2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 1999 Poul-Henning Kamp.
5 * Copyright (c) 2008 Bjoern A. Zeeb.
6 * Copyright (c) 2009 James Gritton.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/kernel.h>
41 #include <sys/systm.h>
42 #include <sys/errno.h>
43 #include <sys/sysproto.h>
44 #include <sys/malloc.h>
48 #include <sys/taskqueue.h>
49 #include <sys/fcntl.h>
51 #include <sys/linker.h>
53 #include <sys/mutex.h>
54 #include <sys/racct.h>
56 #include <sys/refcount.h>
58 #include <sys/sysent.h>
59 #include <sys/namei.h>
60 #include <sys/mount.h>
61 #include <sys/queue.h>
62 #include <sys/socket.h>
63 #include <sys/syscallsubr.h>
64 #include <sys/sysctl.h>
66 #include <sys/vnode.h>
71 #include <netinet/in.h>
77 #include <security/mac/mac_framework.h>
79 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
80 #define PRISON0_HOSTUUID_MODULE "hostuuid"
82 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
83 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
85 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
90 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
94 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
96 #define _PR_IP_SADDRSEL 0
100 /* prison0 describes what is "real" about the system. */
101 struct prison prison0 = {
107 .pr_securelevel = -1,
109 .pr_childmax = JAIL_MAX,
110 .pr_hostuuid = DEFAULT_HOSTUUID,
111 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
113 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
115 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
117 .pr_allow = PR_ALLOW_ALL_STATIC,
119 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
126 struct jailsys_flags {
132 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
133 struct sx allprison_lock;
134 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
135 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
136 LIST_HEAD(, prison_racct) allprison_racct;
139 static int get_next_prid(struct prison **insprp);
140 static int do_jail_attach(struct thread *td, struct prison *pr);
141 static void prison_complete(void *context, int pending);
142 static void prison_deref(struct prison *pr, int flags);
143 static char *prison_path(struct prison *pr1, struct prison *pr2);
144 static void prison_remove_one(struct prison *pr);
146 static void prison_racct_attach(struct prison *pr);
147 static void prison_racct_modify(struct prison *pr);
148 static void prison_racct_detach(struct prison *pr);
151 /* Flags for prison_deref */
152 #define PD_DEREF 0x01
153 #define PD_DEUREF 0x02
154 #define PD_LOCKED 0x04
155 #define PD_LIST_SLOCKED 0x08
156 #define PD_LIST_XLOCKED 0x10
159 * Parameter names corresponding to PR_* flag values. Size values are for kvm
160 * as we cannot figure out the size of a sparse array, or an array without a
163 static struct bool_flags pr_flag_bool[] = {
164 {"persist", "nopersist", PR_PERSIST},
166 {"ip4.saddrsel", "ip4.nosaddrsel", PR_IP4_SADDRSEL},
169 {"ip6.saddrsel", "ip6.nosaddrsel", PR_IP6_SADDRSEL},
172 const size_t pr_flag_bool_size = sizeof(pr_flag_bool);
174 static struct jailsys_flags pr_flag_jailsys[] = {
175 {"host", 0, PR_HOST},
177 {"vnet", 0, PR_VNET},
180 {"ip4", PR_IP4_USER, PR_IP4_USER},
183 {"ip6", PR_IP6_USER, PR_IP6_USER},
186 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
189 * Make this array full-size so dynamic parameters can be added.
190 * It is protected by prison0.mtx, but lockless reading is allowed
191 * with an atomic check of the flag values.
193 static struct bool_flags pr_flag_allow[NBBY * NBPW] = {
194 {"allow.set_hostname", "allow.noset_hostname", PR_ALLOW_SET_HOSTNAME},
195 {"allow.sysvipc", "allow.nosysvipc", PR_ALLOW_SYSVIPC},
196 {"allow.raw_sockets", "allow.noraw_sockets", PR_ALLOW_RAW_SOCKETS},
197 {"allow.chflags", "allow.nochflags", PR_ALLOW_CHFLAGS},
198 {"allow.mount", "allow.nomount", PR_ALLOW_MOUNT},
199 {"allow.quotas", "allow.noquotas", PR_ALLOW_QUOTAS},
200 {"allow.socket_af", "allow.nosocket_af", PR_ALLOW_SOCKET_AF},
201 {"allow.mlock", "allow.nomlock", PR_ALLOW_MLOCK},
202 {"allow.reserved_ports", "allow.noreserved_ports",
203 PR_ALLOW_RESERVED_PORTS},
204 {"allow.read_msgbuf", "allow.noread_msgbuf", PR_ALLOW_READ_MSGBUF},
205 {"allow.unprivileged_proc_debug", "allow.nounprivileged_proc_debug",
206 PR_ALLOW_UNPRIV_DEBUG},
207 {"allow.suser", "allow.nosuser", PR_ALLOW_SUSER},
209 static unsigned pr_allow_all = PR_ALLOW_ALL_STATIC;
210 const size_t pr_flag_allow_size = sizeof(pr_flag_allow);
212 #define JAIL_DEFAULT_ALLOW (PR_ALLOW_SET_HOSTNAME | \
213 PR_ALLOW_RESERVED_PORTS | \
214 PR_ALLOW_UNPRIV_DEBUG | \
216 #define JAIL_DEFAULT_ENFORCE_STATFS 2
217 #define JAIL_DEFAULT_DEVFS_RSNUM 0
218 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
219 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
220 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
221 #if defined(INET) || defined(INET6)
222 static unsigned jail_max_af_ips = 255;
226 * Initialize the parts of prison0 that can't be static-initialized with
227 * constants. This is called from proc0_init() after creating thread0 cpuset.
232 uint8_t *file, *data;
235 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
236 prison0.pr_osreldate = osreldate;
237 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
239 /* If we have a preloaded hostuuid, use it. */
240 file = preload_search_by_type(PRISON0_HOSTUUID_MODULE);
242 data = preload_fetch_addr(file);
243 size = preload_fetch_size(file);
246 * The preloaded data may include trailing whitespace, almost
247 * certainly a newline; skip over any whitespace or
248 * non-printable characters to be safe.
250 while (size > 0 && data[size - 1] <= 0x20) {
253 if (validate_uuid(data, size, NULL, 0) == 0) {
254 (void)strlcpy(prison0.pr_hostuuid, data,
256 } else if (bootverbose) {
257 printf("hostuuid: preload data malformed: '%s'",
263 printf("hostuuid: using %s\n", prison0.pr_hostuuid);
272 sys_jail(struct thread *td, struct jail_args *uap)
278 error = copyin(uap->jail, &version, sizeof(uint32_t));
287 /* FreeBSD single IPv4 jails. */
288 bzero(&j, sizeof(struct jail));
289 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
292 j.version = j0.version;
294 j.hostname = j0.hostname;
295 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
301 * Version 1 was used by multi-IPv4 jail implementations
302 * that never made it into the official kernel.
