xref: /freebsd/sys/kern/kern_jail.c (revision ec65e4f8d0654361df5e97d4de3518edebf76b46)
1 /*-
2  * Copyright (c) 1999 Poul-Henning Kamp.
3  * Copyright (c) 2008 Bjoern A. Zeeb.
4  * Copyright (c) 2009 James Gritton.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include "opt_compat.h"
33 #include "opt_ddb.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
44 #include <sys/osd.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
49 #include <sys/jail.h>
50 #include <sys/lock.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
54 #include <sys/sx.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
63 
64 #include <net/if.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 
69 #ifdef DDB
70 #include <ddb/ddb.h>
71 #endif /* DDB */
72 
73 #include <security/mac/mac_framework.h>
74 
75 #define	DEFAULT_HOSTUUID	"00000000-0000-0000-0000-000000000000"
76 
77 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
78 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
79 
80 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
81 #ifdef INET
82 #ifdef INET6
83 #define	_PR_IP_SADDRSEL	PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
84 #else
85 #define	_PR_IP_SADDRSEL	PR_IP4_SADDRSEL
86 #endif
87 #else /* !INET */
88 #ifdef INET6
89 #define	_PR_IP_SADDRSEL	PR_IP6_SADDRSEL
90 #else
91 #define	_PR_IP_SADDRSEL	0
92 #endif
93 #endif
94 
95 /* prison0 describes what is "real" about the system. */
96 struct prison prison0 = {
97 	.pr_id		= 0,
98 	.pr_name	= "0",
99 	.pr_ref		= 1,
100 	.pr_uref	= 1,
101 	.pr_path	= "/",
102 	.pr_securelevel	= -1,
103 	.pr_devfs_rsnum = 0,
104 	.pr_childmax	= JAIL_MAX,
105 	.pr_hostuuid	= DEFAULT_HOSTUUID,
106 	.pr_children	= LIST_HEAD_INITIALIZER(prison0.pr_children),
107 #ifdef VIMAGE
108 	.pr_flags	= PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
109 #else
110 	.pr_flags	= PR_HOST|_PR_IP_SADDRSEL,
111 #endif
112 	.pr_allow	= PR_ALLOW_ALL,
113 };
114 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
115 
116 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
117 struct	sx allprison_lock;
118 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
119 struct	prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
120 LIST_HEAD(, prison_racct) allprison_racct;
121 int	lastprid = 0;
122 
123 static int do_jail_attach(struct thread *td, struct prison *pr);
124 static void prison_complete(void *context, int pending);
125 static void prison_deref(struct prison *pr, int flags);
126 static char *prison_path(struct prison *pr1, struct prison *pr2);
127 static void prison_remove_one(struct prison *pr);
128 #ifdef RACCT
129 static void prison_racct_attach(struct prison *pr);
130 static void prison_racct_modify(struct prison *pr);
131 static void prison_racct_detach(struct prison *pr);
132 #endif
133 #ifdef INET
134 static int _prison_check_ip4(const struct prison *, const struct in_addr *);
135 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
136 #endif
137 #ifdef INET6
138 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
139 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
140 #endif
141 
142 /* Flags for prison_deref */
143 #define	PD_DEREF	0x01
144 #define	PD_DEUREF	0x02
145 #define	PD_LOCKED	0x04
146 #define	PD_LIST_SLOCKED	0x08
147 #define	PD_LIST_XLOCKED	0x10
148 
149 /*
150  * Parameter names corresponding to PR_* flag values.  Size values are for kvm
151  * as we cannot figure out the size of a sparse array, or an array without a
152  * terminating entry.
153  */
154 static char *pr_flag_names[] = {
155 	[0] = "persist",
156 #ifdef INET
157 	[7] = "ip4.saddrsel",
158 #endif
159 #ifdef INET6
160 	[8] = "ip6.saddrsel",
161 #endif
162 };
163 const size_t pr_flag_names_size = sizeof(pr_flag_names);
164 
165 static char *pr_flag_nonames[] = {
166 	[0] = "nopersist",
167 #ifdef INET
168 	[7] = "ip4.nosaddrsel",
169 #endif
170 #ifdef INET6
171 	[8] = "ip6.nosaddrsel",
172 #endif
173 };
174 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
175 
176 struct jailsys_flags {
177 	const char	*name;
178 	unsigned	 disable;
179 	unsigned	 new;
180 } pr_flag_jailsys[] = {
181 	{ "host", 0, PR_HOST },
182 #ifdef VIMAGE
183 	{ "vnet", 0, PR_VNET },
184 #endif
185 #ifdef INET
186 	{ "ip4", PR_IP4_USER, PR_IP4_USER },
187 #endif
188 #ifdef INET6
189 	{ "ip6", PR_IP6_USER, PR_IP6_USER },
190 #endif
191 };
192 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
193 
194 static char *pr_allow_names[] = {
195 	"allow.set_hostname",
196 	"allow.sysvipc",
197 	"allow.raw_sockets",
198 	"allow.chflags",
199 	"allow.mount",
200 	"allow.quotas",
201 	"allow.socket_af",
202 	"allow.mount.devfs",
203 	"allow.mount.nullfs",
204 	"allow.mount.zfs",
205 	"allow.mount.procfs",
206 	"allow.mount.tmpfs",
207 	"allow.mount.fdescfs",
208 	"allow.mount.linprocfs",
209 	"allow.mount.linsysfs",
210 };
211 const size_t pr_allow_names_size = sizeof(pr_allow_names);
212 
213 static char *pr_allow_nonames[] = {
214 	"allow.noset_hostname",
215 	"allow.nosysvipc",
216 	"allow.noraw_sockets",
217 	"allow.nochflags",
218 	"allow.nomount",
219 	"allow.noquotas",
220 	"allow.nosocket_af",
221 	"allow.mount.nodevfs",
222 	"allow.mount.nonullfs",
223 	"allow.mount.nozfs",
224 	"allow.mount.noprocfs",
225 	"allow.mount.notmpfs",
226 	"allow.mount.nofdescfs",
227 	"allow.mount.nolinprocfs",
228 	"allow.mount.nolinsysfs",
229 };
230 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
231 
232 #define	JAIL_DEFAULT_ALLOW		PR_ALLOW_SET_HOSTNAME
233 #define	JAIL_DEFAULT_ENFORCE_STATFS	2
234 #define	JAIL_DEFAULT_DEVFS_RSNUM	0
235 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
236 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
237 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
238 #if defined(INET) || defined(INET6)
239 static unsigned jail_max_af_ips = 255;
240 #endif
241 
242 /*
243  * Initialize the parts of prison0 that can't be static-initialized with
244  * constants.  This is called from proc0_init() after creating thread0 cpuset.
245  */
246 void
247 prison0_init(void)
248 {
249 
250 	prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
251 	prison0.pr_osreldate = osreldate;
252 	strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
253 }
254 
255 #ifdef INET
256 static int
257 qcmp_v4(const void *ip1, const void *ip2)
258 {
259 	in_addr_t iaa, iab;
260 
261 	/*
262 	 * We need to compare in HBO here to get the list sorted as expected
263 	 * by the result of the code.  Sorting NBO addresses gives you
264 	 * interesting results.  If you do not understand, do not try.
265 	 */
266 	iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
267 	iab = ntohl(((const struct in_addr *)ip2)->s_addr);
268 
269 	/*
270 	 * Do not simply return the difference of the two numbers, the int is
271 	 * not wide enough.
272 	 */
273 	if (iaa > iab)
274 		return (1);
275 	else if (iaa < iab)
276 		return (-1);
277 	else
278 		return (0);
279 }
280 #endif
281 
282 #ifdef INET6
283 static int
284 qcmp_v6(const void *ip1, const void *ip2)
285 {
286 	const struct in6_addr *ia6a, *ia6b;
287 	int i, rc;
288 
289 	ia6a = (const struct in6_addr *)ip1;
290 	ia6b = (const struct in6_addr *)ip2;
291 
292 	rc = 0;
293 	for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
294 		if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
295 			rc = 1;
296 		else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
297 			rc = -1;
298 	}
299 	return (rc);
300 }
301 #endif
302 
303 /*
304  * struct jail_args {
305  *	struct jail *jail;
306  * };
307  */
308 int
309 sys_jail(struct thread *td, struct jail_args *uap)
310 {
311 	uint32_t version;
312 	int error;
313 	struct jail j;
314 
315 	error = copyin(uap->jail, &version, sizeof(uint32_t));
316 	if (error)
317 		return (error);
318 
319 	switch (version) {
320 	case 0:
321 	{
322 		struct jail_v0 j0;
323 
324 		/* FreeBSD single IPv4 jails. */
325 		bzero(&j, sizeof(struct jail));
326 		error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
327 		if (error)
328 			return (error);
329 		j.version = j0.version;
330 		j.path = j0.path;
331 		j.hostname = j0.hostname;
332 		j.ip4s = htonl(j0.ip_number);	/* jail_v0 is host order */
333 		break;
334 	}
335 
336 	case 1:
337 		/*
338 		 * Version 1 was used by multi-IPv4 jail implementations
339 		 * that never made it into the official kernel.
340 		 */
341 		return (EINVAL);
342 
343 	case 2:	/* JAIL_API_VERSION */
344 		/* FreeBSD multi-IPv4/IPv6,noIP jails. */
345 		error = copyin(uap->jail, &j, sizeof(struct jail));
346 		if (error)
347 			return (error);
348 		break;
349 
350 	default:
351 		/* Sci-Fi jails are not supported, sorry. */
352 		return (EINVAL);
353 	}
354 	return (kern_jail(td, &j));
355 }
356 
357 int
358 kern_jail(struct thread *td, struct jail *j)
359 {
360 	struct iovec optiov[2 * (4 + nitems(pr_allow_names)
361 #ifdef INET
362 			    + 1
363 #endif
364 #ifdef INET6
365 			    + 1
366 #endif
367 			    )];
368 	struct uio opt;
369 	char *u_path, *u_hostname, *u_name;
370 #ifdef INET
371 	uint32_t ip4s;
372 	struct in_addr *u_ip4;
373 #endif
374 #ifdef INET6
375 	struct in6_addr *u_ip6;
376 #endif
377 	size_t tmplen;
378 	int error, enforce_statfs, fi;
379 
380 	bzero(&optiov, sizeof(optiov));
381 	opt.uio_iov = optiov;
382 	opt.uio_iovcnt = 0;
383 	opt.uio_offset = -1;
384 	opt.uio_resid = -1;
385 	opt.uio_segflg = UIO_SYSSPACE;
386 	opt.uio_rw = UIO_READ;
387 	opt.uio_td = td;
388 
389 	/* Set permissions for top-level jails from sysctls. */
390 	if (!jailed(td->td_ucred)) {
391 		for (fi = 0; fi < nitems(pr_allow_names); fi++) {
392 			optiov[opt.uio_iovcnt].iov_base =
393 			    (jail_default_allow & (1 << fi))
394 			    ? pr_allow_names[fi] : pr_allow_nonames[fi];
395 			optiov[opt.uio_iovcnt].iov_len =
396 			    strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
397 			opt.uio_iovcnt += 2;
398 		}
399 		optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
400 		optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
401 		opt.uio_iovcnt++;
402 		enforce_statfs = jail_default_enforce_statfs;
403 		optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
404 		optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
405 		opt.uio_iovcnt++;
406 	}
407 
408 	tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
409 #ifdef INET
410 	ip4s = (j->version == 0) ? 1 : j->ip4s;
411 	if (ip4s > jail_max_af_ips)
412 		return (EINVAL);
413 	tmplen += ip4s * sizeof(struct in_addr);
414 #else
415 	if (j->ip4s > 0)
416 		return (EINVAL);
417 #endif
418 #ifdef INET6
419 	if (j->ip6s > jail_max_af_ips)
420 		return (EINVAL);
421 	tmplen += j->ip6s * sizeof(struct in6_addr);
422 #else
423 	if (j->ip6s > 0)
424 		return (EINVAL);
425 #endif
426 	u_path = malloc(tmplen, M_TEMP, M_WAITOK);
427 	u_hostname = u_path + MAXPATHLEN;
428 	u_name = u_hostname + MAXHOSTNAMELEN;
429 #ifdef INET
430 	u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
431 #endif
432 #ifdef INET6
433 #ifdef INET
434 	u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
435 #else
436 	u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
437 #endif
438 #endif
439 	optiov[opt.uio_iovcnt].iov_base = "path";
440 	optiov[opt.uio_iovcnt].iov_len = sizeof("path");
441 	opt.uio_iovcnt++;
442 	optiov[opt.uio_iovcnt].iov_base = u_path;
443 	error = copyinstr(j->path, u_path, MAXPATHLEN,
444 	    &optiov[opt.uio_iovcnt].iov_len);
445 	if (error) {
446 		free(u_path, M_TEMP);
447 		return (error);
448 	}
449 	opt.uio_iovcnt++;
450 	optiov[opt.uio_iovcnt].iov_base = "host.hostname";
451 	optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
452 	opt.uio_iovcnt++;
453 	optiov[opt.uio_iovcnt].iov_base = u_hostname;
454 	error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
455 	    &optiov[opt.uio_iovcnt].iov_len);
456 	if (error) {
457 		free(u_path, M_TEMP);
458 		return (error);
459 	}
460 	opt.uio_iovcnt++;
461 	if (j->jailname != NULL) {
462 		optiov[opt.uio_iovcnt].iov_base = "name";
463 		optiov[opt.uio_iovcnt].iov_len = sizeof("name");
464 		opt.uio_iovcnt++;
465 		optiov[opt.uio_iovcnt].iov_base = u_name;
466 		error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
467 		    &optiov[opt.uio_iovcnt].iov_len);
468 		if (error) {
469 			free(u_path, M_TEMP);
470 			return (error);
471 		}
472 		opt.uio_iovcnt++;
473 	}
474 #ifdef INET
475 	optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
476 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
477 	opt.uio_iovcnt++;
478 	optiov[opt.uio_iovcnt].iov_base = u_ip4;
479 	optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
480 	if (j->version == 0)
481 		u_ip4->s_addr = j->ip4s;
482 	else {
483 		error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
484 		if (error) {
485 			free(u_path, M_TEMP);
486 			return (error);
487 		}
488 	}
489 	opt.uio_iovcnt++;
490 #endif
491 #ifdef INET6
492 	optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
493 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
494 	opt.uio_iovcnt++;
495 	optiov[opt.uio_iovcnt].iov_base = u_ip6;
496 	optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
497 	error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
498 	if (error) {
499 		free(u_path, M_TEMP);
500 		return (error);
501 	}
502 	opt.uio_iovcnt++;
503 #endif
504 	KASSERT(opt.uio_iovcnt <= nitems(optiov),
505 		("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
506 	error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
507 	free(u_path, M_TEMP);
508 	return (error);
509 }
510 
511 
512 /*
513  * struct jail_set_args {
514  *	struct iovec *iovp;
515  *	unsigned int iovcnt;
516  *	int flags;
517  * };
518  */
519 int
520 sys_jail_set(struct thread *td, struct jail_set_args *uap)
521 {
522 	struct uio *auio;
523 	int error;
524 
525 	/* Check that we have an even number of iovecs. */
526 	if (uap->iovcnt & 1)
527 		return (EINVAL);
528 
529 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
530 	if (error)
531 		return (error);
532 	error = kern_jail_set(td, auio, uap->flags);
533 	free(auio, M_IOV);
534 	return (error);
535 }
536 
537 int
538 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
539 {
540 	struct nameidata nd;
541 #ifdef INET
542 	struct in_addr *ip4;
543 #endif
544 #ifdef INET6
545 	struct in6_addr *ip6;
546 #endif
547 	struct vfsopt *opt;
548 	struct vfsoptlist *opts;
549 	struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
550 	struct vnode *root;
551 	char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
552 	char *g_path, *osrelstr;
553 #if defined(INET) || defined(INET6)
554 	struct prison *tppr;
555 	void *op;
556 #endif
557 	unsigned long hid;
558 	size_t namelen, onamelen, pnamelen;
559 	int born, created, cuflags, descend, enforce;
560 	int error, errmsg_len, errmsg_pos;
561 	int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
562 	int fi, jid, jsys, len, level;
563 	int childmax, osreldt, rsnum, slevel;
564 	int fullpath_disabled;
565 #if defined(INET) || defined(INET6)
566 	int ii, ij;
567 #endif
568 #ifdef INET
569 	int ip4s, redo_ip4;
570 #endif
571 #ifdef INET6
572 	int ip6s, redo_ip6;
573 #endif
574 	uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
575 	unsigned tallow;
576 	char numbuf[12];
577 
578 	error = priv_check(td, PRIV_JAIL_SET);
579 	if (!error && (flags & JAIL_ATTACH))
580 		error = priv_check(td, PRIV_JAIL_ATTACH);
581 	if (error)
582 		return (error);
583 	mypr = td->td_ucred->cr_prison;
584 	if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
585 		return (EPERM);
586 	if (flags & ~JAIL_SET_MASK)
587 		return (EINVAL);
588 
589 	/*
590 	 * Check all the parameters before committing to anything.  Not all
591 	 * errors can be caught early, but we may as well try.  Also, this
592 	 * takes care of some expensive stuff (path lookup) before getting
593 	 * the allprison lock.
594 	 *
595 	 * XXX Jails are not filesystems, and jail parameters are not mount
596 	 *     options.  But it makes more sense to re-use the vfsopt code
597 	 *     than duplicate it under a different name.
