xref: /freebsd/sys/kern/kern_jail.c (revision d65cd7a57bf0600b722afc770838a5d0c1c3a8e1)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1999 Poul-Henning Kamp.
5  * Copyright (c) 2008 Bjoern A. Zeeb.
6  * Copyright (c) 2009 James Gritton.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include "opt_ddb.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/kernel.h>
41 #include <sys/systm.h>
42 #include <sys/errno.h>
43 #include <sys/sysproto.h>
44 #include <sys/malloc.h>
45 #include <sys/osd.h>
46 #include <sys/priv.h>
47 #include <sys/proc.h>
48 #include <sys/taskqueue.h>
49 #include <sys/fcntl.h>
50 #include <sys/jail.h>
51 #include <sys/linker.h>
52 #include <sys/lock.h>
53 #include <sys/mutex.h>
54 #include <sys/racct.h>
55 #include <sys/rctl.h>
56 #include <sys/refcount.h>
57 #include <sys/sx.h>
58 #include <sys/sysent.h>
59 #include <sys/namei.h>
60 #include <sys/mount.h>
61 #include <sys/queue.h>
62 #include <sys/socket.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
65 #include <sys/uuid.h>
66 #include <sys/vnode.h>
67 
68 #include <net/if.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 
73 #ifdef DDB
74 #include <ddb/ddb.h>
75 #endif /* DDB */
76 
77 #include <security/mac/mac_framework.h>
78 
79 #define	DEFAULT_HOSTUUID	"00000000-0000-0000-0000-000000000000"
80 #define	PRISON0_HOSTUUID_MODULE	"hostuuid"
81 
82 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
83 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
84 
85 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
86 #ifdef INET
87 #ifdef INET6
88 #define	_PR_IP_SADDRSEL	PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
89 #else
90 #define	_PR_IP_SADDRSEL	PR_IP4_SADDRSEL
91 #endif
92 #else /* !INET */
93 #ifdef INET6
94 #define	_PR_IP_SADDRSEL	PR_IP6_SADDRSEL
95 #else
96 #define	_PR_IP_SADDRSEL	0
97 #endif
98 #endif
99 
100 /* prison0 describes what is "real" about the system. */
101 struct prison prison0 = {
102 	.pr_id		= 0,
103 	.pr_name	= "0",
104 	.pr_ref		= 1,
105 	.pr_uref	= 1,
106 	.pr_path	= "/",
107 	.pr_securelevel	= -1,
108 	.pr_devfs_rsnum = 0,
109 	.pr_childmax	= JAIL_MAX,
110 	.pr_hostuuid	= DEFAULT_HOSTUUID,
111 	.pr_children	= LIST_HEAD_INITIALIZER(prison0.pr_children),
112 #ifdef VIMAGE
113 	.pr_flags	= PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
114 #else
115 	.pr_flags	= PR_HOST|_PR_IP_SADDRSEL,
116 #endif
117 	.pr_allow	= PR_ALLOW_ALL_STATIC,
118 };
119 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
120 
121 struct bool_flags {
122 	const char	*name;
123 	const char	*noname;
124 	unsigned	 flag;
125 };
126 struct jailsys_flags {
127 	const char	*name;
128 	unsigned	 disable;
129 	unsigned	 new;
130 };
131 
132 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
133 struct	sx allprison_lock;
134 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
135 struct	prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
136 LIST_HEAD(, prison_racct) allprison_racct;
137 int	lastprid = 0;
138 
139 static int do_jail_attach(struct thread *td, struct prison *pr);
140 static void prison_complete(void *context, int pending);
141 static void prison_deref(struct prison *pr, int flags);
142 static char *prison_path(struct prison *pr1, struct prison *pr2);
143 static void prison_remove_one(struct prison *pr);
144 #ifdef RACCT
145 static void prison_racct_attach(struct prison *pr);
146 static void prison_racct_modify(struct prison *pr);
147 static void prison_racct_detach(struct prison *pr);
148 #endif
149 
150 /* Flags for prison_deref */
151 #define	PD_DEREF	0x01
152 #define	PD_DEUREF	0x02
153 #define	PD_LOCKED	0x04
154 #define	PD_LIST_SLOCKED	0x08
155 #define	PD_LIST_XLOCKED	0x10
156 
157 /*
158  * Parameter names corresponding to PR_* flag values.  Size values are for kvm
159  * as we cannot figure out the size of a sparse array, or an array without a
160  * terminating entry.
161  */
162 static struct bool_flags pr_flag_bool[] = {
163 	{"persist", "nopersist", PR_PERSIST},
164 #ifdef INET
165 	{"ip4.saddrsel", "ip4.nosaddrsel", PR_IP4_SADDRSEL},
166 #endif
167 #ifdef INET6
168 	{"ip6.saddrsel", "ip6.nosaddrsel", PR_IP6_SADDRSEL},
169 #endif
170 };
171 const size_t pr_flag_bool_size = sizeof(pr_flag_bool);
172 
173 static struct jailsys_flags pr_flag_jailsys[] = {
174 	{"host", 0, PR_HOST},
175 #ifdef VIMAGE
176 	{"vnet", 0, PR_VNET},
177 #endif
178 #ifdef INET
179 	{"ip4", PR_IP4_USER, PR_IP4_USER},
180 #endif
181 #ifdef INET6
182 	{"ip6", PR_IP6_USER, PR_IP6_USER},
183 #endif
184 };
185 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
186 
187 /* Make this array full-size so dynamic parameters can be added. */
188 static struct bool_flags pr_flag_allow[NBBY * NBPW] = {
189 	{"allow.set_hostname", "allow.noset_hostname", PR_ALLOW_SET_HOSTNAME},
190 	{"allow.sysvipc", "allow.nosysvipc", PR_ALLOW_SYSVIPC},
191 	{"allow.raw_sockets", "allow.noraw_sockets", PR_ALLOW_RAW_SOCKETS},
192 	{"allow.chflags", "allow.nochflags", PR_ALLOW_CHFLAGS},
193 	{"allow.mount", "allow.nomount", PR_ALLOW_MOUNT},
194 	{"allow.quotas", "allow.noquotas", PR_ALLOW_QUOTAS},
195 	{"allow.socket_af", "allow.nosocket_af", PR_ALLOW_SOCKET_AF},
196 	{"allow.mlock", "allow.nomlock", PR_ALLOW_MLOCK},
197 	{"allow.reserved_ports", "allow.noreserved_ports",
198 	 PR_ALLOW_RESERVED_PORTS},
199 	{"allow.read_msgbuf", "allow.noread_msgbuf", PR_ALLOW_READ_MSGBUF},
200 	{"allow.unprivileged_proc_debug", "allow.nounprivileged_proc_debug",
201 	 PR_ALLOW_UNPRIV_DEBUG},
202 };
203 const size_t pr_flag_allow_size = sizeof(pr_flag_allow);
204 
205 #define	JAIL_DEFAULT_ALLOW		(PR_ALLOW_SET_HOSTNAME | \
206 					 PR_ALLOW_RESERVED_PORTS | \
207 					 PR_ALLOW_UNPRIV_DEBUG)
208 #define	JAIL_DEFAULT_ENFORCE_STATFS	2
209 #define	JAIL_DEFAULT_DEVFS_RSNUM	0
210 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
211 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
212 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
213 #if defined(INET) || defined(INET6)
214 static unsigned jail_max_af_ips = 255;
215 #endif
216 
217 /*
218  * Initialize the parts of prison0 that can't be static-initialized with
219  * constants.  This is called from proc0_init() after creating thread0 cpuset.
220  */
221 void
222 prison0_init(void)
223 {
224 	uint8_t *file, *data;
225 	size_t size;
226 
227 	prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
228 	prison0.pr_osreldate = osreldate;
229 	strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
230 
231 	/* If we have a preloaded hostuuid, use it. */
232 	file = preload_search_by_type(PRISON0_HOSTUUID_MODULE);
233 	if (file != NULL) {
234 		data = preload_fetch_addr(file);
235 		size = preload_fetch_size(file);
236 		if (data != NULL) {
237 			/*
238 			 * The preloaded data may include trailing whitespace, almost
239 			 * certainly a newline; skip over any whitespace or
240 			 * non-printable characters to be safe.
241 			 */
242 			while (size > 0 && data[size - 1] <= 0x20) {
243 				data[size--] = '\0';
244 			}
245 			if (validate_uuid(data, size, NULL, 0) == 0) {
246 				(void)strlcpy(prison0.pr_hostuuid, data,
247 				    size + 1);
248 			} else if (bootverbose) {
249 				printf("hostuuid: preload data malformed: '%s'",
250 				    data);
251 			}
252 		}
253 	}
254 	if (bootverbose)
255 		printf("hostuuid: using %s\n", prison0.pr_hostuuid);
256 }
257 
258 /*
259  * struct jail_args {
260  *	struct jail *jail;
261  * };
262  */
263 int
264 sys_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 = htonl(j0.ip_number);	/* jail_v0 is host order */
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 + nitems(pr_flag_allow)
316 #ifdef INET
317 			    + 1
318 #endif
319 #ifdef INET6
320 			    + 1
321 #endif
322 			    )];
323 	struct uio opt;
324 	char *u_path, *u_hostname, *u_name;
325 	struct bool_flags *bf;
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;
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 (bf = pr_flag_allow;
348 		     bf < pr_flag_allow + nitems(pr_flag_allow) &&
349 			bf->flag != 0;
350 		     bf++) {
351 			optiov[opt.uio_iovcnt].iov_base = __DECONST(char *,
352 			    (jail_default_allow & bf->flag)
353 			    ? bf->name : bf->noname);
354 			optiov[opt.uio_iovcnt].iov_len =
355 			    strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
356 			opt.uio_iovcnt += 2;
357 		}
358 		optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
359 		optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
360 		opt.uio_iovcnt++;
361 		enforce_statfs = jail_default_enforce_statfs;
362 		optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
363 		optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
364 		opt.uio_iovcnt++;
365 	}
366 
367 	tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
368 #ifdef INET
369 	ip4s = (j->version == 0) ? 1 : j->ip4s;
370 	if (ip4s > jail_max_af_ips)
371 		return (EINVAL);
372 	tmplen += ip4s * sizeof(struct in_addr);
373 #else
374 	if (j->ip4s > 0)
375 		return (EINVAL);
376 #endif
377 #ifdef INET6
378 	if (j->ip6s > jail_max_af_ips)
379 		return (EINVAL);
380 	tmplen += j->ip6s * sizeof(struct in6_addr);
381 #else
382 	if (j->ip6s > 0)
383 		return (EINVAL);
384 #endif
385 	u_path = malloc(tmplen, M_TEMP, M_WAITOK);
386 	u_hostname = u_path + MAXPATHLEN;
387 	u_name = u_hostname + MAXHOSTNAMELEN;
388 #ifdef INET
389 	u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
390 #endif
391 #ifdef INET6
392 #ifdef INET
393 	u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
394 #else
395 	u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
396 #endif
397 #endif
398 	optiov[opt.uio_iovcnt].iov_base = "path";
399 	optiov[opt.uio_iovcnt].iov_len = sizeof("path");
400 	opt.uio_iovcnt++;
401 	optiov[opt.uio_iovcnt].iov_base = u_path;
402 	error = copyinstr(j->path, u_path, MAXPATHLEN,
403 	    &optiov[opt.uio_iovcnt].iov_len);
404 	if (error) {
405 		free(u_path, M_TEMP);
406 		return (error);
407 	}
408 	opt.uio_iovcnt++;
409 	optiov[opt.uio_iovcnt].iov_base = "host.hostname";
410 	optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
411 	opt.uio_iovcnt++;
412 	optiov[opt.uio_iovcnt].iov_base = u_hostname;
413 	error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
414 	    &optiov[opt.uio_iovcnt].iov_len);
415 	if (error) {
416 		free(u_path, M_TEMP);
417 		return (error);
418 	}
419 	opt.uio_iovcnt++;
420 	if (j->jailname != NULL) {
421 		optiov[opt.uio_iovcnt].iov_base = "name";
422 		optiov[opt.uio_iovcnt].iov_len = sizeof("name");
423 		opt.uio_iovcnt++;
424 		optiov[opt.uio_iovcnt].iov_base = u_name;
425 		error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
426 		    &optiov[opt.uio_iovcnt].iov_len);
427 		if (error) {
428 			free(u_path, M_TEMP);
429 			return (error);
430 		}
431 		opt.uio_iovcnt++;
432 	}
433 #ifdef INET
434 	optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
435 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
436 	opt.uio_iovcnt++;
437 	optiov[opt.uio_iovcnt].iov_base = u_ip4;
438 	optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
439 	if (j->version == 0)
440 		u_ip4->s_addr = j->ip4s;
441 	else {
442 		error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
443 		if (error) {
444 			free(u_path, M_TEMP);
445 			return (error);
446 		}
447 	}
448 	opt.uio_iovcnt++;
449 #endif
450 #ifdef INET6
451 	optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
452 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
453 	opt.uio_iovcnt++;
454 	optiov[opt.uio_iovcnt].iov_base = u_ip6;
455 	optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
456 	error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
457 	if (error) {
458 		free(u_path, M_TEMP);
459 		return (error);
460 	}
461 	opt.uio_iovcnt++;
462 #endif
463 	KASSERT(opt.uio_iovcnt <= nitems(optiov),
464 		("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
465 	error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
466 	free(u_path, M_TEMP);
467 	return (error);
468 }
469 
470 /*
471  * struct jail_set_args {
472  *	struct iovec *iovp;
473  *	unsigned int iovcnt;
474  *	int flags;
475  * };
476  */
477 int
478 sys_jail_set(struct thread *td, struct jail_set_args *uap)
479 {
480 	struct uio *auio;
481 	int error;
482 
483 	/* Check that we have an even number of iovecs. */
484 	if (uap->iovcnt & 1)
485 		return (EINVAL);
486 
487 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
488 	if (error)
489 		return (error);
490 	error = kern_jail_set(td, auio, uap->flags);
491 	free(auio, M_IOV);
492 	return (error);
493 }
494 
495 int
496 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
497 {
498 	struct nameidata nd;
499 #ifdef INET
500 	struct in_addr *ip4;
501 #endif
502 #ifdef INET6
503 	struct in6_addr *ip6;
504 #endif
505 	struct vfsopt *opt;
506 	struct vfsoptlist *opts;
507 	struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
508 	struct vnode *root;
509 	char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
510 	char *g_path, *osrelstr;
511 	struct bool_flags *bf;
512 	struct jailsys_flags *jsf;
513 #if defined(INET) || defined(INET6)
514 	struct prison *tppr;
515 	void *op;
516 #endif
517 	unsigned long hid;
518 	size_t namelen, onamelen, pnamelen;
519 	int born, created, cuflags, descend, enforce;
520 	int error, errmsg_len, errmsg_pos;
521 	int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
522 	int jid, jsys, len, level;
523 	int childmax, osreldt, rsnum, slevel;
524 #if defined(INET) || defined(INET6)
525 	int ii, ij;
526 #endif
527 #ifdef INET
528 	int ip4s, redo_ip4;
529 #endif
530 #ifdef INET6
531 	int ip6s, redo_ip6;
532 #endif
533 	uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
534 	uint64_t pr_allow_diff;
535 	unsigned tallow;
536 	char numbuf[12];
537 
538 	error = priv_check(td, PRIV_JAIL_SET);
539 	if (!error && (flags & JAIL_ATTACH))
540 		error = priv_check(td, PRIV_JAIL_ATTACH);
541 	if (error)
542 		return (error);
543 	mypr = td->td_ucred->cr_prison;
544 	if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
545 		return (EPERM);
546 	if (flags & ~JAIL_SET_MASK)
547 		return (EINVAL);
548 
549 	/*
550 	 * Check all the parameters before committing to anything.  Not all
551 	 * errors can be caught early, but we may as well try.  Also, this
552 	 * takes care of some expensive stuff (path lookup) before getting
553 	 * the allprison lock.
554 	 *
555 	 * XXX Jails are not filesystems, and jail parameters are not mount
556 	 *     options.  But it makes more sense to re-use the vfsopt code
557 	 *     than duplicate it under a different name.
