xref: /openbsd-src/sys/kern/kern_sysctl.c (revision acdebe0390e889e10c40319e8cdc977aae8f04f5)
1 /*	$OpenBSD: kern_sysctl.c,v 1.445 2024/08/26 08:24:25 mvs Exp $	*/
2 /*	$NetBSD: kern_sysctl.c,v 1.17 1996/05/20 17:49:05 mrg Exp $	*/
3 
4 /*-
5  * Copyright (c) 1982, 1986, 1989, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Mike Karels at Berkeley Software Design, Inc.
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. 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 /*
39  * sysctl system call.
40  */
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/atomic.h>
45 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/pool.h>
48 #include <sys/proc.h>
49 #include <sys/resourcevar.h>
50 #include <sys/signalvar.h>
51 #include <sys/fcntl.h>
52 #include <sys/file.h>
53 #include <sys/filedesc.h>
54 #include <sys/vnode.h>
55 #include <sys/unistd.h>
56 #include <sys/buf.h>
57 #include <sys/clockintr.h>
58 #include <sys/tty.h>
59 #include <sys/disklabel.h>
60 #include <sys/disk.h>
61 #include <sys/sysctl.h>
62 #include <sys/msgbuf.h>
63 #include <sys/vmmeter.h>
64 #include <sys/namei.h>
65 #include <sys/exec.h>
66 #include <sys/mbuf.h>
67 #include <sys/percpu.h>
68 #include <sys/sensors.h>
69 #include <sys/pipe.h>
70 #include <sys/eventvar.h>
71 #include <sys/socketvar.h>
72 #include <sys/socket.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/pledge.h>
76 #include <sys/timetc.h>
77 #include <sys/evcount.h>
78 #include <sys/un.h>
79 #include <sys/unpcb.h>
80 #include <sys/sched.h>
81 #include <sys/mount.h>
82 #include <sys/syscallargs.h>
83 #include <sys/wait.h>
84 #include <sys/witness.h>
85 
86 #include <uvm/uvm_extern.h>
87 
88 #include <dev/cons.h>
89 
90 #include <dev/usb/ucomvar.h>
91 
92 #include <net/route.h>
93 #include <netinet/in.h>
94 #include <netinet/ip.h>
95 #include <netinet/ip_var.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet/ip6.h>
98 #include <netinet/tcp.h>
99 #include <netinet/tcp_timer.h>
100 #include <netinet/tcp_var.h>
101 #include <netinet/udp.h>
102 #include <netinet/udp_var.h>
103 #include <netinet6/ip6_var.h>
104 
105 #ifdef DDB
106 #include <ddb/db_var.h>
107 #endif
108 
109 #ifdef SYSVMSG
110 #include <sys/msg.h>
111 #endif
112 #ifdef SYSVSEM
113 #include <sys/sem.h>
114 #endif
115 #ifdef SYSVSHM
116 #include <sys/shm.h>
117 #endif
118 
119 #include "audio.h"
120 #include "dt.h"
121 #include "pf.h"
122 #include "ucom.h"
123 #include "video.h"
124 
125 extern struct forkstat forkstat;
126 extern struct nchstats nchstats;
127 extern int fscale;
128 extern fixpt_t ccpu;
129 extern long numvnodes;
130 extern int allowdt;
131 extern int audio_record_enable;
132 extern int video_record_enable;
133 extern int autoconf_serial;
134 
135 int allowkmem;
136 
137 int sysctl_securelevel(void *, size_t *, void *, size_t, struct proc *);
138 int sysctl_diskinit(int, struct proc *);
139 int sysctl_proc_args(int *, u_int, void *, size_t *, struct proc *);
140 int sysctl_proc_cwd(int *, u_int, void *, size_t *, struct proc *);
141 int sysctl_proc_nobroadcastkill(int *, u_int, void *, size_t, void *, size_t *,
142 	struct proc *);
143 int sysctl_proc_vmmap(int *, u_int, void *, size_t *, struct proc *);
144 int sysctl_intrcnt(int *, u_int, void *, size_t *);
145 int sysctl_sensors(int *, u_int, void *, size_t *, void *, size_t);
146 int sysctl_cptime2(int *, u_int, void *, size_t *, void *, size_t);
147 int sysctl_audio(int *, u_int, void *, size_t *, void *, size_t);
148 int sysctl_video(int *, u_int, void *, size_t *, void *, size_t);
149 int sysctl_cpustats(int *, u_int, void *, size_t *, void *, size_t);
150 int sysctl_utc_offset(void *, size_t *, void *, size_t);
151 int sysctl_hwbattery(int *, u_int, void *, size_t *, void *, size_t);
152 
153 void fill_file(struct kinfo_file *, struct file *, struct filedesc *, int,
154     struct vnode *, struct process *, struct proc *, struct socket *, int);
155 void fill_kproc(struct process *, struct kinfo_proc *, struct proc *, int);
156 
157 int kern_sysctl_locked(int *, u_int, void *, size_t *, void *, size_t,
158 	struct proc *);
159 int hw_sysctl_locked(int *, u_int, void *, size_t *,void *, size_t,
160 	struct proc *);
161 
162 int (*cpu_cpuspeed)(int *);
163 
164 /*
165  * Lock to avoid too many processes vslocking a large amount of memory
166  * at the same time.
167  */
168 struct rwlock sysctl_lock = RWLOCK_INITIALIZER("sysctllk");
169 struct rwlock sysctl_disklock = RWLOCK_INITIALIZER("sysctldlk");
170 
171 int
172 sysctl_vslock(void *addr, size_t len)
173 {
174 	int error;
175 
176 	error = rw_enter(&sysctl_lock, RW_WRITE|RW_INTR);
177 	if (error)
178 		return (error);
179 	KERNEL_LOCK();
180 
181 	if (addr) {
182 		if (atop(len) > uvmexp.wiredmax - uvmexp.wired) {
183 			error = ENOMEM;
184 			goto out;
185 		}
186 		error = uvm_vslock(curproc, addr, len, PROT_READ | PROT_WRITE);
187 		if (error)
188 			goto out;
189 	}
190 
191 	return (0);
192 out:
193 	KERNEL_UNLOCK();
194 	rw_exit_write(&sysctl_lock);
195 	return (error);
196 }
197 
198 void
199 sysctl_vsunlock(void *addr, size_t len)
200 {
201 	KERNEL_ASSERT_LOCKED();
202 
203 	if (addr)
204 		uvm_vsunlock(curproc, addr, len);
205 	KERNEL_UNLOCK();
206 	rw_exit_write(&sysctl_lock);
207 }
208 
209 int
210 sys_sysctl(struct proc *p, void *v, register_t *retval)
211 {
212 	struct sys_sysctl_args /* {
213 		syscallarg(const int *) name;
214 		syscallarg(u_int) namelen;
215 		syscallarg(void *) old;
216 		syscallarg(size_t *) oldlenp;
217 		syscallarg(void *) new;
218 		syscallarg(size_t) newlen;
219 	} */ *uap = v;
220 	int error, dolock = 1;
221 	size_t savelen = 0, oldlen = 0;
222 	sysctlfn *fn;
223 	int name[CTL_MAXNAME];
224 
225 	if (SCARG(uap, new) != NULL &&
226 	    (error = suser(p)))
227 		return (error);
228 	/*
229 	 * all top-level sysctl names are non-terminal
230 	 */
231 	if (SCARG(uap, namelen) > CTL_MAXNAME || SCARG(uap, namelen) < 2)
232 		return (EINVAL);
233 	error = copyin(SCARG(uap, name), name,
234 		       SCARG(uap, namelen) * sizeof(int));
235 	if (error)
236 		return (error);
237 
238 	error = pledge_sysctl(p, SCARG(uap, namelen),
239 	    name, SCARG(uap, new));
240 	if (error)
241 		return (error);
242 
243 	switch (name[0]) {
244 	case CTL_KERN:
245 		dolock = 0;
246 		fn = kern_sysctl;
247 		break;
248 	case CTL_HW:
249 		dolock = 0;
250 		fn = hw_sysctl;
251 		break;
252 	case CTL_VM:
253 		fn = uvm_sysctl;
254 		break;
255 	case CTL_NET:
256 		dolock = 0;
257 		fn = net_sysctl;
258 		break;
259 	case CTL_FS:
260 		fn = fs_sysctl;
261 		break;
262 	case CTL_VFS:
263 		fn = vfs_sysctl;
264 		break;
265 	case CTL_MACHDEP:
266 		fn = cpu_sysctl;
267 		break;
268 #ifdef DEBUG_SYSCTL
269 	case CTL_DEBUG:
270 		fn = debug_sysctl;
271 		break;
272 #endif
273 #ifdef DDB
274 	case CTL_DDB:
275 		fn = ddb_sysctl;
276 		break;
277 #endif
278 	default:
279 		return (EOPNOTSUPP);
280 	}
281 
282 	if (SCARG(uap, oldlenp) &&
283 	    (error = copyin(SCARG(uap, oldlenp), &oldlen, sizeof(oldlen))))
284 		return (error);
285 
286 	if (dolock) {
287 		error = sysctl_vslock(SCARG(uap, old), oldlen);
288 		if (error)
289 			return (error);
290 		savelen = oldlen;
291 	}
292 	error = (*fn)(&name[1], SCARG(uap, namelen) - 1, SCARG(uap, old),
293 	    &oldlen, SCARG(uap, new), SCARG(uap, newlen), p);
294 	if (dolock)
295 		sysctl_vsunlock(SCARG(uap, old), savelen);
296 
297 	if (error)
298 		return (error);
299 	if (SCARG(uap, oldlenp))
300 		error = copyout(&oldlen, SCARG(uap, oldlenp), sizeof(oldlen));
301 	return (error);
302 }
303 
304 /*
305  * Attributes stored in the kernel.
306  */
307 char hostname[MAXHOSTNAMELEN];
308 int hostnamelen;
309 char domainname[MAXHOSTNAMELEN];
310 int domainnamelen;
311 int hostid;
312 char *disknames = NULL;
313 size_t disknameslen;
314 struct diskstats *diskstats = NULL;
315 size_t diskstatslen;
316 int securelevel;
317 
318 /* morally const values reported by sysctl_bounded_arr */
319 static int arg_max = ARG_MAX;
320 static int openbsd = OpenBSD;
321 static int posix_version = _POSIX_VERSION;
322 static int ngroups_max = NGROUPS_MAX;
323 static int int_zero = 0;
324 static int int_one = 1;
325 static int maxpartitions = MAXPARTITIONS;
326 static int raw_part = RAW_PART;
327 
328 extern int somaxconn, sominconn;
329 extern int nosuidcoredump;
330 extern int maxlocksperuid;
331 extern int uvm_wxabort;
332 extern int global_ptrace;
333 
334 const struct sysctl_bounded_args kern_vars[] = {
335 	{KERN_OSREV, &openbsd, SYSCTL_INT_READONLY},
336 	{KERN_MAXVNODES, &maxvnodes, 0, INT_MAX},
337 	{KERN_MAXPROC, &maxprocess, 0, INT_MAX},
338 	{KERN_MAXFILES, &maxfiles, 0, INT_MAX},
339 	{KERN_NFILES, &numfiles, SYSCTL_INT_READONLY},
340 	{KERN_TTYCOUNT, &tty_count, SYSCTL_INT_READONLY},
341 	{KERN_ARGMAX, &arg_max, SYSCTL_INT_READONLY},
342 	{KERN_POSIX1, &posix_version, SYSCTL_INT_READONLY},
343 	{KERN_NGROUPS, &ngroups_max, SYSCTL_INT_READONLY},
344 	{KERN_JOB_CONTROL, &int_one, SYSCTL_INT_READONLY},
345 	{KERN_SAVED_IDS, &int_one, SYSCTL_INT_READONLY},
346 	{KERN_MAXPARTITIONS, &maxpartitions, SYSCTL_INT_READONLY},
347 	{KERN_RAWPARTITION, &raw_part, SYSCTL_INT_READONLY},
348 	{KERN_MAXTHREAD, &maxthread, 0, INT_MAX},
349 	{KERN_NTHREADS, &nthreads, SYSCTL_INT_READONLY},
350 	{KERN_SOMAXCONN, &somaxconn, 0, SHRT_MAX},
351 	{KERN_SOMINCONN, &sominconn, 0, SHRT_MAX},
352 	{KERN_NOSUIDCOREDUMP, &nosuidcoredump, 0, 3},
353 	{KERN_FSYNC, &int_one, SYSCTL_INT_READONLY},
354 	{KERN_SYSVMSG,
355 #ifdef SYSVMSG
356 	 &int_one,
357 #else
358 	 &int_zero,
359 #endif
360 	 SYSCTL_INT_READONLY},
361 	{KERN_SYSVSEM,
362 #ifdef SYSVSEM
363 	 &int_one,
364 #else
365 	 &int_zero,
366 #endif
367 	 SYSCTL_INT_READONLY},
368 	{KERN_SYSVSHM,
369 #ifdef SYSVSHM
370 	 &int_one,
371 #else
372 	 &int_zero,
373 #endif
374 	 SYSCTL_INT_READONLY},
375 	{KERN_FSCALE, &fscale, SYSCTL_INT_READONLY},
376 	{KERN_CCPU, &ccpu, SYSCTL_INT_READONLY},
377 	{KERN_NPROCS, &nprocesses, SYSCTL_INT_READONLY},
378 	{KERN_SPLASSERT, &splassert_ctl, 0, 3},
379 	{KERN_MAXLOCKSPERUID, &maxlocksperuid, 0, INT_MAX},
380 	{KERN_WXABORT, &uvm_wxabort, 0, 1},
381 	{KERN_NETLIVELOCKS, &int_zero, SYSCTL_INT_READONLY},
382 #ifdef PTRACE
383 	{KERN_GLOBAL_PTRACE, &global_ptrace, 0, 1},
384 #endif
385 	{KERN_AUTOCONF_SERIAL, &autoconf_serial, SYSCTL_INT_READONLY},
386 };
387 
388 int
389 kern_sysctl_dirs(int top_name, int *name, u_int namelen,
390     void *oldp, size_t *oldlenp, void *newp, size_t newlen, struct proc *p)
391 {
392 	switch (top_name) {
393 #ifndef SMALL_KERNEL
394 	case KERN_PROC:
395 		return (sysctl_doproc(name, namelen, oldp, oldlenp));
396 	case KERN_PROC_ARGS:
397 		return (sysctl_proc_args(name, namelen, oldp, oldlenp, p));
398 	case KERN_PROC_CWD:
399 		return (sysctl_proc_cwd(name, namelen, oldp, oldlenp, p));
400 	case KERN_PROC_NOBROADCASTKILL:
401 		return (sysctl_proc_nobroadcastkill(name, namelen,
402 		     newp, newlen, oldp, oldlenp, p));
403 	case KERN_PROC_VMMAP:
404 		return (sysctl_proc_vmmap(name, namelen, oldp, oldlenp, p));
405 	case KERN_FILE:
406 		return (sysctl_file(name, namelen, oldp, oldlenp, p));
407 #endif
408 #if defined(GPROF) || defined(DDBPROF)
409 	case KERN_PROF:
410 		return (sysctl_doprof(name, namelen, oldp, oldlenp,
411 		    newp, newlen));
412 #endif
413 	case KERN_MALLOCSTATS:
414 		return (sysctl_malloc(name, namelen, oldp, oldlenp,
415 		    newp, newlen, p));
416 	case KERN_TTY:
417 		return (sysctl_tty(name, namelen, oldp, oldlenp,
418 		    newp, newlen));
419 	case KERN_POOL:
420 		return (sysctl_dopool(name, namelen, oldp, oldlenp));
421 #if defined(SYSVMSG) || defined(SYSVSEM) || defined(SYSVSHM)
422 	case KERN_SYSVIPC_INFO:
423 		return (sysctl_sysvipc(name, namelen, oldp, oldlenp));
424 #endif
425 #ifdef SYSVSEM
426 	case KERN_SEMINFO:
427 		return (sysctl_sysvsem(name, namelen, oldp, oldlenp,
428 		    newp, newlen));
429 #endif
430 #ifdef SYSVSHM
431 	case KERN_SHMINFO:
432 		return (sysctl_sysvshm(name, namelen, oldp, oldlenp,
433 		    newp, newlen));
434 #endif
435 #ifndef SMALL_KERNEL
436 	case KERN_INTRCNT:
437 		return (sysctl_intrcnt(name, namelen, oldp, oldlenp));
438 	case KERN_WATCHDOG:
439 		return (sysctl_wdog(name, namelen, oldp, oldlenp,
440 		    newp, newlen));
441 #endif
442 #ifndef SMALL_KERNEL
443 	case KERN_EVCOUNT:
444 		return (evcount_sysctl(name, namelen, oldp, oldlenp,
445 		    newp, newlen));
446 #endif
447 	case KERN_TIMECOUNTER:
448 		return (sysctl_tc(name, namelen, oldp, oldlenp, newp, newlen));
449 	case KERN_CPTIME2:
450 		return (sysctl_cptime2(name, namelen, oldp, oldlenp,
451 		    newp, newlen));
452 #ifdef WITNESS
453 	case KERN_WITNESSWATCH:
454 		return witness_sysctl_watch(oldp, oldlenp, newp, newlen);
455 	case KERN_WITNESS:
456 		return witness_sysctl(name, namelen, oldp, oldlenp,
457 		    newp, newlen);
458 #endif
459 #if NVIDEO > 0
460 	case KERN_VIDEO:
461 		return (sysctl_video(name, namelen, oldp, oldlenp,
462 		    newp, newlen));
463 #endif
464 	case KERN_CPUSTATS:
465 		return (sysctl_cpustats(name, namelen, oldp, oldlenp,
466 		    newp, newlen));
467 	case KERN_CLOCKINTR:
468 		return sysctl_clockintr(name, namelen, oldp, oldlenp, newp,
469 		    newlen);
470 	default:
471 		return (ENOTDIR);	/* overloaded */
472 	}
473 }
474 
475 /*
476  * kernel related system variables.
