userland_sysctl(td, name, 6, uap->where, uap->size,
1508 			0, 0, 0, &size, 0);
1509 		break;
1510 
1511 	case KINFO_VNODE:
1512 		name[0] = CTL_KERN;
1513 		name[1] = KERN_VNODE;
1514 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1515 			0, 0, 0, &size, 0);
1516 		break;
1517 
1518 	case KINFO_PROC:
1519 		name[0] = CTL_KERN;
1520 		name[1] = KERN_PROC;
1521 		name[2] = uap->op & 0xff;
1522 		name[3] = uap->arg;
1523 		error = userland_sysctl(td, name, 4, uap->where, uap->size,
1524 			0, 0, 0, &size, 0);
1525 		break;
1526 
1527 	case KINFO_FILE:
1528 		name[0] = CTL_KERN;
1529 		name[1] = KERN_FILE;
1530 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1531 			0, 0, 0, &size, 0);
1532 		break;
1533 
1534 	case KINFO_METER:
1535 		name[0] = CTL_VM;
1536 		name[1] = VM_TOTAL;
1537 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1538 			0, 0, 0, &size, 0);
1539 		break;
1540 
1541 	case KINFO_LOADAVG:
1542 		name[0] = CTL_VM;
1543 		name[1] = VM_LOADAVG;
1544 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1545 			0, 0, 0, &size, 0);
1546 		break;
1547 
1548 	case KINFO_CLOCKRATE:
1549 		name[0] = CTL_KERN;
1550 		name[1] = KERN_CLOCKRATE;
1551 		error = userland_sysctl(td, name, 2, uap->where, uap->size,
1552 			0, 0, 0, &size, 0);
1553 		break;
1554 
1555 	case KINFO_BSDI_SYSINFO: {
1556 		/*
1557 		 * this is pretty crude, but it's just enough for uname()
1558 		 * from BSDI's 1.x libc to work.
1559 		 *
1560 		 * *size gives the size of the buffer before the call, and
1561 		 * the amount of data copied after a successful call.
1562 		 * If successful, the return value is the amount of data
1563 		 * available, which can be larger than *size.
1564 		 *
1565 		 * BSDI's 2.x product apparently fails with ENOMEM if *size
1566 		 * is too small.
1567 		 */
1568 
1569 		u_int left;
1570 		char *s;
1571 
1572 		bzero((char *)&bsdi_si, sizeof(bsdi_si));
1573 		bzero(bsdi_strings, sizeof(bsdi_strings));
1574 
1575 		s = bsdi_strings;
1576 
1577 		bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1578 		strcpy(s, ostype);
1579 		s += strlen(s) + 1;
1580 
1581 		bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1582 		strcpy(s, osrelease);
1583 		s += strlen(s) + 1;
1584 
1585 		bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1586 		strcpy(s, machine);
1587 		s += strlen(s) + 1;
1588 
1589 		needed = sizeof(bsdi_si) + (s - bsdi_strings);
1590 
1591 		if ((uap->where == NULL) || (uap->size == NULL)) {
1592 			/* process is asking how much buffer to supply.. */
1593 			size = needed;
1594 			error = 0;
1595 			break;
1596 		}
1597 
1598 		if ((error = copyin(uap->size, &size, sizeof(size))) != 0)
1599 			break;
1600 
1601 		/* if too much buffer supplied, trim it down */
1602 		if (size > needed)
1603 			size = needed;
1604 
1605 		/* how much of the buffer is remaining */
1606 		left = size;
1607 
1608 		if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1609 			break;
1610 
1611 		/* is there any point in continuing? */
1612 		if (left > sizeof(bsdi_si)) {
1613 			left -= sizeof(bsdi_si);
1614 			error = copyout(&bsdi_strings,
1615 					uap->where + sizeof(bsdi_si), left);
1616 		}
1617 		break;
1618 	}
1619 
1620 	default:
1621 		error = EOPNOTSUPP;
1622 		break;
1623 	}
1624 	if (error == 0) {
1625 		td->td_retval[0] = needed ? needed : size;
1626 		if (uap->size) {
1627 			error = copyout(&size, uap->size, sizeof(size));
1628 		}
1629 	}
1630 	mtx_unlock(&Giant);
1631 	return (error);
1632 }
1633 #endif /* COMPAT_43 */
1634