kern_sysctl.c revision 12243 
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 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 *    must display the following acknowledgement:
18 *	This product includes software developed by the University of
19 *	California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 *	@(#)kern_sysctl.c	8.4 (Berkeley) 4/14/94
37 * $Id: kern_sysctl.c,v 1.37 1995/11/12 06:43:01 bde Exp $
38 */
39
40/*
41 * sysctl system call.
42 */
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/sysproto.h>
47#include <sys/kernel.h>
48#include <sys/malloc.h>
49#include <sys/proc.h>
50#include <sys/file.h>
51#include <sys/vnode.h>
52#include <sys/unistd.h>
53#include <sys/buf.h>
54#include <sys/ioctl.h>
55#include <sys/tty.h>
56#include <sys/conf.h>
57#include <vm/vm.h>
58#include <sys/sysctl.h>
59#include <sys/user.h>
60
61extern struct linker_set sysctl_;
62
63/* BEGIN_MIB */
64SYSCTL_NODE(, 0,	  sysctl, CTLFLAG_RW, 0,
65	"Sysctl internal magic");
66SYSCTL_NODE(, CTL_KERN,	  kern,   CTLFLAG_RW, 0,
67	"High kernel, proc, limits &c");
68SYSCTL_NODE(, CTL_VM,	  vm,     CTLFLAG_RW, 0,
69	"Virtual memory");
70SYSCTL_NODE(, CTL_FS,	  fs,     CTLFLAG_RW, 0,
71	"File system");
72SYSCTL_NODE(, CTL_NET,	  net,    CTLFLAG_RW, 0,
73	"Network, (see socket.h)");
74SYSCTL_NODE(, CTL_DEBUG,  debug,  CTLFLAG_RW, 0,
75	"Debugging");
76SYSCTL_NODE(, CTL_HW,	  hw,     CTLFLAG_RW, 0,
77	"hardware");
78SYSCTL_NODE(, CTL_MACHDEP, machdep, CTLFLAG_RW, 0,
79	"machine dependent");
80SYSCTL_NODE(, CTL_USER,	  user,   CTLFLAG_RW, 0,
81	"user-level");
82
83SYSCTL_STRING(_kern, KERN_OSRELEASE, osrelease, CTLFLAG_RD, osrelease, 0, "");
84
85SYSCTL_INT(_kern, KERN_OSREV, osrevision, CTLFLAG_RD, 0, BSD, "");
86
87SYSCTL_STRING(_kern, KERN_VERSION, version, CTLFLAG_RD, version, 0, "");
88
89SYSCTL_STRING(_kern, KERN_OSTYPE, ostype, CTLFLAG_RD, ostype, 0, "");
90
91extern int osreldate;
92SYSCTL_INT(_kern, KERN_OSRELDATE, osreldate, CTLFLAG_RD, &osreldate, 0, "");
93
94SYSCTL_INT(_kern, KERN_MAXVNODES, maxvnodes, CTLFLAG_RD, &desiredvnodes, 0, "");
95
96SYSCTL_INT(_kern, KERN_MAXPROC, maxproc, CTLFLAG_RD, &maxproc, 0, "");
97
98SYSCTL_INT(_kern, KERN_MAXPROCPERUID, maxprocperuid,
99	CTLFLAG_RD, &maxprocperuid, 0, "");
100
101SYSCTL_INT(_kern, KERN_MAXFILESPERPROC, maxfilesperproc,
102	CTLFLAG_RD, &maxfilesperproc, 0, "");
103
104SYSCTL_INT(_kern, KERN_ARGMAX, argmax, CTLFLAG_RD, 0, ARG_MAX, "");
105
106SYSCTL_INT(_kern, KERN_POSIX1, posix1version, CTLFLAG_RD, 0, _POSIX_VERSION, "");
107
108SYSCTL_INT(_kern, KERN_NGROUPS, ngroups, CTLFLAG_RD, 0, NGROUPS_MAX, "");
109
110SYSCTL_INT(_kern, KERN_JOB_CONTROL, job_control, CTLFLAG_RD, 0, 1, "");
111
112SYSCTL_INT(_kern, KERN_MAXFILES, maxfiles, CTLFLAG_RW, &maxfiles, 0, "");
113
114#ifdef _POSIX_SAVED_IDS
115SYSCTL_INT(_kern, KERN_SAVED_IDS, saved_ids, CTLFLAG_RD, 0, 1, "");
116#else
117SYSCTL_INT(_kern, KERN_SAVED_IDS, saved_ids, CTLFLAG_RD, 0, 0, "");
118#endif
119
120char kernelname[MAXPATHLEN] = "/kernel";	/* XXX bloat */
121
122SYSCTL_STRING(_kern, KERN_BOOTFILE, bootfile,
123	CTLFLAG_RW, kernelname, sizeof kernelname, "");
124
125SYSCTL_STRUCT(_kern, KERN_BOOTTIME, boottime,
126	CTLFLAG_RW, &boottime, timeval, "");
127
128SYSCTL_STRING(_hw, HW_MACHINE, machine, CTLFLAG_RD, machine, 0, "");
129
130SYSCTL_STRING(_hw, HW_MODEL, model, CTLFLAG_RD, cpu_model, 0, "");
131
132SYSCTL_INT(_hw, HW_NCPU, ncpu, CTLFLAG_RD, 0, 1, "");
133
134SYSCTL_INT(_hw, HW_BYTEORDER, byteorder, CTLFLAG_RD, 0, BYTE_ORDER, "");
135
136SYSCTL_INT(_hw, HW_PAGESIZE, pagesize, CTLFLAG_RD, 0, PAGE_SIZE, "");
137
138/* END_MIB */
139
140extern int vfs_update_wakeup;
141extern int vfs_update_interval;
142static int
143sysctl_kern_updateinterval SYSCTL_HANDLER_ARGS
144{
145	int error = sysctl_handle_int(oidp,
146		oidp->oid_arg1, oidp->oid_arg2, req);
147	if (!error)
148		wakeup(&vfs_update_wakeup);
149	return error;