558 	 */
559 	error = vfs_buildopts(optuio, &opts);
560 	if (error)
561 		return (error);
562 #ifdef INET
563 	ip4 = NULL;
564 #endif
565 #ifdef INET6
566 	ip6 = NULL;
567 #endif
568 	g_path = NULL;
569 
570 	cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
571 	if (!cuflags) {
572 		error = EINVAL;
573 		vfs_opterror(opts, "no valid operation (create or update)");
574 		goto done_errmsg;
575 	}
576 
577 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
578 	if (error == ENOENT)
579 		jid = 0;
580 	else if (error != 0)
581 		goto done_free;
582 
583 	error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
584 	if (error == ENOENT)
585 		gotslevel = 0;
586 	else if (error != 0)
587 		goto done_free;
588 	else
589 		gotslevel = 1;
590 
591 	error =
592 	    vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
593 	if (error == ENOENT)
594 		gotchildmax = 0;
595 	else if (error != 0)
596 		goto done_free;
597 	else
598 		gotchildmax = 1;
599 
600 	error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
601 	if (error == ENOENT)
602 		gotenforce = 0;
603 	else if (error != 0)
604 		goto done_free;
605 	else if (enforce < 0 || enforce > 2) {
606 		error = EINVAL;
607 		goto done_free;
608 	} else
609 		gotenforce = 1;
610 
611 	error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
612 	if (error == ENOENT)
613 		gotrsnum = 0;
614 	else if (error != 0)
615 		goto done_free;
616 	else
617 		gotrsnum = 1;
618 
619 	pr_flags = ch_flags = 0;
620 	for (bf = pr_flag_bool;
621 	     bf < pr_flag_bool + nitems(pr_flag_bool);
622 	     bf++) {
623 		vfs_flagopt(opts, bf->name, &pr_flags, bf->flag);
624 		vfs_flagopt(opts, bf->noname, &ch_flags, bf->flag);
625 	}
626 	ch_flags |= pr_flags;
627 	for (jsf = pr_flag_jailsys;
628 	     jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
629 	     jsf++) {
630 		error = vfs_copyopt(opts, jsf->name, &jsys, sizeof(jsys));
631 		if (error == ENOENT)
632 			continue;
633 		if (error != 0)
634 			goto done_free;
635 		switch (jsys) {
636 		case JAIL_SYS_DISABLE:
637 			if (!jsf->disable) {
638 				error = EINVAL;
639 				goto done_free;
640 			}
641 			pr_flags |= jsf->disable;
642 			break;
643 		case JAIL_SYS_NEW:
644 			pr_flags |= jsf->new;
645 			break;
646 		case JAIL_SYS_INHERIT:
647 			break;
648 		default:
649 			error = EINVAL;
650 			goto done_free;
651 		}
652 		ch_flags |= jsf->new | jsf->disable;
653 	}
654 	if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
655 	    && !(pr_flags & PR_PERSIST)) {
656 		error = EINVAL;
657 		vfs_opterror(opts, "new jail must persist or attach");
658 		goto done_errmsg;
659 	}
660 #ifdef VIMAGE
661 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
662 		error = EINVAL;
663 		vfs_opterror(opts, "vnet cannot be changed after creation");
664 		goto done_errmsg;
665 	}
666 #endif
667 #ifdef INET
668 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
669 		error = EINVAL;
670 		vfs_opterror(opts, "ip4 cannot be changed after creation");
671 		goto done_errmsg;
672 	}
673 #endif
674 #ifdef INET6
675 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
676 		error = EINVAL;
677 		vfs_opterror(opts, "ip6 cannot be changed after creation");
678 		goto done_errmsg;
679 	}
680 #endif
681 
682 	pr_allow = ch_allow = 0;
683 	for (bf = pr_flag_allow;
684 	     bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
685 	     bf++) {
686 		vfs_flagopt(opts, bf->name, &pr_allow, bf->flag);
687 		vfs_flagopt(opts, bf->noname, &ch_allow, bf->flag);
688 	}
689 	ch_allow |= pr_allow;
690 
691 	error = vfs_getopt(opts, "name", (void **)&name, &len);
692 	if (error == ENOENT)
693 		name = NULL;
694 	else if (error != 0)
695 		goto done_free;
696 	else {
697 		if (len == 0 || name[len - 1] != '\0') {
698 			error = EINVAL;
699 			goto done_free;
700 		}
701 		if (len > MAXHOSTNAMELEN) {
702 			error = ENAMETOOLONG;
703 			goto done_free;
704 		}
705 	}
706 
707 	error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
708 	if (error == ENOENT)
709 		host = NULL;
710 	else if (error != 0)
711 		goto done_free;
712 	else {
713 		ch_flags |= PR_HOST;
714 		pr_flags |= PR_HOST;
715 		if (len == 0 || host[len - 1] != '\0') {
716 			error = EINVAL;
717 			goto done_free;
718 		}
719 		if (len > MAXHOSTNAMELEN) {
720 			error = ENAMETOOLONG;
721 			goto done_free;
722 		}
723 	}
724 
725 	error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
726 	if (error == ENOENT)
727 		domain = NULL;
728 	else if (error != 0)
729 		goto done_free;
730 	else {
731 		ch_flags |= PR_HOST;
732 		pr_flags |= PR_HOST;
733 		if (len == 0 || domain[len - 1] != '\0') {
734 			error = EINVAL;
735 			goto done_free;
736 		}
737 		if (len > MAXHOSTNAMELEN) {
738 			error = ENAMETOOLONG;
739 			goto done_free;
740 		}
741 	}
742 
743 	error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
744 	if (error == ENOENT)
745 		uuid = NULL;
746 	else if (error != 0)
747 		goto done_free;
748 	else {
749 		ch_flags |= PR_HOST;
750 		pr_flags |= PR_HOST;
751 		if (len == 0 || uuid[len - 1] != '\0') {
752 			error = EINVAL;
753 			goto done_free;
754 		}
755 		if (len > HOSTUUIDLEN) {
756 			error = ENAMETOOLONG;
757 			goto done_free;
758 		}
759 	}
760 
761 #ifdef COMPAT_FREEBSD32
762 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
763 		uint32_t hid32;
764 
765 		error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
766 		hid = hid32;
767 	} else
768 #endif
769 		error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
770 	if (error == ENOENT)
771 		gothid = 0;
772 	else if (error != 0)
773 		goto done_free;
774 	else {
775 		gothid = 1;
776 		ch_flags |= PR_HOST;
777 		pr_flags |= PR_HOST;
778 	}
779 
780 #ifdef INET
781 	error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
782 	if (error == ENOENT)
783 		ip4s = 0;
784 	else if (error != 0)
785 		goto done_free;
786 	else if (ip4s & (sizeof(*ip4) - 1)) {
787 		error = EINVAL;
788 		goto done_free;
789 	} else {
790 		ch_flags |= PR_IP4_USER;
791 		pr_flags |= PR_IP4_USER;
792 		if (ip4s > 0) {
793 			ip4s /= sizeof(*ip4);
794 			if (ip4s > jail_max_af_ips) {
795 				error = EINVAL;
796 				vfs_opterror(opts, "too many IPv4 addresses");
797 				goto done_errmsg;
798 			}
799 			ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
800 			bcopy(op, ip4, ip4s * sizeof(*ip4));
801 			/*
802 			 * IP addresses are all sorted but ip[0] to preserve
803 			 * the primary IP address as given from userland.
804 			 * This special IP is used for unbound outgoing
805 			 * connections as well for "loopback" traffic in case
806 			 * source address selection cannot find any more fitting
807 			 * address to connect from.
808 			 */
809 			if (ip4s > 1)
810 				qsort(ip4 + 1, ip4s - 1, sizeof(*ip4),
811 				    prison_qcmp_v4);
812 			/*
813 			 * Check for duplicate addresses and do some simple
814 			 * zero and broadcast checks. If users give other bogus
815 			 * addresses it is their problem.
816 			 *
817 			 * We do not have to care about byte order for these
818 			 * checks so we will do them in NBO.
819 			 */
820 			for (ii = 0; ii < ip4s; ii++) {
821 				if (ip4[ii].s_addr == INADDR_ANY ||
822 				    ip4[ii].s_addr == INADDR_BROADCAST) {
823 					error = EINVAL;
824 					goto done_free;
825 				}
826 				if ((ii+1) < ip4s &&
827 				    (ip4[0].s_addr == ip4[ii+1].s_addr ||
828 				     ip4[ii].s_addr == ip4[ii+1].s_addr)) {
829 					error = EINVAL;
830 					goto done_free;
831 				}
832 			}
833 		}
834 	}
835 #endif
836 
837 #ifdef INET6
838 	error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
839 	if (error == ENOENT)
840 		ip6s = 0;
841 	else if (error != 0)
842 		goto done_free;
843 	else if (ip6s & (sizeof(*ip6) - 1)) {
844 		error = EINVAL;
845 		goto done_free;
846 	} else {
847 		ch_flags |= PR_IP6_USER;
848 		pr_flags |= PR_IP6_USER;
849 		if (ip6s > 0) {
850 			ip6s /= sizeof(*ip6);
851 			if (ip6s > jail_max_af_ips) {
852 				error = EINVAL;
853 				vfs_opterror(opts, "too many IPv6 addresses");
854 				goto done_errmsg;
855 			}
856 			ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
857 			bcopy(op, ip6, ip6s * sizeof(*ip6));
858 			if (ip6s > 1)
859 				qsort(ip6 + 1, ip6s - 1, sizeof(*ip6),
860 				    prison_qcmp_v6);
861 			for (ii = 0; ii < ip6s; ii++) {
862 				if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
863 					error = EINVAL;
864 					goto done_free;
865 				}
866 				if ((ii+1) < ip6s &&
867 				    (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
868 				     IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
869 				{
870 					error = EINVAL;
871 					goto done_free;
872 				}
873 			}
874 		}
875 	}
876 #endif
877 
878 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
879 	if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
880 		error = EINVAL;
881 		vfs_opterror(opts,
882 		    "vnet jails cannot have IP address restrictions");
883 		goto done_errmsg;
884 	}
885 #endif
886 
887 	error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
888 	if (error == ENOENT)
889 		osrelstr = NULL;
890 	else if (error != 0)
891 		goto done_free;
892 	else {
893 		if (flags & JAIL_UPDATE) {
894 			error = EINVAL;
895 			vfs_opterror(opts,
896 			    "osrelease cannot be changed after creation");
897 			goto done_errmsg;
898 		}
899 		if (len == 0 || osrelstr[len - 1] != '\0') {
900 			error = EINVAL;
901 			goto done_free;
902 		}
903 		if (len >= OSRELEASELEN) {
904 			error = ENAMETOOLONG;
905 			vfs_opterror(opts,
906 			    "osrelease string must be 1-%d bytes long",
907 			    OSRELEASELEN - 1);
908 			goto done_errmsg;
909 		}
910 	}
911 
912 	error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
913 	if (error == ENOENT)
914 		osreldt = 0;
915 	else if (error != 0)
916 		goto done_free;
917 	else {
918 		if (flags & JAIL_UPDATE) {
919 			error = EINVAL;
920 			vfs_opterror(opts,
921 			    "osreldate cannot be changed after creation");
922 			goto done_errmsg;
923 		}
924 		if (osreldt == 0) {
925 			error = EINVAL;
926 			vfs_opterror(opts, "osreldate cannot be 0");
927 			goto done_errmsg;
928 		}
929 	}
930 
931 	root = NULL;
932 	error = vfs_getopt(opts, "path", (void **)&path, &len);
933 	if (error == ENOENT)
934 		path = NULL;
935 	else if (error != 0)
936 		goto done_free;
937 	else {
938 		if (flags & JAIL_UPDATE) {
939 			error = EINVAL;
940 			vfs_opterror(opts,
941 			    "path cannot be changed after creation");
942 			goto done_errmsg;
943 		}
944 		if (len == 0 || path[len - 1] != '\0') {
945 			error = EINVAL;
946 			goto done_free;
947 		}
948 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
949 		    path, td);
950 		error = namei(&nd);
951 		if (error)
952 			goto done_free;
953 		root = nd.ni_vp;
954 		NDFREE(&nd, NDF_ONLY_PNBUF);
955 		g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
956 		strlcpy(g_path, path, MAXPATHLEN);
957 		error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
958 		if (error == 0) {
959 			path = g_path;
960 		} else {
961 			/* exit on other errors */
962 			goto done_free;
963 		}
964 		if (root->v_type != VDIR) {
965 			error = ENOTDIR;
966 			vput(root);
967 			goto done_free;
968 		}
969 		VOP_UNLOCK(root);
970 	}
971 
972 	/*
973 	 * Find the specified jail, or at least its parent.
974 	 * This abuses the file error codes ENOENT and EEXIST.
975 	 */
976 	pr = NULL;
977 	ppr = mypr;
978 	if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
979 		namelc = strrchr(name, '.');
980 		jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
981 		if (*p != '\0')
982 			jid = 0;
983 	}
984 	sx_xlock(&allprison_lock);
985 	if (jid != 0) {
986 		/*
987 		 * See if a requested jid already exists.  There is an
988 		 * information leak here if the jid exists but is not within
989 		 * the caller's jail hierarchy.  Jail creators will get EEXIST
990 		 * even though they cannot see the jail, and CREATE | UPDATE
991 		 * will return ENOENT which is not normally a valid error.
992 		 */
993 		if (jid < 0) {
994 			error = EINVAL;
995 			vfs_opterror(opts, "negative jid");
996 			goto done_unlock_list;
997 		}
998 		pr = prison_find(jid);
999 		if (pr != NULL) {
1000 			ppr = pr->pr_parent;
1001 			/* Create: jid must not exist. */
1002 			if (cuflags == JAIL_CREATE) {
1003 				mtx_unlock(&pr->pr_mtx);
1004 				error = EEXIST;
1005 				vfs_opterror(opts, "jail %d already exists",
1006 				    jid);
1007 				goto done_unlock_list;
1008 			}
1009 			if (!prison_ischild(mypr, pr)) {
1010 				mtx_unlock(&pr->pr_mtx);
1011 				pr = NULL;
1012 			} else if (pr->pr_uref == 0) {
1013 				if (!(flags & JAIL_DYING)) {
1014 					mtx_unlock(&pr->pr_mtx);
1015 					error = ENOENT;
1016 					vfs_opterror(opts, "jail %d is dying",
1017 					    jid);
1018 					goto done_unlock_list;
1019 				} else if ((flags & JAIL_ATTACH) ||
1020 				    (pr_flags & PR_PERSIST)) {
1021 					/*
1022 					 * A dying jail might be resurrected
1023 					 * (via attach or persist), but first
1024 					 * it must determine if another jail
1025 					 * has claimed its name.  Accomplish
1026 					 * this by implicitly re-setting the
1027 					 * name.
1028 					 */
1029 					if (name == NULL)
1030 						name = prison_name(mypr, pr);
1031 				}
1032 			}
1033 		}
1034 		if (pr == NULL) {
1035 			/* Update: jid must exist. */
1036 			if (cuflags == JAIL_UPDATE) {
1037 				error = ENOENT;
1038 				vfs_opterror(opts, "jail %d not found", jid);
1039 				goto done_unlock_list;
1040 			}
1041 		}
1042 	}
1043 	/*
1044 	 * If the caller provided a name, look for a jail by that name.
1045 	 * This has different semantics for creates and updates keyed by jid
1046 	 * (where the name must not already exist in a different jail),
1047 	 * and updates keyed by the name itself (where the name must exist
1048 	 * because that is the jail being updated).
1049 	 */
1050 	namelc = NULL;
1051 	if (name != NULL) {
1052 		namelc = strrchr(name, '.');
1053 		if (namelc == NULL)
1054 			namelc = name;
1055 		else {
1056 			/*
1057 			 * This is a hierarchical name.  Split it into the
1058 			 * parent and child names, and make sure the parent
1059 			 * exists or matches an already found jail.
1060 			 */
1061 			if (pr != NULL) {
1062 				if (strncmp(name, ppr->pr_name, namelc - name)
1063 				    || ppr->pr_name[namelc - name] != '\0') {
1064 					mtx_unlock(&pr->pr_mtx);
1065 					error = EINVAL;
1066 					vfs_opterror(opts,
1067 					    "cannot change jail's parent");
1068 					goto done_unlock_list;
1069 				}
1070 			} else {
1071 				*namelc = '\0';
1072 				ppr = prison_find_name(mypr, name);
1073 				if (ppr == NULL) {
1074 					error = ENOENT;
1075 					vfs_opterror(opts,
1076 					    "jail \"%s\" not found", name);
1077 					goto done_unlock_list;
1078 				}
1079 				mtx_unlock(&ppr->pr_mtx);
1080 				*namelc = '.';
1081 			}
1082 			namelc++;
1083 		}
1084 		if (namelc[0] != '\0') {
1085 			pnamelen =
1086 			    (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1087  name_again:
1088 			deadpr = NULL;
1089 			FOREACH_PRISON_CHILD(ppr, tpr) {
1090 				if (tpr != pr && tpr->pr_ref > 0 &&
1091 				    !strcmp(tpr->pr_name + pnamelen, namelc)) {
1092 					if (pr == NULL &&
1093 					    cuflags != JAIL_CREATE) {
1094 						mtx_lock(&tpr->pr_mtx);
1095 						if (tpr->pr_ref > 0) {
1096 							/*
1097 							 * Use this jail
1098 							 * for updates.
1099 							 */
1100 							if (tpr->pr_uref > 0) {
1101 								pr = tpr;
1102 								break;
1103 							}
1104 							deadpr = tpr;
1105 						}
1106 						mtx_unlock(&tpr->pr_mtx);
1107 					} else if (tpr->pr_uref > 0) {
1108 						/*
1109 						 * Create, or update(jid):
1110 						 * name must not exist in an
1111 						 * active sibling jail.
