xref: /freebsd/sys/kern/kern_jail.c (revision 7a0a89d2cb29ee2c383600fa59e42d714a6dcbcb)
1 /*-
2  * Copyright (c) 1999 Poul-Henning Kamp.
3  * Copyright (c) 2008 Bjoern A. Zeeb.
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
30 
31 #include "opt_ddb.h"
32 #include "opt_inet.h"
33 #include "opt_inet6.h"
34 #include "opt_mac.h"
35 
36 #include <sys/param.h>
37 #include <sys/types.h>
38 #include <sys/kernel.h>
39 #include <sys/systm.h>
40 #include <sys/errno.h>
41 #include <sys/sysproto.h>
42 #include <sys/malloc.h>
43 #include <sys/priv.h>
44 #include <sys/proc.h>
45 #include <sys/taskqueue.h>
46 #include <sys/fcntl.h>
47 #include <sys/jail.h>
48 #include <sys/lock.h>
49 #include <sys/mutex.h>
50 #include <sys/sx.h>
51 #include <sys/namei.h>
52 #include <sys/mount.h>
53 #include <sys/queue.h>
54 #include <sys/socket.h>
55 #include <sys/syscallsubr.h>
56 #include <sys/sysctl.h>
57 #include <sys/vnode.h>
58 #include <sys/vimage.h>
59 #include <sys/osd.h>
60 #include <net/if.h>
61 #include <netinet/in.h>
62 #ifdef DDB
63 #include <ddb/ddb.h>
64 #ifdef INET6
65 #include <netinet6/in6_var.h>
66 #endif /* INET6 */
67 #endif /* DDB */
68 
69 #include <security/mac/mac_framework.h>
70 
71 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
72 
73 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
74     "Jail rules");
75 
76 int	jail_set_hostname_allowed = 1;
77 SYSCTL_INT(_security_jail, OID_AUTO, set_hostname_allowed, CTLFLAG_RW,
78     &jail_set_hostname_allowed, 0,
79     "Processes in jail can set their hostnames");
80 
81 int	jail_socket_unixiproute_only = 1;
82 SYSCTL_INT(_security_jail, OID_AUTO, socket_unixiproute_only, CTLFLAG_RW,
83     &jail_socket_unixiproute_only, 0,
84     "Processes in jail are limited to creating UNIX/IP/route sockets only");
85 
86 int	jail_sysvipc_allowed = 0;
87 SYSCTL_INT(_security_jail, OID_AUTO, sysvipc_allowed, CTLFLAG_RW,
88     &jail_sysvipc_allowed, 0,
89     "Processes in jail can use System V IPC primitives");
90 
91 static int jail_enforce_statfs = 2;
92 SYSCTL_INT(_security_jail, OID_AUTO, enforce_statfs, CTLFLAG_RW,
93     &jail_enforce_statfs, 0,
94     "Processes in jail cannot see all mounted file systems");
95 
96 int	jail_allow_raw_sockets = 0;
97 SYSCTL_INT(_security_jail, OID_AUTO, allow_raw_sockets, CTLFLAG_RW,
98     &jail_allow_raw_sockets, 0,
99     "Prison root can create raw sockets");
100 
101 int	jail_chflags_allowed = 0;
102 SYSCTL_INT(_security_jail, OID_AUTO, chflags_allowed, CTLFLAG_RW,
103     &jail_chflags_allowed, 0,
104     "Processes in jail can alter system file flags");
105 
106 int	jail_mount_allowed = 0;
107 SYSCTL_INT(_security_jail, OID_AUTO, mount_allowed, CTLFLAG_RW,
108     &jail_mount_allowed, 0,
109     "Processes in jail can mount/unmount jail-friendly file systems");
110 
111 int	jail_max_af_ips = 255;
112 SYSCTL_INT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
113     &jail_max_af_ips, 0,
114     "Number of IP addresses a jail may have at most per address family");
115 
116 /* allprison, lastprid, and prisoncount are protected by allprison_lock. */
117 struct	prisonlist allprison;
118 struct	sx allprison_lock;
119 int	lastprid = 0;
120 int	prisoncount = 0;
121 
122 static void		 init_prison(void *);
123 static void		 prison_complete(void *context, int pending);
124 static int		 sysctl_jail_list(SYSCTL_HANDLER_ARGS);
125 #ifdef INET
126 static int		_prison_check_ip4(struct prison *, struct in_addr *);
127 #endif
128 #ifdef INET6
129 static int		_prison_check_ip6(struct prison *, struct in6_addr *);
130 #endif
131 
132 static void
133 init_prison(void *data __unused)
134 {
135 
136 	sx_init(&allprison_lock, "allprison");
137 	LIST_INIT(&allprison);
138 }
139 
140 SYSINIT(prison, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_prison, NULL);
141 
142 #ifdef INET
143 static int
144 qcmp_v4(const void *ip1, const void *ip2)
145 {
146 	in_addr_t iaa, iab;
147 
148 	/*
149 	 * We need to compare in HBO here to get the list sorted as expected
150 	 * by the result of the code.  Sorting NBO addresses gives you
151 	 * interesting results.  If you do not understand, do not try.
152 	 */
153 	iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
154 	iab = ntohl(((const struct in_addr *)ip2)->s_addr);
155 
156 	/*
157 	 * Do not simply return the difference of the two numbers, the int is
158 	 * not wide enough.
159 	 */
160 	if (iaa > iab)
161 		return (1);
162 	else if (iaa < iab)
163 		return (-1);
164 	else
165 		return (0);
166 }
167 #endif
168 
169 #ifdef INET6
170 static int
171 qcmp_v6(const void *ip1, const void *ip2)
172 {
173 	const struct in6_addr *ia6a, *ia6b;
174 	int i, rc;
175 
176 	ia6a = (const struct in6_addr *)ip1;
177 	ia6b = (const struct in6_addr *)ip2;
178 
179 	rc = 0;
180 	for (i=0; rc == 0 && i < sizeof(struct in6_addr); i++) {
181 		if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
182 			rc = 1;
183 		else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
184 			rc = -1;
185 	}
186 	return (rc);
187 }
188 #endif
189 
190 #if defined(INET) || defined(INET6)
191 static int
192 prison_check_conflicting_ips(struct prison *p)
193 {
194 	struct prison *pr;
195 	int i;
196 
197 	sx_assert(&allprison_lock, SX_LOCKED);
198 
199 	if (p->pr_ip4s == 0 && p->pr_ip6s == 0)
200 		return (0);
201 
202 	LIST_FOREACH(pr, &allprison, pr_list) {
203 		/*
204 		 * Skip 'dying' prisons to avoid problems when
205 		 * restarting multi-IP jails.
