xref: /freebsd/sys/kern/kern_jail.c (revision aa64588d28258aef88cc33b8043112e8856948d0)
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/sx.h>
53 #include <sys/sysent.h>
54 #include <sys/namei.h>
55 #include <sys/mount.h>
56 #include <sys/queue.h>
57 #include <sys/socket.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/vnode.h>
61 
62 #include <net/if.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/in.h>
66 
67 #ifdef DDB
68 #include <ddb/ddb.h>
69 #ifdef INET6
70 #include <netinet6/in6_var.h>
71 #endif /* INET6 */
72 #endif /* DDB */
73 
74 #include <security/mac/mac_framework.h>
75 
76 #define	DEFAULT_HOSTUUID	"00000000-0000-0000-0000-000000000000"
77 
78 MALLOC_DEFINE(M_PRISON, "prison", "Prison 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_childmax	= JAIL_MAX,
104 	.pr_hostuuid	= DEFAULT_HOSTUUID,
105 	.pr_children	= LIST_HEAD_INITIALIZER(prison0.pr_children),
106 #ifdef VIMAGE
107 	.pr_flags	= PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
108 #else
109 	.pr_flags	= PR_HOST|_PR_IP_SADDRSEL,
110 #endif
111 	.pr_allow	= PR_ALLOW_ALL,
112 };
113 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
114 
115 /* allprison and lastprid are protected by allprison_lock. */
116 struct	sx allprison_lock;
117 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
118 struct	prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
119 int	lastprid = 0;
120 
121 static int do_jail_attach(struct thread *td, struct prison *pr);
122 static void prison_complete(void *context, int pending);
123 static void prison_deref(struct prison *pr, int flags);
124 static char *prison_path(struct prison *pr1, struct prison *pr2);
125 static void prison_remove_one(struct prison *pr);
126 #ifdef INET
127 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
128 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
129 #endif
130 #ifdef INET6
131 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
132 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
133 #endif
134 
135 /* Flags for prison_deref */
136 #define	PD_DEREF	0x01
137 #define	PD_DEUREF	0x02
138 #define	PD_LOCKED	0x04
139 #define	PD_LIST_SLOCKED	0x08
140 #define	PD_LIST_XLOCKED	0x10
141 
142 /*
143  * Parameter names corresponding to PR_* flag values
144  */
145 static char *pr_flag_names[] = {
146 	[0] = "persist",
147 #ifdef INET
148 	[7] = "ip4.saddrsel",
149 #endif
150 #ifdef INET6
151 	[8] = "ip6.saddrsel",
152 #endif
153 };
154 
155 static char *pr_flag_nonames[] = {
156 	[0] = "nopersist",
157 #ifdef INET
158 	[7] = "ip4.nosaddrsel",
159 #endif
160 #ifdef INET6
161 	[8] = "ip6.nosaddrsel",
162 #endif
163 };
164 
165 struct jailsys_flags {
166 	const char	*name;
167 	unsigned	 disable;
168 	unsigned	 new;
169 } pr_flag_jailsys[] = {
170 	{ "host", 0, PR_HOST },
171 #ifdef VIMAGE
172 	{ "vnet", 0, PR_VNET },
173 #endif
174 #ifdef INET
175 	{ "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
176 #endif
177 #ifdef INET6
178 	{ "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
179 #endif
180 };
181 
182 static char *pr_allow_names[] = {
183 	"allow.set_hostname",
184 	"allow.sysvipc",
185 	"allow.raw_sockets",
186 	"allow.chflags",
187 	"allow.mount",
188 	"allow.quotas",
189 	"allow.socket_af",
190 };
191 
192 static char *pr_allow_nonames[] = {
193 	"allow.noset_hostname",
194 	"allow.nosysvipc",
195 	"allow.noraw_sockets",
196 	"allow.nochflags",
197 	"allow.nomount",
198 	"allow.noquotas",
199 	"allow.nosocket_af",
200 };
201 
202 #define	JAIL_DEFAULT_ALLOW		PR_ALLOW_SET_HOSTNAME
203 #define	JAIL_DEFAULT_ENFORCE_STATFS	2
204 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
205 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
206 #if defined(INET) || defined(INET6)
207 static unsigned jail_max_af_ips = 255;
208 #endif
209 
210 #ifdef INET
211 static int
212 qcmp_v4(const void *ip1, const void *ip2)
213 {
214 	in_addr_t iaa, iab;
215 
216 	/*
217 	 * We need to compare in HBO here to get the list sorted as expected
218 	 * by the result of the code.  Sorting NBO addresses gives you
219 	 * interesting results.  If you do not understand, do not try.
220 	 */
221 	iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
222 	iab = ntohl(((const struct in_addr *)ip2)->s_addr);
223 
224 	/*
225 	 * Do not simply return the difference of the two numbers, the int is
226 	 * not wide enough.
227 	 */
228 	if (iaa > iab)
229 		return (1);
230 	else if (iaa < iab)
231 		return (-1);
232 	else
233 		return (0);
234 }
235 #endif
236 
237 #ifdef INET6
238 static int
239 qcmp_v6(const void *ip1, const void *ip2)
240 {
241 	const struct in6_addr *ia6a, *ia6b;
242 	int i, rc;
243 
244 	ia6a = (const struct in6_addr *)ip1;
245 	ia6b = (const struct in6_addr *)ip2;
246 
247 	rc = 0;
248 	for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
249 		if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
250 			rc = 1;
251 		else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
252 			rc = -1;
253 	}
254 	return (rc);
255 }
256 #endif
257 
258 /*
259  * struct jail_args {
260  *	struct jail *jail;
261  * };
262  */
263 int
264 jail(struct thread *td, struct jail_args *uap)
265 {
266 	uint32_t version;
267 	int error;
268 	struct jail j;
269 
270 	error = copyin(uap->jail, &version, sizeof(uint32_t));
271 	if (error)
272 		return (error);
273 
274 	switch (version) {
275 	case 0:
276 	{
277 		struct jail_v0 j0;
278 
279 		/* FreeBSD single IPv4 jails. */
280 		bzero(&j, sizeof(struct jail));
281 		error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
282 		if (error)
283 			return (error);
284 		j.version = j0.version;
285 		j.path = j0.path;
286 		j.hostname = j0.hostname;
287 		j.ip4s = j0.ip_number;
288 		break;
289 	}
290 
291 	case 1:
292 		/*
293 		 * Version 1 was used by multi-IPv4 jail implementations
294 		 * that never made it into the official kernel.
295 		 */
296 		return (EINVAL);
297 
298 	case 2:	/* JAIL_API_VERSION */
299 		/* FreeBSD multi-IPv4/IPv6,noIP jails. */
300 		error = copyin(uap->jail, &j, sizeof(struct jail));
301 		if (error)
302 			return (error);
303 		break;
304 
305 	default:
306 		/* Sci-Fi jails are not supported, sorry. */
307 		return (EINVAL);
308 	}
309 	return (kern_jail(td, &j));
310 }
311 
312 int
313 kern_jail(struct thread *td, struct jail *j)
314 {
315 	struct iovec optiov[2 * (4
316 			    + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
317 #ifdef INET
318 			    + 1
319 #endif
320 #ifdef INET6
321 			    + 1
322 #endif
323 			    )];
324 	struct uio opt;
325 	char *u_path, *u_hostname, *u_name;
326 #ifdef INET
327 	uint32_t ip4s;
328 	struct in_addr *u_ip4;
329 #endif
330 #ifdef INET6
331 	struct in6_addr *u_ip6;
332 #endif
333 	size_t tmplen;
334 	int error, enforce_statfs, fi;
335 
336 	bzero(&optiov, sizeof(optiov));
337 	opt.uio_iov = optiov;
338 	opt.uio_iovcnt = 0;
339 	opt.uio_offset = -1;
340 	opt.uio_resid = -1;
341 	opt.uio_segflg = UIO_SYSSPACE;
342 	opt.uio_rw = UIO_READ;
343 	opt.uio_td = td;
344 
345 	/* Set permissions for top-level jails from sysctls. */
346 	if (!jailed(td->td_ucred)) {
347 		for (fi = 0; fi < sizeof(pr_allow_names) /
348 		     sizeof(pr_allow_names[0]); fi++) {
349 			optiov[opt.uio_iovcnt].iov_base =
350 			    (jail_default_allow & (1 << fi))
351 			    ? pr_allow_names[fi] : pr_allow_nonames[fi];
352 			optiov[opt.uio_iovcnt].iov_len =
353 			    strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
354 			opt.uio_iovcnt += 2;
355 		}
356 		optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
357 		optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
358 		opt.uio_iovcnt++;
359 		enforce_statfs = jail_default_enforce_statfs;
360 		optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
361 		optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
362 		opt.uio_iovcnt++;
363 	}
364 
365 	tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
366 #ifdef INET
367 	ip4s = (j->version == 0) ? 1 : j->ip4s;
368 	if (ip4s > jail_max_af_ips)
369 		return (EINVAL);
370 	tmplen += ip4s * sizeof(struct in_addr);
371 #else
372 	if (j->ip4s > 0)
373 		return (EINVAL);
374 #endif
375 #ifdef INET6
376 	if (j->ip6s > jail_max_af_ips)
377 		return (EINVAL);
378 	tmplen += j->ip6s * sizeof(struct in6_addr);
379 #else
380 	if (j->ip6s > 0)
381 		return (EINVAL);
382 #endif
383 	u_path = malloc(tmplen, M_TEMP, M_WAITOK);
384 	u_hostname = u_path + MAXPATHLEN;
385 	u_name = u_hostname + MAXHOSTNAMELEN;
386 #ifdef INET
387 	u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
388 #endif
389 #ifdef INET6
390 #ifdef INET
391 	u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
392 #else
393 	u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
394 #endif
395 #endif
396 	optiov[opt.uio_iovcnt].iov_base = "path";
397 	optiov[opt.uio_iovcnt].iov_len = sizeof("path");
398 	opt.uio_iovcnt++;
399 	optiov[opt.uio_iovcnt].iov_base = u_path;
400 	error = copyinstr(j->path, u_path, MAXPATHLEN,
401 	    &optiov[opt.uio_iovcnt].iov_len);
402 	if (error) {
403 		free(u_path, M_TEMP);
404 		return (error);
405 	}
406 	opt.uio_iovcnt++;
407 	optiov[opt.uio_iovcnt].iov_base = "host.hostname";
408 	optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
409 	opt.uio_iovcnt++;
410 	optiov[opt.uio_iovcnt].iov_base = u_hostname;
411 	error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
412 	    &optiov[opt.uio_iovcnt].iov_len);
413 	if (error) {
414 		free(u_path, M_TEMP);
415 		return (error);
416 	}
417 	opt.uio_iovcnt++;
418 	if (j->jailname != NULL) {
419 		optiov[opt.uio_iovcnt].iov_base = "name";
420 		optiov[opt.uio_iovcnt].iov_len = sizeof("name");
421 		opt.uio_iovcnt++;
422 		optiov[opt.uio_iovcnt].iov_base = u_name;
423 		error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
424 		    &optiov[opt.uio_iovcnt].iov_len);
425 		if (error) {
426 			free(u_path, M_TEMP);
427 			return (error);
428 		}
429 		opt.uio_iovcnt++;
430 	}
431 #ifdef INET
432 	optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
433 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
434 	opt.uio_iovcnt++;
435 	optiov[opt.uio_iovcnt].iov_base = u_ip4;
436 	optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
437 	if (j->version == 0)
438 		u_ip4->s_addr = j->ip4s;
439 	else {
440 		error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
441 		if (error) {
442 			free(u_path, M_TEMP);
443 			return (error);
444 		}
445 	}
446 	opt.uio_iovcnt++;
447 #endif
448 #ifdef INET6
449 	optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
450 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
451 	opt.uio_iovcnt++;
452 	optiov[opt.uio_iovcnt].iov_base = u_ip6;
453 	optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
454 	error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
455 	if (error) {
456 		free(u_path, M_TEMP);
457 		return (error);
458 	}
459 	opt.uio_iovcnt++;
460 #endif
461 	KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
462 	    ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
463 	error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
464 	free(u_path, M_TEMP);
465 	return (error);
466 }
467 
468 
469 /*
470  * struct jail_set_args {
471  *	struct iovec *iovp;
472  *	unsigned int iovcnt;
473  *	int flags;
474  * };
475  */
476 int
477 jail_set(struct thread *td, struct jail_set_args *uap)
478 {
479 	struct uio *auio;
480 	int error;
481 
482 	/* Check that we have an even number of iovecs. */
483 	if (uap->iovcnt & 1)
484 		return (EINVAL);
485 
486 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
487 	if (error)
488 		return (error);
489 	error = kern_jail_set(td, auio, uap->flags);
490 	free(auio, M_IOV);
491 	return (error);
492 }
493 
494 int
495 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
496 {
497 	struct nameidata nd;
498 #ifdef INET
499 	struct in_addr *ip4;
500 #endif
501 #ifdef INET6
502 	struct in6_addr *ip6;
503 #endif
504 	struct vfsopt *opt;
505 	struct vfsoptlist *opts;
506 	struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
507 	struct vnode *root;
508 	char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
509 #if defined(INET) || defined(INET6)
510 	struct prison *tppr;
511 	void *op;
512 #endif
513 	unsigned long hid;
514 	size_t namelen, onamelen;
515 	int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
516 	int gotchildmax, gotenforce, gothid, gotslevel;
517 	int fi, jid, jsys, len, level;
518 	int childmax, slevel, vfslocked;
519 #if defined(INET) || defined(INET6)
520 	int ii, ij;
521 #endif
522 #ifdef INET
523 	int ip4s, redo_ip4;
524 #endif
525 #ifdef INET6
526 	int ip6s, redo_ip6;
527 #endif
528 	unsigned pr_flags, ch_flags;
529 	unsigned pr_allow, ch_allow, tallow;
530 	char numbuf[12];
531 
532 	error = priv_check(td, PRIV_JAIL_SET);
533 	if (!error && (flags & JAIL_ATTACH))
534 		error = priv_check(td, PRIV_JAIL_ATTACH);
535 	if (error)
536 		return (error);
537 	mypr = ppr = td->td_ucred->cr_prison;
538 	if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
539 		return (EPERM);
540 	if (flags & ~JAIL_SET_MASK)
541 		return (EINVAL);
542 
543 	/*
544 	 * Check all the parameters before committing to anything.  Not all
545 	 * errors can be caught early, but we may as well try.  Also, this
546 	 * takes care of some expensive stuff (path lookup) before getting
547 	 * the allprison lock.
548 	 *
549 	 * XXX Jails are not filesystems, and jail parameters are not mount
550 	 *     options.  But it makes more sense to re-use the vfsopt code
551 	 *     than duplicate it under a different name.