1112 						 */
1113 						error = EEXIST;
1114 						if (pr != NULL)
1115 							mtx_unlock(&pr->pr_mtx);
1116 						vfs_opterror(opts,
1117 						   "jail \"%s\" already exists",
1118 						   name);
1119 						goto done_unlock_list;
1120 					}
1121 				}
1122 			}
1123 			/* If no active jail is found, use a dying one. */
1124 			if (deadpr != NULL && pr == NULL) {
1125 				if (flags & JAIL_DYING) {
1126 					mtx_lock(&deadpr->pr_mtx);
1127 					if (deadpr->pr_ref == 0) {
1128 						mtx_unlock(&deadpr->pr_mtx);
1129 						goto name_again;
1130 					}
1131 					pr = deadpr;
1132 				} else if (cuflags == JAIL_UPDATE) {
1133 					error = ENOENT;
1134 					vfs_opterror(opts,
1135 					    "jail \"%s\" is dying", name);
1136 					goto done_unlock_list;
1137 				}
1138 			}
1139 			/* Update: name must exist if no jid. */
1140 			else if (cuflags == JAIL_UPDATE && pr == NULL) {
1141 				error = ENOENT;
1142 				vfs_opterror(opts, "jail \"%s\" not found",
1143 				    name);
1144 				goto done_unlock_list;
1145 			}
1146 		}
1147 	}
1148 	/* Update: must provide a jid or name. */
1149 	else if (cuflags == JAIL_UPDATE && pr == NULL) {
1150 		error = ENOENT;
1151 		vfs_opterror(opts, "update specified no jail");
1152 		goto done_unlock_list;
1153 	}
1154 
1155 	/* If there's no prison to update, create a new one and link it in. */
1156 	if (pr == NULL) {
1157 		for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1158 			if (tpr->pr_childcount >= tpr->pr_childmax) {
1159 				error = EPERM;
1160 				vfs_opterror(opts, "prison limit exceeded");
1161 				goto done_unlock_list;
1162 			}
1163 		created = 1;
1164 		mtx_lock(&ppr->pr_mtx);
1165 		if (ppr->pr_ref == 0) {
1166 			mtx_unlock(&ppr->pr_mtx);
1167 			error = ENOENT;
1168 			vfs_opterror(opts, "jail \"%s\" not found",
1169 			    prison_name(mypr, ppr));
1170 			goto done_unlock_list;
1171 		}
1172 		ppr->pr_ref++;
1173 		ppr->pr_uref++;
1174 		mtx_unlock(&ppr->pr_mtx);
1175 		pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1176 		if (jid == 0) {
1177 			/* Find the next free jid. */
1178 			jid = lastprid + 1;
1179  findnext:
1180 			if (jid == JAIL_MAX)
1181 				jid = 1;
1182 			TAILQ_FOREACH(tpr, &allprison, pr_list) {
1183 				if (tpr->pr_id < jid)
1184 					continue;
1185 				if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1186 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1187 					break;
1188 				}
1189 				if (jid == lastprid) {
1190 					error = EAGAIN;
1191 					vfs_opterror(opts,
1192 					    "no available jail IDs");
1193 					free(pr, M_PRISON);
1194 					prison_deref(ppr, PD_DEREF |
1195 					    PD_DEUREF | PD_LIST_XLOCKED);
1196 					goto done_releroot;
1197 				}
1198 				jid++;
1199 				goto findnext;
1200 			}
1201 			lastprid = jid;
1202 		} else {
1203 			/*
1204 			 * The jail already has a jid (that did not yet exist),
1205 			 * so just find where to insert it.
1206 			 */
1207 			TAILQ_FOREACH(tpr, &allprison, pr_list)
1208 				if (tpr->pr_id >= jid) {
1209 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1210 					break;
1211 				}
1212 		}
1213 		if (tpr == NULL)
1214 			TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1215 		LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1216 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1217 			tpr->pr_childcount++;
1218 
1219 		pr->pr_parent = ppr;
1220 		pr->pr_id = jid;
1221 
1222 		/* Set some default values, and inherit some from the parent. */
1223 		if (namelc == NULL)
1224 			namelc = "";
1225 		if (path == NULL) {
1226 			path = "/";
1227 			root = mypr->pr_root;
1228 			vref(root);
1229 		}
1230 		strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1231 		pr->pr_flags |= PR_HOST;
1232 #if defined(INET) || defined(INET6)
1233 #ifdef VIMAGE
1234 		if (!(pr_flags & PR_VNET))
1235 #endif
1236 		{
1237 #ifdef INET
1238 			if (!(ch_flags & PR_IP4_USER))
1239 				pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1240 			else if (!(pr_flags & PR_IP4_USER)) {
1241 				pr->pr_flags |= ppr->pr_flags & PR_IP4;
1242 				if (ppr->pr_ip4 != NULL) {
1243 					pr->pr_ip4s = ppr->pr_ip4s;
1244 					pr->pr_ip4 = malloc(pr->pr_ip4s *
1245 					    sizeof(struct in_addr), M_PRISON,
1246 					    M_WAITOK);
1247 					bcopy(ppr->pr_ip4, pr->pr_ip4,
1248 					    pr->pr_ip4s * sizeof(*pr->pr_ip4));
1249 				}
1250 			}
1251 #endif
1252 #ifdef INET6
1253 			if (!(ch_flags & PR_IP6_USER))
1254 				pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1255 			else if (!(pr_flags & PR_IP6_USER)) {
1256 				pr->pr_flags |= ppr->pr_flags & PR_IP6;
1257 				if (ppr->pr_ip6 != NULL) {
1258 					pr->pr_ip6s = ppr->pr_ip6s;
1259 					pr->pr_ip6 = malloc(pr->pr_ip6s *
1260 					    sizeof(struct in6_addr), M_PRISON,
1261 					    M_WAITOK);
1262 					bcopy(ppr->pr_ip6, pr->pr_ip6,
1263 					    pr->pr_ip6s * sizeof(*pr->pr_ip6));
1264 				}
1265 			}
1266 #endif
1267 		}
1268 #endif
1269 		/* Source address selection is always on by default. */
1270 		pr->pr_flags |= _PR_IP_SADDRSEL;
1271 
1272 		pr->pr_securelevel = ppr->pr_securelevel;
1273 		pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1274 		pr->pr_enforce_statfs = jail_default_enforce_statfs;
1275 		pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1276 
1277 		pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1278 		if (osrelstr == NULL)
1279 			strlcpy(pr->pr_osrelease, ppr->pr_osrelease,
1280 			    sizeof(pr->pr_osrelease));
1281 		else
1282 			strlcpy(pr->pr_osrelease, osrelstr,
1283 			    sizeof(pr->pr_osrelease));
1284 
1285 		LIST_INIT(&pr->pr_children);
1286 		mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1287 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1288 
1289 #ifdef VIMAGE
1290 		/* Allocate a new vnet if specified. */
1291 		pr->pr_vnet = (pr_flags & PR_VNET)
1292 		    ? vnet_alloc() : ppr->pr_vnet;
1293 #endif
1294 		/*
1295 		 * Allocate a dedicated cpuset for each jail.
1296 		 * Unlike other initial settings, this may return an erorr.
1297 		 */
1298 		error = cpuset_create_root(ppr, &pr->pr_cpuset);
1299 		if (error) {
1300 			prison_deref(pr, PD_LIST_XLOCKED);
1301 			goto done_releroot;
1302 		}
1303 
1304 		mtx_lock(&pr->pr_mtx);
1305 		/*
1306 		 * New prisons do not yet have a reference, because we do not
1307 		 * want others to see the incomplete prison once the
1308 		 * allprison_lock is downgraded.
1309 		 */
1310 	} else {
1311 		created = 0;
1312 		/*
1313 		 * Grab a reference for existing prisons, to ensure they
1314 		 * continue to exist for the duration of the call.
1315 		 */
1316 		pr->pr_ref++;
1317 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1318 		if ((pr->pr_flags & PR_VNET) &&
1319 		    (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1320 			error = EINVAL;
1321 			vfs_opterror(opts,
1322 			    "vnet jails cannot have IP address restrictions");
1323 			goto done_deref_locked;
1324 		}
1325 #endif
1326 #ifdef INET
1327 		if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1328 			error = EINVAL;
1329 			vfs_opterror(opts,
1330 			    "ip4 cannot be changed after creation");
1331 			goto done_deref_locked;
1332 		}
1333 #endif
1334 #ifdef INET6
1335 		if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1336 			error = EINVAL;
1337 			vfs_opterror(opts,
1338 			    "ip6 cannot be changed after creation");
1339 			goto done_deref_locked;
1340 		}
1341 #endif
1342 	}
1343 
1344 	/* Do final error checking before setting anything. */
1345 	if (gotslevel) {
1346 		if (slevel < ppr->pr_securelevel) {
1347 			error = EPERM;
1348 			goto done_deref_locked;
1349 		}
1350 	}
1351 	if (gotchildmax) {
1352 		if (childmax >= ppr->pr_childmax) {
1353 			error = EPERM;
1354 			goto done_deref_locked;
1355 		}
1356 	}
1357 	if (gotenforce) {
1358 		if (enforce < ppr->pr_enforce_statfs) {
1359 			error = EPERM;
1360 			goto done_deref_locked;
1361 		}
1362 	}
1363 	if (gotrsnum) {
1364 		/*
1365 		 * devfs_rsnum is a uint16_t
1366 		 */
1367 		if (rsnum < 0 || rsnum > 65535) {
1368 			error = EINVAL;
1369 			goto done_deref_locked;
1370 		}
1371 		/*
1372 		 * Nested jails always inherit parent's devfs ruleset
1373 		 */
1374 		if (jailed(td->td_ucred)) {
1375 			if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1376 				error = EPERM;
1377 				goto done_deref_locked;
1378 			} else
1379 				rsnum = ppr->pr_devfs_rsnum;
1380 		}
1381 	}
1382 #ifdef INET
1383 	if (ip4s > 0) {
1384 		if (ppr->pr_flags & PR_IP4) {
1385 			/*
1386 			 * Make sure the new set of IP addresses is a
1387 			 * subset of the parent's list.  Don't worry
1388 			 * about the parent being unlocked, as any
1389 			 * setting is done with allprison_lock held.
1390 			 */
1391 			for (ij = 0; ij < ppr->pr_ip4s; ij++)
1392 				if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1393 					break;
1394 			if (ij == ppr->pr_ip4s) {
1395 				error = EPERM;
1396 				goto done_deref_locked;
1397 			}
1398 			if (ip4s > 1) {
1399 				for (ii = ij = 1; ii < ip4s; ii++) {
1400 					if (ip4[ii].s_addr ==
1401 					    ppr->pr_ip4[0].s_addr)
1402 						continue;
1403 					for (; ij < ppr->pr_ip4s; ij++)
1404 						if (ip4[ii].s_addr ==
1405 						    ppr->pr_ip4[ij].s_addr)
1406 							break;
1407 					if (ij == ppr->pr_ip4s)
1408 						break;
1409 				}
1410 				if (ij == ppr->pr_ip4s) {
1411 					error = EPERM;
1412 					goto done_deref_locked;
1413 				}
1414 			}
1415 		}
1416 		/*
1417 		 * Check for conflicting IP addresses.  We permit them
1418 		 * if there is no more than one IP on each jail.  If
1419 		 * there is a duplicate on a jail with more than one
1420 		 * IP stop checking and return error.
1421 		 */
1422 #ifdef VIMAGE
1423 		for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1424 			if (tppr->pr_flags & PR_VNET)
1425 				break;
1426 #else
1427 		tppr = &prison0;
1428 #endif
1429 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1430 			if (tpr == pr ||
1431 #ifdef VIMAGE
1432 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1433 #endif
1434 			    tpr->pr_uref == 0) {
1435 				descend = 0;
1436 				continue;
1437 			}
1438 			if (!(tpr->pr_flags & PR_IP4_USER))
1439 				continue;
1440 			descend = 0;
1441 			if (tpr->pr_ip4 == NULL ||
1442 			    (ip4s == 1 && tpr->pr_ip4s == 1))
1443 				continue;
1444 			for (ii = 0; ii < ip4s; ii++) {
1445 				if (prison_check_ip4_locked(tpr, &ip4[ii]) ==
1446 				    0) {
1447 					error = EADDRINUSE;
1448 					vfs_opterror(opts,
1449 					    "IPv4 addresses clash");
1450 					goto done_deref_locked;
1451 				}
1452 			}
1453 		}
1454 	}
1455 #endif
1456 #ifdef INET6
1457 	if (ip6s > 0) {
1458 		if (ppr->pr_flags & PR_IP6) {
1459 			/*
1460 			 * Make sure the new set of IP addresses is a
1461 			 * subset of the parent's list.
1462 			 */
1463 			for (ij = 0; ij < ppr->pr_ip6s; ij++)
1464 				if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1465 				    &ppr->pr_ip6[ij]))
1466 					break;
1467 			if (ij == ppr->pr_ip6s) {
1468 				error = EPERM;
1469 				goto done_deref_locked;
1470 			}
1471 			if (ip6s > 1) {
1472 				for (ii = ij = 1; ii < ip6s; ii++) {
1473 					if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1474 					     &ppr->pr_ip6[0]))
1475 						continue;
1476 					for (; ij < ppr->pr_ip6s; ij++)
1477 						if (IN6_ARE_ADDR_EQUAL(
1478 						    &ip6[ii], &ppr->pr_ip6[ij]))
1479 							break;
1480 					if (ij == ppr->pr_ip6s)
1481 						break;
1482 				}
1483 				if (ij == ppr->pr_ip6s) {
1484 					error = EPERM;
1485 					goto done_deref_locked;
1486 				}
1487 			}
1488 		}
1489 		/* Check for conflicting IP addresses. */
1490 #ifdef VIMAGE
1491 		for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1492 			if (tppr->pr_flags & PR_VNET)
1493 				break;
1494 #else
1495 		tppr = &prison0;
1496 #endif
1497 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1498 			if (tpr == pr ||
1499 #ifdef VIMAGE
1500 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1501 #endif
1502 			    tpr->pr_uref == 0) {
1503 				descend = 0;
1504 				continue;
1505 			}
1506 			if (!(tpr->pr_flags & PR_IP6_USER))
1507 				continue;
1508 			descend = 0;
1509 			if (tpr->pr_ip6 == NULL ||
1510 			    (ip6s == 1 && tpr->pr_ip6s == 1))
1511 				continue;
1512 			for (ii = 0; ii < ip6s; ii++) {
1513 				if (prison_check_ip6_locked(tpr, &ip6[ii]) ==
1514 				    0) {
1515 					error = EADDRINUSE;
1516 					vfs_opterror(opts,
1517 					    "IPv6 addresses clash");
1518 					goto done_deref_locked;
1519 				}
1520 			}
1521 		}
1522 	}
1523 #endif
1524 	onamelen = namelen = 0;
1525 	if (namelc != NULL) {
1526 		/* Give a default name of the jid.  Also allow the name to be
1527 		 * explicitly the jid - but not any other number, and only in
1528 		 * normal form (no leading zero/etc).
1529 		 */
1530 		if (namelc[0] == '\0')
1531 			snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1532 		else if ((strtoul(namelc, &p, 10) != jid ||
1533 			  namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1534 			error = EINVAL;
1535 			vfs_opterror(opts,
1536 			    "name cannot be numeric (unless it is the jid)");
1537 			goto done_deref_locked;
1538 		}
1539 		/*
1540 		 * Make sure the name isn't too long for the prison or its
1541 		 * children.
1542 		 */
1543 		pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1544 		onamelen = strlen(pr->pr_name + pnamelen);
1545 		namelen = strlen(namelc);
1546 		if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1547 			error = ENAMETOOLONG;
1548 			goto done_deref_locked;
1549 		}
1550 		FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1551 			if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1552 			    sizeof(pr->pr_name)) {
1553 				error = ENAMETOOLONG;
1554 				goto done_deref_locked;
1555 			}
1556 		}
1557 	}
1558 	pr_allow_diff = pr_allow & ~ppr->pr_allow;
1559 	if (pr_allow_diff & ~PR_ALLOW_DIFFERENCES) {
1560 		error = EPERM;
1561 		goto done_deref_locked;
1562 	}
1563 
1564 	/*
1565 	 * Let modules check their parameters.  This requires unlocking and
1566 	 * then re-locking the prison, but this is still a valid state as long
1567 	 * as allprison_lock remains xlocked.
1568 	 */
1569 	mtx_unlock(&pr->pr_mtx);
1570 	error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1571 	if (error != 0) {
1572 		prison_deref(pr, created
1573 		    ? PD_LIST_XLOCKED
1574 		    : PD_DEREF | PD_LIST_XLOCKED);
1575 		goto done_releroot;
1576 	}
1577 	mtx_lock(&pr->pr_mtx);
1578 
1579 	/* At this point, all valid parameters should have been noted. */
1580 	TAILQ_FOREACH(opt, opts, link) {
1581 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
1582 			error = EINVAL;
1583 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
1584 			goto done_deref_locked;
1585 		}
1586 	}
1587 
1588 	/* Set the parameters of the prison. */
1589 #ifdef INET
1590 	redo_ip4 = 0;
1591 	if (pr_flags & PR_IP4_USER) {
1592 		pr->pr_flags |= PR_IP4;
1593 		free(pr->pr_ip4, M_PRISON);
1594 		pr->pr_ip4s = ip4s;
1595 		pr->pr_ip4 = ip4;
1596 		ip4 = NULL;
1597 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1598 #ifdef VIMAGE
1599 			if (tpr->pr_flags & PR_VNET) {
1600 				descend = 0;
1601 				continue;
1602 			}
1603 #endif
1604 			if (prison_restrict_ip4(tpr, NULL)) {
1605 				redo_ip4 = 1;
1606 				descend = 0;
1607 			}
1608 		}
1609 	}
1610 #endif
1611 #ifdef INET6
1612 	redo_ip6 = 0;
1613 	if (pr_flags & PR_IP6_USER) {
1614 		pr->pr_flags |= PR_IP6;
1615 		free(pr->pr_ip6, M_PRISON);
1616 		pr->pr_ip6s = ip6s;
1617 		pr->pr_ip6 = ip6;
1618 		ip6 = NULL;
1619 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1620 #ifdef VIMAGE
1621 			if (tpr->pr_flags & PR_VNET) {
1622 				descend = 0;
1623 				continue;
1624 			}
1625 #endif
1626 			if (prison_restrict_ip6(tpr, NULL)) {
1627 				redo_ip6 = 1;
1628 				descend = 0;
1629 			}
1630 		}
1631 	}
1632 #endif
1633 	if (gotslevel) {
1634 		pr->pr_securelevel = slevel;
1635 		/* Set all child jails to be at least this level. */
1636 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1637 			if (tpr->pr_securelevel < slevel)
1638 				tpr->pr_securelevel = slevel;
1639 	}
1640 	if (gotchildmax) {
1641 		pr->pr_childmax = childmax;
1642 		/* Set all child jails to under this limit. */
1643 		FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1644 			if (tpr->pr_childmax > childmax - level)
1645 				tpr->pr_childmax = childmax > level
1646 				    ? childmax - level : 0;
1647 	}
1648 	if (gotenforce) {
1649 		pr->pr_enforce_statfs = enforce;
1650 		/* Pass this restriction on to the children. */
1651 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1652 			if (tpr->pr_enforce_statfs < enforce)
1653 				tpr->pr_enforce_statfs = enforce;
1654 	}
1655 	if (gotrsnum) {
1656 		pr->pr_devfs_rsnum = rsnum;
1657 		/* Pass this restriction on to the children. */
1658 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1659 			tpr->pr_devfs_rsnum = rsnum;
1660 	}
1661 	if (namelc != NULL) {
1662 		if (ppr == &prison0)
1663 			strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1664 		else
1665 			snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1666 			    ppr->pr_name, namelc);
1667 		/* Change this component of child names. */
1668 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1669 			bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1670 			    strlen(tpr->pr_name + onamelen) + 1);
1671 			bcopy(pr->pr_name, tpr->pr_name, namelen);
1672 		}
1673 	}
1674 	if (path != NULL) {
1675 		/* Try to keep a real-rooted full pathname. */
1676 		strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1677 		pr->pr_root = root;
1678 	}
1679 	if (PR_HOST & ch_flags & ~pr_flags) {
1680 		if (pr->pr_flags & PR_HOST) {
1681 			/*
1682 			 * Copy the parent's host info.  As with pr_ip4 above,
1683 			 * the lack of a lock on the parent is not a problem;
1684 			 * it is always set with allprison_lock at least
1685 			 * shared, and is held exclusively here.