598 	 */
599 	error = vfs_buildopts(optuio, &opts);
600 	if (error)
601 		return (error);
602 #ifdef INET
603 	ip4 = NULL;
604 #endif
605 #ifdef INET6
606 	ip6 = NULL;
607 #endif
608 	g_path = NULL;
609 
610 	cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
611 	if (!cuflags) {
612 		error = EINVAL;
613 		vfs_opterror(opts, "no valid operation (create or update)");
614 		goto done_errmsg;
615 	}
616 
617 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
618 	if (error == ENOENT)
619 		jid = 0;
620 	else if (error != 0)
621 		goto done_free;
622 
623 	error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
624 	if (error == ENOENT)
625 		gotslevel = 0;
626 	else if (error != 0)
627 		goto done_free;
628 	else
629 		gotslevel = 1;
630 
631 	error =
632 	    vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
633 	if (error == ENOENT)
634 		gotchildmax = 0;
635 	else if (error != 0)
636 		goto done_free;
637 	else
638 		gotchildmax = 1;
639 
640 	error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
641 	if (error == ENOENT)
642 		gotenforce = 0;
643 	else if (error != 0)
644 		goto done_free;
645 	else if (enforce < 0 || enforce > 2) {
646 		error = EINVAL;
647 		goto done_free;
648 	} else
649 		gotenforce = 1;
650 
651 	error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
652 	if (error == ENOENT)
653 		gotrsnum = 0;
654 	else if (error != 0)
655 		goto done_free;
656 	else
657 		gotrsnum = 1;
658 
659 	pr_flags = ch_flags = 0;
660 	for (fi = 0; fi < nitems(pr_flag_names); fi++) {
661 		if (pr_flag_names[fi] == NULL)
662 			continue;
663 		vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
664 		vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
665 	}
666 	ch_flags |= pr_flags;
667 	for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
668 		error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
669 		    sizeof(jsys));
670 		if (error == ENOENT)
671 			continue;
672 		if (error != 0)
673 			goto done_free;
674 		switch (jsys) {
675 		case JAIL_SYS_DISABLE:
676 			if (!pr_flag_jailsys[fi].disable) {
677 				error = EINVAL;
678 				goto done_free;
679 			}
680 			pr_flags |= pr_flag_jailsys[fi].disable;
681 			break;
682 		case JAIL_SYS_NEW:
683 			pr_flags |= pr_flag_jailsys[fi].new;
684 			break;
685 		case JAIL_SYS_INHERIT:
686 			break;
687 		default:
688 			error = EINVAL;
689 			goto done_free;
690 		}
691 		ch_flags |=
692 		    pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
693 	}
694 	if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
695 	    && !(pr_flags & PR_PERSIST)) {
696 		error = EINVAL;
697 		vfs_opterror(opts, "new jail must persist or attach");
698 		goto done_errmsg;
699 	}
700 #ifdef VIMAGE
701 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
702 		error = EINVAL;
703 		vfs_opterror(opts, "vnet cannot be changed after creation");
704 		goto done_errmsg;
705 	}
706 #endif
707 #ifdef INET
708 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
709 		error = EINVAL;
710 		vfs_opterror(opts, "ip4 cannot be changed after creation");
711 		goto done_errmsg;
712 	}
713 #endif
714 #ifdef INET6
715 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
716 		error = EINVAL;
717 		vfs_opterror(opts, "ip6 cannot be changed after creation");
718 		goto done_errmsg;
719 	}
720 #endif
721 
722 	pr_allow = ch_allow = 0;
723 	for (fi = 0; fi < nitems(pr_allow_names); fi++) {
724 		vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
725 		vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
726 	}
727 	ch_allow |= pr_allow;
728 
729 	error = vfs_getopt(opts, "name", (void **)&name, &len);
730 	if (error == ENOENT)
731 		name = NULL;
732 	else if (error != 0)
733 		goto done_free;
734 	else {
735 		if (len == 0 || name[len - 1] != '\0') {
736 			error = EINVAL;
737 			goto done_free;
738 		}
739 		if (len > MAXHOSTNAMELEN) {
740 			error = ENAMETOOLONG;
741 			goto done_free;
742 		}
743 	}
744 
745 	error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
746 	if (error == ENOENT)
747 		host = NULL;
748 	else if (error != 0)
749 		goto done_free;
750 	else {
751 		ch_flags |= PR_HOST;
752 		pr_flags |= PR_HOST;
753 		if (len == 0 || host[len - 1] != '\0') {
754 			error = EINVAL;
755 			goto done_free;
756 		}
757 		if (len > MAXHOSTNAMELEN) {
758 			error = ENAMETOOLONG;
759 			goto done_free;
760 		}
761 	}
762 
763 	error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
764 	if (error == ENOENT)
765 		domain = NULL;
766 	else if (error != 0)
767 		goto done_free;
768 	else {
769 		ch_flags |= PR_HOST;
770 		pr_flags |= PR_HOST;
771 		if (len == 0 || domain[len - 1] != '\0') {
772 			error = EINVAL;
773 			goto done_free;
774 		}
775 		if (len > MAXHOSTNAMELEN) {
776 			error = ENAMETOOLONG;
777 			goto done_free;
778 		}
779 	}
780 
781 	error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
782 	if (error == ENOENT)
783 		uuid = NULL;
784 	else if (error != 0)
785 		goto done_free;
786 	else {
787 		ch_flags |= PR_HOST;
788 		pr_flags |= PR_HOST;
789 		if (len == 0 || uuid[len - 1] != '\0') {
790 			error = EINVAL;
791 			goto done_free;
792 		}
793 		if (len > HOSTUUIDLEN) {
794 			error = ENAMETOOLONG;
795 			goto done_free;
796 		}
797 	}
798 
799 #ifdef COMPAT_FREEBSD32
800 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
801 		uint32_t hid32;
802 
803 		error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
804 		hid = hid32;
805 	} else
806 #endif
807 		error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
808 	if (error == ENOENT)
809 		gothid = 0;
810 	else if (error != 0)
811 		goto done_free;
812 	else {
813 		gothid = 1;
814 		ch_flags |= PR_HOST;
815 		pr_flags |= PR_HOST;
816 	}
817 
818 #ifdef INET
819 	error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
820 	if (error == ENOENT)
821 		ip4s = 0;
822 	else if (error != 0)
823 		goto done_free;
824 	else if (ip4s & (sizeof(*ip4) - 1)) {
825 		error = EINVAL;
826 		goto done_free;
827 	} else {
828 		ch_flags |= PR_IP4_USER;
829 		pr_flags |= PR_IP4_USER;
830 		if (ip4s > 0) {
831 			ip4s /= sizeof(*ip4);
832 			if (ip4s > jail_max_af_ips) {
833 				error = EINVAL;
834 				vfs_opterror(opts, "too many IPv4 addresses");
835 				goto done_errmsg;
836 			}
837 			ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
838 			bcopy(op, ip4, ip4s * sizeof(*ip4));
839 			/*
840 			 * IP addresses are all sorted but ip[0] to preserve
841 			 * the primary IP address as given from userland.
842 			 * This special IP is used for unbound outgoing
843 			 * connections as well for "loopback" traffic in case
844 			 * source address selection cannot find any more fitting
845 			 * address to connect from.
846 			 */
847 			if (ip4s > 1)
848 				qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
849 			/*
850 			 * Check for duplicate addresses and do some simple
851 			 * zero and broadcast checks. If users give other bogus
852 			 * addresses it is their problem.
853 			 *
854 			 * We do not have to care about byte order for these
855 			 * checks so we will do them in NBO.
856 			 */
857 			for (ii = 0; ii < ip4s; ii++) {
858 				if (ip4[ii].s_addr == INADDR_ANY ||
859 				    ip4[ii].s_addr == INADDR_BROADCAST) {
860 					error = EINVAL;
861 					goto done_free;
862 				}
863 				if ((ii+1) < ip4s &&
864 				    (ip4[0].s_addr == ip4[ii+1].s_addr ||
865 				     ip4[ii].s_addr == ip4[ii+1].s_addr)) {
866 					error = EINVAL;
867 					goto done_free;
868 				}
869 			}
870 		}
871 	}
872 #endif
873 
874 #ifdef INET6
875 	error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
876 	if (error == ENOENT)
877 		ip6s = 0;
878 	else if (error != 0)
879 		goto done_free;
880 	else if (ip6s & (sizeof(*ip6) - 1)) {
881 		error = EINVAL;
882 		goto done_free;
883 	} else {
884 		ch_flags |= PR_IP6_USER;
885 		pr_flags |= PR_IP6_USER;
886 		if (ip6s > 0) {
887 			ip6s /= sizeof(*ip6);
888 			if (ip6s > jail_max_af_ips) {
889 				error = EINVAL;
890 				vfs_opterror(opts, "too many IPv6 addresses");
891 				goto done_errmsg;
892 			}
893 			ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
894 			bcopy(op, ip6, ip6s * sizeof(*ip6));
895 			if (ip6s > 1)
896 				qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
897 			for (ii = 0; ii < ip6s; ii++) {
898 				if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
899 					error = EINVAL;
900 					goto done_free;
901 				}
902 				if ((ii+1) < ip6s &&
903 				    (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
904 				     IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
905 				{
906 					error = EINVAL;
907 					goto done_free;
908 				}
909 			}
910 		}
911 	}
912 #endif
913 
914 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
915 	if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
916 		error = EINVAL;
917 		vfs_opterror(opts,
918 		    "vnet jails cannot have IP address restrictions");
919 		goto done_errmsg;
920 	}
921 #endif
922 
923 	fullpath_disabled = 0;
924 	root = NULL;
925 	error = vfs_getopt(opts, "path", (void **)&path, &len);
926 	if (error == ENOENT)
927 		path = NULL;
928 	else if (error != 0)
929 		goto done_free;
930 	else {
931 		if (flags & JAIL_UPDATE) {
932 			error = EINVAL;
933 			vfs_opterror(opts,
934 			    "path cannot be changed after creation");
935 			goto done_errmsg;
936 		}
937 		if (len == 0 || path[len - 1] != '\0') {
938 			error = EINVAL;
939 			goto done_free;
940 		}
941 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
942 		    path, td);
943 		error = namei(&nd);
944 		if (error)
945 			goto done_free;
946 		root = nd.ni_vp;
947 		NDFREE(&nd, NDF_ONLY_PNBUF);
948 		g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
949 		strlcpy(g_path, path, MAXPATHLEN);
950 		error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
951 		if (error == 0)
952 			path = g_path;
953 		else if (error == ENODEV) {
954 			/* proceed if sysctl debug.disablefullpath == 1 */
955 			fullpath_disabled = 1;
956 			if (len < 2 || (len == 2 && path[0] == '/'))
957 				path = NULL;
958 		} else {
959 			/* exit on other errors */
960 			goto done_free;
961 		}
962 		if (root->v_type != VDIR) {
963 			error = ENOTDIR;
964 			vput(root);
965 			goto done_free;
966 		}
967 		VOP_UNLOCK(root, 0);
968 		if (fullpath_disabled) {
969 			/* Leave room for a real-root full pathname. */
970 			if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
971 			    ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
972 				error = ENAMETOOLONG;
973 				goto done_free;
974 			}
975 		}
976 	}
977 
978 	error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
979 	if (error == ENOENT)
980 		osrelstr = NULL;
981 	else if (error != 0)
982 		goto done_free;
983 	else {
984 		if (flags & JAIL_UPDATE) {
985 			error = EINVAL;
986 			vfs_opterror(opts,
987 			    "osrelease cannot be changed after creation");
988 			goto done_errmsg;
989 		}
990 		if (len == 0 || len >= OSRELEASELEN) {
991 			error = EINVAL;
992 			vfs_opterror(opts,
993 			    "osrelease string must be 1-%d bytes long",
994 			    OSRELEASELEN - 1);
995 			goto done_errmsg;
996 		}
997 	}
998 
999 	error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
1000 	if (error == ENOENT)
1001 		osreldt = 0;
1002 	else if (error != 0)
1003 		goto done_free;
1004 	else {
1005 		if (flags & JAIL_UPDATE) {
1006 			error = EINVAL;
1007 			vfs_opterror(opts,
1008 			    "osreldate cannot be changed after creation");
1009 			goto done_errmsg;
1010 		}
1011 		if (osreldt == 0) {
1012 			error = EINVAL;
1013 			vfs_opterror(opts, "osreldate cannot be 0");
1014 			goto done_errmsg;
1015 		}
1016 	}
1017 
1018 	/*
1019 	 * Find the specified jail, or at least its parent.
1020 	 * This abuses the file error codes ENOENT and EEXIST.
1021 	 */
1022 	pr = NULL;
1023 	ppr = mypr;
1024 	if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1025 		namelc = strrchr(name, '.');
1026 		jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1027 		if (*p != '\0')
1028 			jid = 0;
1029 	}
1030 	sx_xlock(&allprison_lock);
1031 	if (jid != 0) {
1032 		/*
1033 		 * See if a requested jid already exists.  There is an
1034 		 * information leak here if the jid exists but is not within
1035 		 * the caller's jail hierarchy.  Jail creators will get EEXIST
1036 		 * even though they cannot see the jail, and CREATE | UPDATE
1037 		 * will return ENOENT which is not normally a valid error.
1038 		 */
1039 		if (jid < 0) {
1040 			error = EINVAL;
1041 			vfs_opterror(opts, "negative jid");
1042 			goto done_unlock_list;
1043 		}
1044 		pr = prison_find(jid);
1045 		if (pr != NULL) {
1046 			ppr = pr->pr_parent;
1047 			/* Create: jid must not exist. */
1048 			if (cuflags == JAIL_CREATE) {
1049 				mtx_unlock(&pr->pr_mtx);
1050 				error = EEXIST;
1051 				vfs_opterror(opts, "jail %d already exists",
1052 				    jid);
1053 				goto done_unlock_list;
1054 			}
1055 			if (!prison_ischild(mypr, pr)) {
1056 				mtx_unlock(&pr->pr_mtx);
1057 				pr = NULL;
1058 			} else if (pr->pr_uref == 0) {
1059 				if (!(flags & JAIL_DYING)) {
1060 					mtx_unlock(&pr->pr_mtx);
1061 					error = ENOENT;
1062 					vfs_opterror(opts, "jail %d is dying",
1063 					    jid);
1064 					goto done_unlock_list;
1065 				} else if ((flags & JAIL_ATTACH) ||
1066 				    (pr_flags & PR_PERSIST)) {
1067 					/*
1068 					 * A dying jail might be resurrected
1069 					 * (via attach or persist), but first
1070 					 * it must determine if another jail
1071 					 * has claimed its name.  Accomplish
1072 					 * this by implicitly re-setting the
1073 					 * name.
1074 					 */
1075 					if (name == NULL)
1076 						name = prison_name(mypr, pr);
1077 				}
1078 			}
1079 		}
1080 		if (pr == NULL) {
1081 			/* Update: jid must exist. */
1082 			if (cuflags == JAIL_UPDATE) {
1083 				error = ENOENT;
1084 				vfs_opterror(opts, "jail %d not found", jid);
1085 				goto done_unlock_list;
1086 			}
1087 		}
1088 	}
1089 	/*
1090 	 * If the caller provided a name, look for a jail by that name.
1091 	 * This has different semantics for creates and updates keyed by jid
1092 	 * (where the name must not already exist in a different jail),
1093 	 * and updates keyed by the name itself (where the name must exist
1094 	 * because that is the jail being updated).
1095 	 */
1096 	namelc = NULL;
1097 	if (name != NULL) {
1098 		namelc = strrchr(name, '.');
1099 		if (namelc == NULL)
1100 			namelc = name;
1101 		else {
1102 			/*
1103 			 * This is a hierarchical name.  Split it into the
1104 			 * parent and child names, and make sure the parent
1105 			 * exists or matches an already found jail.
1106 			 */
1107 			if (pr != NULL) {
1108 				if (strncmp(name, ppr->pr_name, namelc - name)
1109 				    || ppr->pr_name[namelc - name] != '\0') {
1110 					mtx_unlock(&pr->pr_mtx);
1111 					error = EINVAL;
1112 					vfs_opterror(opts,
1113 					    "cannot change jail's parent");
1114 					goto done_unlock_list;
1115 				}
1116 			} else {
1117 				*namelc = '\0';
1118 				ppr = prison_find_name(mypr, name);
1119 				if (ppr == NULL) {
1120 					error = ENOENT;
1121 					vfs_opterror(opts,
1122 					    "jail \"%s\" not found", name);
1123 					goto done_unlock_list;
1124 				}
1125 				mtx_unlock(&ppr->pr_mtx);
1126 				*namelc = '.';
1127 			}
1128 			namelc++;
1129 		}
1130 		if (namelc[0] != '\0') {
1131 			pnamelen =
1132 			    (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1133  name_again:
1134 			deadpr = NULL;
1135 			FOREACH_PRISON_CHILD(ppr, tpr) {
1136 				if (tpr != pr && tpr->pr_ref > 0 &&
1137 				    !strcmp(tpr->pr_name + pnamelen, namelc)) {
1138 					if (pr == NULL &&
1139 					    cuflags != JAIL_CREATE) {
1140 						mtx_lock(&tpr->pr_mtx);
1141 						if (tpr->pr_ref > 0) {
1142 							/*
1143 							 * Use this jail
1144 							 * for updates.
1145 							 */
1146 							if (tpr->pr_uref > 0) {
1147 								pr = tpr;
1148 								break;
1149 							}
1150 							deadpr = tpr;
1151 						}
1152 						mtx_unlock(&tpr->pr_mtx);
1153 					} else if (tpr->pr_uref > 0) {
1154 						/*
1155 						 * Create, or update(jid):
1156 						 * name must not exist in an
1157 						 * active sibling jail.
1158 						 */
1159 						error = EEXIST;
1160 						if (pr != NULL)
1161 							mtx_unlock(&pr->pr_mtx);
1162 						vfs_opterror(opts,
1163 						   "jail \"%s\" already exists",
1164 						   name);
1165 						goto done_unlock_list;
1166 					}
1167 				}
1168 			}
1169 			/* If no active jail is found, use a dying one. */
1170 			if (deadpr != NULL && pr == NULL) {
1171 				if (flags & JAIL_DYING) {
1172 					mtx_lock(&deadpr->pr_mtx);
1173 					if (deadpr->pr_ref == 0) {
1174 						mtx_unlock(&deadpr->pr_mtx);
1175 						goto name_again;
1176 					}
1177 					pr = deadpr;
1178 				} else if (cuflags == JAIL_UPDATE) {
1179 					error = ENOENT;
1180 					vfs_opterror(opts,
1181 					    "jail \"%s\" is dying", name);
1182 					goto done_unlock_list;
1183 				}
1184 			}
1185 			/* Update: name must exist if no jid. */
1186 			else if (cuflags == JAIL_UPDATE && pr == NULL) {
1187 				error = ENOENT;
1188 				vfs_opterror(opts, "jail \"%s\" not found",
1189 				    name);
1190 				goto done_unlock_list;
1191 			}
1192 		}
1193 	}
1194 	/* Update: must provide a jid or name. */
1195 	else if (cuflags == JAIL_UPDATE && pr == NULL) {
1196 		error = ENOENT;
1197 		vfs_opterror(opts, "update specified no jail");
1198 		goto done_unlock_list;
1199 	}
1200 
1201 	/* If there's no prison to update, create a new one and link it in. */
1202 	if (pr == NULL) {
1203 		for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1204 			if (tpr->pr_childcount >= tpr->pr_childmax) {
1205 				error = EPERM;
1206 				vfs_opterror(opts, "prison limit exceeded");
1207 				goto done_unlock_list;
1208 			}
1209 		created = 1;
1210 		mtx_lock(&ppr->pr_mtx);
1211 		if (ppr->pr_ref == 0) {
1212 			mtx_unlock(&ppr->pr_mtx);
1213 			error = ENOENT;
1214 			vfs_opterror(opts, "jail \"%s\" not found",
1215 			    prison_name(mypr, ppr));
1216 			goto done_unlock_list;
1217 		}
1218 		ppr->pr_ref++;
1219 		ppr->pr_uref++;
1220 		mtx_unlock(&ppr->pr_mtx);
1221 		pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1222 		if (jid == 0) {
1223 			/* Find the next free jid. */
1224 			jid = lastprid + 1;
1225  findnext:
1226 			if (jid == JAIL_MAX)
1227 				jid = 1;
1228 			TAILQ_FOREACH(tpr, &allprison, pr_list) {
1229 				if (tpr->pr_id < jid)
1230 					continue;
1231 				if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1232 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1233 					break;
1234 				}
1235 				if (jid == lastprid) {
1236 					error = EAGAIN;
1237 					vfs_opterror(opts,
1238 					    "no available jail IDs");
1239 					free(pr, M_PRISON);
1240 					prison_deref(ppr, PD_DEREF |
1241 					    PD_DEUREF | PD_LIST_XLOCKED);
1242 					goto done_releroot;
1243 				}
1244 				jid++;
1245 				goto findnext;
1246 			}
1247 			lastprid = jid;
1248 		} else {
1249 			/*
1250 			 * The jail already has a jid (that did not yet exist),
1251 			 * so just find where to insert it.