206 		 */
207 		if (pr->pr_state == PRISON_STATE_DYING)
208 			continue;
209 
210 		/*
211 		 * We permit conflicting IPs if there is no
212 		 * more than 1 IP on eeach jail.
213 		 * In case there is one duplicate on a jail with
214 		 * more than one IP stop checking and return error.
215 		 */
216 #ifdef INET
217 		if ((p->pr_ip4s >= 1 && pr->pr_ip4s > 1) ||
218 		    (p->pr_ip4s > 1 && pr->pr_ip4s >= 1)) {
219 			for (i = 0; i < p->pr_ip4s; i++) {
220 				if (_prison_check_ip4(pr, &p->pr_ip4[i]))
221 					return (EINVAL);
222 			}
223 		}
224 #endif
225 #ifdef INET6
226 		if ((p->pr_ip6s >= 1 && pr->pr_ip6s > 1) ||
227 		    (p->pr_ip6s > 1 && pr->pr_ip6s >= 1)) {
228 			for (i = 0; i < p->pr_ip6s; i++) {
229 				if (_prison_check_ip6(pr, &p->pr_ip6[i]))
230 					return (EINVAL);
231 			}
232 		}
233 #endif
234 	}
235 
236 	return (0);
237 }
238 
239 static int
240 jail_copyin_ips(struct jail *j)
241 {
242 #ifdef INET
243 	struct in_addr  *ip4;
244 #endif
245 #ifdef INET6
246 	struct in6_addr *ip6;
247 #endif
248 	int error, i;
249 
250 	/*
251 	 * Copy in addresses, check for duplicate addresses and do some
252 	 * simple 0 and broadcast checks. If users give other bogus addresses
253 	 * it is their problem.
254 	 *
255 	 * IP addresses are all sorted but ip[0] to preserve the primary IP
256 	 * address as given from userland.  This special IP is used for
257 	 * unbound outgoing connections as well for "loopback" traffic.
258 	 */
259 #ifdef INET
260 	ip4 = NULL;
261 #endif
262 #ifdef INET6
263 	ip6 = NULL;
264 #endif
265 #ifdef INET
266 	if (j->ip4s > 0) {
267 		ip4 = (struct in_addr *)malloc(j->ip4s * sizeof(struct in_addr),
268 		    M_PRISON, M_WAITOK | M_ZERO);
269 		error = copyin(j->ip4, ip4, j->ip4s * sizeof(struct in_addr));
270 		if (error)
271 			goto e_free_ip;
272 		/* Sort all but the first IPv4 address. */
273 		if (j->ip4s > 1)
274 			qsort((ip4 + 1), j->ip4s - 1,
275 			    sizeof(struct in_addr), qcmp_v4);
276 
277 		/*
278 		 * We do not have to care about byte order for these checks
279 		 * so we will do them in NBO.
280 		 */
281 		for (i=0; i<j->ip4s; i++) {
282 			if (ip4[i].s_addr == htonl(INADDR_ANY) ||
283 			    ip4[i].s_addr == htonl(INADDR_BROADCAST)) {
284 				error = EINVAL;
285 				goto e_free_ip;
286 			}
287 			if ((i+1) < j->ip4s &&
288 			    (ip4[0].s_addr == ip4[i+1].s_addr ||
289 			    ip4[i].s_addr == ip4[i+1].s_addr)) {
290 				error = EINVAL;
291 				goto e_free_ip;
292 			}
293 		}
294 
295 		j->ip4 = ip4;
296 	} else
297 		j->ip4 = NULL;
298 #endif
299 #ifdef INET6
300 	if (j->ip6s > 0) {
301 		ip6 = (struct in6_addr *)malloc(j->ip6s * sizeof(struct in6_addr),
302 		    M_PRISON, M_WAITOK | M_ZERO);
303 		error = copyin(j->ip6, ip6, j->ip6s * sizeof(struct in6_addr));
304 		if (error)
305 			goto e_free_ip;
306 		/* Sort all but the first IPv6 address. */
307 		if (j->ip6s > 1)
308 			qsort((ip6 + 1), j->ip6s - 1,
309 			    sizeof(struct in6_addr), qcmp_v6);
310 		for (i=0; i<j->ip6s; i++) {
311 			if (IN6_IS_ADDR_UNSPECIFIED(&ip6[i])) {
312 				error = EINVAL;
313 				goto e_free_ip;
314 			}
315 			if ((i+1) < j->ip6s &&
316 			    (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[i+1]) ||
317 			    IN6_ARE_ADDR_EQUAL(&ip6[i], &ip6[i+1]))) {
318 				error = EINVAL;
319 				goto e_free_ip;
320 			}
321 		}
322 
323 		j->ip6 = ip6;
324 	} else
325 		j->ip6 = NULL;
326 #endif
327 	return (0);
328 
329 e_free_ip:
330 #ifdef INET6
331 	free(ip6, M_PRISON);
332 	j->ip6 = NULL;
333 #endif
334 #ifdef INET
335 	free(ip4, M_PRISON);
336 	j->ip4 = NULL;
337 #endif
338 	return (error);
339 }
340 #endif /* INET || INET6 */
341 
342 static int
343 jail_handle_ips(struct jail *j)
344 {
345 #if defined(INET) || defined(INET6)
346 	int error;
347 #endif
348 
349 	/*
350 	 * Finish conversion for older versions, copyin and setup IPs.
351 	 */
352 	switch (j->version) {
353 	case 0:
354 	{
355 #ifdef INET
356 		/* FreeBSD single IPv4 jails. */
357 		struct in_addr *ip4;
358 
359 		if (j->ip4s == INADDR_ANY || j->ip4s == INADDR_BROADCAST)
360 			return (EINVAL);
361 		ip4 = (struct in_addr *)malloc(sizeof(struct in_addr),
362 		    M_PRISON, M_WAITOK | M_ZERO);
363 
364 		/*
365 		 * Jail version 0 still used HBO for the IPv4 address.
366 		 */
367 		ip4->s_addr = htonl(j->ip4s);
368 		j->ip4s = 1;
369 		j->ip4 = ip4;
370 		break;
371 #else
372 		return (EINVAL);
373 #endif
374 	}
375 
376 	case 1:
377 		/*
378 		 * Version 1 was used by multi-IPv4 jail implementations
379 		 * that never made it into the official kernel.
380 		 * We should never hit this here; jail() should catch it.
