xref: /freebsd/sys/kern/kern_sysctl.c (revision ef5d438ed4bc17ad7ece3e40fe4d1f9baf3aadf7)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Mike Karels at Berkeley Software Design, Inc.
7  *
8  * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
9  * project, to make these variables more userfriendly.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  *
39  *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
40  * $Id: kern_sysctl.c,v 1.58 1995/12/17 21:11:22 phk Exp $
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/sysproto.h>
46 #include <sys/kernel.h>
47 #include <sys/vnode.h>
48 #include <sys/unistd.h>
49 #include <sys/conf.h>
50 #include <sys/sysctl.h>
51 #include <sys/malloc.h>
52 #include <sys/proc.h>
53 
54 #include <vm/vm.h>
55 #include <vm/vm_param.h>
56 #include <vm/vm_extern.h>
57 
58 /*
59  * Locking and stats
60  */
61 static struct sysctl_lock {
62 	int	sl_lock;
63 	int	sl_want;
64 	int	sl_locked;
65 } memlock;
66 
67 static int sysctl_root SYSCTL_HANDLER_ARGS;
68 
69 extern struct linker_set sysctl_;
70 
71 /*
72  * MIB definitions.  XXX Very few of these, if any, belong here.
73  */
74 SYSCTL_NODE(, 0,	  sysctl, CTLFLAG_RW, 0,
75 	"Sysctl internal magic");
76 SYSCTL_NODE(, CTL_KERN,	  kern,   CTLFLAG_RW, 0,
77 	"High kernel, proc, limits &c");
78 SYSCTL_NODE(, CTL_VM,	  vm,     CTLFLAG_RW, 0,
79 	"Virtual memory");
80 SYSCTL_NODE(, CTL_FS,	  fs,     CTLFLAG_RW, 0,
81 	"File system");
82 SYSCTL_NODE(, CTL_NET,	  net,    CTLFLAG_RW, 0,
83 	"Network, (see socket.h)");
84 SYSCTL_NODE(, CTL_DEBUG,  debug,  CTLFLAG_RW, 0,
85 	"Debugging");
86 SYSCTL_NODE(, CTL_HW,	  hw,     CTLFLAG_RW, 0,
87 	"hardware");
88 SYSCTL_NODE(, CTL_MACHDEP, machdep, CTLFLAG_RW, 0,
89 	"machine dependent");
90 SYSCTL_NODE(, CTL_USER,	  user,   CTLFLAG_RW, 0,
91 	"user-level");
92 
93 SYSCTL_STRING(_kern, KERN_OSRELEASE, osrelease, CTLFLAG_RD, osrelease, 0, "");
94 
95 SYSCTL_INT(_kern, KERN_OSREV, osrevision, CTLFLAG_RD, 0, BSD, "");
96 
97 SYSCTL_STRING(_kern, KERN_VERSION, version, CTLFLAG_RD, version, 0, "");
98 
99 SYSCTL_STRING(_kern, KERN_OSTYPE, ostype, CTLFLAG_RD, ostype, 0, "");
100 
101 extern int osreldate;
102 SYSCTL_INT(_kern, KERN_OSRELDATE, osreldate, CTLFLAG_RD, &osreldate, 0, "");
103 
104 SYSCTL_INT(_kern, KERN_MAXPROC, maxproc, CTLFLAG_RD, &maxproc, 0, "");
105 
106 SYSCTL_INT(_kern, KERN_MAXPROCPERUID, maxprocperuid,
107 	CTLFLAG_RD, &maxprocperuid, 0, "");
108 
109 SYSCTL_INT(_kern, KERN_ARGMAX, argmax, CTLFLAG_RD, 0, ARG_MAX, "");
110 
111 SYSCTL_INT(_kern, KERN_POSIX1, posix1version, CTLFLAG_RD, 0, _POSIX_VERSION, "");
112 
113 SYSCTL_INT(_kern, KERN_NGROUPS, ngroups, CTLFLAG_RD, 0, NGROUPS_MAX, "");
114 
115 SYSCTL_INT(_kern, KERN_JOB_CONTROL, job_control, CTLFLAG_RD, 0, 1, "");
116 
117 #ifdef _POSIX_SAVED_IDS
118 SYSCTL_INT(_kern, KERN_SAVED_IDS, saved_ids, CTLFLAG_RD, 0, 1, "");
119 #else
120 SYSCTL_INT(_kern, KERN_SAVED_IDS, saved_ids, CTLFLAG_RD, 0, 0, "");
121 #endif
122 
123 char kernelname[MAXPATHLEN] = "/kernel";	/* XXX bloat */
124 
125 SYSCTL_STRING(_kern, KERN_BOOTFILE, bootfile,
126 	CTLFLAG_RW, kernelname, sizeof kernelname, "");
127 