1252 			 */
1253 			TAILQ_FOREACH(tpr, &allprison, pr_list)
1254 				if (tpr->pr_id >= jid) {
1255 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1256 					break;
1257 				}
1258 		}
1259 		if (tpr == NULL)
1260 			TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1261 		LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1262 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1263 			tpr->pr_childcount++;
1264 
1265 		pr->pr_parent = ppr;
1266 		pr->pr_id = jid;
1267 
1268 		/* Set some default values, and inherit some from the parent. */
1269 		if (namelc == NULL)
1270 			namelc = "";
1271 		if (path == NULL) {
1272 			path = "/";
1273 			root = mypr->pr_root;
1274 			vref(root);
1275 		}
1276 		strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1277 		pr->pr_flags |= PR_HOST;
1278 #if defined(INET) || defined(INET6)
1279 #ifdef VIMAGE
1280 		if (!(pr_flags & PR_VNET))
1281 #endif
1282 		{
1283 #ifdef INET
1284 			if (!(ch_flags & PR_IP4_USER))
1285 				pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1286 			else if (!(pr_flags & PR_IP4_USER)) {
1287 				pr->pr_flags |= ppr->pr_flags & PR_IP4;
1288 				if (ppr->pr_ip4 != NULL) {
1289 					pr->pr_ip4s = ppr->pr_ip4s;
1290 					pr->pr_ip4 = malloc(pr->pr_ip4s *
1291 					    sizeof(struct in_addr), M_PRISON,
1292 					    M_WAITOK);
1293 					bcopy(ppr->pr_ip4, pr->pr_ip4,
1294 					    pr->pr_ip4s * sizeof(*pr->pr_ip4));
1295 				}
1296 			}
1297 #endif
1298 #ifdef INET6
1299 			if (!(ch_flags & PR_IP6_USER))
1300 				pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1301 			else if (!(pr_flags & PR_IP6_USER)) {
1302 				pr->pr_flags |= ppr->pr_flags & PR_IP6;
1303 				if (ppr->pr_ip6 != NULL) {
1304 					pr->pr_ip6s = ppr->pr_ip6s;
1305 					pr->pr_ip6 = malloc(pr->pr_ip6s *
1306 					    sizeof(struct in6_addr), M_PRISON,
1307 					    M_WAITOK);
1308 					bcopy(ppr->pr_ip6, pr->pr_ip6,
1309 					    pr->pr_ip6s * sizeof(*pr->pr_ip6));
1310 				}
1311 			}
1312 #endif
1313 		}
1314 #endif
1315 		/* Source address selection is always on by default. */
1316 		pr->pr_flags |= _PR_IP_SADDRSEL;
1317 
1318 		pr->pr_securelevel = ppr->pr_securelevel;
1319 		pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1320 		pr->pr_enforce_statfs = jail_default_enforce_statfs;
1321 		pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1322 
1323 		pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1324 		if (osrelstr == NULL)
1325 		    strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1326 		else
1327 		    strcpy(pr->pr_osrelease, osrelstr);
1328 
1329 		LIST_INIT(&pr->pr_children);
1330 		mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1331 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1332 
1333 #ifdef VIMAGE
1334 		/* Allocate a new vnet if specified. */
1335 		pr->pr_vnet = (pr_flags & PR_VNET)
1336 		    ? vnet_alloc() : ppr->pr_vnet;
1337 #endif
1338 		/*
1339 		 * Allocate a dedicated cpuset for each jail.
1340 		 * Unlike other initial settings, this may return an erorr.
1341 		 */
1342 		error = cpuset_create_root(ppr, &pr->pr_cpuset);
1343 		if (error) {
1344 			prison_deref(pr, PD_LIST_XLOCKED);
1345 			goto done_releroot;
1346 		}
1347 
1348 		mtx_lock(&pr->pr_mtx);
1349 		/*
1350 		 * New prisons do not yet have a reference, because we do not
1351 		 * want others to see the incomplete prison once the
1352 		 * allprison_lock is downgraded.
1353 		 */
1354 	} else {
1355 		created = 0;
1356 		/*
1357 		 * Grab a reference for existing prisons, to ensure they
1358 		 * continue to exist for the duration of the call.
1359 		 */
1360 		pr->pr_ref++;
1361 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1362 		if ((pr->pr_flags & PR_VNET) &&
1363 		    (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1364 			error = EINVAL;
1365 			vfs_opterror(opts,
1366 			    "vnet jails cannot have IP address restrictions");
1367 			goto done_deref_locked;
1368 		}
1369 #endif
1370 #ifdef INET
1371 		if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1372 			error = EINVAL;
1373 			vfs_opterror(opts,
1374 			    "ip4 cannot be changed after creation");
1375 			goto done_deref_locked;
1376 		}
1377 #endif
1378 #ifdef INET6
1379 		if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1380 			error = EINVAL;
1381 			vfs_opterror(opts,
1382 			    "ip6 cannot be changed after creation");
1383 			goto done_deref_locked;
1384 		}
1385 #endif
1386 	}
1387 
1388 	/* Do final error checking before setting anything. */
1389 	if (gotslevel) {
1390 		if (slevel < ppr->pr_securelevel) {
1391 			error = EPERM;
1392 			goto done_deref_locked;
1393 		}
1394 	}
1395 	if (gotchildmax) {
1396 		if (childmax >= ppr->pr_childmax) {
1397 			error = EPERM;
1398 			goto done_deref_locked;
1399 		}
1400 	}
1401 	if (gotenforce) {
1402 		if (enforce < ppr->pr_enforce_statfs) {
1403 			error = EPERM;
1404 			goto done_deref_locked;
1405 		}
1406 	}
1407 	if (gotrsnum) {
1408 		/*
1409 		 * devfs_rsnum is a uint16_t
1410 		 */
1411 		if (rsnum < 0 || rsnum > 65535) {
1412 			error = EINVAL;
1413 			goto done_deref_locked;
1414 		}
1415 		/*
1416 		 * Nested jails always inherit parent's devfs ruleset
1417 		 */
1418 		if (jailed(td->td_ucred)) {
1419 			if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1420 				error = EPERM;
1421 				goto done_deref_locked;
1422 			} else
1423 				rsnum = ppr->pr_devfs_rsnum;
1424 		}
1425 	}
1426 #ifdef INET
1427 	if (ip4s > 0) {
1428 		if (ppr->pr_flags & PR_IP4) {
1429 			/*
1430 			 * Make sure the new set of IP addresses is a
1431 			 * subset of the parent's list.  Don't worry
1432 			 * about the parent being unlocked, as any
1433 			 * setting is done with allprison_lock held.
1434 			 */
1435 			for (ij = 0; ij < ppr->pr_ip4s; ij++)
1436 				if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1437 					break;
1438 			if (ij == ppr->pr_ip4s) {
1439 				error = EPERM;
1440 				goto done_deref_locked;
1441 			}
1442 			if (ip4s > 1) {
1443 				for (ii = ij = 1; ii < ip4s; ii++) {
1444 					if (ip4[ii].s_addr ==
1445 					    ppr->pr_ip4[0].s_addr)
1446 						continue;
1447 					for (; ij < ppr->pr_ip4s; ij++)
1448 						if (ip4[ii].s_addr ==
1449 						    ppr->pr_ip4[ij].s_addr)
1450 							break;
1451 					if (ij == ppr->pr_ip4s)
1452 						break;
1453 				}
1454 				if (ij == ppr->pr_ip4s) {
1455 					error = EPERM;
1456 					goto done_deref_locked;
1457 				}
1458 			}
1459 		}
1460 		/*
1461 		 * Check for conflicting IP addresses.  We permit them
1462 		 * if there is no more than one IP on each jail.  If
1463 		 * there is a duplicate on a jail with more than one
1464 		 * IP stop checking and return error.
1465 		 */
1466 		tppr = ppr;
1467 #ifdef VIMAGE
1468 		for (; tppr != &prison0; tppr = tppr->pr_parent)
1469 			if (tppr->pr_flags & PR_VNET)
1470 				break;
1471 #endif
1472 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1473 			if (tpr == pr ||
1474 #ifdef VIMAGE
1475 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1476 #endif
1477 			    tpr->pr_uref == 0) {
1478 				descend = 0;
1479 				continue;
1480 			}
1481 			if (!(tpr->pr_flags & PR_IP4_USER))
1482 				continue;
1483 			descend = 0;
1484 			if (tpr->pr_ip4 == NULL ||
1485 			    (ip4s == 1 && tpr->pr_ip4s == 1))
1486 				continue;
1487 			for (ii = 0; ii < ip4s; ii++) {
1488 				if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1489 					error = EADDRINUSE;
1490 					vfs_opterror(opts,
1491 					    "IPv4 addresses clash");
1492 					goto done_deref_locked;
1493 				}
1494 			}
1495 		}
1496 	}
1497 #endif
1498 #ifdef INET6
1499 	if (ip6s > 0) {
1500 		if (ppr->pr_flags & PR_IP6) {
1501 			/*
1502 			 * Make sure the new set of IP addresses is a
1503 			 * subset of the parent's list.
1504 			 */
1505 			for (ij = 0; ij < ppr->pr_ip6s; ij++)
1506 				if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1507 				    &ppr->pr_ip6[ij]))
1508 					break;
1509 			if (ij == ppr->pr_ip6s) {
1510 				error = EPERM;
1511 				goto done_deref_locked;
1512 			}
1513 			if (ip6s > 1) {
1514 				for (ii = ij = 1; ii < ip6s; ii++) {
1515 					if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1516 					     &ppr->pr_ip6[0]))
1517 						continue;
1518 					for (; ij < ppr->pr_ip6s; ij++)
1519 						if (IN6_ARE_ADDR_EQUAL(
1520 						    &ip6[ii], &ppr->pr_ip6[ij]))
1521 							break;
1522 					if (ij == ppr->pr_ip6s)
1523 						break;
1524 				}
1525 				if (ij == ppr->pr_ip6s) {
1526 					error = EPERM;
1527 					goto done_deref_locked;
1528 				}
1529 			}
1530 		}
1531 		/* Check for conflicting IP addresses. */
1532 		tppr = ppr;
1533 #ifdef VIMAGE
1534 		for (; tppr != &prison0; tppr = tppr->pr_parent)
1535 			if (tppr->pr_flags & PR_VNET)
1536 				break;
1537 #endif
1538 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1539 			if (tpr == pr ||
1540 #ifdef VIMAGE
1541 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1542 #endif
1543 			    tpr->pr_uref == 0) {
1544 				descend = 0;
1545 				continue;
1546 			}
1547 			if (!(tpr->pr_flags & PR_IP6_USER))
1548 				continue;
1549 			descend = 0;
1550 			if (tpr->pr_ip6 == NULL ||
1551 			    (ip6s == 1 && tpr->pr_ip6s == 1))
1552 				continue;
1553 			for (ii = 0; ii < ip6s; ii++) {
1554 				if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1555 					error = EADDRINUSE;
1556 					vfs_opterror(opts,
1557 					    "IPv6 addresses clash");
1558 					goto done_deref_locked;
1559 				}
1560 			}
1561 		}
1562 	}
1563 #endif
1564 	onamelen = namelen = 0;
1565 	if (namelc != NULL) {
1566 		/* Give a default name of the jid.  Also allow the name to be
1567 		 * explicitly the jid - but not any other number, and only in
1568 		 * normal form (no leading zero/etc).
1569 		 */
1570 		if (namelc[0] == '\0')
1571 			snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1572 		else if ((strtoul(namelc, &p, 10) != jid ||
1573 			  namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1574 			error = EINVAL;
1575 			vfs_opterror(opts,
1576 			    "name cannot be numeric (unless it is the jid)");
1577 			goto done_deref_locked;
1578 		}
1579 		/*
1580 		 * Make sure the name isn't too long for the prison or its
1581 		 * children.
1582 		 */
1583 		pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1584 		onamelen = strlen(pr->pr_name + pnamelen);
1585 		namelen = strlen(namelc);
1586 		if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1587 			error = ENAMETOOLONG;
1588 			goto done_deref_locked;
1589 		}
1590 		FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1591 			if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1592 			    sizeof(pr->pr_name)) {
1593 				error = ENAMETOOLONG;
1594 				goto done_deref_locked;
1595 			}
1596 		}
1597 	}
1598 	if (pr_allow & ~ppr->pr_allow) {
1599 		error = EPERM;
1600 		goto done_deref_locked;
1601 	}
1602 
1603 	/*
1604 	 * Let modules check their parameters.  This requires unlocking and
1605 	 * then re-locking the prison, but this is still a valid state as long
1606 	 * as allprison_lock remains xlocked.
1607 	 */
1608 	mtx_unlock(&pr->pr_mtx);
1609 	error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1610 	if (error != 0) {
1611 		prison_deref(pr, created
1612 		    ? PD_LIST_XLOCKED
1613 		    : PD_DEREF | PD_LIST_XLOCKED);
1614 		goto done_releroot;
1615 	}
1616 	mtx_lock(&pr->pr_mtx);
1617 
1618 	/* At this point, all valid parameters should have been noted. */
1619 	TAILQ_FOREACH(opt, opts, link) {
1620 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
1621 			error = EINVAL;
1622 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
1623 			goto done_deref_locked;
1624 		}
1625 	}
1626 
1627 	/* Set the parameters of the prison. */
1628 #ifdef INET
1629 	redo_ip4 = 0;
1630 	if (pr_flags & PR_IP4_USER) {
1631 		pr->pr_flags |= PR_IP4;
1632 		free(pr->pr_ip4, M_PRISON);
1633 		pr->pr_ip4s = ip4s;
1634 		pr->pr_ip4 = ip4;
1635 		ip4 = NULL;
1636 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1637 #ifdef VIMAGE
1638 			if (tpr->pr_flags & PR_VNET) {
1639 				descend = 0;
1640 				continue;
1641 			}
1642 #endif
1643 			if (prison_restrict_ip4(tpr, NULL)) {
1644 				redo_ip4 = 1;
1645 				descend = 0;
1646 			}
1647 		}
1648 	}
1649 #endif
1650 #ifdef INET6
1651 	redo_ip6 = 0;
1652 	if (pr_flags & PR_IP6_USER) {
1653 		pr->pr_flags |= PR_IP6;
1654 		free(pr->pr_ip6, M_PRISON);
1655 		pr->pr_ip6s = ip6s;
1656 		pr->pr_ip6 = ip6;
1657 		ip6 = NULL;
1658 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1659 #ifdef VIMAGE
1660 			if (tpr->pr_flags & PR_VNET) {
1661 				descend = 0;
1662 				continue;
1663 			}
1664 #endif
1665 			if (prison_restrict_ip6(tpr, NULL)) {
1666 				redo_ip6 = 1;
1667 				descend = 0;
1668 			}
1669 		}
1670 	}
1671 #endif
1672 	if (gotslevel) {
1673 		pr->pr_securelevel = slevel;
1674 		/* Set all child jails to be at least this level. */
1675 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1676 			if (tpr->pr_securelevel < slevel)
1677 				tpr->pr_securelevel = slevel;
1678 	}
1679 	if (gotchildmax) {
1680 		pr->pr_childmax = childmax;
1681 		/* Set all child jails to under this limit. */
1682 		FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1683 			if (tpr->pr_childmax > childmax - level)
1684 				tpr->pr_childmax = childmax > level
1685 				    ? childmax - level : 0;
1686 	}
1687 	if (gotenforce) {
1688 		pr->pr_enforce_statfs = enforce;
1689 		/* Pass this restriction on to the children. */
1690 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1691 			if (tpr->pr_enforce_statfs < enforce)
1692 				tpr->pr_enforce_statfs = enforce;
1693 	}
1694 	if (gotrsnum) {
1695 		pr->pr_devfs_rsnum = rsnum;
1696 		/* Pass this restriction on to the children. */
1697 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1698 			tpr->pr_devfs_rsnum = rsnum;
1699 	}
1700 	if (namelc != NULL) {
1701 		if (ppr == &prison0)
1702 			strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1703 		else
1704 			snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1705 			    ppr->pr_name, namelc);
1706 		/* Change this component of child names. */
1707 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1708 			bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1709 			    strlen(tpr->pr_name + onamelen) + 1);
1710 			bcopy(pr->pr_name, tpr->pr_name, namelen);
1711 		}
1712 	}
1713 	if (path != NULL) {
1714 		/* Try to keep a real-rooted full pathname. */
1715 		if (fullpath_disabled && path[0] == '/' &&
1716 		    strcmp(mypr->pr_path, "/"))
1717 			snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1718 			    mypr->pr_path, path);
1719 		else
1720 			strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1721 		pr->pr_root = root;
1722 	}
1723 	if (PR_HOST & ch_flags & ~pr_flags) {
1724 		if (pr->pr_flags & PR_HOST) {
1725 			/*
1726 			 * Copy the parent's host info.  As with pr_ip4 above,
1727 			 * the lack of a lock on the parent is not a problem;
1728 			 * it is always set with allprison_lock at least
1729 			 * shared, and is held exclusively here.
1730 			 */
1731 			strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1732 			    sizeof(pr->pr_hostname));
1733 			strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1734 			    sizeof(pr->pr_domainname));
1735 			strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1736 			    sizeof(pr->pr_hostuuid));
1737 			pr->pr_hostid = pr->pr_parent->pr_hostid;
1738 		}
1739 	} else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1740 		/* Set this prison, and any descendants without PR_HOST. */
1741 		if (host != NULL)
1742 			strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1743 		if (domain != NULL)
1744 			strlcpy(pr->pr_domainname, domain,
1745 			    sizeof(pr->pr_domainname));
1746 		if (uuid != NULL)
1747 			strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1748 		if (gothid)
1749 			pr->pr_hostid = hid;
1750 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1751 			if (tpr->pr_flags & PR_HOST)
1752 				descend = 0;
1753 			else {
1754 				if (host != NULL)
1755 					strlcpy(tpr->pr_hostname,
1756 					    pr->pr_hostname,
1757 					    sizeof(tpr->pr_hostname));
1758 				if (domain != NULL)
1759 					strlcpy(tpr->pr_domainname,
1760 					    pr->pr_domainname,
1761 					    sizeof(tpr->pr_domainname));
1762 				if (uuid != NULL)
1763 					strlcpy(tpr->pr_hostuuid,
1764 					    pr->pr_hostuuid,
1765 					    sizeof(tpr->pr_hostuuid));
1766 				if (gothid)
1767 					tpr->pr_hostid = hid;
1768 			}
1769 		}
1770 	}
1771 	if ((tallow = ch_allow & ~pr_allow)) {
1772 		/* Clear allow bits in all children. */
1773 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1774 			tpr->pr_allow &= ~tallow;
1775 	}
1776 	pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1777 	/*
1778 	 * Persistent prisons get an extra reference, and prisons losing their
1779 	 * persist flag lose that reference.  Only do this for existing prisons
1780 	 * for now, so new ones will remain unseen until after the module
1781 	 * handlers have completed.
1782 	 */
1783 	born = pr->pr_uref == 0;
1784 	if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1785 		if (pr_flags & PR_PERSIST) {
1786 			pr->pr_ref++;
1787 			pr->pr_uref++;
1788 		} else {
1789 			pr->pr_ref--;
1790 			pr->pr_uref--;
1791 		}
1792 	}
1793 	pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1794 	mtx_unlock(&pr->pr_mtx);
1795 
1796 #ifdef RACCT
1797 	if (racct_enable && created)
1798 		prison_racct_attach(pr);
1799 #endif
1800 
1801 	/* Locks may have prevented a complete restriction of child IP
1802 	 * addresses.  If so, allocate some more memory and try again.