381 		 */
382 		return (EINVAL);
383 
384 	case 2:	/* JAIL_API_VERSION */
385 		/* FreeBSD multi-IPv4/IPv6,noIP jails. */
386 #if defined(INET) || defined(INET6)
387 #ifdef INET
388 		if (j->ip4s > jail_max_af_ips)
389 			return (EINVAL);
390 #else
391 		if (j->ip4s != 0)
392 			return (EINVAL);
393 #endif
394 #ifdef INET6
395 		if (j->ip6s > jail_max_af_ips)
396 			return (EINVAL);
397 #else
398 		if (j->ip6s != 0)
399 			return (EINVAL);
400 #endif
401 		error = jail_copyin_ips(j);
402 		if (error)
403 			return (error);
404 #endif
405 		break;
406 
407 	default:
408 		/* Sci-Fi jails are not supported, sorry. */
409 		return (EINVAL);
410 	}
411 
412 	return (0);
413 }
414 
415 
416 /*
417  * struct jail_args {
418  *	struct jail *jail;
419  * };
420  */
421 int
422 jail(struct thread *td, struct jail_args *uap)
423 {
424 	uint32_t version;
425 	int error;
426 	struct jail j;
427 
428 	error = copyin(uap->jail, &version, sizeof(uint32_t));
429 	if (error)
430 		return (error);
431 
432 	switch (version) {
433 	case 0:
434 		/* FreeBSD single IPv4 jails. */
435 	{
436 		struct jail_v0 j0;
437 
438 		bzero(&j, sizeof(struct jail));
439 		error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
440 		if (error)
441 			return (error);
442 		j.version = j0.version;
443 		j.path = j0.path;
444 		j.hostname = j0.hostname;
445 		j.ip4s = j0.ip_number;
446 		break;
447 	}
448 
449 	case 1:
450 		/*
451 		 * Version 1 was used by multi-IPv4 jail implementations
452 		 * that never made it into the official kernel.
453 		 */
454 		return (EINVAL);
455 
456 	case 2:	/* JAIL_API_VERSION */
457 		/* FreeBSD multi-IPv4/IPv6,noIP jails. */
458 		error = copyin(uap->jail, &j, sizeof(struct jail));
459 		if (error)
460 			return (error);
461 		break;
462 
463 	default:
464 		/* Sci-Fi jails are not supported, sorry. */
465 		return (EINVAL);
466 	}
467 	return (kern_jail(td, &j));
468 }
469 
470 int
471 kern_jail(struct thread *td, struct jail *j)
472 {
473 	struct nameidata nd;
474 	struct prison *pr, *tpr;
475 	struct jail_attach_args jaa;
476 	int vfslocked, error, tryprid;
477 
478 	KASSERT(j != NULL, ("%s: j is NULL", __func__));
479 
480 	/* Handle addresses - convert old structs, copyin, check IPs. */
481 	error = jail_handle_ips(j);
482 	if (error)
483 		return (error);
484 
485 	/* Allocate struct prison and fill it with life. */
486 	pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
487 	mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF);
488 	pr->pr_ref = 1;
489 	error = copyinstr(j->path, &pr->pr_path, sizeof(pr->pr_path), NULL);
490 	if (error)
491 		goto e_killmtx;
492 	NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW | LOCKLEAF, UIO_SYSSPACE,
493 	    pr->pr_path, td);
494 	error = namei(&nd);
495 	if (error)
496 		goto e_killmtx;
497 	vfslocked = NDHASGIANT(&nd);
498 	pr->pr_root = nd.ni_vp;
499 	VOP_UNLOCK(nd.ni_vp, 0);
500 	NDFREE(&nd, NDF_ONLY_PNBUF);
501 	VFS_UNLOCK_GIANT(vfslocked);
502 	error = copyinstr(j->hostname, &pr->pr_host, sizeof(pr->pr_host), NULL);
503 	if (error)
504 		goto e_dropvnref;
505 	if (j->jailname != NULL) {
506 		error = copyinstr(j->jailname, &pr->pr_name,
507 		    sizeof(pr->pr_name), NULL);
508 		if (error)
509 			goto e_dropvnref;
510 	}
511 	if (j->ip4s > 0) {
512 		pr->pr_ip4 = j->ip4;
513 		pr->pr_ip4s = j->ip4s;
514 	}
515 #ifdef INET6
516 	if (j->ip6s > 0) {
517 		pr->pr_ip6 = j->ip6;
518 		pr->pr_ip6s = j->ip6s;
519 	}
520 #endif
521 	pr->pr_linux = NULL;
522 	pr->pr_securelevel = securelevel;
523 	bzero(&pr->pr_osd, sizeof(pr->pr_osd));
524 
525 	/*
526 	 * Pre-set prison state to ALIVE upon cration.  This is needed so we
527 	 * can later attach the process to it, etc (avoiding another extra
528 	 * state for ther process of creation, complicating things).
529 	 */
530 	pr->pr_state = PRISON_STATE_ALIVE;
531 
532 	/* Allocate a dedicated cpuset for each jail. */
533 	error = cpuset_create_root(td, &pr->pr_cpuset);
534 	if (error)
535 		goto e_dropvnref;
536 
537 	sx_xlock(&allprison_lock);
538 	/* Make sure we cannot run into problems with ambiguous bind()ings. */
539 #if defined(INET) || defined(INET6)
540 	error = prison_check_conflicting_ips(pr);
541 	if (error) {
542 		sx_xunlock(&allprison_lock);
543 		goto e_dropcpuset;
544 	}
545 #endif
546 
547 	/* Determine next pr_id and add prison to allprison list. */
548 	tryprid = lastprid + 1;
549 	if (tryprid == JAIL_MAX)
550 		tryprid = 1;
551 next:
552 	LIST_FOREACH(tpr, &allprison, pr_list) {
553 		if (tpr->pr_id == tryprid) {
554 			tryprid++;
555 			if (tryprid == JAIL_MAX) {
556 				sx_xunlock(&allprison_lock);
557 				error = EAGAIN;
558 				goto e_dropcpuset;
559 			}
560 			goto next;
561 		}
562 	}
563 	pr->pr_id = jaa.jid = lastprid = tryprid;
564 	LIST_INSERT_HEAD(&allprison, pr, pr_list);
565 	prisoncount++;
566 	sx_xunlock(&allprison_lock);
567 
568 	error = jail_attach(td, &jaa);
569 	if (error)
570 		goto e_dropprref;
571 	mtx_lock(&pr->pr_mtx);
572 	pr->pr_ref--;
573 	mtx_unlock(&pr->pr_mtx);
574 	td->td_retval[0] = jaa.jid;
575 	return (0);
576 e_dropprref:
577 	sx_xlock(&allprison_lock);
578 	LIST_REMOVE(pr, pr_list);
579 	prisoncount--;
580 	sx_xunlock(&allprison_lock);
581 e_dropcpuset:
582 	cpuset_rel(pr->pr_cpuset);
583 e_dropvnref:
584 	vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
585 	vrele(pr->pr_root);
586 	VFS_UNLOCK_GIANT(vfslocked);
587 e_killmtx:
588 	mtx_destroy(&pr->pr_mtx);
589 	free(pr, M_PRISON);
590 #ifdef INET6
591 	free(j->ip6, M_PRISON);
592 #endif
593 #ifdef INET
594 	free(j->ip4, M_PRISON);
595 #endif
596 	return (error);
597 }
598 
599 /*
600  * struct jail_attach_args {
601  *	int jid;
602  * };
603  */
604 int
605 jail_attach(struct thread *td, struct jail_attach_args *uap)
606 {
607 	struct proc *p;
608 	struct ucred *newcred, *oldcred;
609 	struct prison *pr;
610 	int vfslocked, error;
611 
612 	/*
613 	 * XXX: Note that there is a slight race here if two threads
614 	 * in the same privileged process attempt to attach to two
615 	 * different jails at the same time.  It is important for
616 	 * user processes not to do this, or they might end up with
617 	 * a process root from one prison, but attached to the jail
618 	 * of another.
