xref: /trueos/sys/kern/kern_jail.c (revision 8cee81c05db1904906f988fe4ecb93dd8565cf85)
1 /*-
2  * Copyright (c) 1999 Poul-Henning Kamp.
3  * Copyright (c) 2008 Bjoern A. Zeeb.
4  * Copyright (c) 2009 James Gritton.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include "opt_compat.h"
33 #include "opt_ddb.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
44 #include <sys/osd.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
49 #include <sys/jail.h>
50 #include <sys/lock.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
54 #include <sys/sx.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
63 
64 #include <net/if.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 
69 #ifdef DDB
70 #include <ddb/ddb.h>
71 #ifdef INET6
72 #include <netinet6/in6_var.h>
73 #endif /* INET6 */
74 #endif /* DDB */
75 
76 #include <security/mac/mac_framework.h>
77 
78 #define	DEFAULT_HOSTUUID	"00000000-0000-0000-0000-000000000000"
79 
80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
81 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
82 
83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
84 #ifdef INET
85 #ifdef INET6
86 #define	_PR_IP_SADDRSEL	PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
87 #else
88 #define	_PR_IP_SADDRSEL	PR_IP4_SADDRSEL
89 #endif
90 #else /* !INET */
91 #ifdef INET6
92 #define	_PR_IP_SADDRSEL	PR_IP6_SADDRSEL
93 #else
94 #define	_PR_IP_SADDRSEL	0
95 #endif
96 #endif
97 
98 /* prison0 describes what is "real" about the system. */
99 struct prison prison0 = {
100 	.pr_id		= 0,
101 	.pr_name	= "0",
102 	.pr_ref		= 1,
103 	.pr_uref	= 1,
104 	.pr_path	= "/",
105 	.pr_securelevel	= -1,
106 	.pr_devfs_rsnum = 0,
107 	.pr_childmax	= JAIL_MAX,
108 	.pr_hostuuid	= DEFAULT_HOSTUUID,
109 	.pr_children	= LIST_HEAD_INITIALIZER(prison0.pr_children),
110 #ifdef VIMAGE
111 	.pr_flags	= PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
112 #else
113 	.pr_flags	= PR_HOST|_PR_IP_SADDRSEL,
114 #endif
115 	.pr_allow	= PR_ALLOW_ALL,
116 };
117 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
118 
119 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
120 struct	sx allprison_lock;
121 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
122 struct	prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
123 LIST_HEAD(, prison_racct) allprison_racct;
124 int	lastprid = 0;
125 
126 static int do_jail_attach(struct thread *td, struct prison *pr);
127 static void prison_complete(void *context, int pending);
128 static void prison_deref(struct prison *pr, int flags);
129 static char *prison_path(struct prison *pr1, struct prison *pr2);
130 static void prison_remove_one(struct prison *pr);
131 #ifdef RACCT
132 static void prison_racct_attach(struct prison *pr);
133 static void prison_racct_modify(struct prison *pr);
134 static void prison_racct_detach(struct prison *pr);
135 #endif
136 #ifdef INET
137 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
138 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
139 #endif
140 #ifdef INET6
141 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
142 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
143 #endif
144 
145 /* Flags for prison_deref */
146 #define	PD_DEREF	0x01
147 #define	PD_DEUREF	0x02
148 #define	PD_LOCKED	0x04
149 #define	PD_LIST_SLOCKED	0x08
150 #define	PD_LIST_XLOCKED	0x10
151 
152 /*
153  * Parameter names corresponding to PR_* flag values.  Size values are for kvm
154  * as we cannot figure out the size of a sparse array, or an array without a
155  * terminating entry.
156  */
157 static char *pr_flag_names[] = {
158 	[0] = "persist",
159 #ifdef INET
160 	[7] = "ip4.saddrsel",
161 #endif
162 #ifdef INET6
163 	[8] = "ip6.saddrsel",
164 #endif
165 };
166 const size_t pr_flag_names_size = sizeof(pr_flag_names);
167 
168 static char *pr_flag_nonames[] = {
169 	[0] = "nopersist",
170 #ifdef INET
171 	[7] = "ip4.nosaddrsel",
172 #endif
173 #ifdef INET6
174 	[8] = "ip6.nosaddrsel",
175 #endif
176 };
177 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
178 
179 struct jailsys_flags {
180 	const char	*name;
181 	unsigned	 disable;
182 	unsigned	 new;
183 } pr_flag_jailsys[] = {
184 	{ "host", 0, PR_HOST },
185 #ifdef VIMAGE
186 	{ "vnet", 0, PR_VNET },
187 #endif
188 #ifdef INET
189 	{ "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
190 #endif
191 #ifdef INET6
192 	{ "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
193 #endif
194 };
195 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
196 
197 static char *pr_allow_names[] = {
198 	"allow.set_hostname",
199 	"allow.sysvipc",
200 	"allow.raw_sockets",
201 	"allow.chflags",
202 	"allow.mount",
203 	"allow.quotas",
204 	"allow.socket_af",
205 	"allow.mount.devfs",
206 	"allow.mount.nullfs",
207 	"allow.mount.zfs",
208 	"allow.mount.procfs",
209 	"allow.mount.tmpfs",
210 	"allow.mount.fdescfs",
211 };
212 const size_t pr_allow_names_size = sizeof(pr_allow_names);
213 
214 static char *pr_allow_nonames[] = {
215 	"allow.noset_hostname",
216 	"allow.nosysvipc",
217 	"allow.noraw_sockets",
218 	"allow.nochflags",
219 	"allow.nomount",
220 	"allow.noquotas",
221 	"allow.nosocket_af",
222 	"allow.mount.nodevfs",
223 	"allow.mount.nonullfs",
224 	"allow.mount.nozfs",
225 	"allow.mount.noprocfs",
226 	"allow.mount.notmpfs",
227 	"allow.mount.nofdescfs",
228 };
229 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
230 
231 #define	JAIL_DEFAULT_ALLOW		PR_ALLOW_SET_HOSTNAME
232 #define	JAIL_DEFAULT_ENFORCE_STATFS	2
233 #define	JAIL_DEFAULT_DEVFS_RSNUM	0
234 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
235 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
236 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
237 #if defined(INET) || defined(INET6)
238 static unsigned jail_max_af_ips = 255;
239 #endif
240 
241 /*
242  * Initialize the parts of prison0 that can't be static-initialized with
243  * constants.  This is called from proc0_init() after creating thread0 cpuset.
244  */
245 void
prison0_init(void)246 prison0_init(void)
247 {
248 
249 	prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
250 	prison0.pr_osreldate = osreldate;
251 	strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
252 }
253 
254 #ifdef INET
255 static int
qcmp_v4(const void * ip1,const void * ip2)256 qcmp_v4(const void *ip1, const void *ip2)
257 {
258 	in_addr_t iaa, iab;
259 
260 	/*
261 	 * We need to compare in HBO here to get the list sorted as expected
262 	 * by the result of the code.  Sorting NBO addresses gives you
263 	 * interesting results.  If you do not understand, do not try.
264 	 */
265 	iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
266 	iab = ntohl(((const struct in_addr *)ip2)->s_addr);
267 
268 	/*
269 	 * Do not simply return the difference of the two numbers, the int is
270 	 * not wide enough.
271 	 */
272 	if (iaa > iab)
273 		return (1);
274 	else if (iaa < iab)
275 		return (-1);
276 	else
277 		return (0);
278 }
279 #endif
280 
281 #ifdef INET6
282 static int
qcmp_v6(const void * ip1,const void * ip2)283 qcmp_v6(const void *ip1, const void *ip2)
284 {
285 	const struct in6_addr *ia6a, *ia6b;
286 	int i, rc;
287 
288 	ia6a = (const struct in6_addr *)ip1;
289 	ia6b = (const struct in6_addr *)ip2;
290 
291 	rc = 0;
292 	for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
293 		if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
294 			rc = 1;
295 		else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
296 			rc = -1;
297 	}
298 	return (rc);
299 }
300 #endif
301 
302 /*
303  * struct jail_args {
304  *	struct jail *jail;
305  * };
306  */
307 int
sys_jail(struct thread * td,struct jail_args * uap)308 sys_jail(struct thread *td, struct jail_args *uap)
309 {
310 	uint32_t version;
311 	int error;
312 	struct jail j;
313 
314 	error = copyin(uap->jail, &version, sizeof(uint32_t));
315 	if (error)
316 		return (error);
317 
318 	switch (version) {
319 	case 0:
320 	{
321 		struct jail_v0 j0;
322 
323 		/* FreeBSD single IPv4 jails. */
324 		bzero(&j, sizeof(struct jail));
325 		error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
326 		if (error)
327 			return (error);
328 		j.version = j0.version;
329 		j.path = j0.path;
330 		j.hostname = j0.hostname;
331 		j.ip4s = htonl(j0.ip_number);	/* jail_v0 is host order */
332 		break;
333 	}
334 
335 	case 1:
336 		/*
337 		 * Version 1 was used by multi-IPv4 jail implementations
338 		 * that never made it into the official kernel.
339 		 */
340 		return (EINVAL);
341 
342 	case 2:	/* JAIL_API_VERSION */
343 		/* FreeBSD multi-IPv4/IPv6,noIP jails. */
344 		error = copyin(uap->jail, &j, sizeof(struct jail));
345 		if (error)
346 			return (error);
347 		break;
348 
349 	default:
350 		/* Sci-Fi jails are not supported, sorry. */
351 		return (EINVAL);
352 	}
353 	return (kern_jail(td, &j));
354 }
355 
356 int
kern_jail(struct thread * td,struct jail * j)357 kern_jail(struct thread *td, struct jail *j)
358 {
359 	struct iovec optiov[2 * (4
360 			    + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
361 #ifdef INET
362 			    + 1
363 #endif
364 #ifdef INET6
365 			    + 1
366 #endif
367 			    )];
368 	struct uio opt;
369 	char *u_path, *u_hostname, *u_name;
370 #ifdef INET
371 	uint32_t ip4s;
372 	struct in_addr *u_ip4;
373 #endif
374 #ifdef INET6
375 	struct in6_addr *u_ip6;
376 #endif
377 	size_t tmplen;
378 	int error, enforce_statfs, fi;
379 
380 	bzero(&optiov, sizeof(optiov));
381 	opt.uio_iov = optiov;
382 	opt.uio_iovcnt = 0;
383 	opt.uio_offset = -1;
384 	opt.uio_resid = -1;
385 	opt.uio_segflg = UIO_SYSSPACE;
386 	opt.uio_rw = UIO_READ;
387 	opt.uio_td = td;
388 
389 	/* Set permissions for top-level jails from sysctls. */
390 	if (!jailed(td->td_ucred)) {
391 		for (fi = 0; fi < sizeof(pr_allow_names) /
392 		     sizeof(pr_allow_names[0]); fi++) {
393 			optiov[opt.uio_iovcnt].iov_base =
394 			    (jail_default_allow & (1 << fi))
395 			    ? pr_allow_names[fi] : pr_allow_nonames[fi];
396 			optiov[opt.uio_iovcnt].iov_len =
397 			    strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
398 			opt.uio_iovcnt += 2;
399 		}
400 		optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
401 		optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
402 		opt.uio_iovcnt++;
403 		enforce_statfs = jail_default_enforce_statfs;
404 		optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
405 		optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
406 		opt.uio_iovcnt++;
407 	}
408 
409 	tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
410 #ifdef INET
411 	ip4s = (j->version == 0) ? 1 : j->ip4s;
412 	if (ip4s > jail_max_af_ips)
413 		return (EINVAL);
414 	tmplen += ip4s * sizeof(struct in_addr);
415 #else
416 	if (j->ip4s > 0)
417 		return (EINVAL);
418 #endif
419 #ifdef INET6
420 	if (j->ip6s > jail_max_af_ips)
421 		return (EINVAL);
422 	tmplen += j->ip6s * sizeof(struct in6_addr);
423 #else
424 	if (j->ip6s > 0)
425 		return (EINVAL);
426 #endif
427 	u_path = malloc(tmplen, M_TEMP, M_WAITOK);
428 	u_hostname = u_path + MAXPATHLEN;
429 	u_name = u_hostname + MAXHOSTNAMELEN;
430 #ifdef INET
431 	u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
432 #endif
433 #ifdef INET6
434 #ifdef INET
435 	u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
436 #else
437 	u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
438 #endif
439 #endif
440 	optiov[opt.uio_iovcnt].iov_base = "path";
441 	optiov[opt.uio_iovcnt].iov_len = sizeof("path");
442 	opt.uio_iovcnt++;
443 	optiov[opt.uio_iovcnt].iov_base = u_path;
444 	error = copyinstr(j->path, u_path, MAXPATHLEN,
445 	    &optiov[opt.uio_iovcnt].iov_len);
446 	if (error) {
447 		free(u_path, M_TEMP);
448 		return (error);
449 	}
450 	opt.uio_iovcnt++;
451 	optiov[opt.uio_iovcnt].iov_base = "host.hostname";
452 	optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
453 	opt.uio_iovcnt++;
454 	optiov[opt.uio_iovcnt].iov_base = u_hostname;
455 	error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
456 	    &optiov[opt.uio_iovcnt].iov_len);
457 	if (error) {
458 		free(u_path, M_TEMP);
459 		return (error);
460 	}
461 	opt.uio_iovcnt++;
462 	if (j->jailname != NULL) {
463 		optiov[opt.uio_iovcnt].iov_base = "name";
464 		optiov[opt.uio_iovcnt].iov_len = sizeof("name");
465 		opt.uio_iovcnt++;
466 		optiov[opt.uio_iovcnt].iov_base = u_name;
467 		error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
468 		    &optiov[opt.uio_iovcnt].iov_len);
469 		if (error) {
470 			free(u_path, M_TEMP);
471 			return (error);
472 		}
473 		opt.uio_iovcnt++;
474 	}
475 #ifdef INET
476 	optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
477 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
478 	opt.uio_iovcnt++;
479 	optiov[opt.uio_iovcnt].iov_base = u_ip4;
480 	optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
481 	if (j->version == 0)
482 		u_ip4->s_addr = j->ip4s;
483 	else {
484 		error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
485 		if (error) {
486 			free(u_path, M_TEMP);
487 			return (error);
488 		}
489 	}
490 	opt.uio_iovcnt++;
491 #endif
492 #ifdef INET6
493 	optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
494 	optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
495 	opt.uio_iovcnt++;
496 	optiov[opt.uio_iovcnt].iov_base = u_ip6;
497 	optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
498 	error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
499 	if (error) {
500 		free(u_path, M_TEMP);
501 		return (error);
502 	}
503 	opt.uio_iovcnt++;
504 #endif
505 	KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
506 	    ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
507 	error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
508 	free(u_path, M_TEMP);
509 	return (error);
510 }
511 
512 
513 /*
514  * struct jail_set_args {
515  *	struct iovec *iovp;
516  *	unsigned int iovcnt;
517  *	int flags;
518  * };
519  */
520 int
sys_jail_set(struct thread * td,struct jail_set_args * uap)521 sys_jail_set(struct thread *td, struct jail_set_args *uap)
522 {
523 	struct uio *auio;
524 	int error;
525 
526 	/* Check that we have an even number of iovecs. */
527 	if (uap->iovcnt & 1)
528 		return (EINVAL);
529 
530 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
531 	if (error)
532 		return (error);
533 	error = kern_jail_set(td, auio, uap->flags);
534 	free(auio, M_IOV);
535 	return (error);
536 }
537 
538 int
kern_jail_set(struct thread * td,struct uio * optuio,int flags)539 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
540 {
541 	struct nameidata nd;
542 #ifdef INET
543 	struct in_addr *ip4;
544 #endif
545 #ifdef INET6
546 	struct in6_addr *ip6;
547 #endif
548 	struct vfsopt *opt;
549 	struct vfsoptlist *opts;
550 	struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
551 	struct vnode *root;
552 	char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
553 	char *g_path, *osrelstr;
554 #if defined(INET) || defined(INET6)
555 	struct prison *tppr;
556 	void *op;
557 #endif
558 	unsigned long hid;
559 	size_t namelen, onamelen;
560 	int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
561 	int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
562 	int fi, jid, jsys, len, level;
563 	int childmax, osreldt, rsnum, slevel;
564 	int fullpath_disabled;
565 #if defined(INET) || defined(INET6)
566 	int ii, ij;
567 #endif
568 #ifdef INET
569 	int ip4s, redo_ip4;
570 #endif
571 #ifdef INET6
572 	int ip6s, redo_ip6;
573 #endif
574 	uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
575 	unsigned tallow;
576 	char numbuf[12];
577 
578 	error = priv_check(td, PRIV_JAIL_SET);
579 	if (!error && (flags & JAIL_ATTACH))
580 		error = priv_check(td, PRIV_JAIL_ATTACH);
581 	if (error)
582 		return (error);
583 	mypr = ppr = td->td_ucred->cr_prison;
584 	if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
585 		return (EPERM);
586 	if (flags & ~JAIL_SET_MASK)
587 		return (EINVAL);
588 
589 	/*
590 	 * Check all the parameters before committing to anything.  Not all
591 	 * errors can be caught early, but we may as well try.  Also, this
592 	 * takes care of some expensive stuff (path lookup) before getting
593 	 * the allprison lock.
594 	 *
595 	 * XXX Jails are not filesystems, and jail parameters are not mount
596 	 *     options.  But it makes more sense to re-use the vfsopt code
597 	 *     than duplicate it under a different name.