150}
151
152SYSCTL_PROC(_kern, KERN_UPDATEINTERVAL, update, CTLTYPE_INT|CTLFLAG_RW,
153	&vfs_update_interval, 0, sysctl_kern_updateinterval, "");
154
155
156char hostname[MAXHOSTNAMELEN];
157int hostnamelen;
158static int
159sysctl_kern_hostname SYSCTL_HANDLER_ARGS
160{
161	int error = sysctl_handle_string(oidp,
162		oidp->oid_arg1, oidp->oid_arg2, req);
163	if (req->newptr && (error == 0 || error == ENOMEM))
164		hostnamelen = req->newlen;
165	return error;
166}
167
168SYSCTL_PROC(_kern, KERN_HOSTNAME, hostname, CTLTYPE_STRING|CTLFLAG_RW,
169	&hostname, sizeof(hostname), sysctl_kern_hostname, "");
170
171static int
172sysctl_order_cmp(const void *a, const void *b)
173{
174	const struct sysctl_oid **pa, **pb;
175
176	pa = (const struct sysctl_oid **)a;
177	pb = (const struct sysctl_oid **)b;
178	if (*pa == NULL)
179		return (1);
180	if (*pb == NULL)
181		return (-1);
182	return ((*pa)->oid_number - (*pb)->oid_number);
183}
184
185static void
186sysctl_order(void *arg)
187{
188	int j;
189	struct linker_set *l = (struct linker_set *) arg;
190	struct sysctl_oid **oidpp;
191
192	j = l->ls_length;
193	oidpp = (struct sysctl_oid **) l->ls_items;
194	for (; j--; oidpp++) {
195		if (!*oidpp)
196			continue;
197		if ((*oidpp)->oid_arg1 == arg) {
198			*oidpp = 0;
199			continue;
200		}
201		if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE)
202			if (!(*oidpp)->oid_handler)
203				sysctl_order((*oidpp)->oid_arg1);
204	}
205	qsort(l->ls_items, l->ls_length, sizeof l->ls_items[0],
206		sysctl_order_cmp);
207}
208
209SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_order, &sysctl_);
210
211static void
212sysctl_sysctl_debug_dump_node(struct linker_set *l, int i)
213{
214	int j, k;
215	struct sysctl_oid **oidpp;
216
217	j = l->ls_length;
218	oidpp = (struct sysctl_oid **) l->ls_items;
219	for (; j--; oidpp++) {
220
221		if (!*oidpp)
222			continue;
223
224		for (k=0; k<i; k++)
225			printf(" ");
226
227		if ((*oidpp)->oid_number > 100) {
228			printf("Junk! %p  # %d  %s  k %x  a1 %p  a2 %x  h %p\n",
229				*oidpp,
230		 		(*oidpp)->oid_number, (*oidpp)->oid_name,
231		 		(*oidpp)->oid_kind, (*oidpp)->oid_arg1,
232		 		(*oidpp)->oid_arg2, (*oidpp)->oid_handler);
233			continue;
234		}
235		printf("%d %s ", (*oidpp)->oid_number, (*oidpp)->oid_name);
236
237		printf("%c%c",
238			(*oidpp)->oid_kind & CTLFLAG_RD ? 'R':' ',
239			(*oidpp)->oid_kind & CTLFLAG_WR ? 'W':' ');
240
241		switch ((*oidpp)->oid_kind & CTLTYPE) {
242			case CTLTYPE_NODE:
243				if ((*oidpp)->oid_handler) {
244					printf(" Node(proc)\n");
245				} else {
246					printf(" Node\n");
247					sysctl_sysctl_debug_dump_node(
248						(*oidpp)->oid_arg1, i+2);
249				}
250				break;
251			case CTLTYPE_INT:    printf(" Int\n"); break;
252			case CTLTYPE_STRING: printf(" String\n"); break;
253			case CTLTYPE_QUAD:   printf(" Quad\n"); break;
254			case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
255			default:	     printf("\n");
256		}
257
258	}
259}
260
261
262static int
263sysctl_sysctl_debug SYSCTL_HANDLER_ARGS
264{
265	sysctl_sysctl_debug_dump_node(&sysctl_, 0);
266	return ENOENT;
267}
268
269SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
270	0, 0, sysctl_sysctl_debug, "");
271
272char domainname[MAXHOSTNAMELEN];
273int domainnamelen;
274static int
275sysctl_kern_domainname SYSCTL_HANDLER_ARGS
276{
277	int error = sysctl_handle_string(oidp,
278		oidp->oid_arg1, oidp->oid_arg2, req);
279	if (req->newptr && (error == 0 || error == ENOMEM))
280		domainnamelen = req->newlen;
281	return error;
282}
283
284SYSCTL_PROC(_kern, KERN_DOMAINNAME, domainname, CTLTYPE_STRING|CTLFLAG_RW,
285	&domainname, sizeof(domainname), sysctl_kern_domainname, "");
286
287long hostid;
288/* Some trouble here, if sizeof (int) != sizeof (long) */
289SYSCTL_INT(_kern, KERN_HOSTID, hostid, CTLFLAG_RW, &hostid, 0, "");
290
291/*
292 * Handle an integer, signed or unsigned.