619 	 */
620 	error = priv_check(td, PRIV_JAIL_ATTACH);
621 	if (error)
622 		return (error);
623 
624 	p = td->td_proc;
625 	sx_slock(&allprison_lock);
626 	pr = prison_find(uap->jid);
627 	if (pr == NULL) {
628 		sx_sunlock(&allprison_lock);
629 		return (EINVAL);
630 	}
631 
632 	/*
633 	 * Do not allow a process to attach to a prison that is not
634 	 * considered to be "ALIVE".
635 	 */
636 	if (pr->pr_state != PRISON_STATE_ALIVE) {
637 		mtx_unlock(&pr->pr_mtx);
638 		sx_sunlock(&allprison_lock);
639 		return (EINVAL);
640 	}
641 	pr->pr_ref++;
642 	mtx_unlock(&pr->pr_mtx);
643 	sx_sunlock(&allprison_lock);
644 
645 	/*
646 	 * Reparent the newly attached process to this jail.
647 	 */
648 	error = cpuset_setproc_update_set(p, pr->pr_cpuset);
649 	if (error)
650 		goto e_unref;
651 
652 	vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
653 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
654 	if ((error = change_dir(pr->pr_root, td)) != 0)
655 		goto e_unlock;
656 #ifdef MAC
657 	if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
658 		goto e_unlock;
659 #endif
660 	VOP_UNLOCK(pr->pr_root, 0);
661 	change_root(pr->pr_root, td);
662 	VFS_UNLOCK_GIANT(vfslocked);
663 
664 	newcred = crget();
665 	PROC_LOCK(p);
666 	oldcred = p->p_ucred;
667 	setsugid(p);
668 	crcopy(newcred, oldcred);
669 	newcred->cr_prison = pr;
670 	p->p_ucred = newcred;
671 	prison_proc_hold(pr);
672 	PROC_UNLOCK(p);
673 	crfree(oldcred);
674 	return (0);
675 e_unlock:
676 	VOP_UNLOCK(pr->pr_root, 0);
677 	VFS_UNLOCK_GIANT(vfslocked);
678 e_unref:
679 	mtx_lock(&pr->pr_mtx);
680 	pr->pr_ref--;
681 	mtx_unlock(&pr->pr_mtx);
682 	return (error);
683 }
684 
685 /*
686  * Returns a locked prison instance, or NULL on failure.
687  */
688 struct prison *
689 prison_find(int prid)
690 {
691 	struct prison *pr;
692 
693 	sx_assert(&allprison_lock, SX_LOCKED);
694 	LIST_FOREACH(pr, &allprison, pr_list) {
695 		if (pr->pr_id == prid) {
696 			mtx_lock(&pr->pr_mtx);
697 			if (pr->pr_ref == 0) {
698 				mtx_unlock(&pr->pr_mtx);
699 				break;
700 			}
701 			return (pr);
702 		}
703 	}
704 	return (NULL);
705 }
706 
707 void
708 prison_free_locked(struct prison *pr)
709 {
710 
711 	mtx_assert(&pr->pr_mtx, MA_OWNED);
712 	pr->pr_ref--;
713 	if (pr->pr_ref == 0) {
714 		mtx_unlock(&pr->pr_mtx);
715 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
716 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
717 		return;
718 	}
719 	mtx_unlock(&pr->pr_mtx);
720 }
721 
722 void
723 prison_free(struct prison *pr)
724 {
725 
726 	mtx_lock(&pr->pr_mtx);
727 	prison_free_locked(pr);
728 }
729 
730 static void
731 prison_complete(void *context, int pending)
732 {
733 	struct prison *pr;
734 	int vfslocked;
735 
736 	pr = (struct prison *)context;
737 
738 	sx_xlock(&allprison_lock);
739 	LIST_REMOVE(pr, pr_list);
740 	prisoncount--;
741 	sx_xunlock(&allprison_lock);
742 
743 	cpuset_rel(pr->pr_cpuset);
744 
745 	/* Free all OSD associated to this jail. */
746 	osd_jail_exit(pr);
747 
748 	vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
749 	vrele(pr->pr_root);
750 	VFS_UNLOCK_GIANT(vfslocked);
751 
752 	mtx_destroy(&pr->pr_mtx);
753 	free(pr->pr_linux, M_PRISON);
754 #ifdef INET6
755 	free(pr->pr_ip6, M_PRISON);
756 #endif
757 #ifdef INET
758 	free(pr->pr_ip4, M_PRISON);
759 #endif
760 	free(pr, M_PRISON);
761 }
762 
763 void
764 prison_hold_locked(struct prison *pr)
765 {
766 
767 	mtx_assert(&pr->pr_mtx, MA_OWNED);
768 	KASSERT(pr->pr_ref > 0,
769 	    ("Trying to hold dead prison (id=%d).", pr->pr_id));
770 	pr->pr_ref++;
771 }
772 
773 void
774 prison_hold(struct prison *pr)
775 {
776 
777 	mtx_lock(&pr->pr_mtx);
778 	prison_hold_locked(pr);
779 	mtx_unlock(&pr->pr_mtx);
780 }
781 
782 void
783 prison_proc_hold(struct prison *pr)
784 {
785 
786 	mtx_lock(&pr->pr_mtx);
787 	KASSERT(pr->pr_state == PRISON_STATE_ALIVE,
788 	    ("Cannot add a process to a non-alive prison (id=%d).", pr->pr_id));
789 	pr->pr_nprocs++;
790 	mtx_unlock(&pr->pr_mtx);
791 }
792 
793 void
794 prison_proc_free(struct prison *pr)
795 {
796 
797 	mtx_lock(&pr->pr_mtx);
798 	KASSERT(pr->pr_state == PRISON_STATE_ALIVE && pr->pr_nprocs > 0,
799 	    ("Trying to kill a process in a dead prison (id=%d).", pr->pr_id));
800 	pr->pr_nprocs--;
801 	if (pr->pr_nprocs == 0)
802 		pr->pr_state = PRISON_STATE_DYING;
803 	mtx_unlock(&pr->pr_mtx);
804 }
805 
806 
807 #ifdef INET
808 /*
809  * Pass back primary IPv4 address of this jail.