128 SYSCTL_INT(_hw, HW_NCPU, ncpu, CTLFLAG_RD, 0, 1, "");
129 
130 SYSCTL_INT(_hw, HW_BYTEORDER, byteorder, CTLFLAG_RD, 0, BYTE_ORDER, "");
131 
132 SYSCTL_INT(_hw, HW_PAGESIZE, pagesize, CTLFLAG_RD, 0, PAGE_SIZE, "");
133 
134 char hostname[MAXHOSTNAMELEN];
135 
136 SYSCTL_STRING(_kern, KERN_HOSTNAME, hostname, CTLFLAG_RW,
137 	hostname, sizeof(hostname), "");
138 
139 int securelevel = -1;
140 
141 static int
142 sysctl_kern_securelvl SYSCTL_HANDLER_ARGS
143 {
144 		int error, level;
145 
146 		level = securelevel;
147 		error = sysctl_handle_int(oidp, &level, 0, req);
148 		if (error || !req->newptr)
149 			return (error);
150 		if (level < securelevel && req->p->p_pid != 1)
151 			return (EPERM);
152 		securelevel = level;
153 		return (error);
154 }
155 
156 SYSCTL_PROC(_kern, KERN_SECURELVL, securelevel, CTLTYPE_INT|CTLFLAG_RW,
157 	0, 0, sysctl_kern_securelvl, "I", "");
158 
159 char domainname[MAXHOSTNAMELEN];
160 SYSCTL_STRING(_kern, KERN_DOMAINNAME, domainname, CTLFLAG_RW,
161 	&domainname, sizeof(domainname), "");
162 
163 long hostid;
164 /* Some trouble here, if sizeof (int) != sizeof (long) */
165 SYSCTL_INT(_kern, KERN_HOSTID, hostid, CTLFLAG_RW, &hostid, 0, "");
166 
167 /*
168  * This is really cheating.  These actually live in the libc, something
169  * which I'm not quite sure is a good idea anyway, but in order for
170  * getnext and friends to actually work, we define dummies here.
171  */
172 
173 SYSCTL_STRING(_user, USER_CS_PATH, cs_path, CTLFLAG_RW, "", 0, "");
174 SYSCTL_INT(_user, USER_BC_BASE_MAX, bc_base_max, CTLFLAG_RW, 0, 0, "");
175 SYSCTL_INT(_user, USER_BC_DIM_MAX, bc_dim_max, CTLFLAG_RW, 0, 0, "");
176 SYSCTL_INT(_user, USER_BC_SCALE_MAX, bc_scale_max, CTLFLAG_RW, 0, 0, "");
177 SYSCTL_INT(_user, USER_BC_STRING_MAX, bc_string_max, CTLFLAG_RW, 0, 0, "");
178 SYSCTL_INT(_user, USER_COLL_WEIGHTS_MAX, coll_weights_max, CTLFLAG_RW, 0, 0, "");
179 SYSCTL_INT(_user, USER_EXPR_NEST_MAX, expr_nest_max, CTLFLAG_RW, 0, 0, "");
180 SYSCTL_INT(_user, USER_LINE_MAX, line_max, CTLFLAG_RW, 0, 0, "");
181 SYSCTL_INT(_user, USER_RE_DUP_MAX, re_dup_max, CTLFLAG_RW, 0, 0, "");
182 SYSCTL_INT(_user, USER_POSIX2_VERSION, posix2_version, CTLFLAG_RW, 0, 0, "");
183 SYSCTL_INT(_user, USER_POSIX2_C_BIND, posix2_c_bind, CTLFLAG_RW, 0, 0, "");
184 SYSCTL_INT(_user, USER_POSIX2_C_DEV, posix2_c_dev, CTLFLAG_RW, 0, 0, "");
185 SYSCTL_INT(_user, USER_POSIX2_CHAR_TERM, posix2_char_term, CTLFLAG_RW, 0, 0, "");
186 SYSCTL_INT(_user, USER_POSIX2_FORT_DEV, posix2_fort_dev, CTLFLAG_RW, 0, 0, "");
187 SYSCTL_INT(_user, USER_POSIX2_FORT_RUN, posix2_fort_run, CTLFLAG_RW, 0, 0, "");
188 SYSCTL_INT(_user, USER_POSIX2_LOCALEDEF, posix2_localedef, CTLFLAG_RW, 0, 0, "");
189 SYSCTL_INT(_user, USER_POSIX2_SW_DEV, posix2_sw_dev, CTLFLAG_RW, 0, 0, "");
190 SYSCTL_INT(_user, USER_POSIX2_UPE, posix2_upe, CTLFLAG_RW, 0, 0, "");
191 SYSCTL_INT(_user, USER_STREAM_MAX, stream_max, CTLFLAG_RW, 0, 0, "");
192 SYSCTL_INT(_user, USER_TZNAME_MAX, tzname_max, CTLFLAG_RW, 0, 0, "");
193 
194 
195 /*
196  * End of MIB definitions.
197  */
198 
199 /*
200  * Initialization of the MIB tree.
201  *
202  * Order by number in each linker_set.