306 case 2: /* JAIL_API_VERSION */
307 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
308 error = copyin(uap->jail, &j, sizeof(struct jail));
314 /* Sci-Fi jails are not supported, sorry. */
317 return (kern_jail(td, &j));
321 kern_jail(struct thread *td, struct jail *j)
323 struct iovec optiov[2 * (4 + nitems(pr_flag_allow)
332 char *u_path, *u_hostname, *u_name;
333 struct bool_flags *bf;
336 struct in_addr *u_ip4;
339 struct in6_addr *u_ip6;
342 int error, enforce_statfs;
344 bzero(&optiov, sizeof(optiov));
345 opt.uio_iov = optiov;
349 opt.uio_segflg = UIO_SYSSPACE;
350 opt.uio_rw = UIO_READ;
353 /* Set permissions for top-level jails from sysctls. */
354 if (!jailed(td->td_ucred)) {
355 for (bf = pr_flag_allow;
356 bf < pr_flag_allow + nitems(pr_flag_allow) &&
357 atomic_load_int(&bf->flag) != 0;
359 optiov[opt.uio_iovcnt].iov_base = __DECONST(char *,
360 (jail_default_allow & bf->flag)
361 ? bf->name : bf->noname);
362 optiov[opt.uio_iovcnt].iov_len =
363 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
366 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
367 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
369 enforce_statfs = jail_default_enforce_statfs;
370 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
371 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
375 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
377 ip4s = (j->version == 0) ? 1 : j->ip4s;
378 if (ip4s > jail_max_af_ips)
380 tmplen += ip4s * sizeof(struct in_addr);
386 if (j->ip6s > jail_max_af_ips)
388 tmplen += j->ip6s * sizeof(struct in6_addr);
393 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
394 u_hostname = u_path + MAXPATHLEN;
395 u_name = u_hostname + MAXHOSTNAMELEN;
397 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
401 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
403 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
406 optiov[opt.uio_iovcnt].iov_base = "path";
407 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
409 optiov[opt.uio_iovcnt].iov_base = u_path;
410 error = copyinstr(j->path, u_path, MAXPATHLEN,
411 &optiov[opt.uio_iovcnt].iov_len);
413 free(u_path, M_TEMP);
417 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
418 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
420 optiov[opt.uio_iovcnt].iov_base = u_hostname;
421 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
422 &optiov[opt.uio_iovcnt].iov_len);
424 free(u_path, M_TEMP);
428 if (j->jailname != NULL) {
429 optiov[opt.uio_iovcnt].iov_base = "name";
430 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
432 optiov[opt.uio_iovcnt].iov_base = u_name;
433 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
434 &optiov[opt.uio_iovcnt].iov_len);
436 free(u_path, M_TEMP);
442 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
443 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
445 optiov[opt.uio_iovcnt].iov_base = u_ip4;
446 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
448 u_ip4->s_addr = j->ip4s;
450 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
452 free(u_path, M_TEMP);
459 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
460 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
462 optiov[opt.uio_iovcnt].iov_base = u_ip6;
463 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
464 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
466 free(u_path, M_TEMP);
471 KASSERT(opt.uio_iovcnt <= nitems(optiov),
472 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
473 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
474 free(u_path, M_TEMP);
479 * struct jail_set_args {
480 * struct iovec *iovp;
481 * unsigned int iovcnt;
486 sys_jail_set(struct thread *td, struct jail_set_args *uap)
491 /* Check that we have an even number of iovecs. */
495 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
498 error = kern_jail_set(td, auio, uap->flags);
504 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
511 struct in6_addr *ip6;
514 struct vfsoptlist *opts;
515 struct prison *pr, *deadpr, *inspr, *mypr, *ppr, *tpr;
517 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
518 char *g_path, *osrelstr;
519 struct bool_flags *bf;
520 struct jailsys_flags *jsf;
521 #if defined(INET) || defined(INET6)
526 size_t namelen, onamelen, pnamelen;
527 int born, created, cuflags, descend, enforce;
528 int error, errmsg_len, errmsg_pos;
529 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
530 int jid, jsys, len, level;
531 int childmax, osreldt, rsnum, slevel;
532 #if defined(INET) || defined(INET6)
541 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
542 uint64_t pr_allow_diff;
546 error = priv_check(td, PRIV_JAIL_SET);
547 if (!error && (flags & JAIL_ATTACH))
548 error = priv_check(td, PRIV_JAIL_ATTACH);
551 mypr = td->td_ucred->cr_prison;
552 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
554 if (flags & ~JAIL_SET_MASK)
558 * Check all the parameters before committing to anything. Not all
559 * errors can be caught early, but we may as well try. Also, this
560 * takes care of some expensive stuff (path lookup) before getting
561 * the allprison lock.
563 * XXX Jails are not filesystems, and jail parameters are not mount
564 * options. But it makes more sense to re-use the vfsopt code
565 * than duplicate it under a different name.