810  *
811  * If not jailed return success but do not alter the address.  Caller has to
812  * make sure to intialize it correctly (INADDR_ANY).
813  *
814  * Returns 0 on success, 1 on error.  Address returned in NBO.
815  */
816 int
817 prison_getip4(struct ucred *cred, struct in_addr *ia)
818 {
819 
820 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
821 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
822 
823 	if (!jailed(cred))
824 		/* Do not change address passed in. */
825 		return (0);
826 
827 	if (cred->cr_prison->pr_ip4 == NULL)
828 		return (1);
829 
830 	ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr;
831 	return (0);
832 }
833 
834 /*
835  * Make sure our (source) address is set to something meaningful to this
836  * jail.
837  *
838  * Returns 0 on success, 1 on error.  Address passed in in NBO and returned
839  * in NBO.
840  */
841 int
842 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
843 {
844 	struct in_addr ia0;
845 
846 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
847 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
848 
849 	if (!jailed(cred))
850 		return (0);
851 	if (cred->cr_prison->pr_ip4 == NULL)
852 		return (1);
853 
854 	ia0.s_addr = ntohl(ia->s_addr);
855 	if (ia0.s_addr == INADDR_LOOPBACK) {
856 		ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr;
857 		return (0);
858 	}
859 
860 	/*
861 	 * In case there is only 1 IPv4 address, bind directly.
862 	 */
863 	if (ia0.s_addr == INADDR_ANY && cred->cr_prison->pr_ip4s == 1) {
864 		ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr;
865 		return (0);
866 	}
867 
868 	if (ia0.s_addr == INADDR_ANY || prison_check_ip4(cred, ia))
869 		return (0);
870 
871 	return (1);
872 }
873 
874 /*
875  * Rewrite destination address in case we will connect to loopback address.
876  *
877  * Returns 0 on success, 1 on error.  Address passed in in NBO and returned
878  * in NBO.
879  */
880 int
881 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
882 {
883 
884 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
885 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
886 
887 	if (!jailed(cred))
888 		return (0);
889 	if (cred->cr_prison->pr_ip4 == NULL)
890 		return (1);
891 	if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
892 		ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr;
893 		return (0);
894 	}
895 
896 	/*
897 	 * Return success because nothing had to be changed.
898 	 */
899 	return (0);
900 }
901 
902 /*
903  * Check if given address belongs to the jail referenced by cred.
904  *
905  * Returns 1 if address belongs to jail, 0 if not.  Address passed in in NBO.
906  */
907 static int
908 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
909 {
910 	int i, a, z, d;
911 
912 	if (pr->pr_ip4 == NULL)
913 		return (0);
914 
915 	/*
916 	 * Check the primary IP.
917 	 */
918 	if (pr->pr_ip4[0].s_addr == ia->s_addr)
919 		return (1);
920 
921 	/*
922 	 * All the other IPs are sorted so we can do a binary search.
923 	 */
924 	a = 0;
925 	z = pr->pr_ip4s - 2;
926 	while (a <= z) {
927 		i = (a + z) / 2;
928 		d = qcmp_v4(&pr->pr_ip4[i+1], ia);
929 		if (d > 0)
930 			z = i - 1;
931 		else if (d < 0)
932 			a = i + 1;
933 		else
934 			return (1);
935 	}
936 	return (0);
937 }
938 
939 int
940 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
941 {
942 
943 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
944 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
945 
946 	if (!jailed(cred))
947 		return (1);
948 
949 	return (_prison_check_ip4(cred->cr_prison, ia));
950 }
951 #endif
952 
953 #ifdef INET6
954 /*
955  * Pass back primary IPv6 address for this jail.
956  *
957  * If not jailed return success but do not alter the address.  Caller has to
958  * make sure to intialize it correctly (IN6ADDR_ANY_INIT).
959  *
960  * Returns 0 on success, 1 on error.
961  */
962 int
963 prison_getip6(struct ucred *cred, struct in6_addr *ia6)
964 {
965 
966 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
967 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
968 
969 	if (!jailed(cred))
970 		return (0);
971 	if (cred->cr_prison->pr_ip6 == NULL)
972 		return (1);
973 	bcopy(&cred->cr_prison->pr_ip6[0], ia6, sizeof(struct in6_addr));
974 	return (0);
975 }
976 
977 /*
978  * Make sure our (source) address is set to something meaningful to this jail.
979  *
980  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
981  * when needed while binding.
982  *
983  * Returns 0 on success, 1 on error.
984  */
985 int
986 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
987 {
988 
989 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
990 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
991 
992 	if (!jailed(cred))
993 		return (0);
994 	if (cred->cr_prison->pr_ip6 == NULL)
995 		return (1);
996 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
997 		bcopy(&cred->cr_prison->pr_ip6[0], ia6,
998 		    sizeof(struct in6_addr));
999 		return (0);
1000 	}
1001 
1002 	/*
1003 	 * In case there is only 1 IPv6 address, and v6only is true, then
1004 	 * bind directly.
1005 	 */
1006 	if (v6only != 0 && IN6_IS_ADDR_UNSPECIFIED(ia6) &&
1007 	    cred->cr_prison->pr_ip6s == 1) {
1008 		bcopy(&cred->cr_prison->pr_ip6[0], ia6,
1009 		    sizeof(struct in6_addr));
1010 		return (0);
1011 	}
1012 	if (IN6_IS_ADDR_UNSPECIFIED(ia6) || prison_check_ip6(cred, ia6))
1013 		return (0);
1014 	return (1);
1015 }
1016 
1017 /*
1018  * Rewrite destination address in case we will connect to loopback address.
1019  *
1020  * Returns 0 on success, 1 on error.
1021  */
1022 int
1023 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
1024 {
1025 
1026 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
1027 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
1028 
1029 	if (!jailed(cred))
1030 		return (0);
1031 	if (cred->cr_prison->pr_ip6 == NULL)
1032 		return (1);
1033 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
1034 		bcopy(&cred->cr_prison->pr_ip6[0], ia6,
1035 		    sizeof(struct in6_addr));
1036 		return (0);
1037 	}
1038 
1039 	/*
1040 	 * Return success because nothing had to be changed.
1041 	 */
1042 	return (0);
1043 }
1044 
1045 /*
1046  * Check if given address belongs to the jail referenced by cred.
1047  *
1048  * Returns 1 if address belongs to jail, 0 if not.