203  */
204 
205 static int
206 sysctl_order_cmp(const void *a, const void *b)
207 {
208 	const struct sysctl_oid **pa, **pb;
209 
210 	pa = (const struct sysctl_oid **)a;
211 	pb = (const struct sysctl_oid **)b;
212 	if (*pa == NULL)
213 		return (1);
214 	if (*pb == NULL)
215 		return (-1);
216 	return ((*pa)->oid_number - (*pb)->oid_number);
217 }
218 
219 static void
220 sysctl_order(void *arg)
221 {
222 	int j, k;
223 	struct linker_set *l = (struct linker_set *) arg;
224 	struct sysctl_oid **oidpp;
225 
226 	/* First, find the highest oid we have */
227 	j = l->ls_length;
228 	oidpp = (struct sysctl_oid **) l->ls_items;
229 	for (k = 0; j--; oidpp++)
230 		if (*oidpp && (*oidpp)->oid_number > k)
231 			k = (*oidpp)->oid_number;
232 
233 	/* Next, replace all OID_AUTO oids with new numbers */
234 	j = l->ls_length;
235 	oidpp = (struct sysctl_oid **) l->ls_items;
236 	k += 100;
237 	for (; j--; oidpp++)
238 		if (*oidpp && (*oidpp)->oid_number == OID_AUTO)
239 			(*oidpp)->oid_number = k++;
240 
241 	/* Finally: sort by oid */
242 	j = l->ls_length;
243 	oidpp = (struct sysctl_oid **) l->ls_items;
244 	for (; j--; oidpp++) {
245 		if (!*oidpp)
246 			continue;
247 		if ((*oidpp)->oid_arg1 == arg) {
248 			*oidpp = 0;
249 			continue;
250 		}
251 		if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE)
252 			if (!(*oidpp)->oid_handler)
253 				sysctl_order((*oidpp)->oid_arg1);
254 	}
255 	qsort(l->ls_items, l->ls_length, sizeof l->ls_items[0],
256 		sysctl_order_cmp);
257 }
258 
259 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_order, &sysctl_);
260 
261 /*
262  * "Staff-functions"
263  *
264  * These functions implement a presently undocumented interface
265  * used by the sysctl program to walk the tree, and get the type
266  * so it can print the value.
267  * This interface is under work and consideration, and should probably
268  * be killed with a big axe by the first person who can find the time.
269  * (be aware though, that the proper interface isn't as obvious as it
270  * may seem, there are various conflicting requirements.
271  *
272  * {0,0}	printf the entire MIB-tree.
273  * {0,1,...}	return the name of the "..." OID.
274  * {0,2,...}	return the next OID.
275  * {0,3}	return the OID of the name in "new"
276  * {0,4,...}	return the kind & format info for the "..." OID.
277  */
278 
279 static void
280 sysctl_sysctl_debug_dump_node(struct linker_set *l, int i)
281 {
282 	int j, k;
283 	struct sysctl_oid **oidpp;
284 
285 	j = l->ls_length;
286 	oidpp = (struct sysctl_oid **) l->ls_items;
287 	for (; j--; oidpp++) {
288 
289 		if (!*oidpp)
290 			continue;
291 
292 		for (k=0; k<i; k++)
293 			printf(" ");
294 
295 		if ((*oidpp)->oid_number > 100) {
296 			printf("Junk! %p  # %d  %s  k %x  a1 %p  a2 %x  h %p\n",
297 				*oidpp,
298 		 		(*oidpp)->oid_number, (*oidpp)->oid_name,
299 		 		(*oidpp)->oid_kind, (*oidpp)->oid_arg1,
300 		 		(*oidpp)->oid_arg2, (*oidpp)->oid_handler);
301 			continue;
302 		}
303 		printf("%d %s ", (*oidpp)->oid_number, (*oidpp)->oid_name);
304 
305 		printf("%c%c",
306 			(*oidpp)->oid_kind & CTLFLAG_RD ? 'R':' ',
307 			(*oidpp)->oid_kind & CTLFLAG_WR ? 'W':' ');
308 
309 		switch ((*oidpp)->oid_kind & CTLTYPE) {
310 			case CTLTYPE_NODE:
311 				if ((*oidpp)->oid_handler) {
312 					printf(" Node(proc)\n");
313 				} else {
314 					printf(" Node\n");
315 					sysctl_sysctl_debug_dump_node(
316 						(*oidpp)->oid_arg1, i+2);
317 				}
318 				break;
319 			case CTLTYPE_INT:    printf(" Int\n"); break;
320 			case CTLTYPE_STRING: printf(" String\n"); break;
321 			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
322 			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
323 			default:	     printf("\n");
324 		}
325 
326 	}
327 }
328 
329 static int
330 sysctl_sysctl_debug SYSCTL_HANDLER_ARGS
331 {
332 	sysctl_sysctl_debug_dump_node(&sysctl_, 0);
333 	return ENOENT;
334 }
335 
336 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