567 error = vfs_buildopts(optuio, &opts);
578 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
581 vfs_opterror(opts, "no valid operation (create or update)");
585 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
591 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
600 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
608 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
613 else if (enforce < 0 || enforce > 2) {
619 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
627 pr_flags = ch_flags = 0;
628 for (bf = pr_flag_bool;
629 bf < pr_flag_bool + nitems(pr_flag_bool);
631 vfs_flagopt(opts, bf->name, &pr_flags, bf->flag);
632 vfs_flagopt(opts, bf->noname, &ch_flags, bf->flag);
634 ch_flags |= pr_flags;
635 for (jsf = pr_flag_jailsys;
636 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
638 error = vfs_copyopt(opts, jsf->name, &jsys, sizeof(jsys));
644 case JAIL_SYS_DISABLE:
649 pr_flags |= jsf->disable;
652 pr_flags |= jsf->new;
654 case JAIL_SYS_INHERIT:
660 ch_flags |= jsf->new | jsf->disable;
662 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
663 && !(pr_flags & PR_PERSIST)) {
665 vfs_opterror(opts, "new jail must persist or attach");
669 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
671 vfs_opterror(opts, "vnet cannot be changed after creation");
676 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
678 vfs_opterror(opts, "ip4 cannot be changed after creation");
683 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
685 vfs_opterror(opts, "ip6 cannot be changed after creation");
690 pr_allow = ch_allow = 0;
691 for (bf = pr_flag_allow;
692 bf < pr_flag_allow + nitems(pr_flag_allow) &&
693 atomic_load_int(&bf->flag) != 0;
695 vfs_flagopt(opts, bf->name, &pr_allow, bf->flag);
696 vfs_flagopt(opts, bf->noname, &ch_allow, bf->flag);
698 ch_allow |= pr_allow;
700 error = vfs_getopt(opts, "name", (void **)&name, &len);
706 if (len == 0 || name[len - 1] != '\0') {
710 if (len > MAXHOSTNAMELEN) {
711 error = ENAMETOOLONG;
716 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
724 if (len == 0 || host[len - 1] != '\0') {
728 if (len > MAXHOSTNAMELEN) {
729 error = ENAMETOOLONG;
734 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
742 if (len == 0 || domain[len - 1] != '\0') {
746 if (len > MAXHOSTNAMELEN) {
747 error = ENAMETOOLONG;
752 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
760 if (len == 0 || uuid[len - 1] != '\0') {
764 if (len > HOSTUUIDLEN) {
765 error = ENAMETOOLONG;
770 #ifdef COMPAT_FREEBSD32
771 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
774 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
778 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
790 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
795 else if (ip4s & (sizeof(*ip4) - 1)) {
799 ch_flags |= PR_IP4_USER;
800 pr_flags |= PR_IP4_USER;
802 ip4s /= sizeof(*ip4);
803 if (ip4s > jail_max_af_ips) {
805 vfs_opterror(opts, "too many IPv4 addresses");
808 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
809 bcopy(op, ip4, ip4s * sizeof(*ip4));
811 * IP addresses are all sorted but ip[0] to preserve
812 * the primary IP address as given from userland.
813 * This special IP is used for unbound outgoing
814 * connections as well for "loopback" traffic in case
815 * source address selection cannot find any more fitting
816 * address to connect from.
819 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4),
822 * Check for duplicate addresses and do some simple
823 * zero and broadcast checks. If users give other bogus
824 * addresses it is their problem.
826 * We do not have to care about byte order for these
827 * checks so we will do them in NBO.
829 for (ii = 0; ii < ip4s; ii++) {
830 if (ip4[ii].s_addr == INADDR_ANY ||
831 ip4[ii].s_addr == INADDR_BROADCAST) {
836 (ip4[0].s_addr == ip4[ii+1].s_addr ||
837 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
847 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
852 else if (ip6s & (sizeof(*ip6) - 1)) {
856 ch_flags |= PR_IP6_USER;
857 pr_flags |= PR_IP6_USER;
859 ip6s /= sizeof(*ip6);
860 if (ip6s > jail_max_af_ips) {
862 vfs_opterror(opts, "too many IPv6 addresses");
865 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
866 bcopy(op, ip6, ip6s * sizeof(*ip6));
868 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6),
870 for (ii = 0; ii < ip6s; ii++) {
871 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
876 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
877 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
887 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
888 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
891 "vnet jails cannot have IP address restrictions");
896 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
902 if (flags & JAIL_UPDATE) {
905 "osrelease cannot be changed after creation");
908 if (len == 0 || osrelstr[len - 1] != '\0') {
912 if (len >= OSRELEASELEN) {
913 error = ENAMETOOLONG;
915 "osrelease string must be 1-%d bytes long",
921 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
927 if (flags & JAIL_UPDATE) {
930 "osreldate cannot be changed after creation");
935 vfs_opterror(opts, "osreldate cannot be 0");
941 error = vfs_getopt(opts, "path", (void **)&path, &len);
947 if (flags & JAIL_UPDATE) {
950 "path cannot be changed after creation");
953 if (len == 0 || path[len - 1] != '\0') {
957 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
963 NDFREE(&nd, NDF_ONLY_PNBUF);
964 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
965 strlcpy(g_path, path, MAXPATHLEN);
966 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
970 /* exit on other errors */
973 if (root->v_type != VDIR) {
982 * Find the specified jail, or at least its parent.
983 * This abuses the file error codes ENOENT and EEXIST.
988 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
989 namelc = strrchr(name, '.');
990 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
994 sx_xlock(&allprison_lock);
998 vfs_opterror(opts, "negative jid");
999 goto done_unlock_list;
1002 * See if a requested jid already exists. Keep track of
1003 * where it can be inserted later.
1005 TAILQ_FOREACH(inspr, &allprison, pr_list) {
1006 if (inspr->pr_id == jid) {
1007 mtx_lock(&inspr->pr_mtx);
1008 if (inspr->pr_ref > 0) {
1012 mtx_unlock(&inspr->pr_mtx);
1015 if (inspr->pr_id > jid)
1019 ppr = pr->pr_parent;
1020 /* Create: jid must not exist. */
1021 if (cuflags == JAIL_CREATE) {
1023 * Even creators that cannot see the jail will
1026 mtx_unlock(&pr->pr_mtx);
1028 vfs_opterror(opts, "jail %d already exists",
1030 goto done_unlock_list;
1032 if (!prison_ischild(mypr, pr)) {
1034 * Updaters get ENOENT if they cannot see the
1035 * jail. This is true even for CREATE | UPDATE,
1036 * which normally cannot give this error.
1038 mtx_unlock(&pr->pr_mtx);
1040 } else if (pr->pr_uref == 0) {
1041 if (!(flags & JAIL_DYING)) {
1042 mtx_unlock(&pr->pr_mtx);
1044 vfs_opterror(opts, "jail %d is dying",
1046 goto done_unlock_list;
1047 } else if ((flags & JAIL_ATTACH) ||
1048 (pr_flags & PR_PERSIST)) {
1050 * A dying jail might be resurrected
1051 * (via attach or persist), but first
1052 * it must determine if another jail
1053 * has claimed its name. Accomplish
1054 * this by implicitly re-setting the
1058 name = prison_name(mypr, pr);
1063 /* Update: jid must exist. */
1064 if (cuflags == JAIL_UPDATE) {
1066 vfs_opterror(opts, "jail %d not found", jid);
1067 goto done_unlock_list;
1072 * If the caller provided a name, look for a jail by that name.
1073 * This has different semantics for creates and updates keyed by jid
1074 * (where the name must not already exist in a different jail),
1075 * and updates keyed by the name itself (where the name must exist
1076 * because that is the jail being updated).
1080 namelc = strrchr(name, '.');
1085 * This is a hierarchical name. Split it into the
1086 * parent and child names, and make sure the parent
1087 * exists or matches an already found jail.