293 * Two cases:
294 *     a variable:  point arg1 at it.
295 *     a constant:  pass it in arg2.
296 */
297
298int
299sysctl_handle_int SYSCTL_HANDLER_ARGS
300{
301	int error = 0;
302
303	if (arg1)
304		error = SYSCTL_OUT(req, arg1, sizeof(int));
305	else if (arg2)
306		error = SYSCTL_OUT(req, &arg2, sizeof(int));
307
308	if (error || !req->newptr)
309		return (error);
310
311	if (!arg1)
312		error = EPERM;
313	else
314		error = SYSCTL_IN(req, arg1, sizeof(int));
315	return (error);
316}
317
318/*
319 * Handle our generic '\0' terminated 'C' string.
320 * Two cases:
321 * 	a variable string:  point arg1 at it, arg2 is max length.
322 * 	a constant string:  point arg1 at it, arg2 is zero.
323 */
324
325int
326sysctl_handle_string SYSCTL_HANDLER_ARGS
327{
328	int error=0;
329
330	if (arg2)
331		error = SYSCTL_OUT(req, arg1, arg2);
332	else
333		error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
334
335	if (error || !req->newptr)
336		return (error);
337
338	if ((req->newlen - req->newidx) > arg2) {
339		error = E2BIG;
340	} else {
341		arg2 = (req->newlen - req->newidx);
342		error = SYSCTL_IN(req, arg1, arg2);
343		((char *)arg1)[arg2] = '\0';
344	}
345
346	return (error);
347}
348
349/*
350 * Handle any kind of opaque data.
351 * arg1 points to it, arg2 is the size.
352 */
353
354int
355sysctl_handle_opaque SYSCTL_HANDLER_ARGS
356{
357	int error;
358
359	error = SYSCTL_OUT(req, arg1, arg2);
360
361	if (error || !req->newptr)
362		return (error);
363
364	error = SYSCTL_IN(req, arg1, arg2);
365
366	return (error);
367}
368
369int
370sysctl_old_kernel(struct sysctl_req *req, void *p, int l)
371{
372	int i = min(req->oldlen - req->oldidx, l);
373	if (i) {
374		bcopy(p, req->oldptr + req->oldidx, i);
375		req->oldidx += i;
376	}
377	if (i != l)
378		return (ENOMEM);
379	else
380		return (0);
381
382}
383
384int
385sysctl_new_kernel(struct sysctl_req *req, void *p, int l)
386{
387	int i = req->newlen - req->newidx;
388	if (i < l)
389		return (EINVAL);
390	bcopy(req->newptr + req->newidx, p, l);
391	req->newidx += l;
392	return (0);
393}
394
395int
396sysctl_old_user(struct sysctl_req *req, void *p, int l)
397{
398	int error , i = min(req->oldlen - req->oldidx, l);
399
400	error  = copyout(p, req->oldptr + req->oldidx, i);
401	req->oldidx += i;
402	if (error)
403		return (error);
404	if (i < l)
405		return (ENOMEM);
406	return (error);
407}
408
409int
410sysctl_new_user(struct sysctl_req *req, void *p, int l)
411{
412	int error, i = req->newlen - req->newidx;
413	if (i < l)
414		return (EINVAL);
415	error = copyin(req->newptr + req->newidx, p, l);
416	req->newidx += l;
417	return (error);
418}
419
420#ifdef DEBUG
421static sysctlfn debug_sysctl;
422#endif
423
424/*
425 * Locking and stats
426 */
427static struct sysctl_lock {
428	int	sl_lock;
429	int	sl_want;
430	int	sl_locked;
431} memlock;
432
433
434
435/*
436 * Traverse our tree, and find the right node, execute whatever it points
437 * at, and return the resulting error code.
438 * We work entirely in kernel-space at this time.
439 */
440
441
442int
443sysctl_root SYSCTL_HANDLER_ARGS
444{
445	int *name = (int *) arg1;
446	int namelen = arg2;
447	int indx, i, j;
448	struct sysctl_oid **oidpp;
449	struct linker_set *lsp = &sysctl_;
450
451	j = lsp->ls_length;
452	oidpp = (struct sysctl_oid **) lsp->ls_items;
453
454	indx = 0;
455	while (j-- && indx < CTL_MAXNAME) {
456		if (*oidpp && ((*oidpp)->oid_number == name[indx])) {
457			indx++;
458			if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
459				if ((*oidpp)->oid_handler)
460					goto found;
461				if (indx == namelen)
462					return ENOENT;
463				lsp = (struct linker_set*)(*oidpp)->oid_arg1;
464				j = lsp->ls_length;
465				oidpp = (struct sysctl_oid **)lsp->ls_items;
466			} else {
467				if (indx != namelen)
468					return EISDIR;
469				goto found;
470			}
471		} else {
472			oidpp++;
473		}
474	}
475	return EJUSTRETURN;
476found:
477
478	/* If writing isn't allowed */
479	if (req->newptr && !((*oidpp)->oid_kind & CTLFLAG_WR))
480		return (EPERM);
481
482	if (!(*oidpp)->oid_handler)
483		return EINVAL;
484
485	if (((*oidpp)->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
486		i = ((*oidpp)->oid_handler) (*oidpp,
487					name + indx, namelen - indx,
488					req);
489	} else {
490		i = ((*oidpp)->oid_handler) (*oidpp,
491					(*oidpp)->oid_arg1, (*oidpp)->oid_arg2,
492					req);
493	}
494	return (i);
495}
496
497#ifndef _SYS_SYSPROTO_H_
498struct sysctl_args {
499	int	*name;
500	u_int	namelen;
501	void	*old;
502	size_t	*oldlenp;
503	void	*new;
504	size_t	newlen;
505};
506#endif
507
508int
509__sysctl(p, uap, retval)
510	struct proc *p;
511	register struct sysctl_args *uap;
512	int *retval;
513{
514	int error, name[CTL_MAXNAME];
515
516	if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
517		return (EINVAL);
518
519 	error = copyin(uap->name, &name, uap->namelen * sizeof(int));
520 	if (error)
521		return (error);
522
523	return (userland_sysctl(p, name, uap->namelen,
524		uap->old, uap->oldlenp, 0,
525		uap->new, uap->newlen, retval));
526}
527
528static sysctlfn kern_sysctl;
529
530/*
531 * This is used from various compatibility syscalls too.  That's why name
532 * must be in kernel space.