552 	 */
553 	error = vfs_buildopts(optuio, &opts);
554 	if (error)
555 		return (error);
556 #ifdef INET
557 	ip4 = NULL;
558 #endif
559 #ifdef INET6
560 	ip6 = NULL;
561 #endif
562 
563 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
564 	if (error == ENOENT)
565 		jid = 0;
566 	else if (error != 0)
567 		goto done_free;
568 
569 	error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
570 	if (error == ENOENT)
571 		gotslevel = 0;
572 	else if (error != 0)
573 		goto done_free;
574 	else
575 		gotslevel = 1;
576 
577 	error =
578 	    vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
579 	if (error == ENOENT)
580 		gotchildmax = 0;
581 	else if (error != 0)
582 		goto done_free;
583 	else
584 		gotchildmax = 1;
585 
586 	error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
587 	gotenforce = (error == 0);
588 	if (gotenforce) {
589 		if (enforce < 0 || enforce > 2)
590 			return (EINVAL);
591 	} else if (error != ENOENT)
592 		goto done_free;
593 
594 	pr_flags = ch_flags = 0;
595 	for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
596 	    fi++) {
597 		if (pr_flag_names[fi] == NULL)
598 			continue;
599 		vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
600 		vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
601 	}
602 	ch_flags |= pr_flags;
603 	for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
604 	    fi++) {
605 		error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
606 		    sizeof(jsys));
607 		if (error == ENOENT)
608 			continue;
609 		if (error != 0)
610 			goto done_free;
611 		switch (jsys) {
612 		case JAIL_SYS_DISABLE:
613 			if (!pr_flag_jailsys[fi].disable) {
614 				error = EINVAL;
615 				goto done_free;
616 			}
617 			pr_flags |= pr_flag_jailsys[fi].disable;
618 			break;
619 		case JAIL_SYS_NEW:
620 			pr_flags |= pr_flag_jailsys[fi].new;
621 			break;
622 		case JAIL_SYS_INHERIT:
623 			break;
624 		default:
625 			error = EINVAL;
626 			goto done_free;
627 		}
628 		ch_flags |=
629 		    pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
630 	}
631 	if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
632 	    && !(pr_flags & PR_PERSIST)) {
633 		error = EINVAL;
634 		vfs_opterror(opts, "new jail must persist or attach");
635 		goto done_errmsg;
636 	}
637 #ifdef VIMAGE
638 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
639 		error = EINVAL;
640 		vfs_opterror(opts, "vnet cannot be changed after creation");
641 		goto done_errmsg;
642 	}
643 #endif
644 #ifdef INET
645 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
646 		error = EINVAL;
647 		vfs_opterror(opts, "ip4 cannot be changed after creation");
648 		goto done_errmsg;
649 	}
650 #endif
651 #ifdef INET6
652 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
653 		error = EINVAL;
654 		vfs_opterror(opts, "ip6 cannot be changed after creation");
655 		goto done_errmsg;
656 	}
657 #endif
658 
659 	pr_allow = ch_allow = 0;
660 	for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
661 	    fi++) {
662 		vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
663 		vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
664 	}
665 	ch_allow |= pr_allow;
666 
667 	error = vfs_getopt(opts, "name", (void **)&name, &len);
668 	if (error == ENOENT)
669 		name = NULL;
670 	else if (error != 0)
671 		goto done_free;
672 	else {
673 		if (len == 0 || name[len - 1] != '\0') {
674 			error = EINVAL;
675 			goto done_free;
676 		}
677 		if (len > MAXHOSTNAMELEN) {
678 			error = ENAMETOOLONG;
679 			goto done_free;
680 		}
681 	}
682 
683 	error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
684 	if (error == ENOENT)
685 		host = NULL;
686 	else if (error != 0)
687 		goto done_free;
688 	else {
689 		ch_flags |= PR_HOST;
690 		pr_flags |= PR_HOST;
691 		if (len == 0 || host[len - 1] != '\0') {
692 			error = EINVAL;
693 			goto done_free;
694 		}
695 		if (len > MAXHOSTNAMELEN) {
696 			error = ENAMETOOLONG;
697 			goto done_free;
698 		}
699 	}
700 
701 	error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
702 	if (error == ENOENT)
703 		domain = NULL;
704 	else if (error != 0)
705 		goto done_free;
706 	else {
707 		ch_flags |= PR_HOST;
708 		pr_flags |= PR_HOST;
709 		if (len == 0 || domain[len - 1] != '\0') {
710 			error = EINVAL;
711 			goto done_free;
712 		}
713 		if (len > MAXHOSTNAMELEN) {
714 			error = ENAMETOOLONG;
715 			goto done_free;
716 		}
717 	}
718 
719 	error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
720 	if (error == ENOENT)
721 		uuid = NULL;
722 	else if (error != 0)
723 		goto done_free;
724 	else {
725 		ch_flags |= PR_HOST;
726 		pr_flags |= PR_HOST;
727 		if (len == 0 || uuid[len - 1] != '\0') {
728 			error = EINVAL;
729 			goto done_free;
730 		}
731 		if (len > HOSTUUIDLEN) {
732 			error = ENAMETOOLONG;
733 			goto done_free;
734 		}
735 	}
736 
737 #ifdef COMPAT_FREEBSD32
738 	if (td->td_proc->p_sysent->sv_flags & SV_ILP32) {
739 		uint32_t hid32;
740 
741 		error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
742 		hid = hid32;
743 	} else
744 #endif
745 		error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
746 	if (error == ENOENT)
747 		gothid = 0;
748 	else if (error != 0)
749 		goto done_free;
750 	else {
751 		gothid = 1;
752 		ch_flags |= PR_HOST;
753 		pr_flags |= PR_HOST;
754 	}
755 
756 #ifdef INET
757 	error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
758 	if (error == ENOENT)
759 		ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
760 	else if (error != 0)
761 		goto done_free;
762 	else if (ip4s & (sizeof(*ip4) - 1)) {
763 		error = EINVAL;
764 		goto done_free;
765 	} else {
766 		ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
767 		if (ip4s == 0)
768 			pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
769 		else {
770 			pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
771 			ip4s /= sizeof(*ip4);
772 			if (ip4s > jail_max_af_ips) {
773 				error = EINVAL;
774 				vfs_opterror(opts, "too many IPv4 addresses");
775 				goto done_errmsg;
776 			}
777 			ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
778 			bcopy(op, ip4, ip4s * sizeof(*ip4));
779 			/*
780 			 * IP addresses are all sorted but ip[0] to preserve
781 			 * the primary IP address as given from userland.
782 			 * This special IP is used for unbound outgoing
783 			 * connections as well for "loopback" traffic in case
784 			 * source address selection cannot find any more fitting
785 			 * address to connect from.
786 			 */
787 			if (ip4s > 1)
788 				qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
789 			/*
790 			 * Check for duplicate addresses and do some simple
791 			 * zero and broadcast checks. If users give other bogus
792 			 * addresses it is their problem.
793 			 *
794 			 * We do not have to care about byte order for these
795 			 * checks so we will do them in NBO.
796 			 */
797 			for (ii = 0; ii < ip4s; ii++) {
798 				if (ip4[ii].s_addr == INADDR_ANY ||
799 				    ip4[ii].s_addr == INADDR_BROADCAST) {
800 					error = EINVAL;
801 					goto done_free;
802 				}
803 				if ((ii+1) < ip4s &&
804 				    (ip4[0].s_addr == ip4[ii+1].s_addr ||
805 				     ip4[ii].s_addr == ip4[ii+1].s_addr)) {
806 					error = EINVAL;
807 					goto done_free;
808 				}
809 			}
810 		}
811 	}
812 #endif
813 
814 #ifdef INET6
815 	error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
816 	if (error == ENOENT)
817 		ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
818 	else if (error != 0)
819 		goto done_free;
820 	else if (ip6s & (sizeof(*ip6) - 1)) {
821 		error = EINVAL;
822 		goto done_free;
823 	} else {
824 		ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
825 		if (ip6s == 0)
826 			pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
827 		else {
828 			pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
829 			ip6s /= sizeof(*ip6);
830 			if (ip6s > jail_max_af_ips) {
831 				error = EINVAL;
832 				vfs_opterror(opts, "too many IPv6 addresses");
833 				goto done_errmsg;
834 			}
835 			ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
836 			bcopy(op, ip6, ip6s * sizeof(*ip6));
837 			if (ip6s > 1)
838 				qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
839 			for (ii = 0; ii < ip6s; ii++) {
840 				if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
841 					error = EINVAL;
842 					goto done_free;
843 				}
844 				if ((ii+1) < ip6s &&
845 				    (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
846 				     IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
847 				{
848 					error = EINVAL;
849 					goto done_free;
850 				}
851 			}
852 		}
853 	}
854 #endif
855 
856 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
857 	if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
858 		error = EINVAL;
859 		vfs_opterror(opts,
860 		    "vnet jails cannot have IP address restrictions");
861 		goto done_errmsg;
862 	}
863 #endif
864 
865 	root = NULL;
866 	error = vfs_getopt(opts, "path", (void **)&path, &len);
867 	if (error == ENOENT)
868 		path = NULL;
869 	else if (error != 0)
870 		goto done_free;
871 	else {
872 		if (flags & JAIL_UPDATE) {
873 			error = EINVAL;
874 			vfs_opterror(opts,
875 			    "path cannot be changed after creation");
876 			goto done_errmsg;
877 		}
878 		if (len == 0 || path[len - 1] != '\0') {
879 			error = EINVAL;
880 			goto done_free;
881 		}
882 		if (len < 2 || (len == 2 && path[0] == '/'))
883 			path = NULL;
884 		else {
885 			/* Leave room for a real-root full pathname. */
886 			if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
887 			    ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
888 				error = ENAMETOOLONG;
889 				goto done_free;
890 			}
891 			NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
892 			    path, td);
893 			error = namei(&nd);
894 			if (error)
895 				goto done_free;
896 			vfslocked = NDHASGIANT(&nd);
897 			root = nd.ni_vp;
898 			NDFREE(&nd, NDF_ONLY_PNBUF);
899 			if (root->v_type != VDIR) {
900 				error = ENOTDIR;
901 				vrele(root);
902 				VFS_UNLOCK_GIANT(vfslocked);
903 				goto done_free;
904 			}
905 			VFS_UNLOCK_GIANT(vfslocked);
906 		}
907 	}
908 
909 	/*
910 	 * Grab the allprison lock before letting modules check their
911 	 * parameters.  Once we have it, do not let go so we'll have a
912 	 * consistent view of the OSD list.
913 	 */
914 	sx_xlock(&allprison_lock);
915 	error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
916 	if (error)
917 		goto done_unlock_list;
918 
919 	/* By now, all parameters should have been noted. */
920 	TAILQ_FOREACH(opt, opts, link) {
921 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
922 			error = EINVAL;
923 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
924 			goto done_unlock_list;
925 		}
926 	}
927 
928 	/*
929 	 * See if we are creating a new record or updating an existing one.
930 	 * This abuses the file error codes ENOENT and EEXIST.
931 	 */
932 	cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
933 	if (!cuflags) {
934 		error = EINVAL;
935 		vfs_opterror(opts, "no valid operation (create or update)");
936 		goto done_unlock_list;
937 	}
938 	pr = NULL;
939 	namelc = NULL;
940 	if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
941 		namelc = strrchr(name, '.');
942 		jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
943 		if (*p != '\0')
944 			jid = 0;
945 	}
946 	if (jid != 0) {
947 		/*
948 		 * See if a requested jid already exists.  There is an
949 		 * information leak here if the jid exists but is not within
950 		 * the caller's jail hierarchy.  Jail creators will get EEXIST
951 		 * even though they cannot see the jail, and CREATE | UPDATE
952 		 * will return ENOENT which is not normally a valid error.
953 		 */
954 		if (jid < 0) {
955 			error = EINVAL;
956 			vfs_opterror(opts, "negative jid");
957 			goto done_unlock_list;
958 		}
959 		pr = prison_find(jid);
960 		if (pr != NULL) {
961 			ppr = pr->pr_parent;
962 			/* Create: jid must not exist. */
963 			if (cuflags == JAIL_CREATE) {
964 				mtx_unlock(&pr->pr_mtx);
965 				error = EEXIST;
966 				vfs_opterror(opts, "jail %d already exists",
967 				    jid);
968 				goto done_unlock_list;
969 			}
970 			if (!prison_ischild(mypr, pr)) {
971 				mtx_unlock(&pr->pr_mtx);
972 				pr = NULL;
973 			} else if (pr->pr_uref == 0) {
974 				if (!(flags & JAIL_DYING)) {
975 					mtx_unlock(&pr->pr_mtx);
976 					error = ENOENT;
977 					vfs_opterror(opts, "jail %d is dying",
978 					    jid);
979 					goto done_unlock_list;
980 				} else if ((flags & JAIL_ATTACH) ||
981 				    (pr_flags & PR_PERSIST)) {
982 					/*
983 					 * A dying jail might be resurrected
984 					 * (via attach or persist), but first
985 					 * it must determine if another jail
986 					 * has claimed its name.  Accomplish
987 					 * this by implicitly re-setting the
988 					 * name.
989 					 */
990 					if (name == NULL)
991 						name = prison_name(mypr, pr);
992 				}
993 			}
994 		}
995 		if (pr == NULL) {
996 			/* Update: jid must exist. */
997 			if (cuflags == JAIL_UPDATE) {
998 				error = ENOENT;
999 				vfs_opterror(opts, "jail %d not found", jid);
1000 				goto done_unlock_list;
1001 			}
1002 		}
1003 	}
1004 	/*
1005 	 * If the caller provided a name, look for a jail by that name.
1006 	 * This has different semantics for creates and updates keyed by jid
1007 	 * (where the name must not already exist in a different jail),
1008 	 * and updates keyed by the name itself (where the name must exist
1009 	 * because that is the jail being updated).
1010 	 */
1011 	if (name != NULL) {
1012 		namelc = strrchr(name, '.');
1013 		if (namelc == NULL)
1014 			namelc = name;
1015 		else {
1016 			/*
1017 			 * This is a hierarchical name.  Split it into the
1018 			 * parent and child names, and make sure the parent
1019 			 * exists or matches an already found jail.
1020 			 */
1021 			*namelc = '\0';
1022 			if (pr != NULL) {
1023 				if (strncmp(name, ppr->pr_name, namelc - name)
1024 				    || ppr->pr_name[namelc - name] != '\0') {
1025 					mtx_unlock(&pr->pr_mtx);
1026 					error = EINVAL;
1027 					vfs_opterror(opts,
1028 					    "cannot change jail's parent");
1029 					goto done_unlock_list;
1030 				}
1031 			} else {
1032 				ppr = prison_find_name(mypr, name);
1033 				if (ppr == NULL) {
1034 					error = ENOENT;
1035 					vfs_opterror(opts,
1036 					    "jail \"%s\" not found", name);
1037 					goto done_unlock_list;
1038 				}
1039 				mtx_unlock(&ppr->pr_mtx);
1040 			}
1041 			name = ++namelc;
1042 		}
1043 		if (name[0] != '\0') {
1044 			namelen =
1045 			    (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1046  name_again:
1047 			deadpr = NULL;
1048 			FOREACH_PRISON_CHILD(ppr, tpr) {
1049 				if (tpr != pr && tpr->pr_ref > 0 &&
1050 				    !strcmp(tpr->pr_name + namelen, name)) {
1051 					if (pr == NULL &&
1052 					    cuflags != JAIL_CREATE) {
1053 						mtx_lock(&tpr->pr_mtx);
1054 						if (tpr->pr_ref > 0) {
1055 							/*
1056 							 * Use this jail
1057 							 * for updates.
1058 							 */
1059 							if (tpr->pr_uref > 0) {
1060 								pr = tpr;
1061 								break;
1062 							}
1063 							deadpr = tpr;
1064 						}
1065 						mtx_unlock(&tpr->pr_mtx);
1066 					} else if (tpr->pr_uref > 0) {
1067 						/*
1068 						 * Create, or update(jid):
1069 						 * name must not exist in an
1070 						 * active sibling jail.
1071 						 */
1072 						error = EEXIST;
1073 						if (pr != NULL)
1074 							mtx_unlock(&pr->pr_mtx);
1075 						vfs_opterror(opts,
1076 						   "jail \"%s\" already exists",
1077 						   name);
1078 						goto done_unlock_list;
1079 					}
1080 				}
1081 			}
1082 			/* If no active jail is found, use a dying one. */
1083 			if (deadpr != NULL && pr == NULL) {
1084 				if (flags & JAIL_DYING) {
1085 					mtx_lock(&deadpr->pr_mtx);
1086 					if (deadpr->pr_ref == 0) {
1087 						mtx_unlock(&deadpr->pr_mtx);
1088 						goto name_again;
1089 					}
1090 					pr = deadpr;
1091 				} else if (cuflags == JAIL_UPDATE) {
1092 					error = ENOENT;
1093 					vfs_opterror(opts,
1094 					    "jail \"%s\" is dying", name);
1095 					goto done_unlock_list;
1096 				}
1097 			}
1098 			/* Update: name must exist if no jid. */
1099 			else if (cuflags == JAIL_UPDATE && pr == NULL) {
1100 				error = ENOENT;
1101 				vfs_opterror(opts, "jail \"%s\" not found",
1102 				    name);
1103 				goto done_unlock_list;
1104 			}
1105 		}
1106 	}
1107 	/* Update: must provide a jid or name. */
1108 	else if (cuflags == JAIL_UPDATE && pr == NULL) {
1109 		error = ENOENT;
1110 		vfs_opterror(opts, "update specified no jail");
1111 		goto done_unlock_list;
1112 	}
1113 
1114 	/* If there's no prison to update, create a new one and link it in. */
1115 	if (pr == NULL) {
1116 		for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1117 			if (tpr->pr_childcount >= tpr->pr_childmax) {
1118 				error = EPERM;
1119 				vfs_opterror(opts, "prison limit exceeded");
1120 				goto done_unlock_list;
1121 			}
1122 		created = 1;
1123 		mtx_lock(&ppr->pr_mtx);
1124 		if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1125 			mtx_unlock(&ppr->pr_mtx);
1126 			error = ENOENT;
1127 			vfs_opterror(opts, "parent jail went away!");
1128 			goto done_unlock_list;
1129 		}
1130 		ppr->pr_ref++;
1131 		ppr->pr_uref++;
1132 		mtx_unlock(&ppr->pr_mtx);
1133 		pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1134 		if (jid == 0) {
1135 			/* Find the next free jid. */
1136 			jid = lastprid + 1;
1137  findnext:
1138 			if (jid == JAIL_MAX)
1139 				jid = 1;
1140 			TAILQ_FOREACH(tpr, &allprison, pr_list) {
1141 				if (tpr->pr_id < jid)
1142 					continue;
1143 				if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1144 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1145 					break;
1146 				}
1147 				if (jid == lastprid) {
1148 					error = EAGAIN;
1149 					vfs_opterror(opts,
1150 					    "no available jail IDs");
1151 					free(pr, M_PRISON);
1152 					prison_deref(ppr, PD_DEREF |
1153 					    PD_DEUREF | PD_LIST_XLOCKED);
1154 					goto done_releroot;
1155 				}
1156 				jid++;
1157 				goto findnext;
1158 			}
1159 			lastprid = jid;
1160 		} else {
1161 			/*
1162 			 * The jail already has a jid (that did not yet exist),
1163 			 * so just find where to insert it.