1090 if (strncmp(name, ppr->pr_name, namelc - name)
1091 || ppr->pr_name[namelc - name] != '\0') {
1092 mtx_unlock(&pr->pr_mtx);
1095 "cannot change jail's parent");
1096 goto done_unlock_list;
1100 ppr = prison_find_name(mypr, name);
1104 "jail \"%s\" not found", name);
1105 goto done_unlock_list;
1107 mtx_unlock(&ppr->pr_mtx);
1112 if (namelc[0] != '\0') {
1114 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1117 FOREACH_PRISON_CHILD(ppr, tpr) {
1118 if (tpr != pr && tpr->pr_ref > 0 &&
1119 !strcmp(tpr->pr_name + pnamelen, namelc)) {
1121 cuflags != JAIL_CREATE) {
1122 mtx_lock(&tpr->pr_mtx);
1123 if (tpr->pr_ref > 0) {
1128 if (tpr->pr_uref > 0) {
1134 mtx_unlock(&tpr->pr_mtx);
1135 } else if (tpr->pr_uref > 0) {
1137 * Create, or update(jid):
1138 * name must not exist in an
1139 * active sibling jail.
1143 mtx_unlock(&pr->pr_mtx);
1145 "jail \"%s\" already exists",
1147 goto done_unlock_list;
1151 /* If no active jail is found, use a dying one. */
1152 if (deadpr != NULL && pr == NULL) {
1153 if (flags & JAIL_DYING) {
1154 mtx_lock(&deadpr->pr_mtx);
1155 if (deadpr->pr_ref == 0) {
1156 mtx_unlock(&deadpr->pr_mtx);
1160 } else if (cuflags == JAIL_UPDATE) {
1163 "jail \"%s\" is dying", name);
1164 goto done_unlock_list;
1167 /* Update: name must exist if no jid. */
1168 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1170 vfs_opterror(opts, "jail \"%s\" not found",
1172 goto done_unlock_list;
1176 /* Update: must provide a jid or name. */
1177 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1179 vfs_opterror(opts, "update specified no jail");
1180 goto done_unlock_list;
1183 /* If there's no prison to update, create a new one and link it in. */
1185 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1186 if (tpr->pr_childcount >= tpr->pr_childmax) {
1188 vfs_opterror(opts, "prison limit exceeded");
1189 goto done_unlock_list;
1192 mtx_lock(&ppr->pr_mtx);
1193 if (ppr->pr_ref == 0) {
1194 mtx_unlock(&ppr->pr_mtx);
1196 vfs_opterror(opts, "jail \"%s\" not found",
1197 prison_name(mypr, ppr));
1198 goto done_unlock_list;
1202 mtx_unlock(&ppr->pr_mtx);
1203 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1205 if (jid == 0 && (jid = get_next_prid(&inspr)) == 0) {
1207 vfs_opterror(opts, "no available jail IDs");
1210 PD_DEREF | PD_DEUREF | PD_LIST_XLOCKED);
1215 TAILQ_INSERT_BEFORE(inspr, pr, pr_list);
1217 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1219 pr->pr_parent = ppr;
1220 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1221 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1222 tpr->pr_childcount++;
1224 /* Set some default values, and inherit some from the parent. */
1229 root = mypr->pr_root;
1232 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1233 pr->pr_flags |= PR_HOST;
1234 #if defined(INET) || defined(INET6)
1236 if (!(pr_flags & PR_VNET))
1240 if (!(ch_flags & PR_IP4_USER))
1241 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1242 else if (!(pr_flags & PR_IP4_USER)) {
1243 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1244 if (ppr->pr_ip4 != NULL) {
1245 pr->pr_ip4s = ppr->pr_ip4s;
1246 pr->pr_ip4 = malloc(pr->pr_ip4s *
1247 sizeof(struct in_addr), M_PRISON,
1249 bcopy(ppr->pr_ip4, pr->pr_ip4,
1250 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1255 if (!(ch_flags & PR_IP6_USER))
1256 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1257 else if (!(pr_flags & PR_IP6_USER)) {
1258 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1259 if (ppr->pr_ip6 != NULL) {
1260 pr->pr_ip6s = ppr->pr_ip6s;
1261 pr->pr_ip6 = malloc(pr->pr_ip6s *
1262 sizeof(struct in6_addr), M_PRISON,
1264 bcopy(ppr->pr_ip6, pr->pr_ip6,
1265 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1271 /* Source address selection is always on by default. */
1272 pr->pr_flags |= _PR_IP_SADDRSEL;
1274 pr->pr_securelevel = ppr->pr_securelevel;
1275 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1276 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1277 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1279 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1280 if (osrelstr == NULL)
1281 strlcpy(pr->pr_osrelease, ppr->pr_osrelease,
1282 sizeof(pr->pr_osrelease));
1284 strlcpy(pr->pr_osrelease, osrelstr,
1285 sizeof(pr->pr_osrelease));
1287 LIST_INIT(&pr->pr_children);
1288 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1289 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1292 /* Allocate a new vnet if specified. */
1293 pr->pr_vnet = (pr_flags & PR_VNET)
1294 ? vnet_alloc() : ppr->pr_vnet;
1297 * Allocate a dedicated cpuset for each jail.
1298 * Unlike other initial settings, this may return an erorr.
1300 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1302 prison_deref(pr, PD_LIST_XLOCKED);
1306 mtx_lock(&pr->pr_mtx);
1308 * New prisons do not yet have a reference, because we do not
1309 * want others to see the incomplete prison once the
1310 * allprison_lock is downgraded.
1315 * Grab a reference for existing prisons, to ensure they
1316 * continue to exist for the duration of the call.
1319 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1320 if ((pr->pr_flags & PR_VNET) &&
1321 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1324 "vnet jails cannot have IP address restrictions");
1325 goto done_deref_locked;
1329 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1332 "ip4 cannot be changed after creation");
1333 goto done_deref_locked;
1337 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1340 "ip6 cannot be changed after creation");
1341 goto done_deref_locked;
1346 /* Do final error checking before setting anything. */
1348 if (slevel < ppr->pr_securelevel) {
1350 goto done_deref_locked;
1354 if (childmax >= ppr->pr_childmax) {
1356 goto done_deref_locked;
1360 if (enforce < ppr->pr_enforce_statfs) {
1362 goto done_deref_locked;
1367 * devfs_rsnum is a uint16_t
1369 if (rsnum < 0 || rsnum > 65535) {
1371 goto done_deref_locked;
1374 * Nested jails always inherit parent's devfs ruleset
1376 if (jailed(td->td_ucred)) {
1377 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1379 goto done_deref_locked;
1381 rsnum = ppr->pr_devfs_rsnum;
1386 if (ppr->pr_flags & PR_IP4) {
1388 * Make sure the new set of IP addresses is a
1389 * subset of the parent's list. Don't worry
1390 * about the parent being unlocked, as any
1391 * setting is done with allprison_lock held.