337 	0, 0, sysctl_sysctl_debug, "-", "");
338 
339 static int
340 sysctl_sysctl_name SYSCTL_HANDLER_ARGS
341 {
342 	int *name = (int *) arg1;
343 	u_int namelen = arg2;
344 	int i, j, error = 0;
345 	struct sysctl_oid **oidpp;
346 	struct linker_set *lsp = &sysctl_;
347 	char buf[10];
348 
349 	while (namelen) {
350 		if (!lsp) {
351 			sprintf(buf,"%d",*name);
352 			if (req->oldidx)
353 				error = SYSCTL_OUT(req, ".", 1);
354 			if (!error)
355 				error = SYSCTL_OUT(req, buf, strlen(buf));
356 			if (error)
357 				return (error);
358 			namelen--;
359 			name++;
360 			continue;
361 		}
362 		oidpp = (struct sysctl_oid **) lsp->ls_items;
363 		j = lsp->ls_length;
364 		lsp = 0;
365 		for (i = 0; i < j; i++, oidpp++) {
366 			if (*oidpp && ((*oidpp)->oid_number != *name))
367 				continue;
368 
369 			if (req->oldidx)
370 				error = SYSCTL_OUT(req, ".", 1);
371 			if (!error)
372 				error = SYSCTL_OUT(req, (*oidpp)->oid_name,
373 					strlen((*oidpp)->oid_name));
374 			if (error)
375 				return (error);
376 
377 			namelen--;
378 			name++;
379 
380 			if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
381 				break;
382 
383 			if ((*oidpp)->oid_handler)
384 				break;
385 
386 			lsp = (struct linker_set*)(*oidpp)->oid_arg1;
387 			break;
388 		}
389 	}
390 	return (SYSCTL_OUT(req, "", 1));
391 }
392 
393 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
394 
395 static int
396 sysctl_sysctl_next_ls (struct linker_set *lsp, int *name, u_int namelen,
397 	int *next, int *len, int level, struct sysctl_oid **oidp)
398 {
399 	int i, j;
400 	struct sysctl_oid **oidpp;
401 
402 	oidpp = (struct sysctl_oid **) lsp->ls_items;
403 	j = lsp->ls_length;
404 	*len = level;
405 	for (i = 0; i < j; i++, oidpp++) {
406 		if (!*oidpp)
407 			continue;
408 
409 		*next = (*oidpp)->oid_number;
410 		*oidp = *oidpp;
411 
412 		if (!namelen) {
413 			if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
414 				return 0;
415 			if ((*oidpp)->oid_handler)
416 				/* We really should call the handler here...*/
417 				return 0;
418 			lsp = (struct linker_set*)(*oidpp)->oid_arg1;
419 			return (sysctl_sysctl_next_ls (lsp, 0, 0, next+1,
420 				len, level+1, oidp));
421 		}
422 
423 		if ((*oidpp)->oid_number < *name)
424 			continue;
425 
426 		if ((*oidpp)->oid_number > *name) {
427 			if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
428 				return 0;
429 			if ((*oidpp)->oid_handler)
430 				return 0;
431 			lsp = (struct linker_set*)(*oidpp)->oid_arg1;
432 			if (!sysctl_sysctl_next_ls (lsp, name+1, namelen-1,
433 				next+1, len, level+1, oidp))
434 				return (0);
435 			namelen = 1;
436 			*len = level;
437 			continue;
438 		}
439 		if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
440 			continue;
441 
442 		if ((*oidpp)->oid_handler)
443 			continue;
444 
445 		lsp = (struct linker_set*)(*oidpp)->oid_arg1;
446 		if (!sysctl_sysctl_next_ls (lsp, name+1, namelen-1, next+1,
447 			len, level+1, oidp))
448 			return (0);
449 		namelen = 1;
450 		*len = level;
451 	}
452 	return 1;
453 }
454 
455 static int
456 sysctl_sysctl_next SYSCTL_HANDLER_ARGS
457 {
458 	int *name = (int *) arg1;
459 	u_int namelen = arg2;
460 	int i, j, error;
461 	struct sysctl_oid *oid;
462 	struct linker_set *lsp = &sysctl_;
463 	int newoid[CTL_MAXNAME];
464 
465 	i = sysctl_sysctl_next_ls (lsp, name, namelen, newoid, &j, 1, &oid);
466 	if (i)
467 		return ENOENT;
468 	error = SYSCTL_OUT(req, newoid, j * sizeof (int));
469 	return (error);
470 }
471 
472 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
473 
474 static int
475 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidp)
476 {
477 	int i, j;
478 	struct sysctl_oid **oidpp;
479 	struct linker_set *lsp = &sysctl_;
480 	char *p;
481 
482 	if (!*name)
483 		return ENOENT;
484 
485 	p = name + strlen(name) - 1 ;
486 	if (*p == '.')
487 		*p = '\0';
488 
489 	*len = 0;
490 
491 	for (p = name; *p && *p != '.'; p++)
492 		;
493 	i = *p;
494 	if (i == '.')