1164 			 */
1165 			TAILQ_FOREACH(tpr, &allprison, pr_list)
1166 				if (tpr->pr_id >= jid) {
1167 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1168 					break;
1169 				}
1170 		}
1171 		if (tpr == NULL)
1172 			TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1173 		LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1174 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1175 			tpr->pr_childcount++;
1176 
1177 		pr->pr_parent = ppr;
1178 		pr->pr_id = jid;
1179 
1180 		/* Set some default values, and inherit some from the parent. */
1181 		if (name == NULL)
1182 			name = "";
1183 		if (path == NULL) {
1184 			path = "/";
1185 			root = mypr->pr_root;
1186 			vref(root);
1187 		}
1188 		strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1189 		pr->pr_flags |= PR_HOST;
1190 #if defined(INET) || defined(INET6)
1191 #ifdef VIMAGE
1192 		if (!(pr_flags & PR_VNET))
1193 #endif
1194 		{
1195 #ifdef INET
1196 			if (!(ch_flags & PR_IP4_USER))
1197 				pr->pr_flags |=
1198 				    PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1199 			else if (!(pr_flags & PR_IP4_USER)) {
1200 				pr->pr_flags |= ppr->pr_flags & PR_IP4;
1201 				if (ppr->pr_ip4 != NULL) {
1202 					pr->pr_ip4s = ppr->pr_ip4s;
1203 					pr->pr_ip4 = malloc(pr->pr_ip4s *
1204 					    sizeof(struct in_addr), M_PRISON,
1205 					    M_WAITOK);
1206 					bcopy(ppr->pr_ip4, pr->pr_ip4,
1207 					    pr->pr_ip4s * sizeof(*pr->pr_ip4));
1208 				}
1209 			}
1210 #endif
1211 #ifdef INET6
1212 			if (!(ch_flags & PR_IP6_USER))
1213 				pr->pr_flags |=
1214 				    PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1215 			else if (!(pr_flags & PR_IP6_USER)) {
1216 				pr->pr_flags |= ppr->pr_flags & PR_IP6;
1217 				if (ppr->pr_ip6 != NULL) {
1218 					pr->pr_ip6s = ppr->pr_ip6s;
1219 					pr->pr_ip6 = malloc(pr->pr_ip6s *
1220 					    sizeof(struct in6_addr), M_PRISON,
1221 					    M_WAITOK);
1222 					bcopy(ppr->pr_ip6, pr->pr_ip6,
1223 					    pr->pr_ip6s * sizeof(*pr->pr_ip6));
1224 				}
1225 			}
1226 #endif
1227 		}
1228 #endif
1229 		/* Source address selection is always on by default. */
1230 		pr->pr_flags |= _PR_IP_SADDRSEL;
1231 
1232 		pr->pr_securelevel = ppr->pr_securelevel;
1233 		pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1234 		pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1235 
1236 		LIST_INIT(&pr->pr_children);
1237 		mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1238 
1239 #ifdef VIMAGE
1240 		/* Allocate a new vnet if specified. */
1241 		pr->pr_vnet = (pr_flags & PR_VNET)
1242 		    ? vnet_alloc() : ppr->pr_vnet;
1243 #endif
1244 		/*
1245 		 * Allocate a dedicated cpuset for each jail.
1246 		 * Unlike other initial settings, this may return an erorr.
1247 		 */
1248 		error = cpuset_create_root(ppr, &pr->pr_cpuset);
1249 		if (error) {
1250 			prison_deref(pr, PD_LIST_XLOCKED);
1251 			goto done_releroot;
1252 		}
1253 
1254 		mtx_lock(&pr->pr_mtx);
1255 		/*
1256 		 * New prisons do not yet have a reference, because we do not
1257 		 * want other to see the incomplete prison once the
1258 		 * allprison_lock is downgraded.
1259 		 */
1260 	} else {
1261 		created = 0;
1262 		/*
1263 		 * Grab a reference for existing prisons, to ensure they
1264 		 * continue to exist for the duration of the call.
1265 		 */
1266 		pr->pr_ref++;
1267 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1268 		if ((pr->pr_flags & PR_VNET) &&
1269 		    (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1270 			error = EINVAL;
1271 			vfs_opterror(opts,
1272 			    "vnet jails cannot have IP address restrictions");
1273 			goto done_deref_locked;
1274 		}
1275 #endif
1276 #ifdef INET
1277 		if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1278 			error = EINVAL;
1279 			vfs_opterror(opts,
1280 			    "ip4 cannot be changed after creation");
1281 			goto done_deref_locked;
1282 		}
1283 #endif
1284 #ifdef INET6
1285 		if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1286 			error = EINVAL;
1287 			vfs_opterror(opts,
1288 			    "ip6 cannot be changed after creation");
1289 			goto done_deref_locked;
1290 		}
1291 #endif
1292 	}
1293 
1294 	/* Do final error checking before setting anything. */
1295 	if (gotslevel) {
1296 		if (slevel < ppr->pr_securelevel) {
1297 			error = EPERM;
1298 			goto done_deref_locked;
1299 		}
1300 	}
1301 	if (gotchildmax) {
1302 		if (childmax >= ppr->pr_childmax) {
1303 			error = EPERM;
1304 			goto done_deref_locked;
1305 		}
1306 	}
1307 	if (gotenforce) {
1308 		if (enforce < ppr->pr_enforce_statfs) {
1309 			error = EPERM;
1310 			goto done_deref_locked;
1311 		}
1312 	}
1313 #ifdef INET
1314 	if (ip4s > 0) {
1315 		if (ppr->pr_flags & PR_IP4) {
1316 			/*
1317 			 * Make sure the new set of IP addresses is a
1318 			 * subset of the parent's list.  Don't worry
1319 			 * about the parent being unlocked, as any
1320 			 * setting is done with allprison_lock held.
1321 			 */
1322 			for (ij = 0; ij < ppr->pr_ip4s; ij++)
1323 				if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1324 					break;
1325 			if (ij == ppr->pr_ip4s) {
1326 				error = EPERM;
1327 				goto done_deref_locked;
1328 			}
1329 			if (ip4s > 1) {
1330 				for (ii = ij = 1; ii < ip4s; ii++) {
1331 					if (ip4[ii].s_addr ==
1332 					    ppr->pr_ip4[0].s_addr)
1333 						continue;
1334 					for (; ij < ppr->pr_ip4s; ij++)
1335 						if (ip4[ii].s_addr ==
1336 						    ppr->pr_ip4[ij].s_addr)
1337 							break;
1338 					if (ij == ppr->pr_ip4s)
1339 						break;
1340 				}
1341 				if (ij == ppr->pr_ip4s) {
1342 					error = EPERM;
1343 					goto done_deref_locked;
1344 				}
1345 			}
1346 		}
1347 		/*
1348 		 * Check for conflicting IP addresses.  We permit them
1349 		 * if there is no more than one IP on each jail.  If
1350 		 * there is a duplicate on a jail with more than one
1351 		 * IP stop checking and return error.
1352 		 */
1353 		tppr = ppr;
1354 #ifdef VIMAGE
1355 		for (; tppr != &prison0; tppr = tppr->pr_parent)
1356 			if (tppr->pr_flags & PR_VNET)
1357 				break;
1358 #endif
1359 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1360 			if (tpr == pr ||
1361 #ifdef VIMAGE
1362 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1363 #endif
1364 			    tpr->pr_uref == 0) {
1365 				descend = 0;
1366 				continue;
1367 			}
1368 			if (!(tpr->pr_flags & PR_IP4_USER))
1369 				continue;
1370 			descend = 0;
1371 			if (tpr->pr_ip4 == NULL ||
1372 			    (ip4s == 1 && tpr->pr_ip4s == 1))
1373 				continue;
1374 			for (ii = 0; ii < ip4s; ii++) {
1375 				if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1376 					error = EADDRINUSE;
1377 					vfs_opterror(opts,
1378 					    "IPv4 addresses clash");
1379 					goto done_deref_locked;
1380 				}
1381 			}
1382 		}
1383 	}
1384 #endif
1385 #ifdef INET6
1386 	if (ip6s > 0) {
1387 		if (ppr->pr_flags & PR_IP6) {
1388 			/*
1389 			 * Make sure the new set of IP addresses is a
1390 			 * subset of the parent's list.
1391 			 */
1392 			for (ij = 0; ij < ppr->pr_ip6s; ij++)
1393 				if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1394 				    &ppr->pr_ip6[ij]))
1395 					break;
1396 			if (ij == ppr->pr_ip6s) {
1397 				error = EPERM;
1398 				goto done_deref_locked;
1399 			}
1400 			if (ip6s > 1) {
1401 				for (ii = ij = 1; ii < ip6s; ii++) {
1402 					if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1403 					     &ppr->pr_ip6[0]))
1404 						continue;
1405 					for (; ij < ppr->pr_ip6s; ij++)
1406 						if (IN6_ARE_ADDR_EQUAL(
1407 						    &ip6[ii], &ppr->pr_ip6[ij]))
1408 							break;
1409 					if (ij == ppr->pr_ip6s)
1410 						break;
1411 				}
1412 				if (ij == ppr->pr_ip6s) {
1413 					error = EPERM;
1414 					goto done_deref_locked;
1415 				}
1416 			}
1417 		}
1418 		/* Check for conflicting IP addresses. */
1419 		tppr = ppr;
1420 #ifdef VIMAGE
1421 		for (; tppr != &prison0; tppr = tppr->pr_parent)
1422 			if (tppr->pr_flags & PR_VNET)
1423 				break;
1424 #endif
1425 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1426 			if (tpr == pr ||
1427 #ifdef VIMAGE
1428 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1429 #endif
1430 			    tpr->pr_uref == 0) {
1431 				descend = 0;
1432 				continue;
1433 			}
1434 			if (!(tpr->pr_flags & PR_IP6_USER))
1435 				continue;
1436 			descend = 0;
1437 			if (tpr->pr_ip6 == NULL ||
1438 			    (ip6s == 1 && tpr->pr_ip6s == 1))
1439 				continue;
1440 			for (ii = 0; ii < ip6s; ii++) {
1441 				if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1442 					error = EADDRINUSE;
1443 					vfs_opterror(opts,
1444 					    "IPv6 addresses clash");
1445 					goto done_deref_locked;
1446 				}
1447 			}
1448 		}
1449 	}
1450 #endif
1451 	onamelen = namelen = 0;
1452 	if (name != NULL) {
1453 		/* Give a default name of the jid. */
1454 		if (name[0] == '\0')
1455 			snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1456 		else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1457 		    *p == '\0')) {
1458 			error = EINVAL;
1459 			vfs_opterror(opts,
1460 			    "name cannot be numeric (unless it is the jid)");
1461 			goto done_deref_locked;
1462 		}
1463 		/*
1464 		 * Make sure the name isn't too long for the prison or its
1465 		 * children.
1466 		 */
1467 		onamelen = strlen(pr->pr_name);
1468 		namelen = strlen(name);
1469 		if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1470 			error = ENAMETOOLONG;
1471 			goto done_deref_locked;
1472 		}
1473 		FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1474 			if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1475 			    sizeof(pr->pr_name)) {
1476 				error = ENAMETOOLONG;
1477 				goto done_deref_locked;
1478 			}
1479 		}
1480 	}
1481 	if (pr_allow & ~ppr->pr_allow) {
1482 		error = EPERM;
1483 		goto done_deref_locked;
1484 	}
1485 
1486 	/* Set the parameters of the prison. */
1487 #ifdef INET
1488 	redo_ip4 = 0;
1489 	if (pr_flags & PR_IP4_USER) {
1490 		pr->pr_flags |= PR_IP4;
1491 		free(pr->pr_ip4, M_PRISON);
1492 		pr->pr_ip4s = ip4s;
1493 		pr->pr_ip4 = ip4;
1494 		ip4 = NULL;
1495 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1496 #ifdef VIMAGE
1497 			if (tpr->pr_flags & PR_VNET) {
1498 				descend = 0;
1499 				continue;
1500 			}
1501 #endif
1502 			if (prison_restrict_ip4(tpr, NULL)) {
1503 				redo_ip4 = 1;
1504 				descend = 0;
1505 			}
1506 		}
1507 	}
1508 #endif
1509 #ifdef INET6
1510 	redo_ip6 = 0;
1511 	if (pr_flags & PR_IP6_USER) {
1512 		pr->pr_flags |= PR_IP6;
1513 		free(pr->pr_ip6, M_PRISON);
1514 		pr->pr_ip6s = ip6s;
1515 		pr->pr_ip6 = ip6;
1516 		ip6 = NULL;
1517 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1518 #ifdef VIMAGE
1519 			if (tpr->pr_flags & PR_VNET) {
1520 				descend = 0;
1521 				continue;
1522 			}
1523 #endif
1524 			if (prison_restrict_ip6(tpr, NULL)) {
1525 				redo_ip6 = 1;
1526 				descend = 0;
1527 			}
1528 		}
1529 	}
1530 #endif
1531 	if (gotslevel) {
1532 		pr->pr_securelevel = slevel;
1533 		/* Set all child jails to be at least this level. */
1534 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1535 			if (tpr->pr_securelevel < slevel)
1536 				tpr->pr_securelevel = slevel;
1537 	}
1538 	if (gotchildmax) {
1539 		pr->pr_childmax = childmax;
1540 		/* Set all child jails to under this limit. */
1541 		FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1542 			if (tpr->pr_childmax > childmax - level)
1543 				tpr->pr_childmax = childmax > level
1544 				    ? childmax - level : 0;
1545 	}
1546 	if (gotenforce) {
1547 		pr->pr_enforce_statfs = enforce;
1548 		/* Pass this restriction on to the children. */
1549 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1550 			if (tpr->pr_enforce_statfs < enforce)
1551 				tpr->pr_enforce_statfs = enforce;
1552 	}
1553 	if (name != NULL) {
1554 		if (ppr == &prison0)
1555 			strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1556 		else
1557 			snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1558 			    ppr->pr_name, name);
1559 		/* Change this component of child names. */
1560 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1561 			bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1562 			    strlen(tpr->pr_name + onamelen) + 1);
1563 			bcopy(pr->pr_name, tpr->pr_name, namelen);
1564 		}
1565 	}
1566 	if (path != NULL) {
1567 		/* Try to keep a real-rooted full pathname. */
1568 		if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
1569 			snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1570 			    mypr->pr_path, path);
1571 		else
1572 			strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1573 		pr->pr_root = root;
1574 	}
1575 	if (PR_HOST & ch_flags & ~pr_flags) {
1576 		if (pr->pr_flags & PR_HOST) {
1577 			/*
1578 			 * Copy the parent's host info.  As with pr_ip4 above,
1579 			 * the lack of a lock on the parent is not a problem;
1580 			 * it is always set with allprison_lock at least
1581 			 * shared, and is held exclusively here.
1582 			 */
1583 			strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1584 			    sizeof(pr->pr_hostname));
1585 			strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1586 			    sizeof(pr->pr_domainname));
1587 			strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1588 			    sizeof(pr->pr_hostuuid));
1589 			pr->pr_hostid = pr->pr_parent->pr_hostid;
1590 		}
1591 	} else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1592 		/* Set this prison, and any descendants without PR_HOST. */
1593 		if (host != NULL)
1594 			strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1595 		if (domain != NULL)
1596 			strlcpy(pr->pr_domainname, domain,
1597 			    sizeof(pr->pr_domainname));
1598 		if (uuid != NULL)
1599 			strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1600 		if (gothid)
1601 			pr->pr_hostid = hid;
1602 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1603 			if (tpr->pr_flags & PR_HOST)
1604 				descend = 0;
1605 			else {
1606 				if (host != NULL)
1607 					strlcpy(tpr->pr_hostname,
1608 					    pr->pr_hostname,
1609 					    sizeof(tpr->pr_hostname));
1610 				if (domain != NULL)
1611 					strlcpy(tpr->pr_domainname,
1612 					    pr->pr_domainname,
1613 					    sizeof(tpr->pr_domainname));
1614 				if (uuid != NULL)
1615 					strlcpy(tpr->pr_hostuuid,
1616 					    pr->pr_hostuuid,
1617 					    sizeof(tpr->pr_hostuuid));
1618 				if (gothid)
1619 					tpr->pr_hostid = hid;
1620 			}
1621 		}
1622 	}
1623 	if ((tallow = ch_allow & ~pr_allow)) {
1624 		/* Clear allow bits in all children. */
1625 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1626 			tpr->pr_allow &= ~tallow;
1627 	}
1628 	pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1629 	/*
1630 	 * Persistent prisons get an extra reference, and prisons losing their
1631 	 * persist flag lose that reference.  Only do this for existing prisons
1632 	 * for now, so new ones will remain unseen until after the module
1633 	 * handlers have completed.
1634 	 */
1635 	if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1636 		if (pr_flags & PR_PERSIST) {
1637 			pr->pr_ref++;
1638 			pr->pr_uref++;
1639 		} else {
1640 			pr->pr_ref--;
1641 			pr->pr_uref--;
1642 		}
1643 	}
1644 	pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1645 	mtx_unlock(&pr->pr_mtx);
1646 
1647 	/* Locks may have prevented a complete restriction of child IP
1648 	 * addresses.  If so, allocate some more memory and try again.