1686 			 */
1687 			strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1688 			    sizeof(pr->pr_hostname));
1689 			strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1690 			    sizeof(pr->pr_domainname));
1691 			strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1692 			    sizeof(pr->pr_hostuuid));
1693 			pr->pr_hostid = pr->pr_parent->pr_hostid;
1694 		}
1695 	} else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1696 		/* Set this prison, and any descendants without PR_HOST. */
1697 		if (host != NULL)
1698 			strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1699 		if (domain != NULL)
1700 			strlcpy(pr->pr_domainname, domain,
1701 			    sizeof(pr->pr_domainname));
1702 		if (uuid != NULL)
1703 			strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1704 		if (gothid)
1705 			pr->pr_hostid = hid;
1706 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1707 			if (tpr->pr_flags & PR_HOST)
1708 				descend = 0;
1709 			else {
1710 				if (host != NULL)
1711 					strlcpy(tpr->pr_hostname,
1712 					    pr->pr_hostname,
1713 					    sizeof(tpr->pr_hostname));
1714 				if (domain != NULL)
1715 					strlcpy(tpr->pr_domainname,
1716 					    pr->pr_domainname,
1717 					    sizeof(tpr->pr_domainname));
1718 				if (uuid != NULL)
1719 					strlcpy(tpr->pr_hostuuid,
1720 					    pr->pr_hostuuid,
1721 					    sizeof(tpr->pr_hostuuid));
1722 				if (gothid)
1723 					tpr->pr_hostid = hid;
1724 			}
1725 		}
1726 	}
1727 	if ((tallow = ch_allow & ~pr_allow)) {
1728 		/* Clear allow bits in all children. */
1729 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1730 			tpr->pr_allow &= ~tallow;
1731 	}
1732 	pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1733 	/*
1734 	 * Persistent prisons get an extra reference, and prisons losing their
1735 	 * persist flag lose that reference.  Only do this for existing prisons
1736 	 * for now, so new ones will remain unseen until after the module
1737 	 * handlers have completed.
1738 	 */
1739 	born = pr->pr_uref == 0;
1740 	if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1741 		if (pr_flags & PR_PERSIST) {
1742 			pr->pr_ref++;
1743 			pr->pr_uref++;
1744 		} else {
1745 			pr->pr_ref--;
1746 			pr->pr_uref--;
1747 		}
1748 	}
1749 	pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1750 	mtx_unlock(&pr->pr_mtx);
1751 
1752 #ifdef RACCT
1753 	if (racct_enable && created)
1754 		prison_racct_attach(pr);
1755 #endif
1756 
1757 	/* Locks may have prevented a complete restriction of child IP
1758 	 * addresses.  If so, allocate some more memory and try again.
1759 	 */
1760 #ifdef INET
1761 	while (redo_ip4) {
1762 		ip4s = pr->pr_ip4s;
1763 		ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1764 		mtx_lock(&pr->pr_mtx);
1765 		redo_ip4 = 0;
1766 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1767 #ifdef VIMAGE
1768 			if (tpr->pr_flags & PR_VNET) {
1769 				descend = 0;
1770 				continue;
1771 			}
1772 #endif
1773 			if (prison_restrict_ip4(tpr, ip4)) {
1774 				if (ip4 != NULL)
1775 					ip4 = NULL;
1776 				else
1777 					redo_ip4 = 1;
1778 			}
1779 		}
1780 		mtx_unlock(&pr->pr_mtx);
1781 	}
1782 #endif
1783 #ifdef INET6
1784 	while (redo_ip6) {
1785 		ip6s = pr->pr_ip6s;
1786 		ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1787 		mtx_lock(&pr->pr_mtx);
1788 		redo_ip6 = 0;
1789 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1790 #ifdef VIMAGE
1791 			if (tpr->pr_flags & PR_VNET) {
1792 				descend = 0;
1793 				continue;
1794 			}
1795 #endif
1796 			if (prison_restrict_ip6(tpr, ip6)) {
1797 				if (ip6 != NULL)
1798 					ip6 = NULL;
1799 				else
1800 					redo_ip6 = 1;
1801 			}
1802 		}
1803 		mtx_unlock(&pr->pr_mtx);
1804 	}
1805 #endif
1806 
1807 	/* Let the modules do their work. */
1808 	sx_downgrade(&allprison_lock);
1809 	if (born) {
1810 		error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1811 		if (error) {
1812 			(void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1813 			prison_deref(pr, created
1814 			    ? PD_LIST_SLOCKED
1815 			    : PD_DEREF | PD_LIST_SLOCKED);
1816 			goto done_errmsg;
1817 		}
1818 	}
1819 	error = osd_jail_call(pr, PR_METHOD_SET, opts);
1820 	if (error) {
1821 		if (born)
1822 			(void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1823 		prison_deref(pr, created
1824 		    ? PD_LIST_SLOCKED
1825 		    : PD_DEREF | PD_LIST_SLOCKED);
1826 		goto done_errmsg;
1827 	}
1828 
1829 	/* Attach this process to the prison if requested. */
1830 	if (flags & JAIL_ATTACH) {
1831 		mtx_lock(&pr->pr_mtx);
1832 		error = do_jail_attach(td, pr);
1833 		if (error) {
1834 			vfs_opterror(opts, "attach failed");
1835 			if (!created)
1836 				prison_deref(pr, PD_DEREF);
1837 			goto done_errmsg;
1838 		}
1839 	}
1840 
1841 #ifdef RACCT
1842 	if (racct_enable && !created) {
1843 		if (!(flags & JAIL_ATTACH))
1844 			sx_sunlock(&allprison_lock);
1845 		prison_racct_modify(pr);
1846 		if (!(flags & JAIL_ATTACH))
1847 			sx_slock(&allprison_lock);
1848 	}
1849 #endif
1850 
1851 	td->td_retval[0] = pr->pr_id;
1852 
1853 	/*
1854 	 * Now that it is all there, drop the temporary reference from existing
1855 	 * prisons.  Or add a reference to newly created persistent prisons
1856 	 * (which was not done earlier so that the prison would not be publicly
1857 	 * visible).
1858 	 */
1859 	if (!created) {
1860 		prison_deref(pr, (flags & JAIL_ATTACH)
1861 		    ? PD_DEREF
1862 		    : PD_DEREF | PD_LIST_SLOCKED);
1863 	} else {
1864 		if (pr_flags & PR_PERSIST) {
1865 			mtx_lock(&pr->pr_mtx);
1866 			pr->pr_ref++;
1867 			pr->pr_uref++;
1868 			mtx_unlock(&pr->pr_mtx);
1869 		}
1870 		if (!(flags & JAIL_ATTACH))
1871 			sx_sunlock(&allprison_lock);
1872 	}
1873 
1874 	goto done_free;
1875 
1876  done_deref_locked:
1877 	prison_deref(pr, created
1878 	    ? PD_LOCKED | PD_LIST_XLOCKED
1879 	    : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1880 	goto done_releroot;
1881  done_unlock_list:
1882 	sx_xunlock(&allprison_lock);
1883  done_releroot:
1884 	if (root != NULL)
1885 		vrele(root);
1886  done_errmsg:
1887 	if (error) {
1888 		if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1889 		    &errmsg_len) == 0 && errmsg_len > 0) {
1890 			errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1891 			if (optuio->uio_segflg == UIO_SYSSPACE)
1892 				bcopy(errmsg,
1893 				    optuio->uio_iov[errmsg_pos].iov_base,
1894 				    errmsg_len);
1895 			else
1896 				copyout(errmsg,
1897 				    optuio->uio_iov[errmsg_pos].iov_base,
1898 				    errmsg_len);
1899 		}
1900 	}
1901  done_free:
1902 #ifdef INET
1903 	free(ip4, M_PRISON);
1904 #endif
1905 #ifdef INET6
1906 	free(ip6, M_PRISON);
1907 #endif
1908 	if (g_path != NULL)
1909 		free(g_path, M_TEMP);
1910 	vfs_freeopts(opts);
1911 	return (error);
1912 }
1913 
1914 /*
1915  * struct jail_get_args {
1916  *	struct iovec *iovp;
1917  *	unsigned int iovcnt;
1918  *	int flags;
1919  * };
1920  */
1921 int
1922 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1923 {
1924 	struct uio *auio;
1925 	int error;
1926 
1927 	/* Check that we have an even number of iovecs. */
1928 	if (uap->iovcnt & 1)
1929 		return (EINVAL);
1930 
1931 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1932 	if (error)
1933 		return (error);
1934 	error = kern_jail_get(td, auio, uap->flags);
1935 	if (error == 0)
1936 		error = copyout(auio->uio_iov, uap->iovp,
1937 		    uap->iovcnt * sizeof (struct iovec));
1938 	free(auio, M_IOV);
1939 	return (error);
1940 }
1941 
1942 int
1943 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1944 {
1945 	struct bool_flags *bf;
1946 	struct jailsys_flags *jsf;
1947 	struct prison *pr, *mypr;
1948 	struct vfsopt *opt;
1949 	struct vfsoptlist *opts;
1950 	char *errmsg, *name;
1951 	int error, errmsg_len, errmsg_pos, i, jid, len, locked, pos;
1952 	unsigned f;
1953 
1954 	if (flags & ~JAIL_GET_MASK)
1955 		return (EINVAL);
1956 
1957 	/* Get the parameter list. */
1958 	error = vfs_buildopts(optuio, &opts);
1959 	if (error)
1960 		return (error);
1961 	errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1962 	mypr = td->td_ucred->cr_prison;
1963 
1964 	/*
1965 	 * Find the prison specified by one of: lastjid, jid, name.
1966 	 */
1967 	sx_slock(&allprison_lock);
1968 	error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1969 	if (error == 0) {
1970 		TAILQ_FOREACH(pr, &allprison, pr_list) {
1971 			if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1972 				mtx_lock(&pr->pr_mtx);
1973 				if (pr->pr_ref > 0 &&
1974 				    (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1975 					break;
1976 				mtx_unlock(&pr->pr_mtx);
1977 			}
1978 		}
1979 		if (pr != NULL)
1980 			goto found_prison;
1981 		error = ENOENT;
1982 		vfs_opterror(opts, "no jail after %d", jid);
1983 		goto done_unlock_list;
1984 	} else if (error != ENOENT)
1985 		goto done_unlock_list;
1986 
1987 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1988 	if (error == 0) {
1989 		if (jid != 0) {
1990 			pr = prison_find_child(mypr, jid);
1991 			if (pr != NULL) {
1992 				if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1993 					mtx_unlock(&pr->pr_mtx);
1994 					error = ENOENT;
1995 					vfs_opterror(opts, "jail %d is dying",
1996 					    jid);
1997 					goto done_unlock_list;
1998 				}
1999 				goto found_prison;
2000 			}
2001 			error = ENOENT;
2002 			vfs_opterror(opts, "jail %d not found", jid);
2003 			goto done_unlock_list;
2004 		}
2005 	} else if (error != ENOENT)
2006 		goto done_unlock_list;
2007 
2008 	error = vfs_getopt(opts, "name", (void **)&name, &len);
2009 	if (error == 0) {
2010 		if (len == 0 || name[len - 1] != '\0') {
2011 			error = EINVAL;
2012 			goto done_unlock_list;
2013 		}
2014 		pr = prison_find_name(mypr, name);
2015 		if (pr != NULL) {
2016 			if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2017 				mtx_unlock(&pr->pr_mtx);
2018 				error = ENOENT;
2019 				vfs_opterror(opts, "jail \"%s\" is dying",
2020 				    name);
2021 				goto done_unlock_list;
2022 			}
2023 			goto found_prison;
2024 		}
2025 		error = ENOENT;
2026 		vfs_opterror(opts, "jail \"%s\" not found", name);
2027 		goto done_unlock_list;
2028 	} else if (error != ENOENT)
2029 		goto done_unlock_list;
2030 
2031 	vfs_opterror(opts, "no jail specified");
2032 	error = ENOENT;
2033 	goto done_unlock_list;
2034 
2035  found_prison:
2036 	/* Get the parameters of the prison. */
2037 	pr->pr_ref++;
2038 	locked = PD_LOCKED;
2039 	td->td_retval[0] = pr->pr_id;
2040 	error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2041 	if (error != 0 && error != ENOENT)
2042 		goto done_deref;
2043 	i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2044 	error = vfs_setopt(opts, "parent", &i, sizeof(i));
2045 	if (error != 0 && error != ENOENT)
2046 		goto done_deref;
2047 	error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2048 	if (error != 0 && error != ENOENT)
2049 		goto done_deref;
2050 	error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2051 	    sizeof(pr->pr_cpuset->cs_id));
2052 	if (error != 0 && error != ENOENT)
2053 		goto done_deref;
2054 	error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2055 	if (error != 0 && error != ENOENT)
2056 		goto done_deref;
2057 #ifdef INET
2058 	error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2059 	    pr->pr_ip4s * sizeof(*pr->pr_ip4));
2060 	if (error != 0 && error != ENOENT)
2061 		goto done_deref;
2062 #endif
2063 #ifdef INET6
2064 	error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2065 	    pr->pr_ip6s * sizeof(*pr->pr_ip6));
2066 	if (error != 0 && error != ENOENT)
2067 		goto done_deref;
2068 #endif
2069 	error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2070 	    sizeof(pr->pr_securelevel));
2071 	if (error != 0 && error != ENOENT)
2072 		goto done_deref;
2073 	error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2074 	    sizeof(pr->pr_childcount));
2075 	if (error != 0 && error != ENOENT)
2076 		goto done_deref;
2077 	error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2078 	    sizeof(pr->pr_childmax));
2079 	if (error != 0 && error != ENOENT)
2080 		goto done_deref;
2081 	error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2082 	if (error != 0 && error != ENOENT)
2083 		goto done_deref;
2084 	error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2085 	if (error != 0 && error != ENOENT)
2086 		goto done_deref;
2087 	error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2088 	if (error != 0 && error != ENOENT)
2089 		goto done_deref;
2090 #ifdef COMPAT_FREEBSD32
2091 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2092 		uint32_t hid32 = pr->pr_hostid;
2093 
2094 		error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2095 	} else
2096 #endif
2097 	error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2098 	    sizeof(pr->pr_hostid));
2099 	if (error != 0 && error != ENOENT)
2100 		goto done_deref;
2101 	error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2102 	    sizeof(pr->pr_enforce_statfs));
2103 	if (error != 0 && error != ENOENT)
2104 		goto done_deref;
2105 	error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2106 	    sizeof(pr->pr_devfs_rsnum));
2107 	if (error != 0 && error != ENOENT)
2108 		goto done_deref;
2109 	for (bf = pr_flag_bool;
2110 	     bf < pr_flag_bool + nitems(pr_flag_bool);
2111 	     bf++) {
2112 		i = (pr->pr_flags & bf->flag) ? 1 : 0;
2113 		error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2114 		if (error != 0 && error != ENOENT)
2115 			goto done_deref;
2116 		i = !i;
2117 		error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2118 		if (error != 0 && error != ENOENT)
2119 			goto done_deref;
2120 	}
2121 	for (jsf = pr_flag_jailsys;
2122 	     jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
2123 	     jsf++) {
2124 		f = pr->pr_flags & (jsf->disable | jsf->new);
2125 		i = (f != 0 && f == jsf->disable) ? JAIL_SYS_DISABLE
2126 		    : (f == jsf->new) ? JAIL_SYS_NEW
2127 		    : JAIL_SYS_INHERIT;
2128 		error = vfs_setopt(opts, jsf->name, &i, sizeof(i));
2129 		if (error != 0 && error != ENOENT)
2130 			goto done_deref;
2131 	}
2132 	for (bf = pr_flag_allow;
2133 	     bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
2134 	     bf++) {
2135 		i = (pr->pr_allow & bf->flag) ? 1 : 0;
2136 		error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2137 		if (error != 0 && error != ENOENT)
2138 			goto done_deref;
2139 		i = !i;
2140 		error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2141 		if (error != 0 && error != ENOENT)
2142 			goto done_deref;
2143 	}
2144 	i = (pr->pr_uref == 0);
2145 	error = vfs_setopt(opts, "dying", &i, sizeof(i));
2146 	if (error != 0 && error != ENOENT)
2147 		goto done_deref;
2148 	i = !i;
2149 	error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2150 	if (error != 0 && error != ENOENT)
2151 		goto done_deref;
2152 	error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2153 	    sizeof(pr->pr_osreldate));
2154 	if (error != 0 && error != ENOENT)
2155 		goto done_deref;
2156 	error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2157 	if (error != 0 && error != ENOENT)
2158 		goto done_deref;
2159 
2160 	/* Get the module parameters. */
2161 	mtx_unlock(&pr->pr_mtx);
2162 	locked = 0;
2163 	error = osd_jail_call(pr, PR_METHOD_GET, opts);
2164 	if (error)
2165 		goto done_deref;
2166 	prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2167 
2168 	/* By now, all parameters should have been noted. */
2169 	TAILQ_FOREACH(opt, opts, link) {
2170 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
2171 			error = EINVAL;
2172 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
2173 			goto done_errmsg;
2174 		}
2175 	}
2176 
2177 	/* Write the fetched parameters back to userspace. */
2178 	error = 0;
2179 	TAILQ_FOREACH(opt, opts, link) {
2180 		if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2181 			pos = 2 * opt->pos + 1;
2182 			optuio->uio_iov[pos].iov_len = opt->len;
2183 			if (opt->value != NULL) {
2184 				if (optuio->uio_segflg == UIO_SYSSPACE) {
2185 					bcopy(opt->value,
2186 					    optuio->uio_iov[pos].iov_base,
2187 					    opt->len);
2188 				} else {
2189 					error = copyout(opt->value,
2190 					    optuio->uio_iov[pos].iov_base,
2191 					    opt->len);
2192 					if (error)
2193 						break;
2194 				}
2195 			}
2196 		}
2197 	}
2198 	goto done_errmsg;
2199 
2200  done_deref:
2201 	prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2202 	goto done_errmsg;
2203 
2204  done_unlock_list:
2205 	sx_sunlock(&allprison_lock);
2206  done_errmsg:
2207 	if (error && errmsg_pos >= 0) {
2208 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2209 		errmsg_pos = 2 * errmsg_pos + 1;
2210 		if (errmsg_len > 0) {
2211 			if (optuio->uio_segflg == UIO_SYSSPACE)
2212 				bcopy(errmsg,
2213 				    optuio->uio_iov[errmsg_pos].iov_base,
2214 				    errmsg_len);
2215 			else
2216 				copyout(errmsg,
2217 				    optuio->uio_iov[errmsg_pos].iov_base,
2218 				    errmsg_len);
2219 		}
2220 	}
2221 	vfs_freeopts(opts);
2222 	return (error);
2223 }
2224 
2225 /*
2226  * struct jail_remove_args {
2227  *	int jid;
2228  * };
2229  */
2230 int
2231 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2232 {
2233 	struct prison *pr, *cpr, *lpr, *tpr;
2234 	int descend, error;
2235 
2236 	error = priv_check(td, PRIV_JAIL_REMOVE);
2237 	if (error)
2238 		return (error);
2239 
2240 	sx_xlock(&allprison_lock);
2241 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2242 	if (pr == NULL) {
2243 		sx_xunlock(&allprison_lock);
2244 		return (EINVAL);
2245 	}
2246 
2247 	/* Remove all descendants of this prison, then remove this prison. */
2248 	pr->pr_ref++;
2249 	if (!LIST_EMPTY(&pr->pr_children)) {
2250 		mtx_unlock(&pr->pr_mtx);
2251 		lpr = NULL;
2252 		FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2253 			mtx_lock(&cpr->pr_mtx);
2254 			if (cpr->pr_ref > 0) {
2255 				tpr = cpr;
2256 				cpr->pr_ref++;
2257 			} else {
2258 				/* Already removed - do not do it again. */
2259 				tpr = NULL;
2260 			}
2261 			mtx_unlock(&cpr->pr_mtx);
2262 			if (lpr != NULL) {
2263 				mtx_lock(&lpr->pr_mtx);
2264 				prison_remove_one(lpr);
2265 				sx_xlock(&allprison_lock);
2266 			}
2267 			lpr = tpr;
2268 		}
2269 		if (lpr != NULL) {
2270 			mtx_lock(&lpr->pr_mtx);
2271 			prison_remove_one(lpr);
2272 			sx_xlock(&allprison_lock);
2273 		}
2274 		mtx_lock(&pr->pr_mtx);
2275 	}
2276 	prison_remove_one(pr);
2277 	return (0);
2278 }
2279 
2280 static void
2281 prison_remove_one(struct prison *pr)
2282 {
2283 	struct proc *p;
2284 	int deuref;
2285 
2286 	/* If the prison was persistent, it is not anymore. */
2287 	deuref = 0;
2288 	if (pr->pr_flags & PR_PERSIST) {
2289 		pr->pr_ref--;
2290 		deuref = PD_DEUREF;
2291 		pr->pr_flags &= ~PR_PERSIST;
2292 	}
2293 
2294 	/*
2295 	 * jail_remove added a reference.  If that's the only one, remove
2296 	 * the prison now.