1049  */
1050 static int
1051 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
1052 {
1053 	int i, a, z, d;
1054 
1055 	if (pr->pr_ip6 == NULL)
1056 		return (0);
1057 
1058 	/*
1059 	 * Check the primary IP.
1060 	 */
1061 	if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
1062 		return (1);
1063 
1064 	/*
1065 	 * All the other IPs are sorted so we can do a binary search.
1066 	 */
1067 	a = 0;
1068 	z = pr->pr_ip6s - 2;
1069 	while (a <= z) {
1070 		i = (a + z) / 2;
1071 		d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
1072 		if (d > 0)
1073 			z = i - 1;
1074 		else if (d < 0)
1075 			a = i + 1;
1076 		else
1077 			return (1);
1078 	}
1079 	return (0);
1080 }
1081 
1082 int
1083 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
1084 {
1085 
1086 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
1087 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
1088 
1089 	if (!jailed(cred))
1090 		return (1);
1091 
1092 	return (_prison_check_ip6(cred->cr_prison, ia6));
1093 }
1094 #endif
1095 
1096 /*
1097  * Check if given address belongs to the jail referenced by cred (wrapper to
1098  * prison_check_ip[46]).
1099  *
1100  * Returns 1 if address belongs to jail, 0 if not.  IPv4 Address passed in in
1101  * NBO.
1102  */
1103 int
1104 prison_if(struct ucred *cred, struct sockaddr *sa)
1105 {
1106 #ifdef INET
1107 	struct sockaddr_in *sai;
1108 #endif
1109 #ifdef INET6
1110 	struct sockaddr_in6 *sai6;
1111 #endif
1112 	int ok;
1113 
1114 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
1115 	KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
1116 
1117 	ok = 0;
1118 	switch(sa->sa_family)
1119 	{
1120 #ifdef INET
1121 	case AF_INET:
1122 		sai = (struct sockaddr_in *)sa;
1123 		if (prison_check_ip4(cred, &sai->sin_addr))
1124 			ok = 1;
1125 		break;
1126 
1127 #endif
1128 #ifdef INET6
1129 	case AF_INET6:
1130 		sai6 = (struct sockaddr_in6 *)sa;
1131 		if (prison_check_ip6(cred, (struct in6_addr *)&sai6->sin6_addr))
1132 			ok = 1;
1133 		break;
1134 
1135 #endif
1136 	default:
1137 		if (!jail_socket_unixiproute_only)
1138 			ok = 1;
1139 	}
1140 	return (ok);
1141 }
1142 
1143 /*
1144  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
1145  */
1146 int
1147 prison_check(struct ucred *cred1, struct ucred *cred2)
1148 {
1149 
1150 	if (jailed(cred1)) {
1151 		if (!jailed(cred2))
1152 			return (ESRCH);
1153 		if (cred2->cr_prison != cred1->cr_prison)
1154 			return (ESRCH);
1155 	}
1156 
1157 	return (0);
1158 }
1159 
1160 /*
1161  * Return 1 if the passed credential is in a jail, otherwise 0.
1162  */
1163 int
1164 jailed(struct ucred *cred)
1165 {
1166 
1167 	return (cred->cr_prison != NULL);
1168 }
1169 
1170 /*
1171  * Return the correct hostname for the passed credential.
1172  */
1173 void
1174 getcredhostname(struct ucred *cred, char *buf, size_t size)
1175 {
1176 	INIT_VPROCG(cred->cr_vimage->v_procg);
1177 
1178 	if (jailed(cred)) {
1179 		mtx_lock(&cred->cr_prison->pr_mtx);
1180 		strlcpy(buf, cred->cr_prison->pr_host, size);
1181 		mtx_unlock(&cred->cr_prison->pr_mtx);
1182 	} else {
1183 		mtx_lock(&hostname_mtx);
1184 		strlcpy(buf, V_hostname, size);
1185 		mtx_unlock(&hostname_mtx);
1186 	}
1187 }
1188 
1189 /*
1190  * Determine whether the subject represented by cred can "see"
1191  * status of a mount point.
1192  * Returns: 0 for permitted, ENOENT otherwise.
1193  * XXX: This function should be called cr_canseemount() and should be
1194  *      placed in kern_prot.c.
1195  */
1196 int
1197 prison_canseemount(struct ucred *cred, struct mount *mp)
1198 {
1199 	struct prison *pr;
1200 	struct statfs *sp;
1201 	size_t len;
1202 
1203 	if (!jailed(cred) || jail_enforce_statfs == 0)
1204 		return (0);
1205 	pr = cred->cr_prison;
1206 	if (pr->pr_root->v_mount == mp)
1207 		return (0);
1208 	if (jail_enforce_statfs == 2)
1209 		return (ENOENT);
1210 	/*
1211 	 * If jail's chroot directory is set to "/" we should be able to see
1212 	 * all mount-points from inside a jail.
1213 	 * This is ugly check, but this is the only situation when jail's
1214 	 * directory ends with '/'.
1215 	 */
1216 	if (strcmp(pr->pr_path, "/") == 0)
1217 		return (0);
1218 	len = strlen(pr->pr_path);
1219 	sp = &mp->mnt_stat;
1220 	if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
1221 		return (ENOENT);
1222 	/*
1223 	 * Be sure that we don't have situation where jail's root directory
1224 	 * is "/some/path" and mount point is "/some/pathpath".
1225 	 */
1226 	if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
1227 		return (ENOENT);
1228 	return (0);
1229 }
1230 
1231 void
1232 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
1233 {
1234 	char jpath[MAXPATHLEN];
1235 	struct prison *pr;
1236 	size_t len;
1237 
1238 	if (!jailed(cred) || jail_enforce_statfs == 0)
1239 		return;
1240 	pr = cred->cr_prison;
1241 	if (prison_canseemount(cred, mp) != 0) {
1242 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1243 		strlcpy(sp->f_mntonname, "[restricted]",
1244 		    sizeof(sp->f_mntonname));
1245 		return;
1246 	}
1247 	if (pr->pr_root->v_mount == mp) {
1248 		/*
1249 		 * Clear current buffer data, so we are sure nothing from
1250 		 * the valid path left there.
1251 		 */
1252 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1253 		*sp->f_mntonname = '/';
1254 		return;
1255 	}
1256 	/*
1257 	 * If jail's chroot directory is set to "/" we should be able to see
1258 	 * all mount-points from inside a jail.
1259 	 */
1260 	if (strcmp(pr->pr_path, "/") == 0)
1261 		return;
1262 	len = strlen(pr->pr_path);
1263 	strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
1264 	/*
1265 	 * Clear current buffer data, so we are sure nothing from
1266 	 * the valid path left there.