1649 	 */
1650 #ifdef INET
1651 	while (redo_ip4) {
1652 		ip4s = pr->pr_ip4s;
1653 		ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1654 		mtx_lock(&pr->pr_mtx);
1655 		redo_ip4 = 0;
1656 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1657 #ifdef VIMAGE
1658 			if (tpr->pr_flags & PR_VNET) {
1659 				descend = 0;
1660 				continue;
1661 			}
1662 #endif
1663 			if (prison_restrict_ip4(tpr, ip4)) {
1664 				if (ip4 != NULL)
1665 					ip4 = NULL;
1666 				else
1667 					redo_ip4 = 1;
1668 			}
1669 		}
1670 		mtx_unlock(&pr->pr_mtx);
1671 	}
1672 #endif
1673 #ifdef INET6
1674 	while (redo_ip6) {
1675 		ip6s = pr->pr_ip6s;
1676 		ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1677 		mtx_lock(&pr->pr_mtx);
1678 		redo_ip6 = 0;
1679 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1680 #ifdef VIMAGE
1681 			if (tpr->pr_flags & PR_VNET) {
1682 				descend = 0;
1683 				continue;
1684 			}
1685 #endif
1686 			if (prison_restrict_ip6(tpr, ip6)) {
1687 				if (ip6 != NULL)
1688 					ip6 = NULL;
1689 				else
1690 					redo_ip6 = 1;
1691 			}
1692 		}
1693 		mtx_unlock(&pr->pr_mtx);
1694 	}
1695 #endif
1696 
1697 	/* Let the modules do their work. */
1698 	sx_downgrade(&allprison_lock);
1699 	if (created) {
1700 		error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1701 		if (error) {
1702 			prison_deref(pr, PD_LIST_SLOCKED);
1703 			goto done_errmsg;
1704 		}
1705 	}
1706 	error = osd_jail_call(pr, PR_METHOD_SET, opts);
1707 	if (error) {
1708 		prison_deref(pr, created
1709 		    ? PD_LIST_SLOCKED
1710 		    : PD_DEREF | PD_LIST_SLOCKED);
1711 		goto done_errmsg;
1712 	}
1713 
1714 	/* Attach this process to the prison if requested. */
1715 	if (flags & JAIL_ATTACH) {
1716 		mtx_lock(&pr->pr_mtx);
1717 		error = do_jail_attach(td, pr);
1718 		if (error) {
1719 			vfs_opterror(opts, "attach failed");
1720 			if (!created)
1721 				prison_deref(pr, PD_DEREF);
1722 			goto done_errmsg;
1723 		}
1724 	}
1725 
1726 	/*
1727 	 * Now that it is all there, drop the temporary reference from existing
1728 	 * prisons.  Or add a reference to newly created persistent prisons
1729 	 * (which was not done earlier so that the prison would not be publicly
1730 	 * visible).
1731 	 */
1732 	if (!created) {
1733 		prison_deref(pr, (flags & JAIL_ATTACH)
1734 		    ? PD_DEREF
1735 		    : PD_DEREF | PD_LIST_SLOCKED);
1736 	} else {
1737 		if (pr_flags & PR_PERSIST) {
1738 			mtx_lock(&pr->pr_mtx);
1739 			pr->pr_ref++;
1740 			pr->pr_uref++;
1741 			mtx_unlock(&pr->pr_mtx);
1742 		}
1743 		if (!(flags & JAIL_ATTACH))
1744 			sx_sunlock(&allprison_lock);
1745 	}
1746 	td->td_retval[0] = pr->pr_id;
1747 	goto done_errmsg;
1748 
1749  done_deref_locked:
1750 	prison_deref(pr, created
1751 	    ? PD_LOCKED | PD_LIST_XLOCKED
1752 	    : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1753 	goto done_releroot;
1754  done_unlock_list:
1755 	sx_xunlock(&allprison_lock);
1756  done_releroot:
1757 	if (root != NULL) {
1758 		vfslocked = VFS_LOCK_GIANT(root->v_mount);
1759 		vrele(root);
1760 		VFS_UNLOCK_GIANT(vfslocked);
1761 	}
1762  done_errmsg:
1763 	if (error) {
1764 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1765 		if (errmsg_len > 0) {
1766 			errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1767 			if (errmsg_pos > 0) {
1768 				if (optuio->uio_segflg == UIO_SYSSPACE)
1769 					bcopy(errmsg,
1770 					   optuio->uio_iov[errmsg_pos].iov_base,
1771 					   errmsg_len);
1772 				else
1773 					copyout(errmsg,
1774 					   optuio->uio_iov[errmsg_pos].iov_base,
1775 					   errmsg_len);
1776 			}
1777 		}
1778 	}
1779  done_free:
1780 #ifdef INET
1781 	free(ip4, M_PRISON);
1782 #endif
1783 #ifdef INET6
1784 	free(ip6, M_PRISON);
1785 #endif
1786 	vfs_freeopts(opts);
1787 	return (error);
1788 }
1789 
1790 
1791 /*
1792  * struct jail_get_args {
1793  *	struct iovec *iovp;
1794  *	unsigned int iovcnt;
1795  *	int flags;
1796  * };
1797  */
1798 int
1799 jail_get(struct thread *td, struct jail_get_args *uap)
1800 {
1801 	struct uio *auio;
1802 	int error;
1803 
1804 	/* Check that we have an even number of iovecs. */
1805 	if (uap->iovcnt & 1)
1806 		return (EINVAL);
1807 
1808 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1809 	if (error)
1810 		return (error);
1811 	error = kern_jail_get(td, auio, uap->flags);
1812 	if (error == 0)
1813 		error = copyout(auio->uio_iov, uap->iovp,
1814 		    uap->iovcnt * sizeof (struct iovec));
1815 	free(auio, M_IOV);
1816 	return (error);
1817 }
1818 
1819 int
1820 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1821 {
1822 	struct prison *pr, *mypr;
1823 	struct vfsopt *opt;
1824 	struct vfsoptlist *opts;
1825 	char *errmsg, *name;
1826 	int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1827 
1828 	if (flags & ~JAIL_GET_MASK)
1829 		return (EINVAL);
1830 
1831 	/* Get the parameter list. */
1832 	error = vfs_buildopts(optuio, &opts);
1833 	if (error)
1834 		return (error);
1835 	errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1836 	mypr = td->td_ucred->cr_prison;
1837 
1838 	/*
1839 	 * Find the prison specified by one of: lastjid, jid, name.
1840 	 */
1841 	sx_slock(&allprison_lock);
1842 	error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1843 	if (error == 0) {
1844 		TAILQ_FOREACH(pr, &allprison, pr_list) {
1845 			if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1846 				mtx_lock(&pr->pr_mtx);
1847 				if (pr->pr_ref > 0 &&
1848 				    (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1849 					break;
1850 				mtx_unlock(&pr->pr_mtx);
1851 			}
1852 		}
1853 		if (pr != NULL)
1854 			goto found_prison;
1855 		error = ENOENT;
1856 		vfs_opterror(opts, "no jail after %d", jid);
1857 		goto done_unlock_list;
1858 	} else if (error != ENOENT)
1859 		goto done_unlock_list;
1860 
1861 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1862 	if (error == 0) {
1863 		if (jid != 0) {
1864 			pr = prison_find_child(mypr, jid);
1865 			if (pr != NULL) {
1866 				if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1867 					mtx_unlock(&pr->pr_mtx);
1868 					error = ENOENT;
1869 					vfs_opterror(opts, "jail %d is dying",
1870 					    jid);
1871 					goto done_unlock_list;
1872 				}
1873 				goto found_prison;
1874 			}
1875 			error = ENOENT;
1876 			vfs_opterror(opts, "jail %d not found", jid);
1877 			goto done_unlock_list;
1878 		}
1879 	} else if (error != ENOENT)
1880 		goto done_unlock_list;
1881 
1882 	error = vfs_getopt(opts, "name", (void **)&name, &len);
1883 	if (error == 0) {
1884 		if (len == 0 || name[len - 1] != '\0') {
1885 			error = EINVAL;
1886 			goto done_unlock_list;
1887 		}
1888 		pr = prison_find_name(mypr, name);
1889 		if (pr != NULL) {
1890 			if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1891 				mtx_unlock(&pr->pr_mtx);
1892 				error = ENOENT;
1893 				vfs_opterror(opts, "jail \"%s\" is dying",
1894 				    name);
1895 				goto done_unlock_list;
1896 			}
1897 			goto found_prison;
1898 		}
1899 		error = ENOENT;
1900 		vfs_opterror(opts, "jail \"%s\" not found", name);
1901 		goto done_unlock_list;
1902 	} else if (error != ENOENT)
1903 		goto done_unlock_list;
1904 
1905 	vfs_opterror(opts, "no jail specified");
1906 	error = ENOENT;
1907 	goto done_unlock_list;
1908 
1909  found_prison:
1910 	/* Get the parameters of the prison. */
1911 	pr->pr_ref++;
1912 	locked = PD_LOCKED;
1913 	td->td_retval[0] = pr->pr_id;
1914 	error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
1915 	if (error != 0 && error != ENOENT)
1916 		goto done_deref;
1917 	i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
1918 	error = vfs_setopt(opts, "parent", &i, sizeof(i));
1919 	if (error != 0 && error != ENOENT)
1920 		goto done_deref;
1921 	error = vfs_setopts(opts, "name", prison_name(mypr, pr));
1922 	if (error != 0 && error != ENOENT)
1923 		goto done_deref;
1924 	error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
1925 	    sizeof(pr->pr_cpuset->cs_id));
1926 	if (error != 0 && error != ENOENT)
1927 		goto done_deref;
1928 	error = vfs_setopts(opts, "path", prison_path(mypr, pr));
1929 	if (error != 0 && error != ENOENT)
1930 		goto done_deref;
1931 #ifdef INET
1932 	error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
1933 	    pr->pr_ip4s * sizeof(*pr->pr_ip4));
1934 	if (error != 0 && error != ENOENT)
1935 		goto done_deref;
1936 #endif
1937 #ifdef INET6
1938 	error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
1939 	    pr->pr_ip6s * sizeof(*pr->pr_ip6));
1940 	if (error != 0 && error != ENOENT)
1941 		goto done_deref;
1942 #endif
1943 	error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
1944 	    sizeof(pr->pr_securelevel));
1945 	if (error != 0 && error != ENOENT)
1946 		goto done_deref;
1947 	error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
1948 	    sizeof(pr->pr_childcount));
1949 	if (error != 0 && error != ENOENT)
1950 		goto done_deref;
1951 	error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
1952 	    sizeof(pr->pr_childmax));
1953 	if (error != 0 && error != ENOENT)
1954 		goto done_deref;
1955 	error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
1956 	if (error != 0 && error != ENOENT)
1957 		goto done_deref;
1958 	error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
1959 	if (error != 0 && error != ENOENT)
1960 		goto done_deref;
1961 	error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
1962 	if (error != 0 && error != ENOENT)
1963 		goto done_deref;
1964 #ifdef COMPAT_FREEBSD32
1965 	if (td->td_proc->p_sysent->sv_flags & SV_ILP32) {
1966 		uint32_t hid32 = pr->pr_hostid;
1967 
1968 		error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
1969 	} else
1970 #endif
1971 	error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
1972 	    sizeof(pr->pr_hostid));
1973 	if (error != 0 && error != ENOENT)
1974 		goto done_deref;
1975 	error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
1976 	    sizeof(pr->pr_enforce_statfs));
1977 	if (error != 0 && error != ENOENT)
1978 		goto done_deref;
1979 	for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
1980 	    fi++) {
1981 		if (pr_flag_names[fi] == NULL)
1982 			continue;
1983 		i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
1984 		error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
1985 		if (error != 0 && error != ENOENT)
1986 			goto done_deref;
1987 		i = !i;
1988 		error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
1989 		if (error != 0 && error != ENOENT)
1990 			goto done_deref;
1991 	}
1992 	for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
1993 	    fi++) {
1994 		i = pr->pr_flags &
1995 		    (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
1996 		i = pr_flag_jailsys[fi].disable &&
1997 		      (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
1998 		    : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
1999 		    : JAIL_SYS_INHERIT;
2000 		error =
2001 		    vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2002 		if (error != 0 && error != ENOENT)
2003 			goto done_deref;
2004 	}
2005 	for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2006 	    fi++) {
2007 		if (pr_allow_names[fi] == NULL)
2008 			continue;
2009 		i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2010 		error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2011 		if (error != 0 && error != ENOENT)
2012 			goto done_deref;
2013 		i = !i;
2014 		error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2015 		if (error != 0 && error != ENOENT)
2016 			goto done_deref;
2017 	}
2018 	i = (pr->pr_uref == 0);
2019 	error = vfs_setopt(opts, "dying", &i, sizeof(i));
2020 	if (error != 0 && error != ENOENT)
2021 		goto done_deref;
2022 	i = !i;
2023 	error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2024 	if (error != 0 && error != ENOENT)
2025 		goto done_deref;
2026 
2027 	/* Get the module parameters. */
2028 	mtx_unlock(&pr->pr_mtx);
2029 	locked = 0;
2030 	error = osd_jail_call(pr, PR_METHOD_GET, opts);
2031 	if (error)
2032 		goto done_deref;
2033 	prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2034 
2035 	/* By now, all parameters should have been noted. */
2036 	TAILQ_FOREACH(opt, opts, link) {
2037 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
2038 			error = EINVAL;
2039 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
2040 			goto done_errmsg;
2041 		}
2042 	}
2043 
2044 	/* Write the fetched parameters back to userspace. */
2045 	error = 0;
2046 	TAILQ_FOREACH(opt, opts, link) {
2047 		if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2048 			pos = 2 * opt->pos + 1;
2049 			optuio->uio_iov[pos].iov_len = opt->len;
2050 			if (opt->value != NULL) {
2051 				if (optuio->uio_segflg == UIO_SYSSPACE) {
2052 					bcopy(opt->value,
2053 					    optuio->uio_iov[pos].iov_base,
2054 					    opt->len);
2055 				} else {
2056 					error = copyout(opt->value,
2057 					    optuio->uio_iov[pos].iov_base,
2058 					    opt->len);
2059 					if (error)
2060 						break;
2061 				}
2062 			}
2063 		}
2064 	}
2065 	goto done_errmsg;
2066 
2067  done_deref:
2068 	prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2069 	goto done_errmsg;
2070 
2071  done_unlock_list:
2072 	sx_sunlock(&allprison_lock);
2073  done_errmsg:
2074 	if (error && errmsg_pos >= 0) {
2075 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2076 		errmsg_pos = 2 * errmsg_pos + 1;
2077 		if (errmsg_len > 0) {
2078 			if (optuio->uio_segflg == UIO_SYSSPACE)
2079 				bcopy(errmsg,
2080 				    optuio->uio_iov[errmsg_pos].iov_base,
2081 				    errmsg_len);
2082 			else
2083 				copyout(errmsg,
2084 				    optuio->uio_iov[errmsg_pos].iov_base,
2085 				    errmsg_len);
2086 		}
2087 	}
2088 	vfs_freeopts(opts);
2089 	return (error);
2090 }
2091 
2092 
2093 /*
2094  * struct jail_remove_args {
2095  *	int jid;
2096  * };
2097  */
2098 int
2099 jail_remove(struct thread *td, struct jail_remove_args *uap)
2100 {
2101 	struct prison *pr, *cpr, *lpr, *tpr;
2102 	int descend, error;
2103 
2104 	error = priv_check(td, PRIV_JAIL_REMOVE);
2105 	if (error)
2106 		return (error);
2107 
2108 	sx_xlock(&allprison_lock);
2109 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2110 	if (pr == NULL) {
2111 		sx_xunlock(&allprison_lock);
2112 		return (EINVAL);
2113 	}
2114 
2115 	/* Remove all descendants of this prison, then remove this prison. */
2116 	pr->pr_ref++;
2117 	pr->pr_flags |= PR_REMOVE;
2118 	if (!LIST_EMPTY(&pr->pr_children)) {
2119 		mtx_unlock(&pr->pr_mtx);
2120 		lpr = NULL;
2121 		FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2122 			mtx_lock(&cpr->pr_mtx);
2123 			if (cpr->pr_ref > 0) {
2124 				tpr = cpr;
2125 				cpr->pr_ref++;
2126 				cpr->pr_flags |= PR_REMOVE;
2127 			} else {
2128 				/* Already removed - do not do it again. */
2129 				tpr = NULL;
2130 			}
2131 			mtx_unlock(&cpr->pr_mtx);
2132 			if (lpr != NULL) {
2133 				mtx_lock(&lpr->pr_mtx);
2134 				prison_remove_one(lpr);
2135 				sx_xlock(&allprison_lock);
2136 			}
2137 			lpr = tpr;
2138 		}
2139 		if (lpr != NULL) {
2140 			mtx_lock(&lpr->pr_mtx);
2141 			prison_remove_one(lpr);
2142 			sx_xlock(&allprison_lock);
2143 		}
2144 		mtx_lock(&pr->pr_mtx);
2145 	}
2146 	prison_remove_one(pr);
2147 	return (0);
2148 }
2149 
2150 static void
2151 prison_remove_one(struct prison *pr)
2152 {
2153 	struct proc *p;
2154 	int deuref;
2155 
2156 	/* If the prison was persistent, it is not anymore. */
2157 	deuref = 0;
2158 	if (pr->pr_flags & PR_PERSIST) {
2159 		pr->pr_ref--;
2160 		deuref = PD_DEUREF;
2161 		pr->pr_flags &= ~PR_PERSIST;
2162 	}
2163 
2164 	/*
2165 	 * jail_remove added a reference.  If that's the only one, remove
2166 	 * the prison now.
2167 	 */
2168 	KASSERT(pr->pr_ref > 0,
2169 	    ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2170 	if (pr->pr_ref == 1) {
2171 		prison_deref(pr,
2172 		    deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2173 		return;
2174 	}
2175 
2176 	mtx_unlock(&pr->pr_mtx);
2177 	sx_xunlock(&allprison_lock);
2178 	/*
2179 	 * Kill all processes unfortunate enough to be attached to this prison.
2180 	 */
2181 	sx_slock(&allproc_lock);
2182 	LIST_FOREACH(p, &allproc, p_list) {
2183 		PROC_LOCK(p);
2184 		if (p->p_state != PRS_NEW && p->p_ucred &&
2185 		    p->p_ucred->cr_prison == pr)
2186 			psignal(p, SIGKILL);
2187 		PROC_UNLOCK(p);
2188 	}
2189 	sx_sunlock(&allproc_lock);
2190 	/* Remove the temporary reference added by jail_remove. */
2191 	prison_deref(pr, deuref | PD_DEREF);
2192 }
2193 
2194 
2195 /*
2196  * struct jail_attach_args {
2197  *	int jid;
2198  * };
2199  */
2200 int
2201 jail_attach(struct thread *td, struct jail_attach_args *uap)
2202 {
2203 	struct prison *pr;
2204 	int error;
2205 
2206 	error = priv_check(td, PRIV_JAIL_ATTACH);
2207 	if (error)
2208 		return (error);
2209 
2210 	sx_slock(&allprison_lock);
2211 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2212 	if (pr == NULL) {
2213 		sx_sunlock(&allprison_lock);
2214 		return (EINVAL);
2215 	}
2216 
2217 	/*
2218 	 * Do not allow a process to attach to a prison that is not
2219 	 * considered to be "alive".