533 */
534int
535userland_sysctl(struct proc *p, int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, int *retval)
536{
537	int error = 0, dolock = 1, i, oldlen = 0;
538	u_int savelen = 0;
539	sysctlfn *fn;
540	struct sysctl_req req;
541
542	bzero(&req, sizeof req);
543
544	if (new != NULL && (error = suser(p->p_ucred, &p->p_acflag)))
545		return (error);
546
547	if (oldlenp) {
548		if (inkernel) {
549			req.oldlen = *oldlenp;
550		} else {
551			error = copyin(oldlenp, &req.oldlen,
552				sizeof(req.oldlen));
553			if (error)
554				return (error);
555		}
556		oldlen = req.oldlen;
557	}
558
559	if (old) {
560		if (!useracc(old, req.oldlen, B_WRITE))
561			return (EFAULT);
562		req.oldptr= old;
563	}
564
565	if (newlen) {
566		if (!useracc(new, req.newlen, B_READ))
567			return (EFAULT);
568		req.newlen = newlen;
569		req.newptr = new;
570	}
571
572	req.oldfunc = sysctl_old_user;
573	req.newfunc = sysctl_new_user;
574
575	error = sysctl_root(0, name, namelen, &req);
576
577        if (!error || error == ENOMEM) {
578		if (retval)
579			*retval = req.oldlen;
580		if (oldlenp) {
581			if (inkernel) {
582				*oldlenp = req.oldlen;
583			} else {
584				i = copyout(&req.oldlen, oldlenp,
585					sizeof(req.oldlen));
586				if (i)
587					error = i;
588			}
589		}
590		return (error);
591	}
592
593	switch (name[0]) {
594	case CTL_KERN:
595		fn = kern_sysctl;
596		if (name[1] != KERN_VNODE)      /* XXX */
597			dolock = 0;
598		break;
599	case CTL_HW:
600		fn = hw_sysctl;
601		break;
602	case CTL_VM:
603		fn = vm_sysctl;
604		break;
605	case CTL_NET:
606		fn = net_sysctl;
607		break;
608	case CTL_FS:
609		fn = fs_sysctl;
610		break;
611#ifdef DEBUG
612	case CTL_DEBUG:
613		fn = debug_sysctl;
614		break;
615#endif
616	default:
617		return (EOPNOTSUPP);
618	}
619	if (old != NULL) {
620		if (!useracc(old, oldlen, B_WRITE))
621			return (EFAULT);
622		while (memlock.sl_lock) {
623			memlock.sl_want = 1;
624			(void) tsleep((caddr_t)&memlock, PRIBIO+1, "sysctl", 0);
625			memlock.sl_locked++;
626		}
627		memlock.sl_lock = 1;
628		if (dolock)
629			vslock(old, oldlen);
630		savelen = oldlen;
631	}
632
633
634	error = (*fn)(name + 1, namelen - 1, old, &oldlen,
635	    new, newlen, p);
636
637
638	if (old != NULL) {
639		if (dolock)
640			vsunlock(old, savelen, B_WRITE);
641		memlock.sl_lock = 0;
642		if (memlock.sl_want) {
643			memlock.sl_want = 0;
644			wakeup((caddr_t)&memlock);
645		}
646	}
647#if 0
648	if (error) {
649		printf("SYSCTL_ERROR: ");
650		for(i=0;i<namelen;i++)
651			printf("%d ", name[i]);
652		printf("= %d\n", error);
653	}
654#endif
655	if (error)
656		return (error);
657	if (retval)
658		*retval = oldlen;
659	if (oldlenp) {
660		if (inkernel) {
661			*oldlenp = oldlen;
662		} else {
663			error = copyout(&oldlen, oldlenp, sizeof(oldlen));
664		}
665	}
666	return (error);
667}
668
669/*
670 * Attributes stored in the kernel.
671 */
672int securelevel = -1;
673
674/*
675 * kernel related system variables.