2297 	 */
2298 	KASSERT(pr->pr_ref > 0,
2299 	    ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2300 	if (pr->pr_ref == 1) {
2301 		prison_deref(pr,
2302 		    deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2303 		return;
2304 	}
2305 
2306 	mtx_unlock(&pr->pr_mtx);
2307 	sx_xunlock(&allprison_lock);
2308 	/*
2309 	 * Kill all processes unfortunate enough to be attached to this prison.
2310 	 */
2311 	sx_slock(&allproc_lock);
2312 	FOREACH_PROC_IN_SYSTEM(p) {
2313 		PROC_LOCK(p);
2314 		if (p->p_state != PRS_NEW && p->p_ucred &&
2315 		    p->p_ucred->cr_prison == pr)
2316 			kern_psignal(p, SIGKILL);
2317 		PROC_UNLOCK(p);
2318 	}
2319 	sx_sunlock(&allproc_lock);
2320 	/* Remove the temporary reference added by jail_remove. */
2321 	prison_deref(pr, deuref | PD_DEREF);
2322 }
2323 
2324 /*
2325  * struct jail_attach_args {
2326  *	int jid;
2327  * };
2328  */
2329 int
2330 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2331 {
2332 	struct prison *pr;
2333 	int error;
2334 
2335 	error = priv_check(td, PRIV_JAIL_ATTACH);
2336 	if (error)
2337 		return (error);
2338 
2339 	/*
2340 	 * Start with exclusive hold on allprison_lock to ensure that a possible
2341 	 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2342 	 * But then immediately downgrade it since we don't need to stop
2343 	 * readers.
2344 	 */
2345 	sx_xlock(&allprison_lock);
2346 	sx_downgrade(&allprison_lock);
2347 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2348 	if (pr == NULL) {
2349 		sx_sunlock(&allprison_lock);
2350 		return (EINVAL);
2351 	}
2352 
2353 	/*
2354 	 * Do not allow a process to attach to a prison that is not
2355 	 * considered to be "alive".
2356 	 */
2357 	if (pr->pr_uref == 0) {
2358 		mtx_unlock(&pr->pr_mtx);
2359 		sx_sunlock(&allprison_lock);
2360 		return (EINVAL);
2361 	}
2362 
2363 	return (do_jail_attach(td, pr));
2364 }
2365 
2366 static int
2367 do_jail_attach(struct thread *td, struct prison *pr)
2368 {
2369 	struct proc *p;
2370 	struct ucred *newcred, *oldcred;
2371 	int error;
2372 
2373 	/*
2374 	 * XXX: Note that there is a slight race here if two threads
2375 	 * in the same privileged process attempt to attach to two
2376 	 * different jails at the same time.  It is important for
2377 	 * user processes not to do this, or they might end up with
2378 	 * a process root from one prison, but attached to the jail
2379 	 * of another.
2380 	 */
2381 	pr->pr_ref++;
2382 	pr->pr_uref++;
2383 	mtx_unlock(&pr->pr_mtx);
2384 
2385 	/* Let modules do whatever they need to prepare for attaching. */
2386 	error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2387 	if (error) {
2388 		prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2389 		return (error);
2390 	}
2391 	sx_sunlock(&allprison_lock);
2392 
2393 	/*
2394 	 * Reparent the newly attached process to this jail.
2395 	 */
2396 	p = td->td_proc;
2397 	error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2398 	if (error)
2399 		goto e_revert_osd;
2400 
2401 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2402 	if ((error = change_dir(pr->pr_root, td)) != 0)
2403 		goto e_unlock;
2404 #ifdef MAC
2405 	if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2406 		goto e_unlock;
2407 #endif
2408 	VOP_UNLOCK(pr->pr_root);
2409 	if ((error = pwd_chroot(td, pr->pr_root)))
2410 		goto e_revert_osd;
2411 
2412 	newcred = crget();
2413 	PROC_LOCK(p);
2414 	oldcred = crcopysafe(p, newcred);
2415 	newcred->cr_prison = pr;
2416 	proc_set_cred(p, newcred);
2417 	setsugid(p);
2418 #ifdef RACCT
2419 	racct_proc_ucred_changed(p, oldcred, newcred);
2420 	crhold(newcred);
2421 #endif
2422 	PROC_UNLOCK(p);
2423 #ifdef RCTL
2424 	rctl_proc_ucred_changed(p, newcred);
2425 	crfree(newcred);
2426 #endif
2427 	prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2428 	crfree(oldcred);
2429 	return (0);
2430 
2431  e_unlock:
2432 	VOP_UNLOCK(pr->pr_root);
2433  e_revert_osd:
2434 	/* Tell modules this thread is still in its old jail after all. */
2435 	(void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2436 	prison_deref(pr, PD_DEREF | PD_DEUREF);
2437 	return (error);
2438 }
2439 
2440 /*
2441  * Returns a locked prison instance, or NULL on failure.
2442  */
2443 struct prison *
2444 prison_find(int prid)
2445 {
2446 	struct prison *pr;
2447 
2448 	sx_assert(&allprison_lock, SX_LOCKED);
2449 	TAILQ_FOREACH(pr, &allprison, pr_list) {
2450 		if (pr->pr_id == prid) {
2451 			mtx_lock(&pr->pr_mtx);
2452 			if (pr->pr_ref > 0)
2453 				return (pr);
2454 			mtx_unlock(&pr->pr_mtx);
2455 		}
2456 	}
2457 	return (NULL);
2458 }
2459 
2460 /*
2461  * Find a prison that is a descendant of mypr.  Returns a locked prison or NULL.
2462  */
2463 struct prison *
2464 prison_find_child(struct prison *mypr, int prid)
2465 {
2466 	struct prison *pr;
2467 	int descend;
2468 
2469 	sx_assert(&allprison_lock, SX_LOCKED);
2470 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2471 		if (pr->pr_id == prid) {
2472 			mtx_lock(&pr->pr_mtx);
2473 			if (pr->pr_ref > 0)
2474 				return (pr);
2475 			mtx_unlock(&pr->pr_mtx);
2476 		}
2477 	}
2478 	return (NULL);
2479 }
2480 
2481 /*
2482  * Look for the name relative to mypr.  Returns a locked prison or NULL.
2483  */
2484 struct prison *
2485 prison_find_name(struct prison *mypr, const char *name)
2486 {
2487 	struct prison *pr, *deadpr;
2488 	size_t mylen;
2489 	int descend;
2490 
2491 	sx_assert(&allprison_lock, SX_LOCKED);
2492 	mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2493  again:
2494 	deadpr = NULL;
2495 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2496 		if (!strcmp(pr->pr_name + mylen, name)) {
2497 			mtx_lock(&pr->pr_mtx);
2498 			if (pr->pr_ref > 0) {
2499 				if (pr->pr_uref > 0)
2500 					return (pr);
2501 				deadpr = pr;
2502 			}
2503 			mtx_unlock(&pr->pr_mtx);
2504 		}
2505 	}
2506 	/* There was no valid prison - perhaps there was a dying one. */
2507 	if (deadpr != NULL) {
2508 		mtx_lock(&deadpr->pr_mtx);
2509 		if (deadpr->pr_ref == 0) {
2510 			mtx_unlock(&deadpr->pr_mtx);
2511 			goto again;
2512 		}
2513 	}
2514 	return (deadpr);
2515 }
2516 
2517 /*
2518  * See if a prison has the specific flag set.
2519  */
2520 int
2521 prison_flag(struct ucred *cred, unsigned flag)
2522 {
2523 
2524 	/* This is an atomic read, so no locking is necessary. */
2525 	return (cred->cr_prison->pr_flags & flag);
2526 }
2527 
2528 int
2529 prison_allow(struct ucred *cred, unsigned flag)
2530 {
2531 
2532 	/* This is an atomic read, so no locking is necessary. */
2533 	return (cred->cr_prison->pr_allow & flag);
2534 }
2535 
2536 /*
2537  * Remove a prison reference.  If that was the last reference, remove the
2538  * prison itself - but not in this context in case there are locks held.
2539  */
2540 void
2541 prison_free_locked(struct prison *pr)
2542 {
2543 	int ref;
2544 
2545 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2546 	ref = --pr->pr_ref;
2547 	mtx_unlock(&pr->pr_mtx);
2548 	if (ref == 0)
2549 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2550 }
2551 
2552 void
2553 prison_free(struct prison *pr)
2554 {
2555 
2556 	mtx_lock(&pr->pr_mtx);
2557 	prison_free_locked(pr);
2558 }
2559 
2560 /*
2561  * Complete a call to either prison_free or prison_proc_free.
2562  */
2563 static void
2564 prison_complete(void *context, int pending)
2565 {
2566 	struct prison *pr = context;
2567 
2568 	sx_xlock(&allprison_lock);
2569 	mtx_lock(&pr->pr_mtx);
2570 	prison_deref(pr, pr->pr_uref
2571 	    ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2572 	    : PD_LOCKED | PD_LIST_XLOCKED);
2573 }
2574 
2575 /*
2576  * Remove a prison reference (usually).  This internal version assumes no
2577  * mutexes are held, except perhaps the prison itself.  If there are no more
2578  * references, release and delist the prison.  On completion, the prison lock
2579  * and the allprison lock are both unlocked.
2580  */
2581 static void
2582 prison_deref(struct prison *pr, int flags)
2583 {
2584 	struct prison *ppr, *tpr;
2585 	int ref, lasturef;
2586 
2587 	if (!(flags & PD_LOCKED))
2588 		mtx_lock(&pr->pr_mtx);
2589 	for (;;) {
2590 		if (flags & PD_DEUREF) {
2591 			KASSERT(pr->pr_uref > 0,
2592 			    ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2593 			     pr->pr_id));
2594 			pr->pr_uref--;
2595 			lasturef = pr->pr_uref == 0;
2596 			if (lasturef)
2597 				pr->pr_ref++;
2598 			KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2599 		} else
2600 			lasturef = 0;
2601 		if (flags & PD_DEREF) {
2602 			KASSERT(pr->pr_ref > 0,
2603 			    ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2604 			     pr->pr_id));
2605 			pr->pr_ref--;
2606 		}
2607 		ref = pr->pr_ref;
2608 		mtx_unlock(&pr->pr_mtx);
2609 
2610 		/*
2611 		 * Tell the modules if the last user reference was removed
2612 		 * (even it sticks around in dying state).
2613 		 */
2614 		if (lasturef) {
2615 			if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2616 				sx_xlock(&allprison_lock);
2617 				flags |= PD_LIST_XLOCKED;
2618 			}
2619 			(void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2620 			mtx_lock(&pr->pr_mtx);
2621 			ref = --pr->pr_ref;
2622 			mtx_unlock(&pr->pr_mtx);
2623 		}
2624 
2625 		/* If the prison still has references, nothing else to do. */
2626 		if (ref > 0) {
2627 			if (flags & PD_LIST_SLOCKED)
2628 				sx_sunlock(&allprison_lock);
2629 			else if (flags & PD_LIST_XLOCKED)
2630 				sx_xunlock(&allprison_lock);
2631 			return;
2632 		}
2633 
2634 		if (flags & PD_LIST_SLOCKED) {
2635 			if (!sx_try_upgrade(&allprison_lock)) {
2636 				sx_sunlock(&allprison_lock);
2637 				sx_xlock(&allprison_lock);
2638 			}
2639 		} else if (!(flags & PD_LIST_XLOCKED))
2640 			sx_xlock(&allprison_lock);
2641 
2642 		TAILQ_REMOVE(&allprison, pr, pr_list);
2643 		LIST_REMOVE(pr, pr_sibling);
2644 		ppr = pr->pr_parent;
2645 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2646 			tpr->pr_childcount--;
2647 		sx_xunlock(&allprison_lock);
2648 
2649 #ifdef VIMAGE
2650 		if (pr->pr_vnet != ppr->pr_vnet)
2651 			vnet_destroy(pr->pr_vnet);
2652 #endif
2653 		if (pr->pr_root != NULL)
2654 			vrele(pr->pr_root);
2655 		mtx_destroy(&pr->pr_mtx);
2656 #ifdef INET
2657 		free(pr->pr_ip4, M_PRISON);
2658 #endif
2659 #ifdef INET6
2660 		free(pr->pr_ip6, M_PRISON);
2661 #endif
2662 		if (pr->pr_cpuset != NULL)
2663 			cpuset_rel(pr->pr_cpuset);
2664 		osd_jail_exit(pr);
2665 #ifdef RACCT
2666 		if (racct_enable)
2667 			prison_racct_detach(pr);
2668 #endif
2669 		free(pr, M_PRISON);
2670 
2671 		/* Removing a prison frees a reference on its parent. */
2672 		pr = ppr;
2673 		mtx_lock(&pr->pr_mtx);
2674 		flags = PD_DEREF | PD_DEUREF;
2675 	}
2676 }
2677 
2678 void
2679 prison_hold_locked(struct prison *pr)
2680 {
2681 
2682 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2683 	KASSERT(pr->pr_ref > 0,
2684 	    ("Trying to hold dead prison %p (jid=%d).", pr, pr->pr_id));
2685 	pr->pr_ref++;
2686 }
2687 
2688 void
2689 prison_hold(struct prison *pr)
2690 {
2691 
2692 	mtx_lock(&pr->pr_mtx);
2693 	prison_hold_locked(pr);
2694 	mtx_unlock(&pr->pr_mtx);
2695 }
2696 
2697 void
2698 prison_proc_hold(struct prison *pr)
2699 {
2700 
2701 	mtx_lock(&pr->pr_mtx);
2702 	KASSERT(pr->pr_uref > 0,
2703 	    ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2704 	pr->pr_uref++;
2705 	mtx_unlock(&pr->pr_mtx);
2706 }
2707 
2708 void
2709 prison_proc_free(struct prison *pr)
2710 {
2711 
2712 	mtx_lock(&pr->pr_mtx);
2713 	KASSERT(pr->pr_uref > 0,
2714 	    ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2715 	if (pr->pr_uref > 1)
2716 		pr->pr_uref--;
2717 	else {
2718 		/*
2719 		 * Don't remove the last user reference in this context, which
2720 		 * is expected to be a process that is not only locked, but
2721 		 * also half dead.