1393 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1394 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1396 if (ij == ppr->pr_ip4s) {
1398 goto done_deref_locked;
1401 for (ii = ij = 1; ii < ip4s; ii++) {
1402 if (ip4[ii].s_addr ==
1403 ppr->pr_ip4[0].s_addr)
1405 for (; ij < ppr->pr_ip4s; ij++)
1406 if (ip4[ii].s_addr ==
1407 ppr->pr_ip4[ij].s_addr)
1409 if (ij == ppr->pr_ip4s)
1412 if (ij == ppr->pr_ip4s) {
1414 goto done_deref_locked;
1419 * Check for conflicting IP addresses. We permit them
1420 * if there is no more than one IP on each jail. If
1421 * there is a duplicate on a jail with more than one
1422 * IP stop checking and return error.
1425 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1426 if (tppr->pr_flags & PR_VNET)
1431 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1434 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1436 tpr->pr_uref == 0) {
1440 if (!(tpr->pr_flags & PR_IP4_USER))
1443 if (tpr->pr_ip4 == NULL ||
1444 (ip4s == 1 && tpr->pr_ip4s == 1))
1446 for (ii = 0; ii < ip4s; ii++) {
1447 if (prison_check_ip4_locked(tpr, &ip4[ii]) ==
1451 "IPv4 addresses clash");
1452 goto done_deref_locked;
1460 if (ppr->pr_flags & PR_IP6) {
1462 * Make sure the new set of IP addresses is a
1463 * subset of the parent's list.
1465 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1466 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1469 if (ij == ppr->pr_ip6s) {
1471 goto done_deref_locked;
1474 for (ii = ij = 1; ii < ip6s; ii++) {
1475 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1478 for (; ij < ppr->pr_ip6s; ij++)
1479 if (IN6_ARE_ADDR_EQUAL(
1480 &ip6[ii], &ppr->pr_ip6[ij]))
1482 if (ij == ppr->pr_ip6s)
1485 if (ij == ppr->pr_ip6s) {
1487 goto done_deref_locked;
1491 /* Check for conflicting IP addresses. */
1493 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1494 if (tppr->pr_flags & PR_VNET)
1499 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1502 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1504 tpr->pr_uref == 0) {
1508 if (!(tpr->pr_flags & PR_IP6_USER))
1511 if (tpr->pr_ip6 == NULL ||
1512 (ip6s == 1 && tpr->pr_ip6s == 1))
1514 for (ii = 0; ii < ip6s; ii++) {
1515 if (prison_check_ip6_locked(tpr, &ip6[ii]) ==
1519 "IPv6 addresses clash");
1520 goto done_deref_locked;
1526 onamelen = namelen = 0;
1527 if (namelc != NULL) {
1528 /* Give a default name of the jid. Also allow the name to be
1529 * explicitly the jid - but not any other number, and only in
1530 * normal form (no leading zero/etc).
1532 if (namelc[0] == '\0')
1533 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1534 else if ((strtoul(namelc, &p, 10) != jid ||
1535 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1538 "name cannot be numeric (unless it is the jid)");
1539 goto done_deref_locked;
1542 * Make sure the name isn't too long for the prison or its
1545 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1546 onamelen = strlen(pr->pr_name + pnamelen);
1547 namelen = strlen(namelc);
1548 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1549 error = ENAMETOOLONG;
1550 goto done_deref_locked;
1552 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1553 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1554 sizeof(pr->pr_name)) {
1555 error = ENAMETOOLONG;
1556 goto done_deref_locked;
1560 pr_allow_diff = pr_allow & ~ppr->pr_allow;
1561 if (pr_allow_diff & ~PR_ALLOW_DIFFERENCES) {
1563 goto done_deref_locked;
1567 * Let modules check their parameters. This requires unlocking and
1568 * then re-locking the prison, but this is still a valid state as long
1569 * as allprison_lock remains xlocked.
1571 mtx_unlock(&pr->pr_mtx);
1572 error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1574 prison_deref(pr, created
1576 : PD_DEREF | PD_LIST_XLOCKED);
1579 mtx_lock(&pr->pr_mtx);
1581 /* At this point, all valid parameters should have been noted. */
1582 TAILQ_FOREACH(opt, opts, link) {
1583 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1585 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1586 goto done_deref_locked;
1590 /* Set the parameters of the prison. */
1593 if (pr_flags & PR_IP4_USER) {
1594 pr->pr_flags |= PR_IP4;
1595 free(pr->pr_ip4, M_PRISON);
1599 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1601 if (tpr->pr_flags & PR_VNET) {
1606 if (prison_restrict_ip4(tpr, NULL)) {
1615 if (pr_flags & PR_IP6_USER) {
1616 pr->pr_flags |= PR_IP6;
1617 free(pr->pr_ip6, M_PRISON);
1621 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1623 if (tpr->pr_flags & PR_VNET) {
1628 if (prison_restrict_ip6(tpr, NULL)) {
1636 pr->pr_securelevel = slevel;
1637 /* Set all child jails to be at least this level. */
1638 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1639 if (tpr->pr_securelevel < slevel)
1640 tpr->pr_securelevel = slevel;
1643 pr->pr_childmax = childmax;
1644 /* Set all child jails to under this limit. */
1645 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1646 if (tpr->pr_childmax > childmax - level)
1647 tpr->pr_childmax = childmax > level
1648 ? childmax - level : 0;
1651 pr->pr_enforce_statfs = enforce;
1652 /* Pass this restriction on to the children. */
1653 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1654 if (tpr->pr_enforce_statfs < enforce)
1655 tpr->pr_enforce_statfs = enforce;
1658 pr->pr_devfs_rsnum = rsnum;
1659 /* Pass this restriction on to the children. */
1660 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1661 tpr->pr_devfs_rsnum = rsnum;
1663 if (namelc != NULL) {
1664 if (ppr == &prison0)
1665 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1667 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1668 ppr->pr_name, namelc);
1669 /* Change this component of child names. */
1670 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1671 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1672 strlen(tpr->pr_name + onamelen) + 1);
1673 bcopy(pr->pr_name, tpr->pr_name, namelen);
1677 /* Try to keep a real-rooted full pathname. */
1678 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1681 if (PR_HOST & ch_flags & ~pr_flags) {
1682 if (pr->pr_flags & PR_HOST) {
1684 * Copy the parent's host info. As with pr_ip4 above,
1685 * the lack of a lock on the parent is not a problem;
1686 * it is always set with allprison_lock at least
1687 * shared, and is held exclusively here.