1803 	 */
1804 #ifdef INET
1805 	while (redo_ip4) {
1806 		ip4s = pr->pr_ip4s;
1807 		ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1808 		mtx_lock(&pr->pr_mtx);
1809 		redo_ip4 = 0;
1810 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1811 #ifdef VIMAGE
1812 			if (tpr->pr_flags & PR_VNET) {
1813 				descend = 0;
1814 				continue;
1815 			}
1816 #endif
1817 			if (prison_restrict_ip4(tpr, ip4)) {
1818 				if (ip4 != NULL)
1819 					ip4 = NULL;
1820 				else
1821 					redo_ip4 = 1;
1822 			}
1823 		}
1824 		mtx_unlock(&pr->pr_mtx);
1825 	}
1826 #endif
1827 #ifdef INET6
1828 	while (redo_ip6) {
1829 		ip6s = pr->pr_ip6s;
1830 		ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1831 		mtx_lock(&pr->pr_mtx);
1832 		redo_ip6 = 0;
1833 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1834 #ifdef VIMAGE
1835 			if (tpr->pr_flags & PR_VNET) {
1836 				descend = 0;
1837 				continue;
1838 			}
1839 #endif
1840 			if (prison_restrict_ip6(tpr, ip6)) {
1841 				if (ip6 != NULL)
1842 					ip6 = NULL;
1843 				else
1844 					redo_ip6 = 1;
1845 			}
1846 		}
1847 		mtx_unlock(&pr->pr_mtx);
1848 	}
1849 #endif
1850 
1851 	/* Let the modules do their work. */
1852 	sx_downgrade(&allprison_lock);
1853 	if (born) {
1854 		error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1855 		if (error) {
1856 			(void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1857 			prison_deref(pr, created
1858 			    ? PD_LIST_SLOCKED
1859 			    : PD_DEREF | PD_LIST_SLOCKED);
1860 			goto done_errmsg;
1861 		}
1862 	}
1863 	error = osd_jail_call(pr, PR_METHOD_SET, opts);
1864 	if (error) {
1865 		if (born)
1866 			(void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1867 		prison_deref(pr, created
1868 		    ? PD_LIST_SLOCKED
1869 		    : PD_DEREF | PD_LIST_SLOCKED);
1870 		goto done_errmsg;
1871 	}
1872 
1873 	/* Attach this process to the prison if requested. */
1874 	if (flags & JAIL_ATTACH) {
1875 		mtx_lock(&pr->pr_mtx);
1876 		error = do_jail_attach(td, pr);
1877 		if (error) {
1878 			vfs_opterror(opts, "attach failed");
1879 			if (!created)
1880 				prison_deref(pr, PD_DEREF);
1881 			goto done_errmsg;
1882 		}
1883 	}
1884 
1885 #ifdef RACCT
1886 	if (racct_enable && !created) {
1887 		if (!(flags & JAIL_ATTACH))
1888 			sx_sunlock(&allprison_lock);
1889 		prison_racct_modify(pr);
1890 		if (!(flags & JAIL_ATTACH))
1891 			sx_slock(&allprison_lock);
1892 	}
1893 #endif
1894 
1895 	td->td_retval[0] = pr->pr_id;
1896 
1897 	/*
1898 	 * Now that it is all there, drop the temporary reference from existing
1899 	 * prisons.  Or add a reference to newly created persistent prisons
1900 	 * (which was not done earlier so that the prison would not be publicly
1901 	 * visible).
1902 	 */
1903 	if (!created) {
1904 		prison_deref(pr, (flags & JAIL_ATTACH)
1905 		    ? PD_DEREF
1906 		    : PD_DEREF | PD_LIST_SLOCKED);
1907 	} else {
1908 		if (pr_flags & PR_PERSIST) {
1909 			mtx_lock(&pr->pr_mtx);
1910 			pr->pr_ref++;
1911 			pr->pr_uref++;
1912 			mtx_unlock(&pr->pr_mtx);
1913 		}
1914 		if (!(flags & JAIL_ATTACH))
1915 			sx_sunlock(&allprison_lock);
1916 	}
1917 
1918 	goto done_free;
1919 
1920  done_deref_locked:
1921 	prison_deref(pr, created
1922 	    ? PD_LOCKED | PD_LIST_XLOCKED
1923 	    : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1924 	goto done_releroot;
1925  done_unlock_list:
1926 	sx_xunlock(&allprison_lock);
1927  done_releroot:
1928 	if (root != NULL)
1929 		vrele(root);
1930  done_errmsg:
1931 	if (error) {
1932 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1933 		if (errmsg_len > 0) {
1934 			errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1935 			if (errmsg_pos > 0) {
1936 				if (optuio->uio_segflg == UIO_SYSSPACE)
1937 					bcopy(errmsg,
1938 					   optuio->uio_iov[errmsg_pos].iov_base,
1939 					   errmsg_len);
1940 				else
1941 					copyout(errmsg,
1942 					   optuio->uio_iov[errmsg_pos].iov_base,
1943 					   errmsg_len);
1944 			}
1945 		}
1946 	}
1947  done_free:
1948 #ifdef INET
1949 	free(ip4, M_PRISON);
1950 #endif
1951 #ifdef INET6
1952 	free(ip6, M_PRISON);
1953 #endif
1954 	if (g_path != NULL)
1955 		free(g_path, M_TEMP);
1956 	vfs_freeopts(opts);
1957 	return (error);
1958 }
1959 
1960 
1961 /*
1962  * struct jail_get_args {
1963  *	struct iovec *iovp;
1964  *	unsigned int iovcnt;
1965  *	int flags;
1966  * };
1967  */
1968 int
1969 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1970 {
1971 	struct uio *auio;
1972 	int error;
1973 
1974 	/* Check that we have an even number of iovecs. */
1975 	if (uap->iovcnt & 1)
1976 		return (EINVAL);
1977 
1978 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1979 	if (error)
1980 		return (error);
1981 	error = kern_jail_get(td, auio, uap->flags);
1982 	if (error == 0)
1983 		error = copyout(auio->uio_iov, uap->iovp,
1984 		    uap->iovcnt * sizeof (struct iovec));
1985 	free(auio, M_IOV);
1986 	return (error);
1987 }
1988 
1989 int
1990 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1991 {
1992 	struct prison *pr, *mypr;
1993 	struct vfsopt *opt;
1994 	struct vfsoptlist *opts;
1995 	char *errmsg, *name;
1996 	int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1997 
1998 	if (flags & ~JAIL_GET_MASK)
1999 		return (EINVAL);
2000 
2001 	/* Get the parameter list. */
2002 	error = vfs_buildopts(optuio, &opts);
2003 	if (error)
2004 		return (error);
2005 	errmsg_pos = vfs_getopt_pos(opts, "errmsg");
2006 	mypr = td->td_ucred->cr_prison;
2007 
2008 	/*
2009 	 * Find the prison specified by one of: lastjid, jid, name.
2010 	 */
2011 	sx_slock(&allprison_lock);
2012 	error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2013 	if (error == 0) {
2014 		TAILQ_FOREACH(pr, &allprison, pr_list) {
2015 			if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2016 				mtx_lock(&pr->pr_mtx);
2017 				if (pr->pr_ref > 0 &&
2018 				    (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2019 					break;
2020 				mtx_unlock(&pr->pr_mtx);
2021 			}
2022 		}
2023 		if (pr != NULL)
2024 			goto found_prison;
2025 		error = ENOENT;
2026 		vfs_opterror(opts, "no jail after %d", jid);
2027 		goto done_unlock_list;
2028 	} else if (error != ENOENT)
2029 		goto done_unlock_list;
2030 
2031 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2032 	if (error == 0) {
2033 		if (jid != 0) {
2034 			pr = prison_find_child(mypr, jid);
2035 			if (pr != NULL) {
2036 				if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2037 					mtx_unlock(&pr->pr_mtx);
2038 					error = ENOENT;
2039 					vfs_opterror(opts, "jail %d is dying",
2040 					    jid);
2041 					goto done_unlock_list;
2042 				}
2043 				goto found_prison;
2044 			}
2045 			error = ENOENT;
2046 			vfs_opterror(opts, "jail %d not found", jid);
2047 			goto done_unlock_list;
2048 		}
2049 	} else if (error != ENOENT)
2050 		goto done_unlock_list;
2051 
2052 	error = vfs_getopt(opts, "name", (void **)&name, &len);
2053 	if (error == 0) {
2054 		if (len == 0 || name[len - 1] != '\0') {
2055 			error = EINVAL;
2056 			goto done_unlock_list;
2057 		}
2058 		pr = prison_find_name(mypr, name);
2059 		if (pr != NULL) {
2060 			if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2061 				mtx_unlock(&pr->pr_mtx);
2062 				error = ENOENT;
2063 				vfs_opterror(opts, "jail \"%s\" is dying",
2064 				    name);
2065 				goto done_unlock_list;
2066 			}
2067 			goto found_prison;
2068 		}
2069 		error = ENOENT;
2070 		vfs_opterror(opts, "jail \"%s\" not found", name);
2071 		goto done_unlock_list;
2072 	} else if (error != ENOENT)
2073 		goto done_unlock_list;
2074 
2075 	vfs_opterror(opts, "no jail specified");
2076 	error = ENOENT;
2077 	goto done_unlock_list;
2078 
2079  found_prison:
2080 	/* Get the parameters of the prison. */
2081 	pr->pr_ref++;
2082 	locked = PD_LOCKED;
2083 	td->td_retval[0] = pr->pr_id;
2084 	error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2085 	if (error != 0 && error != ENOENT)
2086 		goto done_deref;
2087 	i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2088 	error = vfs_setopt(opts, "parent", &i, sizeof(i));
2089 	if (error != 0 && error != ENOENT)
2090 		goto done_deref;
2091 	error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2092 	if (error != 0 && error != ENOENT)
2093 		goto done_deref;
2094 	error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2095 	    sizeof(pr->pr_cpuset->cs_id));
2096 	if (error != 0 && error != ENOENT)
2097 		goto done_deref;
2098 	error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2099 	if (error != 0 && error != ENOENT)
2100 		goto done_deref;
2101 #ifdef INET
2102 	error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2103 	    pr->pr_ip4s * sizeof(*pr->pr_ip4));
2104 	if (error != 0 && error != ENOENT)
2105 		goto done_deref;
2106 #endif
2107 #ifdef INET6
2108 	error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2109 	    pr->pr_ip6s * sizeof(*pr->pr_ip6));
2110 	if (error != 0 && error != ENOENT)
2111 		goto done_deref;
2112 #endif
2113 	error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2114 	    sizeof(pr->pr_securelevel));
2115 	if (error != 0 && error != ENOENT)
2116 		goto done_deref;
2117 	error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2118 	    sizeof(pr->pr_childcount));
2119 	if (error != 0 && error != ENOENT)
2120 		goto done_deref;
2121 	error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2122 	    sizeof(pr->pr_childmax));
2123 	if (error != 0 && error != ENOENT)
2124 		goto done_deref;
2125 	error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2126 	if (error != 0 && error != ENOENT)
2127 		goto done_deref;
2128 	error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2129 	if (error != 0 && error != ENOENT)
2130 		goto done_deref;
2131 	error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2132 	if (error != 0 && error != ENOENT)
2133 		goto done_deref;
2134 #ifdef COMPAT_FREEBSD32
2135 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2136 		uint32_t hid32 = pr->pr_hostid;
2137 
2138 		error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2139 	} else
2140 #endif
2141 	error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2142 	    sizeof(pr->pr_hostid));
2143 	if (error != 0 && error != ENOENT)
2144 		goto done_deref;
2145 	error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2146 	    sizeof(pr->pr_enforce_statfs));
2147 	if (error != 0 && error != ENOENT)
2148 		goto done_deref;
2149 	error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2150 	    sizeof(pr->pr_devfs_rsnum));
2151 	if (error != 0 && error != ENOENT)
2152 		goto done_deref;
2153 	for (fi = 0; fi < nitems(pr_flag_names); fi++) {
2154 		if (pr_flag_names[fi] == NULL)
2155 			continue;
2156 		i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2157 		error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2158 		if (error != 0 && error != ENOENT)
2159 			goto done_deref;
2160 		i = !i;
2161 		error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2162 		if (error != 0 && error != ENOENT)
2163 			goto done_deref;
2164 	}
2165 	for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
2166 		i = pr->pr_flags &
2167 		    (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2168 		i = pr_flag_jailsys[fi].disable &&
2169 		      (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2170 		    : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2171 		    : JAIL_SYS_INHERIT;
2172 		error =
2173 		    vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2174 		if (error != 0 && error != ENOENT)
2175 			goto done_deref;
2176 	}
2177 	for (fi = 0; fi < nitems(pr_allow_names); fi++) {
2178 		if (pr_allow_names[fi] == NULL)
2179 			continue;
2180 		i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2181 		error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2182 		if (error != 0 && error != ENOENT)
2183 			goto done_deref;
2184 		i = !i;
2185 		error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2186 		if (error != 0 && error != ENOENT)
2187 			goto done_deref;
2188 	}
2189 	i = (pr->pr_uref == 0);
2190 	error = vfs_setopt(opts, "dying", &i, sizeof(i));
2191 	if (error != 0 && error != ENOENT)
2192 		goto done_deref;
2193 	i = !i;
2194 	error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2195 	if (error != 0 && error != ENOENT)
2196 		goto done_deref;
2197 	error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2198 	    sizeof(pr->pr_osreldate));
2199 	if (error != 0 && error != ENOENT)
2200 		goto done_deref;
2201 	error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2202 	if (error != 0 && error != ENOENT)
2203 		goto done_deref;
2204 
2205 	/* Get the module parameters. */
2206 	mtx_unlock(&pr->pr_mtx);
2207 	locked = 0;
2208 	error = osd_jail_call(pr, PR_METHOD_GET, opts);
2209 	if (error)
2210 		goto done_deref;
2211 	prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2212 
2213 	/* By now, all parameters should have been noted. */
2214 	TAILQ_FOREACH(opt, opts, link) {
2215 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
2216 			error = EINVAL;
2217 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
2218 			goto done_errmsg;
2219 		}
2220 	}
2221 
2222 	/* Write the fetched parameters back to userspace. */
2223 	error = 0;
2224 	TAILQ_FOREACH(opt, opts, link) {
2225 		if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2226 			pos = 2 * opt->pos + 1;
2227 			optuio->uio_iov[pos].iov_len = opt->len;
2228 			if (opt->value != NULL) {
2229 				if (optuio->uio_segflg == UIO_SYSSPACE) {
2230 					bcopy(opt->value,
2231 					    optuio->uio_iov[pos].iov_base,
2232 					    opt->len);
2233 				} else {
2234 					error = copyout(opt->value,
2235 					    optuio->uio_iov[pos].iov_base,
2236 					    opt->len);
2237 					if (error)
2238 						break;
2239 				}
2240 			}
2241 		}
2242 	}
2243 	goto done_errmsg;
2244 
2245  done_deref:
2246 	prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2247 	goto done_errmsg;
2248 
2249  done_unlock_list:
2250 	sx_sunlock(&allprison_lock);
2251  done_errmsg:
2252 	if (error && errmsg_pos >= 0) {
2253 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2254 		errmsg_pos = 2 * errmsg_pos + 1;
2255 		if (errmsg_len > 0) {
2256 			if (optuio->uio_segflg == UIO_SYSSPACE)
2257 				bcopy(errmsg,
2258 				    optuio->uio_iov[errmsg_pos].iov_base,
2259 				    errmsg_len);
2260 			else
2261 				copyout(errmsg,
2262 				    optuio->uio_iov[errmsg_pos].iov_base,
2263 				    errmsg_len);
2264 		}
2265 	}
2266 	vfs_freeopts(opts);
2267 	return (error);
2268 }
2269 
2270 
2271 /*
2272  * struct jail_remove_args {
2273  *	int jid;
2274  * };
2275  */
2276 int
2277 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2278 {
2279 	struct prison *pr, *cpr, *lpr, *tpr;
2280 	int descend, error;
2281 
2282 	error = priv_check(td, PRIV_JAIL_REMOVE);
2283 	if (error)
2284 		return (error);
2285 
2286 	sx_xlock(&allprison_lock);
2287 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2288 	if (pr == NULL) {
2289 		sx_xunlock(&allprison_lock);
2290 		return (EINVAL);
2291 	}
2292 
2293 	/* Remove all descendants of this prison, then remove this prison. */
2294 	pr->pr_ref++;
2295 	if (!LIST_EMPTY(&pr->pr_children)) {
2296 		mtx_unlock(&pr->pr_mtx);
2297 		lpr = NULL;
2298 		FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2299 			mtx_lock(&cpr->pr_mtx);
2300 			if (cpr->pr_ref > 0) {
2301 				tpr = cpr;
2302 				cpr->pr_ref++;
2303 			} else {
2304 				/* Already removed - do not do it again. */
2305 				tpr = NULL;
2306 			}
2307 			mtx_unlock(&cpr->pr_mtx);
2308 			if (lpr != NULL) {
2309 				mtx_lock(&lpr->pr_mtx);
2310 				prison_remove_one(lpr);
2311 				sx_xlock(&allprison_lock);
2312 			}
2313 			lpr = tpr;
2314 		}
2315 		if (lpr != NULL) {
2316 			mtx_lock(&lpr->pr_mtx);
2317 			prison_remove_one(lpr);
2318 			sx_xlock(&allprison_lock);
2319 		}
2320 		mtx_lock(&pr->pr_mtx);
2321 	}
2322 	prison_remove_one(pr);
2323 	return (0);
2324 }
2325 
2326 static void
2327 prison_remove_one(struct prison *pr)
2328 {
2329 	struct proc *p;
2330 	int deuref;
2331 
2332 	/* If the prison was persistent, it is not anymore. */
2333 	deuref = 0;
2334 	if (pr->pr_flags & PR_PERSIST) {
2335 		pr->pr_ref--;
2336 		deuref = PD_DEUREF;
2337 		pr->pr_flags &= ~PR_PERSIST;
2338 	}
2339 
2340 	/*
2341 	 * jail_remove added a reference.  If that's the only one, remove
2342 	 * the prison now.
2343 	 */
2344 	KASSERT(pr->pr_ref > 0,
2345 	    ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2346 	if (pr->pr_ref == 1) {
2347 		prison_deref(pr,
2348 		    deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2349 		return;
2350 	}
2351 
2352 	mtx_unlock(&pr->pr_mtx);
2353 	sx_xunlock(&allprison_lock);
2354 	/*
2355 	 * Kill all processes unfortunate enough to be attached to this prison.
2356 	 */
2357 	sx_slock(&allproc_lock);
2358 	LIST_FOREACH(p, &allproc, p_list) {
2359 		PROC_LOCK(p);
2360 		if (p->p_state != PRS_NEW && p->p_ucred &&
2361 		    p->p_ucred->cr_prison == pr)
2362 			kern_psignal(p, SIGKILL);
2363 		PROC_UNLOCK(p);
2364 	}
2365 	sx_sunlock(&allproc_lock);
2366 	/* Remove the temporary reference added by jail_remove. */
2367 	prison_deref(pr, deuref | PD_DEREF);
2368 }
2369 
2370 
2371 /*
2372  * struct jail_attach_args {
2373  *	int jid;
2374  * };
2375  */
2376 int
2377 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2378 {
2379 	struct prison *pr;
2380 	int error;
2381 
2382 	error = priv_check(td, PRIV_JAIL_ATTACH);
2383 	if (error)
2384 		return (error);
2385 
2386 	sx_slock(&allprison_lock);
2387 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2388 	if (pr == NULL) {
2389 		sx_sunlock(&allprison_lock);
2390 		return (EINVAL);
2391 	}
2392 
2393 	/*
2394 	 * Do not allow a process to attach to a prison that is not
2395 	 * considered to be "alive".