676 */
677static int
678kern_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
679	int *name;
680	u_int namelen;
681	void *oldp;
682	size_t *oldlenp;
683	void *newp;
684	size_t newlen;
685	struct proc *p;
686{
687	int error, level;
688	dev_t ndumpdev;
689
690	/* all sysctl names at this level are terminal */
691	if (namelen != 1 && !(name[0] == KERN_PROC || name[0] == KERN_PROF
692			      || name[0] == KERN_NTP_PLL))
693		return (ENOTDIR);		/* overloaded */
694
695	switch (name[0]) {
696
697	case KERN_SECURELVL:
698		level = securelevel;
699		if ((error = sysctl_int(oldp, oldlenp, newp, newlen, &level)) ||
700		    newp == NULL)
701			return (error);
702		if (level < securelevel && p->p_pid != 1)
703			return (EPERM);
704		securelevel = level;
705		return (0);
706	case KERN_VNODE:
707		return (sysctl_vnode(oldp, oldlenp));
708	case KERN_PROC:
709		return (sysctl_doproc(name + 1, namelen - 1, oldp, oldlenp));
710	case KERN_FILE:
711		return (sysctl_file(oldp, oldlenp));
712#ifdef GPROF
713	case KERN_PROF:
714		return (sysctl_doprof(name + 1, namelen - 1, oldp, oldlenp,
715		    newp, newlen));
716#endif
717	case KERN_NTP_PLL:
718		return (ntp_sysctl(name + 1, namelen - 1, oldp, oldlenp,
719				   newp, newlen, p));
720	case KERN_DUMPDEV:
721		ndumpdev = dumpdev;
722		error = sysctl_struct(oldp, oldlenp, newp, newlen, &ndumpdev,
723				      sizeof ndumpdev);
724		if (!error && ndumpdev != dumpdev) {
725			error = setdumpdev(ndumpdev);
726		}
727		return error;
728	default:
729		return (EOPNOTSUPP);
730	}
731	/* NOTREACHED */
732}
733
734/*
735 * hardware related system variables.
736 */
737int
738hw_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
739	int *name;
740	u_int namelen;
741	void *oldp;
742	size_t *oldlenp;
743	void *newp;
744	size_t newlen;
745	struct proc *p;
746{
747	/* almost all sysctl names at this level are terminal */
748	if (namelen != 1 && name[0] != HW_DEVCONF)
749		return (ENOTDIR);		/* overloaded */
750
751	switch (name[0]) {
752	case HW_PHYSMEM:
753		return (sysctl_rdint(oldp, oldlenp, newp, ctob(physmem)));
754	case HW_USERMEM:
755		return (sysctl_rdint(oldp, oldlenp, newp,
756		    ctob(physmem - cnt.v_wire_count)));
757	case HW_DEVCONF:
758		return (dev_sysctl(name + 1, namelen - 1, oldp, oldlenp,
759				   newp, newlen, p));
760	default:
761		return (EOPNOTSUPP);
762	}
763	/* NOTREACHED */
764}
765
766#ifdef DEBUG
767/*
768 * Debugging related system variables.
769 */
770struct ctldebug debug0, debug1, debug2, debug3, debug4;
771struct ctldebug debug5, debug6, debug7, debug8, debug9;
772struct ctldebug debug10, debug11, debug12, debug13, debug14;
773struct ctldebug debug15, debug16, debug17, debug18, debug19;
774static struct ctldebug *debugvars[CTL_DEBUG_MAXID] = {
775	&debug0, &debug1, &debug2, &debug3, &debug4,
776	&debug5, &debug6, &debug7, &debug8, &debug9,
777	&debug10, &debug11, &debug12, &debug13, &debug14,
778	&debug15, &debug16, &debug17, &debug18, &debug19,
779};
780static int
781debug_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
782	int *name;
783	u_int namelen;
784	void *oldp;
785	size_t *oldlenp;
786	void *newp;
787	size_t newlen;
788	struct proc *p;
789{
790	struct ctldebug *cdp;
791
792	/* all sysctl names at this level are name and field */
793	if (namelen != 2)
794		return (ENOTDIR);		/* overloaded */
795	cdp = debugvars[name[0]];
796	if (cdp->debugname == 0)
797		return (EOPNOTSUPP);
798	switch (name[1]) {
799	case CTL_DEBUG_NAME:
800		return (sysctl_rdstring(oldp, oldlenp, newp, cdp->debugname));
801	case CTL_DEBUG_VALUE:
802		return (sysctl_int(oldp, oldlenp, newp, newlen, cdp->debugvar));
803	default:
804		return (EOPNOTSUPP);
805	}
806	/* NOTREACHED */
807}
808#endif /* DEBUG */
809
810/*
811 * Validate parameters and get old / set new parameters
812 * for an integer-valued sysctl function.
813 */
814int
815sysctl_int(oldp, oldlenp, newp, newlen, valp)
816	void *oldp;
817	size_t *oldlenp;
818	void *newp;
819	size_t newlen;
820	int *valp;
821{
822	int error = 0;
823
824	if (oldp && *oldlenp < sizeof(int))
825		return (ENOMEM);
826	if (newp && newlen != sizeof(int))
827		return (EINVAL);
828	*oldlenp = sizeof(int);
829	if (oldp)
830		error = copyout(valp, oldp, sizeof(int));
831	if (error == 0 && newp)
832		error = copyin(newp, valp, sizeof(int));
833	return (error);
834}
835
836/*
837 * As above, but read-only.
838 */
839int
840sysctl_rdint(oldp, oldlenp, newp, val)
841	void *oldp;
842	size_t *oldlenp;
843	void *newp;
844	int val;
845{
846	int error = 0;
847
848	if (oldp && *oldlenp < sizeof(int))
849		return (ENOMEM);
850	if (newp)
851		return (EPERM);
852	*oldlenp = sizeof(int);
853	if (oldp)
854		error = copyout((caddr_t)&val, oldp, sizeof(int));
855	return (error);
856}
857
858/*
859 * Validate parameters and get old / set new parameters
860 * for a string-valued sysctl function.