2722 		 */
2723 		pr->pr_ref++;
2724 		mtx_unlock(&pr->pr_mtx);
2725 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2726 		return;
2727 	}
2728 	mtx_unlock(&pr->pr_mtx);
2729 }
2730 
2731 /*
2732  * Check if a jail supports the given address family.
2733  *
2734  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
2735  * if not.
2736  */
2737 int
2738 prison_check_af(struct ucred *cred, int af)
2739 {
2740 	struct prison *pr;
2741 	int error;
2742 
2743 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2744 
2745 	pr = cred->cr_prison;
2746 #ifdef VIMAGE
2747 	/* Prisons with their own network stack are not limited. */
2748 	if (prison_owns_vnet(cred))
2749 		return (0);
2750 #endif
2751 
2752 	error = 0;
2753 	switch (af)
2754 	{
2755 #ifdef INET
2756 	case AF_INET:
2757 		if (pr->pr_flags & PR_IP4)
2758 		{
2759 			mtx_lock(&pr->pr_mtx);
2760 			if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
2761 				error = EAFNOSUPPORT;
2762 			mtx_unlock(&pr->pr_mtx);
2763 		}
2764 		break;
2765 #endif
2766 #ifdef INET6
2767 	case AF_INET6:
2768 		if (pr->pr_flags & PR_IP6)
2769 		{
2770 			mtx_lock(&pr->pr_mtx);
2771 			if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
2772 				error = EAFNOSUPPORT;
2773 			mtx_unlock(&pr->pr_mtx);
2774 		}
2775 		break;
2776 #endif
2777 	case AF_LOCAL:
2778 	case AF_ROUTE:
2779 		break;
2780 	default:
2781 		if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
2782 			error = EAFNOSUPPORT;
2783 	}
2784 	return (error);
2785 }
2786 
2787 /*
2788  * Check if given address belongs to the jail referenced by cred (wrapper to
2789  * prison_check_ip[46]).
2790  *
2791  * Returns 0 if jail doesn't restrict the address family or if address belongs
2792  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
2793  * the jail doesn't allow the address family.  IPv4 Address passed in in NBO.
2794  */
2795 int
2796 prison_if(struct ucred *cred, const struct sockaddr *sa)
2797 {
2798 #ifdef INET
2799 	const struct sockaddr_in *sai;
2800 #endif
2801 #ifdef INET6
2802 	const struct sockaddr_in6 *sai6;
2803 #endif
2804 	int error;
2805 
2806 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2807 	KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
2808 
2809 #ifdef VIMAGE
2810 	if (prison_owns_vnet(cred))
2811 		return (0);
2812 #endif
2813 
2814 	error = 0;
2815 	switch (sa->sa_family)
2816 	{
2817 #ifdef INET
2818 	case AF_INET:
2819 		sai = (const struct sockaddr_in *)sa;
2820 		error = prison_check_ip4(cred, &sai->sin_addr);
2821 		break;
2822 #endif
2823 #ifdef INET6
2824 	case AF_INET6:
2825 		sai6 = (const struct sockaddr_in6 *)sa;
2826 		error = prison_check_ip6(cred, &sai6->sin6_addr);
2827 		break;
2828 #endif
2829 	default:
2830 		if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
2831 			error = EAFNOSUPPORT;
2832 	}
2833 	return (error);
2834 }
2835 
2836 /*
2837  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
2838  */
2839 int
2840 prison_check(struct ucred *cred1, struct ucred *cred2)
2841 {
2842 
2843 	return ((cred1->cr_prison == cred2->cr_prison ||
2844 	    prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
2845 }
2846 
2847 /*
2848  * Return 1 if p2 is a child of p1, otherwise 0.
2849  */
2850 int
2851 prison_ischild(struct prison *pr1, struct prison *pr2)
2852 {
2853 
2854 	for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
2855 		if (pr1 == pr2)
2856 			return (1);
2857 	return (0);
2858 }
2859 
2860 /*
2861  * Return 1 if the passed credential is in a jail and that jail does not
2862  * have its own virtual network stack, otherwise 0.
2863  */
2864 int
2865 jailed_without_vnet(struct ucred *cred)
2866 {
2867 
2868 	if (!jailed(cred))
2869 		return (0);
2870 #ifdef VIMAGE
2871 	if (prison_owns_vnet(cred))
2872 		return (0);
2873 #endif
2874 
2875 	return (1);
2876 }
2877 
2878 /*
2879  * Return the correct hostname (domainname, et al) for the passed credential.
2880  */
2881 void
2882 getcredhostname(struct ucred *cred, char *buf, size_t size)
2883 {
2884 	struct prison *pr;
2885 
2886 	/*
2887 	 * A NULL credential can be used to shortcut to the physical
2888 	 * system's hostname.
2889 	 */
2890 	pr = (cred != NULL) ? cred->cr_prison : &prison0;
2891 	mtx_lock(&pr->pr_mtx);
2892 	strlcpy(buf, pr->pr_hostname, size);
2893 	mtx_unlock(&pr->pr_mtx);
2894 }
2895 
2896 void
2897 getcreddomainname(struct ucred *cred, char *buf, size_t size)
2898 {
2899 
2900 	mtx_lock(&cred->cr_prison->pr_mtx);
2901 	strlcpy(buf, cred->cr_prison->pr_domainname, size);
2902 	mtx_unlock(&cred->cr_prison->pr_mtx);
2903 }
2904 
2905 void
2906 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
2907 {
2908 
2909 	mtx_lock(&cred->cr_prison->pr_mtx);
2910 	strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
2911 	mtx_unlock(&cred->cr_prison->pr_mtx);
2912 }
2913 
2914 void
2915 getcredhostid(struct ucred *cred, unsigned long *hostid)
2916 {
2917 
2918 	mtx_lock(&cred->cr_prison->pr_mtx);
2919 	*hostid = cred->cr_prison->pr_hostid;
2920 	mtx_unlock(&cred->cr_prison->pr_mtx);
2921 }
2922 
2923 void
2924 getjailname(struct ucred *cred, char *name, size_t len)
2925 {
2926 
2927 	mtx_lock(&cred->cr_prison->pr_mtx);
2928 	strlcpy(name, cred->cr_prison->pr_name, len);
2929 	mtx_unlock(&cred->cr_prison->pr_mtx);
2930 }
2931 
2932 #ifdef VIMAGE
2933 /*
2934  * Determine whether the prison represented by cred owns
2935  * its vnet rather than having it inherited.
2936  *
2937  * Returns 1 in case the prison owns the vnet, 0 otherwise.
2938  */
2939 int
2940 prison_owns_vnet(struct ucred *cred)
2941 {
2942 
2943 	/*
2944 	 * vnets cannot be added/removed after jail creation,
2945 	 * so no need to lock here.
2946 	 */
2947 	return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
2948 }
2949 #endif
2950 
2951 /*
2952  * Determine whether the subject represented by cred can "see"
2953  * status of a mount point.
2954  * Returns: 0 for permitted, ENOENT otherwise.
2955  * XXX: This function should be called cr_canseemount() and should be
2956  *      placed in kern_prot.c.
2957  */
2958 int
2959 prison_canseemount(struct ucred *cred, struct mount *mp)
2960 {
2961 	struct prison *pr;
2962 	struct statfs *sp;
2963 	size_t len;
2964 
2965 	pr = cred->cr_prison;
2966 	if (pr->pr_enforce_statfs == 0)
2967 		return (0);
2968 	if (pr->pr_root->v_mount == mp)
2969 		return (0);
2970 	if (pr->pr_enforce_statfs == 2)
2971 		return (ENOENT);
2972 	/*
2973 	 * If jail's chroot directory is set to "/" we should be able to see
2974 	 * all mount-points from inside a jail.
2975 	 * This is ugly check, but this is the only situation when jail's
2976 	 * directory ends with '/'.
2977 	 */
2978 	if (strcmp(pr->pr_path, "/") == 0)
2979 		return (0);
2980 	len = strlen(pr->pr_path);
2981 	sp = &mp->mnt_stat;
2982 	if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
2983 		return (ENOENT);
2984 	/*
2985 	 * Be sure that we don't have situation where jail's root directory
2986 	 * is "/some/path" and mount point is "/some/pathpath".
2987 	 */
2988 	if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
2989 		return (ENOENT);
2990 	return (0);
2991 }
2992 
2993 void
2994 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
2995 {
2996 	char jpath[MAXPATHLEN];
2997 	struct prison *pr;
2998 	size_t len;
2999 
3000 	pr = cred->cr_prison;
3001 	if (pr->pr_enforce_statfs == 0)
3002 		return;
3003 	if (prison_canseemount(cred, mp) != 0) {
3004 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3005 		strlcpy(sp->f_mntonname, "[restricted]",
3006 		    sizeof(sp->f_mntonname));
3007 		return;
3008 	}
3009 	if (pr->pr_root->v_mount == mp) {
3010 		/*
3011 		 * Clear current buffer data, so we are sure nothing from
3012 		 * the valid path left there.
3013 		 */
3014 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3015 		*sp->f_mntonname = '/';
3016 		return;
3017 	}
3018 	/*
3019 	 * If jail's chroot directory is set to "/" we should be able to see
3020 	 * all mount-points from inside a jail.
3021 	 */
3022 	if (strcmp(pr->pr_path, "/") == 0)
3023 		return;
3024 	len = strlen(pr->pr_path);
3025 	strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3026 	/*
3027 	 * Clear current buffer data, so we are sure nothing from
3028 	 * the valid path left there.
3029 	 */
3030 	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3031 	if (*jpath == '\0') {
3032 		/* Should never happen. */
3033 		*sp->f_mntonname = '/';
3034 	} else {
3035 		strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3036 	}
3037 }
3038 
3039 /*
3040  * Check with permission for a specific privilege is granted within jail.  We
3041  * have a specific list of accepted privileges; the rest are denied.
3042  */
3043 int
3044 prison_priv_check(struct ucred *cred, int priv)
3045 {
3046 
3047 	/*
3048 	 * Some policies have custom handlers. This routine should not be
3049 	 * called for them. See priv_check_cred().
3050 	 */
3051 	switch (priv) {
3052 	case PRIV_VFS_GENERATION:
3053 		KASSERT(0, ("prison_priv_check instead of a custom handler "
3054 		    "called for %d\n", priv));
3055 	}
3056 
3057 	if (!jailed(cred))
3058 		return (0);
3059 
3060 #ifdef VIMAGE
3061 	/*
3062 	 * Privileges specific to prisons with a virtual network stack.
3063 	 * There might be a duplicate entry here in case the privilege
3064 	 * is only granted conditionally in the legacy jail case.
3065 	 */
3066 	switch (priv) {
3067 #ifdef notyet
3068 		/*
3069 		 * NFS-specific privileges.
3070 		 */
3071 	case PRIV_NFS_DAEMON:
3072 	case PRIV_NFS_LOCKD:
3073 #endif
3074 		/*
3075 		 * Network stack privileges.
3076 		 */
3077 	case PRIV_NET_BRIDGE:
3078 	case PRIV_NET_GRE:
3079 	case PRIV_NET_BPF:
3080 	case PRIV_NET_RAW:		/* Dup, cond. in legacy jail case. */
3081 	case PRIV_NET_ROUTE:
3082 	case PRIV_NET_TAP:
3083 	case PRIV_NET_SETIFMTU:
3084 	case PRIV_NET_SETIFFLAGS:
3085 	case PRIV_NET_SETIFCAP:
3086 	case PRIV_NET_SETIFDESCR:
3087 	case PRIV_NET_SETIFNAME	:
3088 	case PRIV_NET_SETIFMETRIC:
3089 	case PRIV_NET_SETIFPHYS:
3090 	case PRIV_NET_SETIFMAC:
3091 	case PRIV_NET_SETLANPCP:
3092 	case PRIV_NET_ADDMULTI:
3093 	case PRIV_NET_DELMULTI:
3094 	case PRIV_NET_HWIOCTL:
3095 	case PRIV_NET_SETLLADDR:
3096 	case PRIV_NET_ADDIFGROUP:
3097 	case PRIV_NET_DELIFGROUP:
3098 	case PRIV_NET_IFCREATE:
3099 	case PRIV_NET_IFDESTROY:
3100 	case PRIV_NET_ADDIFADDR:
3101 	case PRIV_NET_DELIFADDR:
3102 	case PRIV_NET_LAGG:
3103 	case PRIV_NET_GIF:
3104 	case PRIV_NET_SETIFVNET:
3105 	case PRIV_NET_SETIFFIB:
3106 
3107 		/*
3108 		 * 802.11-related privileges.
3109 		 */
3110 	case PRIV_NET80211_GETKEY:
3111 #ifdef notyet
3112 	case PRIV_NET80211_MANAGE:		/* XXX-BZ discuss with sam@ */
3113 #endif
3114 
3115 #ifdef notyet
3116 		/*
3117 		 * ATM privileges.
3118 		 */
3119 	case PRIV_NETATM_CFG:
3120 	case PRIV_NETATM_ADD:
3121 	case PRIV_NETATM_DEL:
3122 	case PRIV_NETATM_SET:
3123 
3124 		/*
3125 		 * Bluetooth privileges.
3126 		 */
3127 	case PRIV_NETBLUETOOTH_RAW:
3128 #endif
3129 
3130 		/*
3131 		 * Netgraph and netgraph module privileges.
3132 		 */
3133 	case PRIV_NETGRAPH_CONTROL:
3134 #ifdef notyet
3135 	case PRIV_NETGRAPH_TTY:
3136 #endif
3137 
3138 		/*
3139 		 * IPv4 and IPv6 privileges.
3140 		 */
3141 	case PRIV_NETINET_IPFW:
3142 	case PRIV_NETINET_DIVERT:
3143 	case PRIV_NETINET_PF:
3144 	case PRIV_NETINET_DUMMYNET:
3145 	case PRIV_NETINET_CARP:
3146 	case PRIV_NETINET_MROUTE:
3147 	case PRIV_NETINET_RAW:
3148 	case PRIV_NETINET_ADDRCTRL6:
3149 	case PRIV_NETINET_ND6:
3150 	case PRIV_NETINET_SCOPE6:
3151 	case PRIV_NETINET_ALIFETIME6:
3152 	case PRIV_NETINET_IPSEC:
3153 	case PRIV_NETINET_BINDANY:
3154 
3155 #ifdef notyet
3156 		/*
3157 		 * NCP privileges.
3158 		 */
3159 	case PRIV_NETNCP:
3160 
3161 		/*
3162 		 * SMB privileges.
3163 		 */
3164 	case PRIV_NETSMB:
3165 #endif
3166 
3167 	/*
3168 	 * No default: or deny here.
3169 	 * In case of no permit fall through to next switch().
3170 	 */
3171 		if (cred->cr_prison->pr_flags & PR_VNET)
3172 			return (0);
3173 	}
3174 #endif /* VIMAGE */
3175 
3176 	switch (priv) {
3177 
3178 		/*
3179 		 * Allow ktrace privileges for root in jail.
3180 		 */
3181 	case PRIV_KTRACE:
3182 
3183 #if 0
3184 		/*
3185 		 * Allow jailed processes to configure audit identity and
3186 		 * submit audit records (login, etc).  In the future we may
3187 		 * want to further refine the relationship between audit and
3188 		 * jail.
3189 		 */
3190 	case PRIV_AUDIT_GETAUDIT:
3191 	case PRIV_AUDIT_SETAUDIT:
3192 	case PRIV_AUDIT_SUBMIT:
3193 #endif
3194 
3195 		/*
3196 		 * Allow jailed processes to manipulate process UNIX
3197 		 * credentials in any way they see fit.
3198 		 */
3199 	case PRIV_CRED_SETUID:
3200 	case PRIV_CRED_SETEUID:
3201 	case PRIV_CRED_SETGID:
3202 	case PRIV_CRED_SETEGID:
3203 	case PRIV_CRED_SETGROUPS:
3204 	case PRIV_CRED_SETREUID:
3205 	case PRIV_CRED_SETREGID:
3206 	case PRIV_CRED_SETRESUID:
3207 	case PRIV_CRED_SETRESGID:
3208 
3209 		/*
3210 		 * Jail implements visibility constraints already, so allow
3211 		 * jailed root to override uid/gid-based constraints.
3212 		 */
3213 	case PRIV_SEEOTHERGIDS:
3214 	case PRIV_SEEOTHERUIDS:
3215 
3216 		/*
3217 		 * Jail implements inter-process debugging limits already, so
3218 		 * allow jailed root various debugging privileges.
3219 		 */
3220 	case PRIV_DEBUG_DIFFCRED:
3221 	case PRIV_DEBUG_SUGID:
3222 	case PRIV_DEBUG_UNPRIV:
3223 
3224 		/*
3225 		 * Allow jail to set various resource limits and login
3226 		 * properties, and for now, exceed process resource limits.
3227 		 */
3228 	case PRIV_PROC_LIMIT:
3229 	case PRIV_PROC_SETLOGIN:
3230 	case PRIV_PROC_SETRLIMIT:
3231 
3232 		/*
3233 		 * System V and POSIX IPC privileges are granted in jail.