477  */
478 int
479 kern_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
480     size_t newlen, struct proc *p)
481 {
482 	int error;
483 	size_t savelen;
484 
485 	/* dispatch the non-terminal nodes first */
486 	if (namelen != 1) {
487 		switch (name[0]) {
488 #if NAUDIO > 0
489 		case KERN_AUDIO:
490 			return (sysctl_audio(name + 1, namelen - 1,
491 			    oldp, oldlenp, newp, newlen));
492 #endif
493 		default:
494 			break;
495 		}
496 
497 		savelen = *oldlenp;
498 		if ((error = sysctl_vslock(oldp, savelen)))
499 			return (error);
500 		error = kern_sysctl_dirs(name[0], name + 1, namelen - 1,
501 		    oldp, oldlenp, newp, newlen, p);
502 		sysctl_vsunlock(oldp, savelen);
503 		return (error);
504 	}
505 
506 	switch (name[0]) {
507 	case KERN_OSTYPE:
508 		return (sysctl_rdstring(oldp, oldlenp, newp, ostype));
509 	case KERN_OSRELEASE:
510 		return (sysctl_rdstring(oldp, oldlenp, newp, osrelease));
511 	case KERN_OSVERSION:
512 		return (sysctl_rdstring(oldp, oldlenp, newp, osversion));
513 	case KERN_VERSION:
514 		return (sysctl_rdstring(oldp, oldlenp, newp, version));
515 	case KERN_NUMVNODES:  /* XXX numvnodes is a long */
516 		return (sysctl_rdint(oldp, oldlenp, newp, numvnodes));
517 #if NDT > 0
518 	case KERN_ALLOWDT:
519 		return (sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
520 		    &allowdt));
521 #endif
522 	case KERN_HOSTID:
523 		return (sysctl_int(oldp, oldlenp, newp, newlen, &hostid));
524 	case KERN_CLOCKRATE:
525 		return (sysctl_clockrate(oldp, oldlenp, newp));
526 	case KERN_BOOTTIME: {
527 		struct timeval bt;
528 		memset(&bt, 0, sizeof bt);
529 		microboottime(&bt);
530 		return (sysctl_rdstruct(oldp, oldlenp, newp, &bt, sizeof bt));
531 	}
532 	case KERN_MBSTAT: {
533 		extern struct cpumem *mbstat;
534 		uint64_t counters[MBSTAT_COUNT];
535 		struct mbstat mbs;
536 		unsigned int i;
537 
538 		memset(&mbs, 0, sizeof(mbs));
539 		counters_read(mbstat, counters, MBSTAT_COUNT, NULL);
540 		for (i = 0; i < MBSTAT_TYPES; i++)
541 			mbs.m_mtypes[i] = counters[i];
542 
543 		mbs.m_drops = counters[MBSTAT_DROPS];
544 		mbs.m_wait = counters[MBSTAT_WAIT];
545 		mbs.m_drain = counters[MBSTAT_DRAIN];
546 
547 		return (sysctl_rdstruct(oldp, oldlenp, newp,
548 		    &mbs, sizeof(mbs)));
549 	}
550 	case KERN_MSGBUFSIZE:
551 	case KERN_CONSBUFSIZE: {
552 		struct msgbuf *mp;
553 		mp = (name[0] == KERN_MSGBUFSIZE) ? msgbufp : consbufp;
554 		/*
555 		 * deal with cases where the message buffer has
556 		 * become corrupted.
557 		 */
558 		if (!mp || mp->msg_magic != MSG_MAGIC)
559 			return (ENXIO);
560 		return (sysctl_rdint(oldp, oldlenp, newp, mp->msg_bufs));
561 	}
562 	case KERN_OSREV:
563 	case KERN_MAXPROC:
564 	case KERN_MAXFILES:
565 	case KERN_NFILES:
566 	case KERN_TTYCOUNT:
567 	case KERN_ARGMAX:
568 	case KERN_POSIX1:
569 	case KERN_NGROUPS:
570 	case KERN_JOB_CONTROL:
571 	case KERN_SAVED_IDS:
572 	case KERN_MAXPARTITIONS:
573 	case KERN_RAWPARTITION:
574 	case KERN_MAXTHREAD:
575 	case KERN_NTHREADS:
576 	case KERN_SOMAXCONN:
577 	case KERN_SOMINCONN:
578 	case KERN_FSYNC:
579 	case KERN_SYSVMSG:
580 	case KERN_SYSVSEM:
581 	case KERN_SYSVSHM:
582 	case KERN_FSCALE:
583 	case KERN_CCPU:
584 	case KERN_NPROCS:
585 	case KERN_NETLIVELOCKS:
586 	case KERN_AUTOCONF_SERIAL:
587 		return (sysctl_bounded_arr(kern_vars, nitems(kern_vars), name,
588 		    namelen, oldp, oldlenp, newp, newlen));
589 	}
590 
591 	savelen = *oldlenp;
592 	if ((error = sysctl_vslock(oldp, savelen)))
593 		return (error);
594 	error = kern_sysctl_locked(name, namelen, oldp, oldlenp,
595 	    newp, newlen, p);
596 	sysctl_vsunlock(oldp, savelen);
597 
598 	return (error);
599 }
600 
601 int
602 kern_sysctl_locked(int *name, u_int namelen, void *oldp, size_t *oldlenp,
603     void *newp, size_t newlen, struct proc *p)
604 {
605 	int error, stackgap;
606 	dev_t dev;
607 	extern int pool_debug;
608 
609 	switch (name[0]) {
610 	case KERN_SECURELVL:
611 		return (sysctl_securelevel(oldp, oldlenp, newp, newlen, p));
612 	case KERN_ALLOWKMEM:
613 		return (sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
614 		    &allowkmem));
615 	case KERN_HOSTNAME:
616 		error = sysctl_tstring(oldp, oldlenp, newp, newlen,
617 		    hostname, sizeof(hostname));
618 		if (newp && !error)
619 			hostnamelen = newlen;
620 		return (error);
621 	case KERN_DOMAINNAME:
622 		if (securelevel >= 1 && domainnamelen && newp)
623 			error = EPERM;
624 		else
625 			error = sysctl_tstring(oldp, oldlenp, newp, newlen,
626 			    domainname, sizeof(domainname));
627 		if (newp && !error)
628 			domainnamelen = newlen;
629 		return (error);
630 	case KERN_CONSBUF:
631 		if ((error = suser(p)))
632 			return (error);
633 		/* FALLTHROUGH */
634 	case KERN_MSGBUF: {
635 		struct msgbuf *mp;
636 		mp = (name[0] == KERN_MSGBUF) ? msgbufp : consbufp;
637 		/*
638 		 * deal with cases where the message buffer has
639 		 * become corrupted.
640 		 */
641 		if (!mp || mp->msg_magic != MSG_MAGIC)
642 			return (ENXIO);
643 		return (sysctl_rdstruct(oldp, oldlenp, newp, mp,
644 		    mp->msg_bufs + offsetof(struct msgbuf, msg_bufc)));
645 	}
646 	case KERN_CPTIME:
647 	{
648 		CPU_INFO_ITERATOR cii;
649 		struct cpu_info *ci;
650 		long cp_time[CPUSTATES];
651 		int i, n = 0;
652 
653 		memset(cp_time, 0, sizeof(cp_time));
654 
655 		CPU_INFO_FOREACH(cii, ci) {
656 			if (!cpu_is_online(ci))
657 				continue;
658 			n++;
659 			for (i = 0; i < CPUSTATES; i++)
660 				cp_time[i] += ci->ci_schedstate.spc_cp_time[i];
661 		}
662 
663 		for (i = 0; i < CPUSTATES; i++)
664 			cp_time[i] /= n;
665 
666 		return (sysctl_rdstruct(oldp, oldlenp, newp, &cp_time,
667 		    sizeof(cp_time)));
668 	}
669 	case KERN_NCHSTATS:
670 		return (sysctl_rdstruct(oldp, oldlenp, newp, &nchstats,
671 		    sizeof(struct nchstats)));
672 	case KERN_FORKSTAT:
673 		return (sysctl_rdstruct(oldp, oldlenp, newp, &forkstat,
674 		    sizeof(struct forkstat)));
675 	case KERN_STACKGAPRANDOM:
676 		stackgap = stackgap_random;
677 		error = sysctl_int(oldp, oldlenp, newp, newlen, &stackgap);
678 		if (error)
679 			return (error);
680 		/*
681 		 * Safety harness.
682 		 */
683 		if ((stackgap < ALIGNBYTES && stackgap != 0) ||
684 		    !powerof2(stackgap) || stackgap >= MAXSSIZ)
685 			return (EINVAL);
686 		stackgap_random = stackgap;
687 		return (0);
688 	case KERN_MAXCLUSTERS: {
689 		int val = nmbclust;
690 		error = sysctl_int(oldp, oldlenp, newp, newlen, &val);
691 		if (error == 0 && val != nmbclust)
692 			error = nmbclust_update(val);
693 		return (error);
694 	}
695 	case KERN_CACHEPCT: {
696 		u_int64_t dmapages;
697 		int opct, pgs;
698 		opct = bufcachepercent;
699 		error = sysctl_int(oldp, oldlenp, newp, newlen,
700 		    &bufcachepercent);
701 		if (error)
702 			return(error);
703 		if (bufcachepercent > 90 || bufcachepercent < 5) {
704 			bufcachepercent = opct;
705 			return (EINVAL);
706 		}
707 		dmapages = uvm_pagecount(&dma_constraint);
708 		if (bufcachepercent != opct) {
709 			pgs = bufcachepercent * dmapages / 100;
710 			bufadjust(pgs); /* adjust bufpages */
711 			bufhighpages = bufpages; /* set high water mark */
712 		}
713 		return(0);
714 	}
715 	case KERN_CONSDEV:
716 		if (cn_tab != NULL)
717 			dev = cn_tab->cn_dev;
718 		else
719 			dev = NODEV;
720 		return sysctl_rdstruct(oldp, oldlenp, newp, &dev, sizeof(dev));
721 	case KERN_POOL_DEBUG: {
722 		int old_pool_debug = pool_debug;
723 
724 		error = sysctl_int(oldp, oldlenp, newp, newlen,
725 		    &pool_debug);
726 		if (error == 0 && pool_debug != old_pool_debug)
727 			pool_reclaim_all();
728 		return (error);
729 	}
730 #if NPF > 0
731 	case KERN_PFSTATUS:
732 		return (pf_sysctl(oldp, oldlenp, newp, newlen));
733 #endif
734 	case KERN_TIMEOUT_STATS:
735 		return (timeout_sysctl(oldp, oldlenp, newp, newlen));
736 	case KERN_UTC_OFFSET:
737 		return (sysctl_utc_offset(oldp, oldlenp, newp, newlen));
738 	default:
739 		return (sysctl_bounded_arr(kern_vars, nitems(kern_vars), name,
740 		    namelen, oldp, oldlenp, newp, newlen));
741 	}
742 	/* NOTREACHED */
743 }
744 
745 /*
746  * hardware related system variables.