861 */
862int
863sysctl_string(oldp, oldlenp, newp, newlen, str, maxlen)
864	void *oldp;
865	size_t *oldlenp;
866	void *newp;
867	size_t newlen;
868	char *str;
869	int maxlen;
870{
871	int len, error = 0, rval = 0;
872
873	len = strlen(str) + 1;
874	if (oldp && *oldlenp < len) {
875		len = *oldlenp;
876		rval = ENOMEM;
877	}
878	if (newp && newlen >= maxlen)
879		return (EINVAL);
880	if (oldp) {
881		*oldlenp = len;
882		error = copyout(str, oldp, len);
883		if (error)
884			rval = error;
885	}
886	if ((error == 0 || error == ENOMEM) && newp) {
887		error = copyin(newp, str, newlen);
888		if (error)
889			rval = error;
890		str[newlen] = 0;
891	}
892	return (rval);
893}
894
895/*
896 * As above, but read-only.
897 */
898int
899sysctl_rdstring(oldp, oldlenp, newp, str)
900	void *oldp;
901	size_t *oldlenp;
902	void *newp;
903	char *str;
904{
905	int len, error = 0, rval = 0;
906
907	len = strlen(str) + 1;
908	if (oldp && *oldlenp < len) {
909		len = *oldlenp;
910		rval = ENOMEM;
911	}
912	if (newp)
913		return (EPERM);
914	*oldlenp = len;
915	if (oldp)
916		error = copyout(str, oldp, len);
917		if (error)
918			rval = error;
919	return (rval);
920}
921
922/*
923 * Validate parameters and get old / set new parameters
924 * for a structure oriented sysctl function.
925 */
926int
927sysctl_struct(oldp, oldlenp, newp, newlen, sp, len)
928	void *oldp;
929	size_t *oldlenp;
930	void *newp;
931	size_t newlen;
932	void *sp;
933	int len;
934{
935	int error = 0;
936
937	if (oldp && *oldlenp < len)
938		return (ENOMEM);
939	if (newp && newlen > len)
940		return (EINVAL);
941	if (oldp) {
942		*oldlenp = len;
943		error = copyout(sp, oldp, len);
944	}
945	if (error == 0 && newp)
946		error = copyin(newp, sp, len);
947	return (error);
948}
949
950/*
951 * Validate parameters and get old parameters
952 * for a structure oriented sysctl function.
953 */
954int
955sysctl_rdstruct(oldp, oldlenp, newp, sp, len)
956	void *oldp;
957	size_t *oldlenp;
958	void *newp, *sp;
959	int len;
960{
961	int error = 0;
962
963	if (oldp && *oldlenp < len)
964		return (ENOMEM);
965	if (newp)
966		return (EPERM);
967	*oldlenp = len;
968	if (oldp)
969		error = copyout(sp, oldp, len);
970	return (error);
971}
972
973/*
974 * Get file structures.
975 */
976int
977sysctl_file(where, sizep)
978	char *where;
979	size_t *sizep;
980{
981	int buflen, error;
982	struct file *fp;
983	char *start = where;
984
985	buflen = *sizep;
986	if (where == NULL) {
987		/*
988		 * overestimate by 10 files
989		 */
990		*sizep = sizeof(filehead) + (nfiles + 10) * sizeof(struct file);
991		return (0);
992	}
993
994	/*
995	 * first copyout filehead
996	 */
997	if (buflen < sizeof(filehead)) {
998		*sizep = 0;
999		return (0);
1000	}
1001	error = copyout((caddr_t)&filehead, where, sizeof(filehead));
1002	if (error)
1003		return (error);
1004	buflen -= sizeof(filehead);
1005	where += sizeof(filehead);
1006
1007	/*
1008	 * followed by an array of file structures
1009	 */
1010	for (fp = filehead; fp != NULL; fp = fp->f_filef) {
1011		if (buflen < sizeof(struct file)) {
1012			*sizep = where - start;
1013			return (ENOMEM);
1014		}
1015		error = copyout((caddr_t)fp, where, sizeof (struct file));
1016		if (error)
1017			return (error);
1018		buflen -= sizeof(struct file);
1019		where += sizeof(struct file);
1020	}
1021	*sizep = where - start;
1022	return (0);
1023}
1024
1025/*
1026 * try over estimating by 5 procs
1027 */
1028#define KERN_PROCSLOP	(5 * sizeof (struct kinfo_proc))
1029
1030int
1031sysctl_doproc(name, namelen, where, sizep)
1032	int *name;
1033	u_int namelen;
1034	char *where;
1035	size_t *sizep;
1036{
1037	register struct proc *p;
1038	register struct kinfo_proc *dp = (struct kinfo_proc *)where;
1039	register int needed = 0;
1040	int buflen = where != NULL ? *sizep : 0;
1041	int doingzomb;
1042	struct eproc eproc;
1043	int error = 0;
1044
1045	if (namelen != 2 && !(namelen == 1 && name[0] == KERN_PROC_ALL))
1046		return (EINVAL);
1047	p = (struct proc *)allproc;
1048	doingzomb = 0;
1049again:
1050	for (; p != NULL; p = p->p_next) {
1051		/*
1052		 * Skip embryonic processes.
1053		 */
1054		if (p->p_stat == SIDL)
1055			continue;
1056		/*
1057		 * TODO - make more efficient (see notes below).
1058		 * do by session.