495 		*p = '\0';
496 
497 	j = lsp->ls_length;
498 	oidpp = (struct sysctl_oid **) lsp->ls_items;
499 
500 	while (j-- && *len < CTL_MAXNAME) {
501 		if (!*oidpp)
502 			continue;
503 		if (strcmp(name, (*oidpp)->oid_name)) {
504 			oidpp++;
505 			continue;
506 		}
507 		*oid++ = (*oidpp)->oid_number;
508 		(*len)++;
509 
510 		if (!i) {
511 			if (oidp)
512 				*oidp = *oidpp;
513 			return (0);
514 		}
515 
516 		if (((*oidpp)->oid_kind & CTLTYPE) != CTLTYPE_NODE)
517 			break;
518 
519 		if ((*oidpp)->oid_handler)
520 			break;
521 
522 		lsp = (struct linker_set*)(*oidpp)->oid_arg1;
523 		j = lsp->ls_length;
524 		oidpp = (struct sysctl_oid **)lsp->ls_items;
525 		name = p+1;
526 		for (p = name; *p && *p != '.'; p++)
527 				;
528 		i = *p;
529 		if (i == '.')
530 			*p = '\0';
531 	}
532 	return ENOENT;
533 }
534 
535 static int
536 sysctl_sysctl_name2oid SYSCTL_HANDLER_ARGS
537 {
538 	char *p;
539 	int error, oid[CTL_MAXNAME], len;
540 	struct sysctl_oid *op = 0;
541 
542 	if (!req->newlen)
543 		return ENOENT;
544 
545 	p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
546 
547 	error = SYSCTL_IN(req, p, req->newlen);
548 	if (error) {
549 		free(p, M_SYSCTL);
550 		return (error);
551 	}
552 
553 	p [req->newlen] = '\0';
554 
555 	error = name2oid(p, oid, &len, &op);
556 
557 	free(p, M_SYSCTL);
558 
559 	if (error)
560 		return (error);
561 
562 	error = SYSCTL_OUT(req, oid, len * sizeof *oid);
563 	return (error);
564 }
565 
566 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
567 	sysctl_sysctl_name2oid, "I", "");
568 
569 static int
570 sysctl_sysctl_oidfmt SYSCTL_HANDLER_ARGS
571 {
572 	int *name = (int *) arg1, error;
573 	u_int namelen = arg2;
574 	int indx, j;
575 	struct sysctl_oid **oidpp;
576 	struct linker_set *lsp = &sysctl_;
577 
578 	j = lsp->ls_length;
579 	oidpp = (struct sysctl_oid **) lsp->ls_items;
580 
581 	indx = 0;
582 	while (j-- && indx < CTL_MAXNAME) {
583 		if (*oidpp && ((*oidpp)->oid_number == name[indx])) {
584 			indx++;
585 			if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
586 				if ((*oidpp)->oid_handler)
587 					goto found;
588 				if (indx == namelen)
589 					goto found;
590 				lsp = (struct linker_set*)(*oidpp)->oid_arg1;
591 				j = lsp->ls_length;
592 				oidpp = (struct sysctl_oid **)lsp->ls_items;
593 			} else {
594 				if (indx != namelen)
595 					return EISDIR;
596 				goto found;
597 			}
598 		} else {
599 			oidpp++;
600 		}
601 	}
602 	return ENOENT;
603 found:
604 	if (!(*oidpp)->oid_fmt)
605 		return ENOENT;
606 	error = SYSCTL_OUT(req,
607 		&(*oidpp)->oid_kind, sizeof((*oidpp)->oid_kind));
608 	if (!error)
609 		error = SYSCTL_OUT(req, (*oidpp)->oid_fmt,
610 			strlen((*oidpp)->oid_fmt)+1);
611 	return (error);
612 }
613 
614 
615 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
616 
617 /*
618  * Default "handler" functions.
619  */
620 
621 /*
622  * Handle an integer, signed or unsigned.
623  * Two cases:
624  *     a variable:  point arg1 at it.
625  *     a constant:  pass it in arg2.
626  */
627 
628 int
629 sysctl_handle_int SYSCTL_HANDLER_ARGS
630 {
631 	int error = 0;
632 
633 	if (arg1)
634 		error = SYSCTL_OUT(req, arg1, sizeof(int));
635 	else if (arg2)
636 		error = SYSCTL_OUT(req, &arg2, sizeof(int));
637 
638 	if (error || !req->newptr)
639 		return (error);
640 
641 	if (!arg1)
642 		error = EPERM;
643 	else
644 		error = SYSCTL_IN(req, arg1, sizeof(int));
645 	return (error);
646 }
647 
648 /*
649  * Handle our generic '\0' terminated 'C' string.
650  * Two cases:
651  * 	a variable string:  point arg1 at it, arg2 is max length.
652  * 	a constant string:  point arg1 at it, arg2 is zero.
653  */
654 
655 int
656 sysctl_handle_string SYSCTL_HANDLER_ARGS
657 {
658 	int error=0;
659 
660 	error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
661 
662 	if (error || !req->newptr || !arg2)
663 		return (error);
664 
665 	if ((req->newlen - req->newidx) > arg2) {
666 		error = E2BIG;
667 	} else {
668 		arg2 = (req->newlen - req->newidx);
669 		error = SYSCTL_IN(req, arg1, arg2);
670 		((char *)arg1)[arg2] = '\0';
671 	}
672 
673 	return (error);
674 }
675 
676 /*
677  * Handle any kind of opaque data.