747  */
748 char *hw_vendor, *hw_prod, *hw_uuid, *hw_serial, *hw_ver;
749 int allowpowerdown = 1;
750 int hw_power = 1;
751 
752 /* morally const values reported by sysctl_bounded_arr */
753 static int byte_order = BYTE_ORDER;
754 
755 const struct sysctl_bounded_args hw_vars[] = {
756 	{HW_NCPU, &ncpus, SYSCTL_INT_READONLY},
757 	{HW_NCPUFOUND, &ncpusfound, SYSCTL_INT_READONLY},
758 	{HW_BYTEORDER, &byte_order, SYSCTL_INT_READONLY},
759 	{HW_PAGESIZE, &uvmexp.pagesize, SYSCTL_INT_READONLY},
760 	{HW_DISKCOUNT, &disk_count, SYSCTL_INT_READONLY},
761 	{HW_POWER, &hw_power, SYSCTL_INT_READONLY},
762 };
763 
764 int
765 hw_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
766     size_t newlen, struct proc *p)
767 {
768 	extern char machine[], cpu_model[];
769 	int err;
770 
771 	/*
772 	 * all sysctl names at this level except sensors and battery
773 	 * are terminal
774 	 */
775 	if (name[0] != HW_SENSORS && name[0] != HW_BATTERY && namelen != 1)
776 		return (ENOTDIR);		/* overloaded */
777 
778 	switch (name[0]) {
779 	case HW_MACHINE:
780 		return (sysctl_rdstring(oldp, oldlenp, newp, machine));
781 	case HW_MODEL:
782 		return (sysctl_rdstring(oldp, oldlenp, newp, cpu_model));
783 	case HW_NCPUONLINE:
784 		return (sysctl_rdint(oldp, oldlenp, newp,
785 		    sysctl_hwncpuonline()));
786 	case HW_PHYSMEM:
787 		return (sysctl_rdint(oldp, oldlenp, newp, ptoa(physmem)));
788 	case HW_USERMEM:
789 		return (sysctl_rdint(oldp, oldlenp, newp,
790 		    ptoa(physmem - uvmexp.wired)));
791 	case HW_DISKNAMES:
792 	case HW_DISKSTATS:
793 	case HW_CPUSPEED:
794 #ifndef	SMALL_KERNEL
795 	case HW_SENSORS:
796 	case HW_SETPERF:
797 	case HW_PERFPOLICY:
798 	case HW_BATTERY:
799 #endif /* !SMALL_KERNEL */
800 	case HW_ALLOWPOWERDOWN:
801 	case HW_UCOMNAMES:
802 #ifdef __HAVE_CPU_TOPOLOGY
803 	case HW_SMT:
804 #endif
805 	{
806 		size_t savelen = *oldlenp;
807 		if ((err = sysctl_vslock(oldp, savelen)))
808 			return (err);
809 		err = hw_sysctl_locked(name, namelen, oldp, oldlenp,
810 		    newp, newlen, p);
811 		sysctl_vsunlock(oldp, savelen);
812 		return (err);
813 	}
814 	case HW_VENDOR:
815 		if (hw_vendor)
816 			return (sysctl_rdstring(oldp, oldlenp, newp,
817 			    hw_vendor));
818 		else
819 			return (EOPNOTSUPP);
820 	case HW_PRODUCT:
821 		if (hw_prod)
822 			return (sysctl_rdstring(oldp, oldlenp, newp, hw_prod));
823 		else
824 			return (EOPNOTSUPP);
825 	case HW_VERSION:
826 		if (hw_ver)
827 			return (sysctl_rdstring(oldp, oldlenp, newp, hw_ver));
828 		else
829 			return (EOPNOTSUPP);
830 	case HW_SERIALNO:
831 		if (hw_serial)
832 			return (sysctl_rdstring(oldp, oldlenp, newp,
833 			    hw_serial));
834 		else
835 			return (EOPNOTSUPP);
836 	case HW_UUID:
837 		if (hw_uuid)
838 			return (sysctl_rdstring(oldp, oldlenp, newp, hw_uuid));
839 		else
840 			return (EOPNOTSUPP);
841 	case HW_PHYSMEM64:
842 		return (sysctl_rdquad(oldp, oldlenp, newp,
843 		    ptoa((psize_t)physmem)));
844 	case HW_USERMEM64:
845 		return (sysctl_rdquad(oldp, oldlenp, newp,
846 		    ptoa((psize_t)physmem - uvmexp.wired)));
847 	default:
848 		return sysctl_bounded_arr(hw_vars, nitems(hw_vars), name,
849 		    namelen, oldp, oldlenp, newp, newlen);
850 	}
851 	/* NOTREACHED */
852 }
853 
854 int
855 hw_sysctl_locked(int *name, u_int namelen, void *oldp, size_t *oldlenp,
856     void *newp, size_t newlen, struct proc *p)
857 {
858 	int err, cpuspeed;
859 
860 	switch (name[0]) {
861 	case HW_DISKNAMES:
862 		err = sysctl_diskinit(0, p);
863 		if (err)
864 			return err;
865 		if (disknames)
866 			return (sysctl_rdstring(oldp, oldlenp, newp,
867 			    disknames));
868 		else
869 			return (sysctl_rdstring(oldp, oldlenp, newp, ""));
870 	case HW_DISKSTATS:
871 		err = sysctl_diskinit(1, p);
872 		if (err)
873 			return err;
874 		return (sysctl_rdstruct(oldp, oldlenp, newp, diskstats,
875 		    disk_count * sizeof(struct diskstats)));
876 	case HW_CPUSPEED:
877 		if (!cpu_cpuspeed)
878 			return (EOPNOTSUPP);
879 		err = cpu_cpuspeed(&cpuspeed);
880 		if (err)
881 			return err;
882 		return (sysctl_rdint(oldp, oldlenp, newp, cpuspeed));
883 #ifndef SMALL_KERNEL
884 	case HW_SENSORS:
885 		return (sysctl_sensors(name + 1, namelen - 1, oldp, oldlenp,
886 		    newp, newlen));
887 	case HW_SETPERF:
888 		return (sysctl_hwsetperf(oldp, oldlenp, newp, newlen));
889 	case HW_PERFPOLICY:
890 		return (sysctl_hwperfpolicy(oldp, oldlenp, newp, newlen));
891 #endif /* !SMALL_KERNEL */
892 	case HW_ALLOWPOWERDOWN:
893 		return (sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
894 		    &allowpowerdown));
895 	case HW_UCOMNAMES: {
896 		const char *str = "";
897 #if NUCOM > 0
898 		str = sysctl_ucominit();
899 #endif	/* NUCOM > 0 */
900 		return (sysctl_rdstring(oldp, oldlenp, newp, str));
901 	}
902 #ifdef __HAVE_CPU_TOPOLOGY
903 	case HW_SMT:
904 		return (sysctl_hwsmt(oldp, oldlenp, newp, newlen));
905 #endif
906 #ifndef SMALL_KERNEL
907 	case HW_BATTERY:
908 		return (sysctl_hwbattery(name + 1, namelen - 1, oldp, oldlenp,
909 		    newp, newlen));
910 #endif
911 	default:
912 		return (EOPNOTSUPP);
913 	}
914 	/* NOTREACHED */
915 }
916 
917 #ifndef SMALL_KERNEL
918 
919 int hw_battery_chargemode;
920 int hw_battery_chargestart;
921 int hw_battery_chargestop;
922 int (*hw_battery_setchargemode)(int);
923 int (*hw_battery_setchargestart)(int);
924 int (*hw_battery_setchargestop)(int);
925 
926 int
927 sysctl_hwchargemode(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
928 {
929 	int mode = hw_battery_chargemode;
930 	int error;
931 
932 	if (!hw_battery_setchargemode)
933 		return EOPNOTSUPP;
934 
935 	error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
936 	    &mode, -1, 1);
937 	if (error)
938 		return error;
939 
940 	if (newp != NULL)
941 		error = hw_battery_setchargemode(mode);
942 
943 	return error;
944 }
945 
946 int
947 sysctl_hwchargestart(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
948 {
949 	int start = hw_battery_chargestart;
950 	int error;
951 
952 	if (!hw_battery_setchargestart)
953 		return EOPNOTSUPP;
954 
955 	error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
956 	    &start, 0, 100);
957 	if (error)
958 		return error;
959 
960 	if (newp != NULL)
961 		error = hw_battery_setchargestart(start);
962 
963 	return error;
964 }
965 
966 int
967 sysctl_hwchargestop(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
968 {
969 	int stop = hw_battery_chargestop;
970 	int error;
971 
972 	if (!hw_battery_setchargestop)
973 		return EOPNOTSUPP;
974 
975 	error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
976 	    &stop, 0, 100);
977 	if (error)
978 		return error;
979 
980 	if (newp != NULL)
981 		error = hw_battery_setchargestop(stop);
982 
983 	return error;
984 }
985 
986 int
987 sysctl_hwbattery(int *name, u_int namelen, void *oldp, size_t *oldlenp,
988     void *newp, size_t newlen)
989 {
990 	if (namelen != 1)
991 		return (ENOTDIR);
992 
993 	switch (name[0]) {
994 	case HW_BATTERY_CHARGEMODE:
995 		return (sysctl_hwchargemode(oldp, oldlenp, newp, newlen));
996 	case HW_BATTERY_CHARGESTART:
997 		return (sysctl_hwchargestart(oldp, oldlenp, newp, newlen));
998 	case HW_BATTERY_CHARGESTOP:
999 		return (sysctl_hwchargestop(oldp, oldlenp, newp, newlen));
1000 	default:
1001 		return (EOPNOTSUPP);
1002 	}
1003 	/* NOTREACHED */
1004 }
1005 
1006 #endif
1007 
1008 #ifdef DEBUG_SYSCTL
1009 /*
1010  * Debugging related system variables.
1011  */
1012 extern struct ctldebug debug_vfs_busyprt;
1013 struct ctldebug debug1, debug2, debug3, debug4;
1014 struct ctldebug debug5, debug6, debug7, debug8, debug9;
1015 struct ctldebug debug10, debug11, debug12, debug13, debug14;
1016 struct ctldebug debug15, debug16, debug17, debug18, debug19;
1017 static struct ctldebug *debugvars[CTL_DEBUG_MAXID] = {
1018 	&debug_vfs_busyprt,
1019 	&debug1, &debug2, &debug3, &debug4,
1020 	&debug5, &debug6, &debug7, &debug8, &debug9,
1021 	&debug10, &debug11, &debug12, &debug13, &debug14,
1022 	&debug15, &debug16, &debug17, &debug18, &debug19,
1023 };
1024 int
1025 debug_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1026     size_t newlen, struct proc *p)
1027 {
1028 	struct ctldebug *cdp;
1029 
1030 	/* all sysctl names at this level are name and field */
1031 	if (namelen != 2)
1032 		return (ENOTDIR);		/* overloaded */
1033 	if (name[0] < 0 || name[0] >= nitems(debugvars))
1034 		return (EOPNOTSUPP);
1035 	cdp = debugvars[name[0]];
1036 	if (cdp->debugname == 0)
1037 		return (EOPNOTSUPP);
1038 	switch (name[1]) {
1039 	case CTL_DEBUG_NAME:
1040 		return (sysctl_rdstring(oldp, oldlenp, newp, cdp->debugname));
1041 	case CTL_DEBUG_VALUE:
1042 		return (sysctl_int(oldp, oldlenp, newp, newlen, cdp->debugvar));
1043 	default:
1044 		return (EOPNOTSUPP);
1045 	}
1046 	/* NOTREACHED */
1047 }
1048 #endif /* DEBUG_SYSCTL */
1049 
1050 /*
1051  * Reads, or writes that lower the value
1052  */
1053 int
1054 sysctl_int_lower(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1055     int *valp)
1056 {
1057 	unsigned int oldval, newval;
1058 	int error;
1059 
1060 	if (oldp && *oldlenp < sizeof(int))
1061 		return (ENOMEM);
1062 	if (newp && newlen != sizeof(int))
1063 		return (EINVAL);
1064 	*oldlenp = sizeof(int);
1065 
1066 	if (newp) {
1067 		if ((error = copyin(newp, &newval, sizeof(int))))
1068 			return (error);
1069 		do {
1070 			oldval = atomic_load_int(valp);
1071 			if (oldval < (unsigned int)newval)
1072 				return (EPERM);	/* do not allow raising */
1073 		} while (atomic_cas_uint(valp, oldval, newval) != oldval);
1074 
1075 		if (oldp) {
1076 			/* new value has been set although user gets error */
1077 			if ((error = copyout(&oldval, oldp, sizeof(int))))
1078 				return (error);
1079 		}
1080 	} else if (oldp) {
1081 		oldval = atomic_load_int(valp);
1082 
1083 		if ((error = copyout(&oldval, oldp, sizeof(int))))
1084 			return (error);
1085 	}
1086 
1087 	return (0);
1088 }
1089 
1090 /*
1091  * Validate parameters and get old / set new parameters
1092  * for an integer-valued sysctl function.