598 	 */
599 	error = vfs_buildopts(optuio, &opts);
600 	if (error)
601 		return (error);
602 #ifdef INET
603 	ip4 = NULL;
604 #endif
605 #ifdef INET6
606 	ip6 = NULL;
607 #endif
608 	g_path = NULL;
609 
610 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
611 	if (error == ENOENT)
612 		jid = 0;
613 	else if (error != 0)
614 		goto done_free;
615 
616 	error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
617 	if (error == ENOENT)
618 		gotslevel = 0;
619 	else if (error != 0)
620 		goto done_free;
621 	else
622 		gotslevel = 1;
623 
624 	error =
625 	    vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
626 	if (error == ENOENT)
627 		gotchildmax = 0;
628 	else if (error != 0)
629 		goto done_free;
630 	else
631 		gotchildmax = 1;
632 
633 	error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
634 	if (error == ENOENT)
635 		gotenforce = 0;
636 	else if (error != 0)
637 		goto done_free;
638 	else if (enforce < 0 || enforce > 2) {
639 		error = EINVAL;
640 		goto done_free;
641 	} else
642 		gotenforce = 1;
643 
644 	error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
645 	if (error == ENOENT)
646 		gotrsnum = 0;
647 	else if (error != 0)
648 		goto done_free;
649 	else
650 		gotrsnum = 1;
651 
652 	pr_flags = ch_flags = 0;
653 	for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
654 	    fi++) {
655 		if (pr_flag_names[fi] == NULL)
656 			continue;
657 		vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
658 		vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
659 	}
660 	ch_flags |= pr_flags;
661 	for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
662 	    fi++) {
663 		error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
664 		    sizeof(jsys));
665 		if (error == ENOENT)
666 			continue;
667 		if (error != 0)
668 			goto done_free;
669 		switch (jsys) {
670 		case JAIL_SYS_DISABLE:
671 			if (!pr_flag_jailsys[fi].disable) {
672 				error = EINVAL;
673 				goto done_free;
674 			}
675 			pr_flags |= pr_flag_jailsys[fi].disable;
676 			break;
677 		case JAIL_SYS_NEW:
678 			pr_flags |= pr_flag_jailsys[fi].new;
679 			break;
680 		case JAIL_SYS_INHERIT:
681 			break;
682 		default:
683 			error = EINVAL;
684 			goto done_free;
685 		}
686 		ch_flags |=
687 		    pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
688 	}
689 	if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
690 	    && !(pr_flags & PR_PERSIST)) {
691 		error = EINVAL;
692 		vfs_opterror(opts, "new jail must persist or attach");
693 		goto done_errmsg;
694 	}
695 #ifdef VIMAGE
696 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
697 		error = EINVAL;
698 		vfs_opterror(opts, "vnet cannot be changed after creation");
699 		goto done_errmsg;
700 	}
701 #endif
702 #ifdef INET
703 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
704 		error = EINVAL;
705 		vfs_opterror(opts, "ip4 cannot be changed after creation");
706 		goto done_errmsg;
707 	}
708 #endif
709 #ifdef INET6
710 	if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
711 		error = EINVAL;
712 		vfs_opterror(opts, "ip6 cannot be changed after creation");
713 		goto done_errmsg;
714 	}
715 #endif
716 
717 	pr_allow = ch_allow = 0;
718 	for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
719 	    fi++) {
720 		vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
721 		vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
722 	}
723 	ch_allow |= pr_allow;
724 
725 	error = vfs_getopt(opts, "name", (void **)&name, &len);
726 	if (error == ENOENT)
727 		name = NULL;
728 	else if (error != 0)
729 		goto done_free;
730 	else {
731 		if (len == 0 || name[len - 1] != '\0') {
732 			error = EINVAL;
733 			goto done_free;
734 		}
735 		if (len > MAXHOSTNAMELEN) {
736 			error = ENAMETOOLONG;
737 			goto done_free;
738 		}
739 	}
740 
741 	error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
742 	if (error == ENOENT)
743 		host = NULL;
744 	else if (error != 0)
745 		goto done_free;
746 	else {
747 		ch_flags |= PR_HOST;
748 		pr_flags |= PR_HOST;
749 		if (len == 0 || host[len - 1] != '\0') {
750 			error = EINVAL;
751 			goto done_free;
752 		}
753 		if (len > MAXHOSTNAMELEN) {
754 			error = ENAMETOOLONG;
755 			goto done_free;
756 		}
757 	}
758 
759 	error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
760 	if (error == ENOENT)
761 		domain = NULL;
762 	else if (error != 0)
763 		goto done_free;
764 	else {
765 		ch_flags |= PR_HOST;
766 		pr_flags |= PR_HOST;
767 		if (len == 0 || domain[len - 1] != '\0') {
768 			error = EINVAL;
769 			goto done_free;
770 		}
771 		if (len > MAXHOSTNAMELEN) {
772 			error = ENAMETOOLONG;
773 			goto done_free;
774 		}
775 	}
776 
777 	error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
778 	if (error == ENOENT)
779 		uuid = NULL;
780 	else if (error != 0)
781 		goto done_free;
782 	else {
783 		ch_flags |= PR_HOST;
784 		pr_flags |= PR_HOST;
785 		if (len == 0 || uuid[len - 1] != '\0') {
786 			error = EINVAL;
787 			goto done_free;
788 		}
789 		if (len > HOSTUUIDLEN) {
790 			error = ENAMETOOLONG;
791 			goto done_free;
792 		}
793 	}
794 
795 #ifdef COMPAT_FREEBSD32
796 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
797 		uint32_t hid32;
798 
799 		error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
800 		hid = hid32;
801 	} else
802 #endif
803 		error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
804 	if (error == ENOENT)
805 		gothid = 0;
806 	else if (error != 0)
807 		goto done_free;
808 	else {
809 		gothid = 1;
810 		ch_flags |= PR_HOST;
811 		pr_flags |= PR_HOST;
812 	}
813 
814 #ifdef INET
815 	error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
816 	if (error == ENOENT)
817 		ip4s = 0;
818 	else if (error != 0)
819 		goto done_free;
820 	else if (ip4s & (sizeof(*ip4) - 1)) {
821 		error = EINVAL;
822 		goto done_free;
823 	} else {
824 		ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
825 		if (ip4s == 0)
826 			pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
827 		else {
828 			pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
829 			ip4s /= sizeof(*ip4);
830 			if (ip4s > jail_max_af_ips) {
831 				error = EINVAL;
832 				vfs_opterror(opts, "too many IPv4 addresses");
833 				goto done_errmsg;
834 			}
835 			ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
836 			bcopy(op, ip4, ip4s * sizeof(*ip4));
837 			/*
838 			 * IP addresses are all sorted but ip[0] to preserve
839 			 * the primary IP address as given from userland.
840 			 * This special IP is used for unbound outgoing
841 			 * connections as well for "loopback" traffic in case
842 			 * source address selection cannot find any more fitting
843 			 * address to connect from.
844 			 */
845 			if (ip4s > 1)
846 				qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
847 			/*
848 			 * Check for duplicate addresses and do some simple
849 			 * zero and broadcast checks. If users give other bogus
850 			 * addresses it is their problem.
851 			 *
852 			 * We do not have to care about byte order for these
853 			 * checks so we will do them in NBO.
854 			 */
855 			for (ii = 0; ii < ip4s; ii++) {
856 				if (ip4[ii].s_addr == INADDR_ANY ||
857 				    ip4[ii].s_addr == INADDR_BROADCAST) {
858 					error = EINVAL;
859 					goto done_free;
860 				}
861 				if ((ii+1) < ip4s &&
862 				    (ip4[0].s_addr == ip4[ii+1].s_addr ||
863 				     ip4[ii].s_addr == ip4[ii+1].s_addr)) {
864 					error = EINVAL;
865 					goto done_free;
866 				}
867 			}
868 		}
869 	}
870 #endif
871 
872 #ifdef INET6
873 	error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
874 	if (error == ENOENT)
875 		ip6s = 0;
876 	else if (error != 0)
877 		goto done_free;
878 	else if (ip6s & (sizeof(*ip6) - 1)) {
879 		error = EINVAL;
880 		goto done_free;
881 	} else {
882 		ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
883 		if (ip6s == 0)
884 			pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
885 		else {
886 			pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
887 			ip6s /= sizeof(*ip6);
888 			if (ip6s > jail_max_af_ips) {
889 				error = EINVAL;
890 				vfs_opterror(opts, "too many IPv6 addresses");
891 				goto done_errmsg;
892 			}
893 			ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
894 			bcopy(op, ip6, ip6s * sizeof(*ip6));
895 			if (ip6s > 1)
896 				qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
897 			for (ii = 0; ii < ip6s; ii++) {
898 				if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
899 					error = EINVAL;
900 					goto done_free;
901 				}
902 				if ((ii+1) < ip6s &&
903 				    (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
904 				     IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
905 				{
906 					error = EINVAL;
907 					goto done_free;
908 				}
909 			}
910 		}
911 	}
912 #endif
913 
914 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
915 	if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
916 		error = EINVAL;
917 		vfs_opterror(opts,
918 		    "vnet jails cannot have IP address restrictions");
919 		goto done_errmsg;
920 	}
921 #endif
922 
923 	fullpath_disabled = 0;
924 	root = NULL;
925 	error = vfs_getopt(opts, "path", (void **)&path, &len);
926 	if (error == ENOENT)
927 		path = NULL;
928 	else if (error != 0)
929 		goto done_free;
930 	else {
931 		if (flags & JAIL_UPDATE) {
932 			error = EINVAL;
933 			vfs_opterror(opts,
934 			    "path cannot be changed after creation");
935 			goto done_errmsg;
936 		}
937 		if (len == 0 || path[len - 1] != '\0') {
938 			error = EINVAL;
939 			goto done_free;
940 		}
941 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
942 		    path, td);
943 		error = namei(&nd);
944 		if (error)
945 			goto done_free;
946 		root = nd.ni_vp;
947 		NDFREE(&nd, NDF_ONLY_PNBUF);
948 		g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
949 		strlcpy(g_path, path, MAXPATHLEN);
950 		error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
951 		if (error == 0)
952 			path = g_path;
953 		else if (error == ENODEV) {
954 			/* proceed if sysctl debug.disablefullpath == 1 */
955 			fullpath_disabled = 1;
956 			if (len < 2 || (len == 2 && path[0] == '/'))
957 				path = NULL;
958 		} else {
959 			/* exit on other errors */
960 			goto done_free;
961 		}
962 		if (root->v_type != VDIR) {
963 			error = ENOTDIR;
964 			vput(root);
965 			goto done_free;
966 		}
967 		VOP_UNLOCK(root, 0);
968 		if (fullpath_disabled) {
969 			/* Leave room for a real-root full pathname. */
970 			if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
971 			    ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
972 				error = ENAMETOOLONG;
973 				goto done_free;
974 			}
975 		}
976 	}
977 
978 	error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
979 	if (error == ENOENT)
980 		osrelstr = NULL;
981 	else if (error != 0)
982 		goto done_free;
983 	else {
984 		if (flags & JAIL_UPDATE) {
985 			error = EINVAL;
986 			vfs_opterror(opts,
987 			    "osrelease cannot be changed after creation");
988 			goto done_errmsg;
989 		}
990 		if (len == 0 || len >= OSRELEASELEN) {
991 			error = EINVAL;
992 			vfs_opterror(opts,
993 			    "osrelease string must be 1-%d bytes long",
994 			    OSRELEASELEN - 1);
995 			goto done_errmsg;
996 		}
997 	}
998 
999 	error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
1000 	if (error == ENOENT)
1001 		osreldt = 0;
1002 	else if (error != 0)
1003 		goto done_free;
1004 	else {
1005 		if (flags & JAIL_UPDATE) {
1006 			error = EINVAL;
1007 			vfs_opterror(opts,
1008 			    "osreldate cannot be changed after creation");
1009 			goto done_errmsg;
1010 		}
1011 		if (osreldt == 0) {
1012 			error = EINVAL;
1013 			vfs_opterror(opts, "osreldate cannot be 0");
1014 			goto done_errmsg;
1015 		}
1016 	}
1017 
1018 	/*
1019 	 * Grab the allprison lock before letting modules check their
1020 	 * parameters.  Once we have it, do not let go so we'll have a
1021 	 * consistent view of the OSD list.
1022 	 */
1023 	sx_xlock(&allprison_lock);
1024 	error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
1025 	if (error)
1026 		goto done_unlock_list;
1027 
1028 	/* By now, all parameters should have been noted. */
1029 	TAILQ_FOREACH(opt, opts, link) {
1030 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
1031 			error = EINVAL;
1032 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
1033 			goto done_unlock_list;
1034 		}
1035 	}
1036 
1037 	/*
1038 	 * See if we are creating a new record or updating an existing one.
1039 	 * This abuses the file error codes ENOENT and EEXIST.
1040 	 */
1041 	cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
1042 	if (!cuflags) {
1043 		error = EINVAL;
1044 		vfs_opterror(opts, "no valid operation (create or update)");
1045 		goto done_unlock_list;
1046 	}
1047 	pr = NULL;
1048 	namelc = NULL;
1049 	if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1050 		namelc = strrchr(name, '.');
1051 		jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1052 		if (*p != '\0')
1053 			jid = 0;
1054 	}
1055 	if (jid != 0) {
1056 		/*
1057 		 * See if a requested jid already exists.  There is an
1058 		 * information leak here if the jid exists but is not within
1059 		 * the caller's jail hierarchy.  Jail creators will get EEXIST
1060 		 * even though they cannot see the jail, and CREATE | UPDATE
1061 		 * will return ENOENT which is not normally a valid error.
1062 		 */
1063 		if (jid < 0) {
1064 			error = EINVAL;
1065 			vfs_opterror(opts, "negative jid");
1066 			goto done_unlock_list;
1067 		}
1068 		pr = prison_find(jid);
1069 		if (pr != NULL) {
1070 			ppr = pr->pr_parent;
1071 			/* Create: jid must not exist. */
1072 			if (cuflags == JAIL_CREATE) {
1073 				mtx_unlock(&pr->pr_mtx);
1074 				error = EEXIST;
1075 				vfs_opterror(opts, "jail %d already exists",
1076 				    jid);
1077 				goto done_unlock_list;
1078 			}
1079 			if (!prison_ischild(mypr, pr)) {
1080 				mtx_unlock(&pr->pr_mtx);
1081 				pr = NULL;
1082 			} else if (pr->pr_uref == 0) {
1083 				if (!(flags & JAIL_DYING)) {
1084 					mtx_unlock(&pr->pr_mtx);
1085 					error = ENOENT;
1086 					vfs_opterror(opts, "jail %d is dying",
1087 					    jid);
1088 					goto done_unlock_list;
1089 				} else if ((flags & JAIL_ATTACH) ||
1090 				    (pr_flags & PR_PERSIST)) {
1091 					/*
1092 					 * A dying jail might be resurrected
1093 					 * (via attach or persist), but first
1094 					 * it must determine if another jail
1095 					 * has claimed its name.  Accomplish
1096 					 * this by implicitly re-setting the
1097 					 * name.
1098 					 */
1099 					if (name == NULL)
1100 						name = prison_name(mypr, pr);
1101 				}
1102 			}
1103 		}
1104 		if (pr == NULL) {
1105 			/* Update: jid must exist. */
1106 			if (cuflags == JAIL_UPDATE) {
1107 				error = ENOENT;
1108 				vfs_opterror(opts, "jail %d not found", jid);
1109 				goto done_unlock_list;
1110 			}
1111 		}
1112 	}
1113 	/*
1114 	 * If the caller provided a name, look for a jail by that name.
1115 	 * This has different semantics for creates and updates keyed by jid
1116 	 * (where the name must not already exist in a different jail),
1117 	 * and updates keyed by the name itself (where the name must exist
1118 	 * because that is the jail being updated).
1119 	 */
1120 	if (name != NULL) {
1121 		namelc = strrchr(name, '.');
1122 		if (namelc == NULL)
1123 			namelc = name;
1124 		else {
1125 			/*
1126 			 * This is a hierarchical name.  Split it into the
1127 			 * parent and child names, and make sure the parent
1128 			 * exists or matches an already found jail.
1129 			 */
1130 			*namelc = '\0';
1131 			if (pr != NULL) {
1132 				if (strncmp(name, ppr->pr_name, namelc - name)
1133 				    || ppr->pr_name[namelc - name] != '\0') {
1134 					mtx_unlock(&pr->pr_mtx);
1135 					error = EINVAL;
1136 					vfs_opterror(opts,
1137 					    "cannot change jail's parent");
1138 					goto done_unlock_list;
1139 				}
1140 			} else {
1141 				ppr = prison_find_name(mypr, name);
1142 				if (ppr == NULL) {
1143 					error = ENOENT;
1144 					vfs_opterror(opts,
1145 					    "jail \"%s\" not found", name);
1146 					goto done_unlock_list;
1147 				}
1148 				mtx_unlock(&ppr->pr_mtx);
1149 			}
1150 			name = ++namelc;
1151 		}
1152 		if (name[0] != '\0') {
1153 			namelen =
1154 			    (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1155  name_again:
1156 			deadpr = NULL;
1157 			FOREACH_PRISON_CHILD(ppr, tpr) {
1158 				if (tpr != pr && tpr->pr_ref > 0 &&
1159 				    !strcmp(tpr->pr_name + namelen, name)) {
1160 					if (pr == NULL &&
1161 					    cuflags != JAIL_CREATE) {
1162 						mtx_lock(&tpr->pr_mtx);
1163 						if (tpr->pr_ref > 0) {
1164 							/*
1165 							 * Use this jail
1166 							 * for updates.
1167 							 */
1168 							if (tpr->pr_uref > 0) {
1169 								pr = tpr;
1170 								break;
1171 							}
1172 							deadpr = tpr;
1173 						}
1174 						mtx_unlock(&tpr->pr_mtx);
1175 					} else if (tpr->pr_uref > 0) {
1176 						/*
1177 						 * Create, or update(jid):
1178 						 * name must not exist in an
1179 						 * active sibling jail.
1180 						 */
1181 						error = EEXIST;
1182 						if (pr != NULL)
1183 							mtx_unlock(&pr->pr_mtx);
1184 						vfs_opterror(opts,
1185 						   "jail \"%s\" already exists",
1186 						   name);
1187 						goto done_unlock_list;
1188 					}
1189 				}
1190 			}
1191 			/* If no active jail is found, use a dying one. */
1192 			if (deadpr != NULL && pr == NULL) {
1193 				if (flags & JAIL_DYING) {
1194 					mtx_lock(&deadpr->pr_mtx);
1195 					if (deadpr->pr_ref == 0) {
1196 						mtx_unlock(&deadpr->pr_mtx);
1197 						goto name_again;
1198 					}
1199 					pr = deadpr;
1200 				} else if (cuflags == JAIL_UPDATE) {
1201 					error = ENOENT;
1202 					vfs_opterror(opts,
1203 					    "jail \"%s\" is dying", name);
1204 					goto done_unlock_list;
1205 				}
1206 			}
1207 			/* Update: name must exist if no jid. */
1208 			else if (cuflags == JAIL_UPDATE && pr == NULL) {
1209 				error = ENOENT;
1210 				vfs_opterror(opts, "jail \"%s\" not found",
1211 				    name);
1212 				goto done_unlock_list;
1213 			}
1214 		}
1215 	}
1216 	/* Update: must provide a jid or name. */
1217 	else if (cuflags == JAIL_UPDATE && pr == NULL) {
1218 		error = ENOENT;
1219 		vfs_opterror(opts, "update specified no jail");
1220 		goto done_unlock_list;
1221 	}
1222 
1223 	/* If there's no prison to update, create a new one and link it in. */
1224 	if (pr == NULL) {
1225 		for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1226 			if (tpr->pr_childcount >= tpr->pr_childmax) {
1227 				error = EPERM;
1228 				vfs_opterror(opts, "prison limit exceeded");
1229 				goto done_unlock_list;
1230 			}
1231 		created = 1;
1232 		mtx_lock(&ppr->pr_mtx);
1233 		if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1234 			mtx_unlock(&ppr->pr_mtx);
1235 			error = ENOENT;
1236 			vfs_opterror(opts, "parent jail went away!");
1237 			goto done_unlock_list;
1238 		}
1239 		ppr->pr_ref++;
1240 		ppr->pr_uref++;
1241 		mtx_unlock(&ppr->pr_mtx);
1242 		pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1243 		if (jid == 0) {
1244 			/* Find the next free jid. */
1245 			jid = lastprid + 1;
1246  findnext:
1247 			if (jid == JAIL_MAX)
1248 				jid = 1;
1249 			TAILQ_FOREACH(tpr, &allprison, pr_list) {
1250 				if (tpr->pr_id < jid)
1251 					continue;
1252 				if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1253 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1254 					break;
1255 				}
1256 				if (jid == lastprid) {
1257 					error = EAGAIN;
1258 					vfs_opterror(opts,
1259 					    "no available jail IDs");
1260 					free(pr, M_PRISON);
1261 					prison_deref(ppr, PD_DEREF |
1262 					    PD_DEUREF | PD_LIST_XLOCKED);
1263 					goto done_releroot;
1264 				}
1265 				jid++;
1266 				goto findnext;
1267 			}
1268 			lastprid = jid;
1269 		} else {
1270 			/*
1271 			 * The jail already has a jid (that did not yet exist),
1272 			 * so just find where to insert it.