3234 		 */
3235 	case PRIV_IPC_READ:
3236 	case PRIV_IPC_WRITE:
3237 	case PRIV_IPC_ADMIN:
3238 	case PRIV_IPC_MSGSIZE:
3239 	case PRIV_MQ_ADMIN:
3240 
3241 		/*
3242 		 * Jail operations within a jail work on child jails.
3243 		 */
3244 	case PRIV_JAIL_ATTACH:
3245 	case PRIV_JAIL_SET:
3246 	case PRIV_JAIL_REMOVE:
3247 
3248 		/*
3249 		 * Jail implements its own inter-process limits, so allow
3250 		 * root processes in jail to change scheduling on other
3251 		 * processes in the same jail.  Likewise for signalling.
3252 		 */
3253 	case PRIV_SCHED_DIFFCRED:
3254 	case PRIV_SCHED_CPUSET:
3255 	case PRIV_SIGNAL_DIFFCRED:
3256 	case PRIV_SIGNAL_SUGID:
3257 
3258 		/*
3259 		 * Allow jailed processes to write to sysctls marked as jail
3260 		 * writable.
3261 		 */
3262 	case PRIV_SYSCTL_WRITEJAIL:
3263 
3264 		/*
3265 		 * Allow root in jail to manage a variety of quota
3266 		 * properties.  These should likely be conditional on a
3267 		 * configuration option.
3268 		 */
3269 	case PRIV_VFS_GETQUOTA:
3270 	case PRIV_VFS_SETQUOTA:
3271 
3272 		/*
3273 		 * Since Jail relies on chroot() to implement file system
3274 		 * protections, grant many VFS privileges to root in jail.
3275 		 * Be careful to exclude mount-related and NFS-related
3276 		 * privileges.
3277 		 */
3278 	case PRIV_VFS_READ:
3279 	case PRIV_VFS_WRITE:
3280 	case PRIV_VFS_ADMIN:
3281 	case PRIV_VFS_EXEC:
3282 	case PRIV_VFS_LOOKUP:
3283 	case PRIV_VFS_BLOCKRESERVE:	/* XXXRW: Slightly surprising. */
3284 	case PRIV_VFS_CHFLAGS_DEV:
3285 	case PRIV_VFS_CHOWN:
3286 	case PRIV_VFS_CHROOT:
3287 	case PRIV_VFS_RETAINSUGID:
3288 	case PRIV_VFS_FCHROOT:
3289 	case PRIV_VFS_LINK:
3290 	case PRIV_VFS_SETGID:
3291 	case PRIV_VFS_STAT:
3292 	case PRIV_VFS_STICKYFILE:
3293 
3294 		/*
3295 		 * As in the non-jail case, non-root users are expected to be
3296 		 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3297 		 * exists in the jail and they have permission to access it).
3298 		 */
3299 	case PRIV_KMEM_READ:
3300 		return (0);
3301 
3302 		/*
3303 		 * Depending on the global setting, allow privilege of
3304 		 * setting system flags.
3305 		 */
3306 	case PRIV_VFS_SYSFLAGS:
3307 		if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3308 			return (0);
3309 		else
3310 			return (EPERM);
3311 
3312 		/*
3313 		 * Depending on the global setting, allow privilege of
3314 		 * mounting/unmounting file systems.
3315 		 */
3316 	case PRIV_VFS_MOUNT:
3317 	case PRIV_VFS_UNMOUNT:
3318 	case PRIV_VFS_MOUNT_NONUSER:
3319 	case PRIV_VFS_MOUNT_OWNER:
3320 		if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3321 		    cred->cr_prison->pr_enforce_statfs < 2)
3322 			return (0);
3323 		else
3324 			return (EPERM);
3325 
3326 		/*
3327 		 * Conditionnaly allow locking (unlocking) physical pages
3328 		 * in memory.
3329 		 */
3330 	case PRIV_VM_MLOCK:
3331 	case PRIV_VM_MUNLOCK:
3332 		if (cred->cr_prison->pr_allow & PR_ALLOW_MLOCK)
3333 			return (0);
3334 		else
3335 			return (EPERM);
3336 
3337 		/*
3338 		 * Conditionally allow jailed root to bind reserved ports.
3339 		 */
3340 	case PRIV_NETINET_RESERVEDPORT:
3341 		if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS)
3342 			return (0);
3343 		else
3344 			return (EPERM);
3345 
3346 		/*
3347 		 * Allow jailed root to reuse in-use ports.
3348 		 */
3349 	case PRIV_NETINET_REUSEPORT:
3350 		return (0);
3351 
3352 		/*
3353 		 * Allow jailed root to set certain IPv4/6 (option) headers.
3354 		 */
3355 	case PRIV_NETINET_SETHDROPTS:
3356 		return (0);
3357 
3358 		/*
3359 		 * Conditionally allow creating raw sockets in jail.
3360 		 */
3361 	case PRIV_NETINET_RAW:
3362 		if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3363 			return (0);
3364 		else
3365 			return (EPERM);
3366 
3367 		/*
3368 		 * Since jail implements its own visibility limits on netstat
3369 		 * sysctls, allow getcred.  This allows identd to work in
3370 		 * jail.
3371 		 */
3372 	case PRIV_NETINET_GETCRED:
3373 		return (0);
3374 
3375 		/*
3376 		 * Allow jailed root to set loginclass.
3377 		 */
3378 	case PRIV_PROC_SETLOGINCLASS:
3379 		return (0);
3380 
3381 		/*
3382 		 * Do not allow a process inside a jail to read the kernel
3383 		 * message buffer unless explicitly permitted.
3384 		 */
3385 	case PRIV_MSGBUF:
3386 		if (cred->cr_prison->pr_allow & PR_ALLOW_READ_MSGBUF)
3387 			return (0);
3388 		return (EPERM);
3389 
3390 	default:
3391 		/*
3392 		 * In all remaining cases, deny the privilege request.  This
3393 		 * includes almost all network privileges, many system
3394 		 * configuration privileges.
3395 		 */
3396 		return (EPERM);
3397 	}
3398 }
3399 
3400 /*
3401  * Return the part of pr2's name that is relative to pr1, or the whole name
3402  * if it does not directly follow.
3403  */
3404 
3405 char *
3406 prison_name(struct prison *pr1, struct prison *pr2)
3407 {
3408 	char *name;
3409 
3410 	/* Jails see themselves as "0" (if they see themselves at all). */
3411 	if (pr1 == pr2)
3412 		return "0";
3413 	name = pr2->pr_name;
3414 	if (prison_ischild(pr1, pr2)) {
3415 		/*
3416 		 * pr1 isn't locked (and allprison_lock may not be either)
3417 		 * so its length can't be counted on.  But the number of dots
3418 		 * can be counted on - and counted.
3419 		 */
3420 		for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3421 			name = strchr(name, '.') + 1;
3422 	}
3423 	return (name);
3424 }
3425 
3426 /*
3427  * Return the part of pr2's path that is relative to pr1, or the whole path
3428  * if it does not directly follow.
3429  */
3430 static char *
3431 prison_path(struct prison *pr1, struct prison *pr2)
3432 {
3433 	char *path1, *path2;
3434 	int len1;
3435 
3436 	path1 = pr1->pr_path;
3437 	path2 = pr2->pr_path;
3438 	if (!strcmp(path1, "/"))
3439 		return (path2);
3440 	len1 = strlen(path1);
3441 	if (strncmp(path1, path2, len1))
3442 		return (path2);
3443 	if (path2[len1] == '\0')
3444 		return "/";
3445 	if (path2[len1] == '/')
3446 		return (path2 + len1);
3447 	return (path2);
3448 }
3449 
3450 /*
3451  * Jail-related sysctls.
3452  */
3453 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
3454     "Jails");
3455 
3456 static int
3457 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3458 {
3459 	struct xprison *xp;
3460 	struct prison *pr, *cpr;
3461 #ifdef INET
3462 	struct in_addr *ip4 = NULL;
3463 	int ip4s = 0;
3464 #endif
3465 #ifdef INET6
3466 	struct in6_addr *ip6 = NULL;
3467 	int ip6s = 0;
3468 #endif
3469 	int descend, error;
3470 
3471 	xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3472 	pr = req->td->td_ucred->cr_prison;
3473 	error = 0;
3474 	sx_slock(&allprison_lock);
3475 	FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3476 #if defined(INET) || defined(INET6)
3477  again:
3478 #endif
3479 		mtx_lock(&cpr->pr_mtx);
3480 #ifdef INET
3481 		if (cpr->pr_ip4s > 0) {
3482 			if (ip4s < cpr->pr_ip4s) {
3483 				ip4s = cpr->pr_ip4s;
3484 				mtx_unlock(&cpr->pr_mtx);
3485 				ip4 = realloc(ip4, ip4s *
3486 				    sizeof(struct in_addr), M_TEMP, M_WAITOK);
3487 				goto again;
3488 			}
3489 			bcopy(cpr->pr_ip4, ip4,
3490 			    cpr->pr_ip4s * sizeof(struct in_addr));
3491 		}
3492 #endif
3493 #ifdef INET6
3494 		if (cpr->pr_ip6s > 0) {
3495 			if (ip6s < cpr->pr_ip6s) {
3496 				ip6s = cpr->pr_ip6s;
3497 				mtx_unlock(&cpr->pr_mtx);
3498 				ip6 = realloc(ip6, ip6s *
3499 				    sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3500 				goto again;
3501 			}
3502 			bcopy(cpr->pr_ip6, ip6,
3503 			    cpr->pr_ip6s * sizeof(struct in6_addr));
3504 		}
3505 #endif
3506 		if (cpr->pr_ref == 0) {
3507 			mtx_unlock(&cpr->pr_mtx);
3508 			continue;
3509 		}
3510 		bzero(xp, sizeof(*xp));
3511 		xp->pr_version = XPRISON_VERSION;
3512 		xp->pr_id = cpr->pr_id;
3513 		xp->pr_state = cpr->pr_uref > 0
3514 		    ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3515 		strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3516 		strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3517 		strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3518 #ifdef INET
3519 		xp->pr_ip4s = cpr->pr_ip4s;
3520 #endif
3521 #ifdef INET6
3522 		xp->pr_ip6s = cpr->pr_ip6s;
3523 #endif
3524 		mtx_unlock(&cpr->pr_mtx);
3525 		error = SYSCTL_OUT(req, xp, sizeof(*xp));
3526 		if (error)
3527 			break;
3528 #ifdef INET
3529 		if (xp->pr_ip4s > 0) {
3530 			error = SYSCTL_OUT(req, ip4,
3531 			    xp->pr_ip4s * sizeof(struct in_addr));
3532 			if (error)
3533 				break;
3534 		}
3535 #endif
3536 #ifdef INET6
3537 		if (xp->pr_ip6s > 0) {
3538 			error = SYSCTL_OUT(req, ip6,
3539 			    xp->pr_ip6s * sizeof(struct in6_addr));
3540 			if (error)
3541 				break;
3542 		}
3543 #endif
3544 	}
3545 	sx_sunlock(&allprison_lock);
3546 	free(xp, M_TEMP);
3547 #ifdef INET
3548 	free(ip4, M_TEMP);
3549 #endif
3550 #ifdef INET6
3551 	free(ip6, M_TEMP);
3552 #endif
3553 	return (error);
3554 }
3555 
3556 SYSCTL_OID(_security_jail, OID_AUTO, list,
3557     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3558     sysctl_jail_list, "S", "List of active jails");
3559 
3560 static int
3561 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3562 {
3563 	int error, injail;
3564 
3565 	injail = jailed(req->td->td_ucred);
3566 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
3567 
3568 	return (error);
3569 }
3570 
3571 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3572     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3573     sysctl_jail_jailed, "I", "Process in jail?");
3574 
3575 static int
3576 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
3577 {
3578 	int error, havevnet;
3579 #ifdef VIMAGE
3580 	struct ucred *cred = req->td->td_ucred;
3581 
3582 	havevnet = jailed(cred) && prison_owns_vnet(cred);
3583 #else
3584 	havevnet = 0;
3585 #endif
3586 	error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
3587 
3588 	return (error);
3589 }
3590 
3591 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
3592     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3593     sysctl_jail_vnet, "I", "Jail owns vnet?");
3594 
3595 #if defined(INET) || defined(INET6)
3596 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3597     &jail_max_af_ips, 0,
3598     "Number of IP addresses a jail may have at most per address family (deprecated)");
3599 #endif
3600 
3601 /*
3602  * Default parameters for jail(2) compatibility.  For historical reasons,
3603  * the sysctl names have varying similarity to the parameter names.  Prisons
3604  * just see their own parameters, and can't change them.
3605  */
3606 static int
3607 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3608 {
3609 	struct prison *pr;
3610 	int allow, error, i;
3611 
3612 	pr = req->td->td_ucred->cr_prison;
3613 	allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3614 
3615 	/* Get the current flag value, and convert it to a boolean. */
3616 	i = (allow & arg2) ? 1 : 0;
3617 	if (arg1 != NULL)
3618 		i = !i;
3619 	error = sysctl_handle_int(oidp, &i, 0, req);
3620 	if (error || !req->newptr)
3621 		return (error);
3622 	i = i ? arg2 : 0;
3623 	if (arg1 != NULL)
3624 		i ^= arg2;
3625 	/*
3626 	 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3627 	 * for writing.
3628 	 */
3629 	mtx_lock(&prison0.pr_mtx);
3630 	jail_default_allow = (jail_default_allow & ~arg2) | i;
3631 	mtx_unlock(&prison0.pr_mtx);
3632 	return (0);
3633 }
3634 
3635 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3636     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3637     NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3638     "Processes in jail can set their hostnames (deprecated)");
3639 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3640     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3641     (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3642     "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
3643 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3644     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3645     NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3646     "Processes in jail can use System V IPC primitives (deprecated)");
3647 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3648     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3649     NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3650     "Prison root can create raw sockets (deprecated)");
3651 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3652     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3653     NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3654     "Processes in jail can alter system file flags (deprecated)");
3655 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3656     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3657     NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3658     "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
3659 
3660 static int
3661 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3662 {
3663 	struct prison *pr;
3664 	int level, error;
3665 
3666 	pr = req->td->td_ucred->cr_prison;
3667 	level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3668 	error = sysctl_handle_int(oidp, &level, 0, req);
3669 	if (error || !req->newptr)
3670 		return (error);
3671 	*(int *)arg1 = level;
3672 	return (0);
3673 }
3674 
3675 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3676     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3677     &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3678     sysctl_jail_default_level, "I",
3679     "Processes in jail cannot see all mounted file systems (deprecated)");
3680 
3681 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
3682     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
3683     &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
3684     sysctl_jail_default_level, "I",
3685     "Ruleset for the devfs filesystem in jail (deprecated)");
3686 
3687 /*
3688  * Nodes to describe jail parameters.  Maximum length of string parameters
3689  * is returned in the string itself, and the other parameters exist merely
3690  * to make themselves and their types known.
3691  */
3692 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
3693     "Jail parameters");
3694 
3695 int
3696 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3697 {
3698 	int i;
3699 	long l;
3700 	size_t s;
3701 	char numbuf[12];
3702 
3703 	switch (oidp->oid_kind & CTLTYPE)
3704 	{
3705 	case CTLTYPE_LONG:
3706 	case CTLTYPE_ULONG:
3707 		l = 0;
3708 #ifdef SCTL_MASK32
3709 		if (!(req->flags & SCTL_MASK32))
3710 #endif
3711 			return (SYSCTL_OUT(req, &l, sizeof(l)));
3712 	case CTLTYPE_INT:
3713 	case CTLTYPE_UINT:
3714 		i = 0;
3715 		return (SYSCTL_OUT(req, &i, sizeof(i)));
3716 	case CTLTYPE_STRING:
3717 		snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
3718 		return
3719 		    (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3720 	case CTLTYPE_STRUCT:
3721 		s = (size_t)arg2;
3722 		return (SYSCTL_OUT(req, &s, sizeof(s)));
3723 	}
3724 	return (0);
3725 }
3726 
3727 /*
3728  * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
3729  * jail creation time but cannot be changed in an existing jail.
3730  */
3731 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3732 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3733 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3734 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3735 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3736     "I", "Jail secure level");
3737 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
3738     "Jail value for kern.osreldate and uname -K");
3739 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
3740     "Jail value for kern.osrelease and uname -r");
3741 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3742     "I", "Jail cannot see all mounted file systems");
3743 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
3744     "I", "Ruleset for in-jail devfs mounts");
3745 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3746     "B", "Jail persistence");
3747 #ifdef VIMAGE
3748 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3749     "E,jailsys", "Virtual network stack");
3750 #endif
3751 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3752     "B", "Jail is in the process of shutting down");
3753 
3754 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
3755 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
3756     "I", "Current number of child jails");
3757 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
3758     "I", "Maximum number of child jails");
3759 
3760 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
3761 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3762     "Jail hostname");
3763 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3764     "Jail NIS domainname");
3765 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3766     "Jail host UUID");
3767 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3768     "LU", "Jail host ID");
3769 
3770 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3771 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3772 
3773 #ifdef INET
3774 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
3775     "Jail IPv4 address virtualization");
3776 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3777     "S,in_addr,a", "Jail IPv4 addresses");
3778 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3779     "B", "Do (not) use IPv4 source address selection rather than the "
3780     "primary jail IPv4 address.");
3781 #endif
3782 #ifdef INET6
3783 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
3784     "Jail IPv6 address virtualization");
3785 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3786     "S,in6_addr,a", "Jail IPv6 addresses");
3787 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3788     "B", "Do (not) use IPv6 source address selection rather than the "
3789     "primary jail IPv6 address.");
3790 #endif
3791 
3792 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3793 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3794     "B", "Jail may set hostname");
3795 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3796     "B", "Jail may use SYSV IPC");
3797 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3798     "B", "Jail may create raw sockets");
3799 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3800     "B", "Jail may alter system file flags");
3801 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3802     "B", "Jail may set file quotas");
3803 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3804     "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3805 SYSCTL_JAIL_PARAM(_allow, mlock, CTLTYPE_INT | CTLFLAG_RW,
3806     "B", "Jail may lock (unlock) physical pages in memory");
3807 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW,
3808     "B", "Jail may bind sockets to reserved ports");
3809 SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW,
3810     "B", "Jail may read the kernel message buffer");
3811 SYSCTL_JAIL_PARAM(_allow, unprivileged_proc_debug, CTLTYPE_INT | CTLFLAG_RW,
3812     "B", "Unprivileged processes may use process debugging facilities");
3813 
3814 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
3815 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
3816     "B", "Jail may mount/unmount jail-friendly file systems in general");
3817 
3818 /*
3819  * Add a dynamic parameter allow.<name>, or allow.<prefix>.<name>.  Return
3820  * its associated bit in the pr_allow bitmask, or zero if the parameter was
3821  * not created.