1093  */
1094 int
1095 sysctl_int(void *oldp, size_t *oldlenp, void *newp, size_t newlen, int *valp)
1096 {
1097 	return (sysctl_int_bounded(oldp, oldlenp, newp, newlen, valp,
1098 	    INT_MIN, INT_MAX));
1099 }
1100 
1101 /*
1102  * As above, but read-only.
1103  */
1104 int
1105 sysctl_rdint(void *oldp, size_t *oldlenp, void *newp, int val)
1106 {
1107 	int error = 0;
1108 
1109 	if (oldp && *oldlenp < sizeof(int))
1110 		return (ENOMEM);
1111 	if (newp)
1112 		return (EPERM);
1113 	*oldlenp = sizeof(int);
1114 	if (oldp)
1115 		error = copyout((caddr_t)&val, oldp, sizeof(int));
1116 	return (error);
1117 }
1118 
1119 int
1120 sysctl_securelevel(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1121     struct proc *p)
1122 {
1123 	int oldval, newval;
1124 	int error;
1125 
1126 	if (oldp && *oldlenp < sizeof(int))
1127 		return (ENOMEM);
1128 	if (newp && newlen != sizeof(int))
1129 		return (EINVAL);
1130 	*oldlenp = sizeof(int);
1131 
1132 	if (newp) {
1133 		if ((error = copyin(newp, &newval, sizeof(int))))
1134 			return (error);
1135 		do {
1136 			oldval = atomic_load_int(&securelevel);
1137 			if ((oldval > 0 || newval < -1) && newval < oldval &&
1138 			    p->p_p->ps_pid != 1)
1139 				return (EPERM);
1140 		} while (atomic_cas_uint(&securelevel, oldval, newval) !=
1141 		    oldval);
1142 
1143 		if (oldp) {
1144 			/* new value has been set although user gets error */
1145 			if ((error = copyout(&oldval, oldp, sizeof(int))))
1146 				return (error);
1147 		}
1148 	} else if (oldp) {
1149 		oldval = atomic_load_int(&securelevel);
1150 
1151 		if ((error = copyout(&oldval, oldp, sizeof(int))))
1152 			return (error);
1153 	}
1154 
1155 	return (0);
1156 }
1157 
1158 /*
1159  * Selects between sysctl_rdint and sysctl_int according to securelevel.
1160  */
1161 int
1162 sysctl_securelevel_int(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1163     int *valp)
1164 {
1165 	if (atomic_load_int(&securelevel) > 0)
1166 		return (sysctl_rdint(oldp, oldlenp, newp, *valp));
1167 	return (sysctl_int(oldp, oldlenp, newp, newlen, valp));
1168 }
1169 
1170 /*
1171  * Read-only or bounded integer values.
1172  */
1173 int
1174 sysctl_int_bounded(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1175     int *valp, int minimum, int maximum)
1176 {
1177 	int oldval, newval;
1178 	int error;
1179 
1180 	/* read only */
1181 	if (newp != NULL && minimum > maximum)
1182 		return (EPERM);
1183 
1184 	if (oldp != NULL && *oldlenp < sizeof(int))
1185 		return (ENOMEM);
1186 	if (newp != NULL && newlen != sizeof(int))
1187 		return (EINVAL);
1188 	*oldlenp = sizeof(int);
1189 
1190 	/* copyin() may sleep, call it first */
1191 	if (newp != NULL) {
1192 		if ((error = copyin(newp, &newval, sizeof(int))))
1193 			return (error);
1194 		/* outside limits */
1195 		if (newval < minimum || maximum < newval)
1196 			return (EINVAL);
1197 	}
1198 	if (oldp != NULL) {
1199 		if (newp != NULL)
1200 			oldval = atomic_swap_uint(valp, newval);
1201 		else
1202 			oldval = atomic_load_int(valp);
1203 		if ((error = copyout(&oldval, oldp, sizeof(int)))) {
1204 			/* new value has been set although user gets error */
1205 			return (error);
1206 		}
1207 	} else if (newp != NULL)
1208 		atomic_store_int(valp, newval);
1209 
1210 	return (0);
1211 }
1212 
1213 /*
1214  * Array of read-only or bounded integer values.
1215  */
1216 int
1217 sysctl_bounded_arr(const struct sysctl_bounded_args *valpp, u_int valplen,
1218     int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1219     size_t newlen)
1220 {
1221 	u_int i;
1222 	if (namelen != 1)
1223 		return (ENOTDIR);
1224 	for (i = 0; i < valplen; ++i) {
1225 		if (valpp[i].mib == name[0]) {
1226 			return (sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1227 			    valpp[i].var, valpp[i].minimum, valpp[i].maximum));
1228 		}
1229 	}
1230 	return (EOPNOTSUPP);
1231 }
1232 
1233 /*
1234  * Validate parameters and get old / set new parameters
1235  * for an integer-valued sysctl function.
1236  */
1237 int
1238 sysctl_quad(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1239     int64_t *valp)
1240 {
1241 	int error = 0;
1242 
1243 	if (oldp && *oldlenp < sizeof(int64_t))
1244 		return (ENOMEM);
1245 	if (newp && newlen != sizeof(int64_t))
1246 		return (EINVAL);
1247 	*oldlenp = sizeof(int64_t);
1248 	if (oldp)
1249 		error = copyout(valp, oldp, sizeof(int64_t));
1250 	if (error == 0 && newp)
1251 		error = copyin(newp, valp, sizeof(int64_t));
1252 	return (error);
1253 }
1254 
1255 /*
1256  * As above, but read-only.
1257  */
1258 int
1259 sysctl_rdquad(void *oldp, size_t *oldlenp, void *newp, int64_t val)
1260 {
1261 	int error = 0;
1262 
1263 	if (oldp && *oldlenp < sizeof(int64_t))
1264 		return (ENOMEM);
1265 	if (newp)
1266 		return (EPERM);
1267 	*oldlenp = sizeof(int64_t);
1268 	if (oldp)
1269 		error = copyout((caddr_t)&val, oldp, sizeof(int64_t));
1270 	return (error);
1271 }
1272 
1273 /*
1274  * Validate parameters and get old / set new parameters
1275  * for a string-valued sysctl function.
1276  */
1277 int
1278 sysctl_string(void *oldp, size_t *oldlenp, void *newp, size_t newlen, char *str,
1279     size_t maxlen)
1280 {
1281 	return sysctl__string(oldp, oldlenp, newp, newlen, str, maxlen, 0);
1282 }
1283 
1284 int
1285 sysctl_tstring(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1286     char *str, size_t maxlen)
1287 {
1288 	return sysctl__string(oldp, oldlenp, newp, newlen, str, maxlen, 1);
1289 }
1290 
1291 int
1292 sysctl__string(void *oldp, size_t *oldlenp, void *newp, size_t newlen,
1293     char *str, size_t maxlen, int trunc)
1294 {
1295 	size_t len;
1296 	int error = 0;
1297 
1298 	len = strlen(str) + 1;
1299 	if (oldp && *oldlenp < len) {
1300 		if (trunc == 0 || *oldlenp == 0)
1301 			return (ENOMEM);
1302 	}
1303 	if (newp && newlen >= maxlen)
1304 		return (EINVAL);
1305 	if (oldp) {
1306 		if (trunc && *oldlenp < len) {
1307 			len = *oldlenp;
1308 			error = copyout(str, oldp, len - 1);
1309 			if (error == 0)
1310 				error = copyout("", (char *)oldp + len - 1, 1);
1311 		} else {
1312 			error = copyout(str, oldp, len);
1313 		}
1314 	}
1315 	*oldlenp = len;
1316 	if (error == 0 && newp) {
1317 		error = copyin(newp, str, newlen);
1318 		str[newlen] = 0;
1319 	}
1320 	return (error);
1321 }
1322 
1323 /*
1324  * As above, but read-only.
1325  */
1326 int
1327 sysctl_rdstring(void *oldp, size_t *oldlenp, void *newp, const char *str)
1328 {
1329 	size_t len;
1330 	int error = 0;
1331 
1332 	len = strlen(str) + 1;
1333 	if (oldp && *oldlenp < len)
1334 		return (ENOMEM);
1335 	if (newp)
1336 		return (EPERM);
1337 	*oldlenp = len;
1338 	if (oldp)
1339 		error = copyout(str, oldp, len);
1340 	return (error);
1341 }
1342 
1343 /*
1344  * Validate parameters and get old / set new parameters
1345  * for a structure oriented sysctl function.
1346  */
1347 int
1348 sysctl_struct(void *oldp, size_t *oldlenp, void *newp, size_t newlen, void *sp,
1349     size_t len)
1350 {
1351 	int error = 0;
1352 
1353 	if (oldp && *oldlenp < len)
1354 		return (ENOMEM);
1355 	if (newp && newlen > len)
1356 		return (EINVAL);
1357 	if (oldp) {
1358 		*oldlenp = len;
1359 		error = copyout(sp, oldp, len);
1360 	}
1361 	if (error == 0 && newp)
1362 		error = copyin(newp, sp, len);
1363 	return (error);
1364 }
1365 
1366 /*
1367  * Validate parameters and get old parameters
1368  * for a structure oriented sysctl function.
1369  */
1370 int
1371 sysctl_rdstruct(void *oldp, size_t *oldlenp, void *newp, const void *sp,
1372     size_t len)
1373 {
1374 	int error = 0;
1375 
1376 	if (oldp && *oldlenp < len)
1377 		return (ENOMEM);
1378 	if (newp)
1379 		return (EPERM);
1380 	*oldlenp = len;
1381 	if (oldp)
1382 		error = copyout(sp, oldp, len);
1383 	return (error);
1384 }
1385 
1386 #ifndef SMALL_KERNEL
1387 void
1388 fill_file(struct kinfo_file *kf, struct file *fp, struct filedesc *fdp,
1389 	  int fd, struct vnode *vp, struct process *pr, struct proc *p,
1390 	  struct socket *so, int show_pointers)
1391 {
1392 	struct vattr va;
1393 
1394 	memset(kf, 0, sizeof(*kf));
1395 
1396 	kf->fd_fd = fd;		/* might not really be an fd */
1397 
1398 	if (fp != NULL) {
1399 		if (show_pointers)
1400 			kf->f_fileaddr = PTRTOINT64(fp);
1401 		kf->f_flag = fp->f_flag;
1402 		kf->f_iflags = fp->f_iflags;
1403 		kf->f_type = fp->f_type;
1404 		kf->f_count = fp->f_count;
1405 		if (show_pointers)
1406 			kf->f_ucred = PTRTOINT64(fp->f_cred);
1407 		kf->f_uid = fp->f_cred->cr_uid;
1408 		kf->f_gid = fp->f_cred->cr_gid;
1409 		if (show_pointers)
1410 			kf->f_ops = PTRTOINT64(fp->f_ops);
1411 		if (show_pointers)
1412 			kf->f_data = PTRTOINT64(fp->f_data);
1413 		kf->f_usecount = 0;
1414 
1415 		if (suser(p) == 0 || p->p_ucred->cr_uid == fp->f_cred->cr_uid) {
1416 			mtx_enter(&fp->f_mtx);
1417 			kf->f_offset = fp->f_offset;
1418 			kf->f_rxfer = fp->f_rxfer;
1419 			kf->f_rwfer = fp->f_wxfer;
1420 			kf->f_seek = fp->f_seek;
1421 			kf->f_rbytes = fp->f_rbytes;
1422 			kf->f_wbytes = fp->f_wbytes;
1423 			mtx_leave(&fp->f_mtx);
1424 		} else
1425 			kf->f_offset = -1;
1426 	} else if (vp != NULL) {
1427 		/* fake it */
1428 		kf->f_type = DTYPE_VNODE;
1429 		kf->f_flag = FREAD;
1430 		if (fd == KERN_FILE_TRACE)
1431 			kf->f_flag |= FWRITE;
1432 	} else if (so != NULL) {
1433 		/* fake it */
1434 		kf->f_type = DTYPE_SOCKET;
1435 	}
1436 
1437 	/* information about the object associated with this file */
1438 	switch (kf->f_type) {
1439 	case DTYPE_VNODE:
1440 		if (fp != NULL)
1441 			vp = (struct vnode *)fp->f_data;
1442 
1443 		if (show_pointers)
1444 			kf->v_un = PTRTOINT64(vp->v_un.vu_socket);
1445 		kf->v_type = vp->v_type;
1446 		kf->v_tag = vp->v_tag;
1447 		kf->v_flag = vp->v_flag;
1448 		if (show_pointers)
1449 			kf->v_data = PTRTOINT64(vp->v_data);
1450 		if (show_pointers)
1451 			kf->v_mount = PTRTOINT64(vp->v_mount);
1452 		if (vp->v_mount)
1453 			strlcpy(kf->f_mntonname,
1454 			    vp->v_mount->mnt_stat.f_mntonname,
1455 			    sizeof(kf->f_mntonname));
1456 
1457 		if (VOP_GETATTR(vp, &va, p->p_ucred, p) == 0) {
1458 			kf->va_fileid = va.va_fileid;
1459 			kf->va_mode = MAKEIMODE(va.va_type, va.va_mode);
1460 			kf->va_size = va.va_size;
1461 			kf->va_rdev = va.va_rdev;
1462 			kf->va_fsid = va.va_fsid & 0xffffffff;
1463 			kf->va_nlink = va.va_nlink;
1464 		}
1465 		break;
1466 
1467 	case DTYPE_SOCKET: {
1468 		int locked = 0;
1469 
1470 		if (so == NULL) {
1471 			so = (struct socket *)fp->f_data;
1472 			/* if so is passed as parameter it is already locked */
1473 			solock(so);
1474 			locked = 1;
1475 		}
1476 
1477 		kf->so_type = so->so_type;
1478 		kf->so_state = so->so_state | so->so_snd.sb_state |
1479 		    so->so_rcv.sb_state;
1480 		if (show_pointers)
1481 			kf->so_pcb = PTRTOINT64(so->so_pcb);
1482 		else
1483 			kf->so_pcb = -1;
1484 		kf->so_protocol = so->so_proto->pr_protocol;
1485 		kf->so_family = so->so_proto->pr_domain->dom_family;
1486 		kf->so_rcv_cc = so->so_rcv.sb_cc;
1487 		kf->so_snd_cc = so->so_snd.sb_cc;
1488 		if (isspliced(so)) {
1489 			if (show_pointers)
1490 				kf->so_splice =
1491 				    PTRTOINT64(so->so_sp->ssp_socket);
1492 			kf->so_splicelen = so->so_sp->ssp_len;