1059		 */
1060		switch (name[0]) {
1061
1062		case KERN_PROC_PID:
1063			/* could do this with just a lookup */
1064			if (p->p_pid != (pid_t)name[1])
1065				continue;
1066			break;
1067
1068		case KERN_PROC_PGRP:
1069			/* could do this by traversing pgrp */
1070			if (p->p_pgrp == NULL || p->p_pgrp->pg_id != (pid_t)name[1])
1071				continue;
1072			break;
1073
1074		case KERN_PROC_TTY:
1075			if ((p->p_flag & P_CONTROLT) == 0 ||
1076			    p->p_session == NULL ||
1077			    p->p_session->s_ttyp == NULL ||
1078			    p->p_session->s_ttyp->t_dev != (dev_t)name[1])
1079				continue;
1080			break;
1081
1082		case KERN_PROC_UID:
1083			if (p->p_ucred == NULL || p->p_ucred->cr_uid != (uid_t)name[1])
1084				continue;
1085			break;
1086
1087		case KERN_PROC_RUID:
1088			if (p->p_ucred == NULL || p->p_cred->p_ruid != (uid_t)name[1])
1089				continue;
1090			break;
1091		}
1092		if (buflen >= sizeof(struct kinfo_proc)) {
1093			fill_eproc(p, &eproc);
1094			error = copyout((caddr_t)p, &dp->kp_proc,
1095			    sizeof(struct proc));
1096			if (error)
1097				return (error);
1098			error = copyout((caddr_t)&eproc, &dp->kp_eproc,
1099			    sizeof(eproc));
1100			if (error)
1101				return (error);
1102			dp++;
1103			buflen -= sizeof(struct kinfo_proc);
1104		}
1105		needed += sizeof(struct kinfo_proc);
1106	}
1107	if (doingzomb == 0) {
1108		p = zombproc;
1109		doingzomb++;
1110		goto again;
1111	}
1112	if (where != NULL) {
1113		*sizep = (caddr_t)dp - where;
1114		if (needed > *sizep)
1115			return (ENOMEM);
1116	} else {
1117		needed += KERN_PROCSLOP;
1118		*sizep = needed;
1119	}
1120	return (0);
1121}
1122
1123/*
1124 * Fill in an eproc structure for the specified process.
1125 */
1126void
1127fill_eproc(p, ep)
1128	register struct proc *p;
1129	register struct eproc *ep;
1130{
1131	register struct tty *tp;
1132
1133	bzero(ep, sizeof(*ep));
1134
1135	ep->e_paddr = p;
1136	if (p->p_cred) {
1137		ep->e_pcred = *p->p_cred;
1138		if (p->p_ucred)
1139			ep->e_ucred = *p->p_ucred;
1140	}
1141	if (p->p_stat != SIDL && p->p_stat != SZOMB && p->p_vmspace != NULL) {
1142		register struct vmspace *vm = p->p_vmspace;
1143
1144#ifdef pmap_resident_count
1145		ep->e_vm.vm_rssize = pmap_resident_count(&vm->vm_pmap); /*XXX*/
1146#else
1147		ep->e_vm.vm_rssize = vm->vm_rssize;
1148#endif
1149		ep->e_vm.vm_tsize = vm->vm_tsize;
1150		ep->e_vm.vm_dsize = vm->vm_dsize;
1151		ep->e_vm.vm_ssize = vm->vm_ssize;
1152#ifndef sparc
1153		ep->e_vm.vm_pmap = vm->vm_pmap;
1154#endif
1155	}
1156	if (p->p_pptr)
1157		ep->e_ppid = p->p_pptr->p_pid;
1158	if (p->p_pgrp) {
1159		ep->e_sess = p->p_pgrp->pg_session;
1160		ep->e_pgid = p->p_pgrp->pg_id;
1161		ep->e_jobc = p->p_pgrp->pg_jobc;
1162	}
1163	if ((p->p_flag & P_CONTROLT) &&
1164	    (ep->e_sess != NULL) &&
1165	    ((tp = ep->e_sess->s_ttyp) != NULL)) {
1166		ep->e_tdev = tp->t_dev;
1167		ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1168		ep->e_tsess = tp->t_session;
1169	} else
1170		ep->e_tdev = NODEV;
1171	if (ep->e_sess && ep->e_sess->s_ttyvp)
1172		ep->e_flag = EPROC_CTTY;
1173	if (SESS_LEADER(p))
1174		ep->e_flag |= EPROC_SLEADER;
1175	if (p->p_wmesg) {
1176		strncpy(ep->e_wmesg, p->p_wmesg, WMESGLEN);
1177		ep->e_wmesg[WMESGLEN] = 0;
1178	}
1179}
1180
1181#ifdef COMPAT_43
1182#include <sys/socket.h>
1183#define	KINFO_PROC		(0<<8)
1184#define	KINFO_RT		(1<<8)
1185#define	KINFO_VNODE		(2<<8)
1186#define	KINFO_FILE		(3<<8)
1187#define	KINFO_METER		(4<<8)
1188#define	KINFO_LOADAVG		(5<<8)
1189#define	KINFO_CLOCKRATE		(6<<8)
1190
1191/* Non-standard BSDI extension - only present on their 4.3 net-2 releases */
1192#define	KINFO_BSDI_SYSINFO	(101<<8)
1193
1194/*
1195 * XXX this is bloat, but I hope it's better here than on the potentially
1196 * limited kernel stack...  -Peter
1197 */
1198
1199struct {
1200	int	bsdi_machine;		/* "i386" on BSD/386 */
1201/*      ^^^ this is an offset to the string, relative to the struct start */
1202	char	*pad0;
1203	long	pad1;
1204	long	pad2;
1205	long	pad3;
1206	u_long	pad4;
1207	u_long	pad5;
1208	u_long	pad6;
1209
1210	int	bsdi_ostype;		/* "BSD/386" on BSD/386 */
1211	int	bsdi_osrelease;		/* "1.1" on BSD/386 */
1212	long	pad7;
1213	long	pad8;
1214	char	*pad9;
1215
1216	long	pad10;
1217	long	pad11;
1218	int	pad12;
1219	long	pad13;
1220	quad_t	pad14;
1221	long	pad15;
1222
1223	struct	timeval pad16;
1224	/* we dont set this, because BSDI's uname used gethostname() instead */
1225	int	bsdi_hostname;		/* hostname on BSD/386 */
1226
1227	/* the actual string data is appended here */
1228
1229} bsdi_si;
1230/*
1231 * this data is appended to the end of the bsdi_si structure during copyout.