678  * arg1 points to it, arg2 is the size.
679  */
680 
681 int
682 sysctl_handle_opaque SYSCTL_HANDLER_ARGS
683 {
684 	int error;
685 
686 	error = SYSCTL_OUT(req, arg1, arg2);
687 
688 	if (error || !req->newptr)
689 		return (error);
690 
691 	error = SYSCTL_IN(req, arg1, arg2);
692 
693 	return (error);
694 }
695 
696 /*
697  * Transfer functions to/from kernel space.
698  * XXX: rather untested at this point
699  */
700 static int
701 sysctl_old_kernel(struct sysctl_req *req, const void *p, int l)
702 {
703 	int i = 0;
704 
705 	if (req->oldptr) {
706 		i = min(req->oldlen - req->oldidx, l);
707 		if (i > 0)
708 			bcopy(p, req->oldptr + req->oldidx, i);
709 	}
710 	req->oldidx += l;
711 	if (i != l)
712 		return (ENOMEM);
713 	return (0);
714 
715 }
716 
717 static int
718 sysctl_new_kernel(struct sysctl_req *req, const void *p, int l)
719 {
720 	if (!req->newptr)
721 		return 0;
722 	if (req->newlen - req->newidx < l)
723 		return (EINVAL);
724 	bcopy(p, req->newptr + req->newidx, l);
725 	req->newidx += l;
726 	return (0);
727 }
728 
729 /*
730  * Transfer function to/from user space.
731  */
732 static int
733 sysctl_old_user(struct sysctl_req *req, const void *p, int l)
734 {
735 	int error = 0, i = 0;
736 
737 	if (req->lock == 1 && req->oldptr) {
738 		vslock(req->oldptr, req->oldlen);
739 		req->lock = 2;
740 	}
741 	if (req->oldptr) {
742 		i = min(req->oldlen - req->oldidx, l);
743 		if (i > 0)
744 			error  = copyout(p, req->oldptr + req->oldidx, i);
745 	}
746 	req->oldidx += l;
747 	if (error)
748 		return (error);
749 	if (req->oldptr && i < l)
750 		return (ENOMEM);
751 	return (0);
752 }
753 
754 static int
755 sysctl_new_user(struct sysctl_req *req, void *p, int l)
756 {
757 	int error;
758 
759 	if (!req->newptr)
760 		return 0;
761 	if (req->newlen - req->newidx < l)
762 		return (EINVAL);
763 	error = copyin(req->newptr + req->newidx, p, l);
764 	req->newidx += l;
765 	return (error);
766 }
767 
768 /*
769  * Traverse our tree, and find the right node, execute whatever it points
770  * at, and return the resulting error code.
771  */
772 
773 int
774 sysctl_root SYSCTL_HANDLER_ARGS
775 {
776 	int *name = (int *) arg1;
777 	u_int namelen = arg2;
778 	int indx, i, j;
779 	struct sysctl_oid **oidpp;
780 	struct linker_set *lsp = &sysctl_;
781 
782 	j = lsp->ls_length;
783 	oidpp = (struct sysctl_oid **) lsp->ls_items;
784 
785 	indx = 0;
786 	while (j-- && indx < CTL_MAXNAME) {
787 		if (*oidpp && ((*oidpp)->oid_number == name[indx])) {
788 			indx++;
789 			if ((*oidpp)->oid_kind & CTLFLAG_NOLOCK)
790 				req->lock = 0;
791 			if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
792 				if ((*oidpp)->oid_handler)
793 					goto found;
794 				if (indx == namelen)
795 					return ENOENT;
796 				lsp = (struct linker_set*)(*oidpp)->oid_arg1;
797 				j = lsp->ls_length;
798 				oidpp = (struct sysctl_oid **)lsp->ls_items;
799 			} else {
800 				if (indx != namelen)
801 					return EISDIR;
802 				goto found;
803 			}
804 		} else {
805 			oidpp++;
806 		}
807 	}
808 	return ENOENT;
809 found:
810 	/* If writing isn't allowed */
811 	if (req->newptr && !((*oidpp)->oid_kind & CTLFLAG_WR))
812 		return (EPERM);
813 
814 	/* Most likely only root can write */
815 	if (!((*oidpp)->oid_kind & CTLFLAG_ANYBODY) &&
816 	    req->newptr && req->p &&
817 	    (i = suser(req->p->p_ucred, &req->p->p_acflag)))
818 		return (i);
819 
820 	if (!(*oidpp)->oid_handler)
821 		return EINVAL;
822 
823 	if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
824 		i = ((*oidpp)->oid_handler) (*oidpp,
825 					name + indx, namelen - indx,
826 					req);
827 	} else {
828 		i = ((*oidpp)->oid_handler) (*oidpp,
829 					(*oidpp)->oid_arg1, (*oidpp)->oid_arg2,
830 					req);
831 	}
832 	return (i);
833 }
834 
835 #ifndef _SYS_SYSPROTO_H_
836 struct sysctl_args {
837 	int	*name;
838 	u_int	namelen;
839 	void	*old;
840 	size_t	*oldlenp;
841 	void	*new;
842 	size_t	newlen;
843 };
844 #endif
845 
846 int
847 __sysctl(struct proc *p, struct sysctl_args *uap, int *retval)
848 {
849 	int error, i, j, name[CTL_MAXNAME];
850 
851 	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
852 		return (EINVAL);
853 
854  	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
855  	if (error)
856 		return (error);
857 
858 	error = userland_sysctl(p, name, uap->namelen,
859 		uap->old, uap->oldlenp, 0,
860 		uap->new, uap->newlen, &j);
861 	if (error && error != ENOMEM)
862 		return (error);
863 	if (uap->oldlenp) {
864 		i = copyout(&j, uap->oldlenp, sizeof(j));
865 		if (i)
866 			return (i);
867 	}
868 	return (error);
869 }
870 
871 /*
872  * This is used from various compatibility syscalls too.  That's why name
873  * must be in kernel space.