2396 	 */
2397 	if (pr->pr_uref == 0) {
2398 		mtx_unlock(&pr->pr_mtx);
2399 		sx_sunlock(&allprison_lock);
2400 		return (EINVAL);
2401 	}
2402 
2403 	return (do_jail_attach(td, pr));
2404 }
2405 
2406 static int
2407 do_jail_attach(struct thread *td, struct prison *pr)
2408 {
2409 	struct proc *p;
2410 	struct ucred *newcred, *oldcred;
2411 	int error;
2412 
2413 	/*
2414 	 * XXX: Note that there is a slight race here if two threads
2415 	 * in the same privileged process attempt to attach to two
2416 	 * different jails at the same time.  It is important for
2417 	 * user processes not to do this, or they might end up with
2418 	 * a process root from one prison, but attached to the jail
2419 	 * of another.
2420 	 */
2421 	pr->pr_ref++;
2422 	pr->pr_uref++;
2423 	mtx_unlock(&pr->pr_mtx);
2424 
2425 	/* Let modules do whatever they need to prepare for attaching. */
2426 	error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2427 	if (error) {
2428 		prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2429 		return (error);
2430 	}
2431 	sx_sunlock(&allprison_lock);
2432 
2433 	/*
2434 	 * Reparent the newly attached process to this jail.
2435 	 */
2436 	p = td->td_proc;
2437 	error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2438 	if (error)
2439 		goto e_revert_osd;
2440 
2441 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2442 	if ((error = change_dir(pr->pr_root, td)) != 0)
2443 		goto e_unlock;
2444 #ifdef MAC
2445 	if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2446 		goto e_unlock;
2447 #endif
2448 	VOP_UNLOCK(pr->pr_root, 0);
2449 	if ((error = pwd_chroot(td, pr->pr_root)))
2450 		goto e_revert_osd;
2451 
2452 	newcred = crget();
2453 	PROC_LOCK(p);
2454 	oldcred = crcopysafe(p, newcred);
2455 	newcred->cr_prison = pr;
2456 	proc_set_cred(p, newcred);
2457 	setsugid(p);
2458 	PROC_UNLOCK(p);
2459 #ifdef RACCT
2460 	racct_proc_ucred_changed(p, oldcred, newcred);
2461 #endif
2462 	prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2463 	crfree(oldcred);
2464 	return (0);
2465 
2466  e_unlock:
2467 	VOP_UNLOCK(pr->pr_root, 0);
2468  e_revert_osd:
2469 	/* Tell modules this thread is still in its old jail after all. */
2470 	(void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2471 	prison_deref(pr, PD_DEREF | PD_DEUREF);
2472 	return (error);
2473 }
2474 
2475 
2476 /*
2477  * Returns a locked prison instance, or NULL on failure.
2478  */
2479 struct prison *
2480 prison_find(int prid)
2481 {
2482 	struct prison *pr;
2483 
2484 	sx_assert(&allprison_lock, SX_LOCKED);
2485 	TAILQ_FOREACH(pr, &allprison, pr_list) {
2486 		if (pr->pr_id == prid) {
2487 			mtx_lock(&pr->pr_mtx);
2488 			if (pr->pr_ref > 0)
2489 				return (pr);
2490 			mtx_unlock(&pr->pr_mtx);
2491 		}
2492 	}
2493 	return (NULL);
2494 }
2495 
2496 /*
2497  * Find a prison that is a descendant of mypr.  Returns a locked prison or NULL.
2498  */
2499 struct prison *
2500 prison_find_child(struct prison *mypr, int prid)
2501 {
2502 	struct prison *pr;
2503 	int descend;
2504 
2505 	sx_assert(&allprison_lock, SX_LOCKED);
2506 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2507 		if (pr->pr_id == prid) {
2508 			mtx_lock(&pr->pr_mtx);
2509 			if (pr->pr_ref > 0)
2510 				return (pr);
2511 			mtx_unlock(&pr->pr_mtx);
2512 		}
2513 	}
2514 	return (NULL);
2515 }
2516 
2517 /*
2518  * Look for the name relative to mypr.  Returns a locked prison or NULL.
2519  */
2520 struct prison *
2521 prison_find_name(struct prison *mypr, const char *name)
2522 {
2523 	struct prison *pr, *deadpr;
2524 	size_t mylen;
2525 	int descend;
2526 
2527 	sx_assert(&allprison_lock, SX_LOCKED);
2528 	mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2529  again:
2530 	deadpr = NULL;
2531 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2532 		if (!strcmp(pr->pr_name + mylen, name)) {
2533 			mtx_lock(&pr->pr_mtx);
2534 			if (pr->pr_ref > 0) {
2535 				if (pr->pr_uref > 0)
2536 					return (pr);
2537 				deadpr = pr;
2538 			}
2539 			mtx_unlock(&pr->pr_mtx);
2540 		}
2541 	}
2542 	/* There was no valid prison - perhaps there was a dying one. */
2543 	if (deadpr != NULL) {
2544 		mtx_lock(&deadpr->pr_mtx);
2545 		if (deadpr->pr_ref == 0) {
2546 			mtx_unlock(&deadpr->pr_mtx);
2547 			goto again;
2548 		}
2549 	}
2550 	return (deadpr);
2551 }
2552 
2553 /*
2554  * See if a prison has the specific flag set.
2555  */
2556 int
2557 prison_flag(struct ucred *cred, unsigned flag)
2558 {
2559 
2560 	/* This is an atomic read, so no locking is necessary. */
2561 	return (cred->cr_prison->pr_flags & flag);
2562 }
2563 
2564 int
2565 prison_allow(struct ucred *cred, unsigned flag)
2566 {
2567 
2568 	/* This is an atomic read, so no locking is necessary. */
2569 	return (cred->cr_prison->pr_allow & flag);
2570 }
2571 
2572 /*
2573  * Remove a prison reference.  If that was the last reference, remove the
2574  * prison itself - but not in this context in case there are locks held.
2575  */
2576 void
2577 prison_free_locked(struct prison *pr)
2578 {
2579 	int ref;
2580 
2581 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2582 	ref = --pr->pr_ref;
2583 	mtx_unlock(&pr->pr_mtx);
2584 	if (ref == 0)
2585 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2586 }
2587 
2588 void
2589 prison_free(struct prison *pr)
2590 {
2591 
2592 	mtx_lock(&pr->pr_mtx);
2593 	prison_free_locked(pr);
2594 }
2595 
2596 /*
2597  * Complete a call to either prison_free or prison_proc_free.
2598  */
2599 static void
2600 prison_complete(void *context, int pending)
2601 {
2602 	struct prison *pr = context;
2603 
2604 	mtx_lock(&pr->pr_mtx);
2605 	prison_deref(pr, pr->pr_uref
2606 	    ? PD_DEREF | PD_DEUREF | PD_LOCKED : PD_LOCKED);
2607 }
2608 
2609 /*
2610  * Remove a prison reference (usually).  This internal version assumes no
2611  * mutexes are held, except perhaps the prison itself.  If there are no more
2612  * references, release and delist the prison.  On completion, the prison lock
2613  * and the allprison lock are both unlocked.
2614  */
2615 static void
2616 prison_deref(struct prison *pr, int flags)
2617 {
2618 	struct prison *ppr, *tpr;
2619 	int ref, lasturef;
2620 
2621 	if (!(flags & PD_LOCKED))
2622 		mtx_lock(&pr->pr_mtx);
2623 	for (;;) {
2624 		if (flags & PD_DEUREF) {
2625 			KASSERT(pr->pr_uref > 0,
2626 			    ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2627 			     pr->pr_id));
2628 			pr->pr_uref--;
2629 			lasturef = pr->pr_uref == 0;
2630 			if (lasturef)
2631 				pr->pr_ref++;
2632 			KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2633 		} else
2634 			lasturef = 0;
2635 		if (flags & PD_DEREF) {
2636 			KASSERT(pr->pr_ref > 0,
2637 			    ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2638 			     pr->pr_id));
2639 			pr->pr_ref--;
2640 		}
2641 		ref = pr->pr_ref;
2642 		mtx_unlock(&pr->pr_mtx);
2643 
2644 		/*
2645 		 * Tell the modules if the last user reference was removed
2646 		 * (even it sticks around in dying state).
2647 		 */
2648 		if (lasturef) {
2649 			if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2650 				if (ref > 1) {
2651 					sx_slock(&allprison_lock);
2652 					flags |= PD_LIST_SLOCKED;
2653 				} else {
2654 					sx_xlock(&allprison_lock);
2655 					flags |= PD_LIST_XLOCKED;
2656 				}
2657 			}
2658 			(void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2659 			mtx_lock(&pr->pr_mtx);
2660 			ref = --pr->pr_ref;
2661 			mtx_unlock(&pr->pr_mtx);
2662 		}
2663 
2664 		/* If the prison still has references, nothing else to do. */
2665 		if (ref > 0) {
2666 			if (flags & PD_LIST_SLOCKED)
2667 				sx_sunlock(&allprison_lock);
2668 			else if (flags & PD_LIST_XLOCKED)
2669 				sx_xunlock(&allprison_lock);
2670 			return;
2671 		}
2672 
2673 		if (flags & PD_LIST_SLOCKED) {
2674 			if (!sx_try_upgrade(&allprison_lock)) {
2675 				sx_sunlock(&allprison_lock);
2676 				sx_xlock(&allprison_lock);
2677 			}
2678 		} else if (!(flags & PD_LIST_XLOCKED))
2679 			sx_xlock(&allprison_lock);
2680 
2681 		TAILQ_REMOVE(&allprison, pr, pr_list);
2682 		LIST_REMOVE(pr, pr_sibling);
2683 		ppr = pr->pr_parent;
2684 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2685 			tpr->pr_childcount--;
2686 		sx_xunlock(&allprison_lock);
2687 
2688 #ifdef VIMAGE
2689 		if (pr->pr_vnet != ppr->pr_vnet)
2690 			vnet_destroy(pr->pr_vnet);
2691 #endif
2692 		if (pr->pr_root != NULL)
2693 			vrele(pr->pr_root);
2694 		mtx_destroy(&pr->pr_mtx);
2695 #ifdef INET
2696 		free(pr->pr_ip4, M_PRISON);
2697 #endif
2698 #ifdef INET6
2699 		free(pr->pr_ip6, M_PRISON);
2700 #endif
2701 		if (pr->pr_cpuset != NULL)
2702 			cpuset_rel(pr->pr_cpuset);
2703 		osd_jail_exit(pr);
2704 #ifdef RACCT
2705 		if (racct_enable)
2706 			prison_racct_detach(pr);
2707 #endif
2708 		free(pr, M_PRISON);
2709 
2710 		/* Removing a prison frees a reference on its parent. */
2711 		pr = ppr;
2712 		mtx_lock(&pr->pr_mtx);
2713 		flags = PD_DEREF | PD_DEUREF;
2714 	}
2715 }
2716 
2717 void
2718 prison_hold_locked(struct prison *pr)
2719 {
2720 
2721 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2722 	KASSERT(pr->pr_ref > 0,
2723 	    ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2724 	pr->pr_ref++;
2725 }
2726 
2727 void
2728 prison_hold(struct prison *pr)
2729 {
2730 
2731 	mtx_lock(&pr->pr_mtx);
2732 	prison_hold_locked(pr);
2733 	mtx_unlock(&pr->pr_mtx);
2734 }
2735 
2736 void
2737 prison_proc_hold(struct prison *pr)
2738 {
2739 
2740 	mtx_lock(&pr->pr_mtx);
2741 	KASSERT(pr->pr_uref > 0,
2742 	    ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2743 	pr->pr_uref++;
2744 	mtx_unlock(&pr->pr_mtx);
2745 }
2746 
2747 void
2748 prison_proc_free(struct prison *pr)
2749 {
2750 
2751 	mtx_lock(&pr->pr_mtx);
2752 	KASSERT(pr->pr_uref > 0,
2753 	    ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2754 	if (pr->pr_uref > 1)
2755 		pr->pr_uref--;
2756 	else {
2757 		/*
2758 		 * Don't remove the last user reference in this context, which
2759 		 * is expected to be a process that is not only locked, but
2760 		 * also half dead.
2761 		 */
2762 		pr->pr_ref++;
2763 		mtx_unlock(&pr->pr_mtx);
2764 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2765 		return;
2766 	}
2767 	mtx_unlock(&pr->pr_mtx);
2768 }
2769 
2770 
2771 #ifdef INET
2772 /*
2773  * Restrict a prison's IP address list with its parent's, possibly replacing
2774  * it.  Return true if the replacement buffer was used (or would have been).
2775  */
2776 static int
2777 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2778 {
2779 	int ii, ij, used;
2780 	struct prison *ppr;
2781 
2782 	ppr = pr->pr_parent;
2783 	if (!(pr->pr_flags & PR_IP4_USER)) {
2784 		/* This has no user settings, so just copy the parent's list. */
2785 		if (pr->pr_ip4s < ppr->pr_ip4s) {
2786 			/*
2787 			 * There's no room for the parent's list.  Use the
2788 			 * new list buffer, which is assumed to be big enough
2789 			 * (if it was passed).  If there's no buffer, try to
2790 			 * allocate one.
2791 			 */
2792 			used = 1;
2793 			if (newip4 == NULL) {
2794 				newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2795 				    M_PRISON, M_NOWAIT);
2796 				if (newip4 != NULL)
2797 					used = 0;
2798 			}
2799 			if (newip4 != NULL) {
2800 				bcopy(ppr->pr_ip4, newip4,
2801 				    ppr->pr_ip4s * sizeof(*newip4));
2802 				free(pr->pr_ip4, M_PRISON);
2803 				pr->pr_ip4 = newip4;
2804 				pr->pr_ip4s = ppr->pr_ip4s;
2805 			}
2806 			return (used);
2807 		}
2808 		pr->pr_ip4s = ppr->pr_ip4s;
2809 		if (pr->pr_ip4s > 0)
2810 			bcopy(ppr->pr_ip4, pr->pr_ip4,
2811 			    pr->pr_ip4s * sizeof(*newip4));
2812 		else if (pr->pr_ip4 != NULL) {
2813 			free(pr->pr_ip4, M_PRISON);
2814 			pr->pr_ip4 = NULL;
2815 		}
2816 	} else if (pr->pr_ip4s > 0) {
2817 		/* Remove addresses that aren't in the parent. */
2818 		for (ij = 0; ij < ppr->pr_ip4s; ij++)
2819 			if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2820 				break;
2821 		if (ij < ppr->pr_ip4s)
2822 			ii = 1;
2823 		else {
2824 			bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2825 			    --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2826 			ii = 0;
2827 		}
2828 		for (ij = 1; ii < pr->pr_ip4s; ) {
2829 			if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2830 				ii++;
2831 				continue;
2832 			}
2833 			switch (ij >= ppr->pr_ip4s ? -1 :
2834 				qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2835 			case -1:
2836 				bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2837 				    (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2838 				break;
2839 			case 0:
2840 				ii++;
2841 				ij++;
2842 				break;
2843 			case 1:
2844 				ij++;
2845 				break;
2846 			}
2847 		}
2848 		if (pr->pr_ip4s == 0) {
2849 			free(pr->pr_ip4, M_PRISON);
2850 			pr->pr_ip4 = NULL;
2851 		}
2852 	}
2853 	return (0);
2854 }
2855 
2856 /*
2857  * Pass back primary IPv4 address of this jail.
2858  *
2859  * If not restricted return success but do not alter the address.  Caller has
2860  * to make sure to initialize it correctly (e.g. INADDR_ANY).
2861  *
2862  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2863  * Address returned in NBO.
2864  */
2865 int
2866 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2867 {
2868 	struct prison *pr;
2869 
2870 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2871 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2872 
2873 	pr = cred->cr_prison;
2874 	if (!(pr->pr_flags & PR_IP4))
2875 		return (0);
2876 	mtx_lock(&pr->pr_mtx);
2877 	if (!(pr->pr_flags & PR_IP4)) {
2878 		mtx_unlock(&pr->pr_mtx);
2879 		return (0);
2880 	}
2881 	if (pr->pr_ip4 == NULL) {
2882 		mtx_unlock(&pr->pr_mtx);
2883 		return (EAFNOSUPPORT);
2884 	}
2885 
2886 	ia->s_addr = pr->pr_ip4[0].s_addr;
2887 	mtx_unlock(&pr->pr_mtx);
2888 	return (0);
2889 }
2890 
2891 /*
2892  * Return 1 if we should do proper source address selection or are not jailed.
2893  * We will return 0 if we should bypass source address selection in favour
2894  * of the primary jail IPv4 address. Only in this case *ia will be updated and
2895  * returned in NBO.
2896  * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2897  */
2898 int
2899 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2900 {
2901 	struct prison *pr;
2902 	struct in_addr lia;
2903 	int error;
2904 
2905 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2906 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2907 
2908 	if (!jailed(cred))
2909 		return (1);
2910 
2911 	pr = cred->cr_prison;
2912 	if (pr->pr_flags & PR_IP4_SADDRSEL)
2913 		return (1);
2914 
2915 	lia.s_addr = INADDR_ANY;
2916 	error = prison_get_ip4(cred, &lia);
2917 	if (error)
2918 		return (error);
2919 	if (lia.s_addr == INADDR_ANY)
2920 		return (1);
2921 
2922 	ia->s_addr = lia.s_addr;
2923 	return (0);
2924 }
2925 
2926 /*
2927  * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2928  */
2929 int
2930 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2931 {
2932 
2933 	if (pr1 == pr2)
2934 		return (1);
2935 
2936 	/*
2937 	 * No need to lock since the PR_IP4_USER flag can't be altered for
2938 	 * existing prisons.
2939 	 */
2940 	while (pr1 != &prison0 &&
2941 #ifdef VIMAGE
2942 	       !(pr1->pr_flags & PR_VNET) &&
2943 #endif
2944 	       !(pr1->pr_flags & PR_IP4_USER))
2945 		pr1 = pr1->pr_parent;
2946 	while (pr2 != &prison0 &&
2947 #ifdef VIMAGE
2948 	       !(pr2->pr_flags & PR_VNET) &&
2949 #endif
2950 	       !(pr2->pr_flags & PR_IP4_USER))
2951 		pr2 = pr2->pr_parent;
2952 	return (pr1 == pr2);
2953 }
2954 
2955 /*
2956  * Make sure our (source) address is set to something meaningful to this
2957  * jail.
2958  *
2959  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2960  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2961  * doesn't allow IPv4.  Address passed in in NBO and returned in NBO.
2962  */
2963 int
2964 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2965 {
2966 	struct prison *pr;
2967 	struct in_addr ia0;
2968 	int error;
2969 
2970 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2971 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2972 
2973 	pr = cred->cr_prison;
2974 	if (!(pr->pr_flags & PR_IP4))
2975 		return (0);
2976 	mtx_lock(&pr->pr_mtx);
2977 	if (!(pr->pr_flags & PR_IP4)) {
2978 		mtx_unlock(&pr->pr_mtx);
2979 		return (0);
2980 	}
2981 	if (pr->pr_ip4 == NULL) {
2982 		mtx_unlock(&pr->pr_mtx);
2983 		return (EAFNOSUPPORT);
2984 	}
2985 
2986 	ia0.s_addr = ntohl(ia->s_addr);
2987 	if (ia0.s_addr == INADDR_LOOPBACK) {
2988 		ia->s_addr = pr->pr_ip4[0].s_addr;
2989 		mtx_unlock(&pr->pr_mtx);
2990 		return (0);
2991 	}
2992 
2993 	if (ia0.s_addr == INADDR_ANY) {
2994 		/*
2995 		 * In case there is only 1 IPv4 address, bind directly.