1493 		} else if (issplicedback(so))
1494 			kf->so_splicelen = -1;
1495 		if (so->so_pcb == NULL) {
1496 			if (locked)
1497 				sounlock(so);
1498 			break;
1499 		}
1500 		switch (kf->so_family) {
1501 		case AF_INET: {
1502 			struct inpcb *inpcb = so->so_pcb;
1503 
1504 			soassertlocked(so);
1505 			if (show_pointers)
1506 				kf->inp_ppcb = PTRTOINT64(inpcb->inp_ppcb);
1507 			kf->inp_lport = inpcb->inp_lport;
1508 			kf->inp_laddru[0] = inpcb->inp_laddr.s_addr;
1509 			kf->inp_fport = inpcb->inp_fport;
1510 			kf->inp_faddru[0] = inpcb->inp_faddr.s_addr;
1511 			kf->inp_rtableid = inpcb->inp_rtableid;
1512 			if (so->so_type == SOCK_RAW)
1513 				kf->inp_proto = inpcb->inp_ip.ip_p;
1514 			if (so->so_proto->pr_protocol == IPPROTO_TCP) {
1515 				struct tcpcb *tcpcb = (void *)inpcb->inp_ppcb;
1516 				kf->t_rcv_wnd = tcpcb->rcv_wnd;
1517 				kf->t_snd_wnd = tcpcb->snd_wnd;
1518 				kf->t_snd_cwnd = tcpcb->snd_cwnd;
1519 				kf->t_state = tcpcb->t_state;
1520 			}
1521 			break;
1522 		    }
1523 		case AF_INET6: {
1524 			struct inpcb *inpcb = so->so_pcb;
1525 
1526 			soassertlocked(so);
1527 			if (show_pointers)
1528 				kf->inp_ppcb = PTRTOINT64(inpcb->inp_ppcb);
1529 			kf->inp_lport = inpcb->inp_lport;
1530 			kf->inp_laddru[0] = inpcb->inp_laddr6.s6_addr32[0];
1531 			kf->inp_laddru[1] = inpcb->inp_laddr6.s6_addr32[1];
1532 			kf->inp_laddru[2] = inpcb->inp_laddr6.s6_addr32[2];
1533 			kf->inp_laddru[3] = inpcb->inp_laddr6.s6_addr32[3];
1534 			kf->inp_fport = inpcb->inp_fport;
1535 			kf->inp_faddru[0] = inpcb->inp_faddr6.s6_addr32[0];
1536 			kf->inp_faddru[1] = inpcb->inp_faddr6.s6_addr32[1];
1537 			kf->inp_faddru[2] = inpcb->inp_faddr6.s6_addr32[2];
1538 			kf->inp_faddru[3] = inpcb->inp_faddr6.s6_addr32[3];
1539 			kf->inp_rtableid = inpcb->inp_rtableid;
1540 			if (so->so_type == SOCK_RAW)
1541 				kf->inp_proto = inpcb->inp_ipv6.ip6_nxt;
1542 			if (so->so_proto->pr_protocol == IPPROTO_TCP) {
1543 				struct tcpcb *tcpcb = (void *)inpcb->inp_ppcb;
1544 				kf->t_rcv_wnd = tcpcb->rcv_wnd;
1545 				kf->t_snd_wnd = tcpcb->snd_wnd;
1546 				kf->t_state = tcpcb->t_state;
1547 			}
1548 			break;
1549 		    }
1550 		case AF_UNIX: {
1551 			struct unpcb *unpcb = so->so_pcb;
1552 
1553 			kf->f_msgcount = unpcb->unp_msgcount;
1554 			if (show_pointers) {
1555 				kf->unp_conn	= PTRTOINT64(unpcb->unp_conn);
1556 				kf->unp_refs	= PTRTOINT64(
1557 				    SLIST_FIRST(&unpcb->unp_refs));
1558 				kf->unp_nextref	= PTRTOINT64(
1559 				    SLIST_NEXT(unpcb, unp_nextref));
1560 				kf->v_un	= PTRTOINT64(unpcb->unp_vnode);
1561 				kf->unp_addr	= PTRTOINT64(unpcb->unp_addr);
1562 			}
1563 			if (unpcb->unp_addr != NULL) {
1564 				struct sockaddr_un *un = mtod(unpcb->unp_addr,
1565 				    struct sockaddr_un *);
1566 				memcpy(kf->unp_path, un->sun_path, un->sun_len
1567 				    - offsetof(struct sockaddr_un,sun_path));
1568 			}
1569 			break;
1570 		    }
1571 		}
1572 		if (locked)
1573 			sounlock(so);
1574 		break;
1575 	    }
1576 
1577 	case DTYPE_PIPE: {
1578 		struct pipe *pipe = (struct pipe *)fp->f_data;
1579 
1580 		if (show_pointers)
1581 			kf->pipe_peer = PTRTOINT64(pipe->pipe_peer);
1582 		kf->pipe_state = pipe->pipe_state;
1583 		break;
1584 	    }
1585 
1586 	case DTYPE_KQUEUE: {
1587 		struct kqueue *kqi = (struct kqueue *)fp->f_data;
1588 
1589 		kf->kq_count = kqi->kq_count;
1590 		kf->kq_state = kqi->kq_state;
1591 		break;
1592 	    }
1593 	}
1594 
1595 	/* per-process information for KERN_FILE_BY[PU]ID */
1596 	if (pr != NULL) {
1597 		kf->p_pid = pr->ps_pid;
1598 		kf->p_uid = pr->ps_ucred->cr_uid;
1599 		kf->p_gid = pr->ps_ucred->cr_gid;
1600 		kf->p_tid = -1;
1601 		strlcpy(kf->p_comm, pr->ps_comm, sizeof(kf->p_comm));
1602 	}
1603 	if (fdp != NULL) {
1604 		fdplock(fdp);
1605 		kf->fd_ofileflags = fdp->fd_ofileflags[fd];
1606 		fdpunlock(fdp);
1607 	}
1608 }
1609 
1610 /*
1611  * Get file structures.
1612  */
1613 int
1614 sysctl_file(int *name, u_int namelen, char *where, size_t *sizep,
1615     struct proc *p)
1616 {
1617 	struct kinfo_file *kf;
1618 	struct filedesc *fdp;
1619 	struct file *fp;
1620 	struct process *pr;
1621 	size_t buflen, elem_size, elem_count, outsize;
1622 	char *dp = where;
1623 	int arg, i, error = 0, needed = 0, matched;
1624 	u_int op;
1625 	int show_pointers;
1626 
1627 	if (namelen > 4)
1628 		return (ENOTDIR);
1629 	if (namelen < 4 || name[2] > sizeof(*kf))
1630 		return (EINVAL);
1631 
1632 	buflen = where != NULL ? *sizep : 0;
1633 	op = name[0];
1634 	arg = name[1];
1635 	elem_size = name[2];
1636 	elem_count = name[3];
1637 	outsize = MIN(sizeof(*kf), elem_size);
1638 
1639 	if (elem_size < 1)
1640 		return (EINVAL);
1641 
1642 	show_pointers = suser(curproc) == 0;
1643 
1644 	kf = malloc(sizeof(*kf), M_TEMP, M_WAITOK);
1645 
1646 #define FILLIT2(fp, fdp, i, vp, pr, so) do {				\
1647 	if (buflen >= elem_size && elem_count > 0) {			\
1648 		fill_file(kf, fp, fdp, i, vp, pr, p, so, show_pointers);\
1649 		error = copyout(kf, dp, outsize);			\
1650 		if (error)						\
1651 			break;						\
1652 		dp += elem_size;					\
1653 		buflen -= elem_size;					\
1654 		elem_count--;						\
1655 	}								\
1656 	needed += elem_size;						\
1657 } while (0)
1658 #define FILLIT(fp, fdp, i, vp, pr) \
1659 	FILLIT2(fp, fdp, i, vp, pr, NULL)
1660 #define FILLSO(so) \
1661 	FILLIT2(NULL, NULL, 0, NULL, NULL, so)
1662 
1663 	switch (op) {
1664 	case KERN_FILE_BYFILE:
1665 		/* use the inp-tables to pick up closed connections, too */
1666 		if (arg == DTYPE_SOCKET) {
1667 			struct inpcb *inp;
1668 
1669 			NET_LOCK();
1670 			mtx_enter(&tcbtable.inpt_mtx);
1671 			TAILQ_FOREACH(inp, &tcbtable.inpt_queue, inp_queue)
1672 				FILLSO(inp->inp_socket);
1673 			mtx_leave(&tcbtable.inpt_mtx);
1674 #ifdef INET6
1675 			mtx_enter(&tcb6table.inpt_mtx);
1676 			TAILQ_FOREACH(inp, &tcb6table.inpt_queue, inp_queue)
1677 				FILLSO(inp->inp_socket);
1678 			mtx_leave(&tcb6table.inpt_mtx);
1679 #endif
1680 			mtx_enter(&udbtable.inpt_mtx);
1681 			TAILQ_FOREACH(inp, &udbtable.inpt_queue, inp_queue)
1682 				FILLSO(inp->inp_socket);
1683 			mtx_leave(&udbtable.inpt_mtx);
1684 #ifdef INET6
1685 			mtx_enter(&udb6table.inpt_mtx);
1686 			TAILQ_FOREACH(inp, &udb6table.inpt_queue, inp_queue)
1687 				FILLSO(inp->inp_socket);
1688 			mtx_leave(&udb6table.inpt_mtx);
1689 #endif
1690 			mtx_enter(&rawcbtable.inpt_mtx);
1691 			TAILQ_FOREACH(inp, &rawcbtable.inpt_queue, inp_queue)
1692 				FILLSO(inp->inp_socket);
1693 			mtx_leave(&rawcbtable.inpt_mtx);
1694 #ifdef INET6
1695 			mtx_enter(&rawin6pcbtable.inpt_mtx);
1696 			TAILQ_FOREACH(inp, &rawin6pcbtable.inpt_queue,
1697 			    inp_queue)
1698 				FILLSO(inp->inp_socket);
1699 			mtx_leave(&rawin6pcbtable.inpt_mtx);
1700 #endif
1701 			NET_UNLOCK();
1702 		}
1703 		fp = NULL;
1704 		while ((fp = fd_iterfile(fp, p)) != NULL) {
1705 			if ((arg == 0 || fp->f_type == arg)) {
1706 				int af, skip = 0;
1707 				if (arg == DTYPE_SOCKET && fp->f_type == arg) {
1708 					af = ((struct socket *)fp->f_data)->
1709 					    so_proto->pr_domain->dom_family;
1710 					if (af == AF_INET || af == AF_INET6)
1711 						skip = 1;
1712 				}
1713 				if (!skip)
1714 					FILLIT(fp, NULL, 0, NULL, NULL);
1715 			}
1716 		}
1717 		break;
1718 	case KERN_FILE_BYPID:
1719 		/* A arg of -1 indicates all processes */
1720 		if (arg < -1) {
1721 			error = EINVAL;
1722 			break;
1723 		}
1724 		matched = 0;
1725 		LIST_FOREACH(pr, &allprocess, ps_list) {
1726 			/*
1727 			 * skip system, exiting, embryonic and undead
1728 			 * processes
1729 			 */
1730 			if (pr->ps_flags & (PS_SYSTEM | PS_EMBRYO | PS_EXITING))
1731 				continue;
1732 			if (arg >= 0 && pr->ps_pid != (pid_t)arg) {
1733 				/* not the pid we are looking for */
1734 				continue;
1735 			}
1736 
1737 			refcnt_take(&pr->ps_refcnt);
1738 
1739 			matched = 1;
1740 			fdp = pr->ps_fd;
1741 			if (pr->ps_textvp)
1742 				FILLIT(NULL, NULL, KERN_FILE_TEXT, pr->ps_textvp, pr);
1743 			if (fdp->fd_cdir)
1744 				FILLIT(NULL, NULL, KERN_FILE_CDIR, fdp->fd_cdir, pr);
1745 			if (fdp->fd_rdir)
1746 				FILLIT(NULL, NULL, KERN_FILE_RDIR, fdp->fd_rdir, pr);
1747 			if (pr->ps_tracevp)
1748 				FILLIT(NULL, NULL, KERN_FILE_TRACE, pr->ps_tracevp, pr);
1749 			for (i = 0; i < fdp->fd_nfiles; i++) {
1750 				if ((fp = fd_getfile(fdp, i)) == NULL)
1751 					continue;
1752 				FILLIT(fp, fdp, i, NULL, pr);
1753 				FRELE(fp, p);
1754 			}
1755 
1756 			refcnt_rele_wake(&pr->ps_refcnt);
1757 
1758 			/* pid is unique, stop searching */
1759 			if (arg >= 0)
1760 				break;
1761 		}
1762 		if (!matched)
1763 			error = ESRCH;
1764 		break;
1765 	case KERN_FILE_BYUID:
1766 		LIST_FOREACH(pr, &allprocess, ps_list) {
1767 			/*
1768 			 * skip system, exiting, embryonic and undead
1769 			 * processes
1770 			 */
1771 			if (pr->ps_flags & (PS_SYSTEM | PS_EMBRYO | PS_EXITING))
1772 				continue;
1773 			if (arg >= 0 && pr->ps_ucred->cr_uid != (uid_t)arg) {
1774 				/* not the uid we are looking for */
1775 				continue;
1776 			}
1777 
1778 			refcnt_take(&pr->ps_refcnt);
1779 
1780 			fdp = pr->ps_fd;
1781 			if (fdp->fd_cdir)
1782 				FILLIT(NULL, NULL, KERN_FILE_CDIR, fdp->fd_cdir, pr);
1783 			if (fdp->fd_rdir)
1784 				FILLIT(NULL, NULL, KERN_FILE_RDIR, fdp->fd_rdir, pr);
1785 			if (pr->ps_tracevp)
1786 				FILLIT(NULL, NULL, KERN_FILE_TRACE, pr->ps_tracevp, pr);
1787 			for (i = 0; i < fdp->fd_nfiles; i++) {
1788 				if ((fp = fd_getfile(fdp, i)) == NULL)
1789 					continue;
1790 				FILLIT(fp, fdp, i, NULL, pr);
1791 				FRELE(fp, p);
1792 			}
1793 
1794 			refcnt_rele_wake(&pr->ps_refcnt);
1795 		}
1796 		break;
1797 	default:
1798 		error = EINVAL;
1799 		break;
1800 	}
1801 	free(kf, M_TEMP, sizeof(*kf));
1802 
1803 	if (!error) {
1804 		if (where == NULL)
1805 			needed += KERN_FILESLOP * elem_size;
1806 		else if (*sizep < needed)
1807 			error = ENOMEM;
1808 		*sizep = needed;
1809 	}
1810 
1811 	return (error);
1812 }
1813 
1814 /*
1815  * try over estimating by 5 procs
1816  */
1817 #define KERN_PROCSLOP	5
1818 
1819 int
1820 sysctl_doproc(int *name, u_int namelen, char *where, size_t *sizep)
1821 {
1822 	struct kinfo_proc *kproc = NULL;
1823 	struct proc *p;
1824 	struct process *pr;
1825 	char *dp;
1826 	int arg, buflen, doingzomb, elem_size, elem_count;
1827 	int error, needed, op;
1828 	int dothreads = 0;
1829 	int show_pointers;
1830 
1831 	dp = where;
1832 	buflen = where != NULL ? *sizep : 0;
1833 	needed = error = 0;
1834 
1835 	if (namelen != 4 || name[2] <= 0 || name[3] < 0 ||
1836 	    name[2] > sizeof(*kproc))
1837 		return (EINVAL);
1838 	op = name[0];
1839 	arg = name[1];
1840 	elem_size = name[2];
1841 	elem_count = name[3];
1842 
1843 	dothreads = op & KERN_PROC_SHOW_THREADS;
1844 	op &= ~KERN_PROC_SHOW_THREADS;
1845 
1846 	show_pointers = suser(curproc) == 0;
1847 
1848 	if (where != NULL)
1849 		kproc = malloc(sizeof(*kproc), M_TEMP, M_WAITOK);
1850 
1851 	pr = LIST_FIRST(&allprocess);
1852 	doingzomb = 0;
1853 again:
1854 	for (; pr != NULL; pr = LIST_NEXT(pr, ps_list)) {
1855 		/* XXX skip processes in the middle of being zapped */
1856 		if (pr->ps_pgrp == NULL)
1857 			continue;
1858 
1859 		/*
1860 		 * Skip embryonic processes.