1273 			 */
1274 			TAILQ_FOREACH(tpr, &allprison, pr_list)
1275 				if (tpr->pr_id >= jid) {
1276 					TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1277 					break;
1278 				}
1279 		}
1280 		if (tpr == NULL)
1281 			TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1282 		LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1283 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1284 			tpr->pr_childcount++;
1285 
1286 		pr->pr_parent = ppr;
1287 		pr->pr_id = jid;
1288 
1289 		/* Set some default values, and inherit some from the parent. */
1290 		if (name == NULL)
1291 			name = "";
1292 		if (path == NULL) {
1293 			path = "/";
1294 			root = mypr->pr_root;
1295 			vref(root);
1296 		}
1297 		strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1298 		pr->pr_flags |= PR_HOST;
1299 #if defined(INET) || defined(INET6)
1300 #ifdef VIMAGE
1301 		if (!(pr_flags & PR_VNET))
1302 #endif
1303 		{
1304 #ifdef INET
1305 			if (!(ch_flags & PR_IP4_USER))
1306 				pr->pr_flags |=
1307 				    PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1308 			else if (!(pr_flags & PR_IP4_USER)) {
1309 				pr->pr_flags |= ppr->pr_flags & PR_IP4;
1310 				if (ppr->pr_ip4 != NULL) {
1311 					pr->pr_ip4s = ppr->pr_ip4s;
1312 					pr->pr_ip4 = malloc(pr->pr_ip4s *
1313 					    sizeof(struct in_addr), M_PRISON,
1314 					    M_WAITOK);
1315 					bcopy(ppr->pr_ip4, pr->pr_ip4,
1316 					    pr->pr_ip4s * sizeof(*pr->pr_ip4));
1317 				}
1318 			}
1319 #endif
1320 #ifdef INET6
1321 			if (!(ch_flags & PR_IP6_USER))
1322 				pr->pr_flags |=
1323 				    PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1324 			else if (!(pr_flags & PR_IP6_USER)) {
1325 				pr->pr_flags |= ppr->pr_flags & PR_IP6;
1326 				if (ppr->pr_ip6 != NULL) {
1327 					pr->pr_ip6s = ppr->pr_ip6s;
1328 					pr->pr_ip6 = malloc(pr->pr_ip6s *
1329 					    sizeof(struct in6_addr), M_PRISON,
1330 					    M_WAITOK);
1331 					bcopy(ppr->pr_ip6, pr->pr_ip6,
1332 					    pr->pr_ip6s * sizeof(*pr->pr_ip6));
1333 				}
1334 			}
1335 #endif
1336 		}
1337 #endif
1338 		/* Source address selection is always on by default. */
1339 		pr->pr_flags |= _PR_IP_SADDRSEL;
1340 
1341 		pr->pr_securelevel = ppr->pr_securelevel;
1342 		pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1343 		pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1344 		pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1345 
1346 		pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1347 		if (osrelstr == NULL)
1348 		    strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1349 		else
1350 		    strcpy(pr->pr_osrelease, osrelstr);
1351 
1352 		LIST_INIT(&pr->pr_children);
1353 		mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1354 
1355 #ifdef VIMAGE
1356 		/* Allocate a new vnet if specified. */
1357 		pr->pr_vnet = (pr_flags & PR_VNET)
1358 		    ? vnet_alloc() : ppr->pr_vnet;
1359 #endif
1360 		/*
1361 		 * Allocate a dedicated cpuset for each jail.
1362 		 * Unlike other initial settings, this may return an erorr.
1363 		 */
1364 		error = cpuset_create_root(ppr, &pr->pr_cpuset);
1365 		if (error) {
1366 			prison_deref(pr, PD_LIST_XLOCKED);
1367 			goto done_releroot;
1368 		}
1369 
1370 		mtx_lock(&pr->pr_mtx);
1371 		/*
1372 		 * New prisons do not yet have a reference, because we do not
1373 		 * want other to see the incomplete prison once the
1374 		 * allprison_lock is downgraded.
1375 		 */
1376 	} else {
1377 		created = 0;
1378 		/*
1379 		 * Grab a reference for existing prisons, to ensure they
1380 		 * continue to exist for the duration of the call.
1381 		 */
1382 		pr->pr_ref++;
1383 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1384 		if ((pr->pr_flags & PR_VNET) &&
1385 		    (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1386 			error = EINVAL;
1387 			vfs_opterror(opts,
1388 			    "vnet jails cannot have IP address restrictions");
1389 			goto done_deref_locked;
1390 		}
1391 #endif
1392 #ifdef INET
1393 		if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1394 			error = EINVAL;
1395 			vfs_opterror(opts,
1396 			    "ip4 cannot be changed after creation");
1397 			goto done_deref_locked;
1398 		}
1399 #endif
1400 #ifdef INET6
1401 		if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1402 			error = EINVAL;
1403 			vfs_opterror(opts,
1404 			    "ip6 cannot be changed after creation");
1405 			goto done_deref_locked;
1406 		}
1407 #endif
1408 	}
1409 
1410 	/* Do final error checking before setting anything. */
1411 	if (gotslevel) {
1412 		if (slevel < ppr->pr_securelevel) {
1413 			error = EPERM;
1414 			goto done_deref_locked;
1415 		}
1416 	}
1417 	if (gotchildmax) {
1418 		if (childmax >= ppr->pr_childmax) {
1419 			error = EPERM;
1420 			goto done_deref_locked;
1421 		}
1422 	}
1423 	if (gotenforce) {
1424 		if (enforce < ppr->pr_enforce_statfs) {
1425 			error = EPERM;
1426 			goto done_deref_locked;
1427 		}
1428 	}
1429 	if (gotrsnum) {
1430 		/*
1431 		 * devfs_rsnum is a uint16_t
1432 		 */
1433 		if (rsnum < 0 || rsnum > 65535) {
1434 			error = EINVAL;
1435 			goto done_deref_locked;
1436 		}
1437 		/*
1438 		 * Nested jails always inherit parent's devfs ruleset
1439 		 */
1440 		if (jailed(td->td_ucred)) {
1441 			if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1442 				error = EPERM;
1443 				goto done_deref_locked;
1444 			} else
1445 				rsnum = ppr->pr_devfs_rsnum;
1446 		}
1447 	}
1448 #ifdef INET
1449 	if (ip4s > 0) {
1450 		if (ppr->pr_flags & PR_IP4) {
1451 			/*
1452 			 * Make sure the new set of IP addresses is a
1453 			 * subset of the parent's list.  Don't worry
1454 			 * about the parent being unlocked, as any
1455 			 * setting is done with allprison_lock held.
1456 			 */
1457 			for (ij = 0; ij < ppr->pr_ip4s; ij++)
1458 				if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1459 					break;
1460 			if (ij == ppr->pr_ip4s) {
1461 				error = EPERM;
1462 				goto done_deref_locked;
1463 			}
1464 			if (ip4s > 1) {
1465 				for (ii = ij = 1; ii < ip4s; ii++) {
1466 					if (ip4[ii].s_addr ==
1467 					    ppr->pr_ip4[0].s_addr)
1468 						continue;
1469 					for (; ij < ppr->pr_ip4s; ij++)
1470 						if (ip4[ii].s_addr ==
1471 						    ppr->pr_ip4[ij].s_addr)
1472 							break;
1473 					if (ij == ppr->pr_ip4s)
1474 						break;
1475 				}
1476 				if (ij == ppr->pr_ip4s) {
1477 					error = EPERM;
1478 					goto done_deref_locked;
1479 				}
1480 			}
1481 		}
1482 		/*
1483 		 * Check for conflicting IP addresses.  We permit them
1484 		 * if there is no more than one IP on each jail.  If
1485 		 * there is a duplicate on a jail with more than one
1486 		 * IP stop checking and return error.
1487 		 */
1488 		tppr = ppr;
1489 #ifdef VIMAGE
1490 		for (; tppr != &prison0; tppr = tppr->pr_parent)
1491 			if (tppr->pr_flags & PR_VNET)
1492 				break;
1493 #endif
1494 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1495 			if (tpr == pr ||
1496 #ifdef VIMAGE
1497 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1498 #endif
1499 			    tpr->pr_uref == 0) {
1500 				descend = 0;
1501 				continue;
1502 			}
1503 			if (!(tpr->pr_flags & PR_IP4_USER))
1504 				continue;
1505 			descend = 0;
1506 			if (tpr->pr_ip4 == NULL ||
1507 			    (ip4s == 1 && tpr->pr_ip4s == 1))
1508 				continue;
1509 			for (ii = 0; ii < ip4s; ii++) {
1510 				if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1511 					error = EADDRINUSE;
1512 					vfs_opterror(opts,
1513 					    "IPv4 addresses clash");
1514 					goto done_deref_locked;
1515 				}
1516 			}
1517 		}
1518 	}
1519 #endif
1520 #ifdef INET6
1521 	if (ip6s > 0) {
1522 		if (ppr->pr_flags & PR_IP6) {
1523 			/*
1524 			 * Make sure the new set of IP addresses is a
1525 			 * subset of the parent's list.
1526 			 */
1527 			for (ij = 0; ij < ppr->pr_ip6s; ij++)
1528 				if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1529 				    &ppr->pr_ip6[ij]))
1530 					break;
1531 			if (ij == ppr->pr_ip6s) {
1532 				error = EPERM;
1533 				goto done_deref_locked;
1534 			}
1535 			if (ip6s > 1) {
1536 				for (ii = ij = 1; ii < ip6s; ii++) {
1537 					if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1538 					     &ppr->pr_ip6[0]))
1539 						continue;
1540 					for (; ij < ppr->pr_ip6s; ij++)
1541 						if (IN6_ARE_ADDR_EQUAL(
1542 						    &ip6[ii], &ppr->pr_ip6[ij]))
1543 							break;
1544 					if (ij == ppr->pr_ip6s)
1545 						break;
1546 				}
1547 				if (ij == ppr->pr_ip6s) {
1548 					error = EPERM;
1549 					goto done_deref_locked;
1550 				}
1551 			}
1552 		}
1553 		/* Check for conflicting IP addresses. */
1554 		tppr = ppr;
1555 #ifdef VIMAGE
1556 		for (; tppr != &prison0; tppr = tppr->pr_parent)
1557 			if (tppr->pr_flags & PR_VNET)
1558 				break;
1559 #endif
1560 		FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1561 			if (tpr == pr ||
1562 #ifdef VIMAGE
1563 			    (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1564 #endif
1565 			    tpr->pr_uref == 0) {
1566 				descend = 0;
1567 				continue;
1568 			}
1569 			if (!(tpr->pr_flags & PR_IP6_USER))
1570 				continue;
1571 			descend = 0;
1572 			if (tpr->pr_ip6 == NULL ||
1573 			    (ip6s == 1 && tpr->pr_ip6s == 1))
1574 				continue;
1575 			for (ii = 0; ii < ip6s; ii++) {
1576 				if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1577 					error = EADDRINUSE;
1578 					vfs_opterror(opts,
1579 					    "IPv6 addresses clash");
1580 					goto done_deref_locked;
1581 				}
1582 			}
1583 		}
1584 	}
1585 #endif
1586 	onamelen = namelen = 0;
1587 	if (name != NULL) {
1588 		/* Give a default name of the jid. */
1589 		if (name[0] == '\0')
1590 			snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1591 		else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1592 		    *p == '\0')) {
1593 			error = EINVAL;
1594 			vfs_opterror(opts,
1595 			    "name cannot be numeric (unless it is the jid)");
1596 			goto done_deref_locked;
1597 		}
1598 		/*
1599 		 * Make sure the name isn't too long for the prison or its
1600 		 * children.
1601 		 */
1602 		onamelen = strlen(pr->pr_name);
1603 		namelen = strlen(name);
1604 		if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1605 			error = ENAMETOOLONG;
1606 			goto done_deref_locked;
1607 		}
1608 		FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1609 			if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1610 			    sizeof(pr->pr_name)) {
1611 				error = ENAMETOOLONG;
1612 				goto done_deref_locked;
1613 			}
1614 		}
1615 	}
1616 	if (pr_allow & ~ppr->pr_allow) {
1617 		error = EPERM;
1618 		goto done_deref_locked;
1619 	}
1620 
1621 	/* Set the parameters of the prison. */
1622 #ifdef INET
1623 	redo_ip4 = 0;
1624 	if (pr_flags & PR_IP4_USER) {
1625 		pr->pr_flags |= PR_IP4;
1626 		free(pr->pr_ip4, M_PRISON);
1627 		pr->pr_ip4s = ip4s;
1628 		pr->pr_ip4 = ip4;
1629 		ip4 = NULL;
1630 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1631 #ifdef VIMAGE
1632 			if (tpr->pr_flags & PR_VNET) {
1633 				descend = 0;
1634 				continue;
1635 			}
1636 #endif
1637 			if (prison_restrict_ip4(tpr, NULL)) {
1638 				redo_ip4 = 1;
1639 				descend = 0;
1640 			}
1641 		}
1642 	}
1643 #endif
1644 #ifdef INET6
1645 	redo_ip6 = 0;
1646 	if (pr_flags & PR_IP6_USER) {
1647 		pr->pr_flags |= PR_IP6;
1648 		free(pr->pr_ip6, M_PRISON);
1649 		pr->pr_ip6s = ip6s;
1650 		pr->pr_ip6 = ip6;
1651 		ip6 = NULL;
1652 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1653 #ifdef VIMAGE
1654 			if (tpr->pr_flags & PR_VNET) {
1655 				descend = 0;
1656 				continue;
1657 			}
1658 #endif
1659 			if (prison_restrict_ip6(tpr, NULL)) {
1660 				redo_ip6 = 1;
1661 				descend = 0;
1662 			}
1663 		}
1664 	}
1665 #endif
1666 	if (gotslevel) {
1667 		pr->pr_securelevel = slevel;
1668 		/* Set all child jails to be at least this level. */
1669 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1670 			if (tpr->pr_securelevel < slevel)
1671 				tpr->pr_securelevel = slevel;
1672 	}
1673 	if (gotchildmax) {
1674 		pr->pr_childmax = childmax;
1675 		/* Set all child jails to under this limit. */
1676 		FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1677 			if (tpr->pr_childmax > childmax - level)
1678 				tpr->pr_childmax = childmax > level
1679 				    ? childmax - level : 0;
1680 	}
1681 	if (gotenforce) {
1682 		pr->pr_enforce_statfs = enforce;
1683 		/* Pass this restriction on to the children. */
1684 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1685 			if (tpr->pr_enforce_statfs < enforce)
1686 				tpr->pr_enforce_statfs = enforce;
1687 	}
1688 	if (gotrsnum) {
1689 		pr->pr_devfs_rsnum = rsnum;
1690 		/* Pass this restriction on to the children. */
1691 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1692 			tpr->pr_devfs_rsnum = rsnum;
1693 	}
1694 	if (name != NULL) {
1695 		if (ppr == &prison0)
1696 			strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1697 		else
1698 			snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1699 			    ppr->pr_name, name);
1700 		/* Change this component of child names. */
1701 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1702 			bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1703 			    strlen(tpr->pr_name + onamelen) + 1);
1704 			bcopy(pr->pr_name, tpr->pr_name, namelen);
1705 		}
1706 	}
1707 	if (path != NULL) {
1708 		/* Try to keep a real-rooted full pathname. */
1709 		if (fullpath_disabled && path[0] == '/' &&
1710 		    strcmp(mypr->pr_path, "/"))
1711 			snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1712 			    mypr->pr_path, path);
1713 		else
1714 			strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1715 		pr->pr_root = root;
1716 	}
1717 	if (PR_HOST & ch_flags & ~pr_flags) {
1718 		if (pr->pr_flags & PR_HOST) {
1719 			/*
1720 			 * Copy the parent's host info.  As with pr_ip4 above,
1721 			 * the lack of a lock on the parent is not a problem;
1722 			 * it is always set with allprison_lock at least
1723 			 * shared, and is held exclusively here.
1724 			 */
1725 			strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1726 			    sizeof(pr->pr_hostname));
1727 			strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1728 			    sizeof(pr->pr_domainname));
1729 			strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1730 			    sizeof(pr->pr_hostuuid));
1731 			pr->pr_hostid = pr->pr_parent->pr_hostid;
1732 		}
1733 	} else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1734 		/* Set this prison, and any descendants without PR_HOST. */
1735 		if (host != NULL)
1736 			strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1737 		if (domain != NULL)
1738 			strlcpy(pr->pr_domainname, domain,
1739 			    sizeof(pr->pr_domainname));
1740 		if (uuid != NULL)
1741 			strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1742 		if (gothid)
1743 			pr->pr_hostid = hid;
1744 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1745 			if (tpr->pr_flags & PR_HOST)
1746 				descend = 0;
1747 			else {
1748 				if (host != NULL)
1749 					strlcpy(tpr->pr_hostname,
1750 					    pr->pr_hostname,
1751 					    sizeof(tpr->pr_hostname));
1752 				if (domain != NULL)
1753 					strlcpy(tpr->pr_domainname,
1754 					    pr->pr_domainname,
1755 					    sizeof(tpr->pr_domainname));
1756 				if (uuid != NULL)
1757 					strlcpy(tpr->pr_hostuuid,
1758 					    pr->pr_hostuuid,
1759 					    sizeof(tpr->pr_hostuuid));
1760 				if (gothid)
1761 					tpr->pr_hostid = hid;
1762 			}
1763 		}
1764 	}
1765 	if ((tallow = ch_allow & ~pr_allow)) {
1766 		/* Clear allow bits in all children. */
1767 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1768 			tpr->pr_allow &= ~tallow;
1769 	}
1770 	pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1771 	/*
1772 	 * Persistent prisons get an extra reference, and prisons losing their
1773 	 * persist flag lose that reference.  Only do this for existing prisons
1774 	 * for now, so new ones will remain unseen until after the module
1775 	 * handlers have completed.
1776 	 */
1777 	if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1778 		if (pr_flags & PR_PERSIST) {
1779 			pr->pr_ref++;
1780 			pr->pr_uref++;
1781 		} else {
1782 			pr->pr_ref--;
1783 			pr->pr_uref--;
1784 		}
1785 	}
1786 	pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1787 	mtx_unlock(&pr->pr_mtx);
1788 
1789 #ifdef RACCT
1790 	if (created)
1791 		prison_racct_attach(pr);
1792 #endif
1793 
1794 	/* Locks may have prevented a complete restriction of child IP
1795 	 * addresses.  If so, allocate some more memory and try again.