1689 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1690 sizeof(pr->pr_hostname));
1691 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1692 sizeof(pr->pr_domainname));
1693 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1694 sizeof(pr->pr_hostuuid));
1695 pr->pr_hostid = pr->pr_parent->pr_hostid;
1697 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1698 /* Set this prison, and any descendants without PR_HOST. */
1700 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1702 strlcpy(pr->pr_domainname, domain,
1703 sizeof(pr->pr_domainname));
1705 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1707 pr->pr_hostid = hid;
1708 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1709 if (tpr->pr_flags & PR_HOST)
1713 strlcpy(tpr->pr_hostname,
1715 sizeof(tpr->pr_hostname));
1717 strlcpy(tpr->pr_domainname,
1719 sizeof(tpr->pr_domainname));
1721 strlcpy(tpr->pr_hostuuid,
1723 sizeof(tpr->pr_hostuuid));
1725 tpr->pr_hostid = hid;
1729 if ((tallow = ch_allow & ~pr_allow)) {
1730 /* Clear allow bits in all children. */
1731 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1732 tpr->pr_allow &= ~tallow;
1734 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1736 * Persistent prisons get an extra reference, and prisons losing their
1737 * persist flag lose that reference. Only do this for existing prisons
1738 * for now, so new ones will remain unseen until after the module
1739 * handlers have completed.
1741 born = pr->pr_uref == 0;
1742 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1743 if (pr_flags & PR_PERSIST) {
1751 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1752 mtx_unlock(&pr->pr_mtx);
1755 if (racct_enable && created)
1756 prison_racct_attach(pr);
1759 /* Locks may have prevented a complete restriction of child IP
1760 * addresses. If so, allocate some more memory and try again.
1765 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1766 mtx_lock(&pr->pr_mtx);
1768 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1770 if (tpr->pr_flags & PR_VNET) {
1775 if (prison_restrict_ip4(tpr, ip4)) {
1782 mtx_unlock(&pr->pr_mtx);
1788 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1789 mtx_lock(&pr->pr_mtx);
1791 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1793 if (tpr->pr_flags & PR_VNET) {
1798 if (prison_restrict_ip6(tpr, ip6)) {
1805 mtx_unlock(&pr->pr_mtx);
1809 /* Let the modules do their work. */
1810 sx_downgrade(&allprison_lock);
1812 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1814 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1815 prison_deref(pr, created
1817 : PD_DEREF | PD_LIST_SLOCKED);
1821 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1824 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1825 prison_deref(pr, created
1827 : PD_DEREF | PD_LIST_SLOCKED);
1831 /* Attach this process to the prison if requested. */
1832 if (flags & JAIL_ATTACH) {
1833 mtx_lock(&pr->pr_mtx);
1834 error = do_jail_attach(td, pr);
1836 vfs_opterror(opts, "attach failed");
1838 prison_deref(pr, PD_DEREF);
1844 if (racct_enable && !created) {
1845 if (!(flags & JAIL_ATTACH))
1846 sx_sunlock(&allprison_lock);
1847 prison_racct_modify(pr);
1848 if (!(flags & JAIL_ATTACH))
1849 sx_slock(&allprison_lock);
1853 td->td_retval[0] = pr->pr_id;
1856 * Now that it is all there, drop the temporary reference from existing
1857 * prisons. Or add a reference to newly created persistent prisons
1858 * (which was not done earlier so that the prison would not be publicly
1862 prison_deref(pr, (flags & JAIL_ATTACH)
1864 : PD_DEREF | PD_LIST_SLOCKED);
1866 if (pr_flags & PR_PERSIST) {
1867 mtx_lock(&pr->pr_mtx);
1870 mtx_unlock(&pr->pr_mtx);
1872 if (!(flags & JAIL_ATTACH))
1873 sx_sunlock(&allprison_lock);
1879 prison_deref(pr, created
1880 ? PD_LOCKED | PD_LIST_XLOCKED
1881 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1884 sx_xunlock(&allprison_lock);
1890 if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1891 &errmsg_len) == 0 && errmsg_len > 0) {
1892 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1893 if (optuio->uio_segflg == UIO_SYSSPACE)
1895 optuio->uio_iov[errmsg_pos].iov_base,
1899 optuio->uio_iov[errmsg_pos].iov_base,
1905 free(ip4, M_PRISON);
1908 free(ip6, M_PRISON);
1911 free(g_path, M_TEMP);
1917 * Find the next available prison ID. Return the ID on success, or zero
1918 * on failure. Also set a pointer to the allprison list entry the prison
1919 * should be inserted before.
1922 get_next_prid(struct prison **insprp)
1924 struct prison *inspr;
1927 jid = lastprid % JAIL_MAX + 1;
1928 if (TAILQ_EMPTY(&allprison) ||
1929 TAILQ_LAST(&allprison, prisonlist)->pr_id < jid) {
1931 * A common case is for all jails to be implicitly numbered,
1932 * which means they'll go on the end of the list, at least
1933 * for the first JAIL_MAX times.
1938 * Take two passes through the allprison list: first starting
1939 * with the proposed jid, then ending with it.
1941 for (maxid = JAIL_MAX; maxid != 0; ) {
1942 TAILQ_FOREACH(inspr, &allprison, pr_list) {
1943 if (inspr->pr_id < jid)
1945 if (inspr->pr_id > jid || inspr->pr_ref == 0) {
1947 * Found an opening. This may be a gap
1948 * in the list, or a dead jail with the
1954 if (++jid > maxid) {
1955 if (lastprid == maxid || lastprid == 0)
1958 * The entire legal range
1959 * has been traversed
1963 /* Try again from the start. */
1969 if (inspr == NULL) {
1970 /* Found room at the end of the list. */
1981 * struct jail_get_args {
1982 * struct iovec *iovp;
1983 * unsigned int iovcnt;
1988 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1993 /* Check that we have an even number of iovecs. */
1994 if (uap->iovcnt & 1)
1997 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
2000 error = kern_jail_get(td, auio, uap->flags);
2002 error = copyout(auio->uio_iov, uap->iovp,
2003 uap->iovcnt * sizeof (struct iovec));
2009 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
2011 struct bool_flags *bf;
2012 struct jailsys_flags *jsf;
2013 struct prison *pr, *mypr;
2015 struct vfsoptlist *opts;
2016 char *errmsg, *name;
2017 int error, errmsg_len, errmsg_pos, i, jid, len, locked, pos;
2020 if (flags & ~JAIL_GET_MASK)
2023 /* Get the parameter list. */
2024 error = vfs_buildopts(optuio, &opts);
2027 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
2028 mypr = td->td_ucred->cr_prison;
2031 * Find the prison specified by one of: lastjid, jid, name.