2996 		 */
2997 		if (pr->pr_ip4s == 1)
2998 			ia->s_addr = pr->pr_ip4[0].s_addr;
2999 		mtx_unlock(&pr->pr_mtx);
3000 		return (0);
3001 	}
3002 
3003 	error = _prison_check_ip4(pr, ia);
3004 	mtx_unlock(&pr->pr_mtx);
3005 	return (error);
3006 }
3007 
3008 /*
3009  * Rewrite destination address in case we will connect to loopback address.
3010  *
3011  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
3012  * Address passed in in NBO and returned in NBO.
3013  */
3014 int
3015 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
3016 {
3017 	struct prison *pr;
3018 
3019 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3020 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3021 
3022 	pr = cred->cr_prison;
3023 	if (!(pr->pr_flags & PR_IP4))
3024 		return (0);
3025 	mtx_lock(&pr->pr_mtx);
3026 	if (!(pr->pr_flags & PR_IP4)) {
3027 		mtx_unlock(&pr->pr_mtx);
3028 		return (0);
3029 	}
3030 	if (pr->pr_ip4 == NULL) {
3031 		mtx_unlock(&pr->pr_mtx);
3032 		return (EAFNOSUPPORT);
3033 	}
3034 
3035 	if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
3036 		ia->s_addr = pr->pr_ip4[0].s_addr;
3037 		mtx_unlock(&pr->pr_mtx);
3038 		return (0);
3039 	}
3040 
3041 	/*
3042 	 * Return success because nothing had to be changed.
3043 	 */
3044 	mtx_unlock(&pr->pr_mtx);
3045 	return (0);
3046 }
3047 
3048 /*
3049  * Check if given address belongs to the jail referenced by cred/prison.
3050  *
3051  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
3052  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3053  * doesn't allow IPv4.  Address passed in in NBO.
3054  */
3055 static int
3056 _prison_check_ip4(const struct prison *pr, const struct in_addr *ia)
3057 {
3058 	int i, a, z, d;
3059 
3060 	/*
3061 	 * Check the primary IP.
3062 	 */
3063 	if (pr->pr_ip4[0].s_addr == ia->s_addr)
3064 		return (0);
3065 
3066 	/*
3067 	 * All the other IPs are sorted so we can do a binary search.
3068 	 */
3069 	a = 0;
3070 	z = pr->pr_ip4s - 2;
3071 	while (a <= z) {
3072 		i = (a + z) / 2;
3073 		d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3074 		if (d > 0)
3075 			z = i - 1;
3076 		else if (d < 0)
3077 			a = i + 1;
3078 		else
3079 			return (0);
3080 	}
3081 
3082 	return (EADDRNOTAVAIL);
3083 }
3084 
3085 int
3086 prison_check_ip4(const struct ucred *cred, const struct in_addr *ia)
3087 {
3088 	struct prison *pr;
3089 	int error;
3090 
3091 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3092 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3093 
3094 	pr = cred->cr_prison;
3095 	if (!(pr->pr_flags & PR_IP4))
3096 		return (0);
3097 	mtx_lock(&pr->pr_mtx);
3098 	if (!(pr->pr_flags & PR_IP4)) {
3099 		mtx_unlock(&pr->pr_mtx);
3100 		return (0);
3101 	}
3102 	if (pr->pr_ip4 == NULL) {
3103 		mtx_unlock(&pr->pr_mtx);
3104 		return (EAFNOSUPPORT);
3105 	}
3106 
3107 	error = _prison_check_ip4(pr, ia);
3108 	mtx_unlock(&pr->pr_mtx);
3109 	return (error);
3110 }
3111 #endif
3112 
3113 #ifdef INET6
3114 static int
3115 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3116 {
3117 	int ii, ij, used;
3118 	struct prison *ppr;
3119 
3120 	ppr = pr->pr_parent;
3121 	if (!(pr->pr_flags & PR_IP6_USER)) {
3122 		/* This has no user settings, so just copy the parent's list. */
3123 		if (pr->pr_ip6s < ppr->pr_ip6s) {
3124 			/*
3125 			 * There's no room for the parent's list.  Use the
3126 			 * new list buffer, which is assumed to be big enough
3127 			 * (if it was passed).  If there's no buffer, try to
3128 			 * allocate one.
3129 			 */
3130 			used = 1;
3131 			if (newip6 == NULL) {
3132 				newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3133 				    M_PRISON, M_NOWAIT);
3134 				if (newip6 != NULL)
3135 					used = 0;
3136 			}
3137 			if (newip6 != NULL) {
3138 				bcopy(ppr->pr_ip6, newip6,
3139 				    ppr->pr_ip6s * sizeof(*newip6));
3140 				free(pr->pr_ip6, M_PRISON);
3141 				pr->pr_ip6 = newip6;
3142 				pr->pr_ip6s = ppr->pr_ip6s;
3143 			}
3144 			return (used);
3145 		}
3146 		pr->pr_ip6s = ppr->pr_ip6s;
3147 		if (pr->pr_ip6s > 0)
3148 			bcopy(ppr->pr_ip6, pr->pr_ip6,
3149 			    pr->pr_ip6s * sizeof(*newip6));
3150 		else if (pr->pr_ip6 != NULL) {
3151 			free(pr->pr_ip6, M_PRISON);
3152 			pr->pr_ip6 = NULL;
3153 		}
3154 	} else if (pr->pr_ip6s > 0) {
3155 		/* Remove addresses that aren't in the parent. */
3156 		for (ij = 0; ij < ppr->pr_ip6s; ij++)
3157 			if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3158 			    &ppr->pr_ip6[ij]))
3159 				break;
3160 		if (ij < ppr->pr_ip6s)
3161 			ii = 1;
3162 		else {
3163 			bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3164 			    --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3165 			ii = 0;
3166 		}
3167 		for (ij = 1; ii < pr->pr_ip6s; ) {
3168 			if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3169 			    &ppr->pr_ip6[0])) {
3170 				ii++;
3171 				continue;
3172 			}
3173 			switch (ij >= ppr->pr_ip6s ? -1 :
3174 				qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3175 			case -1:
3176 				bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3177 				    (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3178 				break;
3179 			case 0:
3180 				ii++;
3181 				ij++;
3182 				break;
3183 			case 1:
3184 				ij++;
3185 				break;
3186 			}
3187 		}
3188 		if (pr->pr_ip6s == 0) {
3189 			free(pr->pr_ip6, M_PRISON);
3190 			pr->pr_ip6 = NULL;
3191 		}
3192 	}
3193 	return 0;
3194 }
3195 
3196 /*
3197  * Pass back primary IPv6 address for this jail.
3198  *
3199  * If not restricted return success but do not alter the address.  Caller has
3200  * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3201  *
3202  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3203  */
3204 int
3205 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3206 {
3207 	struct prison *pr;
3208 
3209 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3210 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3211 
3212 	pr = cred->cr_prison;
3213 	if (!(pr->pr_flags & PR_IP6))
3214 		return (0);
3215 	mtx_lock(&pr->pr_mtx);
3216 	if (!(pr->pr_flags & PR_IP6)) {
3217 		mtx_unlock(&pr->pr_mtx);
3218 		return (0);
3219 	}
3220 	if (pr->pr_ip6 == NULL) {
3221 		mtx_unlock(&pr->pr_mtx);
3222 		return (EAFNOSUPPORT);
3223 	}
3224 
3225 	bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3226 	mtx_unlock(&pr->pr_mtx);
3227 	return (0);
3228 }
3229 
3230 /*
3231  * Return 1 if we should do proper source address selection or are not jailed.
3232  * We will return 0 if we should bypass source address selection in favour
3233  * of the primary jail IPv6 address. Only in this case *ia will be updated and
3234  * returned in NBO.
3235  * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3236  */
3237 int
3238 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3239 {
3240 	struct prison *pr;
3241 	struct in6_addr lia6;
3242 	int error;
3243 
3244 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3245 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3246 
3247 	if (!jailed(cred))
3248 		return (1);
3249 
3250 	pr = cred->cr_prison;
3251 	if (pr->pr_flags & PR_IP6_SADDRSEL)
3252 		return (1);
3253 
3254 	lia6 = in6addr_any;
3255 	error = prison_get_ip6(cred, &lia6);
3256 	if (error)
3257 		return (error);
3258 	if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3259 		return (1);
3260 
3261 	bcopy(&lia6, ia6, sizeof(struct in6_addr));
3262 	return (0);
3263 }
3264 
3265 /*
3266  * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3267  */
3268 int
3269 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3270 {
3271 
3272 	if (pr1 == pr2)
3273 		return (1);
3274 
3275 	while (pr1 != &prison0 &&
3276 #ifdef VIMAGE
3277 	       !(pr1->pr_flags & PR_VNET) &&
3278 #endif
3279 	       !(pr1->pr_flags & PR_IP6_USER))
3280 		pr1 = pr1->pr_parent;
3281 	while (pr2 != &prison0 &&
3282 #ifdef VIMAGE
3283 	       !(pr2->pr_flags & PR_VNET) &&
3284 #endif
3285 	       !(pr2->pr_flags & PR_IP6_USER))
3286 		pr2 = pr2->pr_parent;
3287 	return (pr1 == pr2);
3288 }
3289 
3290 /*
3291  * Make sure our (source) address is set to something meaningful to this jail.
3292  *
3293  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3294  * when needed while binding.
3295  *
3296  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3297  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3298  * doesn't allow IPv6.
3299  */
3300 int
3301 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3302 {
3303 	struct prison *pr;
3304 	int error;
3305 
3306 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3307 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3308 
3309 	pr = cred->cr_prison;
3310 	if (!(pr->pr_flags & PR_IP6))
3311 		return (0);
3312 	mtx_lock(&pr->pr_mtx);
3313 	if (!(pr->pr_flags & PR_IP6)) {
3314 		mtx_unlock(&pr->pr_mtx);
3315 		return (0);
3316 	}
3317 	if (pr->pr_ip6 == NULL) {
3318 		mtx_unlock(&pr->pr_mtx);
3319 		return (EAFNOSUPPORT);
3320 	}
3321 
3322 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3323 		bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3324 		mtx_unlock(&pr->pr_mtx);
3325 		return (0);
3326 	}
3327 
3328 	if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3329 		/*
3330 		 * In case there is only 1 IPv6 address, and v6only is true,
3331 		 * then bind directly.
3332 		 */
3333 		if (v6only != 0 && pr->pr_ip6s == 1)
3334 			bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3335 		mtx_unlock(&pr->pr_mtx);
3336 		return (0);
3337 	}
3338 
3339 	error = _prison_check_ip6(pr, ia6);
3340 	mtx_unlock(&pr->pr_mtx);
3341 	return (error);
3342 }
3343 
3344 /*
3345  * Rewrite destination address in case we will connect to loopback address.
3346  *
3347  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3348  */
3349 int
3350 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3351 {
3352 	struct prison *pr;
3353 
3354 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3355 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3356 
3357 	pr = cred->cr_prison;
3358 	if (!(pr->pr_flags & PR_IP6))
3359 		return (0);
3360 	mtx_lock(&pr->pr_mtx);
3361 	if (!(pr->pr_flags & PR_IP6)) {
3362 		mtx_unlock(&pr->pr_mtx);
3363 		return (0);
3364 	}
3365 	if (pr->pr_ip6 == NULL) {
3366 		mtx_unlock(&pr->pr_mtx);
3367 		return (EAFNOSUPPORT);
3368 	}
3369 
3370 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3371 		bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3372 		mtx_unlock(&pr->pr_mtx);
3373 		return (0);
3374 	}
3375 
3376 	/*
3377 	 * Return success because nothing had to be changed.
3378 	 */
3379 	mtx_unlock(&pr->pr_mtx);
3380 	return (0);
3381 }
3382 
3383 /*
3384  * Check if given address belongs to the jail referenced by cred/prison.
3385  *
3386  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3387  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3388  * doesn't allow IPv6.
3389  */
3390 static int
3391 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3392 {
3393 	int i, a, z, d;
3394 
3395 	/*
3396 	 * Check the primary IP.
3397 	 */
3398 	if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3399 		return (0);
3400 
3401 	/*
3402 	 * All the other IPs are sorted so we can do a binary search.
3403 	 */
3404 	a = 0;
3405 	z = pr->pr_ip6s - 2;
3406 	while (a <= z) {
3407 		i = (a + z) / 2;
3408 		d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3409 		if (d > 0)
3410 			z = i - 1;
3411 		else if (d < 0)
3412 			a = i + 1;
3413 		else
3414 			return (0);
3415 	}
3416 
3417 	return (EADDRNOTAVAIL);
3418 }
3419 
3420 int
3421 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3422 {
3423 	struct prison *pr;
3424 	int error;
3425 
3426 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3427 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3428 
3429 	pr = cred->cr_prison;
3430 	if (!(pr->pr_flags & PR_IP6))
3431 		return (0);
3432 	mtx_lock(&pr->pr_mtx);
3433 	if (!(pr->pr_flags & PR_IP6)) {
3434 		mtx_unlock(&pr->pr_mtx);
3435 		return (0);
3436 	}
3437 	if (pr->pr_ip6 == NULL) {
3438 		mtx_unlock(&pr->pr_mtx);
3439 		return (EAFNOSUPPORT);
3440 	}
3441 
3442 	error = _prison_check_ip6(pr, ia6);
3443 	mtx_unlock(&pr->pr_mtx);
3444 	return (error);
3445 }
3446 #endif
3447 
3448 /*
3449  * Check if a jail supports the given address family.
3450  *
3451  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3452  * if not.
3453  */
3454 int
3455 prison_check_af(struct ucred *cred, int af)
3456 {
3457 	struct prison *pr;
3458 	int error;
3459 
3460 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3461 
3462 	pr = cred->cr_prison;
3463 #ifdef VIMAGE
3464 	/* Prisons with their own network stack are not limited. */
3465 	if (prison_owns_vnet(cred))
3466 		return (0);
3467 #endif
3468 
3469 	error = 0;
3470 	switch (af)
3471 	{
3472 #ifdef INET
3473 	case AF_INET:
3474 		if (pr->pr_flags & PR_IP4)
3475 		{
3476 			mtx_lock(&pr->pr_mtx);
3477 			if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3478 				error = EAFNOSUPPORT;
3479 			mtx_unlock(&pr->pr_mtx);
3480 		}
3481 		break;
3482 #endif
3483 #ifdef INET6
3484 	case AF_INET6:
3485 		if (pr->pr_flags & PR_IP6)
3486 		{
3487 			mtx_lock(&pr->pr_mtx);
3488 			if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3489 				error = EAFNOSUPPORT;
3490 			mtx_unlock(&pr->pr_mtx);
3491 		}
3492 		break;
3493 #endif
3494 	case AF_LOCAL:
3495 	case AF_ROUTE:
3496 		break;
3497 	default:
3498 		if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3499 			error = EAFNOSUPPORT;
3500 	}
3501 	return (error);
3502 }
3503 
3504 /*
3505  * Check if given address belongs to the jail referenced by cred (wrapper to
3506  * prison_check_ip[46]).
3507  *
3508  * Returns 0 if jail doesn't restrict the address family or if address belongs
3509  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3510  * the jail doesn't allow the address family.  IPv4 Address passed in in NBO.
3511  */
3512 int
3513 prison_if(struct ucred *cred, struct sockaddr *sa)
3514 {
3515 #ifdef INET
3516 	struct sockaddr_in *sai;
3517 #endif
3518 #ifdef INET6
3519 	struct sockaddr_in6 *sai6;
3520 #endif
3521 	int error;
3522 
3523 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3524 	KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3525 
3526 #ifdef VIMAGE
3527 	if (prison_owns_vnet(cred))
3528 		return (0);
3529 #endif
3530 
3531 	error = 0;
3532 	switch (sa->sa_family)
3533 	{
3534 #ifdef INET
3535 	case AF_INET:
3536 		sai = (struct sockaddr_in *)sa;
3537 		error = prison_check_ip4(cred, &sai->sin_addr);
3538 		break;
3539 #endif
3540 #ifdef INET6
3541 	case AF_INET6:
3542 		sai6 = (struct sockaddr_in6 *)sa;
3543 		error = prison_check_ip6(cred, &sai6->sin6_addr);
3544 		break;
3545 #endif
3546 	default:
3547 		if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3548 			error = EAFNOSUPPORT;
3549 	}
3550 	return (error);
3551 }
3552 
3553 /*
3554  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3555  */
3556 int
3557 prison_check(struct ucred *cred1, struct ucred *cred2)
3558 {
3559 
3560 	return ((cred1->cr_prison == cred2->cr_prison ||
3561 	    prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3562 }
3563 
3564 /*
3565  * Return 1 if p2 is a child of p1, otherwise 0.
3566  */
3567 int
3568 prison_ischild(struct prison *pr1, struct prison *pr2)
3569 {
3570 
3571 	for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3572 		if (pr1 == pr2)
3573 			return (1);
3574 	return (0);
3575 }
3576 
3577 /*
3578  * Return 1 if the passed credential is in a jail, otherwise 0.
3579  */
3580 int
3581 jailed(struct ucred *cred)
3582 {
3583 
3584 	return (cred->cr_prison != &prison0);
3585 }
3586 
3587 /*
3588  * Return 1 if the passed credential is in a jail and that jail does not
3589  * have its own virtual network stack, otherwise 0.
3590  */
3591 int
3592 jailed_without_vnet(struct ucred *cred)
3593 {
3594 
3595 	if (!jailed(cred))
3596 		return (0);
3597 #ifdef VIMAGE
3598 	if (prison_owns_vnet(cred))
3599 		return (0);
3600 #endif
3601 
3602 	return (1);
3603 }
3604 
3605 /*
3606  * Return the correct hostname (domainname, et al) for the passed credential.
3607  */
3608 void
3609 getcredhostname(struct ucred *cred, char *buf, size_t size)
3610 {
3611 	struct prison *pr;
3612 
3613 	/*
3614 	 * A NULL credential can be used to shortcut to the physical
3615 	 * system's hostname.
3616 	 */
3617 	pr = (cred != NULL) ? cred->cr_prison : &prison0;
3618 	mtx_lock(&pr->pr_mtx);
3619 	strlcpy(buf, pr->pr_hostname, size);
3620 	mtx_unlock(&pr->pr_mtx);
3621 }
3622 
3623 void
3624 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3625 {
3626 
3627 	mtx_lock(&cred->cr_prison->pr_mtx);
3628 	strlcpy(buf, cred->cr_prison->pr_domainname, size);
3629 	mtx_unlock(&cred->cr_prison->pr_mtx);
3630 }
3631 
3632 void
3633 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3634 {
3635 
3636 	mtx_lock(&cred->cr_prison->pr_mtx);
3637 	strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3638 	mtx_unlock(&cred->cr_prison->pr_mtx);
3639 }
3640 
3641 void
3642 getcredhostid(struct ucred *cred, unsigned long *hostid)
3643 {
3644 
3645 	mtx_lock(&cred->cr_prison->pr_mtx);
3646 	*hostid = cred->cr_prison->pr_hostid;
3647 	mtx_unlock(&cred->cr_prison->pr_mtx);
3648 }
3649 
3650 #ifdef VIMAGE
3651 /*
3652  * Determine whether the prison represented by cred owns
3653  * its vnet rather than having it inherited.
3654  *
3655  * Returns 1 in case the prison owns the vnet, 0 otherwise.
3656  */
3657 int
3658 prison_owns_vnet(struct ucred *cred)
3659 {
3660 
3661 	/*
3662 	 * vnets cannot be added/removed after jail creation,
3663 	 * so no need to lock here.