1796 	 */
1797 #ifdef INET
1798 	while (redo_ip4) {
1799 		ip4s = pr->pr_ip4s;
1800 		ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1801 		mtx_lock(&pr->pr_mtx);
1802 		redo_ip4 = 0;
1803 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1804 #ifdef VIMAGE
1805 			if (tpr->pr_flags & PR_VNET) {
1806 				descend = 0;
1807 				continue;
1808 			}
1809 #endif
1810 			if (prison_restrict_ip4(tpr, ip4)) {
1811 				if (ip4 != NULL)
1812 					ip4 = NULL;
1813 				else
1814 					redo_ip4 = 1;
1815 			}
1816 		}
1817 		mtx_unlock(&pr->pr_mtx);
1818 	}
1819 #endif
1820 #ifdef INET6
1821 	while (redo_ip6) {
1822 		ip6s = pr->pr_ip6s;
1823 		ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1824 		mtx_lock(&pr->pr_mtx);
1825 		redo_ip6 = 0;
1826 		FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1827 #ifdef VIMAGE
1828 			if (tpr->pr_flags & PR_VNET) {
1829 				descend = 0;
1830 				continue;
1831 			}
1832 #endif
1833 			if (prison_restrict_ip6(tpr, ip6)) {
1834 				if (ip6 != NULL)
1835 					ip6 = NULL;
1836 				else
1837 					redo_ip6 = 1;
1838 			}
1839 		}
1840 		mtx_unlock(&pr->pr_mtx);
1841 	}
1842 #endif
1843 
1844 	/* Let the modules do their work. */
1845 	sx_downgrade(&allprison_lock);
1846 	if (created) {
1847 		error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1848 		if (error) {
1849 			prison_deref(pr, PD_LIST_SLOCKED);
1850 			goto done_errmsg;
1851 		}
1852 	}
1853 	error = osd_jail_call(pr, PR_METHOD_SET, opts);
1854 	if (error) {
1855 		prison_deref(pr, created
1856 		    ? PD_LIST_SLOCKED
1857 		    : PD_DEREF | PD_LIST_SLOCKED);
1858 		goto done_errmsg;
1859 	}
1860 
1861 	/* Attach this process to the prison if requested. */
1862 	if (flags & JAIL_ATTACH) {
1863 		mtx_lock(&pr->pr_mtx);
1864 		error = do_jail_attach(td, pr);
1865 		if (error) {
1866 			vfs_opterror(opts, "attach failed");
1867 			if (!created)
1868 				prison_deref(pr, PD_DEREF);
1869 			goto done_errmsg;
1870 		}
1871 	}
1872 
1873 #ifdef RACCT
1874 	if (!created) {
1875 		if (!(flags & JAIL_ATTACH))
1876 			sx_sunlock(&allprison_lock);
1877 		prison_racct_modify(pr);
1878 		if (!(flags & JAIL_ATTACH))
1879 			sx_slock(&allprison_lock);
1880 	}
1881 #endif
1882 
1883 	td->td_retval[0] = pr->pr_id;
1884 
1885 	/*
1886 	 * Now that it is all there, drop the temporary reference from existing
1887 	 * prisons.  Or add a reference to newly created persistent prisons
1888 	 * (which was not done earlier so that the prison would not be publicly
1889 	 * visible).
1890 	 */
1891 	if (!created) {
1892 		prison_deref(pr, (flags & JAIL_ATTACH)
1893 		    ? PD_DEREF
1894 		    : PD_DEREF | PD_LIST_SLOCKED);
1895 	} else {
1896 		if (pr_flags & PR_PERSIST) {
1897 			mtx_lock(&pr->pr_mtx);
1898 			pr->pr_ref++;
1899 			pr->pr_uref++;
1900 			mtx_unlock(&pr->pr_mtx);
1901 		}
1902 		if (!(flags & JAIL_ATTACH))
1903 			sx_sunlock(&allprison_lock);
1904 	}
1905 
1906 	goto done_errmsg;
1907 
1908  done_deref_locked:
1909 	prison_deref(pr, created
1910 	    ? PD_LOCKED | PD_LIST_XLOCKED
1911 	    : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1912 	goto done_releroot;
1913  done_unlock_list:
1914 	sx_xunlock(&allprison_lock);
1915  done_releroot:
1916 	if (root != NULL)
1917 		vrele(root);
1918  done_errmsg:
1919 	if (error) {
1920 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1921 		if (errmsg_len > 0) {
1922 			errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1923 			if (errmsg_pos > 0) {
1924 				if (optuio->uio_segflg == UIO_SYSSPACE)
1925 					bcopy(errmsg,
1926 					   optuio->uio_iov[errmsg_pos].iov_base,
1927 					   errmsg_len);
1928 				else
1929 					copyout(errmsg,
1930 					   optuio->uio_iov[errmsg_pos].iov_base,
1931 					   errmsg_len);
1932 			}
1933 		}
1934 	}
1935  done_free:
1936 #ifdef INET
1937 	free(ip4, M_PRISON);
1938 #endif
1939 #ifdef INET6
1940 	free(ip6, M_PRISON);
1941 #endif
1942 	if (g_path != NULL)
1943 		free(g_path, M_TEMP);
1944 	vfs_freeopts(opts);
1945 	return (error);
1946 }
1947 
1948 
1949 /*
1950  * struct jail_get_args {
1951  *	struct iovec *iovp;
1952  *	unsigned int iovcnt;
1953  *	int flags;
1954  * };
1955  */
1956 int
sys_jail_get(struct thread * td,struct jail_get_args * uap)1957 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1958 {
1959 	struct uio *auio;
1960 	int error;
1961 
1962 	/* Check that we have an even number of iovecs. */
1963 	if (uap->iovcnt & 1)
1964 		return (EINVAL);
1965 
1966 	error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1967 	if (error)
1968 		return (error);
1969 	error = kern_jail_get(td, auio, uap->flags);
1970 	if (error == 0)
1971 		error = copyout(auio->uio_iov, uap->iovp,
1972 		    uap->iovcnt * sizeof (struct iovec));
1973 	free(auio, M_IOV);
1974 	return (error);
1975 }
1976 
1977 int
kern_jail_get(struct thread * td,struct uio * optuio,int flags)1978 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1979 {
1980 	struct prison *pr, *mypr;
1981 	struct vfsopt *opt;
1982 	struct vfsoptlist *opts;
1983 	char *errmsg, *name;
1984 	int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1985 
1986 	if (flags & ~JAIL_GET_MASK)
1987 		return (EINVAL);
1988 
1989 	/* Get the parameter list. */
1990 	error = vfs_buildopts(optuio, &opts);
1991 	if (error)
1992 		return (error);
1993 	errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1994 	mypr = td->td_ucred->cr_prison;
1995 
1996 	/*
1997 	 * Find the prison specified by one of: lastjid, jid, name.
1998 	 */
1999 	sx_slock(&allprison_lock);
2000 	error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2001 	if (error == 0) {
2002 		TAILQ_FOREACH(pr, &allprison, pr_list) {
2003 			if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2004 				mtx_lock(&pr->pr_mtx);
2005 				if (pr->pr_ref > 0 &&
2006 				    (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2007 					break;
2008 				mtx_unlock(&pr->pr_mtx);
2009 			}
2010 		}
2011 		if (pr != NULL)
2012 			goto found_prison;
2013 		error = ENOENT;
2014 		vfs_opterror(opts, "no jail after %d", jid);
2015 		goto done_unlock_list;
2016 	} else if (error != ENOENT)
2017 		goto done_unlock_list;
2018 
2019 	error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2020 	if (error == 0) {
2021 		if (jid != 0) {
2022 			pr = prison_find_child(mypr, jid);
2023 			if (pr != NULL) {
2024 				if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2025 					mtx_unlock(&pr->pr_mtx);
2026 					error = ENOENT;
2027 					vfs_opterror(opts, "jail %d is dying",
2028 					    jid);
2029 					goto done_unlock_list;
2030 				}
2031 				goto found_prison;
2032 			}
2033 			error = ENOENT;
2034 			vfs_opterror(opts, "jail %d not found", jid);
2035 			goto done_unlock_list;
2036 		}
2037 	} else if (error != ENOENT)
2038 		goto done_unlock_list;
2039 
2040 	error = vfs_getopt(opts, "name", (void **)&name, &len);
2041 	if (error == 0) {
2042 		if (len == 0 || name[len - 1] != '\0') {
2043 			error = EINVAL;
2044 			goto done_unlock_list;
2045 		}
2046 		pr = prison_find_name(mypr, name);
2047 		if (pr != NULL) {
2048 			if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2049 				mtx_unlock(&pr->pr_mtx);
2050 				error = ENOENT;
2051 				vfs_opterror(opts, "jail \"%s\" is dying",
2052 				    name);
2053 				goto done_unlock_list;
2054 			}
2055 			goto found_prison;
2056 		}
2057 		error = ENOENT;
2058 		vfs_opterror(opts, "jail \"%s\" not found", name);
2059 		goto done_unlock_list;
2060 	} else if (error != ENOENT)
2061 		goto done_unlock_list;
2062 
2063 	vfs_opterror(opts, "no jail specified");
2064 	error = ENOENT;
2065 	goto done_unlock_list;
2066 
2067  found_prison:
2068 	/* Get the parameters of the prison. */
2069 	pr->pr_ref++;
2070 	locked = PD_LOCKED;
2071 	td->td_retval[0] = pr->pr_id;
2072 	error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2073 	if (error != 0 && error != ENOENT)
2074 		goto done_deref;
2075 	i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2076 	error = vfs_setopt(opts, "parent", &i, sizeof(i));
2077 	if (error != 0 && error != ENOENT)
2078 		goto done_deref;
2079 	error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2080 	if (error != 0 && error != ENOENT)
2081 		goto done_deref;
2082 	error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2083 	    sizeof(pr->pr_cpuset->cs_id));
2084 	if (error != 0 && error != ENOENT)
2085 		goto done_deref;
2086 	error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2087 	if (error != 0 && error != ENOENT)
2088 		goto done_deref;
2089 #ifdef INET
2090 	error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2091 	    pr->pr_ip4s * sizeof(*pr->pr_ip4));
2092 	if (error != 0 && error != ENOENT)
2093 		goto done_deref;
2094 #endif
2095 #ifdef INET6
2096 	error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2097 	    pr->pr_ip6s * sizeof(*pr->pr_ip6));
2098 	if (error != 0 && error != ENOENT)
2099 		goto done_deref;
2100 #endif
2101 	error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2102 	    sizeof(pr->pr_securelevel));
2103 	if (error != 0 && error != ENOENT)
2104 		goto done_deref;
2105 	error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2106 	    sizeof(pr->pr_childcount));
2107 	if (error != 0 && error != ENOENT)
2108 		goto done_deref;
2109 	error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2110 	    sizeof(pr->pr_childmax));
2111 	if (error != 0 && error != ENOENT)
2112 		goto done_deref;
2113 	error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2114 	if (error != 0 && error != ENOENT)
2115 		goto done_deref;
2116 	error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2117 	if (error != 0 && error != ENOENT)
2118 		goto done_deref;
2119 	error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2120 	if (error != 0 && error != ENOENT)
2121 		goto done_deref;
2122 #ifdef COMPAT_FREEBSD32
2123 	if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2124 		uint32_t hid32 = pr->pr_hostid;
2125 
2126 		error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2127 	} else
2128 #endif
2129 	error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2130 	    sizeof(pr->pr_hostid));
2131 	if (error != 0 && error != ENOENT)
2132 		goto done_deref;
2133 	error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2134 	    sizeof(pr->pr_enforce_statfs));
2135 	if (error != 0 && error != ENOENT)
2136 		goto done_deref;
2137 	error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2138 	    sizeof(pr->pr_devfs_rsnum));
2139 	if (error != 0 && error != ENOENT)
2140 		goto done_deref;
2141 	for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2142 	    fi++) {
2143 		if (pr_flag_names[fi] == NULL)
2144 			continue;
2145 		i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2146 		error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2147 		if (error != 0 && error != ENOENT)
2148 			goto done_deref;
2149 		i = !i;
2150 		error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2151 		if (error != 0 && error != ENOENT)
2152 			goto done_deref;
2153 	}
2154 	for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2155 	    fi++) {
2156 		i = pr->pr_flags &
2157 		    (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2158 		i = pr_flag_jailsys[fi].disable &&
2159 		      (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2160 		    : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2161 		    : JAIL_SYS_INHERIT;
2162 		error =
2163 		    vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2164 		if (error != 0 && error != ENOENT)
2165 			goto done_deref;
2166 	}
2167 	for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2168 	    fi++) {
2169 		if (pr_allow_names[fi] == NULL)
2170 			continue;
2171 		i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2172 		error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2173 		if (error != 0 && error != ENOENT)
2174 			goto done_deref;
2175 		i = !i;
2176 		error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2177 		if (error != 0 && error != ENOENT)
2178 			goto done_deref;
2179 	}
2180 	i = (pr->pr_uref == 0);
2181 	error = vfs_setopt(opts, "dying", &i, sizeof(i));
2182 	if (error != 0 && error != ENOENT)
2183 		goto done_deref;
2184 	i = !i;
2185 	error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2186 	if (error != 0 && error != ENOENT)
2187 		goto done_deref;
2188 	error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2189 	    sizeof(pr->pr_osreldate));
2190 	if (error != 0 && error != ENOENT)
2191 		goto done_deref;
2192 	error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2193 	if (error != 0 && error != ENOENT)
2194 		goto done_deref;
2195 
2196 	/* Get the module parameters. */
2197 	mtx_unlock(&pr->pr_mtx);
2198 	locked = 0;
2199 	error = osd_jail_call(pr, PR_METHOD_GET, opts);
2200 	if (error)
2201 		goto done_deref;
2202 	prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2203 
2204 	/* By now, all parameters should have been noted. */
2205 	TAILQ_FOREACH(opt, opts, link) {
2206 		if (!opt->seen && strcmp(opt->name, "errmsg")) {
2207 			error = EINVAL;
2208 			vfs_opterror(opts, "unknown parameter: %s", opt->name);
2209 			goto done_errmsg;
2210 		}
2211 	}
2212 
2213 	/* Write the fetched parameters back to userspace. */
2214 	error = 0;
2215 	TAILQ_FOREACH(opt, opts, link) {
2216 		if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2217 			pos = 2 * opt->pos + 1;
2218 			optuio->uio_iov[pos].iov_len = opt->len;
2219 			if (opt->value != NULL) {
2220 				if (optuio->uio_segflg == UIO_SYSSPACE) {
2221 					bcopy(opt->value,
2222 					    optuio->uio_iov[pos].iov_base,
2223 					    opt->len);
2224 				} else {
2225 					error = copyout(opt->value,
2226 					    optuio->uio_iov[pos].iov_base,
2227 					    opt->len);
2228 					if (error)
2229 						break;
2230 				}
2231 			}
2232 		}
2233 	}
2234 	goto done_errmsg;
2235 
2236  done_deref:
2237 	prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2238 	goto done_errmsg;
2239 
2240  done_unlock_list:
2241 	sx_sunlock(&allprison_lock);
2242  done_errmsg:
2243 	if (error && errmsg_pos >= 0) {
2244 		vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2245 		errmsg_pos = 2 * errmsg_pos + 1;
2246 		if (errmsg_len > 0) {
2247 			if (optuio->uio_segflg == UIO_SYSSPACE)
2248 				bcopy(errmsg,
2249 				    optuio->uio_iov[errmsg_pos].iov_base,
2250 				    errmsg_len);
2251 			else
2252 				copyout(errmsg,
2253 				    optuio->uio_iov[errmsg_pos].iov_base,
2254 				    errmsg_len);
2255 		}
2256 	}
2257 	vfs_freeopts(opts);
2258 	return (error);
2259 }
2260 
2261 
2262 /*
2263  * struct jail_remove_args {
2264  *	int jid;
2265  * };
2266  */
2267 int
sys_jail_remove(struct thread * td,struct jail_remove_args * uap)2268 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2269 {
2270 	struct prison *pr, *cpr, *lpr, *tpr;
2271 	int descend, error;
2272 
2273 	error = priv_check(td, PRIV_JAIL_REMOVE);
2274 	if (error)
2275 		return (error);
2276 
2277 	sx_xlock(&allprison_lock);
2278 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2279 	if (pr == NULL) {
2280 		sx_xunlock(&allprison_lock);
2281 		return (EINVAL);
2282 	}
2283 
2284 	/* Remove all descendants of this prison, then remove this prison. */
2285 	pr->pr_ref++;
2286 	pr->pr_flags |= PR_REMOVE;
2287 	if (!LIST_EMPTY(&pr->pr_children)) {
2288 		mtx_unlock(&pr->pr_mtx);
2289 		lpr = NULL;
2290 		FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2291 			mtx_lock(&cpr->pr_mtx);
2292 			if (cpr->pr_ref > 0) {
2293 				tpr = cpr;
2294 				cpr->pr_ref++;
2295 				cpr->pr_flags |= PR_REMOVE;
2296 			} else {
2297 				/* Already removed - do not do it again. */
2298 				tpr = NULL;
2299 			}
2300 			mtx_unlock(&cpr->pr_mtx);
2301 			if (lpr != NULL) {
2302 				mtx_lock(&lpr->pr_mtx);
2303 				prison_remove_one(lpr);
2304 				sx_xlock(&allprison_lock);
2305 			}
2306 			lpr = tpr;
2307 		}
2308 		if (lpr != NULL) {
2309 			mtx_lock(&lpr->pr_mtx);
2310 			prison_remove_one(lpr);
2311 			sx_xlock(&allprison_lock);
2312 		}
2313 		mtx_lock(&pr->pr_mtx);
2314 	}
2315 	prison_remove_one(pr);
2316 	return (0);
2317 }
2318 
2319 static void
prison_remove_one(struct prison * pr)2320 prison_remove_one(struct prison *pr)
2321 {
2322 	struct proc *p;
2323 	int deuref;
2324 
2325 	/* If the prison was persistent, it is not anymore. */
2326 	deuref = 0;
2327 	if (pr->pr_flags & PR_PERSIST) {
2328 		pr->pr_ref--;
2329 		deuref = PD_DEUREF;
2330 		pr->pr_flags &= ~PR_PERSIST;
2331 	}
2332 
2333 	/*
2334 	 * jail_remove added a reference.  If that's the only one, remove
2335 	 * the prison now.
2336 	 */
2337 	KASSERT(pr->pr_ref > 0,
2338 	    ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2339 	if (pr->pr_ref == 1) {
2340 		prison_deref(pr,
2341 		    deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2342 		return;
2343 	}
2344 
2345 	mtx_unlock(&pr->pr_mtx);
2346 	sx_xunlock(&allprison_lock);
2347 	/*
2348 	 * Kill all processes unfortunate enough to be attached to this prison.
2349 	 */
2350 	sx_slock(&allproc_lock);
2351 	LIST_FOREACH(p, &allproc, p_list) {
2352 		PROC_LOCK(p);
2353 		if (p->p_state != PRS_NEW && p->p_ucred &&
2354 		    p->p_ucred->cr_prison == pr)
2355 			kern_psignal(p, SIGKILL);
2356 		PROC_UNLOCK(p);
2357 	}
2358 	sx_sunlock(&allproc_lock);
2359 	/* Remove the temporary reference added by jail_remove. */
2360 	prison_deref(pr, deuref | PD_DEREF);
2361 }
2362 
2363 
2364 /*
2365  * struct jail_attach_args {
2366  *	int jid;
2367  * };
2368  */
2369 int
sys_jail_attach(struct thread * td,struct jail_attach_args * uap)2370 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2371 {
2372 	struct prison *pr;
2373 	int error;
2374 
2375 	error = priv_check(td, PRIV_JAIL_ATTACH);
2376 	if (error)
2377 		return (error);
2378 
2379 	sx_slock(&allprison_lock);
2380 	pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2381 	if (pr == NULL) {
2382 		sx_sunlock(&allprison_lock);
2383 		return (EINVAL);
2384 	}
2385 
2386 	/*
2387 	 * Do not allow a process to attach to a prison that is not
2388 	 * considered to be "alive".
2389 	 */
2390 	if (pr->pr_uref == 0) {
2391 		mtx_unlock(&pr->pr_mtx);
2392 		sx_sunlock(&allprison_lock);
2393 		return (EINVAL);
2394 	}
2395 
2396 	return (do_jail_attach(td, pr));
2397 }
2398 
2399 static int
do_jail_attach(struct thread * td,struct prison * pr)2400 do_jail_attach(struct thread *td, struct prison *pr)
2401 {
2402 	struct prison *ppr;
2403 	struct proc *p;
2404 	struct ucred *newcred, *oldcred;
2405 	int error;
2406 
2407 	/*
2408 	 * XXX: Note that there is a slight race here if two threads
2409 	 * in the same privileged process attempt to attach to two
2410 	 * different jails at the same time.  It is important for
2411 	 * user processes not to do this, or they might end up with
2412 	 * a process root from one prison, but attached to the jail
2413 	 * of another.