2220 	 */
2221 	if (pr->pr_uref == 0) {
2222 		mtx_unlock(&pr->pr_mtx);
2223 		sx_sunlock(&allprison_lock);
2224 		return (EINVAL);
2225 	}
2226 
2227 	return (do_jail_attach(td, pr));
2228 }
2229 
2230 static int
2231 do_jail_attach(struct thread *td, struct prison *pr)
2232 {
2233 	struct prison *ppr;
2234 	struct proc *p;
2235 	struct ucred *newcred, *oldcred;
2236 	int vfslocked, error;
2237 
2238 	/*
2239 	 * XXX: Note that there is a slight race here if two threads
2240 	 * in the same privileged process attempt to attach to two
2241 	 * different jails at the same time.  It is important for
2242 	 * user processes not to do this, or they might end up with
2243 	 * a process root from one prison, but attached to the jail
2244 	 * of another.
2245 	 */
2246 	pr->pr_ref++;
2247 	pr->pr_uref++;
2248 	mtx_unlock(&pr->pr_mtx);
2249 
2250 	/* Let modules do whatever they need to prepare for attaching. */
2251 	error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2252 	if (error) {
2253 		prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2254 		return (error);
2255 	}
2256 	sx_sunlock(&allprison_lock);
2257 
2258 	/*
2259 	 * Reparent the newly attached process to this jail.
2260 	 */
2261 	ppr = td->td_ucred->cr_prison;
2262 	p = td->td_proc;
2263 	error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2264 	if (error)
2265 		goto e_revert_osd;
2266 
2267 	vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2268 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2269 	if ((error = change_dir(pr->pr_root, td)) != 0)
2270 		goto e_unlock;
2271 #ifdef MAC
2272 	if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2273 		goto e_unlock;
2274 #endif
2275 	VOP_UNLOCK(pr->pr_root, 0);
2276 	if ((error = change_root(pr->pr_root, td)))
2277 		goto e_unlock_giant;
2278 	VFS_UNLOCK_GIANT(vfslocked);
2279 
2280 	newcred = crget();
2281 	PROC_LOCK(p);
2282 	oldcred = p->p_ucred;
2283 	setsugid(p);
2284 	crcopy(newcred, oldcred);
2285 	newcred->cr_prison = pr;
2286 	p->p_ucred = newcred;
2287 	PROC_UNLOCK(p);
2288 	crfree(oldcred);
2289 	prison_deref(ppr, PD_DEREF | PD_DEUREF);
2290 	return (0);
2291  e_unlock:
2292 	VOP_UNLOCK(pr->pr_root, 0);
2293  e_unlock_giant:
2294 	VFS_UNLOCK_GIANT(vfslocked);
2295  e_revert_osd:
2296 	/* Tell modules this thread is still in its old jail after all. */
2297 	(void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2298 	prison_deref(pr, PD_DEREF | PD_DEUREF);
2299 	return (error);
2300 }
2301 
2302 
2303 /*
2304  * Returns a locked prison instance, or NULL on failure.
2305  */
2306 struct prison *
2307 prison_find(int prid)
2308 {
2309 	struct prison *pr;
2310 
2311 	sx_assert(&allprison_lock, SX_LOCKED);
2312 	TAILQ_FOREACH(pr, &allprison, pr_list) {
2313 		if (pr->pr_id == prid) {
2314 			mtx_lock(&pr->pr_mtx);
2315 			if (pr->pr_ref > 0)
2316 				return (pr);
2317 			mtx_unlock(&pr->pr_mtx);
2318 		}
2319 	}
2320 	return (NULL);
2321 }
2322 
2323 /*
2324  * Find a prison that is a descendant of mypr.  Returns a locked prison or NULL.
2325  */
2326 struct prison *
2327 prison_find_child(struct prison *mypr, int prid)
2328 {
2329 	struct prison *pr;
2330 	int descend;
2331 
2332 	sx_assert(&allprison_lock, SX_LOCKED);
2333 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2334 		if (pr->pr_id == prid) {
2335 			mtx_lock(&pr->pr_mtx);
2336 			if (pr->pr_ref > 0)
2337 				return (pr);
2338 			mtx_unlock(&pr->pr_mtx);
2339 		}
2340 	}
2341 	return (NULL);
2342 }
2343 
2344 /*
2345  * Look for the name relative to mypr.  Returns a locked prison or NULL.
2346  */
2347 struct prison *
2348 prison_find_name(struct prison *mypr, const char *name)
2349 {
2350 	struct prison *pr, *deadpr;
2351 	size_t mylen;
2352 	int descend;
2353 
2354 	sx_assert(&allprison_lock, SX_LOCKED);
2355 	mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2356  again:
2357 	deadpr = NULL;
2358 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2359 		if (!strcmp(pr->pr_name + mylen, name)) {
2360 			mtx_lock(&pr->pr_mtx);
2361 			if (pr->pr_ref > 0) {
2362 				if (pr->pr_uref > 0)
2363 					return (pr);
2364 				deadpr = pr;
2365 			}
2366 			mtx_unlock(&pr->pr_mtx);
2367 		}
2368 	}
2369 	/* There was no valid prison - perhaps there was a dying one. */
2370 	if (deadpr != NULL) {
2371 		mtx_lock(&deadpr->pr_mtx);
2372 		if (deadpr->pr_ref == 0) {
2373 			mtx_unlock(&deadpr->pr_mtx);
2374 			goto again;
2375 		}
2376 	}
2377 	return (deadpr);
2378 }
2379 
2380 /*
2381  * See if a prison has the specific flag set.
2382  */
2383 int
2384 prison_flag(struct ucred *cred, unsigned flag)
2385 {
2386 
2387 	/* This is an atomic read, so no locking is necessary. */
2388 	return (cred->cr_prison->pr_flags & flag);
2389 }
2390 
2391 int
2392 prison_allow(struct ucred *cred, unsigned flag)
2393 {
2394 
2395 	/* This is an atomic read, so no locking is necessary. */
2396 	return (cred->cr_prison->pr_allow & flag);
2397 }
2398 
2399 /*
2400  * Remove a prison reference.  If that was the last reference, remove the
2401  * prison itself - but not in this context in case there are locks held.
2402  */
2403 void
2404 prison_free_locked(struct prison *pr)
2405 {
2406 
2407 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2408 	pr->pr_ref--;
2409 	if (pr->pr_ref == 0) {
2410 		mtx_unlock(&pr->pr_mtx);
2411 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2412 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2413 		return;
2414 	}
2415 	mtx_unlock(&pr->pr_mtx);
2416 }
2417 
2418 void
2419 prison_free(struct prison *pr)
2420 {
2421 
2422 	mtx_lock(&pr->pr_mtx);
2423 	prison_free_locked(pr);
2424 }
2425 
2426 static void
2427 prison_complete(void *context, int pending)
2428 {
2429 
2430 	prison_deref((struct prison *)context, 0);
2431 }
2432 
2433 /*
2434  * Remove a prison reference (usually).  This internal version assumes no
2435  * mutexes are held, except perhaps the prison itself.  If there are no more
2436  * references, release and delist the prison.  On completion, the prison lock
2437  * and the allprison lock are both unlocked.
2438  */
2439 static void
2440 prison_deref(struct prison *pr, int flags)
2441 {
2442 	struct prison *ppr, *tpr;
2443 	int vfslocked;
2444 
2445 	if (!(flags & PD_LOCKED))
2446 		mtx_lock(&pr->pr_mtx);
2447 	/* Decrement the user references in a separate loop. */
2448 	if (flags & PD_DEUREF) {
2449 		for (tpr = pr;; tpr = tpr->pr_parent) {
2450 			if (tpr != pr)
2451 				mtx_lock(&tpr->pr_mtx);
2452 			if (--tpr->pr_uref > 0)
2453 				break;
2454 			KASSERT(tpr != &prison0, ("prison0 pr_uref=0"));
2455 			mtx_unlock(&tpr->pr_mtx);
2456 		}
2457 		/* Done if there were only user references to remove. */
2458 		if (!(flags & PD_DEREF)) {
2459 			mtx_unlock(&tpr->pr_mtx);
2460 			if (flags & PD_LIST_SLOCKED)
2461 				sx_sunlock(&allprison_lock);
2462 			else if (flags & PD_LIST_XLOCKED)
2463 				sx_xunlock(&allprison_lock);
2464 			return;
2465 		}
2466 		if (tpr != pr) {
2467 			mtx_unlock(&tpr->pr_mtx);
2468 			mtx_lock(&pr->pr_mtx);
2469 		}
2470 	}
2471 
2472 	for (;;) {
2473 		if (flags & PD_DEREF)
2474 			pr->pr_ref--;
2475 		/* If the prison still has references, nothing else to do. */
2476 		if (pr->pr_ref > 0) {
2477 			mtx_unlock(&pr->pr_mtx);
2478 			if (flags & PD_LIST_SLOCKED)
2479 				sx_sunlock(&allprison_lock);
2480 			else if (flags & PD_LIST_XLOCKED)
2481 				sx_xunlock(&allprison_lock);
2482 			return;
2483 		}
2484 
2485 		mtx_unlock(&pr->pr_mtx);
2486 		if (flags & PD_LIST_SLOCKED) {
2487 			if (!sx_try_upgrade(&allprison_lock)) {
2488 				sx_sunlock(&allprison_lock);
2489 				sx_xlock(&allprison_lock);
2490 			}
2491 		} else if (!(flags & PD_LIST_XLOCKED))
2492 			sx_xlock(&allprison_lock);
2493 
2494 		TAILQ_REMOVE(&allprison, pr, pr_list);
2495 		LIST_REMOVE(pr, pr_sibling);
2496 		ppr = pr->pr_parent;
2497 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2498 			tpr->pr_childcount--;
2499 		sx_xunlock(&allprison_lock);
2500 
2501 #ifdef VIMAGE
2502 		if (pr->pr_vnet != ppr->pr_vnet)
2503 			vnet_destroy(pr->pr_vnet);
2504 #endif
2505 		if (pr->pr_root != NULL) {
2506 			vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2507 			vrele(pr->pr_root);
2508 			VFS_UNLOCK_GIANT(vfslocked);
2509 		}
2510 		mtx_destroy(&pr->pr_mtx);
2511 #ifdef INET
2512 		free(pr->pr_ip4, M_PRISON);
2513 #endif
2514 #ifdef INET6
2515 		free(pr->pr_ip6, M_PRISON);
2516 #endif
2517 		if (pr->pr_cpuset != NULL)
2518 			cpuset_rel(pr->pr_cpuset);
2519 		osd_jail_exit(pr);
2520 		free(pr, M_PRISON);
2521 
2522 		/* Removing a prison frees a reference on its parent. */
2523 		pr = ppr;
2524 		mtx_lock(&pr->pr_mtx);
2525 		flags = PD_DEREF;
2526 	}
2527 }
2528 
2529 void
2530 prison_hold_locked(struct prison *pr)
2531 {
2532 
2533 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2534 	KASSERT(pr->pr_ref > 0,
2535 	    ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2536 	pr->pr_ref++;
2537 }
2538 
2539 void
2540 prison_hold(struct prison *pr)
2541 {
2542 
2543 	mtx_lock(&pr->pr_mtx);
2544 	prison_hold_locked(pr);
2545 	mtx_unlock(&pr->pr_mtx);
2546 }
2547 
2548 void
2549 prison_proc_hold(struct prison *pr)
2550 {
2551 
2552 	mtx_lock(&pr->pr_mtx);
2553 	KASSERT(pr->pr_uref > 0,
2554 	    ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2555 	pr->pr_uref++;
2556 	mtx_unlock(&pr->pr_mtx);
2557 }
2558 
2559 void
2560 prison_proc_free(struct prison *pr)
2561 {
2562 
2563 	mtx_lock(&pr->pr_mtx);
2564 	KASSERT(pr->pr_uref > 0,
2565 	    ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2566 	prison_deref(pr, PD_DEUREF | PD_LOCKED);
2567 }
2568 
2569 
2570 #ifdef INET
2571 /*
2572  * Restrict a prison's IP address list with its parent's, possibly replacing
2573  * it.  Return true if the replacement buffer was used (or would have been).
2574  */
2575 static int
2576 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2577 {
2578 	int ii, ij, used;
2579 	struct prison *ppr;
2580 
2581 	ppr = pr->pr_parent;
2582 	if (!(pr->pr_flags & PR_IP4_USER)) {
2583 		/* This has no user settings, so just copy the parent's list. */
2584 		if (pr->pr_ip4s < ppr->pr_ip4s) {
2585 			/*
2586 			 * There's no room for the parent's list.  Use the
2587 			 * new list buffer, which is assumed to be big enough
2588 			 * (if it was passed).  If there's no buffer, try to
2589 			 * allocate one.
2590 			 */
2591 			used = 1;
2592 			if (newip4 == NULL) {
2593 				newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2594 				    M_PRISON, M_NOWAIT);
2595 				if (newip4 != NULL)
2596 					used = 0;
2597 			}
2598 			if (newip4 != NULL) {
2599 				bcopy(ppr->pr_ip4, newip4,
2600 				    ppr->pr_ip4s * sizeof(*newip4));
2601 				free(pr->pr_ip4, M_PRISON);
2602 				pr->pr_ip4 = newip4;
2603 				pr->pr_ip4s = ppr->pr_ip4s;
2604 			}
2605 			return (used);
2606 		}
2607 		pr->pr_ip4s = ppr->pr_ip4s;
2608 		if (pr->pr_ip4s > 0)
2609 			bcopy(ppr->pr_ip4, pr->pr_ip4,
2610 			    pr->pr_ip4s * sizeof(*newip4));
2611 		else if (pr->pr_ip4 != NULL) {
2612 			free(pr->pr_ip4, M_PRISON);
2613 			pr->pr_ip4 = NULL;
2614 		}
2615 	} else if (pr->pr_ip4s > 0) {
2616 		/* Remove addresses that aren't in the parent. */
2617 		for (ij = 0; ij < ppr->pr_ip4s; ij++)
2618 			if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2619 				break;
2620 		if (ij < ppr->pr_ip4s)
2621 			ii = 1;
2622 		else {
2623 			bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2624 			    --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2625 			ii = 0;
2626 		}
2627 		for (ij = 1; ii < pr->pr_ip4s; ) {
2628 			if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2629 				ii++;
2630 				continue;
2631 			}
2632 			switch (ij >= ppr->pr_ip4s ? -1 :
2633 				qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2634 			case -1:
2635 				bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2636 				    (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2637 				break;
2638 			case 0:
2639 				ii++;
2640 				ij++;
2641 				break;
2642 			case 1:
2643 				ij++;
2644 				break;
2645 			}
2646 		}
2647 		if (pr->pr_ip4s == 0) {
2648 			pr->pr_flags |= PR_IP4_DISABLE;
2649 			free(pr->pr_ip4, M_PRISON);
2650 			pr->pr_ip4 = NULL;
2651 		}
2652 	}
2653 	return (0);
2654 }
2655 
2656 /*
2657  * Pass back primary IPv4 address of this jail.
2658  *
2659  * If not restricted return success but do not alter the address.  Caller has
2660  * to make sure to initialize it correctly (e.g. INADDR_ANY).
2661  *
2662  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2663  * Address returned in NBO.
2664  */
2665 int
2666 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2667 {
2668 	struct prison *pr;
2669 
2670 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2671 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2672 
2673 	pr = cred->cr_prison;
2674 	if (!(pr->pr_flags & PR_IP4))
2675 		return (0);
2676 	mtx_lock(&pr->pr_mtx);
2677 	if (!(pr->pr_flags & PR_IP4)) {
2678 		mtx_unlock(&pr->pr_mtx);
2679 		return (0);
2680 	}
2681 	if (pr->pr_ip4 == NULL) {
2682 		mtx_unlock(&pr->pr_mtx);
2683 		return (EAFNOSUPPORT);
2684 	}
2685 
2686 	ia->s_addr = pr->pr_ip4[0].s_addr;
2687 	mtx_unlock(&pr->pr_mtx);
2688 	return (0);
2689 }
2690 
2691 /*
2692  * Return 1 if we should do proper source address selection or are not jailed.
2693  * We will return 0 if we should bypass source address selection in favour
2694  * of the primary jail IPv4 address. Only in this case *ia will be updated and
2695  * returned in NBO.
2696  * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2697  */
2698 int
2699 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2700 {
2701 	struct prison *pr;
2702 	struct in_addr lia;
2703 	int error;
2704 
2705 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2706 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2707 
2708 	if (!jailed(cred))
2709 		return (1);
2710 
2711 	pr = cred->cr_prison;
2712 	if (pr->pr_flags & PR_IP4_SADDRSEL)
2713 		return (1);
2714 
2715 	lia.s_addr = INADDR_ANY;
2716 	error = prison_get_ip4(cred, &lia);
2717 	if (error)
2718 		return (error);
2719 	if (lia.s_addr == INADDR_ANY)
2720 		return (1);
2721 
2722 	ia->s_addr = lia.s_addr;
2723 	return (0);
2724 }
2725 
2726 /*
2727  * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2728  */
2729 int
2730 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2731 {
2732 
2733 	if (pr1 == pr2)
2734 		return (1);
2735 
2736 	/*
2737 	 * No need to lock since the PR_IP4_USER flag can't be altered for
2738 	 * existing prisons.