1267 	 */
1268 	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
1269 	if (*jpath == '\0') {
1270 		/* Should never happen. */
1271 		*sp->f_mntonname = '/';
1272 	} else {
1273 		strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
1274 	}
1275 }
1276 
1277 /*
1278  * Check with permission for a specific privilege is granted within jail.  We
1279  * have a specific list of accepted privileges; the rest are denied.
1280  */
1281 int
1282 prison_priv_check(struct ucred *cred, int priv)
1283 {
1284 
1285 	if (!jailed(cred))
1286 		return (0);
1287 
1288 	switch (priv) {
1289 
1290 		/*
1291 		 * Allow ktrace privileges for root in jail.
1292 		 */
1293 	case PRIV_KTRACE:
1294 
1295 #if 0
1296 		/*
1297 		 * Allow jailed processes to configure audit identity and
1298 		 * submit audit records (login, etc).  In the future we may
1299 		 * want to further refine the relationship between audit and
1300 		 * jail.
1301 		 */
1302 	case PRIV_AUDIT_GETAUDIT:
1303 	case PRIV_AUDIT_SETAUDIT:
1304 	case PRIV_AUDIT_SUBMIT:
1305 #endif
1306 
1307 		/*
1308 		 * Allow jailed processes to manipulate process UNIX
1309 		 * credentials in any way they see fit.
1310 		 */
1311 	case PRIV_CRED_SETUID:
1312 	case PRIV_CRED_SETEUID:
1313 	case PRIV_CRED_SETGID:
1314 	case PRIV_CRED_SETEGID:
1315 	case PRIV_CRED_SETGROUPS:
1316 	case PRIV_CRED_SETREUID:
1317 	case PRIV_CRED_SETREGID:
1318 	case PRIV_CRED_SETRESUID:
1319 	case PRIV_CRED_SETRESGID:
1320 
1321 		/*
1322 		 * Jail implements visibility constraints already, so allow
1323 		 * jailed root to override uid/gid-based constraints.
1324 		 */
1325 	case PRIV_SEEOTHERGIDS:
1326 	case PRIV_SEEOTHERUIDS:
1327 
1328 		/*
1329 		 * Jail implements inter-process debugging limits already, so
1330 		 * allow jailed root various debugging privileges.
1331 		 */
1332 	case PRIV_DEBUG_DIFFCRED:
1333 	case PRIV_DEBUG_SUGID:
1334 	case PRIV_DEBUG_UNPRIV:
1335 
1336 		/*
1337 		 * Allow jail to set various resource limits and login
1338 		 * properties, and for now, exceed process resource limits.
1339 		 */
1340 	case PRIV_PROC_LIMIT:
1341 	case PRIV_PROC_SETLOGIN:
1342 	case PRIV_PROC_SETRLIMIT:
1343 
1344 		/*
1345 		 * System V and POSIX IPC privileges are granted in jail.
1346 		 */
1347 	case PRIV_IPC_READ:
1348 	case PRIV_IPC_WRITE:
1349 	case PRIV_IPC_ADMIN:
1350 	case PRIV_IPC_MSGSIZE:
1351 	case PRIV_MQ_ADMIN:
1352 
1353 		/*
1354 		 * Jail implements its own inter-process limits, so allow
1355 		 * root processes in jail to change scheduling on other
1356 		 * processes in the same jail.  Likewise for signalling.
1357 		 */
1358 	case PRIV_SCHED_DIFFCRED:
1359 	case PRIV_SCHED_CPUSET:
1360 	case PRIV_SIGNAL_DIFFCRED:
1361 	case PRIV_SIGNAL_SUGID:
1362 
1363 		/*
1364 		 * Allow jailed processes to write to sysctls marked as jail
1365 		 * writable.
1366 		 */
1367 	case PRIV_SYSCTL_WRITEJAIL:
1368 
1369 		/*
1370 		 * Allow root in jail to manage a variety of quota
1371 		 * properties.  These should likely be conditional on a
1372 		 * configuration option.
1373 		 */
1374 	case PRIV_VFS_GETQUOTA:
1375 	case PRIV_VFS_SETQUOTA:
1376 
1377 		/*
1378 		 * Since Jail relies on chroot() to implement file system
1379 		 * protections, grant many VFS privileges to root in jail.
1380 		 * Be careful to exclude mount-related and NFS-related
1381 		 * privileges.
1382 		 */
1383 	case PRIV_VFS_READ:
1384 	case PRIV_VFS_WRITE:
1385 	case PRIV_VFS_ADMIN:
1386 	case PRIV_VFS_EXEC:
1387 	case PRIV_VFS_LOOKUP:
1388 	case PRIV_VFS_BLOCKRESERVE:	/* XXXRW: Slightly surprising. */
1389 	case PRIV_VFS_CHFLAGS_DEV:
1390 	case PRIV_VFS_CHOWN:
1391 	case PRIV_VFS_CHROOT:
1392 	case PRIV_VFS_RETAINSUGID:
1393 	case PRIV_VFS_FCHROOT:
1394 	case PRIV_VFS_LINK:
1395 	case PRIV_VFS_SETGID:
1396 	case PRIV_VFS_STAT:
1397 	case PRIV_VFS_STICKYFILE:
1398 		return (0);
1399 
1400 		/*
1401 		 * Depending on the global setting, allow privilege of
1402 		 * setting system flags.
1403 		 */
1404 	case PRIV_VFS_SYSFLAGS:
1405 		if (jail_chflags_allowed)
1406 			return (0);
1407 		else
1408 			return (EPERM);
1409 
1410 		/*
1411 		 * Depending on the global setting, allow privilege of
1412 		 * mounting/unmounting file systems.
1413 		 */
1414 	case PRIV_VFS_MOUNT:
1415 	case PRIV_VFS_UNMOUNT:
1416 	case PRIV_VFS_MOUNT_NONUSER:
1417 	case PRIV_VFS_MOUNT_OWNER:
1418 		if (jail_mount_allowed)
1419 			return (0);
1420 		else
1421 			return (EPERM);
1422 
1423 		/*
1424 		 * Allow jailed root to bind reserved ports and reuse in-use
1425 		 * ports.
1426 		 */
1427 	case PRIV_NETINET_RESERVEDPORT:
1428 	case PRIV_NETINET_REUSEPORT:
1429 		return (0);
1430 
1431 		/*
1432 		 * Allow jailed root to set certian IPv4/6 (option) headers.
1433 		 */
1434 	case PRIV_NETINET_SETHDROPTS:
1435 		return (0);
1436 
1437 		/*
1438 		 * Conditionally allow creating raw sockets in jail.