2414 	 */
2415 	pr->pr_ref++;
2416 	pr->pr_uref++;
2417 	mtx_unlock(&pr->pr_mtx);
2418 
2419 	/* Let modules do whatever they need to prepare for attaching. */
2420 	error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2421 	if (error) {
2422 		prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2423 		return (error);
2424 	}
2425 	sx_sunlock(&allprison_lock);
2426 
2427 	/*
2428 	 * Reparent the newly attached process to this jail.
2429 	 */
2430 	ppr = td->td_ucred->cr_prison;
2431 	p = td->td_proc;
2432 	error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2433 	if (error)
2434 		goto e_revert_osd;
2435 
2436 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2437 	if ((error = change_dir(pr->pr_root, td)) != 0)
2438 		goto e_unlock;
2439 #ifdef MAC
2440 	if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2441 		goto e_unlock;
2442 #endif
2443 	VOP_UNLOCK(pr->pr_root, 0);
2444 	if ((error = change_root(pr->pr_root, td)))
2445 		goto e_revert_osd;
2446 
2447 	newcred = crget();
2448 	PROC_LOCK(p);
2449 	oldcred = p->p_ucred;
2450 	setsugid(p);
2451 	crcopy(newcred, oldcred);
2452 	newcred->cr_prison = pr;
2453 	p->p_ucred = newcred;
2454 	PROC_UNLOCK(p);
2455 #ifdef RACCT
2456 	racct_proc_ucred_changed(p, oldcred, newcred);
2457 #endif
2458 	crfree(oldcred);
2459 	prison_deref(ppr, PD_DEREF | PD_DEUREF);
2460 	return (0);
2461  e_unlock:
2462 	VOP_UNLOCK(pr->pr_root, 0);
2463  e_revert_osd:
2464 	/* Tell modules this thread is still in its old jail after all. */
2465 	(void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2466 	prison_deref(pr, PD_DEREF | PD_DEUREF);
2467 	return (error);
2468 }
2469 
2470 
2471 /*
2472  * Returns a locked prison instance, or NULL on failure.
2473  */
2474 struct prison *
prison_find(int prid)2475 prison_find(int prid)
2476 {
2477 	struct prison *pr;
2478 
2479 	sx_assert(&allprison_lock, SX_LOCKED);
2480 	TAILQ_FOREACH(pr, &allprison, pr_list) {
2481 		if (pr->pr_id == prid) {
2482 			mtx_lock(&pr->pr_mtx);
2483 			if (pr->pr_ref > 0)
2484 				return (pr);
2485 			mtx_unlock(&pr->pr_mtx);
2486 		}
2487 	}
2488 	return (NULL);
2489 }
2490 
2491 /*
2492  * Find a prison that is a descendant of mypr.  Returns a locked prison or NULL.
2493  */
2494 struct prison *
prison_find_child(struct prison * mypr,int prid)2495 prison_find_child(struct prison *mypr, int prid)
2496 {
2497 	struct prison *pr;
2498 	int descend;
2499 
2500 	sx_assert(&allprison_lock, SX_LOCKED);
2501 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2502 		if (pr->pr_id == prid) {
2503 			mtx_lock(&pr->pr_mtx);
2504 			if (pr->pr_ref > 0)
2505 				return (pr);
2506 			mtx_unlock(&pr->pr_mtx);
2507 		}
2508 	}
2509 	return (NULL);
2510 }
2511 
2512 /*
2513  * Look for the name relative to mypr.  Returns a locked prison or NULL.
2514  */
2515 struct prison *
prison_find_name(struct prison * mypr,const char * name)2516 prison_find_name(struct prison *mypr, const char *name)
2517 {
2518 	struct prison *pr, *deadpr;
2519 	size_t mylen;
2520 	int descend;
2521 
2522 	sx_assert(&allprison_lock, SX_LOCKED);
2523 	mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2524  again:
2525 	deadpr = NULL;
2526 	FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2527 		if (!strcmp(pr->pr_name + mylen, name)) {
2528 			mtx_lock(&pr->pr_mtx);
2529 			if (pr->pr_ref > 0) {
2530 				if (pr->pr_uref > 0)
2531 					return (pr);
2532 				deadpr = pr;
2533 			}
2534 			mtx_unlock(&pr->pr_mtx);
2535 		}
2536 	}
2537 	/* There was no valid prison - perhaps there was a dying one. */
2538 	if (deadpr != NULL) {
2539 		mtx_lock(&deadpr->pr_mtx);
2540 		if (deadpr->pr_ref == 0) {
2541 			mtx_unlock(&deadpr->pr_mtx);
2542 			goto again;
2543 		}
2544 	}
2545 	return (deadpr);
2546 }
2547 
2548 /*
2549  * See if a prison has the specific flag set.
2550  */
2551 int
prison_flag(struct ucred * cred,unsigned flag)2552 prison_flag(struct ucred *cred, unsigned flag)
2553 {
2554 
2555 	/* This is an atomic read, so no locking is necessary. */
2556 	return (cred->cr_prison->pr_flags & flag);
2557 }
2558 
2559 int
prison_allow(struct ucred * cred,unsigned flag)2560 prison_allow(struct ucred *cred, unsigned flag)
2561 {
2562 
2563 	/* This is an atomic read, so no locking is necessary. */
2564 	return (cred->cr_prison->pr_allow & flag);
2565 }
2566 
2567 /*
2568  * Remove a prison reference.  If that was the last reference, remove the
2569  * prison itself - but not in this context in case there are locks held.
2570  */
2571 void
prison_free_locked(struct prison * pr)2572 prison_free_locked(struct prison *pr)
2573 {
2574 
2575 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2576 	pr->pr_ref--;
2577 	if (pr->pr_ref == 0) {
2578 		mtx_unlock(&pr->pr_mtx);
2579 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2580 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2581 		return;
2582 	}
2583 	mtx_unlock(&pr->pr_mtx);
2584 }
2585 
2586 void
prison_free(struct prison * pr)2587 prison_free(struct prison *pr)
2588 {
2589 
2590 	mtx_lock(&pr->pr_mtx);
2591 	prison_free_locked(pr);
2592 }
2593 
2594 static void
prison_complete(void * context,int pending)2595 prison_complete(void *context, int pending)
2596 {
2597 
2598 	prison_deref((struct prison *)context, 0);
2599 }
2600 
2601 /*
2602  * Remove a prison reference (usually).  This internal version assumes no
2603  * mutexes are held, except perhaps the prison itself.  If there are no more
2604  * references, release and delist the prison.  On completion, the prison lock
2605  * and the allprison lock are both unlocked.
2606  */
2607 static void
prison_deref(struct prison * pr,int flags)2608 prison_deref(struct prison *pr, int flags)
2609 {
2610 	struct prison *ppr, *tpr;
2611 
2612 	if (!(flags & PD_LOCKED))
2613 		mtx_lock(&pr->pr_mtx);
2614 	for (;;) {
2615 		if (flags & PD_DEUREF) {
2616 			pr->pr_uref--;
2617 			KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2618 		}
2619 		if (flags & PD_DEREF)
2620 			pr->pr_ref--;
2621 		/* If the prison still has references, nothing else to do. */
2622 		if (pr->pr_ref > 0) {
2623 			mtx_unlock(&pr->pr_mtx);
2624 			if (flags & PD_LIST_SLOCKED)
2625 				sx_sunlock(&allprison_lock);
2626 			else if (flags & PD_LIST_XLOCKED)
2627 				sx_xunlock(&allprison_lock);
2628 			return;
2629 		}
2630 
2631 		mtx_unlock(&pr->pr_mtx);
2632 		if (flags & PD_LIST_SLOCKED) {
2633 			if (!sx_try_upgrade(&allprison_lock)) {
2634 				sx_sunlock(&allprison_lock);
2635 				sx_xlock(&allprison_lock);
2636 			}
2637 		} else if (!(flags & PD_LIST_XLOCKED))
2638 			sx_xlock(&allprison_lock);
2639 
2640 		TAILQ_REMOVE(&allprison, pr, pr_list);
2641 		LIST_REMOVE(pr, pr_sibling);
2642 		ppr = pr->pr_parent;
2643 		for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2644 			tpr->pr_childcount--;
2645 		sx_xunlock(&allprison_lock);
2646 
2647 #ifdef VIMAGE
2648 		if (pr->pr_vnet != ppr->pr_vnet)
2649 			vnet_destroy(pr->pr_vnet);
2650 #endif
2651 		if (pr->pr_root != NULL)
2652 			vrele(pr->pr_root);
2653 		mtx_destroy(&pr->pr_mtx);
2654 #ifdef INET
2655 		free(pr->pr_ip4, M_PRISON);
2656 #endif
2657 #ifdef INET6
2658 		free(pr->pr_ip6, M_PRISON);
2659 #endif
2660 		if (pr->pr_cpuset != NULL)
2661 			cpuset_rel(pr->pr_cpuset);
2662 		osd_jail_exit(pr);
2663 #ifdef RACCT
2664 		prison_racct_detach(pr);
2665 #endif
2666 		free(pr, M_PRISON);
2667 
2668 		/* Removing a prison frees a reference on its parent. */
2669 		pr = ppr;
2670 		mtx_lock(&pr->pr_mtx);
2671 		flags = PD_DEREF | PD_DEUREF;
2672 	}
2673 }
2674 
2675 void
prison_hold_locked(struct prison * pr)2676 prison_hold_locked(struct prison *pr)
2677 {
2678 
2679 	mtx_assert(&pr->pr_mtx, MA_OWNED);
2680 	KASSERT(pr->pr_ref > 0,
2681 	    ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2682 	pr->pr_ref++;
2683 }
2684 
2685 void
prison_hold(struct prison * pr)2686 prison_hold(struct prison *pr)
2687 {
2688 
2689 	mtx_lock(&pr->pr_mtx);
2690 	prison_hold_locked(pr);
2691 	mtx_unlock(&pr->pr_mtx);
2692 }
2693 
2694 void
prison_proc_hold(struct prison * pr)2695 prison_proc_hold(struct prison *pr)
2696 {
2697 
2698 	mtx_lock(&pr->pr_mtx);
2699 	KASSERT(pr->pr_uref > 0,
2700 	    ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2701 	pr->pr_uref++;
2702 	mtx_unlock(&pr->pr_mtx);
2703 }
2704 
2705 void
prison_proc_free(struct prison * pr)2706 prison_proc_free(struct prison *pr)
2707 {
2708 
2709 	mtx_lock(&pr->pr_mtx);
2710 	KASSERT(pr->pr_uref > 0,
2711 	    ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2712 	prison_deref(pr, PD_DEUREF | PD_LOCKED);
2713 }
2714 
2715 
2716 #ifdef INET
2717 /*
2718  * Restrict a prison's IP address list with its parent's, possibly replacing
2719  * it.  Return true if the replacement buffer was used (or would have been).
2720  */
2721 static int
prison_restrict_ip4(struct prison * pr,struct in_addr * newip4)2722 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2723 {
2724 	int ii, ij, used;
2725 	struct prison *ppr;
2726 
2727 	ppr = pr->pr_parent;
2728 	if (!(pr->pr_flags & PR_IP4_USER)) {
2729 		/* This has no user settings, so just copy the parent's list. */
2730 		if (pr->pr_ip4s < ppr->pr_ip4s) {
2731 			/*
2732 			 * There's no room for the parent's list.  Use the
2733 			 * new list buffer, which is assumed to be big enough
2734 			 * (if it was passed).  If there's no buffer, try to
2735 			 * allocate one.
2736 			 */
2737 			used = 1;
2738 			if (newip4 == NULL) {
2739 				newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2740 				    M_PRISON, M_NOWAIT);
2741 				if (newip4 != NULL)
2742 					used = 0;
2743 			}
2744 			if (newip4 != NULL) {
2745 				bcopy(ppr->pr_ip4, newip4,
2746 				    ppr->pr_ip4s * sizeof(*newip4));
2747 				free(pr->pr_ip4, M_PRISON);
2748 				pr->pr_ip4 = newip4;
2749 				pr->pr_ip4s = ppr->pr_ip4s;
2750 			}
2751 			return (used);
2752 		}
2753 		pr->pr_ip4s = ppr->pr_ip4s;
2754 		if (pr->pr_ip4s > 0)
2755 			bcopy(ppr->pr_ip4, pr->pr_ip4,
2756 			    pr->pr_ip4s * sizeof(*newip4));
2757 		else if (pr->pr_ip4 != NULL) {
2758 			free(pr->pr_ip4, M_PRISON);
2759 			pr->pr_ip4 = NULL;
2760 		}
2761 	} else if (pr->pr_ip4s > 0) {
2762 		/* Remove addresses that aren't in the parent. */
2763 		for (ij = 0; ij < ppr->pr_ip4s; ij++)
2764 			if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2765 				break;
2766 		if (ij < ppr->pr_ip4s)
2767 			ii = 1;
2768 		else {
2769 			bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2770 			    --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2771 			ii = 0;
2772 		}
2773 		for (ij = 1; ii < pr->pr_ip4s; ) {
2774 			if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2775 				ii++;
2776 				continue;
2777 			}
2778 			switch (ij >= ppr->pr_ip4s ? -1 :
2779 				qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2780 			case -1:
2781 				bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2782 				    (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2783 				break;
2784 			case 0:
2785 				ii++;
2786 				ij++;
2787 				break;
2788 			case 1:
2789 				ij++;
2790 				break;
2791 			}
2792 		}
2793 		if (pr->pr_ip4s == 0) {
2794 			pr->pr_flags |= PR_IP4_DISABLE;
2795 			free(pr->pr_ip4, M_PRISON);
2796 			pr->pr_ip4 = NULL;
2797 		}
2798 	}
2799 	return (0);
2800 }
2801 
2802 /*
2803  * Pass back primary IPv4 address of this jail.
2804  *
2805  * If not restricted return success but do not alter the address.  Caller has
2806  * to make sure to initialize it correctly (e.g. INADDR_ANY).
2807  *
2808  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2809  * Address returned in NBO.
2810  */
2811 int
prison_get_ip4(struct ucred * cred,struct in_addr * ia)2812 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2813 {
2814 	struct prison *pr;
2815 
2816 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2817 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2818 
2819 	pr = cred->cr_prison;
2820 	if (!(pr->pr_flags & PR_IP4))
2821 		return (0);
2822 	mtx_lock(&pr->pr_mtx);
2823 	if (!(pr->pr_flags & PR_IP4)) {
2824 		mtx_unlock(&pr->pr_mtx);
2825 		return (0);
2826 	}
2827 	if (pr->pr_ip4 == NULL) {
2828 		mtx_unlock(&pr->pr_mtx);
2829 		return (EAFNOSUPPORT);
2830 	}
2831 
2832 	ia->s_addr = pr->pr_ip4[0].s_addr;
2833 	mtx_unlock(&pr->pr_mtx);
2834 	return (0);
2835 }
2836 
2837 /*
2838  * Return 1 if we should do proper source address selection or are not jailed.
2839  * We will return 0 if we should bypass source address selection in favour
2840  * of the primary jail IPv4 address. Only in this case *ia will be updated and
2841  * returned in NBO.
2842  * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2843  */
2844 int
prison_saddrsel_ip4(struct ucred * cred,struct in_addr * ia)2845 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2846 {
2847 	struct prison *pr;
2848 	struct in_addr lia;
2849 	int error;
2850 
2851 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2852 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2853 
2854 	if (!jailed(cred))
2855 		return (1);
2856 
2857 	pr = cred->cr_prison;
2858 	if (pr->pr_flags & PR_IP4_SADDRSEL)
2859 		return (1);
2860 
2861 	lia.s_addr = INADDR_ANY;
2862 	error = prison_get_ip4(cred, &lia);
2863 	if (error)
2864 		return (error);
2865 	if (lia.s_addr == INADDR_ANY)
2866 		return (1);
2867 
2868 	ia->s_addr = lia.s_addr;
2869 	return (0);
2870 }
2871 
2872 /*
2873  * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2874  */
2875 int
prison_equal_ip4(struct prison * pr1,struct prison * pr2)2876 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2877 {
2878 
2879 	if (pr1 == pr2)
2880 		return (1);
2881 
2882 	/*
2883 	 * No need to lock since the PR_IP4_USER flag can't be altered for
2884 	 * existing prisons.
2885 	 */
2886 	while (pr1 != &prison0 &&
2887 #ifdef VIMAGE
2888 	       !(pr1->pr_flags & PR_VNET) &&
2889 #endif
2890 	       !(pr1->pr_flags & PR_IP4_USER))
2891 		pr1 = pr1->pr_parent;
2892 	while (pr2 != &prison0 &&
2893 #ifdef VIMAGE
2894 	       !(pr2->pr_flags & PR_VNET) &&
2895 #endif
2896 	       !(pr2->pr_flags & PR_IP4_USER))
2897 		pr2 = pr2->pr_parent;
2898 	return (pr1 == pr2);
2899 }
2900 
2901 /*
2902  * Make sure our (source) address is set to something meaningful to this
2903  * jail.
2904  *
2905  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2906  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2907  * doesn't allow IPv4.  Address passed in in NBO and returned in NBO.
2908  */
2909 int
prison_local_ip4(struct ucred * cred,struct in_addr * ia)2910 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2911 {
2912 	struct prison *pr;
2913 	struct in_addr ia0;
2914 	int error;
2915 
2916 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2917 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2918 
2919 	pr = cred->cr_prison;
2920 	if (!(pr->pr_flags & PR_IP4))
2921 		return (0);
2922 	mtx_lock(&pr->pr_mtx);
2923 	if (!(pr->pr_flags & PR_IP4)) {
2924 		mtx_unlock(&pr->pr_mtx);
2925 		return (0);
2926 	}
2927 	if (pr->pr_ip4 == NULL) {
2928 		mtx_unlock(&pr->pr_mtx);
2929 		return (EAFNOSUPPORT);
2930 	}
2931 
2932 	ia0.s_addr = ntohl(ia->s_addr);
2933 	if (ia0.s_addr == INADDR_LOOPBACK) {
2934 		ia->s_addr = pr->pr_ip4[0].s_addr;
2935 		mtx_unlock(&pr->pr_mtx);
2936 		return (0);
2937 	}
2938 
2939 	if (ia0.s_addr == INADDR_ANY) {
2940 		/*
2941 		 * In case there is only 1 IPv4 address, bind directly.
2942 		 */
2943 		if (pr->pr_ip4s == 1)
2944 			ia->s_addr = pr->pr_ip4[0].s_addr;
2945 		mtx_unlock(&pr->pr_mtx);
2946 		return (0);
2947 	}
2948 
2949 	error = _prison_check_ip4(pr, ia);
2950 	mtx_unlock(&pr->pr_mtx);
2951 	return (error);
2952 }
2953 
2954 /*
2955  * Rewrite destination address in case we will connect to loopback address.
2956  *
2957  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2958  * Address passed in in NBO and returned in NBO.
2959  */
2960 int
prison_remote_ip4(struct ucred * cred,struct in_addr * ia)2961 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2962 {
2963 	struct prison *pr;
2964 
2965 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2966 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2967 
2968 	pr = cred->cr_prison;
2969 	if (!(pr->pr_flags & PR_IP4))
2970 		return (0);
2971 	mtx_lock(&pr->pr_mtx);
2972 	if (!(pr->pr_flags & PR_IP4)) {
2973 		mtx_unlock(&pr->pr_mtx);
2974 		return (0);
2975 	}
2976 	if (pr->pr_ip4 == NULL) {
2977 		mtx_unlock(&pr->pr_mtx);
2978 		return (EAFNOSUPPORT);
2979 	}
2980 
2981 	if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2982 		ia->s_addr = pr->pr_ip4[0].s_addr;
2983 		mtx_unlock(&pr->pr_mtx);
2984 		return (0);
2985 	}
2986 
2987 	/*
2988 	 * Return success because nothing had to be changed.