2739 	 */
2740 	while (pr1 != &prison0 &&
2741 #ifdef VIMAGE
2742 	       !(pr1->pr_flags & PR_VNET) &&
2743 #endif
2744 	       !(pr1->pr_flags & PR_IP4_USER))
2745 		pr1 = pr1->pr_parent;
2746 	while (pr2 != &prison0 &&
2747 #ifdef VIMAGE
2748 	       !(pr2->pr_flags & PR_VNET) &&
2749 #endif
2750 	       !(pr2->pr_flags & PR_IP4_USER))
2751 		pr2 = pr2->pr_parent;
2752 	return (pr1 == pr2);
2753 }
2754 
2755 /*
2756  * Make sure our (source) address is set to something meaningful to this
2757  * jail.
2758  *
2759  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2760  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2761  * doesn't allow IPv4.  Address passed in in NBO and returned in NBO.
2762  */
2763 int
2764 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2765 {
2766 	struct prison *pr;
2767 	struct in_addr ia0;
2768 	int error;
2769 
2770 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2771 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2772 
2773 	pr = cred->cr_prison;
2774 	if (!(pr->pr_flags & PR_IP4))
2775 		return (0);
2776 	mtx_lock(&pr->pr_mtx);
2777 	if (!(pr->pr_flags & PR_IP4)) {
2778 		mtx_unlock(&pr->pr_mtx);
2779 		return (0);
2780 	}
2781 	if (pr->pr_ip4 == NULL) {
2782 		mtx_unlock(&pr->pr_mtx);
2783 		return (EAFNOSUPPORT);
2784 	}
2785 
2786 	ia0.s_addr = ntohl(ia->s_addr);
2787 	if (ia0.s_addr == INADDR_LOOPBACK) {
2788 		ia->s_addr = pr->pr_ip4[0].s_addr;
2789 		mtx_unlock(&pr->pr_mtx);
2790 		return (0);
2791 	}
2792 
2793 	if (ia0.s_addr == INADDR_ANY) {
2794 		/*
2795 		 * In case there is only 1 IPv4 address, bind directly.
2796 		 */
2797 		if (pr->pr_ip4s == 1)
2798 			ia->s_addr = pr->pr_ip4[0].s_addr;
2799 		mtx_unlock(&pr->pr_mtx);
2800 		return (0);
2801 	}
2802 
2803 	error = _prison_check_ip4(pr, ia);
2804 	mtx_unlock(&pr->pr_mtx);
2805 	return (error);
2806 }
2807 
2808 /*
2809  * Rewrite destination address in case we will connect to loopback address.
2810  *
2811  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2812  * Address passed in in NBO and returned in NBO.
2813  */
2814 int
2815 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2816 {
2817 	struct prison *pr;
2818 
2819 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2820 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2821 
2822 	pr = cred->cr_prison;
2823 	if (!(pr->pr_flags & PR_IP4))
2824 		return (0);
2825 	mtx_lock(&pr->pr_mtx);
2826 	if (!(pr->pr_flags & PR_IP4)) {
2827 		mtx_unlock(&pr->pr_mtx);
2828 		return (0);
2829 	}
2830 	if (pr->pr_ip4 == NULL) {
2831 		mtx_unlock(&pr->pr_mtx);
2832 		return (EAFNOSUPPORT);
2833 	}
2834 
2835 	if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2836 		ia->s_addr = pr->pr_ip4[0].s_addr;
2837 		mtx_unlock(&pr->pr_mtx);
2838 		return (0);
2839 	}
2840 
2841 	/*
2842 	 * Return success because nothing had to be changed.
2843 	 */
2844 	mtx_unlock(&pr->pr_mtx);
2845 	return (0);
2846 }
2847 
2848 /*
2849  * Check if given address belongs to the jail referenced by cred/prison.
2850  *
2851  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2852  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2853  * doesn't allow IPv4.  Address passed in in NBO.
2854  */
2855 static int
2856 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2857 {
2858 	int i, a, z, d;
2859 
2860 	/*
2861 	 * Check the primary IP.
2862 	 */
2863 	if (pr->pr_ip4[0].s_addr == ia->s_addr)
2864 		return (0);
2865 
2866 	/*
2867 	 * All the other IPs are sorted so we can do a binary search.
2868 	 */
2869 	a = 0;
2870 	z = pr->pr_ip4s - 2;
2871 	while (a <= z) {
2872 		i = (a + z) / 2;
2873 		d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2874 		if (d > 0)
2875 			z = i - 1;
2876 		else if (d < 0)
2877 			a = i + 1;
2878 		else
2879 			return (0);
2880 	}
2881 
2882 	return (EADDRNOTAVAIL);
2883 }
2884 
2885 int
2886 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2887 {
2888 	struct prison *pr;
2889 	int error;
2890 
2891 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2892 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2893 
2894 	pr = cred->cr_prison;
2895 	if (!(pr->pr_flags & PR_IP4))
2896 		return (0);
2897 	mtx_lock(&pr->pr_mtx);
2898 	if (!(pr->pr_flags & PR_IP4)) {
2899 		mtx_unlock(&pr->pr_mtx);
2900 		return (0);
2901 	}
2902 	if (pr->pr_ip4 == NULL) {
2903 		mtx_unlock(&pr->pr_mtx);
2904 		return (EAFNOSUPPORT);
2905 	}
2906 
2907 	error = _prison_check_ip4(pr, ia);
2908 	mtx_unlock(&pr->pr_mtx);
2909 	return (error);
2910 }
2911 #endif
2912 
2913 #ifdef INET6
2914 static int
2915 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2916 {
2917 	int ii, ij, used;
2918 	struct prison *ppr;
2919 
2920 	ppr = pr->pr_parent;
2921 	if (!(pr->pr_flags & PR_IP6_USER)) {
2922 		/* This has no user settings, so just copy the parent's list. */
2923 		if (pr->pr_ip6s < ppr->pr_ip6s) {
2924 			/*
2925 			 * There's no room for the parent's list.  Use the
2926 			 * new list buffer, which is assumed to be big enough
2927 			 * (if it was passed).  If there's no buffer, try to
2928 			 * allocate one.
2929 			 */
2930 			used = 1;
2931 			if (newip6 == NULL) {
2932 				newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
2933 				    M_PRISON, M_NOWAIT);
2934 				if (newip6 != NULL)
2935 					used = 0;
2936 			}
2937 			if (newip6 != NULL) {
2938 				bcopy(ppr->pr_ip6, newip6,
2939 				    ppr->pr_ip6s * sizeof(*newip6));
2940 				free(pr->pr_ip6, M_PRISON);
2941 				pr->pr_ip6 = newip6;
2942 				pr->pr_ip6s = ppr->pr_ip6s;
2943 			}
2944 			return (used);
2945 		}
2946 		pr->pr_ip6s = ppr->pr_ip6s;
2947 		if (pr->pr_ip6s > 0)
2948 			bcopy(ppr->pr_ip6, pr->pr_ip6,
2949 			    pr->pr_ip6s * sizeof(*newip6));
2950 		else if (pr->pr_ip6 != NULL) {
2951 			free(pr->pr_ip6, M_PRISON);
2952 			pr->pr_ip6 = NULL;
2953 		}
2954 	} else if (pr->pr_ip6s > 0) {
2955 		/* Remove addresses that aren't in the parent. */
2956 		for (ij = 0; ij < ppr->pr_ip6s; ij++)
2957 			if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
2958 			    &ppr->pr_ip6[ij]))
2959 				break;
2960 		if (ij < ppr->pr_ip6s)
2961 			ii = 1;
2962 		else {
2963 			bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
2964 			    --pr->pr_ip6s * sizeof(*pr->pr_ip6));
2965 			ii = 0;
2966 		}
2967 		for (ij = 1; ii < pr->pr_ip6s; ) {
2968 			if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
2969 			    &ppr->pr_ip6[0])) {
2970 				ii++;
2971 				continue;
2972 			}
2973 			switch (ij >= ppr->pr_ip4s ? -1 :
2974 				qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
2975 			case -1:
2976 				bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
2977 				    (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
2978 				break;
2979 			case 0:
2980 				ii++;
2981 				ij++;
2982 				break;
2983 			case 1:
2984 				ij++;
2985 				break;
2986 			}
2987 		}
2988 		if (pr->pr_ip6s == 0) {
2989 			pr->pr_flags |= PR_IP6_DISABLE;
2990 			free(pr->pr_ip6, M_PRISON);
2991 			pr->pr_ip6 = NULL;
2992 		}
2993 	}
2994 	return 0;
2995 }
2996 
2997 /*
2998  * Pass back primary IPv6 address for this jail.
2999  *
3000  * If not restricted return success but do not alter the address.  Caller has
3001  * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3002  *
3003  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3004  */
3005 int
3006 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3007 {
3008 	struct prison *pr;
3009 
3010 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3011 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3012 
3013 	pr = cred->cr_prison;
3014 	if (!(pr->pr_flags & PR_IP6))
3015 		return (0);
3016 	mtx_lock(&pr->pr_mtx);
3017 	if (!(pr->pr_flags & PR_IP6)) {
3018 		mtx_unlock(&pr->pr_mtx);
3019 		return (0);
3020 	}
3021 	if (pr->pr_ip6 == NULL) {
3022 		mtx_unlock(&pr->pr_mtx);
3023 		return (EAFNOSUPPORT);
3024 	}
3025 
3026 	bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3027 	mtx_unlock(&pr->pr_mtx);
3028 	return (0);
3029 }
3030 
3031 /*
3032  * Return 1 if we should do proper source address selection or are not jailed.
3033  * We will return 0 if we should bypass source address selection in favour
3034  * of the primary jail IPv6 address. Only in this case *ia will be updated and
3035  * returned in NBO.
3036  * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3037  */
3038 int
3039 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3040 {
3041 	struct prison *pr;
3042 	struct in6_addr lia6;
3043 	int error;
3044 
3045 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3046 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3047 
3048 	if (!jailed(cred))
3049 		return (1);
3050 
3051 	pr = cred->cr_prison;
3052 	if (pr->pr_flags & PR_IP6_SADDRSEL)
3053 		return (1);
3054 
3055 	lia6 = in6addr_any;
3056 	error = prison_get_ip6(cred, &lia6);
3057 	if (error)
3058 		return (error);
3059 	if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3060 		return (1);
3061 
3062 	bcopy(&lia6, ia6, sizeof(struct in6_addr));
3063 	return (0);
3064 }
3065 
3066 /*
3067  * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3068  */
3069 int
3070 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3071 {
3072 
3073 	if (pr1 == pr2)
3074 		return (1);
3075 
3076 	while (pr1 != &prison0 &&
3077 #ifdef VIMAGE
3078 	       !(pr1->pr_flags & PR_VNET) &&
3079 #endif
3080 	       !(pr1->pr_flags & PR_IP6_USER))
3081 		pr1 = pr1->pr_parent;
3082 	while (pr2 != &prison0 &&
3083 #ifdef VIMAGE
3084 	       !(pr2->pr_flags & PR_VNET) &&
3085 #endif
3086 	       !(pr2->pr_flags & PR_IP6_USER))
3087 		pr2 = pr2->pr_parent;
3088 	return (pr1 == pr2);
3089 }
3090 
3091 /*
3092  * Make sure our (source) address is set to something meaningful to this jail.
3093  *
3094  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3095  * when needed while binding.
3096  *
3097  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3098  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3099  * doesn't allow IPv6.
3100  */
3101 int
3102 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3103 {
3104 	struct prison *pr;
3105 	int error;
3106 
3107 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3108 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3109 
3110 	pr = cred->cr_prison;
3111 	if (!(pr->pr_flags & PR_IP6))
3112 		return (0);
3113 	mtx_lock(&pr->pr_mtx);
3114 	if (!(pr->pr_flags & PR_IP6)) {
3115 		mtx_unlock(&pr->pr_mtx);
3116 		return (0);
3117 	}
3118 	if (pr->pr_ip6 == NULL) {
3119 		mtx_unlock(&pr->pr_mtx);
3120 		return (EAFNOSUPPORT);
3121 	}
3122 
3123 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3124 		bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3125 		mtx_unlock(&pr->pr_mtx);
3126 		return (0);
3127 	}
3128 
3129 	if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3130 		/*
3131 		 * In case there is only 1 IPv6 address, and v6only is true,
3132 		 * then bind directly.
3133 		 */
3134 		if (v6only != 0 && pr->pr_ip6s == 1)
3135 			bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3136 		mtx_unlock(&pr->pr_mtx);
3137 		return (0);
3138 	}
3139 
3140 	error = _prison_check_ip6(pr, ia6);
3141 	mtx_unlock(&pr->pr_mtx);
3142 	return (error);
3143 }
3144 
3145 /*
3146  * Rewrite destination address in case we will connect to loopback address.
3147  *
3148  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3149  */
3150 int
3151 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3152 {
3153 	struct prison *pr;
3154 
3155 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3156 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3157 
3158 	pr = cred->cr_prison;
3159 	if (!(pr->pr_flags & PR_IP6))
3160 		return (0);
3161 	mtx_lock(&pr->pr_mtx);
3162 	if (!(pr->pr_flags & PR_IP6)) {
3163 		mtx_unlock(&pr->pr_mtx);
3164 		return (0);
3165 	}
3166 	if (pr->pr_ip6 == NULL) {
3167 		mtx_unlock(&pr->pr_mtx);
3168 		return (EAFNOSUPPORT);
3169 	}
3170 
3171 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3172 		bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3173 		mtx_unlock(&pr->pr_mtx);
3174 		return (0);
3175 	}
3176 
3177 	/*
3178 	 * Return success because nothing had to be changed.
3179 	 */
3180 	mtx_unlock(&pr->pr_mtx);
3181 	return (0);
3182 }
3183 
3184 /*
3185  * Check if given address belongs to the jail referenced by cred/prison.
3186  *
3187  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3188  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3189  * doesn't allow IPv6.
3190  */
3191 static int
3192 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3193 {
3194 	int i, a, z, d;
3195 
3196 	/*
3197 	 * Check the primary IP.
3198 	 */
3199 	if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3200 		return (0);
3201 
3202 	/*
3203 	 * All the other IPs are sorted so we can do a binary search.
3204 	 */
3205 	a = 0;
3206 	z = pr->pr_ip6s - 2;
3207 	while (a <= z) {
3208 		i = (a + z) / 2;
3209 		d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3210 		if (d > 0)
3211 			z = i - 1;
3212 		else if (d < 0)
3213 			a = i + 1;
3214 		else
3215 			return (0);
3216 	}
3217 
3218 	return (EADDRNOTAVAIL);
3219 }
3220 
3221 int
3222 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3223 {
3224 	struct prison *pr;
3225 	int error;
3226 
3227 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3228 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3229 
3230 	pr = cred->cr_prison;
3231 	if (!(pr->pr_flags & PR_IP6))
3232 		return (0);
3233 	mtx_lock(&pr->pr_mtx);
3234 	if (!(pr->pr_flags & PR_IP6)) {
3235 		mtx_unlock(&pr->pr_mtx);
3236 		return (0);
3237 	}
3238 	if (pr->pr_ip6 == NULL) {
3239 		mtx_unlock(&pr->pr_mtx);
3240 		return (EAFNOSUPPORT);
3241 	}
3242 
3243 	error = _prison_check_ip6(pr, ia6);
3244 	mtx_unlock(&pr->pr_mtx);
3245 	return (error);
3246 }
3247 #endif
3248 
3249 /*
3250  * Check if a jail supports the given address family.
3251  *
3252  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3253  * if not.
3254  */
3255 int
3256 prison_check_af(struct ucred *cred, int af)
3257 {
3258 	struct prison *pr;
3259 	int error;
3260 
3261 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3262 
3263 	pr = cred->cr_prison;
3264 #ifdef VIMAGE
3265 	/* Prisons with their own network stack are not limited. */
3266 	if (prison_owns_vnet(cred))
3267 		return (0);
3268 #endif
3269 
3270 	error = 0;
3271 	switch (af)
3272 	{
3273 #ifdef INET
3274 	case AF_INET:
3275 		if (pr->pr_flags & PR_IP4)
3276 		{
3277 			mtx_lock(&pr->pr_mtx);
3278 			if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3279 				error = EAFNOSUPPORT;
3280 			mtx_unlock(&pr->pr_mtx);
3281 		}
3282 		break;
3283 #endif
3284 #ifdef INET6
3285 	case AF_INET6:
3286 		if (pr->pr_flags & PR_IP6)
3287 		{
3288 			mtx_lock(&pr->pr_mtx);
3289 			if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3290 				error = EAFNOSUPPORT;
3291 			mtx_unlock(&pr->pr_mtx);
3292 		}
3293 		break;
3294 #endif
3295 	case AF_LOCAL:
3296 	case AF_ROUTE:
3297 		break;
3298 	default:
3299 		if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3300 			error = EAFNOSUPPORT;
3301 	}
3302 	return (error);
3303 }
3304 
3305 /*
3306  * Check if given address belongs to the jail referenced by cred (wrapper to
3307  * prison_check_ip[46]).
3308  *
3309  * Returns 0 if jail doesn't restrict the address family or if address belongs
3310  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3311  * the jail doesn't allow the address family.  IPv4 Address passed in in NBO.