3664 	 */
3665 	return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3666 }
3667 #endif
3668 
3669 /*
3670  * Determine whether the subject represented by cred can "see"
3671  * status of a mount point.
3672  * Returns: 0 for permitted, ENOENT otherwise.
3673  * XXX: This function should be called cr_canseemount() and should be
3674  *      placed in kern_prot.c.
3675  */
3676 int
3677 prison_canseemount(struct ucred *cred, struct mount *mp)
3678 {
3679 	struct prison *pr;
3680 	struct statfs *sp;
3681 	size_t len;
3682 
3683 	pr = cred->cr_prison;
3684 	if (pr->pr_enforce_statfs == 0)
3685 		return (0);
3686 	if (pr->pr_root->v_mount == mp)
3687 		return (0);
3688 	if (pr->pr_enforce_statfs == 2)
3689 		return (ENOENT);
3690 	/*
3691 	 * If jail's chroot directory is set to "/" we should be able to see
3692 	 * all mount-points from inside a jail.
3693 	 * This is ugly check, but this is the only situation when jail's
3694 	 * directory ends with '/'.
3695 	 */
3696 	if (strcmp(pr->pr_path, "/") == 0)
3697 		return (0);
3698 	len = strlen(pr->pr_path);
3699 	sp = &mp->mnt_stat;
3700 	if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3701 		return (ENOENT);
3702 	/*
3703 	 * Be sure that we don't have situation where jail's root directory
3704 	 * is "/some/path" and mount point is "/some/pathpath".
3705 	 */
3706 	if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3707 		return (ENOENT);
3708 	return (0);
3709 }
3710 
3711 void
3712 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3713 {
3714 	char jpath[MAXPATHLEN];
3715 	struct prison *pr;
3716 	size_t len;
3717 
3718 	pr = cred->cr_prison;
3719 	if (pr->pr_enforce_statfs == 0)
3720 		return;
3721 	if (prison_canseemount(cred, mp) != 0) {
3722 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3723 		strlcpy(sp->f_mntonname, "[restricted]",
3724 		    sizeof(sp->f_mntonname));
3725 		return;
3726 	}
3727 	if (pr->pr_root->v_mount == mp) {
3728 		/*
3729 		 * Clear current buffer data, so we are sure nothing from
3730 		 * the valid path left there.
3731 		 */
3732 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3733 		*sp->f_mntonname = '/';
3734 		return;
3735 	}
3736 	/*
3737 	 * If jail's chroot directory is set to "/" we should be able to see
3738 	 * all mount-points from inside a jail.
3739 	 */
3740 	if (strcmp(pr->pr_path, "/") == 0)
3741 		return;
3742 	len = strlen(pr->pr_path);
3743 	strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3744 	/*
3745 	 * Clear current buffer data, so we are sure nothing from
3746 	 * the valid path left there.
3747 	 */
3748 	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3749 	if (*jpath == '\0') {
3750 		/* Should never happen. */
3751 		*sp->f_mntonname = '/';
3752 	} else {
3753 		strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3754 	}
3755 }
3756 
3757 /*
3758  * Check with permission for a specific privilege is granted within jail.  We
3759  * have a specific list of accepted privileges; the rest are denied.
3760  */
3761 int
3762 prison_priv_check(struct ucred *cred, int priv)
3763 {
3764 
3765 	if (!jailed(cred))
3766 		return (0);
3767 
3768 #ifdef VIMAGE
3769 	/*
3770 	 * Privileges specific to prisons with a virtual network stack.
3771 	 * There might be a duplicate entry here in case the privilege
3772 	 * is only granted conditionally in the legacy jail case.
3773 	 */
3774 	switch (priv) {
3775 #ifdef notyet
3776 		/*
3777 		 * NFS-specific privileges.
3778 		 */
3779 	case PRIV_NFS_DAEMON:
3780 	case PRIV_NFS_LOCKD:
3781 #endif
3782 		/*
3783 		 * Network stack privileges.
3784 		 */
3785 	case PRIV_NET_BRIDGE:
3786 	case PRIV_NET_GRE:
3787 	case PRIV_NET_BPF:
3788 	case PRIV_NET_RAW:		/* Dup, cond. in legacy jail case. */
3789 	case PRIV_NET_ROUTE:
3790 	case PRIV_NET_TAP:
3791 	case PRIV_NET_SETIFMTU:
3792 	case PRIV_NET_SETIFFLAGS:
3793 	case PRIV_NET_SETIFCAP:
3794 	case PRIV_NET_SETIFDESCR:
3795 	case PRIV_NET_SETIFNAME	:
3796 	case PRIV_NET_SETIFMETRIC:
3797 	case PRIV_NET_SETIFPHYS:
3798 	case PRIV_NET_SETIFMAC:
3799 	case PRIV_NET_ADDMULTI:
3800 	case PRIV_NET_DELMULTI:
3801 	case PRIV_NET_HWIOCTL:
3802 	case PRIV_NET_SETLLADDR:
3803 	case PRIV_NET_ADDIFGROUP:
3804 	case PRIV_NET_DELIFGROUP:
3805 	case PRIV_NET_IFCREATE:
3806 	case PRIV_NET_IFDESTROY:
3807 	case PRIV_NET_ADDIFADDR:
3808 	case PRIV_NET_DELIFADDR:
3809 	case PRIV_NET_LAGG:
3810 	case PRIV_NET_GIF:
3811 	case PRIV_NET_SETIFVNET:
3812 	case PRIV_NET_SETIFFIB:
3813 
3814 		/*
3815 		 * 802.11-related privileges.
3816 		 */
3817 	case PRIV_NET80211_GETKEY:
3818 #ifdef notyet
3819 	case PRIV_NET80211_MANAGE:		/* XXX-BZ discuss with sam@ */
3820 #endif
3821 
3822 #ifdef notyet
3823 		/*
3824 		 * ATM privileges.
3825 		 */
3826 	case PRIV_NETATM_CFG:
3827 	case PRIV_NETATM_ADD:
3828 	case PRIV_NETATM_DEL:
3829 	case PRIV_NETATM_SET:
3830 
3831 		/*
3832 		 * Bluetooth privileges.
3833 		 */
3834 	case PRIV_NETBLUETOOTH_RAW:
3835 #endif
3836 
3837 		/*
3838 		 * Netgraph and netgraph module privileges.
3839 		 */
3840 	case PRIV_NETGRAPH_CONTROL:
3841 #ifdef notyet
3842 	case PRIV_NETGRAPH_TTY:
3843 #endif
3844 
3845 		/*
3846 		 * IPv4 and IPv6 privileges.
3847 		 */
3848 	case PRIV_NETINET_IPFW:
3849 	case PRIV_NETINET_DIVERT:
3850 	case PRIV_NETINET_PF:
3851 	case PRIV_NETINET_DUMMYNET:
3852 	case PRIV_NETINET_CARP:
3853 	case PRIV_NETINET_MROUTE:
3854 	case PRIV_NETINET_RAW:
3855 	case PRIV_NETINET_ADDRCTRL6:
3856 	case PRIV_NETINET_ND6:
3857 	case PRIV_NETINET_SCOPE6:
3858 	case PRIV_NETINET_ALIFETIME6:
3859 	case PRIV_NETINET_IPSEC:
3860 	case PRIV_NETINET_BINDANY:
3861 
3862 #ifdef notyet
3863 		/*
3864 		 * NCP privileges.
3865 		 */
3866 	case PRIV_NETNCP:
3867 
3868 		/*
3869 		 * SMB privileges.
3870 		 */
3871 	case PRIV_NETSMB:
3872 #endif
3873 
3874 	/*
3875 	 * No default: or deny here.
3876 	 * In case of no permit fall through to next switch().
3877 	 */
3878 		if (cred->cr_prison->pr_flags & PR_VNET)
3879 			return (0);
3880 	}
3881 #endif /* VIMAGE */
3882 
3883 	switch (priv) {
3884 
3885 		/*
3886 		 * Allow ktrace privileges for root in jail.
3887 		 */
3888 	case PRIV_KTRACE:
3889 
3890 #if 0
3891 		/*
3892 		 * Allow jailed processes to configure audit identity and
3893 		 * submit audit records (login, etc).  In the future we may
3894 		 * want to further refine the relationship between audit and
3895 		 * jail.
3896 		 */
3897 	case PRIV_AUDIT_GETAUDIT:
3898 	case PRIV_AUDIT_SETAUDIT:
3899 	case PRIV_AUDIT_SUBMIT:
3900 #endif
3901 
3902 		/*
3903 		 * Allow jailed processes to manipulate process UNIX
3904 		 * credentials in any way they see fit.
3905 		 */
3906 	case PRIV_CRED_SETUID:
3907 	case PRIV_CRED_SETEUID:
3908 	case PRIV_CRED_SETGID:
3909 	case PRIV_CRED_SETEGID:
3910 	case PRIV_CRED_SETGROUPS:
3911 	case PRIV_CRED_SETREUID:
3912 	case PRIV_CRED_SETREGID:
3913 	case PRIV_CRED_SETRESUID:
3914 	case PRIV_CRED_SETRESGID:
3915 
3916 		/*
3917 		 * Jail implements visibility constraints already, so allow
3918 		 * jailed root to override uid/gid-based constraints.
3919 		 */
3920 	case PRIV_SEEOTHERGIDS:
3921 	case PRIV_SEEOTHERUIDS:
3922 
3923 		/*
3924 		 * Jail implements inter-process debugging limits already, so
3925 		 * allow jailed root various debugging privileges.
3926 		 */
3927 	case PRIV_DEBUG_DIFFCRED:
3928 	case PRIV_DEBUG_SUGID:
3929 	case PRIV_DEBUG_UNPRIV:
3930 
3931 		/*
3932 		 * Allow jail to set various resource limits and login
3933 		 * properties, and for now, exceed process resource limits.
3934 		 */
3935 	case PRIV_PROC_LIMIT:
3936 	case PRIV_PROC_SETLOGIN:
3937 	case PRIV_PROC_SETRLIMIT:
3938 
3939 		/*
3940 		 * System V and POSIX IPC privileges are granted in jail.
3941 		 */
3942 	case PRIV_IPC_READ:
3943 	case PRIV_IPC_WRITE:
3944 	case PRIV_IPC_ADMIN:
3945 	case PRIV_IPC_MSGSIZE:
3946 	case PRIV_MQ_ADMIN:
3947 
3948 		/*
3949 		 * Jail operations within a jail work on child jails.
3950 		 */
3951 	case PRIV_JAIL_ATTACH:
3952 	case PRIV_JAIL_SET:
3953 	case PRIV_JAIL_REMOVE:
3954 
3955 		/*
3956 		 * Jail implements its own inter-process limits, so allow
3957 		 * root processes in jail to change scheduling on other
3958 		 * processes in the same jail.  Likewise for signalling.
3959 		 */
3960 	case PRIV_SCHED_DIFFCRED:
3961 	case PRIV_SCHED_CPUSET:
3962 	case PRIV_SIGNAL_DIFFCRED:
3963 	case PRIV_SIGNAL_SUGID:
3964 
3965 		/*
3966 		 * Allow jailed processes to write to sysctls marked as jail
3967 		 * writable.
3968 		 */
3969 	case PRIV_SYSCTL_WRITEJAIL:
3970 
3971 		/*
3972 		 * Allow root in jail to manage a variety of quota
3973 		 * properties.  These should likely be conditional on a
3974 		 * configuration option.
3975 		 */
3976 	case PRIV_VFS_GETQUOTA:
3977 	case PRIV_VFS_SETQUOTA:
3978 
3979 		/*
3980 		 * Since Jail relies on chroot() to implement file system
3981 		 * protections, grant many VFS privileges to root in jail.
3982 		 * Be careful to exclude mount-related and NFS-related
3983 		 * privileges.
3984 		 */
3985 	case PRIV_VFS_READ:
3986 	case PRIV_VFS_WRITE:
3987 	case PRIV_VFS_ADMIN:
3988 	case PRIV_VFS_EXEC:
3989 	case PRIV_VFS_LOOKUP:
3990 	case PRIV_VFS_BLOCKRESERVE:	/* XXXRW: Slightly surprising. */
3991 	case PRIV_VFS_CHFLAGS_DEV:
3992 	case PRIV_VFS_CHOWN:
3993 	case PRIV_VFS_CHROOT:
3994 	case PRIV_VFS_RETAINSUGID:
3995 	case PRIV_VFS_FCHROOT:
3996 	case PRIV_VFS_LINK:
3997 	case PRIV_VFS_SETGID:
3998 	case PRIV_VFS_STAT:
3999 	case PRIV_VFS_STICKYFILE:
4000 
4001 		/*
4002 		 * As in the non-jail case, non-root users are expected to be
4003 		 * able to read kernel/phyiscal memory (provided /dev/[k]mem
4004 		 * exists in the jail and they have permission to access it).
4005 		 */
4006 	case PRIV_KMEM_READ:
4007 		return (0);
4008 
4009 		/*
4010 		 * Depending on the global setting, allow privilege of
4011 		 * setting system flags.
4012 		 */
4013 	case PRIV_VFS_SYSFLAGS:
4014 		if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
4015 			return (0);
4016 		else
4017 			return (EPERM);
4018 
4019 		/*
4020 		 * Depending on the global setting, allow privilege of
4021 		 * mounting/unmounting file systems.
4022 		 */
4023 	case PRIV_VFS_MOUNT:
4024 	case PRIV_VFS_UNMOUNT:
4025 	case PRIV_VFS_MOUNT_NONUSER:
4026 	case PRIV_VFS_MOUNT_OWNER:
4027 		if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
4028 		    cred->cr_prison->pr_enforce_statfs < 2)
4029 			return (0);
4030 		else
4031 			return (EPERM);
4032 
4033 		/*
4034 		 * Allow jailed root to bind reserved ports and reuse in-use
4035 		 * ports.
4036 		 */
4037 	case PRIV_NETINET_RESERVEDPORT:
4038 	case PRIV_NETINET_REUSEPORT:
4039 		return (0);
4040 
4041 		/*
4042 		 * Allow jailed root to set certain IPv4/6 (option) headers.
4043 		 */
4044 	case PRIV_NETINET_SETHDROPTS:
4045 		return (0);
4046 
4047 		/*
4048 		 * Conditionally allow creating raw sockets in jail.
4049 		 */
4050 	case PRIV_NETINET_RAW:
4051 		if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4052 			return (0);
4053 		else
4054 			return (EPERM);
4055 
4056 		/*
4057 		 * Since jail implements its own visibility limits on netstat
4058 		 * sysctls, allow getcred.  This allows identd to work in
4059 		 * jail.
4060 		 */
4061 	case PRIV_NETINET_GETCRED:
4062 		return (0);
4063 
4064 		/*
4065 		 * Allow jailed root to set loginclass.
4066 		 */
4067 	case PRIV_PROC_SETLOGINCLASS:
4068 		return (0);
4069 
4070 	default:
4071 		/*
4072 		 * In all remaining cases, deny the privilege request.  This
4073 		 * includes almost all network privileges, many system
4074 		 * configuration privileges.
4075 		 */
4076 		return (EPERM);
4077 	}
4078 }
4079 
4080 /*
4081  * Return the part of pr2's name that is relative to pr1, or the whole name
4082  * if it does not directly follow.
4083  */
4084 
4085 char *
4086 prison_name(struct prison *pr1, struct prison *pr2)
4087 {
4088 	char *name;
4089 
4090 	/* Jails see themselves as "0" (if they see themselves at all). */
4091 	if (pr1 == pr2)
4092 		return "0";
4093 	name = pr2->pr_name;
4094 	if (prison_ischild(pr1, pr2)) {
4095 		/*
4096 		 * pr1 isn't locked (and allprison_lock may not be either)
4097 		 * so its length can't be counted on.  But the number of dots
4098 		 * can be counted on - and counted.
4099 		 */
4100 		for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4101 			name = strchr(name, '.') + 1;
4102 	}
4103 	return (name);
4104 }
4105 
4106 /*
4107  * Return the part of pr2's path that is relative to pr1, or the whole path
4108  * if it does not directly follow.
4109  */
4110 static char *
4111 prison_path(struct prison *pr1, struct prison *pr2)
4112 {
4113 	char *path1, *path2;
4114 	int len1;
4115 
4116 	path1 = pr1->pr_path;
4117 	path2 = pr2->pr_path;
4118 	if (!strcmp(path1, "/"))
4119 		return (path2);
4120 	len1 = strlen(path1);
4121 	if (strncmp(path1, path2, len1))
4122 		return (path2);
4123 	if (path2[len1] == '\0')
4124 		return "/";
4125 	if (path2[len1] == '/')
4126 		return (path2 + len1);
4127 	return (path2);
4128 }
4129 
4130 
4131 /*
4132  * Jail-related sysctls.
4133  */
4134 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4135     "Jails");
4136 
4137 static int
4138 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4139 {
4140 	struct xprison *xp;
4141 	struct prison *pr, *cpr;
4142 #ifdef INET
4143 	struct in_addr *ip4 = NULL;
4144 	int ip4s = 0;
4145 #endif
4146 #ifdef INET6
4147 	struct in6_addr *ip6 = NULL;
4148 	int ip6s = 0;
4149 #endif
4150 	int descend, error;
4151 
4152 	xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4153 	pr = req->td->td_ucred->cr_prison;
4154 	error = 0;
4155 	sx_slock(&allprison_lock);
4156 	FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4157 #if defined(INET) || defined(INET6)
4158  again:
4159 #endif
4160 		mtx_lock(&cpr->pr_mtx);
4161 #ifdef INET
4162 		if (cpr->pr_ip4s > 0) {
4163 			if (ip4s < cpr->pr_ip4s) {
4164 				ip4s = cpr->pr_ip4s;
4165 				mtx_unlock(&cpr->pr_mtx);
4166 				ip4 = realloc(ip4, ip4s *
4167 				    sizeof(struct in_addr), M_TEMP, M_WAITOK);
4168 				goto again;
4169 			}
4170 			bcopy(cpr->pr_ip4, ip4,
4171 			    cpr->pr_ip4s * sizeof(struct in_addr));
4172 		}
4173 #endif
4174 #ifdef INET6
4175 		if (cpr->pr_ip6s > 0) {
4176 			if (ip6s < cpr->pr_ip6s) {
4177 				ip6s = cpr->pr_ip6s;
4178 				mtx_unlock(&cpr->pr_mtx);
4179 				ip6 = realloc(ip6, ip6s *
4180 				    sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4181 				goto again;
4182 			}
4183 			bcopy(cpr->pr_ip6, ip6,
4184 			    cpr->pr_ip6s * sizeof(struct in6_addr));
4185 		}
4186 #endif
4187 		if (cpr->pr_ref == 0) {
4188 			mtx_unlock(&cpr->pr_mtx);
4189 			continue;
4190 		}
4191 		bzero(xp, sizeof(*xp));
4192 		xp->pr_version = XPRISON_VERSION;
4193 		xp->pr_id = cpr->pr_id;
4194 		xp->pr_state = cpr->pr_uref > 0
4195 		    ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4196 		strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4197 		strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4198 		strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4199 #ifdef INET
4200 		xp->pr_ip4s = cpr->pr_ip4s;
4201 #endif
4202 #ifdef INET6
4203 		xp->pr_ip6s = cpr->pr_ip6s;
4204 #endif
4205 		mtx_unlock(&cpr->pr_mtx);
4206 		error = SYSCTL_OUT(req, xp, sizeof(*xp));
4207 		if (error)
4208 			break;
4209 #ifdef INET
4210 		if (xp->pr_ip4s > 0) {
4211 			error = SYSCTL_OUT(req, ip4,
4212 			    xp->pr_ip4s * sizeof(struct in_addr));
4213 			if (error)
4214 				break;
4215 		}
4216 #endif
4217 #ifdef INET6
4218 		if (xp->pr_ip6s > 0) {
4219 			error = SYSCTL_OUT(req, ip6,
4220 			    xp->pr_ip6s * sizeof(struct in6_addr));
4221 			if (error)
4222 				break;
4223 		}
4224 #endif
4225 	}
4226 	sx_sunlock(&allprison_lock);
4227 	free(xp, M_TEMP);
4228 #ifdef INET
4229 	free(ip4, M_TEMP);
4230 #endif
4231 #ifdef INET6
4232 	free(ip6, M_TEMP);
4233 #endif
4234 	return (error);
4235 }
4236 
4237 SYSCTL_OID(_security_jail, OID_AUTO, list,
4238     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4239     sysctl_jail_list, "S", "List of active jails");
4240 
4241 static int
4242 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4243 {
4244 	int error, injail;
4245 
4246 	injail = jailed(req->td->td_ucred);
4247 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
4248 
4249 	return (error);
4250 }
4251 
4252 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4253     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4254     sysctl_jail_jailed, "I", "Process in jail?");
4255 
4256 static int
4257 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4258 {
4259 	int error, havevnet;
4260 #ifdef VIMAGE
4261 	struct ucred *cred = req->td->td_ucred;
4262 
4263 	havevnet = jailed(cred) && prison_owns_vnet(cred);
4264 #else
4265 	havevnet = 0;
4266 #endif
4267 	error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4268 
4269 	return (error);
4270 }
4271 
4272 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4273     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4274     sysctl_jail_vnet, "I", "Jail owns VNET?");
4275 
4276 #if defined(INET) || defined(INET6)
4277 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4278     &jail_max_af_ips, 0,
4279     "Number of IP addresses a jail may have at most per address family");
4280 #endif
4281 
4282 /*
4283  * Default parameters for jail(2) compatibility.  For historical reasons,
4284  * the sysctl names have varying similarity to the parameter names.  Prisons
4285  * just see their own parameters, and can't change them.