2989 	 */
2990 	mtx_unlock(&pr->pr_mtx);
2991 	return (0);
2992 }
2993 
2994 /*
2995  * Check if given address belongs to the jail referenced by cred/prison.
2996  *
2997  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2998  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2999  * doesn't allow IPv4.  Address passed in in NBO.
3000  */
3001 static int
_prison_check_ip4(struct prison * pr,struct in_addr * ia)3002 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
3003 {
3004 	int i, a, z, d;
3005 
3006 	/*
3007 	 * Check the primary IP.
3008 	 */
3009 	if (pr->pr_ip4[0].s_addr == ia->s_addr)
3010 		return (0);
3011 
3012 	/*
3013 	 * All the other IPs are sorted so we can do a binary search.
3014 	 */
3015 	a = 0;
3016 	z = pr->pr_ip4s - 2;
3017 	while (a <= z) {
3018 		i = (a + z) / 2;
3019 		d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3020 		if (d > 0)
3021 			z = i - 1;
3022 		else if (d < 0)
3023 			a = i + 1;
3024 		else
3025 			return (0);
3026 	}
3027 
3028 	return (EADDRNOTAVAIL);
3029 }
3030 
3031 int
prison_check_ip4(struct ucred * cred,struct in_addr * ia)3032 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
3033 {
3034 	struct prison *pr;
3035 	int error;
3036 
3037 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3038 	KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3039 
3040 	pr = cred->cr_prison;
3041 	if (!(pr->pr_flags & PR_IP4))
3042 		return (0);
3043 	mtx_lock(&pr->pr_mtx);
3044 	if (!(pr->pr_flags & PR_IP4)) {
3045 		mtx_unlock(&pr->pr_mtx);
3046 		return (0);
3047 	}
3048 	if (pr->pr_ip4 == NULL) {
3049 		mtx_unlock(&pr->pr_mtx);
3050 		return (EAFNOSUPPORT);
3051 	}
3052 
3053 	error = _prison_check_ip4(pr, ia);
3054 	mtx_unlock(&pr->pr_mtx);
3055 	return (error);
3056 }
3057 #endif
3058 
3059 #ifdef INET6
3060 static int
prison_restrict_ip6(struct prison * pr,struct in6_addr * newip6)3061 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3062 {
3063 	int ii, ij, used;
3064 	struct prison *ppr;
3065 
3066 	ppr = pr->pr_parent;
3067 	if (!(pr->pr_flags & PR_IP6_USER)) {
3068 		/* This has no user settings, so just copy the parent's list. */
3069 		if (pr->pr_ip6s < ppr->pr_ip6s) {
3070 			/*
3071 			 * There's no room for the parent's list.  Use the
3072 			 * new list buffer, which is assumed to be big enough
3073 			 * (if it was passed).  If there's no buffer, try to
3074 			 * allocate one.
3075 			 */
3076 			used = 1;
3077 			if (newip6 == NULL) {
3078 				newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3079 				    M_PRISON, M_NOWAIT);
3080 				if (newip6 != NULL)
3081 					used = 0;
3082 			}
3083 			if (newip6 != NULL) {
3084 				bcopy(ppr->pr_ip6, newip6,
3085 				    ppr->pr_ip6s * sizeof(*newip6));
3086 				free(pr->pr_ip6, M_PRISON);
3087 				pr->pr_ip6 = newip6;
3088 				pr->pr_ip6s = ppr->pr_ip6s;
3089 			}
3090 			return (used);
3091 		}
3092 		pr->pr_ip6s = ppr->pr_ip6s;
3093 		if (pr->pr_ip6s > 0)
3094 			bcopy(ppr->pr_ip6, pr->pr_ip6,
3095 			    pr->pr_ip6s * sizeof(*newip6));
3096 		else if (pr->pr_ip6 != NULL) {
3097 			free(pr->pr_ip6, M_PRISON);
3098 			pr->pr_ip6 = NULL;
3099 		}
3100 	} else if (pr->pr_ip6s > 0) {
3101 		/* Remove addresses that aren't in the parent. */
3102 		for (ij = 0; ij < ppr->pr_ip6s; ij++)
3103 			if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3104 			    &ppr->pr_ip6[ij]))
3105 				break;
3106 		if (ij < ppr->pr_ip6s)
3107 			ii = 1;
3108 		else {
3109 			bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3110 			    --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3111 			ii = 0;
3112 		}
3113 		for (ij = 1; ii < pr->pr_ip6s; ) {
3114 			if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3115 			    &ppr->pr_ip6[0])) {
3116 				ii++;
3117 				continue;
3118 			}
3119 			switch (ij >= ppr->pr_ip6s ? -1 :
3120 				qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3121 			case -1:
3122 				bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3123 				    (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3124 				break;
3125 			case 0:
3126 				ii++;
3127 				ij++;
3128 				break;
3129 			case 1:
3130 				ij++;
3131 				break;
3132 			}
3133 		}
3134 		if (pr->pr_ip6s == 0) {
3135 			pr->pr_flags |= PR_IP6_DISABLE;
3136 			free(pr->pr_ip6, M_PRISON);
3137 			pr->pr_ip6 = NULL;
3138 		}
3139 	}
3140 	return 0;
3141 }
3142 
3143 /*
3144  * Pass back primary IPv6 address for this jail.
3145  *
3146  * If not restricted return success but do not alter the address.  Caller has
3147  * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3148  *
3149  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3150  */
3151 int
prison_get_ip6(struct ucred * cred,struct in6_addr * ia6)3152 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3153 {
3154 	struct prison *pr;
3155 
3156 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3157 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3158 
3159 	pr = cred->cr_prison;
3160 	if (!(pr->pr_flags & PR_IP6))
3161 		return (0);
3162 	mtx_lock(&pr->pr_mtx);
3163 	if (!(pr->pr_flags & PR_IP6)) {
3164 		mtx_unlock(&pr->pr_mtx);
3165 		return (0);
3166 	}
3167 	if (pr->pr_ip6 == NULL) {
3168 		mtx_unlock(&pr->pr_mtx);
3169 		return (EAFNOSUPPORT);
3170 	}
3171 
3172 	bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3173 	mtx_unlock(&pr->pr_mtx);
3174 	return (0);
3175 }
3176 
3177 /*
3178  * Return 1 if we should do proper source address selection or are not jailed.
3179  * We will return 0 if we should bypass source address selection in favour
3180  * of the primary jail IPv6 address. Only in this case *ia will be updated and
3181  * returned in NBO.
3182  * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3183  */
3184 int
prison_saddrsel_ip6(struct ucred * cred,struct in6_addr * ia6)3185 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3186 {
3187 	struct prison *pr;
3188 	struct in6_addr lia6;
3189 	int error;
3190 
3191 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3192 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3193 
3194 	if (!jailed(cred))
3195 		return (1);
3196 
3197 	pr = cred->cr_prison;
3198 	if (pr->pr_flags & PR_IP6_SADDRSEL)
3199 		return (1);
3200 
3201 	lia6 = in6addr_any;
3202 	error = prison_get_ip6(cred, &lia6);
3203 	if (error)
3204 		return (error);
3205 	if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3206 		return (1);
3207 
3208 	bcopy(&lia6, ia6, sizeof(struct in6_addr));
3209 	return (0);
3210 }
3211 
3212 /*
3213  * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3214  */
3215 int
prison_equal_ip6(struct prison * pr1,struct prison * pr2)3216 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3217 {
3218 
3219 	if (pr1 == pr2)
3220 		return (1);
3221 
3222 	while (pr1 != &prison0 &&
3223 #ifdef VIMAGE
3224 	       !(pr1->pr_flags & PR_VNET) &&
3225 #endif
3226 	       !(pr1->pr_flags & PR_IP6_USER))
3227 		pr1 = pr1->pr_parent;
3228 	while (pr2 != &prison0 &&
3229 #ifdef VIMAGE
3230 	       !(pr2->pr_flags & PR_VNET) &&
3231 #endif
3232 	       !(pr2->pr_flags & PR_IP6_USER))
3233 		pr2 = pr2->pr_parent;
3234 	return (pr1 == pr2);
3235 }
3236 
3237 /*
3238  * Make sure our (source) address is set to something meaningful to this jail.
3239  *
3240  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3241  * when needed while binding.
3242  *
3243  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3244  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3245  * doesn't allow IPv6.
3246  */
3247 int
prison_local_ip6(struct ucred * cred,struct in6_addr * ia6,int v6only)3248 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3249 {
3250 	struct prison *pr;
3251 	int error;
3252 
3253 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3254 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3255 
3256 	pr = cred->cr_prison;
3257 	if (!(pr->pr_flags & PR_IP6))
3258 		return (0);
3259 	mtx_lock(&pr->pr_mtx);
3260 	if (!(pr->pr_flags & PR_IP6)) {
3261 		mtx_unlock(&pr->pr_mtx);
3262 		return (0);
3263 	}
3264 	if (pr->pr_ip6 == NULL) {
3265 		mtx_unlock(&pr->pr_mtx);
3266 		return (EAFNOSUPPORT);
3267 	}
3268 
3269 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3270 		bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3271 		mtx_unlock(&pr->pr_mtx);
3272 		return (0);
3273 	}
3274 
3275 	if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3276 		/*
3277 		 * In case there is only 1 IPv6 address, and v6only is true,
3278 		 * then bind directly.
3279 		 */
3280 		if (v6only != 0 && pr->pr_ip6s == 1)
3281 			bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3282 		mtx_unlock(&pr->pr_mtx);
3283 		return (0);
3284 	}
3285 
3286 	error = _prison_check_ip6(pr, ia6);
3287 	mtx_unlock(&pr->pr_mtx);
3288 	return (error);
3289 }
3290 
3291 /*
3292  * Rewrite destination address in case we will connect to loopback address.
3293  *
3294  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3295  */
3296 int
prison_remote_ip6(struct ucred * cred,struct in6_addr * ia6)3297 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3298 {
3299 	struct prison *pr;
3300 
3301 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3302 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3303 
3304 	pr = cred->cr_prison;
3305 	if (!(pr->pr_flags & PR_IP6))
3306 		return (0);
3307 	mtx_lock(&pr->pr_mtx);
3308 	if (!(pr->pr_flags & PR_IP6)) {
3309 		mtx_unlock(&pr->pr_mtx);
3310 		return (0);
3311 	}
3312 	if (pr->pr_ip6 == NULL) {
3313 		mtx_unlock(&pr->pr_mtx);
3314 		return (EAFNOSUPPORT);
3315 	}
3316 
3317 	if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3318 		bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3319 		mtx_unlock(&pr->pr_mtx);
3320 		return (0);
3321 	}
3322 
3323 	/*
3324 	 * Return success because nothing had to be changed.
3325 	 */
3326 	mtx_unlock(&pr->pr_mtx);
3327 	return (0);
3328 }
3329 
3330 /*
3331  * Check if given address belongs to the jail referenced by cred/prison.
3332  *
3333  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3334  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3335  * doesn't allow IPv6.
3336  */
3337 static int
_prison_check_ip6(struct prison * pr,struct in6_addr * ia6)3338 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3339 {
3340 	int i, a, z, d;
3341 
3342 	/*
3343 	 * Check the primary IP.
3344 	 */
3345 	if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3346 		return (0);
3347 
3348 	/*
3349 	 * All the other IPs are sorted so we can do a binary search.
3350 	 */
3351 	a = 0;
3352 	z = pr->pr_ip6s - 2;
3353 	while (a <= z) {
3354 		i = (a + z) / 2;
3355 		d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3356 		if (d > 0)
3357 			z = i - 1;
3358 		else if (d < 0)
3359 			a = i + 1;
3360 		else
3361 			return (0);
3362 	}
3363 
3364 	return (EADDRNOTAVAIL);
3365 }
3366 
3367 int
prison_check_ip6(struct ucred * cred,struct in6_addr * ia6)3368 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3369 {
3370 	struct prison *pr;
3371 	int error;
3372 
3373 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3374 	KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3375 
3376 	pr = cred->cr_prison;
3377 	if (!(pr->pr_flags & PR_IP6))
3378 		return (0);
3379 	mtx_lock(&pr->pr_mtx);
3380 	if (!(pr->pr_flags & PR_IP6)) {
3381 		mtx_unlock(&pr->pr_mtx);
3382 		return (0);
3383 	}
3384 	if (pr->pr_ip6 == NULL) {
3385 		mtx_unlock(&pr->pr_mtx);
3386 		return (EAFNOSUPPORT);
3387 	}
3388 
3389 	error = _prison_check_ip6(pr, ia6);
3390 	mtx_unlock(&pr->pr_mtx);
3391 	return (error);
3392 }
3393 #endif
3394 
3395 /*
3396  * Check if a jail supports the given address family.
3397  *
3398  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3399  * if not.
3400  */
3401 int
prison_check_af(struct ucred * cred,int af)3402 prison_check_af(struct ucred *cred, int af)
3403 {
3404 	struct prison *pr;
3405 	int error;
3406 
3407 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3408 
3409 	pr = cred->cr_prison;
3410 #ifdef VIMAGE
3411 	/* Prisons with their own network stack are not limited. */
3412 	if (prison_owns_vnet(cred))
3413 		return (0);
3414 #endif
3415 
3416 	error = 0;
3417 	switch (af)
3418 	{
3419 #ifdef INET
3420 	case AF_INET:
3421 		if (pr->pr_flags & PR_IP4)
3422 		{
3423 			mtx_lock(&pr->pr_mtx);
3424 			if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3425 				error = EAFNOSUPPORT;
3426 			mtx_unlock(&pr->pr_mtx);
3427 		}
3428 		break;
3429 #endif
3430 #ifdef INET6
3431 	case AF_INET6:
3432 		if (pr->pr_flags & PR_IP6)
3433 		{
3434 			mtx_lock(&pr->pr_mtx);
3435 			if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3436 				error = EAFNOSUPPORT;
3437 			mtx_unlock(&pr->pr_mtx);
3438 		}
3439 		break;
3440 #endif
3441 	case AF_LOCAL:
3442 	case AF_ROUTE:
3443 		break;
3444 	default:
3445 		if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3446 			error = EAFNOSUPPORT;
3447 	}
3448 	return (error);
3449 }
3450 
3451 /*
3452  * Check if given address belongs to the jail referenced by cred (wrapper to
3453  * prison_check_ip[46]).
3454  *
3455  * Returns 0 if jail doesn't restrict the address family or if address belongs
3456  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3457  * the jail doesn't allow the address family.  IPv4 Address passed in in NBO.
3458  */
3459 int
prison_if(struct ucred * cred,struct sockaddr * sa)3460 prison_if(struct ucred *cred, struct sockaddr *sa)
3461 {
3462 #ifdef INET
3463 	struct sockaddr_in *sai;
3464 #endif
3465 #ifdef INET6
3466 	struct sockaddr_in6 *sai6;
3467 #endif
3468 	int error;
3469 
3470 	KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3471 	KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3472 
3473 #ifdef VIMAGE
3474 	if (prison_owns_vnet(cred))
3475 		return (0);
3476 #endif
3477 
3478 	error = 0;
3479 	switch (sa->sa_family)
3480 	{
3481 #ifdef INET
3482 	case AF_INET:
3483 		sai = (struct sockaddr_in *)sa;
3484 		error = prison_check_ip4(cred, &sai->sin_addr);
3485 		break;
3486 #endif
3487 #ifdef INET6
3488 	case AF_INET6:
3489 		sai6 = (struct sockaddr_in6 *)sa;
3490 		error = prison_check_ip6(cred, &sai6->sin6_addr);
3491 		break;
3492 #endif
3493 	default:
3494 		if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3495 			error = EAFNOSUPPORT;
3496 	}
3497 	return (error);
3498 }
3499 
3500 /*
3501  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3502  */
3503 int
prison_check(struct ucred * cred1,struct ucred * cred2)3504 prison_check(struct ucred *cred1, struct ucred *cred2)
3505 {
3506 
3507 	return ((cred1->cr_prison == cred2->cr_prison ||
3508 	    prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3509 }
3510 
3511 /*
3512  * Return 1 if p2 is a child of p1, otherwise 0.
3513  */
3514 int
prison_ischild(struct prison * pr1,struct prison * pr2)3515 prison_ischild(struct prison *pr1, struct prison *pr2)
3516 {
3517 
3518 	for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3519 		if (pr1 == pr2)
3520 			return (1);
3521 	return (0);
3522 }
3523 
3524 /*
3525  * Return 1 if the passed credential is in a jail, otherwise 0.
3526  */
3527 int
jailed(struct ucred * cred)3528 jailed(struct ucred *cred)
3529 {
3530 
3531 	return (cred->cr_prison != &prison0);
3532 }
3533 
3534 /*
3535  * Return 1 if the passed credential is in a jail and that jail does not
3536  * have its own virtual network stack, otherwise 0.
3537  */
3538 int
jailed_without_vnet(struct ucred * cred)3539 jailed_without_vnet(struct ucred *cred)
3540 {
3541 
3542 	if (!jailed(cred))
3543 		return (0);
3544 #ifdef VIMAGE
3545 	if (prison_owns_vnet(cred))
3546 		return (0);
3547 #endif
3548 
3549 	return (1);
3550 }
3551 
3552 /*
3553  * Return the correct hostname (domainname, et al) for the passed credential.
3554  */
3555 void
getcredhostname(struct ucred * cred,char * buf,size_t size)3556 getcredhostname(struct ucred *cred, char *buf, size_t size)
3557 {
3558 	struct prison *pr;
3559 
3560 	/*
3561 	 * A NULL credential can be used to shortcut to the physical
3562 	 * system's hostname.
3563 	 */
3564 	pr = (cred != NULL) ? cred->cr_prison : &prison0;
3565 	mtx_lock(&pr->pr_mtx);
3566 	strlcpy(buf, pr->pr_hostname, size);
3567 	mtx_unlock(&pr->pr_mtx);
3568 }
3569 
3570 void
getcreddomainname(struct ucred * cred,char * buf,size_t size)3571 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3572 {
3573 
3574 	mtx_lock(&cred->cr_prison->pr_mtx);
3575 	strlcpy(buf, cred->cr_prison->pr_domainname, size);
3576 	mtx_unlock(&cred->cr_prison->pr_mtx);
3577 }
3578 
3579 void
getcredhostuuid(struct ucred * cred,char * buf,size_t size)3580 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3581 {
3582 
3583 	mtx_lock(&cred->cr_prison->pr_mtx);
3584 	strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3585 	mtx_unlock(&cred->cr_prison->pr_mtx);
3586 }
3587 
3588 void
getcredhostid(struct ucred * cred,unsigned long * hostid)3589 getcredhostid(struct ucred *cred, unsigned long *hostid)
3590 {
3591 
3592 	mtx_lock(&cred->cr_prison->pr_mtx);
3593 	*hostid = cred->cr_prison->pr_hostid;
3594 	mtx_unlock(&cred->cr_prison->pr_mtx);
3595 }
3596 
3597 #ifdef VIMAGE
3598 /*
3599  * Determine whether the prison represented by cred owns
3600  * its vnet rather than having it inherited.
3601  *
3602  * Returns 1 in case the prison owns the vnet, 0 otherwise.
3603  */
3604 int
prison_owns_vnet(struct ucred * cred)3605 prison_owns_vnet(struct ucred *cred)
3606 {
3607 
3608 	/*
3609 	 * vnets cannot be added/removed after jail creation,
3610 	 * so no need to lock here.
3611 	 */
3612 	return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3613 }
3614 #endif
3615 
3616 /*
3617  * Determine whether the subject represented by cred can "see"
3618  * status of a mount point.