2033 sx_slock(&allprison_lock);
2034 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2036 TAILQ_FOREACH(pr, &allprison, pr_list) {
2037 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2038 mtx_lock(&pr->pr_mtx);
2039 if (pr->pr_ref > 0 &&
2040 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2042 mtx_unlock(&pr->pr_mtx);
2048 vfs_opterror(opts, "no jail after %d", jid);
2049 goto done_unlock_list;
2050 } else if (error != ENOENT)
2051 goto done_unlock_list;
2053 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2056 pr = prison_find_child(mypr, jid);
2058 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2059 mtx_unlock(&pr->pr_mtx);
2061 vfs_opterror(opts, "jail %d is dying",
2063 goto done_unlock_list;
2068 vfs_opterror(opts, "jail %d not found", jid);
2069 goto done_unlock_list;
2071 } else if (error != ENOENT)
2072 goto done_unlock_list;
2074 error = vfs_getopt(opts, "name", (void **)&name, &len);
2076 if (len == 0 || name[len - 1] != '\0') {
2078 goto done_unlock_list;
2080 pr = prison_find_name(mypr, name);
2082 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2083 mtx_unlock(&pr->pr_mtx);
2085 vfs_opterror(opts, "jail \"%s\" is dying",
2087 goto done_unlock_list;
2092 vfs_opterror(opts, "jail \"%s\" not found", name);
2093 goto done_unlock_list;
2094 } else if (error != ENOENT)
2095 goto done_unlock_list;
2097 vfs_opterror(opts, "no jail specified");
2099 goto done_unlock_list;
2102 /* Get the parameters of the prison. */
2105 td->td_retval[0] = pr->pr_id;
2106 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2107 if (error != 0 && error != ENOENT)
2109 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2110 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2111 if (error != 0 && error != ENOENT)
2113 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2114 if (error != 0 && error != ENOENT)
2116 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2117 sizeof(pr->pr_cpuset->cs_id));
2118 if (error != 0 && error != ENOENT)
2120 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2121 if (error != 0 && error != ENOENT)
2124 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2125 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2126 if (error != 0 && error != ENOENT)
2130 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2131 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2132 if (error != 0 && error != ENOENT)
2135 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2136 sizeof(pr->pr_securelevel));
2137 if (error != 0 && error != ENOENT)
2139 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2140 sizeof(pr->pr_childcount));
2141 if (error != 0 && error != ENOENT)
2143 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2144 sizeof(pr->pr_childmax));
2145 if (error != 0 && error != ENOENT)
2147 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2148 if (error != 0 && error != ENOENT)
2150 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2151 if (error != 0 && error != ENOENT)
2153 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2154 if (error != 0 && error != ENOENT)
2156 #ifdef COMPAT_FREEBSD32
2157 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2158 uint32_t hid32 = pr->pr_hostid;
2160 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2163 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2164 sizeof(pr->pr_hostid));
2165 if (error != 0 && error != ENOENT)
2167 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2168 sizeof(pr->pr_enforce_statfs));
2169 if (error != 0 && error != ENOENT)
2171 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2172 sizeof(pr->pr_devfs_rsnum));
2173 if (error != 0 && error != ENOENT)
2175 for (bf = pr_flag_bool;
2176 bf < pr_flag_bool + nitems(pr_flag_bool);
2178 i = (pr->pr_flags & bf->flag) ? 1 : 0;
2179 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2180 if (error != 0 && error != ENOENT)
2183 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2184 if (error != 0 && error != ENOENT)
2187 for (jsf = pr_flag_jailsys;
2188 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
2190 f = pr->pr_flags & (jsf->disable | jsf->new);
2191 i = (f != 0 && f == jsf->disable) ? JAIL_SYS_DISABLE
2192 : (f == jsf->new) ? JAIL_SYS_NEW
2194 error = vfs_setopt(opts, jsf->name, &i, sizeof(i));
2195 if (error != 0 && error != ENOENT)
2198 for (bf = pr_flag_allow;
2199 bf < pr_flag_allow + nitems(pr_flag_allow) &&
2200 atomic_load_int(&bf->flag) != 0;
2202 i = (pr->pr_allow & bf->flag) ? 1 : 0;
2203 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2204 if (error != 0 && error != ENOENT)
2207 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2208 if (error != 0 && error != ENOENT)
2211 i = (pr->pr_uref == 0);
2212 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2213 if (error != 0 && error != ENOENT)
2216 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2217 if (error != 0 && error != ENOENT)
2219 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2220 sizeof(pr->pr_osreldate));
2221 if (error != 0 && error != ENOENT)
2223 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2224 if (error != 0 && error != ENOENT)
2227 /* Get the module parameters. */
2228 mtx_unlock(&pr->pr_mtx);
2230 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2233 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2235 /* By now, all parameters should have been noted. */
2236 TAILQ_FOREACH(opt, opts, link) {
2237 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2239 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2244 /* Write the fetched parameters back to userspace. */
2246 TAILQ_FOREACH(opt, opts, link) {
2247 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2248 pos = 2 * opt->pos + 1;
2249 optuio->uio_iov[pos].iov_len = opt->len;
2250 if (opt->value != NULL) {
2251 if (optuio->uio_segflg == UIO_SYSSPACE) {
2253 optuio->uio_iov[pos].iov_base,
2256 error = copyout(opt->value,
2257 optuio->uio_iov[pos].iov_base,
2268 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2272 sx_sunlock(&allprison_lock);
2274 if (error && errmsg_pos >= 0) {
2275 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2276 errmsg_pos = 2 * errmsg_pos + 1;
2277 if (errmsg_len > 0) {
2278 if (optuio->uio_segflg == UIO_SYSSPACE)
2280 optuio->uio_iov[errmsg_pos].iov_base,
2284 optuio->uio_iov[errmsg_pos].iov_base,
2293 * struct jail_remove_args {
2298 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2300 struct prison *pr, *cpr, *lpr, *tpr;
2303 error = priv_check(td, PRIV_JAIL_REMOVE);
2307 sx_xlock(&allprison_lock);
2308 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2310 sx_xunlock(&allprison_lock);
2314 /* Remove all descendants of this prison, then remove this prison. */
2316 if (!LIST_EMPTY(&pr->pr_children)) {
2317 mtx_unlock(&pr->pr_mtx);
2319 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2320 mtx_lock(&cpr->pr_mtx);
2321 if (cpr->pr_ref > 0) {
2325 /* Already removed - do not do it again. */
2328 mtx_unlock(&cpr->pr_mtx);
2330 mtx_lock(&lpr->pr_mtx);
2331 prison_remove_one(lpr);
2332 sx_xlock(&allprison_lock);
2337 mtx_lock(&lpr->pr_mtx);
2338 prison_remove_one(lpr);
2339 sx_xlock(&allprison_lock);
2341 mtx_lock(&pr->pr_mtx);
2343 prison_remove_one(pr);
2348 prison_remove_one(struct prison *pr)
2353 /* If the prison was persistent, it is not anymore. */
2355 if (pr->pr_flags & PR_PERSIST) {
2358 pr->pr_flags &= ~PR_PERSIST;
2362 * jail_remove added a reference. If that's the only one, remove
2365 KASSERT(pr->pr_ref > 0,
2366 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2367 if (pr->pr_ref == 1) {
2369 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2373 mtx_unlock(&pr->pr_mtx);
2374 sx_xunlock(&allprison_lock);
2376 * Kill all processes unfortunate enough to be attached to this prison.