1861 		 */
1862 		if (pr->ps_flags & PS_EMBRYO)
1863 			continue;
1864 
1865 		/*
1866 		 * TODO - make more efficient (see notes below).
1867 		 */
1868 		switch (op) {
1869 
1870 		case KERN_PROC_PID:
1871 			/* could do this with just a lookup */
1872 			if (pr->ps_pid != (pid_t)arg)
1873 				continue;
1874 			break;
1875 
1876 		case KERN_PROC_PGRP:
1877 			/* could do this by traversing pgrp */
1878 			if (pr->ps_pgrp->pg_id != (pid_t)arg)
1879 				continue;
1880 			break;
1881 
1882 		case KERN_PROC_SESSION:
1883 			if (pr->ps_session->s_leader == NULL ||
1884 			    pr->ps_session->s_leader->ps_pid != (pid_t)arg)
1885 				continue;
1886 			break;
1887 
1888 		case KERN_PROC_TTY:
1889 			if ((pr->ps_flags & PS_CONTROLT) == 0 ||
1890 			    pr->ps_session->s_ttyp == NULL ||
1891 			    pr->ps_session->s_ttyp->t_dev != (dev_t)arg)
1892 				continue;
1893 			break;
1894 
1895 		case KERN_PROC_UID:
1896 			if (pr->ps_ucred->cr_uid != (uid_t)arg)
1897 				continue;
1898 			break;
1899 
1900 		case KERN_PROC_RUID:
1901 			if (pr->ps_ucred->cr_ruid != (uid_t)arg)
1902 				continue;
1903 			break;
1904 
1905 		case KERN_PROC_ALL:
1906 			if (pr->ps_flags & PS_SYSTEM)
1907 				continue;
1908 			break;
1909 
1910 		case KERN_PROC_KTHREAD:
1911 			/* no filtering */
1912 			break;
1913 
1914 		default:
1915 			error = EINVAL;
1916 			goto err;
1917 		}
1918 
1919 		if (buflen >= elem_size && elem_count > 0) {
1920 			fill_kproc(pr, kproc, NULL, show_pointers);
1921 			error = copyout(kproc, dp, elem_size);
1922 			if (error)
1923 				goto err;
1924 			dp += elem_size;
1925 			buflen -= elem_size;
1926 			elem_count--;
1927 		}
1928 		needed += elem_size;
1929 
1930 		/* Skip per-thread entries if not required by op */
1931 		if (!dothreads)
1932 			continue;
1933 
1934 		TAILQ_FOREACH(p, &pr->ps_threads, p_thr_link) {
1935 			if (buflen >= elem_size && elem_count > 0) {
1936 				fill_kproc(pr, kproc, p, show_pointers);
1937 				error = copyout(kproc, dp, elem_size);
1938 				if (error)
1939 					goto err;
1940 				dp += elem_size;
1941 				buflen -= elem_size;
1942 				elem_count--;
1943 			}
1944 			needed += elem_size;
1945 		}
1946 	}
1947 	if (doingzomb == 0) {
1948 		pr = LIST_FIRST(&zombprocess);
1949 		doingzomb++;
1950 		goto again;
1951 	}
1952 	if (where != NULL) {
1953 		*sizep = dp - where;
1954 		if (needed > *sizep) {
1955 			error = ENOMEM;
1956 			goto err;
1957 		}
1958 	} else {
1959 		needed += KERN_PROCSLOP * elem_size;
1960 		*sizep = needed;
1961 	}
1962 err:
1963 	if (kproc)
1964 		free(kproc, M_TEMP, sizeof(*kproc));
1965 	return (error);
1966 }
1967 
1968 /*
1969  * Fill in a kproc structure for the specified process.
1970  */
1971 void
1972 fill_kproc(struct process *pr, struct kinfo_proc *ki, struct proc *p,
1973     int show_pointers)
1974 {
1975 	struct session *s = pr->ps_session;
1976 	struct tty *tp;
1977 	struct vmspace *vm = pr->ps_vmspace;
1978 	struct timespec booted, st, ut, utc;
1979 	struct tusage tu;
1980 	int isthread;
1981 
1982 	isthread = p != NULL;
1983 	if (!isthread) {
1984 		p = pr->ps_mainproc;		/* XXX */
1985 		tuagg_get_process(&tu, pr);
1986 	} else
1987 		tuagg_get_proc(&tu, p);
1988 
1989 	FILL_KPROC(ki, strlcpy, p, pr, pr->ps_ucred, pr->ps_pgrp,
1990 	    p, pr, s, vm, pr->ps_limit, pr->ps_sigacts, &tu, isthread,
1991 	    show_pointers);
1992 
1993 	/* stuff that's too painful to generalize into the macros */
1994 	if (pr->ps_pptr)
1995 		ki->p_ppid = pr->ps_ppid;
1996 	if (s->s_leader)
1997 		ki->p_sid = s->s_leader->ps_pid;
1998 
1999 	if ((pr->ps_flags & PS_CONTROLT) && (tp = s->s_ttyp)) {
2000 		ki->p_tdev = tp->t_dev;
2001 		ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : -1;
2002 		if (show_pointers)
2003 			ki->p_tsess = PTRTOINT64(tp->t_session);
2004 	} else {
2005 		ki->p_tdev = NODEV;
2006 		ki->p_tpgid = -1;
2007 	}
2008 
2009 	/* fixups that can only be done in the kernel */
2010 	if ((pr->ps_flags & PS_ZOMBIE) == 0) {
2011 		if ((pr->ps_flags & PS_EMBRYO) == 0 && vm != NULL)
2012 			ki->p_vm_rssize = vm_resident_count(vm);
2013 		calctsru(&tu, &ut, &st, NULL);
2014 		ki->p_uutime_sec = ut.tv_sec;
2015 		ki->p_uutime_usec = ut.tv_nsec/1000;
2016 		ki->p_ustime_sec = st.tv_sec;
2017 		ki->p_ustime_usec = st.tv_nsec/1000;
2018 
2019 		/* Convert starting uptime to a starting UTC time. */
2020 		nanoboottime(&booted);
2021 		timespecadd(&booted, &pr->ps_start, &utc);
2022 		ki->p_ustart_sec = utc.tv_sec;
2023 		ki->p_ustart_usec = utc.tv_nsec / 1000;
2024 
2025 #ifdef MULTIPROCESSOR
2026 		if (p->p_cpu != NULL)
2027 			ki->p_cpuid = CPU_INFO_UNIT(p->p_cpu);
2028 #endif
2029 	}
2030 
2031 	/* get %cpu and schedule state: just one thread or sum of all? */
2032 	if (isthread) {
2033 		ki->p_pctcpu = p->p_pctcpu;
2034 		ki->p_stat   = p->p_stat;
2035 	} else {
2036 		ki->p_pctcpu = 0;
2037 		ki->p_stat = (pr->ps_flags & PS_ZOMBIE) ? SDEAD : SIDL;
2038 		TAILQ_FOREACH(p, &pr->ps_threads, p_thr_link) {
2039 			ki->p_pctcpu += p->p_pctcpu;
2040 			/* find best state: ONPROC > RUN > STOP > SLEEP > .. */
2041 			if (p->p_stat == SONPROC || ki->p_stat == SONPROC)
2042 				ki->p_stat = SONPROC;
2043 			else if (p->p_stat == SRUN || ki->p_stat == SRUN)
2044 				ki->p_stat = SRUN;
2045 			else if (p->p_stat == SSTOP || ki->p_stat == SSTOP)
2046 				ki->p_stat = SSTOP;
2047 			else if (p->p_stat == SSLEEP)
2048 				ki->p_stat = SSLEEP;
2049 		}
2050 	}
2051 }
2052 
2053 int
2054 sysctl_proc_args(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2055     struct proc *cp)
2056 {
2057 	struct process *vpr;
2058 	pid_t pid;
2059 	struct ps_strings pss;
2060 	struct iovec iov;
2061 	struct uio uio;
2062 	int error, cnt, op;
2063 	size_t limit;
2064 	char **rargv, **vargv;		/* reader vs. victim */
2065 	char *rarg, *varg, *buf;
2066 	struct vmspace *vm;
2067 	vaddr_t ps_strings;
2068 
2069 	if (namelen > 2)
2070 		return (ENOTDIR);
2071 	if (namelen < 2)
2072 		return (EINVAL);
2073 
2074 	pid = name[0];
2075 	op = name[1];
2076 
2077 	switch (op) {
2078 	case KERN_PROC_ARGV:
2079 	case KERN_PROC_NARGV:
2080 	case KERN_PROC_ENV:
2081 	case KERN_PROC_NENV:
2082 		break;
2083 	default:
2084 		return (EOPNOTSUPP);
2085 	}
2086 
2087 	if ((vpr = prfind(pid)) == NULL)
2088 		return (ESRCH);
2089 
2090 	if (oldp == NULL) {
2091 		if (op == KERN_PROC_NARGV || op == KERN_PROC_NENV)
2092 			*oldlenp = sizeof(int);
2093 		else
2094 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
2095 		return (0);
2096 	}
2097 
2098 	/* Either system process or exiting/zombie */
2099 	if (vpr->ps_flags & (PS_SYSTEM | PS_EXITING))
2100 		return (EINVAL);
2101 
2102 	/* Execing - danger. */
2103 	if ((vpr->ps_flags & PS_INEXEC))
2104 		return (EBUSY);
2105 
2106 	/* Only owner or root can get env */
2107 	if ((op == KERN_PROC_NENV || op == KERN_PROC_ENV) &&
2108 	    (vpr->ps_ucred->cr_uid != cp->p_ucred->cr_uid &&
2109 	    (error = suser(cp)) != 0))
2110 		return (error);
2111 
2112 	ps_strings = vpr->ps_strings;
2113 	vm = vpr->ps_vmspace;
2114 	uvmspace_addref(vm);
2115 	vpr = NULL;
2116 
2117 	buf = malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
2118 
2119 	iov.iov_base = &pss;
2120 	iov.iov_len = sizeof(pss);
2121 	uio.uio_iov = &iov;
2122 	uio.uio_iovcnt = 1;
2123 	uio.uio_offset = (off_t)ps_strings;
2124 	uio.uio_resid = sizeof(pss);
2125 	uio.uio_segflg = UIO_SYSSPACE;
2126 	uio.uio_rw = UIO_READ;
2127 	uio.uio_procp = cp;
2128 
2129 	if ((error = uvm_io(&vm->vm_map, &uio, 0)) != 0)
2130 		goto out;
2131 
2132 	if (op == KERN_PROC_NARGV) {
2133 		error = sysctl_rdint(oldp, oldlenp, NULL, pss.ps_nargvstr);
2134 		goto out;
2135 	}
2136 	if (op == KERN_PROC_NENV) {
2137 		error = sysctl_rdint(oldp, oldlenp, NULL, pss.ps_nenvstr);
2138 		goto out;
2139 	}
2140 
2141 	if (op == KERN_PROC_ARGV) {
2142 		cnt = pss.ps_nargvstr;
2143 		vargv = pss.ps_argvstr;
2144 	} else {
2145 		cnt = pss.ps_nenvstr;
2146 		vargv = pss.ps_envstr;
2147 	}
2148 
2149 	/* -1 to have space for a terminating NUL */
2150 	limit = *oldlenp - 1;
2151 	*oldlenp = 0;
2152 
2153 	rargv = oldp;
2154 
2155 	/*
2156 	 * *oldlenp - number of bytes copied out into readers buffer.