1232 * The "char *" offsets are relative to the base of the bsdi_si struct.
1233 * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings
1234 * should not exceed the length of the buffer here... (or else!! :-)
1235 */
1236char bsdi_strings[80];	/* It had better be less than this! */
1237
1238#ifndef _SYS_SYSPROTO_H_
1239struct getkerninfo_args {
1240	int	op;
1241	char	*where;
1242	int	*size;
1243	int	arg;
1244};
1245#endif
1246
1247int
1248ogetkerninfo(p, uap, retval)
1249	struct proc *p;
1250	register struct getkerninfo_args *uap;
1251	int *retval;
1252{
1253	int error, name[6];
1254	u_int size;
1255
1256	switch (uap->op & 0xff00) {
1257
1258	case KINFO_RT:
1259		name[0] = CTL_NET;
1260		name[1] = PF_ROUTE;
1261		name[2] = 0;
1262		name[3] = (uap->op & 0xff0000) >> 16;
1263		name[4] = uap->op & 0xff;
1264		name[5] = uap->arg;
1265		error = userland_sysctl(p, name, 6, uap->where, uap->size,
1266			0, 0, 0, 0);
1267		break;
1268
1269	case KINFO_VNODE:
1270		name[0] = CTL_KERN;
1271		name[1] = KERN_VNODE;
1272		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1273			0, 0, 0, 0);
1274		break;
1275
1276	case KINFO_PROC:
1277		name[0] = CTL_KERN;
1278		name[1] = KERN_PROC;
1279		name[2] = uap->op & 0xff;
1280		name[3] = uap->arg;
1281		error = userland_sysctl(p, name, 4, uap->where, uap->size,
1282			0, 0, 0, 0);
1283		break;
1284
1285	case KINFO_FILE:
1286		name[0] = CTL_KERN;
1287		name[1] = KERN_FILE;
1288		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1289			0, 0, 0, 0);
1290		break;
1291
1292	case KINFO_METER:
1293		name[0] = CTL_VM;
1294		name[1] = VM_METER;
1295		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1296			0, 0, 0, 0);
1297		break;
1298
1299	case KINFO_LOADAVG:
1300		name[0] = CTL_VM;
1301		name[1] = VM_LOADAVG;
1302		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1303			0, 0, 0, 0);
1304		break;
1305
1306	case KINFO_CLOCKRATE:
1307		name[0] = CTL_KERN;
1308		name[1] = KERN_CLOCKRATE;
1309		error = userland_sysctl(p, name, 2, uap->where, uap->size,
1310			0, 0, 0, 0);
1311		break;
1312
1313	case KINFO_BSDI_SYSINFO: {
1314		/*
1315		 * this is pretty crude, but it's just enough for uname()
1316		 * from BSDI's 1.x libc to work.
1317		 *
1318		 * In particular, it doesn't return the same results when
1319		 * the supplied buffer is too small.  BSDI's version apparently
1320		 * will return the amount copied, and set the *size to how
1321		 * much was needed.  The emulation framework here isn't capable
1322		 * of that, so we just set both to the amount copied.
1323		 * BSDI's 2.x product apparently fails with ENOMEM in this
1324		 * scenario.
1325		 */
1326
1327		u_int needed;
1328		u_int left;
1329		char *s;
1330
1331		bzero((char *)&bsdi_si, sizeof(bsdi_si));
1332		bzero(bsdi_strings, sizeof(bsdi_strings));
1333
1334		s = bsdi_strings;
1335
1336		bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si);
1337		strcpy(s, ostype);
1338		s += strlen(s) + 1;
1339
1340		bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si);
1341		strcpy(s, osrelease);
1342		s += strlen(s) + 1;
1343
1344		bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si);
1345		strcpy(s, machine);
1346		s += strlen(s) + 1;
1347
1348		needed = sizeof(bsdi_si) + (s - bsdi_strings);
1349
1350		if (uap->where == NULL) {
1351			/* process is asking how much buffer to supply.. */
1352			size = needed;
1353			error = 0;
1354			break;
1355		}
1356
1357
1358		/* if too much buffer supplied, trim it down */
1359		if (size > needed)
1360			size = needed;
1361
1362		/* how much of the buffer is remaining */
1363		left = size;
1364
1365		if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0)
1366			break;
1367
1368		/* is there any point in continuing? */
1369		if (left > sizeof(bsdi_si)) {
1370			left -= sizeof(bsdi_si);
1371			error = copyout(&bsdi_strings,
1372					uap->where + sizeof(bsdi_si), left);
1373		}
1374		break;
1375	}
1376
1377	default:
1378		return (EOPNOTSUPP);
1379	}
1380	if (error)
1381		return (error);
1382	*retval = size;
1383	if (uap->size)
1384		error = copyout((caddr_t)&size, (caddr_t)uap->size,
1385		    sizeof(size));
1386	return (error);
1387}
1388#endif /* COMPAT_43 */
1389