3312  */
3313 int
3314 prison_if(struct ucred *cred, struct sockaddr *sa)
3315 {
3316 #ifdef INET
3317 	struct sockaddr_in *sai;
3318 #endif
3319 #ifdef INET6
3320 	struct sockaddr_in6 *sai6;
3321 #endif
3322 	int error;
3323 
3324 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3325 	KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3326 
3327 #ifdef VIMAGE
3328 	if (prison_owns_vnet(cred))
3329 		return (0);
3330 #endif
3331 
3332 	error = 0;
3333 	switch (sa->sa_family)
3334 	{
3335 #ifdef INET
3336 	case AF_INET:
3337 		sai = (struct sockaddr_in *)sa;
3338 		error = prison_check_ip4(cred, &sai->sin_addr);
3339 		break;
3340 #endif
3341 #ifdef INET6
3342 	case AF_INET6:
3343 		sai6 = (struct sockaddr_in6 *)sa;
3344 		error = prison_check_ip6(cred, &sai6->sin6_addr);
3345 		break;
3346 #endif
3347 	default:
3348 		if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3349 			error = EAFNOSUPPORT;
3350 	}
3351 	return (error);
3352 }
3353 
3354 /*
3355  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3356  */
3357 int
3358 prison_check(struct ucred *cred1, struct ucred *cred2)
3359 {
3360 
3361 	return ((cred1->cr_prison == cred2->cr_prison ||
3362 	    prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3363 }
3364 
3365 /*
3366  * Return 1 if p2 is a child of p1, otherwise 0.
3367  */
3368 int
3369 prison_ischild(struct prison *pr1, struct prison *pr2)
3370 {
3371 
3372 	for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3373 		if (pr1 == pr2)
3374 			return (1);
3375 	return (0);
3376 }
3377 
3378 /*
3379  * Return 1 if the passed credential is in a jail, otherwise 0.
3380  */
3381 int
3382 jailed(struct ucred *cred)
3383 {
3384 
3385 	return (cred->cr_prison != &prison0);
3386 }
3387 
3388 /*
3389  * Return 1 if the passed credential is in a jail and that jail does not
3390  * have its own virtual network stack, otherwise 0.
3391  */
3392 int
3393 jailed_without_vnet(struct ucred *cred)
3394 {
3395 
3396 	if (!jailed(cred))
3397 		return (0);
3398 #ifdef VIMAGE
3399 	if (prison_owns_vnet(cred))
3400 		return (0);
3401 #endif
3402 
3403 	return (1);
3404 }
3405 
3406 /*
3407  * Return the correct hostname (domainname, et al) for the passed credential.
3408  */
3409 void
3410 getcredhostname(struct ucred *cred, char *buf, size_t size)
3411 {
3412 	struct prison *pr;
3413 
3414 	/*
3415 	 * A NULL credential can be used to shortcut to the physical
3416 	 * system's hostname.
3417 	 */
3418 	pr = (cred != NULL) ? cred->cr_prison : &prison0;
3419 	mtx_lock(&pr->pr_mtx);
3420 	strlcpy(buf, pr->pr_hostname, size);
3421 	mtx_unlock(&pr->pr_mtx);
3422 }
3423 
3424 void
3425 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3426 {
3427 
3428 	mtx_lock(&cred->cr_prison->pr_mtx);
3429 	strlcpy(buf, cred->cr_prison->pr_domainname, size);
3430 	mtx_unlock(&cred->cr_prison->pr_mtx);
3431 }
3432 
3433 void
3434 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3435 {
3436 
3437 	mtx_lock(&cred->cr_prison->pr_mtx);
3438 	strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3439 	mtx_unlock(&cred->cr_prison->pr_mtx);
3440 }
3441 
3442 void
3443 getcredhostid(struct ucred *cred, unsigned long *hostid)
3444 {
3445 
3446 	mtx_lock(&cred->cr_prison->pr_mtx);
3447 	*hostid = cred->cr_prison->pr_hostid;
3448 	mtx_unlock(&cred->cr_prison->pr_mtx);
3449 }
3450 
3451 #ifdef VIMAGE
3452 /*
3453  * Determine whether the prison represented by cred owns
3454  * its vnet rather than having it inherited.
3455  *
3456  * Returns 1 in case the prison owns the vnet, 0 otherwise.
3457  */
3458 int
3459 prison_owns_vnet(struct ucred *cred)
3460 {
3461 
3462 	/*
3463 	 * vnets cannot be added/removed after jail creation,
3464 	 * so no need to lock here.
3465 	 */
3466 	return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3467 }
3468 #endif
3469 
3470 /*
3471  * Determine whether the subject represented by cred can "see"
3472  * status of a mount point.
3473  * Returns: 0 for permitted, ENOENT otherwise.
3474  * XXX: This function should be called cr_canseemount() and should be
3475  *      placed in kern_prot.c.
3476  */
3477 int
3478 prison_canseemount(struct ucred *cred, struct mount *mp)
3479 {
3480 	struct prison *pr;
3481 	struct statfs *sp;
3482 	size_t len;
3483 
3484 	pr = cred->cr_prison;
3485 	if (pr->pr_enforce_statfs == 0)
3486 		return (0);
3487 	if (pr->pr_root->v_mount == mp)
3488 		return (0);
3489 	if (pr->pr_enforce_statfs == 2)
3490 		return (ENOENT);
3491 	/*
3492 	 * If jail's chroot directory is set to "/" we should be able to see
3493 	 * all mount-points from inside a jail.
3494 	 * This is ugly check, but this is the only situation when jail's
3495 	 * directory ends with '/'.
3496 	 */
3497 	if (strcmp(pr->pr_path, "/") == 0)
3498 		return (0);
3499 	len = strlen(pr->pr_path);
3500 	sp = &mp->mnt_stat;
3501 	if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3502 		return (ENOENT);
3503 	/*
3504 	 * Be sure that we don't have situation where jail's root directory
3505 	 * is "/some/path" and mount point is "/some/pathpath".
3506 	 */
3507 	if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3508 		return (ENOENT);
3509 	return (0);
3510 }
3511 
3512 void
3513 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3514 {
3515 	char jpath[MAXPATHLEN];
3516 	struct prison *pr;
3517 	size_t len;
3518 
3519 	pr = cred->cr_prison;
3520 	if (pr->pr_enforce_statfs == 0)
3521 		return;
3522 	if (prison_canseemount(cred, mp) != 0) {
3523 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3524 		strlcpy(sp->f_mntonname, "[restricted]",
3525 		    sizeof(sp->f_mntonname));
3526 		return;
3527 	}
3528 	if (pr->pr_root->v_mount == mp) {
3529 		/*
3530 		 * Clear current buffer data, so we are sure nothing from
3531 		 * the valid path left there.
3532 		 */
3533 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3534 		*sp->f_mntonname = '/';
3535 		return;
3536 	}
3537 	/*
3538 	 * If jail's chroot directory is set to "/" we should be able to see
3539 	 * all mount-points from inside a jail.
3540 	 */
3541 	if (strcmp(pr->pr_path, "/") == 0)
3542 		return;
3543 	len = strlen(pr->pr_path);
3544 	strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3545 	/*
3546 	 * Clear current buffer data, so we are sure nothing from
3547 	 * the valid path left there.
3548 	 */
3549 	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3550 	if (*jpath == '\0') {
3551 		/* Should never happen. */
3552 		*sp->f_mntonname = '/';
3553 	} else {
3554 		strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3555 	}
3556 }
3557 
3558 /*
3559  * Check with permission for a specific privilege is granted within jail.  We
3560  * have a specific list of accepted privileges; the rest are denied.
3561  */
3562 int
3563 prison_priv_check(struct ucred *cred, int priv)
3564 {
3565 
3566 	if (!jailed(cred))
3567 		return (0);
3568 
3569 #ifdef VIMAGE
3570 	/*
3571 	 * Privileges specific to prisons with a virtual network stack.
3572 	 * There might be a duplicate entry here in case the privilege
3573 	 * is only granted conditionally in the legacy jail case.
3574 	 */
3575 	switch (priv) {
3576 #ifdef notyet
3577 		/*
3578 		 * NFS-specific privileges.
3579 		 */
3580 	case PRIV_NFS_DAEMON:
3581 	case PRIV_NFS_LOCKD:
3582 #endif
3583 		/*
3584 		 * Network stack privileges.
3585 		 */
3586 	case PRIV_NET_BRIDGE:
3587 	case PRIV_NET_GRE:
3588 	case PRIV_NET_BPF:
3589 	case PRIV_NET_RAW:		/* Dup, cond. in legacy jail case. */
3590 	case PRIV_NET_ROUTE:
3591 	case PRIV_NET_TAP:
3592 	case PRIV_NET_SETIFMTU:
3593 	case PRIV_NET_SETIFFLAGS:
3594 	case PRIV_NET_SETIFCAP:
3595 	case PRIV_NET_SETIFDESCR:
3596 	case PRIV_NET_SETIFNAME	:
3597 	case PRIV_NET_SETIFMETRIC:
3598 	case PRIV_NET_SETIFPHYS:
3599 	case PRIV_NET_SETIFMAC:
3600 	case PRIV_NET_ADDMULTI:
3601 	case PRIV_NET_DELMULTI:
3602 	case PRIV_NET_HWIOCTL:
3603 	case PRIV_NET_SETLLADDR:
3604 	case PRIV_NET_ADDIFGROUP:
3605 	case PRIV_NET_DELIFGROUP:
3606 	case PRIV_NET_IFCREATE:
3607 	case PRIV_NET_IFDESTROY:
3608 	case PRIV_NET_ADDIFADDR:
3609 	case PRIV_NET_DELIFADDR:
3610 	case PRIV_NET_LAGG:
3611 	case PRIV_NET_GIF:
3612 	case PRIV_NET_SETIFVNET:
3613 
3614 		/*
3615 		 * 802.11-related privileges.
3616 		 */
3617 	case PRIV_NET80211_GETKEY:
3618 #ifdef notyet
3619 	case PRIV_NET80211_MANAGE:		/* XXX-BZ discuss with sam@ */
3620 #endif
3621 
3622 #ifdef notyet
3623 		/*
3624 		 * AppleTalk privileges.
3625 		 */
3626 	case PRIV_NETATALK_RESERVEDPORT:
3627 
3628 		/*
3629 		 * ATM privileges.
3630 		 */
3631 	case PRIV_NETATM_CFG:
3632 	case PRIV_NETATM_ADD:
3633 	case PRIV_NETATM_DEL:
3634 	case PRIV_NETATM_SET:
3635 
3636 		/*
3637 		 * Bluetooth privileges.
3638 		 */
3639 	case PRIV_NETBLUETOOTH_RAW:
3640 #endif
3641 
3642 		/*
3643 		 * Netgraph and netgraph module privileges.
3644 		 */
3645 	case PRIV_NETGRAPH_CONTROL:
3646 #ifdef notyet
3647 	case PRIV_NETGRAPH_TTY:
3648 #endif
3649 
3650 		/*
3651 		 * IPv4 and IPv6 privileges.
3652 		 */
3653 	case PRIV_NETINET_IPFW:
3654 	case PRIV_NETINET_DIVERT:
3655 	case PRIV_NETINET_PF:
3656 	case PRIV_NETINET_DUMMYNET:
3657 	case PRIV_NETINET_CARP:
3658 	case PRIV_NETINET_MROUTE:
3659 	case PRIV_NETINET_RAW:
3660 	case PRIV_NETINET_ADDRCTRL6:
3661 	case PRIV_NETINET_ND6:
3662 	case PRIV_NETINET_SCOPE6:
3663 	case PRIV_NETINET_ALIFETIME6:
3664 	case PRIV_NETINET_IPSEC:
3665 	case PRIV_NETINET_BINDANY:
3666 
3667 #ifdef notyet
3668 		/*
3669 		 * IPX/SPX privileges.
3670 		 */
3671 	case PRIV_NETIPX_RESERVEDPORT:
3672 	case PRIV_NETIPX_RAW:
3673 
3674 		/*
3675 		 * NCP privileges.
3676 		 */
3677 	case PRIV_NETNCP:
3678 
3679 		/*
3680 		 * SMB privileges.
3681 		 */
3682 	case PRIV_NETSMB:
3683 #endif
3684 
3685 	/*
3686 	 * No default: or deny here.
3687 	 * In case of no permit fall through to next switch().
3688 	 */
3689 		if (cred->cr_prison->pr_flags & PR_VNET)
3690 			return (0);
3691 	}
3692 #endif /* VIMAGE */
3693 
3694 	switch (priv) {
3695 
3696 		/*
3697 		 * Allow ktrace privileges for root in jail.
3698 		 */
3699 	case PRIV_KTRACE:
3700 
3701 #if 0
3702 		/*
3703 		 * Allow jailed processes to configure audit identity and
3704 		 * submit audit records (login, etc).  In the future we may
3705 		 * want to further refine the relationship between audit and
3706 		 * jail.
3707 		 */
3708 	case PRIV_AUDIT_GETAUDIT:
3709 	case PRIV_AUDIT_SETAUDIT:
3710 	case PRIV_AUDIT_SUBMIT:
3711 #endif
3712 
3713 		/*
3714 		 * Allow jailed processes to manipulate process UNIX
3715 		 * credentials in any way they see fit.
3716 		 */
3717 	case PRIV_CRED_SETUID:
3718 	case PRIV_CRED_SETEUID:
3719 	case PRIV_CRED_SETGID:
3720 	case PRIV_CRED_SETEGID:
3721 	case PRIV_CRED_SETGROUPS:
3722 	case PRIV_CRED_SETREUID:
3723 	case PRIV_CRED_SETREGID:
3724 	case PRIV_CRED_SETRESUID:
3725 	case PRIV_CRED_SETRESGID:
3726 
3727 		/*
3728 		 * Jail implements visibility constraints already, so allow
3729 		 * jailed root to override uid/gid-based constraints.
3730 		 */
3731 	case PRIV_SEEOTHERGIDS:
3732 	case PRIV_SEEOTHERUIDS:
3733 
3734 		/*
3735 		 * Jail implements inter-process debugging limits already, so
3736 		 * allow jailed root various debugging privileges.
3737 		 */
3738 	case PRIV_DEBUG_DIFFCRED:
3739 	case PRIV_DEBUG_SUGID:
3740 	case PRIV_DEBUG_UNPRIV:
3741 
3742 		/*
3743 		 * Allow jail to set various resource limits and login
3744 		 * properties, and for now, exceed process resource limits.
3745 		 */
3746 	case PRIV_PROC_LIMIT:
3747 	case PRIV_PROC_SETLOGIN:
3748 	case PRIV_PROC_SETRLIMIT:
3749 
3750 		/*
3751 		 * System V and POSIX IPC privileges are granted in jail.
3752 		 */
3753 	case PRIV_IPC_READ:
3754 	case PRIV_IPC_WRITE:
3755 	case PRIV_IPC_ADMIN:
3756 	case PRIV_IPC_MSGSIZE:
3757 	case PRIV_MQ_ADMIN:
3758 
3759 		/*
3760 		 * Jail operations within a jail work on child jails.
3761 		 */
3762 	case PRIV_JAIL_ATTACH:
3763 	case PRIV_JAIL_SET:
3764 	case PRIV_JAIL_REMOVE:
3765 
3766 		/*
3767 		 * Jail implements its own inter-process limits, so allow
3768 		 * root processes in jail to change scheduling on other
3769 		 * processes in the same jail.  Likewise for signalling.
3770 		 */
3771 	case PRIV_SCHED_DIFFCRED:
3772 	case PRIV_SCHED_CPUSET:
3773 	case PRIV_SIGNAL_DIFFCRED:
3774 	case PRIV_SIGNAL_SUGID:
3775 
3776 		/*
3777 		 * Allow jailed processes to write to sysctls marked as jail
3778 		 * writable.
3779 		 */
3780 	case PRIV_SYSCTL_WRITEJAIL:
3781 
3782 		/*
3783 		 * Allow root in jail to manage a variety of quota
3784 		 * properties.  These should likely be conditional on a
3785 		 * configuration option.
3786 		 */
3787 	case PRIV_VFS_GETQUOTA:
3788 	case PRIV_VFS_SETQUOTA:
3789 
3790 		/*
3791 		 * Since Jail relies on chroot() to implement file system
3792 		 * protections, grant many VFS privileges to root in jail.
3793 		 * Be careful to exclude mount-related and NFS-related
3794 		 * privileges.
3795 		 */
3796 	case PRIV_VFS_READ:
3797 	case PRIV_VFS_WRITE:
3798 	case PRIV_VFS_ADMIN:
3799 	case PRIV_VFS_EXEC:
3800 	case PRIV_VFS_LOOKUP:
3801 	case PRIV_VFS_BLOCKRESERVE:	/* XXXRW: Slightly surprising. */
3802 	case PRIV_VFS_CHFLAGS_DEV:
3803 	case PRIV_VFS_CHOWN:
3804 	case PRIV_VFS_CHROOT:
3805 	case PRIV_VFS_RETAINSUGID:
3806 	case PRIV_VFS_FCHROOT:
3807 	case PRIV_VFS_LINK:
3808 	case PRIV_VFS_SETGID:
3809 	case PRIV_VFS_STAT:
3810 	case PRIV_VFS_STICKYFILE:
3811 		return (0);
3812 
3813 		/*
3814 		 * Depending on the global setting, allow privilege of
3815 		 * setting system flags.
3816 		 */
3817 	case PRIV_VFS_SYSFLAGS:
3818 		if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3819 			return (0);
3820 		else
3821 			return (EPERM);
3822 
3823 		/*
3824 		 * Depending on the global setting, allow privilege of
3825 		 * mounting/unmounting file systems.