1439 		 */
1440 	case PRIV_NETINET_RAW:
1441 		if (jail_allow_raw_sockets)
1442 			return (0);
1443 		else
1444 			return (EPERM);
1445 
1446 		/*
1447 		 * Since jail implements its own visibility limits on netstat
1448 		 * sysctls, allow getcred.  This allows identd to work in
1449 		 * jail.
1450 		 */
1451 	case PRIV_NETINET_GETCRED:
1452 		return (0);
1453 
1454 	default:
1455 		/*
1456 		 * In all remaining cases, deny the privilege request.  This
1457 		 * includes almost all network privileges, many system
1458 		 * configuration privileges.
1459 		 */
1460 		return (EPERM);
1461 	}
1462 }
1463 
1464 static int
1465 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
1466 {
1467 	struct xprison *xp, *sxp;
1468 	struct prison *pr;
1469 	char *p;
1470 	size_t len;
1471 	int count, error;
1472 
1473 	if (jailed(req->td->td_ucred))
1474 		return (0);
1475 
1476 	sx_slock(&allprison_lock);
1477 	if ((count = prisoncount) == 0) {
1478 		sx_sunlock(&allprison_lock);
1479 		return (0);
1480 	}
1481 
1482 	len = sizeof(*xp) * count;
1483 	LIST_FOREACH(pr, &allprison, pr_list) {
1484 #ifdef INET
1485 		len += pr->pr_ip4s * sizeof(struct in_addr);
1486 #endif
1487 #ifdef INET6
1488 		len += pr->pr_ip6s * sizeof(struct in6_addr);
1489 #endif
1490 	}
1491 
1492 	sxp = xp = malloc(len, M_TEMP, M_WAITOK | M_ZERO);
1493 
1494 	LIST_FOREACH(pr, &allprison, pr_list) {
1495 		xp->pr_version = XPRISON_VERSION;
1496 		xp->pr_id = pr->pr_id;
1497 		xp->pr_state = pr->pr_state;
1498 		xp->pr_cpusetid = pr->pr_cpuset->cs_id;
1499 		strlcpy(xp->pr_path, pr->pr_path, sizeof(xp->pr_path));
1500 		mtx_lock(&pr->pr_mtx);
1501 		strlcpy(xp->pr_host, pr->pr_host, sizeof(xp->pr_host));
1502 		strlcpy(xp->pr_name, pr->pr_name, sizeof(xp->pr_name));
1503 		mtx_unlock(&pr->pr_mtx);
1504 #ifdef INET
1505 		xp->pr_ip4s = pr->pr_ip4s;
1506 #endif
1507 #ifdef INET6
1508 		xp->pr_ip6s = pr->pr_ip6s;
1509 #endif
1510 		p = (char *)(xp + 1);
1511 #ifdef INET
1512 		if (pr->pr_ip4s > 0) {
1513 			bcopy(pr->pr_ip4, (struct in_addr *)p,
1514 			    pr->pr_ip4s * sizeof(struct in_addr));
1515 			p += (pr->pr_ip4s * sizeof(struct in_addr));
1516 		}
1517 #endif
1518 #ifdef INET6
1519 		if (pr->pr_ip6s > 0) {
1520 			bcopy(pr->pr_ip6, (struct in6_addr *)p,
1521 			    pr->pr_ip6s * sizeof(struct in6_addr));
1522 			p += (pr->pr_ip6s * sizeof(struct in6_addr));
1523 		}
1524 #endif
1525 		xp = (struct xprison *)p;
1526 	}
1527 	sx_sunlock(&allprison_lock);
1528 
1529 	error = SYSCTL_OUT(req, sxp, len);
1530 	free(sxp, M_TEMP);
1531 	return (error);
1532 }
1533 
1534 SYSCTL_OID(_security_jail, OID_AUTO, list, CTLTYPE_STRUCT | CTLFLAG_RD,
1535     NULL, 0, sysctl_jail_list, "S", "List of active jails");
1536 
1537 static int
1538 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
1539 {
1540 	int error, injail;
1541 
1542 	injail = jailed(req->td->td_ucred);
1543 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
1544 
1545 	return (error);
1546 }
1547 SYSCTL_PROC(_security_jail, OID_AUTO, jailed, CTLTYPE_INT | CTLFLAG_RD,
1548     NULL, 0, sysctl_jail_jailed, "I", "Process in jail?");
1549 
1550 #ifdef DDB
1551 DB_SHOW_COMMAND(jails, db_show_jails)
1552 {
1553 	struct prison *pr;
1554 #ifdef INET
1555 	struct in_addr ia;
1556 #endif
1557 #ifdef INET6
1558 	char ip6buf[INET6_ADDRSTRLEN];
1559 #endif
1560 	const char *state;
1561 #if defined(INET) || defined(INET6)
1562 	int i;
1563 #endif
1564 
1565 	db_printf(
1566 	    "   JID  pr_ref  pr_nprocs  pr_ip4s  pr_ip6s\n");
1567 	db_printf(
1568 	    "        Hostname                      Path\n");
1569 	db_printf(
1570 	    "        Name                          State\n");
1571 	db_printf(
1572 	    "        Cpusetid\n");
1573 	db_printf(
1574 	    "        IP Address(es)\n");
1575 	LIST_FOREACH(pr, &allprison, pr_list) {
1576 		db_printf("%6d  %6d  %9d  %7d  %7d\n",
1577 		    pr->pr_id, pr->pr_ref, pr->pr_nprocs,
1578 		    pr->pr_ip4s, pr->pr_ip6s);
1579 		db_printf("%6s  %-29.29s %.74s\n",
1580 		    "", pr->pr_host, pr->pr_path);
1581 		if (pr->pr_state < 0 || pr->pr_state >= (int)((sizeof(
1582 		    prison_states) / sizeof(struct prison_state))))
1583 			state = "(bogus)";
1584 		else
1585 			state = prison_states[pr->pr_state].state_name;
1586 		db_printf("%6s  %-29.29s %.74s\n",
1587 		    "", (pr->pr_name[0] != '\0') ? pr->pr_name : "", state);
1588 		db_printf("%6s  %-6d\n",
1589 		    "", pr->pr_cpuset->cs_id);
1590 #ifdef INET
1591 		for (i=0; i < pr->pr_ip4s; i++) {
1592 			ia.s_addr = pr->pr_ip4[i].s_addr;
1593 			db_printf("%6s  %s\n", "", inet_ntoa(ia));
1594 		}
1595 #endif
1596 #ifdef INET6
1597 		for (i=0; i < pr->pr_ip6s; i++)
1598 			db_printf("%6s  %s\n",
1599 			    "", ip6_sprintf(ip6buf, &pr->pr_ip6[i]));
1600 #endif /* INET6 */
1601 		if (db_pager_quit)
1602 			break;
1603 	}
1604 }
1605 #endif /* DDB */
1606