2157 	 * limit - maximal number of bytes allowed into readers buffer.
2158 	 * rarg - pointer into readers buffer where next arg will be stored.
2159 	 * rargv - pointer into readers buffer where the next rarg pointer
2160 	 *  will be stored.
2161 	 * vargv - pointer into victim address space where the next argument
2162 	 *  will be read.
2163 	 */
2164 
2165 	/* space for cnt pointers and a NULL */
2166 	rarg = (char *)(rargv + cnt + 1);
2167 	*oldlenp += (cnt + 1) * sizeof(char **);
2168 
2169 	while (cnt > 0 && *oldlenp < limit) {
2170 		size_t len, vstrlen;
2171 
2172 		/* Write to readers argv */
2173 		if ((error = copyout(&rarg, rargv, sizeof(rarg))) != 0)
2174 			goto out;
2175 
2176 		/* read the victim argv */
2177 		iov.iov_base = &varg;
2178 		iov.iov_len = sizeof(varg);
2179 		uio.uio_iov = &iov;
2180 		uio.uio_iovcnt = 1;
2181 		uio.uio_offset = (off_t)(vaddr_t)vargv;
2182 		uio.uio_resid = sizeof(varg);
2183 		uio.uio_segflg = UIO_SYSSPACE;
2184 		uio.uio_rw = UIO_READ;
2185 		uio.uio_procp = cp;
2186 		if ((error = uvm_io(&vm->vm_map, &uio, 0)) != 0)
2187 			goto out;
2188 
2189 		if (varg == NULL)
2190 			break;
2191 
2192 		/*
2193 		 * read the victim arg. We must jump through hoops to avoid
2194 		 * crossing a page boundary too much and returning an error.
2195 		 */
2196 more:
2197 		len = PAGE_SIZE - (((vaddr_t)varg) & PAGE_MASK);
2198 		/* leave space for the terminating NUL */
2199 		iov.iov_base = buf;
2200 		iov.iov_len = len;
2201 		uio.uio_iov = &iov;
2202 		uio.uio_iovcnt = 1;
2203 		uio.uio_offset = (off_t)(vaddr_t)varg;
2204 		uio.uio_resid = len;
2205 		uio.uio_segflg = UIO_SYSSPACE;
2206 		uio.uio_rw = UIO_READ;
2207 		uio.uio_procp = cp;
2208 		if ((error = uvm_io(&vm->vm_map, &uio, 0)) != 0)
2209 			goto out;
2210 
2211 		for (vstrlen = 0; vstrlen < len; vstrlen++) {
2212 			if (buf[vstrlen] == '\0')
2213 				break;
2214 		}
2215 
2216 		/* Don't overflow readers buffer. */
2217 		if (*oldlenp + vstrlen + 1 >= limit) {
2218 			error = ENOMEM;
2219 			goto out;
2220 		}
2221 
2222 		if ((error = copyout(buf, rarg, vstrlen)) != 0)
2223 			goto out;
2224 
2225 		*oldlenp += vstrlen;
2226 		rarg += vstrlen;
2227 
2228 		/* The string didn't end in this page? */
2229 		if (vstrlen == len) {
2230 			varg += vstrlen;
2231 			goto more;
2232 		}
2233 
2234 		/* End of string. Terminate it with a NUL */
2235 		buf[0] = '\0';
2236 		if ((error = copyout(buf, rarg, 1)) != 0)
2237 			goto out;
2238 		*oldlenp += 1;
2239 		rarg += 1;
2240 
2241 		vargv++;
2242 		rargv++;
2243 		cnt--;
2244 	}
2245 
2246 	if (*oldlenp >= limit) {
2247 		error = ENOMEM;
2248 		goto out;
2249 	}
2250 
2251 	/* Write the terminating null */
2252 	rarg = NULL;
2253 	error = copyout(&rarg, rargv, sizeof(rarg));
2254 
2255 out:
2256 	uvmspace_free(vm);
2257 	free(buf, M_TEMP, PAGE_SIZE);
2258 	return (error);
2259 }
2260 
2261 int
2262 sysctl_proc_cwd(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2263     struct proc *cp)
2264 {
2265 	struct process *findpr;
2266 	struct vnode *vp;
2267 	pid_t pid;
2268 	int error;
2269 	size_t lenused, len;
2270 	char *path, *bp, *bend;
2271 
2272 	if (namelen > 1)
2273 		return (ENOTDIR);
2274 	if (namelen < 1)
2275 		return (EINVAL);
2276 
2277 	pid = name[0];
2278 	if ((findpr = prfind(pid)) == NULL)
2279 		return (ESRCH);
2280 
2281 	if (oldp == NULL) {
2282 		*oldlenp = MAXPATHLEN * 4;
2283 		return (0);
2284 	}
2285 
2286 	/* Either system process or exiting/zombie */
2287 	if (findpr->ps_flags & (PS_SYSTEM | PS_EXITING))
2288 		return (EINVAL);
2289 
2290 	/* Only owner or root can get cwd */
2291 	if (findpr->ps_ucred->cr_uid != cp->p_ucred->cr_uid &&
2292 	    (error = suser(cp)) != 0)
2293 		return (error);
2294 
2295 	len = *oldlenp;
2296 	if (len > MAXPATHLEN * 4)
2297 		len = MAXPATHLEN * 4;
2298 	else if (len < 2)
2299 		return (ERANGE);
2300 	*oldlenp = 0;
2301 
2302 	/* snag a reference to the vnode before we can sleep */
2303 	vp = findpr->ps_fd->fd_cdir;
2304 	vref(vp);
2305 
2306 	path = malloc(len, M_TEMP, M_WAITOK);
2307 
2308 	bp = &path[len];
2309 	bend = bp;
2310 	*(--bp) = '\0';
2311 
2312 	/* Same as sys__getcwd */
2313 	error = vfs_getcwd_common(vp, NULL,
2314 	    &bp, path, len / 2, GETCWD_CHECK_ACCESS, cp);
2315 	if (error == 0) {
2316 		*oldlenp = lenused = bend - bp;
2317 		error = copyout(bp, oldp, lenused);
2318 	}
2319 
2320 	vrele(vp);
2321 	free(path, M_TEMP, len);
2322 
2323 	return (error);
2324 }
2325 
2326 int
2327 sysctl_proc_nobroadcastkill(int *name, u_int namelen, void *newp, size_t newlen,
2328     void *oldp, size_t *oldlenp, struct proc *cp)
2329 {
2330 	struct process *findpr;
2331 	pid_t pid;
2332 	int error, flag;
2333 
2334 	if (namelen > 1)
2335 		return (ENOTDIR);
2336 	if (namelen < 1)
2337 		return (EINVAL);
2338 
2339 	pid = name[0];
2340 	if ((findpr = prfind(pid)) == NULL)
2341 		return (ESRCH);
2342 
2343 	/* Either system process or exiting/zombie */
2344 	if (findpr->ps_flags & (PS_SYSTEM | PS_EXITING))
2345 		return (EINVAL);
2346 
2347 	/* Only root can change PS_NOBROADCASTKILL */
2348 	if (newp != NULL && (error = suser(cp)) != 0)
2349 		return (error);
2350 
2351 	/* get the PS_NOBROADCASTKILL flag */
2352 	flag = findpr->ps_flags & PS_NOBROADCASTKILL ? 1 : 0;
2353 
2354 	error = sysctl_int(oldp, oldlenp, newp, newlen, &flag);
2355 	if (error == 0 && newp) {
2356 		if (flag)
2357 			atomic_setbits_int(&findpr->ps_flags,
2358 			    PS_NOBROADCASTKILL);
2359 		else
2360 			atomic_clearbits_int(&findpr->ps_flags,
2361 			    PS_NOBROADCASTKILL);
2362 	}
2363 
2364 	return (error);
2365 }
2366 
2367 /* Arbitrary but reasonable limit for one iteration. */
2368 #define	VMMAP_MAXLEN	MAXPHYS
2369 
2370 int
2371 sysctl_proc_vmmap(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2372     struct proc *cp)
2373 {
2374 	struct process *findpr;
2375 	pid_t pid;
2376 	int error;
2377 	size_t oldlen, len;
2378 	struct kinfo_vmentry *kve, *ukve;
2379 	u_long *ustart, start;
2380 
2381 	if (namelen > 1)
2382 		return (ENOTDIR);
2383 	if (namelen < 1)
2384 		return (EINVAL);
2385 
2386 	/* Provide max buffer length as hint. */
2387 	if (oldp == NULL) {
2388 		if (oldlenp == NULL)
2389 			return (EINVAL);
2390 		else {
2391 			*oldlenp = VMMAP_MAXLEN;
2392 			return (0);
2393 		}
2394 	}
2395 
2396 	pid = name[0];
2397 	if (pid == cp->p_p->ps_pid) {
2398 		/* Self process mapping. */
2399 		findpr = cp->p_p;
2400 	} else if (pid > 0) {
2401 		if ((findpr = prfind(pid)) == NULL)
2402 			return (ESRCH);
2403 
2404 		/* Either system process or exiting/zombie */
2405 		if (findpr->ps_flags & (PS_SYSTEM | PS_EXITING))
2406 			return (EINVAL);
2407 
2408 #if 1
2409 		/* XXX Allow only root for now */
2410 		if ((error = suser(cp)) != 0)
2411 			return (error);
2412 #else
2413 		/* Only owner or root can get vmmap */
2414 		if (findpr->ps_ucred->cr_uid != cp->p_ucred->cr_uid &&
2415 		    (error = suser(cp)) != 0)
2416 			return (error);
2417 #endif
2418 	} else {
2419 		/* Only root can get kernel_map */
2420 		if ((error = suser(cp)) != 0)
2421 			return (error);
2422 		findpr = NULL;
2423 	}
2424 
2425 	/* Check the given size. */
2426 	oldlen = *oldlenp;
2427 	if (oldlen == 0 || oldlen % sizeof(*kve) != 0)
2428 		return (EINVAL);
2429 
2430 	/* Deny huge allocation. */
2431 	if (oldlen > VMMAP_MAXLEN)
2432 		return (EINVAL);
2433 
2434 	/*
2435 	 * Iterate from the given address passed as the first element's
2436 	 * kve_start via oldp.
2437 	 */
2438 	ukve = (struct kinfo_vmentry *)oldp;
2439 	ustart = &ukve->kve_start;
2440 	error = copyin(ustart, &start, sizeof(start));
2441 	if (error != 0)
2442 		return (error);
2443 
2444 	/* Allocate wired memory to not block. */
2445 	kve = malloc(oldlen, M_TEMP, M_WAITOK);
2446 
2447 	/* Set the base address and read entries. */
2448 	kve[0].kve_start = start;
2449 	len = oldlen;
2450 	error = fill_vmmap(findpr, kve, &len);
2451 	if (error != 0 && error != ENOMEM)
2452 		goto done;
2453 	if (len == 0)
2454 		goto done;
2455 
2456 	KASSERT(len <= oldlen);
2457 	KASSERT((len % sizeof(struct kinfo_vmentry)) == 0);
2458 
2459 	error = copyout(kve, oldp, len);
2460 
2461 done:
2462 	*oldlenp = len;
2463 
2464 	free(kve, M_TEMP, oldlen);
2465 
2466 	return (error);
2467 }
2468 #endif
2469 
2470 /*
2471  * Initialize disknames/diskstats for export by sysctl. If update is set,
2472  * then we simply update the disk statistics information.
2473  */
2474 int
2475 sysctl_diskinit(int update, struct proc *p)
2476 {
2477 	struct diskstats *sdk;
2478 	struct disk *dk;
2479 	const char *duid;
2480 	int error, changed = 0;
2481 
2482 	KERNEL_ASSERT_LOCKED();
2483 
2484 	if ((error = rw_enter(&sysctl_disklock, RW_WRITE|RW_INTR)) != 0)
2485 		return error;
2486 
2487 	/* Run in a loop, disks may change while malloc sleeps. */
2488 	while (disk_change) {
2489 		int tlen;
2490 
2491 		disk_change = 0;
2492 
2493 		tlen = 0;
2494 		TAILQ_FOREACH(dk, &disklist, dk_link) {
2495 			if (dk->dk_name)
2496 				tlen += strlen(dk->dk_name);
2497 			tlen += 18;	/* label uid + separators */
2498 		}
2499 		tlen++;
2500 
2501 		/*
2502 		 * The sysctl_disklock ensures that no other process can
2503 		 * allocate disknames and diskstats while our malloc sleeps.