4286  */
4287 static int
4288 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4289 {
4290 	struct prison *pr;
4291 	int allow, error, i;
4292 
4293 	pr = req->td->td_ucred->cr_prison;
4294 	allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4295 
4296 	/* Get the current flag value, and convert it to a boolean. */
4297 	i = (allow & arg2) ? 1 : 0;
4298 	if (arg1 != NULL)
4299 		i = !i;
4300 	error = sysctl_handle_int(oidp, &i, 0, req);
4301 	if (error || !req->newptr)
4302 		return (error);
4303 	i = i ? arg2 : 0;
4304 	if (arg1 != NULL)
4305 		i ^= arg2;
4306 	/*
4307 	 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4308 	 * for writing.
4309 	 */
4310 	mtx_lock(&prison0.pr_mtx);
4311 	jail_default_allow = (jail_default_allow & ~arg2) | i;
4312 	mtx_unlock(&prison0.pr_mtx);
4313 	return (0);
4314 }
4315 
4316 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4317     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4318     NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4319     "Processes in jail can set their hostnames");
4320 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4321     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4322     (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4323     "Processes in jail are limited to creating UNIX/IP/route sockets only");
4324 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4325     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4326     NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4327     "Processes in jail can use System V IPC primitives");
4328 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4329     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4330     NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4331     "Prison root can create raw sockets");
4332 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4333     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4334     NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4335     "Processes in jail can alter system file flags");
4336 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4337     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4338     NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4339     "Processes in jail can mount/unmount jail-friendly file systems");
4340 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4341     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4342     NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4343     "Processes in jail can mount the devfs file system");
4344 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4345     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4346     NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4347     "Processes in jail can mount the fdescfs file system");
4348 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4349     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4350     NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4351     "Processes in jail can mount the nullfs file system");
4352 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4353     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4354     NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4355     "Processes in jail can mount the procfs file system");
4356 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
4357     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4358     NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
4359     "Processes in jail can mount the linprocfs file system");
4360 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
4361     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4362     NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
4363     "Processes in jail can mount the linsysfs file system");
4364 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4365     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4366     NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4367     "Processes in jail can mount the tmpfs file system");
4368 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4369     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4370     NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4371     "Processes in jail can mount the zfs file system");
4372 
4373 static int
4374 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4375 {
4376 	struct prison *pr;
4377 	int level, error;
4378 
4379 	pr = req->td->td_ucred->cr_prison;
4380 	level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4381 	error = sysctl_handle_int(oidp, &level, 0, req);
4382 	if (error || !req->newptr)
4383 		return (error);
4384 	*(int *)arg1 = level;
4385 	return (0);
4386 }
4387 
4388 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4389     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4390     &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4391     sysctl_jail_default_level, "I",
4392     "Processes in jail cannot see all mounted file systems");
4393 
4394 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4395     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4396     &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4397     sysctl_jail_default_level, "I",
4398     "Ruleset for the devfs filesystem in jail");
4399 
4400 /*
4401  * Nodes to describe jail parameters.  Maximum length of string parameters
4402  * is returned in the string itself, and the other parameters exist merely
4403  * to make themselves and their types known.
4404  */
4405 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4406     "Jail parameters");
4407 
4408 int
4409 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4410 {
4411 	int i;
4412 	long l;
4413 	size_t s;
4414 	char numbuf[12];
4415 
4416 	switch (oidp->oid_kind & CTLTYPE)
4417 	{
4418 	case CTLTYPE_LONG:
4419 	case CTLTYPE_ULONG:
4420 		l = 0;
4421 #ifdef SCTL_MASK32
4422 		if (!(req->flags & SCTL_MASK32))
4423 #endif
4424 			return (SYSCTL_OUT(req, &l, sizeof(l)));
4425 	case CTLTYPE_INT:
4426 	case CTLTYPE_UINT:
4427 		i = 0;
4428 		return (SYSCTL_OUT(req, &i, sizeof(i)));
4429 	case CTLTYPE_STRING:
4430 		snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4431 		return
4432 		    (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4433 	case CTLTYPE_STRUCT:
4434 		s = (size_t)arg2;
4435 		return (SYSCTL_OUT(req, &s, sizeof(s)));
4436 	}
4437 	return (0);
4438 }
4439 
4440 /*
4441  * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4442  * jail creation time but cannot be changed in an existing jail.
4443  */
4444 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4445 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4446 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4447 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4448 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4449     "I", "Jail secure level");
4450 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4451     "Jail value for kern.osreldate and uname -K");
4452 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4453     "Jail value for kern.osrelease and uname -r");
4454 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4455     "I", "Jail cannot see all mounted file systems");
4456 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4457     "I", "Ruleset for in-jail devfs mounts");
4458 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4459     "B", "Jail persistence");
4460 #ifdef VIMAGE
4461 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4462     "E,jailsys", "Virtual network stack");
4463 #endif
4464 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4465     "B", "Jail is in the process of shutting down");
4466 
4467 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4468 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4469     "I", "Current number of child jails");
4470 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4471     "I", "Maximum number of child jails");
4472 
4473 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4474 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4475     "Jail hostname");
4476 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4477     "Jail NIS domainname");
4478 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4479     "Jail host UUID");
4480 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4481     "LU", "Jail host ID");
4482 
4483 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4484 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4485 
4486 #ifdef INET
4487 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4488     "Jail IPv4 address virtualization");
4489 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4490     "S,in_addr,a", "Jail IPv4 addresses");
4491 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4492     "B", "Do (not) use IPv4 source address selection rather than the "
4493     "primary jail IPv4 address.");
4494 #endif
4495 #ifdef INET6
4496 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4497     "Jail IPv6 address virtualization");
4498 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4499     "S,in6_addr,a", "Jail IPv6 addresses");
4500 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4501     "B", "Do (not) use IPv6 source address selection rather than the "
4502     "primary jail IPv6 address.");
4503 #endif
4504 
4505 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4506 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4507     "B", "Jail may set hostname");
4508 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4509     "B", "Jail may use SYSV IPC");
4510 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4511     "B", "Jail may create raw sockets");
4512 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4513     "B", "Jail may alter system file flags");
4514 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4515     "B", "Jail may set file quotas");
4516 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4517     "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4518 
4519 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4520 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4521     "B", "Jail may mount/unmount jail-friendly file systems in general");
4522 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4523     "B", "Jail may mount the devfs file system");
4524 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4525     "B", "Jail may mount the fdescfs file system");
4526 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4527     "B", "Jail may mount the nullfs file system");
4528 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4529     "B", "Jail may mount the procfs file system");
4530 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
4531     "B", "Jail may mount the linprocfs file system");
4532 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
4533     "B", "Jail may mount the linsysfs file system");
4534 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4535     "B", "Jail may mount the tmpfs file system");
4536 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4537     "B", "Jail may mount the zfs file system");
4538 
4539 #ifdef RACCT
4540 void
4541 prison_racct_foreach(void (*callback)(struct racct *racct,
4542     void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
4543     void *arg2, void *arg3)
4544 {
4545 	struct prison_racct *prr;
4546 
4547 	ASSERT_RACCT_ENABLED();
4548 
4549 	sx_slock(&allprison_lock);
4550 	if (pre != NULL)
4551 		(pre)();
4552 	LIST_FOREACH(prr, &allprison_racct, prr_next)
4553 		(callback)(prr->prr_racct, arg2, arg3);
4554 	if (post != NULL)
4555 		(post)();
4556 	sx_sunlock(&allprison_lock);
4557 }
4558 
4559 static struct prison_racct *
4560 prison_racct_find_locked(const char *name)
4561 {
4562 	struct prison_racct *prr;
4563 
4564 	ASSERT_RACCT_ENABLED();
4565 	sx_assert(&allprison_lock, SA_XLOCKED);
4566 
4567 	if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4568 		return (NULL);
4569 
4570 	LIST_FOREACH(prr, &allprison_racct, prr_next) {
4571 		if (strcmp(name, prr->prr_name) != 0)
4572 			continue;
4573 
4574 		/* Found prison_racct with a matching name? */
4575 		prison_racct_hold(prr);
4576 		return (prr);
4577 	}
4578 
4579 	/* Add new prison_racct. */
4580 	prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4581 	racct_create(&prr->prr_racct);
4582 
4583 	strcpy(prr->prr_name, name);
4584 	refcount_init(&prr->prr_refcount, 1);
4585 	LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4586 
4587 	return (prr);
4588 }
4589 
4590 struct prison_racct *
4591 prison_racct_find(const char *name)
4592 {
4593 	struct prison_racct *prr;
4594 
4595 	ASSERT_RACCT_ENABLED();
4596 
4597 	sx_xlock(&allprison_lock);
4598 	prr = prison_racct_find_locked(name);
4599 	sx_xunlock(&allprison_lock);
4600 	return (prr);
4601 }
4602 
4603 void
4604 prison_racct_hold(struct prison_racct *prr)
4605 {
4606 
4607 	ASSERT_RACCT_ENABLED();
4608 
4609 	refcount_acquire(&prr->prr_refcount);
4610 }
4611 
4612 static void
4613 prison_racct_free_locked(struct prison_racct *prr)
4614 {
4615 
4616 	ASSERT_RACCT_ENABLED();
4617 	sx_assert(&allprison_lock, SA_XLOCKED);
4618 
4619 	if (refcount_release(&prr->prr_refcount)) {
4620 		racct_destroy(&prr->prr_racct);
4621 		LIST_REMOVE(prr, prr_next);
4622 		free(prr, M_PRISON_RACCT);
4623 	}
4624 }
4625 
4626 void
4627 prison_racct_free(struct prison_racct *prr)
4628 {
4629 	int old;
4630 
4631 	ASSERT_RACCT_ENABLED();
4632 	sx_assert(&allprison_lock, SA_UNLOCKED);
4633 
4634 	old = prr->prr_refcount;
4635 	if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4636 		return;
4637 
4638 	sx_xlock(&allprison_lock);
4639 	prison_racct_free_locked(prr);
4640 	sx_xunlock(&allprison_lock);
4641 }
4642 
4643 static void
4644 prison_racct_attach(struct prison *pr)
4645 {
4646 	struct prison_racct *prr;
4647 
4648 	ASSERT_RACCT_ENABLED();
4649 	sx_assert(&allprison_lock, SA_XLOCKED);
4650 
4651 	prr = prison_racct_find_locked(pr->pr_name);
4652 	KASSERT(prr != NULL, ("cannot find prison_racct"));
4653 
4654 	pr->pr_prison_racct = prr;
4655 }
4656 
4657 /*
4658  * Handle jail renaming.  From the racct point of view, renaming means
4659  * moving from one prison_racct to another.
4660  */
4661 static void
4662 prison_racct_modify(struct prison *pr)
4663 {
4664 	struct proc *p;
4665 	struct ucred *cred;
4666 	struct prison_racct *oldprr;
4667 
4668 	ASSERT_RACCT_ENABLED();
4669 
4670 	sx_slock(&allproc_lock);
4671 	sx_xlock(&allprison_lock);
4672 
4673 	if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4674 		sx_xunlock(&allprison_lock);
4675 		sx_sunlock(&allproc_lock);
4676 		return;
4677 	}
4678 
4679 	oldprr = pr->pr_prison_racct;
4680 	pr->pr_prison_racct = NULL;
4681 
4682 	prison_racct_attach(pr);
4683 
4684 	/*
4685 	 * Move resource utilisation records.
4686 	 */
4687 	racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4688 
4689 	/*
4690 	 * Force rctl to reattach rules to processes.
4691 	 */
4692 	FOREACH_PROC_IN_SYSTEM(p) {
4693 		PROC_LOCK(p);
4694 		cred = crhold(p->p_ucred);
4695 		PROC_UNLOCK(p);
4696 		racct_proc_ucred_changed(p, cred, cred);
4697 		crfree(cred);
4698 	}
4699 
4700 	sx_sunlock(&allproc_lock);
4701 	prison_racct_free_locked(oldprr);
4702 	sx_xunlock(&allprison_lock);
4703 }
4704 
4705 static void
4706 prison_racct_detach(struct prison *pr)
4707 {
4708 
4709 	ASSERT_RACCT_ENABLED();
4710 	sx_assert(&allprison_lock, SA_UNLOCKED);
4711 
4712 	if (pr->pr_prison_racct == NULL)
4713 		return;
4714 	prison_racct_free(pr->pr_prison_racct);
4715 	pr->pr_prison_racct = NULL;
4716 }
4717 #endif /* RACCT */
4718 
4719 #ifdef DDB
4720 
4721 static void
4722 db_show_prison(struct prison *pr)
4723 {
4724 	int fi;
4725 #if defined(INET) || defined(INET6)
4726 	int ii;
4727 #endif
4728 	unsigned jsf;
4729 #ifdef INET6
4730 	char ip6buf[INET6_ADDRSTRLEN];
4731 #endif
4732 
4733 	db_printf("prison %p:\n", pr);
4734 	db_printf(" jid             = %d\n", pr->pr_id);
4735 	db_printf(" name            = %s\n", pr->pr_name);
4736 	db_printf(" parent          = %p\n", pr->pr_parent);
4737 	db_printf(" ref             = %d\n", pr->pr_ref);
4738 	db_printf(" uref            = %d\n", pr->pr_uref);
4739 	db_printf(" path            = %s\n", pr->pr_path);
4740 	db_printf(" cpuset          = %d\n", pr->pr_cpuset
4741 	    ? pr->pr_cpuset->cs_id : -1);
4742 #ifdef VIMAGE
4743 	db_printf(" vnet            = %p\n", pr->pr_vnet);
4744 #endif
4745 	db_printf(" root            = %p\n", pr->pr_root);
4746 	db_printf(" securelevel     = %d\n", pr->pr_securelevel);
4747 	db_printf(" devfs_rsnum     = %d\n", pr->pr_devfs_rsnum);
4748 	db_printf(" children.max    = %d\n", pr->pr_childmax);
4749 	db_printf(" children.cur    = %d\n", pr->pr_childcount);
4750 	db_printf(" child           = %p\n", LIST_FIRST(&pr->pr_children));
4751 	db_printf(" sibling         = %p\n", LIST_NEXT(pr, pr_sibling));
4752 	db_printf(" flags           = 0x%x", pr->pr_flags);
4753 	for (fi = 0; fi < nitems(pr_flag_names); fi++)
4754 		if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4755 			db_printf(" %s", pr_flag_names[fi]);
4756 	for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
4757 		jsf = pr->pr_flags &
4758 		    (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4759 		db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4760 		    pr_flag_jailsys[fi].disable &&
4761 		      (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4762 		    : (jsf == pr_flag_jailsys[fi].new) ? "new"
4763 		    : "inherit");
4764 	}
4765 	db_printf(" allow           = 0x%x", pr->pr_allow);
4766 	for (fi = 0; fi < nitems(pr_allow_names); fi++)
4767 		if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4768 			db_printf(" %s", pr_allow_names[fi]);
4769 	db_printf("\n");
4770 	db_printf(" enforce_statfs  = %d\n", pr->pr_enforce_statfs);
4771 	db_printf(" host.hostname   = %s\n", pr->pr_hostname);
4772 	db_printf(" host.domainname = %s\n", pr->pr_domainname);
4773 	db_printf(" host.hostuuid   = %s\n", pr->pr_hostuuid);
4774 	db_printf(" host.hostid     = %lu\n", pr->pr_hostid);
4775 #ifdef INET
4776 	db_printf(" ip4s            = %d\n", pr->pr_ip4s);
4777 	for (ii = 0; ii < pr->pr_ip4s; ii++)
4778 		db_printf(" %s %s\n",
4779 		    ii == 0 ? "ip4.addr        =" : "                 ",
4780 		    inet_ntoa(pr->pr_ip4[ii]));
4781 #endif
4782 #ifdef INET6
4783 	db_printf(" ip6s            = %d\n", pr->pr_ip6s);
4784 	for (ii = 0; ii < pr->pr_ip6s; ii++)
4785 		db_printf(" %s %s\n",
4786 		    ii == 0 ? "ip6.addr        =" : "                 ",
4787 		    ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4788 #endif
4789 }
4790 
4791 DB_SHOW_COMMAND(prison, db_show_prison_command)
4792 {
4793 	struct prison *pr;
4794 
4795 	if (!have_addr) {
4796 		/*
4797 		 * Show all prisons in the list, and prison0 which is not
4798 		 * listed.
4799 		 */
4800 		db_show_prison(&prison0);
4801 		if (!db_pager_quit) {
4802 			TAILQ_FOREACH(pr, &allprison, pr_list) {
4803 				db_show_prison(pr);
4804 				if (db_pager_quit)
4805 					break;
4806 			}
4807 		}
4808 		return;
4809 	}
4810 
4811 	if (addr == 0)
4812 		pr = &prison0;
4813 	else {
4814 		/* Look for a prison with the ID and with references. */
4815 		TAILQ_FOREACH(pr, &allprison, pr_list)
4816 			if (pr->pr_id == addr && pr->pr_ref > 0)
4817 				break;
4818 		if (pr == NULL)
4819 			/* Look again, without requiring a reference. */
4820 			TAILQ_FOREACH(pr, &allprison, pr_list)
4821 				if (pr->pr_id == addr)
4822 					break;
4823 		if (pr == NULL)
4824 			/* Assume address points to a valid prison. */
4825 			pr = (struct prison *)addr;
4826 	}
4827 	db_show_prison(pr);
4828 }
4829 
4830 #endif /* DDB */
4831