3619  * Returns: 0 for permitted, ENOENT otherwise.
3620  * XXX: This function should be called cr_canseemount() and should be
3621  *      placed in kern_prot.c.
3622  */
3623 int
prison_canseemount(struct ucred * cred,struct mount * mp)3624 prison_canseemount(struct ucred *cred, struct mount *mp)
3625 {
3626 	struct prison *pr;
3627 	struct statfs *sp;
3628 	size_t len;
3629 
3630 	pr = cred->cr_prison;
3631 	if (pr->pr_enforce_statfs == 0)
3632 		return (0);
3633 	if (pr->pr_root->v_mount == mp)
3634 		return (0);
3635 	if (pr->pr_enforce_statfs == 2)
3636 		return (ENOENT);
3637 	/*
3638 	 * If jail's chroot directory is set to "/" we should be able to see
3639 	 * all mount-points from inside a jail.
3640 	 * This is ugly check, but this is the only situation when jail's
3641 	 * directory ends with '/'.
3642 	 */
3643 	if (strcmp(pr->pr_path, "/") == 0)
3644 		return (0);
3645 	len = strlen(pr->pr_path);
3646 	sp = &mp->mnt_stat;
3647 	if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3648 		return (ENOENT);
3649 	/*
3650 	 * Be sure that we don't have situation where jail's root directory
3651 	 * is "/some/path" and mount point is "/some/pathpath".
3652 	 */
3653 	if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3654 		return (ENOENT);
3655 	return (0);
3656 }
3657 
3658 void
prison_enforce_statfs(struct ucred * cred,struct mount * mp,struct statfs * sp)3659 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3660 {
3661 	char jpath[MAXPATHLEN];
3662 	struct prison *pr;
3663 	size_t len;
3664 
3665 	pr = cred->cr_prison;
3666 	if (pr->pr_enforce_statfs == 0)
3667 		return;
3668 	if (prison_canseemount(cred, mp) != 0) {
3669 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3670 		strlcpy(sp->f_mntonname, "[restricted]",
3671 		    sizeof(sp->f_mntonname));
3672 		return;
3673 	}
3674 	if (pr->pr_root->v_mount == mp) {
3675 		/*
3676 		 * Clear current buffer data, so we are sure nothing from
3677 		 * the valid path left there.
3678 		 */
3679 		bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3680 		*sp->f_mntonname = '/';
3681 		return;
3682 	}
3683 	/*
3684 	 * If jail's chroot directory is set to "/" we should be able to see
3685 	 * all mount-points from inside a jail.
3686 	 */
3687 	if (strcmp(pr->pr_path, "/") == 0)
3688 		return;
3689 	len = strlen(pr->pr_path);
3690 	strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3691 	/*
3692 	 * Clear current buffer data, so we are sure nothing from
3693 	 * the valid path left there.
3694 	 */
3695 	bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3696 	if (*jpath == '\0') {
3697 		/* Should never happen. */
3698 		*sp->f_mntonname = '/';
3699 	} else {
3700 		strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3701 	}
3702 }
3703 
3704 /*
3705  * Check with permission for a specific privilege is granted within jail.  We
3706  * have a specific list of accepted privileges; the rest are denied.
3707  */
3708 int
prison_priv_check(struct ucred * cred,int priv)3709 prison_priv_check(struct ucred *cred, int priv)
3710 {
3711 
3712 	if (!jailed(cred))
3713 		return (0);
3714 
3715 #ifdef VIMAGE
3716 	/*
3717 	 * Privileges specific to prisons with a virtual network stack.
3718 	 * There might be a duplicate entry here in case the privilege
3719 	 * is only granted conditionally in the legacy jail case.
3720 	 */
3721 	switch (priv) {
3722 #ifdef notyet
3723 		/*
3724 		 * NFS-specific privileges.
3725 		 */
3726 	case PRIV_NFS_DAEMON:
3727 	case PRIV_NFS_LOCKD:
3728 #endif
3729 		/*
3730 		 * Network stack privileges.
3731 		 */
3732 	case PRIV_NET_BRIDGE:
3733 	case PRIV_NET_GRE:
3734 	case PRIV_NET_BPF:
3735 	case PRIV_NET_RAW:		/* Dup, cond. in legacy jail case. */
3736 	case PRIV_NET_ROUTE:
3737 	case PRIV_NET_TAP:
3738 	case PRIV_NET_SETIFMTU:
3739 	case PRIV_NET_SETIFFLAGS:
3740 	case PRIV_NET_SETIFCAP:
3741 	case PRIV_NET_SETIFDESCR:
3742 	case PRIV_NET_SETIFNAME	:
3743 	case PRIV_NET_SETIFMETRIC:
3744 	case PRIV_NET_SETIFPHYS:
3745 	case PRIV_NET_SETIFMAC:
3746 	case PRIV_NET_ADDMULTI:
3747 	case PRIV_NET_DELMULTI:
3748 	case PRIV_NET_HWIOCTL:
3749 	case PRIV_NET_SETLLADDR:
3750 	case PRIV_NET_ADDIFGROUP:
3751 	case PRIV_NET_DELIFGROUP:
3752 	case PRIV_NET_IFCREATE:
3753 	case PRIV_NET_IFDESTROY:
3754 	case PRIV_NET_ADDIFADDR:
3755 	case PRIV_NET_DELIFADDR:
3756 	case PRIV_NET_LAGG:
3757 	case PRIV_NET_GIF:
3758 	case PRIV_NET_SETIFVNET:
3759 	case PRIV_NET_SETIFFIB:
3760 
3761 		/*
3762 		 * 802.11-related privileges.
3763 		 */
3764 	case PRIV_NET80211_GETKEY:
3765 #ifdef notyet
3766 	case PRIV_NET80211_MANAGE:		/* XXX-BZ discuss with sam@ */
3767 #endif
3768 
3769 #ifdef notyet
3770 		/*
3771 		 * AppleTalk privileges.
3772 		 */
3773 	case PRIV_NETATALK_RESERVEDPORT:
3774 
3775 		/*
3776 		 * ATM privileges.
3777 		 */
3778 	case PRIV_NETATM_CFG:
3779 	case PRIV_NETATM_ADD:
3780 	case PRIV_NETATM_DEL:
3781 	case PRIV_NETATM_SET:
3782 
3783 		/*
3784 		 * Bluetooth privileges.
3785 		 */
3786 	case PRIV_NETBLUETOOTH_RAW:
3787 #endif
3788 
3789 		/*
3790 		 * Netgraph and netgraph module privileges.
3791 		 */
3792 	case PRIV_NETGRAPH_CONTROL:
3793 #ifdef notyet
3794 	case PRIV_NETGRAPH_TTY:
3795 #endif
3796 
3797 		/*
3798 		 * IPv4 and IPv6 privileges.
3799 		 */
3800 	case PRIV_NETINET_IPFW:
3801 	case PRIV_NETINET_DIVERT:
3802 	case PRIV_NETINET_PF:
3803 	case PRIV_NETINET_DUMMYNET:
3804 	case PRIV_NETINET_CARP:
3805 	case PRIV_NETINET_MROUTE:
3806 	case PRIV_NETINET_RAW:
3807 	case PRIV_NETINET_ADDRCTRL6:
3808 	case PRIV_NETINET_ND6:
3809 	case PRIV_NETINET_SCOPE6:
3810 	case PRIV_NETINET_ALIFETIME6:
3811 	case PRIV_NETINET_IPSEC:
3812 	case PRIV_NETINET_BINDANY:
3813 
3814 #ifdef notyet
3815 		/*
3816 		 * IPX/SPX privileges.
3817 		 */
3818 	case PRIV_NETIPX_RESERVEDPORT:
3819 	case PRIV_NETIPX_RAW:
3820 
3821 		/*
3822 		 * NCP privileges.
3823 		 */
3824 	case PRIV_NETNCP:
3825 
3826 		/*
3827 		 * SMB privileges.
3828 		 */
3829 	case PRIV_NETSMB:
3830 #endif
3831 
3832 	/*
3833 	 * No default: or deny here.
3834 	 * In case of no permit fall through to next switch().
3835 	 */
3836 		if (cred->cr_prison->pr_flags & PR_VNET)
3837 			return (0);
3838 	}
3839 #endif /* VIMAGE */
3840 
3841 	switch (priv) {
3842 
3843 		/*
3844 		 * Allow ktrace privileges for root in jail.
3845 		 */
3846 	case PRIV_KTRACE:
3847 
3848 #if 0
3849 		/*
3850 		 * Allow jailed processes to configure audit identity and
3851 		 * submit audit records (login, etc).  In the future we may
3852 		 * want to further refine the relationship between audit and
3853 		 * jail.
3854 		 */
3855 	case PRIV_AUDIT_GETAUDIT:
3856 	case PRIV_AUDIT_SETAUDIT:
3857 	case PRIV_AUDIT_SUBMIT:
3858 #endif
3859 
3860 		/*
3861 		 * Allow jailed processes to manipulate process UNIX
3862 		 * credentials in any way they see fit.
3863 		 */
3864 	case PRIV_CRED_SETUID:
3865 	case PRIV_CRED_SETEUID:
3866 	case PRIV_CRED_SETGID:
3867 	case PRIV_CRED_SETEGID:
3868 	case PRIV_CRED_SETGROUPS:
3869 	case PRIV_CRED_SETREUID:
3870 	case PRIV_CRED_SETREGID:
3871 	case PRIV_CRED_SETRESUID:
3872 	case PRIV_CRED_SETRESGID:
3873 
3874 		/*
3875 		 * Jail implements visibility constraints already, so allow
3876 		 * jailed root to override uid/gid-based constraints.
3877 		 */
3878 	case PRIV_SEEOTHERGIDS:
3879 	case PRIV_SEEOTHERUIDS:
3880 
3881 		/*
3882 		 * Jail implements inter-process debugging limits already, so
3883 		 * allow jailed root various debugging privileges.
3884 		 */
3885 	case PRIV_DEBUG_DIFFCRED:
3886 	case PRIV_DEBUG_SUGID:
3887 	case PRIV_DEBUG_UNPRIV:
3888 
3889 		/*
3890 		 * Allow jail to set various resource limits and login
3891 		 * properties, and for now, exceed process resource limits.
3892 		 */
3893 	case PRIV_PROC_LIMIT:
3894 	case PRIV_PROC_SETLOGIN:
3895 	case PRIV_PROC_SETRLIMIT:
3896 
3897 		/*
3898 		 * System V and POSIX IPC privileges are granted in jail.
3899 		 */
3900 	case PRIV_IPC_READ:
3901 	case PRIV_IPC_WRITE:
3902 	case PRIV_IPC_ADMIN:
3903 	case PRIV_IPC_MSGSIZE:
3904 	case PRIV_MQ_ADMIN:
3905 
3906 		/*
3907 		 * Jail operations within a jail work on child jails.
3908 		 */
3909 	case PRIV_JAIL_ATTACH:
3910 	case PRIV_JAIL_SET:
3911 	case PRIV_JAIL_REMOVE:
3912 
3913 		/*
3914 		 * Jail implements its own inter-process limits, so allow
3915 		 * root processes in jail to change scheduling on other
3916 		 * processes in the same jail.  Likewise for signalling.
3917 		 */
3918 	case PRIV_SCHED_DIFFCRED:
3919 	case PRIV_SCHED_CPUSET:
3920 	case PRIV_SIGNAL_DIFFCRED:
3921 	case PRIV_SIGNAL_SUGID:
3922 
3923 		/*
3924 		 * Allow jailed processes to write to sysctls marked as jail
3925 		 * writable.
3926 		 */
3927 	case PRIV_SYSCTL_WRITEJAIL:
3928 
3929 		/*
3930 		 * Allow root in jail to manage a variety of quota
3931 		 * properties.  These should likely be conditional on a
3932 		 * configuration option.
3933 		 */
3934 	case PRIV_VFS_GETQUOTA:
3935 	case PRIV_VFS_SETQUOTA:
3936 
3937 		/*
3938 		 * Since Jail relies on chroot() to implement file system
3939 		 * protections, grant many VFS privileges to root in jail.
3940 		 * Be careful to exclude mount-related and NFS-related
3941 		 * privileges.
3942 		 */
3943 	case PRIV_VFS_READ:
3944 	case PRIV_VFS_WRITE:
3945 	case PRIV_VFS_ADMIN:
3946 	case PRIV_VFS_EXEC:
3947 	case PRIV_VFS_LOOKUP:
3948 	case PRIV_VFS_BLOCKRESERVE:	/* XXXRW: Slightly surprising. */
3949 	case PRIV_VFS_CHFLAGS_DEV:
3950 	case PRIV_VFS_CHOWN:
3951 	case PRIV_VFS_CHROOT:
3952 	case PRIV_VFS_RETAINSUGID:
3953 	case PRIV_VFS_FCHROOT:
3954 	case PRIV_VFS_LINK:
3955 	case PRIV_VFS_SETGID:
3956 	case PRIV_VFS_STAT:
3957 	case PRIV_VFS_STICKYFILE:
3958 
3959 		/*
3960 		 * As in the non-jail case, non-root users are expected to be
3961 		 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3962 		 * exists in the jail and they have permission to access it).
3963 		 */
3964 	case PRIV_KMEM_READ:
3965 		return (0);
3966 
3967 		/*
3968 		 * Depending on the global setting, allow privilege of
3969 		 * setting system flags.
3970 		 */
3971 	case PRIV_VFS_SYSFLAGS:
3972 		if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3973 			return (0);
3974 		else
3975 			return (EPERM);
3976 
3977 		/*
3978 		 * Depending on the global setting, allow privilege of
3979 		 * mounting/unmounting file systems.
3980 		 */
3981 	case PRIV_VFS_MOUNT:
3982 	case PRIV_VFS_UNMOUNT:
3983 	case PRIV_VFS_MOUNT_NONUSER:
3984 	case PRIV_VFS_MOUNT_OWNER:
3985 		if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3986 		    cred->cr_prison->pr_enforce_statfs < 2)
3987 			return (0);
3988 		else
3989 			return (EPERM);
3990 
3991 		/*
3992 		 * Allow jailed root to bind reserved ports and reuse in-use
3993 		 * ports.
3994 		 */
3995 	case PRIV_NETINET_RESERVEDPORT:
3996 	case PRIV_NETINET_REUSEPORT:
3997 		return (0);
3998 
3999 		/*
4000 		 * Allow jailed root to set certian IPv4/6 (option) headers.
4001 		 */
4002 	case PRIV_NETINET_SETHDROPTS:
4003 		return (0);
4004 
4005 		/*
4006 		 * Conditionally allow creating raw sockets in jail.
4007 		 */
4008 	case PRIV_NETINET_RAW:
4009 		if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4010 			return (0);
4011 		else
4012 			return (EPERM);
4013 
4014 		/*
4015 		 * Since jail implements its own visibility limits on netstat
4016 		 * sysctls, allow getcred.  This allows identd to work in
4017 		 * jail.
4018 		 */
4019 	case PRIV_NETINET_GETCRED:
4020 		return (0);
4021 
4022 		/*
4023 		 * Allow jailed root to set loginclass.
4024 		 */
4025 	case PRIV_PROC_SETLOGINCLASS:
4026 		return (0);
4027 
4028 	default:
4029 		/*
4030 		 * In all remaining cases, deny the privilege request.  This
4031 		 * includes almost all network privileges, many system
4032 		 * configuration privileges.
4033 		 */
4034 		return (EPERM);
4035 	}
4036 }
4037 
4038 /*
4039  * Return the part of pr2's name that is relative to pr1, or the whole name
4040  * if it does not directly follow.
4041  */
4042 
4043 char *
prison_name(struct prison * pr1,struct prison * pr2)4044 prison_name(struct prison *pr1, struct prison *pr2)
4045 {
4046 	char *name;
4047 
4048 	/* Jails see themselves as "0" (if they see themselves at all). */
4049 	if (pr1 == pr2)
4050 		return "0";
4051 	name = pr2->pr_name;
4052 	if (prison_ischild(pr1, pr2)) {
4053 		/*
4054 		 * pr1 isn't locked (and allprison_lock may not be either)
4055 		 * so its length can't be counted on.  But the number of dots
4056 		 * can be counted on - and counted.
4057 		 */
4058 		for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4059 			name = strchr(name, '.') + 1;
4060 	}
4061 	return (name);
4062 }
4063 
4064 /*
4065  * Return the part of pr2's path that is relative to pr1, or the whole path
4066  * if it does not directly follow.
4067  */
4068 static char *
prison_path(struct prison * pr1,struct prison * pr2)4069 prison_path(struct prison *pr1, struct prison *pr2)
4070 {
4071 	char *path1, *path2;
4072 	int len1;
4073 
4074 	path1 = pr1->pr_path;
4075 	path2 = pr2->pr_path;
4076 	if (!strcmp(path1, "/"))
4077 		return (path2);
4078 	len1 = strlen(path1);
4079 	if (strncmp(path1, path2, len1))
4080 		return (path2);
4081 	if (path2[len1] == '\0')
4082 		return "/";
4083 	if (path2[len1] == '/')
4084 		return (path2 + len1);
4085 	return (path2);
4086 }
4087 
4088 
4089 /*
4090  * Jail-related sysctls.