874  */
875 int
876 userland_sysctl(struct proc *p, int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, int *retval)
877 {
878 	int error = 0;
879 	struct sysctl_req req;
880 
881 	bzero(&req, sizeof req);
882 
883 	req.p = p;
884 
885 	if (oldlenp) {
886 		if (inkernel) {
887 			req.oldlen = *oldlenp;
888 		} else {
889 			error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
890 			if (error)
891 				return (error);
892 		}
893 	}
894 
895 	if (old) {
896 		if (!useracc(old, req.oldlen, B_WRITE))
897 			return (EFAULT);
898 		req.oldptr= old;
899 	}
900 
901 	if (newlen) {
902 		if (!useracc(new, req.newlen, B_READ))
903 			return (EFAULT);
904 		req.newlen = newlen;
905 		req.newptr = new;
906 	}
907 
908 	req.oldfunc = sysctl_old_user;
909 	req.newfunc = sysctl_new_user;
910 	req.lock = 1;
911 
912 	/* XXX this should probably be done in a general way */
913 	while (memlock.sl_lock) {
914 		memlock.sl_want = 1;
915 		(void) tsleep((caddr_t)&memlock, PRIBIO+1, "sysctl", 0);
916 		memlock.sl_locked++;
917 	}
918 	memlock.sl_lock = 1;
919 
920 	error = sysctl_root(0, name, namelen, &req);
921 
922 	if (req.lock == 2)
923 		vsunlock(req.oldptr, req.oldlen, B_WRITE);
924 
925 	memlock.sl_lock = 0;
926 
927 	if (memlock.sl_want) {
928 		memlock.sl_want = 0;
929 		wakeup((caddr_t)&memlock);
930 	}
931 
932 	if (error && error != ENOMEM)
933 		return (error);
934 
935 	if (retval) {
936 		if (req.oldptr && req.oldidx > req.oldlen)
937 			*retval = req.oldlen;
938 		else
939 			*retval = req.oldidx;
940 	}
941 	return (error);
942 }
943 
944 #ifdef COMPAT_43
945 #include <sys/socket.h>
946 #define	KINFO_PROC		(0<<8)
947 #define	KINFO_RT		(1<<8)
948 #define	KINFO_VNODE		(2<<8)
949 #define	KINFO_FILE		(3<<8)
950 #define	KINFO_METER		(4<<8)
951 #define	KINFO_LOADAVG		(5<<8)
952 #define	KINFO_CLOCKRATE		(6<<8)
953 
954 /* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
955 #define	KINFO_BSDI_SYSINFO	(101<<8)
956 
957 /*
958  * XXX this is bloat, but I hope it's better here than on the potentially
959  * limited kernel stack...  -Peter
960  */
961 
962 static struct {
963 	int	bsdi_machine;		/* "i386" on BSD/386 */
964 /*      ^^^ this is an offset to the string, relative to the struct start */
965 	char	*pad0;
966 	long	pad1;
967 	long	pad2;
968 	long	pad3;
969 	u_long	pad4;
970 	u_long	pad5;
971 	u_long	pad6;
972 
973 	int	bsdi_ostype;		/* "BSD/386" on BSD/386 */
974 	int	bsdi_osrelease;		/* "1.1" on BSD/386 */
975 	long	pad7;
976 	long	pad8;
977 	char	*pad9;
978 
979 	long	pad10;
980 	long	pad11;
981 	int	pad12;
982 	long	pad13;
983 	quad_t	pad14;
984 	long	pad15;
985 
986 	struct	timeval pad16;
987 	/* we dont set this, because BSDI's uname used gethostname() instead */
988 	int	bsdi_hostname;		/* hostname on BSD/386 */
989 
990 	/* the actual string data is appended here */
991 
992 } bsdi_si;
993 /*
994  * this data is appended to the end of the bsdi_si structure during copyout.
995  * The "char *" offsets are relative to the base of the bsdi_si struct.