2504 		 */
2505 		free(disknames, M_SYSCTL, disknameslen);
2506 		free(diskstats, M_SYSCTL, diskstatslen);
2507 		diskstats = NULL;
2508 		disknames = NULL;
2509 		diskstats = mallocarray(disk_count, sizeof(struct diskstats),
2510 		    M_SYSCTL, M_WAITOK|M_ZERO);
2511 		diskstatslen = disk_count * sizeof(struct diskstats);
2512 		disknames = malloc(tlen, M_SYSCTL, M_WAITOK|M_ZERO);
2513 		disknameslen = tlen;
2514 		disknames[0] = '\0';
2515 		changed = 1;
2516 	}
2517 
2518 	if (changed) {
2519 		int l;
2520 
2521 		l = 0;
2522 		sdk = diskstats;
2523 		TAILQ_FOREACH(dk, &disklist, dk_link) {
2524 			duid = NULL;
2525 			if (dk->dk_label && !duid_iszero(dk->dk_label->d_uid))
2526 				duid = duid_format(dk->dk_label->d_uid);
2527 			snprintf(disknames + l, disknameslen - l, "%s:%s,",
2528 			    dk->dk_name ? dk->dk_name : "",
2529 			    duid ? duid : "");
2530 			l += strlen(disknames + l);
2531 			strlcpy(sdk->ds_name, dk->dk_name,
2532 			    sizeof(sdk->ds_name));
2533 			mtx_enter(&dk->dk_mtx);
2534 			sdk->ds_busy = dk->dk_busy;
2535 			sdk->ds_rxfer = dk->dk_rxfer;
2536 			sdk->ds_wxfer = dk->dk_wxfer;
2537 			sdk->ds_seek = dk->dk_seek;
2538 			sdk->ds_rbytes = dk->dk_rbytes;
2539 			sdk->ds_wbytes = dk->dk_wbytes;
2540 			sdk->ds_attachtime = dk->dk_attachtime;
2541 			sdk->ds_timestamp = dk->dk_timestamp;
2542 			sdk->ds_time = dk->dk_time;
2543 			mtx_leave(&dk->dk_mtx);
2544 			sdk++;
2545 		}
2546 
2547 		/* Eliminate trailing comma */
2548 		if (l != 0)
2549 			disknames[l - 1] = '\0';
2550 	} else if (update) {
2551 		/* Just update, number of drives hasn't changed */
2552 		sdk = diskstats;
2553 		TAILQ_FOREACH(dk, &disklist, dk_link) {
2554 			strlcpy(sdk->ds_name, dk->dk_name,
2555 			    sizeof(sdk->ds_name));
2556 			mtx_enter(&dk->dk_mtx);
2557 			sdk->ds_busy = dk->dk_busy;
2558 			sdk->ds_rxfer = dk->dk_rxfer;
2559 			sdk->ds_wxfer = dk->dk_wxfer;
2560 			sdk->ds_seek = dk->dk_seek;
2561 			sdk->ds_rbytes = dk->dk_rbytes;
2562 			sdk->ds_wbytes = dk->dk_wbytes;
2563 			sdk->ds_attachtime = dk->dk_attachtime;
2564 			sdk->ds_timestamp = dk->dk_timestamp;
2565 			sdk->ds_time = dk->dk_time;
2566 			mtx_leave(&dk->dk_mtx);
2567 			sdk++;
2568 		}
2569 	}
2570 	rw_exit_write(&sysctl_disklock);
2571 	return 0;
2572 }
2573 
2574 #if defined(SYSVMSG) || defined(SYSVSEM) || defined(SYSVSHM)
2575 int
2576 sysctl_sysvipc(int *name, u_int namelen, void *where, size_t *sizep)
2577 {
2578 #ifdef SYSVSEM
2579 	struct sem_sysctl_info *semsi;
2580 #endif
2581 #ifdef SYSVSHM
2582 	struct shm_sysctl_info *shmsi;
2583 #endif
2584 	size_t infosize, dssize, tsize, buflen, bufsiz;
2585 	int i, nds, error, ret;
2586 	void *buf;
2587 
2588 	if (namelen != 1)
2589 		return (EINVAL);
2590 
2591 	buflen = *sizep;
2592 
2593 	switch (*name) {
2594 	case KERN_SYSVIPC_MSG_INFO:
2595 #ifdef SYSVMSG
2596 		return (sysctl_sysvmsg(name, namelen, where, sizep));
2597 #else
2598 		return (EOPNOTSUPP);
2599 #endif
2600 	case KERN_SYSVIPC_SEM_INFO:
2601 #ifdef SYSVSEM
2602 		infosize = sizeof(semsi->seminfo);
2603 		nds = seminfo.semmni;
2604 		dssize = sizeof(semsi->semids[0]);
2605 		break;
2606 #else
2607 		return (EOPNOTSUPP);
2608 #endif
2609 	case KERN_SYSVIPC_SHM_INFO:
2610 #ifdef SYSVSHM
2611 		infosize = sizeof(shmsi->shminfo);
2612 		nds = shminfo.shmmni;
2613 		dssize = sizeof(shmsi->shmids[0]);
2614 		break;
2615 #else
2616 		return (EOPNOTSUPP);
2617 #endif
2618 	default:
2619 		return (EINVAL);
2620 	}
2621 	tsize = infosize + (nds * dssize);
2622 
2623 	/* Return just the total size required. */
2624 	if (where == NULL) {
2625 		*sizep = tsize;
2626 		return (0);
2627 	}
2628 
2629 	/* Not enough room for even the info struct. */
2630 	if (buflen < infosize) {
2631 		*sizep = 0;
2632 		return (ENOMEM);
2633 	}
2634 	bufsiz = min(tsize, buflen);
2635 	buf = malloc(bufsiz, M_TEMP, M_WAITOK|M_ZERO);
2636 
2637 	switch (*name) {
2638 #ifdef SYSVSEM
2639 	case KERN_SYSVIPC_SEM_INFO:
2640 		semsi = (struct sem_sysctl_info *)buf;
2641 		semsi->seminfo = seminfo;
2642 		break;
2643 #endif
2644 #ifdef SYSVSHM
2645 	case KERN_SYSVIPC_SHM_INFO:
2646 		shmsi = (struct shm_sysctl_info *)buf;
2647 		shmsi->shminfo = shminfo;
2648 		break;
2649 #endif
2650 	}
2651 	buflen -= infosize;
2652 
2653 	ret = 0;
2654 	if (buflen > 0) {
2655 		/* Fill in the IPC data structures.  */
2656 		for (i = 0; i < nds; i++) {
2657 			if (buflen < dssize) {
2658 				ret = ENOMEM;
2659 				break;
2660 			}
2661 			switch (*name) {
2662 #ifdef SYSVSEM
2663 			case KERN_SYSVIPC_SEM_INFO:
2664 				if (sema[i] != NULL)
2665 					memcpy(&semsi->semids[i], sema[i],
2666 					    dssize);
2667 				else
2668 					memset(&semsi->semids[i], 0, dssize);
2669 				break;
2670 #endif
2671 #ifdef SYSVSHM
2672 			case KERN_SYSVIPC_SHM_INFO:
2673 				if (shmsegs[i] != NULL)
2674 					memcpy(&shmsi->shmids[i], shmsegs[i],
2675 					    dssize);
2676 				else
2677 					memset(&shmsi->shmids[i], 0, dssize);
2678 				break;
2679 #endif
2680 			}
2681 			buflen -= dssize;
2682 		}
2683 	}
2684 	*sizep -= buflen;
2685 	error = copyout(buf, where, *sizep);
2686 	free(buf, M_TEMP, bufsiz);
2687 	/* If copyout succeeded, use return code set earlier. */
2688 	return (error ? error : ret);
2689 }
2690 #endif /* SYSVMSG || SYSVSEM || SYSVSHM */
2691 
2692 #ifndef	SMALL_KERNEL
2693 
2694 int
2695 sysctl_intrcnt(int *name, u_int namelen, void *oldp, size_t *oldlenp)
2696 {
2697 	return (evcount_sysctl(name, namelen, oldp, oldlenp, NULL, 0));
2698 }
2699 
2700 
2701 int
2702 sysctl_sensors(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2703     void *newp, size_t newlen)
2704 {
2705 	struct ksensor *ks;
2706 	struct sensor *us;
2707 	struct ksensordev *ksd;
2708 	struct sensordev *usd;
2709 	int dev, numt, ret;
2710 	enum sensor_type type;
2711 
2712 	if (namelen != 1 && namelen != 3)
2713 		return (ENOTDIR);
2714 
2715 	dev = name[0];
2716 	if (namelen == 1) {
2717 		ret = sensordev_get(dev, &ksd);
2718 		if (ret)
2719 			return (ret);
2720 
2721 		/* Grab a copy, to clear the kernel pointers */
2722 		usd = malloc(sizeof(*usd), M_TEMP, M_WAITOK|M_ZERO);
2723 		usd->num = ksd->num;
2724 		strlcpy(usd->xname, ksd->xname, sizeof(usd->xname));
2725 		memcpy(usd->maxnumt, ksd->maxnumt, sizeof(usd->maxnumt));
2726 		usd->sensors_count = ksd->sensors_count;
2727 
2728 		ret = sysctl_rdstruct(oldp, oldlenp, newp, usd,
2729 		    sizeof(struct sensordev));
2730 
2731 		free(usd, M_TEMP, sizeof(*usd));
2732 		return (ret);
2733 	}
2734 
2735 	type = name[1];
2736 	numt = name[2];
2737 
2738 	ret = sensor_find(dev, type, numt, &ks);
2739 	if (ret)
2740 		return (ret);
2741 
2742 	/* Grab a copy, to clear the kernel pointers */
2743 	us = malloc(sizeof(*us), M_TEMP, M_WAITOK|M_ZERO);
2744 	memcpy(us->desc, ks->desc, sizeof(us->desc));
2745 	us->tv = ks->tv;
2746 	us->value = ks->value;
2747 	us->type = ks->type;
2748 	us->status = ks->status;
2749 	us->numt = ks->numt;
2750 	us->flags = ks->flags;
2751 
2752 	ret = sysctl_rdstruct(oldp, oldlenp, newp, us,
2753 	    sizeof(struct sensor));
2754 	free(us, M_TEMP, sizeof(*us));
2755 	return (ret);
2756 }
2757 #endif	/* SMALL_KERNEL */
2758 
2759 int
2760 sysctl_cptime2(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2761     void *newp, size_t newlen)
2762 {
2763 	CPU_INFO_ITERATOR cii;
2764 	struct cpu_info *ci;
2765 	int found = 0;
2766 
2767 	if (namelen != 1)
2768 		return (ENOTDIR);
2769 
2770 	CPU_INFO_FOREACH(cii, ci) {
2771 		if (name[0] == CPU_INFO_UNIT(ci)) {
2772 			found = 1;
2773 			break;
2774 		}
2775 	}
2776 	if (!found)
2777 		return (ENOENT);
2778 
2779 	return (sysctl_rdstruct(oldp, oldlenp, newp,
2780 	    &ci->ci_schedstate.spc_cp_time,
2781 	    sizeof(ci->ci_schedstate.spc_cp_time)));
2782 }
2783 
2784 #if NAUDIO > 0
2785 int
2786 sysctl_audio(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2787     void *newp, size_t newlen)
2788 {
2789 	if (namelen != 1)
2790 		return (ENOTDIR);
2791 
2792 	if (name[0] != KERN_AUDIO_RECORD)
2793 		return (ENOENT);
2794 
2795 	return (sysctl_int(oldp, oldlenp, newp, newlen, &audio_record_enable));
2796 }
2797 #endif
2798 
2799 #if NVIDEO > 0
2800 int
2801 sysctl_video(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2802     void *newp, size_t newlen)
2803 {
2804 	if (namelen != 1)
2805 		return (ENOTDIR);
2806 
2807 	if (name[0] != KERN_VIDEO_RECORD)
2808 		return (ENOENT);
2809 
2810 	return (sysctl_int(oldp, oldlenp, newp, newlen, &video_record_enable));
2811 }
2812 #endif
2813 
2814 int
2815 sysctl_cpustats(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2816     void *newp, size_t newlen)
2817 {
2818 	CPU_INFO_ITERATOR cii;
2819 	struct cpustats cs;
2820 	struct cpu_info *ci;
2821 	int found = 0;
2822 
2823 	if (namelen != 1)
2824 		return (ENOTDIR);
2825 
2826 	CPU_INFO_FOREACH(cii, ci) {
2827 		if (name[0] == CPU_INFO_UNIT(ci)) {
2828 			found = 1;
2829 			break;
2830 		}
2831 	}
2832 	if (!found)
2833 		return (ENOENT);
2834 
2835 	memset(&cs, 0, sizeof cs);
2836 	memcpy(&cs.cs_time, &ci->ci_schedstate.spc_cp_time, sizeof(cs.cs_time));
2837 	cs.cs_flags = 0;
2838 	if (cpu_is_online(ci))
2839 		cs.cs_flags |= CPUSTATS_ONLINE;
2840 
2841 	return (sysctl_rdstruct(oldp, oldlenp, newp, &cs, sizeof(cs)));
2842 }
2843 
2844 int
2845 sysctl_utc_offset(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
2846 {
2847 	struct timespec adjusted, now;
2848 	int adjustment_seconds, error, new_offset_minutes, old_offset_minutes;
2849 
2850 	old_offset_minutes = utc_offset / 60;	/* seconds -> minutes */
2851 	new_offset_minutes = old_offset_minutes;
2852 	error = sysctl_securelevel_int(oldp, oldlenp, newp, newlen,
2853 	     &new_offset_minutes);
2854 	if (error)
2855 		return error;
2856 	if (new_offset_minutes < -24 * 60 || new_offset_minutes > 24 * 60)
2857 		return EINVAL;
2858 	if (new_offset_minutes == old_offset_minutes)
2859 		return 0;
2860 
2861 	utc_offset = new_offset_minutes * 60;	/* minutes -> seconds */
2862 	adjustment_seconds = (new_offset_minutes - old_offset_minutes) * 60;
2863 
2864 	nanotime(&now);
2865 	adjusted = now;
2866 	adjusted.tv_sec -= adjustment_seconds;
2867 	tc_setrealtimeclock(&adjusted);
2868 	resettodr();
2869 
2870 	return 0;
2871 }
2872