4091  */
4092 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4093     "Jails");
4094 
4095 static int
sysctl_jail_list(SYSCTL_HANDLER_ARGS)4096 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4097 {
4098 	struct xprison *xp;
4099 	struct prison *pr, *cpr;
4100 #ifdef INET
4101 	struct in_addr *ip4 = NULL;
4102 	int ip4s = 0;
4103 #endif
4104 #ifdef INET6
4105 	struct in6_addr *ip6 = NULL;
4106 	int ip6s = 0;
4107 #endif
4108 	int descend, error;
4109 
4110 	xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4111 	pr = req->td->td_ucred->cr_prison;
4112 	error = 0;
4113 	sx_slock(&allprison_lock);
4114 	FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4115 #if defined(INET) || defined(INET6)
4116  again:
4117 #endif
4118 		mtx_lock(&cpr->pr_mtx);
4119 #ifdef INET
4120 		if (cpr->pr_ip4s > 0) {
4121 			if (ip4s < cpr->pr_ip4s) {
4122 				ip4s = cpr->pr_ip4s;
4123 				mtx_unlock(&cpr->pr_mtx);
4124 				ip4 = realloc(ip4, ip4s *
4125 				    sizeof(struct in_addr), M_TEMP, M_WAITOK);
4126 				goto again;
4127 			}
4128 			bcopy(cpr->pr_ip4, ip4,
4129 			    cpr->pr_ip4s * sizeof(struct in_addr));
4130 		}
4131 #endif
4132 #ifdef INET6
4133 		if (cpr->pr_ip6s > 0) {
4134 			if (ip6s < cpr->pr_ip6s) {
4135 				ip6s = cpr->pr_ip6s;
4136 				mtx_unlock(&cpr->pr_mtx);
4137 				ip6 = realloc(ip6, ip6s *
4138 				    sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4139 				goto again;
4140 			}
4141 			bcopy(cpr->pr_ip6, ip6,
4142 			    cpr->pr_ip6s * sizeof(struct in6_addr));
4143 		}
4144 #endif
4145 		if (cpr->pr_ref == 0) {
4146 			mtx_unlock(&cpr->pr_mtx);
4147 			continue;
4148 		}
4149 		bzero(xp, sizeof(*xp));
4150 		xp->pr_version = XPRISON_VERSION;
4151 		xp->pr_id = cpr->pr_id;
4152 		xp->pr_state = cpr->pr_uref > 0
4153 		    ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4154 		strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4155 		strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4156 		strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4157 #ifdef INET
4158 		xp->pr_ip4s = cpr->pr_ip4s;
4159 #endif
4160 #ifdef INET6
4161 		xp->pr_ip6s = cpr->pr_ip6s;
4162 #endif
4163 		mtx_unlock(&cpr->pr_mtx);
4164 		error = SYSCTL_OUT(req, xp, sizeof(*xp));
4165 		if (error)
4166 			break;
4167 #ifdef INET
4168 		if (xp->pr_ip4s > 0) {
4169 			error = SYSCTL_OUT(req, ip4,
4170 			    xp->pr_ip4s * sizeof(struct in_addr));
4171 			if (error)
4172 				break;
4173 		}
4174 #endif
4175 #ifdef INET6
4176 		if (xp->pr_ip6s > 0) {
4177 			error = SYSCTL_OUT(req, ip6,
4178 			    xp->pr_ip6s * sizeof(struct in6_addr));
4179 			if (error)
4180 				break;
4181 		}
4182 #endif
4183 	}
4184 	sx_sunlock(&allprison_lock);
4185 	free(xp, M_TEMP);
4186 #ifdef INET
4187 	free(ip4, M_TEMP);
4188 #endif
4189 #ifdef INET6
4190 	free(ip6, M_TEMP);
4191 #endif
4192 	return (error);
4193 }
4194 
4195 SYSCTL_OID(_security_jail, OID_AUTO, list,
4196     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4197     sysctl_jail_list, "S", "List of active jails");
4198 
4199 static int
sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)4200 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4201 {
4202 	int error, injail;
4203 
4204 	injail = jailed(req->td->td_ucred);
4205 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
4206 
4207 	return (error);
4208 }
4209 
4210 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4211     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4212     sysctl_jail_jailed, "I", "Process in jail?");
4213 
4214 static int
sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)4215 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4216 {
4217 	int error, havevnet;
4218 #ifdef VIMAGE
4219 	struct ucred *cred = req->td->td_ucred;
4220 
4221 	havevnet = jailed(cred) && prison_owns_vnet(cred);
4222 #else
4223 	havevnet = 0;
4224 #endif
4225 	error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4226 
4227 	return (error);
4228 }
4229 
4230 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4231     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4232     sysctl_jail_vnet, "I", "Jail owns VNET?");
4233 
4234 #if defined(INET) || defined(INET6)
4235 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4236     &jail_max_af_ips, 0,
4237     "Number of IP addresses a jail may have at most per address family");
4238 #endif
4239 
4240 /*
4241  * Default parameters for jail(2) compatability.  For historical reasons,
4242  * the sysctl names have varying similarity to the parameter names.  Prisons
4243  * just see their own parameters, and can't change them.
4244  */
4245 static int
sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)4246 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4247 {
4248 	struct prison *pr;
4249 	int allow, error, i;
4250 
4251 	pr = req->td->td_ucred->cr_prison;
4252 	allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4253 
4254 	/* Get the current flag value, and convert it to a boolean. */
4255 	i = (allow & arg2) ? 1 : 0;
4256 	if (arg1 != NULL)
4257 		i = !i;
4258 	error = sysctl_handle_int(oidp, &i, 0, req);
4259 	if (error || !req->newptr)
4260 		return (error);
4261 	i = i ? arg2 : 0;
4262 	if (arg1 != NULL)
4263 		i ^= arg2;
4264 	/*
4265 	 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4266 	 * for writing.
4267 	 */
4268 	mtx_lock(&prison0.pr_mtx);
4269 	jail_default_allow = (jail_default_allow & ~arg2) | i;
4270 	mtx_unlock(&prison0.pr_mtx);
4271 	return (0);
4272 }
4273 
4274 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4275     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4276     NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4277     "Processes in jail can set their hostnames");
4278 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4279     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4280     (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4281     "Processes in jail are limited to creating UNIX/IP/route sockets only");
4282 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4283     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4284     NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4285     "Processes in jail can use System V IPC primitives");
4286 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4287     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4288     NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4289     "Prison root can create raw sockets");
4290 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4291     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4292     NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4293     "Processes in jail can alter system file flags");
4294 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4295     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4296     NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4297     "Processes in jail can mount/unmount jail-friendly file systems");
4298 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4299     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4300     NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4301     "Processes in jail can mount the devfs file system");
4302 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4303     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4304     NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4305     "Processes in jail can mount the fdescfs file system");
4306 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4307     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4308     NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4309     "Processes in jail can mount the nullfs file system");
4310 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4311     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4312     NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4313     "Processes in jail can mount the procfs file system");
4314 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4315     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4316     NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4317     "Processes in jail can mount the tmpfs file system");
4318 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4319     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4320     NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4321     "Processes in jail can mount the zfs file system");
4322 
4323 static int
sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)4324 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4325 {
4326 	struct prison *pr;
4327 	int level, error;
4328 
4329 	pr = req->td->td_ucred->cr_prison;
4330 	level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4331 	error = sysctl_handle_int(oidp, &level, 0, req);
4332 	if (error || !req->newptr)
4333 		return (error);
4334 	*(int *)arg1 = level;
4335 	return (0);
4336 }
4337 
4338 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4339     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4340     &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4341     sysctl_jail_default_level, "I",
4342     "Processes in jail cannot see all mounted file systems");
4343 
4344 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4345     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4346     &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4347     sysctl_jail_default_level, "I",
4348     "Ruleset for the devfs filesystem in jail");
4349 
4350 /*
4351  * Nodes to describe jail parameters.  Maximum length of string parameters
4352  * is returned in the string itself, and the other parameters exist merely
4353  * to make themselves and their types known.
4354  */
4355 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4356     "Jail parameters");
4357 
4358 int
sysctl_jail_param(SYSCTL_HANDLER_ARGS)4359 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4360 {
4361 	int i;
4362 	long l;
4363 	size_t s;
4364 	char numbuf[12];
4365 
4366 	switch (oidp->oid_kind & CTLTYPE)
4367 	{
4368 	case CTLTYPE_LONG:
4369 	case CTLTYPE_ULONG:
4370 		l = 0;
4371 #ifdef SCTL_MASK32
4372 		if (!(req->flags & SCTL_MASK32))
4373 #endif
4374 			return (SYSCTL_OUT(req, &l, sizeof(l)));
4375 	case CTLTYPE_INT:
4376 	case CTLTYPE_UINT:
4377 		i = 0;
4378 		return (SYSCTL_OUT(req, &i, sizeof(i)));
4379 	case CTLTYPE_STRING:
4380 		snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4381 		return
4382 		    (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4383 	case CTLTYPE_STRUCT:
4384 		s = (size_t)arg2;
4385 		return (SYSCTL_OUT(req, &s, sizeof(s)));
4386 	}
4387 	return (0);
4388 }
4389 
4390 /*
4391  * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4392  * jail creation time but cannot be changed in an existing jail.
4393  */
4394 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4395 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4396 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4397 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4398 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4399     "I", "Jail secure level");
4400 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4401     "Jail value for kern.osreldate and uname -K");
4402 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4403     "Jail value for kern.osrelease and uname -r");
4404 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4405     "I", "Jail cannot see all mounted file systems");
4406 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4407     "I", "Ruleset for in-jail devfs mounts");
4408 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4409     "B", "Jail persistence");
4410 #ifdef VIMAGE
4411 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4412     "E,jailsys", "Virtual network stack");
4413 #endif
4414 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4415     "B", "Jail is in the process of shutting down");
4416 
4417 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4418 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4419     "I", "Current number of child jails");
4420 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4421     "I", "Maximum number of child jails");
4422 
4423 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4424 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4425     "Jail hostname");
4426 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4427     "Jail NIS domainname");
4428 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4429     "Jail host UUID");
4430 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4431     "LU", "Jail host ID");
4432 
4433 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4434 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4435 
4436 #ifdef INET
4437 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4438     "Jail IPv4 address virtualization");
4439 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4440     "S,in_addr,a", "Jail IPv4 addresses");
4441 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4442     "B", "Do (not) use IPv4 source address selection rather than the "
4443     "primary jail IPv4 address.");
4444 #endif
4445 #ifdef INET6
4446 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4447     "Jail IPv6 address virtualization");
4448 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4449     "S,in6_addr,a", "Jail IPv6 addresses");
4450 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4451     "B", "Do (not) use IPv6 source address selection rather than the "
4452     "primary jail IPv6 address.");
4453 #endif
4454 
4455 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4456 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4457     "B", "Jail may set hostname");
4458 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4459     "B", "Jail may use SYSV IPC");
4460 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4461     "B", "Jail may create raw sockets");
4462 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4463     "B", "Jail may alter system file flags");
4464 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4465     "B", "Jail may set file quotas");
4466 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4467     "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4468 
4469 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4470 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4471     "B", "Jail may mount/unmount jail-friendly file systems in general");
4472 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4473     "B", "Jail may mount the devfs file system");
4474 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4475     "B", "Jail may mount the fdescfs file system");
4476 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4477     "B", "Jail may mount the nullfs file system");
4478 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4479     "B", "Jail may mount the procfs file system");
4480 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4481     "B", "Jail may mount the tmpfs file system");
4482 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4483     "B", "Jail may mount the zfs file system");
4484 
4485 void
prison_racct_foreach(void (* callback)(struct racct * racct,void * arg2,void * arg3),void * arg2,void * arg3)4486 prison_racct_foreach(void (*callback)(struct racct *racct,
4487     void *arg2, void *arg3), void *arg2, void *arg3)
4488 {
4489 	struct prison_racct *prr;
4490 
4491 	sx_slock(&allprison_lock);
4492 	LIST_FOREACH(prr, &allprison_racct, prr_next)
4493 		(callback)(prr->prr_racct, arg2, arg3);
4494 	sx_sunlock(&allprison_lock);
4495 }
4496 
4497 static struct prison_racct *
prison_racct_find_locked(const char * name)4498 prison_racct_find_locked(const char *name)
4499 {
4500 	struct prison_racct *prr;
4501 
4502 	sx_assert(&allprison_lock, SA_XLOCKED);
4503 
4504 	if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4505 		return (NULL);
4506 
4507 	LIST_FOREACH(prr, &allprison_racct, prr_next) {
4508 		if (strcmp(name, prr->prr_name) != 0)
4509 			continue;
4510 
4511 		/* Found prison_racct with a matching name? */
4512 		prison_racct_hold(prr);
4513 		return (prr);
4514 	}
4515 
4516 	/* Add new prison_racct. */
4517 	prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4518 	racct_create(&prr->prr_racct);
4519 
4520 	strcpy(prr->prr_name, name);
4521 	refcount_init(&prr->prr_refcount, 1);
4522 	LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4523 
4524 	return (prr);
4525 }
4526 
4527 struct prison_racct *
prison_racct_find(const char * name)4528 prison_racct_find(const char *name)
4529 {
4530 	struct prison_racct *prr;
4531 
4532 	sx_xlock(&allprison_lock);
4533 	prr = prison_racct_find_locked(name);
4534 	sx_xunlock(&allprison_lock);
4535 	return (prr);
4536 }
4537 
4538 void
prison_racct_hold(struct prison_racct * prr)4539 prison_racct_hold(struct prison_racct *prr)
4540 {
4541 
4542 	refcount_acquire(&prr->prr_refcount);
4543 }
4544 
4545 static void
prison_racct_free_locked(struct prison_racct * prr)4546 prison_racct_free_locked(struct prison_racct *prr)
4547 {
4548 
4549 	sx_assert(&allprison_lock, SA_XLOCKED);
4550 
4551 	if (refcount_release(&prr->prr_refcount)) {
4552 		racct_destroy(&prr->prr_racct);
4553 		LIST_REMOVE(prr, prr_next);
4554 		free(prr, M_PRISON_RACCT);
4555 	}
4556 }
4557 
4558 void
prison_racct_free(struct prison_racct * prr)4559 prison_racct_free(struct prison_racct *prr)
4560 {
4561 	int old;
4562 
4563 	sx_assert(&allprison_lock, SA_UNLOCKED);
4564 
4565 	old = prr->prr_refcount;
4566 	if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4567 		return;
4568 
4569 	sx_xlock(&allprison_lock);
4570 	prison_racct_free_locked(prr);
4571 	sx_xunlock(&allprison_lock);
4572 }
4573 
4574 #ifdef RACCT
4575 static void
prison_racct_attach(struct prison * pr)4576 prison_racct_attach(struct prison *pr)
4577 {
4578 	struct prison_racct *prr;
4579 
4580 	sx_assert(&allprison_lock, SA_XLOCKED);
4581 
4582 	prr = prison_racct_find_locked(pr->pr_name);
4583 	KASSERT(prr != NULL, ("cannot find prison_racct"));
4584 
4585 	pr->pr_prison_racct = prr;
4586 }
4587 
4588 /*
4589  * Handle jail renaming.  From the racct point of view, renaming means
4590  * moving from one prison_racct to another.
4591  */
4592 static void
prison_racct_modify(struct prison * pr)4593 prison_racct_modify(struct prison *pr)
4594 {
4595 	struct proc *p;
4596 	struct ucred *cred;
4597 	struct prison_racct *oldprr;
4598 
4599 	sx_slock(&allproc_lock);
4600 	sx_xlock(&allprison_lock);
4601 
4602 	if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4603 		sx_xunlock(&allprison_lock);
4604 		sx_sunlock(&allproc_lock);
4605 		return;
4606 	}
4607 
4608 	oldprr = pr->pr_prison_racct;
4609 	pr->pr_prison_racct = NULL;
4610 
4611 	prison_racct_attach(pr);
4612 
4613 	/*
4614 	 * Move resource utilisation records.
4615 	 */
4616 	racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4617 
4618 	/*
4619 	 * Force rctl to reattach rules to processes.
4620 	 */
4621 	FOREACH_PROC_IN_SYSTEM(p) {
4622 		PROC_LOCK(p);
4623 		cred = crhold(p->p_ucred);
4624 		PROC_UNLOCK(p);
4625 		racct_proc_ucred_changed(p, cred, cred);
4626 		crfree(cred);
4627 	}
4628 
4629 	sx_sunlock(&allproc_lock);
4630 	prison_racct_free_locked(oldprr);
4631 	sx_xunlock(&allprison_lock);
4632 }
4633 
4634 static void
prison_racct_detach(struct prison * pr)4635 prison_racct_detach(struct prison *pr)
4636 {
4637 
4638 	sx_assert(&allprison_lock, SA_UNLOCKED);
4639 
4640 	if (pr->pr_prison_racct == NULL)
4641 		return;
4642 	prison_racct_free(pr->pr_prison_racct);
4643 	pr->pr_prison_racct = NULL;
4644 }
4645 #endif /* RACCT */
4646 
4647 #ifdef DDB
4648 
4649 static void
db_show_prison(struct prison * pr)4650 db_show_prison(struct prison *pr)
4651 {
4652 	int fi;
4653 #if defined(INET) || defined(INET6)
4654 	int ii;
4655 #endif
4656 	unsigned jsf;
4657 #ifdef INET6
4658 	char ip6buf[INET6_ADDRSTRLEN];
4659 #endif
4660 
4661 	db_printf("prison %p:\n", pr);
4662 	db_printf(" jid             = %d\n", pr->pr_id);
4663 	db_printf(" name            = %s\n", pr->pr_name);
4664 	db_printf(" parent          = %p\n", pr->pr_parent);
4665 	db_printf(" ref             = %d\n", pr->pr_ref);
4666 	db_printf(" uref            = %d\n", pr->pr_uref);
4667 	db_printf(" path            = %s\n", pr->pr_path);
4668 	db_printf(" cpuset          = %d\n", pr->pr_cpuset
4669 	    ? pr->pr_cpuset->cs_id : -1);
4670 #ifdef VIMAGE
4671 	db_printf(" vnet            = %p\n", pr->pr_vnet);
4672 #endif
4673 	db_printf(" root            = %p\n", pr->pr_root);
4674 	db_printf(" securelevel     = %d\n", pr->pr_securelevel);
4675 	db_printf(" devfs_rsnum     = %d\n", pr->pr_devfs_rsnum);
4676 	db_printf(" children.max    = %d\n", pr->pr_childmax);
4677 	db_printf(" children.cur    = %d\n", pr->pr_childcount);
4678 	db_printf(" child           = %p\n", LIST_FIRST(&pr->pr_children));
4679 	db_printf(" sibling         = %p\n", LIST_NEXT(pr, pr_sibling));
4680 	db_printf(" flags           = 0x%x", pr->pr_flags);
4681 	for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4682 	    fi++)
4683 		if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4684 			db_printf(" %s", pr_flag_names[fi]);
4685 	for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4686 	    fi++) {
4687 		jsf = pr->pr_flags &
4688 		    (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4689 		db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4690 		    pr_flag_jailsys[fi].disable &&
4691 		      (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4692 		    : (jsf == pr_flag_jailsys[fi].new) ? "new"
4693 		    : "inherit");
4694 	}
4695 	db_printf(" allow           = 0x%x", pr->pr_allow);
4696 	for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4697 	    fi++)
4698 		if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4699 			db_printf(" %s", pr_allow_names[fi]);
4700 	db_printf("\n");
4701 	db_printf(" enforce_statfs  = %d\n", pr->pr_enforce_statfs);
4702 	db_printf(" host.hostname   = %s\n", pr->pr_hostname);
4703 	db_printf(" host.domainname = %s\n", pr->pr_domainname);
4704 	db_printf(" host.hostuuid   = %s\n", pr->pr_hostuuid);
4705 	db_printf(" host.hostid     = %lu\n", pr->pr_hostid);
4706 #ifdef INET
4707 	db_printf(" ip4s            = %d\n", pr->pr_ip4s);
4708 	for (ii = 0; ii < pr->pr_ip4s; ii++)
4709 		db_printf(" %s %s\n",
4710 		    ii == 0 ? "ip4.addr        =" : "                 ",
4711 		    inet_ntoa(pr->pr_ip4[ii]));
4712 #endif
4713 #ifdef INET6
4714 	db_printf(" ip6s            = %d\n", pr->pr_ip6s);
4715 	for (ii = 0; ii < pr->pr_ip6s; ii++)
4716 		db_printf(" %s %s\n",
4717 		    ii == 0 ? "ip6.addr        =" : "                 ",
4718 		    ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4719 #endif
4720 }
4721 
DB_SHOW_COMMAND(prison,db_show_prison_command)4722 DB_SHOW_COMMAND(prison, db_show_prison_command)
4723 {
4724 	struct prison *pr;
4725 
4726 	if (!have_addr) {
4727 		/*
4728 		 * Show all prisons in the list, and prison0 which is not
4729 		 * listed.
4730 		 */
4731 		db_show_prison(&prison0);
4732 		if (!db_pager_quit) {
4733 			TAILQ_FOREACH(pr, &allprison, pr_list) {
4734 				db_show_prison(pr);
4735 				if (db_pager_quit)
4736 					break;
4737 			}
4738 		}
4739 		return;
4740 	}
4741 
4742 	if (addr == 0)
4743 		pr = &prison0;
4744 	else {
4745 		/* Look for a prison with the ID and with references. */
4746 		TAILQ_FOREACH(pr, &allprison, pr_list)
4747 			if (pr->pr_id == addr && pr->pr_ref > 0)
4748 				break;
4749 		if (pr == NULL)
4750 			/* Look again, without requiring a reference. */
4751 			TAILQ_FOREACH(pr, &allprison, pr_list)
4752 				if (pr->pr_id == addr)
4753 					break;
4754 		if (pr == NULL)
4755 			/* Assume address points to a valid prison. */
4756 			pr = (struct prison *)addr;
4757 	}
4758 	db_show_prison(pr);
4759 }
4760 
4761 #endif /* DDB */
4762