3826 		 */
3827 	case PRIV_VFS_MOUNT:
3828 	case PRIV_VFS_UNMOUNT:
3829 	case PRIV_VFS_MOUNT_NONUSER:
3830 	case PRIV_VFS_MOUNT_OWNER:
3831 		if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT)
3832 			return (0);
3833 		else
3834 			return (EPERM);
3835 
3836 		/*
3837 		 * Allow jailed root to bind reserved ports and reuse in-use
3838 		 * ports.
3839 		 */
3840 	case PRIV_NETINET_RESERVEDPORT:
3841 	case PRIV_NETINET_REUSEPORT:
3842 		return (0);
3843 
3844 		/*
3845 		 * Allow jailed root to set certian IPv4/6 (option) headers.
3846 		 */
3847 	case PRIV_NETINET_SETHDROPTS:
3848 		return (0);
3849 
3850 		/*
3851 		 * Conditionally allow creating raw sockets in jail.
3852 		 */
3853 	case PRIV_NETINET_RAW:
3854 		if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3855 			return (0);
3856 		else
3857 			return (EPERM);
3858 
3859 		/*
3860 		 * Since jail implements its own visibility limits on netstat
3861 		 * sysctls, allow getcred.  This allows identd to work in
3862 		 * jail.
3863 		 */
3864 	case PRIV_NETINET_GETCRED:
3865 		return (0);
3866 
3867 	default:
3868 		/*
3869 		 * In all remaining cases, deny the privilege request.  This
3870 		 * includes almost all network privileges, many system
3871 		 * configuration privileges.
3872 		 */
3873 		return (EPERM);
3874 	}
3875 }
3876 
3877 /*
3878  * Return the part of pr2's name that is relative to pr1, or the whole name
3879  * if it does not directly follow.
3880  */
3881 
3882 char *
3883 prison_name(struct prison *pr1, struct prison *pr2)
3884 {
3885 	char *name;
3886 
3887 	/* Jails see themselves as "0" (if they see themselves at all). */
3888 	if (pr1 == pr2)
3889 		return "0";
3890 	name = pr2->pr_name;
3891 	if (prison_ischild(pr1, pr2)) {
3892 		/*
3893 		 * pr1 isn't locked (and allprison_lock may not be either)
3894 		 * so its length can't be counted on.  But the number of dots
3895 		 * can be counted on - and counted.
3896 		 */
3897 		for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3898 			name = strchr(name, '.') + 1;
3899 	}
3900 	return (name);
3901 }
3902 
3903 /*
3904  * Return the part of pr2's path that is relative to pr1, or the whole path
3905  * if it does not directly follow.
3906  */
3907 static char *
3908 prison_path(struct prison *pr1, struct prison *pr2)
3909 {
3910 	char *path1, *path2;
3911 	int len1;
3912 
3913 	path1 = pr1->pr_path;
3914 	path2 = pr2->pr_path;
3915 	if (!strcmp(path1, "/"))
3916 		return (path2);
3917 	len1 = strlen(path1);
3918 	if (strncmp(path1, path2, len1))
3919 		return (path2);
3920 	if (path2[len1] == '\0')
3921 		return "/";
3922 	if (path2[len1] == '/')
3923 		return (path2 + len1);
3924 	return (path2);
3925 }
3926 
3927 
3928 /*
3929  * Jail-related sysctls.
3930  */
3931 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3932     "Jails");
3933 
3934 static int
3935 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3936 {
3937 	struct xprison *xp;
3938 	struct prison *pr, *cpr;
3939 #ifdef INET
3940 	struct in_addr *ip4 = NULL;
3941 	int ip4s = 0;
3942 #endif
3943 #ifdef INET6
3944 	struct in6_addr *ip6 = NULL;
3945 	int ip6s = 0;
3946 #endif
3947 	int descend, error;
3948 
3949 	xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3950 	pr = req->td->td_ucred->cr_prison;
3951 	error = 0;
3952 	sx_slock(&allprison_lock);
3953 	FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3954 #if defined(INET) || defined(INET6)
3955  again:
3956 #endif
3957 		mtx_lock(&cpr->pr_mtx);
3958 #ifdef INET
3959 		if (cpr->pr_ip4s > 0) {
3960 			if (ip4s < cpr->pr_ip4s) {
3961 				ip4s = cpr->pr_ip4s;
3962 				mtx_unlock(&cpr->pr_mtx);
3963 				ip4 = realloc(ip4, ip4s *
3964 				    sizeof(struct in_addr), M_TEMP, M_WAITOK);
3965 				goto again;
3966 			}
3967 			bcopy(cpr->pr_ip4, ip4,
3968 			    cpr->pr_ip4s * sizeof(struct in_addr));
3969 		}
3970 #endif
3971 #ifdef INET6
3972 		if (cpr->pr_ip6s > 0) {
3973 			if (ip6s < cpr->pr_ip6s) {
3974 				ip6s = cpr->pr_ip6s;
3975 				mtx_unlock(&cpr->pr_mtx);
3976 				ip6 = realloc(ip6, ip6s *
3977 				    sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3978 				goto again;
3979 			}
3980 			bcopy(cpr->pr_ip6, ip6,
3981 			    cpr->pr_ip6s * sizeof(struct in6_addr));
3982 		}
3983 #endif
3984 		if (cpr->pr_ref == 0) {
3985 			mtx_unlock(&cpr->pr_mtx);
3986 			continue;
3987 		}
3988 		bzero(xp, sizeof(*xp));
3989 		xp->pr_version = XPRISON_VERSION;
3990 		xp->pr_id = cpr->pr_id;
3991 		xp->pr_state = cpr->pr_uref > 0
3992 		    ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3993 		strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3994 		strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3995 		strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3996 #ifdef INET
3997 		xp->pr_ip4s = cpr->pr_ip4s;
3998 #endif
3999 #ifdef INET6
4000 		xp->pr_ip6s = cpr->pr_ip6s;
4001 #endif
4002 		mtx_unlock(&cpr->pr_mtx);
4003 		error = SYSCTL_OUT(req, xp, sizeof(*xp));
4004 		if (error)
4005 			break;
4006 #ifdef INET
4007 		if (xp->pr_ip4s > 0) {
4008 			error = SYSCTL_OUT(req, ip4,
4009 			    xp->pr_ip4s * sizeof(struct in_addr));
4010 			if (error)
4011 				break;
4012 		}
4013 #endif
4014 #ifdef INET6
4015 		if (xp->pr_ip6s > 0) {
4016 			error = SYSCTL_OUT(req, ip6,
4017 			    xp->pr_ip6s * sizeof(struct in6_addr));
4018 			if (error)
4019 				break;
4020 		}
4021 #endif
4022 	}
4023 	sx_sunlock(&allprison_lock);
4024 	free(xp, M_TEMP);
4025 #ifdef INET
4026 	free(ip4, M_TEMP);
4027 #endif
4028 #ifdef INET6
4029 	free(ip6, M_TEMP);
4030 #endif
4031 	return (error);
4032 }
4033 
4034 SYSCTL_OID(_security_jail, OID_AUTO, list,
4035     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4036     sysctl_jail_list, "S", "List of active jails");
4037 
4038 static int
4039 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4040 {
4041 	int error, injail;
4042 
4043 	injail = jailed(req->td->td_ucred);
4044 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
4045 
4046 	return (error);
4047 }
4048 
4049 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4050     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4051     sysctl_jail_jailed, "I", "Process in jail?");
4052 
4053 #if defined(INET) || defined(INET6)
4054 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4055     &jail_max_af_ips, 0,
4056     "Number of IP addresses a jail may have at most per address family");
4057 #endif
4058 
4059 /*
4060  * Default parameters for jail(2) compatability.  For historical reasons,
4061  * the sysctl names have varying similarity to the parameter names.  Prisons
4062  * just see their own parameters, and can't change them.
4063  */
4064 static int
4065 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4066 {
4067 	struct prison *pr;
4068 	int allow, error, i;
4069 
4070 	pr = req->td->td_ucred->cr_prison;
4071 	allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4072 
4073 	/* Get the current flag value, and convert it to a boolean. */
4074 	i = (allow & arg2) ? 1 : 0;
4075 	if (arg1 != NULL)
4076 		i = !i;
4077 	error = sysctl_handle_int(oidp, &i, 0, req);
4078 	if (error || !req->newptr)
4079 		return (error);
4080 	i = i ? arg2 : 0;
4081 	if (arg1 != NULL)
4082 		i ^= arg2;
4083 	/*
4084 	 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4085 	 * for writing.
4086 	 */
4087 	mtx_lock(&prison0.pr_mtx);
4088 	jail_default_allow = (jail_default_allow & ~arg2) | i;
4089 	mtx_unlock(&prison0.pr_mtx);
4090 	return (0);
4091 }
4092 
4093 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4094     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4095     NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4096     "Processes in jail can set their hostnames");
4097 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4098     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4099     (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4100     "Processes in jail are limited to creating UNIX/IP/route sockets only");
4101 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4102     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4103     NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4104     "Processes in jail can use System V IPC primitives");
4105 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4106     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4107     NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4108     "Prison root can create raw sockets");
4109 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4110     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4111     NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4112     "Processes in jail can alter system file flags");
4113 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4114     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4115     NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4116     "Processes in jail can mount/unmount jail-friendly file systems");
4117 
4118 static int
4119 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4120 {
4121 	struct prison *pr;
4122 	int level, error;
4123 
4124 	pr = req->td->td_ucred->cr_prison;
4125 	level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4126 	error = sysctl_handle_int(oidp, &level, 0, req);
4127 	if (error || !req->newptr)
4128 		return (error);
4129 	*(int *)arg1 = level;
4130 	return (0);
4131 }
4132 
4133 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4134     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4135     &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4136     sysctl_jail_default_level, "I",
4137     "Processes in jail cannot see all mounted file systems");
4138 
4139 /*
4140  * Nodes to describe jail parameters.  Maximum length of string parameters
4141  * is returned in the string itself, and the other parameters exist merely
4142  * to make themselves and their types known.
4143  */
4144 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4145     "Jail parameters");
4146 
4147 int
4148 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4149 {
4150 	int i;
4151 	long l;
4152 	size_t s;
4153 	char numbuf[12];
4154 
4155 	switch (oidp->oid_kind & CTLTYPE)
4156 	{
4157 	case CTLTYPE_LONG:
4158 	case CTLTYPE_ULONG:
4159 		l = 0;
4160 #ifdef SCTL_MASK32
4161 		if (!(req->flags & SCTL_MASK32))
4162 #endif
4163 			return (SYSCTL_OUT(req, &l, sizeof(l)));
4164 	case CTLTYPE_INT:
4165 	case CTLTYPE_UINT:
4166 		i = 0;
4167 		return (SYSCTL_OUT(req, &i, sizeof(i)));
4168 	case CTLTYPE_STRING:
4169 		snprintf(numbuf, sizeof(numbuf), "%d", arg2);
4170 		return
4171 		    (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4172 	case CTLTYPE_STRUCT:
4173 		s = (size_t)arg2;
4174 		return (SYSCTL_OUT(req, &s, sizeof(s)));
4175 	}
4176 	return (0);
4177 }
4178 
4179 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4180 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4181 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4182 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4183 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4184     "I", "Jail secure level");
4185 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4186     "I", "Jail cannot see all mounted file systems");
4187 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4188     "B", "Jail persistence");
4189 #ifdef VIMAGE
4190 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4191     "E,jailsys", "Virtual network stack");
4192 #endif
4193 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4194     "B", "Jail is in the process of shutting down");
4195 
4196 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4197 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4198     "I", "Current number of child jails");
4199 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4200     "I", "Maximum number of child jails");
4201 
4202 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4203 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4204     "Jail hostname");
4205 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4206     "Jail NIS domainname");
4207 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4208     "Jail host UUID");
4209 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4210     "LU", "Jail host ID");
4211 
4212 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4213 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4214 
4215 #ifdef INET
4216 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4217     "Jail IPv4 address virtualization");
4218 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4219     "S,in_addr,a", "Jail IPv4 addresses");
4220 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4221     "B", "Do (not) use IPv4 source address selection rather than the "
4222     "primary jail IPv4 address.");
4223 #endif
4224 #ifdef INET6
4225 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4226     "Jail IPv6 address virtualization");
4227 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4228     "S,in6_addr,a", "Jail IPv6 addresses");
4229 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4230     "B", "Do (not) use IPv6 source address selection rather than the "
4231     "primary jail IPv6 address.");
4232 #endif
4233 
4234 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4235 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4236     "B", "Jail may set hostname");
4237 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4238     "B", "Jail may use SYSV IPC");
4239 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4240     "B", "Jail may create raw sockets");
4241 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4242     "B", "Jail may alter system file flags");
4243 SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
4244     "B", "Jail may mount/unmount jail-friendly file systems");
4245 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4246     "B", "Jail may set file quotas");
4247 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4248     "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4249 
4250 
4251 #ifdef DDB
4252 
4253 static void
4254 db_show_prison(struct prison *pr)
4255 {
4256 	int fi;
4257 #if defined(INET) || defined(INET6)
4258 	int ii;
4259 #endif
4260 	unsigned jsf;
4261 #ifdef INET6
4262 	char ip6buf[INET6_ADDRSTRLEN];
4263 #endif
4264 
4265 	db_printf("prison %p:\n", pr);
4266 	db_printf(" jid             = %d\n", pr->pr_id);
4267 	db_printf(" name            = %s\n", pr->pr_name);
4268 	db_printf(" parent          = %p\n", pr->pr_parent);
4269 	db_printf(" ref             = %d\n", pr->pr_ref);
4270 	db_printf(" uref            = %d\n", pr->pr_uref);
4271 	db_printf(" path            = %s\n", pr->pr_path);
4272 	db_printf(" cpuset          = %d\n", pr->pr_cpuset
4273 	    ? pr->pr_cpuset->cs_id : -1);
4274 #ifdef VIMAGE
4275 	db_printf(" vnet            = %p\n", pr->pr_vnet);
4276 #endif
4277 	db_printf(" root            = %p\n", pr->pr_root);
4278 	db_printf(" securelevel     = %d\n", pr->pr_securelevel);
4279 	db_printf(" children.max    = %d\n", pr->pr_childmax);
4280 	db_printf(" children.cur    = %d\n", pr->pr_childcount);
4281 	db_printf(" child           = %p\n", LIST_FIRST(&pr->pr_children));
4282 	db_printf(" sibling         = %p\n", LIST_NEXT(pr, pr_sibling));
4283 	db_printf(" flags           = 0x%x", pr->pr_flags);
4284 	for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4285 	    fi++)
4286 		if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4287 			db_printf(" %s", pr_flag_names[fi]);
4288 	for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4289 	    fi++) {
4290 		jsf = pr->pr_flags &
4291 		    (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4292 		db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4293 		    pr_flag_jailsys[fi].disable &&
4294 		      (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4295 		    : (jsf == pr_flag_jailsys[fi].new) ? "new"
4296 		    : "inherit");
4297 	}
4298 	db_printf(" allow           = 0x%x", pr->pr_allow);
4299 	for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4300 	    fi++)
4301 		if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4302 			db_printf(" %s", pr_allow_names[fi]);
4303 	db_printf("\n");
4304 	db_printf(" enforce_statfs  = %d\n", pr->pr_enforce_statfs);
4305 	db_printf(" host.hostname   = %s\n", pr->pr_hostname);
4306 	db_printf(" host.domainname = %s\n", pr->pr_domainname);
4307 	db_printf(" host.hostuuid   = %s\n", pr->pr_hostuuid);
4308 	db_printf(" host.hostid     = %lu\n", pr->pr_hostid);
4309 #ifdef INET
4310 	db_printf(" ip4s            = %d\n", pr->pr_ip4s);
4311 	for (ii = 0; ii < pr->pr_ip4s; ii++)
4312 		db_printf(" %s %s\n",
4313 		    ii == 0 ? "ip4.addr        =" : "                 ",
4314 		    inet_ntoa(pr->pr_ip4[ii]));
4315 #endif
4316 #ifdef INET6
4317 	db_printf(" ip6s            = %d\n", pr->pr_ip6s);
4318 	for (ii = 0; ii < pr->pr_ip6s; ii++)
4319 		db_printf(" %s %s\n",
4320 		    ii == 0 ? "ip6.addr        =" : "                 ",
4321 		    ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4322 #endif
4323 }
4324 
4325 DB_SHOW_COMMAND(prison, db_show_prison_command)
4326 {
4327 	struct prison *pr;
4328 
4329 	if (!have_addr) {
4330 		/*
4331 		 * Show all prisons in the list, and prison0 which is not
4332 		 * listed.
4333 		 */
4334 		db_show_prison(&prison0);
4335 		if (!db_pager_quit) {
4336 			TAILQ_FOREACH(pr, &allprison, pr_list) {
4337 				db_show_prison(pr);
4338 				if (db_pager_quit)
4339 					break;
4340 			}
4341 		}
4342 		return;
4343 	}
4344 
4345 	if (addr == 0)
4346 		pr = &prison0;
4347 	else {
4348 		/* Look for a prison with the ID and with references. */
4349 		TAILQ_FOREACH(pr, &allprison, pr_list)
4350 			if (pr->pr_id == addr && pr->pr_ref > 0)
4351 				break;
4352 		if (pr == NULL)
4353 			/* Look again, without requiring a reference. */
4354 			TAILQ_FOREACH(pr, &allprison, pr_list)
4355 				if (pr->pr_id == addr)
4356 					break;
4357 		if (pr == NULL)
4358 			/* Assume address points to a valid prison. */
4359 			pr = (struct prison *)addr;
4360 	}
4361 	db_show_prison(pr);
4362 }
4363 
4364 #endif /* DDB */
4365