996  * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
997  * should not exceed the length of the buffer here... (or else!! :-)
998  */
999 static char bsdi_strings[80];	/* It had better be less than this! */
1000 
1001 #ifndef _SYS_SYSPROTO_H_
1002 struct getkerninfo_args {
1003 	int	op;
1004 	char	*where;
1005 	int	*size;
1006 	int	arg;
1007 };
1008 #endif
1009 
1010 int
1011 ogetkerninfo(struct proc *p, struct getkerninfo_args *uap, int *retval)
1012 {
1013 	int error, name[6];
1014 	u_int size;
1015 
1016 	switch (uap->op & 0xff00) {
1017 
1018 	case KINFO_RT:
1019 		name[0] = CTL_NET;
1020 		name[1] = PF_ROUTE;
1021 		name[2] = 0;
1022 		name[3] = (uap->op & 0xff0000) >> 16;
1023 		name[4] = uap->op & 0xff;
1024 		name[5] = uap->arg;
1025 		error = userland_sysctl(p, name, 6, uap->where, uap->size,
1026 			0, 0, 0, &size);
1027 		break;
1028 
1029 	case KINFO_VNODE:
1030 		name[0] = CTL_KERN;
1031 		name[1] = KERN_VNODE;
1032 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1033 			0, 0, 0, &size);
1034 		break;
1035 
1036 	case KINFO_PROC:
1037 		name[0] = CTL_KERN;
1038 		name[1] = KERN_PROC;
1039 		name[2] = uap->op & 0xff;
1040 		name[3] = uap->arg;
1041 		error = userland_sysctl(p, name, 4, uap->where, uap->size,
1042 			0, 0, 0, &size);
1043 		break;
1044 
1045 	case KINFO_FILE:
1046 		name[0] = CTL_KERN;
1047 		name[1] = KERN_FILE;
1048 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1049 			0, 0, 0, &size);
1050 		break;
1051 
1052 	case KINFO_METER:
1053 		name[0] = CTL_VM;
1054 		name[1] = VM_METER;
1055 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1056 			0, 0, 0, &size);
1057 		break;
1058 
1059 	case KINFO_LOADAVG:
1060 		name[0] = CTL_VM;
1061 		name[1] = VM_LOADAVG;
1062 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1063 			0, 0, 0, &size);
1064 		break;
1065 
1066 	case KINFO_CLOCKRATE:
1067 		name[0] = CTL_KERN;
1068 		name[1] = KERN_CLOCKRATE;
1069 		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1070 			0, 0, 0, &size);
1071 		break;
1072 
1073 	case KINFO_BSDI_SYSINFO: {
1074 		/*
1075 		 * this is pretty crude, but it's just enough for uname()
1076 		 * from BSDI's 1.x libc to work.
1077 		 *
1078 		 * In particular, it doesn't return the same results when
1079 		 * the supplied buffer is too small.  BSDI's version apparently
1080 		 * will return the amount copied, and set the *size to how
1081 		 * much was needed.  The emulation framework here isn't capable
1082 		 * of that, so we just set both to the amount copied.
1083 		 * BSDI's 2.x product apparently fails with ENOMEM in this
1084 		 * scenario.
1085 		 */
1086 
1087 		u_int needed;
1088 		u_int left;
1089 		char *s;
1090 
1091 		bzero((char *)&bsdi_si, sizeof(bsdi_si));
1092 		bzero(bsdi_strings, sizeof(bsdi_strings));
1093 
1094 		s = bsdi_strings;
1095 
1096 		bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1097 		strcpy(s, ostype);
1098 		s += strlen(s) + 1;
1099 
1100 		bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1101 		strcpy(s, osrelease);
1102 		s += strlen(s) + 1;
1103 
1104 		bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1105 		strcpy(s, machine);
1106 		s += strlen(s) + 1;
1107 
1108 		needed = sizeof(bsdi_si) + (s - bsdi_strings);
1109 
1110 		if (uap->where == NULL) {
1111 			/* process is asking how much buffer to supply.. */
1112 			size = needed;
1113 			error = 0;
1114 			break;
1115 		}
1116 
1117 
1118 		/* if too much buffer supplied, trim it down */
1119 		if (size > needed)
1120 			size = needed;
1121 
1122 		/* how much of the buffer is remaining */
1123 		left = size;
1124 
1125 		if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1126 			break;
1127 
1128 		/* is there any point in continuing? */
1129 		if (left > sizeof(bsdi_si)) {
1130 			left -= sizeof(bsdi_si);
1131 			error = copyout(&bsdi_strings,
1132 					uap->where + sizeof(bsdi_si), left);
1133 		}
1134 		break;
1135 	}
1136 
1137 	default:
1138 		return (EOPNOTSUPP);
1139 	}
1140 	if (error)
1141 		return (error);
1142 	*retval = size;
1143 	if (uap->size)
1144 		error = copyout((caddr_t)&size, (caddr_t)uap->size,
1145 		    sizeof(size));
1146 	return (error);
1147 }
1148 #endif /* COMPAT_43 */
1149