3822  */
3823 unsigned
3824 prison_add_allow(const char *prefix, const char *name, const char *prefix_descr,
3825     const char *descr)
3826 {
3827 	struct bool_flags *bf;
3828 	struct sysctl_oid *parent;
3829 	char *allow_name, *allow_noname, *allowed;
3830 #ifndef NO_SYSCTL_DESCR
3831 	char *descr_deprecated;
3832 #endif
3833 	unsigned allow_flag;
3834 
3835 	if (prefix
3836 	    ? asprintf(&allow_name, M_PRISON, "allow.%s.%s", prefix, name)
3837 		< 0 ||
3838 	      asprintf(&allow_noname, M_PRISON, "allow.%s.no%s", prefix, name)
3839 		< 0
3840 	    : asprintf(&allow_name, M_PRISON, "allow.%s", name) < 0 ||
3841 	      asprintf(&allow_noname, M_PRISON, "allow.no%s", name) < 0) {
3842 		free(allow_name, M_PRISON);
3843 		return 0;
3844 	}
3845 
3846 	/*
3847 	 * See if this parameter has already beed added, i.e. a module was
3848 	 * previously loaded/unloaded.
3849 	 */
3850 	mtx_lock(&prison0.pr_mtx);
3851 	for (bf = pr_flag_allow;
3852 	     bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
3853 	     bf++) {
3854 		if (strcmp(bf->name, allow_name) == 0) {
3855 			allow_flag = bf->flag;
3856 			goto no_add;
3857 		}
3858 	}
3859 
3860 	/*
3861 	 * Find a free bit in prison0's pr_allow, failing if there are none
3862 	 * (which shouldn't happen as long as we keep track of how many
3863 	 * potential dynamic flags exist).
3864 	 *
3865 	 * Due to per-jail unprivileged process debugging support
3866 	 * using pr_allow, also verify against PR_ALLOW_ALL_STATIC.
3867 	 * prison0 may have unprivileged process debugging unset.
3868 	 */
3869 	for (allow_flag = 1;; allow_flag <<= 1) {
3870 		if (allow_flag == 0)
3871 			goto no_add;
3872 		if (allow_flag & PR_ALLOW_ALL_STATIC)
3873 			continue;
3874 		if ((prison0.pr_allow & allow_flag) == 0)
3875 			break;
3876 	}
3877 
3878 	/*
3879 	 * Note the parameter in the next open slot in pr_flag_allow.
3880 	 * Set the flag last so code that checks pr_flag_allow can do so
3881 	 * without locking.
3882 	 */
3883 	for (bf = pr_flag_allow; bf->flag != 0; bf++)
3884 		if (bf == pr_flag_allow + nitems(pr_flag_allow)) {
3885 			/* This should never happen, but is not fatal. */
3886 			allow_flag = 0;
3887 			goto no_add;
3888 		}
3889 	prison0.pr_allow |= allow_flag;
3890 	bf->name = allow_name;
3891 	bf->noname = allow_noname;
3892 	bf->flag = allow_flag;
3893 	mtx_unlock(&prison0.pr_mtx);
3894 
3895 	/*
3896 	 * Create sysctls for the paramter, and the back-compat global
3897 	 * permission.
3898 	 */
3899 	parent = prefix
3900 	    ? SYSCTL_ADD_NODE(NULL,
3901 		  SYSCTL_CHILDREN(&sysctl___security_jail_param_allow),
3902 		  OID_AUTO, prefix, CTLFLAG_MPSAFE, 0, prefix_descr)
3903 	    : &sysctl___security_jail_param_allow;
3904 	(void)SYSCTL_ADD_PROC(NULL, SYSCTL_CHILDREN(parent), OID_AUTO,
3905 	    name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3906 	    NULL, 0, sysctl_jail_param, "B", descr);
3907 	if ((prefix
3908 	     ? asprintf(&allowed, M_TEMP, "%s_%s_allowed", prefix, name)
3909 	     : asprintf(&allowed, M_TEMP, "%s_allowed", name)) >= 0) {
3910 #ifndef NO_SYSCTL_DESCR
3911 		(void)asprintf(&descr_deprecated, M_TEMP, "%s (deprecated)",
3912 		    descr);
3913 #endif
3914 		(void)SYSCTL_ADD_PROC(NULL,
3915 		    SYSCTL_CHILDREN(&sysctl___security_jail), OID_AUTO, allowed,
3916 		    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, allow_flag,
3917 		    sysctl_jail_default_allow, "I", descr_deprecated);
3918 #ifndef NO_SYSCTL_DESCR
3919 		free(descr_deprecated, M_TEMP);
3920 #endif
3921 		free(allowed, M_TEMP);
3922 	}
3923 	return allow_flag;
3924 
3925  no_add:
3926 	mtx_unlock(&prison0.pr_mtx);
3927 	free(allow_name, M_PRISON);
3928 	free(allow_noname, M_PRISON);
3929 	return allow_flag;
3930 }
3931 
3932 /*
3933  * The VFS system will register jail-aware filesystems here.  They each get
3934  * a parameter allow.mount.xxxfs and a flag to check when a jailed user
3935  * attempts to mount.
3936  */
3937 void
3938 prison_add_vfs(struct vfsconf *vfsp)
3939 {
3940 #ifdef NO_SYSCTL_DESCR
3941 
3942 	vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
3943 	    NULL, NULL);
3944 #else
3945 	char *descr;
3946 
3947 	(void)asprintf(&descr, M_TEMP, "Jail may mount the %s file system",
3948 	    vfsp->vfc_name);
3949 	vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
3950 	    NULL, descr);
3951 	free(descr, M_TEMP);
3952 #endif
3953 }
3954 
3955 #ifdef RACCT
3956 void
3957 prison_racct_foreach(void (*callback)(struct racct *racct,
3958     void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
3959     void *arg2, void *arg3)
3960 {
3961 	struct prison_racct *prr;
3962 
3963 	ASSERT_RACCT_ENABLED();
3964 
3965 	sx_slock(&allprison_lock);
3966 	if (pre != NULL)
3967 		(pre)();
3968 	LIST_FOREACH(prr, &allprison_racct, prr_next)
3969 		(callback)(prr->prr_racct, arg2, arg3);
3970 	if (post != NULL)
3971 		(post)();
3972 	sx_sunlock(&allprison_lock);
3973 }
3974 
3975 static struct prison_racct *
3976 prison_racct_find_locked(const char *name)
3977 {
3978 	struct prison_racct *prr;
3979 
3980 	ASSERT_RACCT_ENABLED();
3981 	sx_assert(&allprison_lock, SA_XLOCKED);
3982 
3983 	if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
3984 		return (NULL);
3985 
3986 	LIST_FOREACH(prr, &allprison_racct, prr_next) {
3987 		if (strcmp(name, prr->prr_name) != 0)
3988 			continue;
3989 
3990 		/* Found prison_racct with a matching name? */
3991 		prison_racct_hold(prr);
3992 		return (prr);
3993 	}
3994 
3995 	/* Add new prison_racct. */
3996 	prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
3997 	racct_create(&prr->prr_racct);
3998 
3999 	strcpy(prr->prr_name, name);
4000 	refcount_init(&prr->prr_refcount, 1);
4001 	LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4002 
4003 	return (prr);
4004 }
4005 
4006 struct prison_racct *
4007 prison_racct_find(const char *name)
4008 {
4009 	struct prison_racct *prr;
4010 
4011 	ASSERT_RACCT_ENABLED();
4012 
4013 	sx_xlock(&allprison_lock);
4014 	prr = prison_racct_find_locked(name);
4015 	sx_xunlock(&allprison_lock);
4016 	return (prr);
4017 }
4018 
4019 void
4020 prison_racct_hold(struct prison_racct *prr)
4021 {
4022 
4023 	ASSERT_RACCT_ENABLED();
4024 
4025 	refcount_acquire(&prr->prr_refcount);
4026 }
4027 
4028 static void
4029 prison_racct_free_locked(struct prison_racct *prr)
4030 {
4031 
4032 	ASSERT_RACCT_ENABLED();
4033 	sx_assert(&allprison_lock, SA_XLOCKED);
4034 
4035 	if (refcount_release(&prr->prr_refcount)) {
4036 		racct_destroy(&prr->prr_racct);
4037 		LIST_REMOVE(prr, prr_next);
4038 		free(prr, M_PRISON_RACCT);
4039 	}
4040 }
4041 
4042 void
4043 prison_racct_free(struct prison_racct *prr)
4044 {
4045 
4046 	ASSERT_RACCT_ENABLED();
4047 	sx_assert(&allprison_lock, SA_UNLOCKED);
4048 
4049 	if (refcount_release_if_not_last(&prr->prr_refcount))
4050 		return;
4051 
4052 	sx_xlock(&allprison_lock);
4053 	prison_racct_free_locked(prr);
4054 	sx_xunlock(&allprison_lock);
4055 }
4056 
4057 static void
4058 prison_racct_attach(struct prison *pr)
4059 {
4060 	struct prison_racct *prr;
4061 
4062 	ASSERT_RACCT_ENABLED();
4063 	sx_assert(&allprison_lock, SA_XLOCKED);
4064 
4065 	prr = prison_racct_find_locked(pr->pr_name);
4066 	KASSERT(prr != NULL, ("cannot find prison_racct"));
4067 
4068 	pr->pr_prison_racct = prr;
4069 }
4070 
4071 /*
4072  * Handle jail renaming.  From the racct point of view, renaming means
4073  * moving from one prison_racct to another.
4074  */
4075 static void
4076 prison_racct_modify(struct prison *pr)
4077 {
4078 #ifdef RCTL
4079 	struct proc *p;
4080 	struct ucred *cred;
4081 #endif
4082 	struct prison_racct *oldprr;
4083 
4084 	ASSERT_RACCT_ENABLED();
4085 
4086 	sx_slock(&allproc_lock);
4087 	sx_xlock(&allprison_lock);
4088 
4089 	if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4090 		sx_xunlock(&allprison_lock);
4091 		sx_sunlock(&allproc_lock);
4092 		return;
4093 	}
4094 
4095 	oldprr = pr->pr_prison_racct;
4096 	pr->pr_prison_racct = NULL;
4097 
4098 	prison_racct_attach(pr);
4099 
4100 	/*
4101 	 * Move resource utilisation records.
4102 	 */
4103 	racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4104 
4105 #ifdef RCTL
4106 	/*
4107 	 * Force rctl to reattach rules to processes.
4108 	 */
4109 	FOREACH_PROC_IN_SYSTEM(p) {
4110 		PROC_LOCK(p);
4111 		cred = crhold(p->p_ucred);
4112 		PROC_UNLOCK(p);
4113 		rctl_proc_ucred_changed(p, cred);
4114 		crfree(cred);
4115 	}
4116 #endif
4117 
4118 	sx_sunlock(&allproc_lock);
4119 	prison_racct_free_locked(oldprr);
4120 	sx_xunlock(&allprison_lock);
4121 }
4122 
4123 static void
4124 prison_racct_detach(struct prison *pr)
4125 {
4126 
4127 	ASSERT_RACCT_ENABLED();
4128 	sx_assert(&allprison_lock, SA_UNLOCKED);
4129 
4130 	if (pr->pr_prison_racct == NULL)
4131 		return;
4132 	prison_racct_free(pr->pr_prison_racct);
4133 	pr->pr_prison_racct = NULL;
4134 }
4135 #endif /* RACCT */
4136 
4137 #ifdef DDB
4138 
4139 static void
4140 db_show_prison(struct prison *pr)
4141 {
4142 	struct bool_flags *bf;
4143 	struct jailsys_flags *jsf;
4144 #if defined(INET) || defined(INET6)
4145 	int ii;
4146 #endif
4147 	unsigned f;
4148 #ifdef INET
4149 	char ip4buf[INET_ADDRSTRLEN];
4150 #endif
4151 #ifdef INET6
4152 	char ip6buf[INET6_ADDRSTRLEN];
4153 #endif
4154 
4155 	db_printf("prison %p:\n", pr);
4156 	db_printf(" jid             = %d\n", pr->pr_id);
4157 	db_printf(" name            = %s\n", pr->pr_name);
4158 	db_printf(" parent          = %p\n", pr->pr_parent);
4159 	db_printf(" ref             = %d\n", pr->pr_ref);
4160 	db_printf(" uref            = %d\n", pr->pr_uref);
4161 	db_printf(" path            = %s\n", pr->pr_path);
4162 	db_printf(" cpuset          = %d\n", pr->pr_cpuset
4163 	    ? pr->pr_cpuset->cs_id : -1);
4164 #ifdef VIMAGE
4165 	db_printf(" vnet            = %p\n", pr->pr_vnet);
4166 #endif
4167 	db_printf(" root            = %p\n", pr->pr_root);
4168 	db_printf(" securelevel     = %d\n", pr->pr_securelevel);
4169 	db_printf(" devfs_rsnum     = %d\n", pr->pr_devfs_rsnum);
4170 	db_printf(" children.max    = %d\n", pr->pr_childmax);
4171 	db_printf(" children.cur    = %d\n", pr->pr_childcount);
4172 	db_printf(" child           = %p\n", LIST_FIRST(&pr->pr_children));
4173 	db_printf(" sibling         = %p\n", LIST_NEXT(pr, pr_sibling));
4174 	db_printf(" flags           = 0x%x", pr->pr_flags);
4175 	for (bf = pr_flag_bool; bf < pr_flag_bool + nitems(pr_flag_bool); bf++)
4176 		if (pr->pr_flags & bf->flag)
4177 			db_printf(" %s", bf->name);
4178 	for (jsf = pr_flag_jailsys;
4179 	     jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
4180 	     jsf++) {
4181 		f = pr->pr_flags & (jsf->disable | jsf->new);
4182 		db_printf(" %-16s= %s\n", jsf->name,
4183 		    (f != 0 && f == jsf->disable) ? "disable"
4184 		    : (f == jsf->new) ? "new"
4185 		    : "inherit");
4186 	}
4187 	db_printf(" allow           = 0x%x", pr->pr_allow);
4188 	for (bf = pr_flag_allow;
4189 	     bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
4190 	     bf++)
4191 		if (pr->pr_allow & bf->flag)
4192 			db_printf(" %s", bf->name);
4193 	db_printf("\n");
4194 	db_printf(" enforce_statfs  = %d\n", pr->pr_enforce_statfs);
4195 	db_printf(" host.hostname   = %s\n", pr->pr_hostname);
4196 	db_printf(" host.domainname = %s\n", pr->pr_domainname);
4197 	db_printf(" host.hostuuid   = %s\n", pr->pr_hostuuid);
4198 	db_printf(" host.hostid     = %lu\n", pr->pr_hostid);
4199 #ifdef INET
4200 	db_printf(" ip4s            = %d\n", pr->pr_ip4s);
4201 	for (ii = 0; ii < pr->pr_ip4s; ii++)
4202 		db_printf(" %s %s\n",
4203 		    ii == 0 ? "ip4.addr        =" : "                 ",
4204 		    inet_ntoa_r(pr->pr_ip4[ii], ip4buf));
4205 #endif
4206 #ifdef INET6
4207 	db_printf(" ip6s            = %d\n", pr->pr_ip6s);
4208 	for (ii = 0; ii < pr->pr_ip6s; ii++)
4209 		db_printf(" %s %s\n",
4210 		    ii == 0 ? "ip6.addr        =" : "                 ",
4211 		    ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4212 #endif
4213 }
4214 
4215 DB_SHOW_COMMAND(prison, db_show_prison_command)
4216 {
4217 	struct prison *pr;
4218 
4219 	if (!have_addr) {
4220 		/*
4221 		 * Show all prisons in the list, and prison0 which is not
4222 		 * listed.
4223 		 */
4224 		db_show_prison(&prison0);
4225 		if (!db_pager_quit) {
4226 			TAILQ_FOREACH(pr, &allprison, pr_list) {
4227 				db_show_prison(pr);
4228 				if (db_pager_quit)
4229 					break;
4230 			}
4231 		}
4232 		return;
4233 	}
4234 
4235 	if (addr == 0)
4236 		pr = &prison0;
4237 	else {
4238 		/* Look for a prison with the ID and with references. */
4239 		TAILQ_FOREACH(pr, &allprison, pr_list)
4240 			if (pr->pr_id == addr && pr->pr_ref > 0)
4241 				break;
4242 		if (pr == NULL)
4243 			/* Look again, without requiring a reference. */
4244 			TAILQ_FOREACH(pr, &allprison, pr_list)
4245 				if (pr->pr_id == addr)
4246 					break;
4247 		if (pr == NULL)
4248 			/* Assume address points to a valid prison. */
4249 			pr = (struct prison *)addr;
4250 	}
4251 	db_show_prison(pr);
4252 }
4253 
4254 #endif /* DDB */
4255