2378 sx_slock(&allproc_lock);
2379 FOREACH_PROC_IN_SYSTEM(p) {
2381 if (p->p_state != PRS_NEW && p->p_ucred &&
2382 p->p_ucred->cr_prison == pr)
2383 kern_psignal(p, SIGKILL);
2386 sx_sunlock(&allproc_lock);
2387 /* Remove the temporary reference added by jail_remove. */
2388 prison_deref(pr, deuref | PD_DEREF);
2392 * struct jail_attach_args {
2397 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2402 error = priv_check(td, PRIV_JAIL_ATTACH);
2407 * Start with exclusive hold on allprison_lock to ensure that a possible
2408 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2409 * But then immediately downgrade it since we don't need to stop
2412 sx_xlock(&allprison_lock);
2413 sx_downgrade(&allprison_lock);
2414 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2416 sx_sunlock(&allprison_lock);
2421 * Do not allow a process to attach to a prison that is not
2422 * considered to be "alive".
2424 if (pr->pr_uref == 0) {
2425 mtx_unlock(&pr->pr_mtx);
2426 sx_sunlock(&allprison_lock);
2430 return (do_jail_attach(td, pr));
2434 do_jail_attach(struct thread *td, struct prison *pr)
2437 struct ucred *newcred, *oldcred;
2441 * XXX: Note that there is a slight race here if two threads
2442 * in the same privileged process attempt to attach to two
2443 * different jails at the same time. It is important for
2444 * user processes not to do this, or they might end up with
2445 * a process root from one prison, but attached to the jail
2450 mtx_unlock(&pr->pr_mtx);
2452 /* Let modules do whatever they need to prepare for attaching. */
2453 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2455 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2458 sx_sunlock(&allprison_lock);
2461 * Reparent the newly attached process to this jail.
2464 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2468 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2469 if ((error = change_dir(pr->pr_root, td)) != 0)
2472 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2475 VOP_UNLOCK(pr->pr_root);
2476 if ((error = pwd_chroot(td, pr->pr_root)))
2481 oldcred = crcopysafe(p, newcred);
2482 newcred->cr_prison = pr;
2483 proc_set_cred(p, newcred);
2486 racct_proc_ucred_changed(p, oldcred, newcred);
2491 rctl_proc_ucred_changed(p, newcred);
2494 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2499 VOP_UNLOCK(pr->pr_root);
2501 /* Tell modules this thread is still in its old jail after all. */
2502 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2503 prison_deref(pr, PD_DEREF | PD_DEUREF);
2508 * Returns a locked prison instance, or NULL on failure.
2511 prison_find(int prid)
2515 sx_assert(&allprison_lock, SX_LOCKED);
2516 TAILQ_FOREACH(pr, &allprison, pr_list) {
2517 if (pr->pr_id == prid) {
2518 mtx_lock(&pr->pr_mtx);
2522 * Any active prison with the same ID would have
2523 * been inserted before a dead one.
2525 mtx_unlock(&pr->pr_mtx);
2528 if (pr->pr_id > prid)
2535 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2538 prison_find_child(struct prison *mypr, int prid)
2543 sx_assert(&allprison_lock, SX_LOCKED);
2544 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2545 if (pr->pr_id == prid) {
2546 mtx_lock(&pr->pr_mtx);
2549 mtx_unlock(&pr->pr_mtx);
2556 * Look for the name relative to mypr. Returns a locked prison or NULL.
2559 prison_find_name(struct prison *mypr, const char *name)
2561 struct prison *pr, *deadpr;
2565 sx_assert(&allprison_lock, SX_LOCKED);
2566 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2569 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2570 if (!strcmp(pr->pr_name + mylen, name)) {
2571 mtx_lock(&pr->pr_mtx);
2572 if (pr->pr_ref > 0) {
2573 if (pr->pr_uref > 0)
2577 mtx_unlock(&pr->pr_mtx);
2580 /* There was no valid prison - perhaps there was a dying one. */
2581 if (deadpr != NULL) {
2582 mtx_lock(&deadpr->pr_mtx);
2583 if (deadpr->pr_ref == 0) {
2584 mtx_unlock(&deadpr->pr_mtx);
2592 * See if a prison has the specific flag set.
2595 prison_flag(struct ucred *cred, unsigned flag)
2598 /* This is an atomic read, so no locking is necessary. */
2599 return (cred->cr_prison->pr_flags & flag);
2603 prison_allow(struct ucred *cred, unsigned flag)
2606 /* This is an atomic read, so no locking is necessary. */
2607 return (cred->cr_prison->pr_allow & flag);
2611 * Remove a prison reference. If that was the last reference, remove the
2612 * prison itself - but not in this context in case there are locks held.
2615 prison_free_locked(struct prison *pr)
2619 mtx_assert(&pr->pr_mtx, MA_OWNED);
2621 mtx_unlock(&pr->pr_mtx);
2623 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2627 prison_free(struct prison *pr)
2630 mtx_lock(&pr->pr_mtx);
2631 prison_free_locked(pr);
2635 * Complete a call to either prison_free or prison_proc_free.
2638 prison_complete(void *context, int pending)
2640 struct prison *pr = context;
2642 sx_xlock(&allprison_lock);
2643 mtx_lock(&pr->pr_mtx);
2644 prison_deref(pr, pr->pr_uref
2645 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2646 : PD_LOCKED | PD_LIST_XLOCKED);
2650 * Remove a prison reference (usually). This internal version assumes no
2651 * mutexes are held, except perhaps the prison itself. If there are no more
2652 * references, release and delist the prison. On completion, the prison lock
2653 * and the allprison lock are both unlocked.
2656 prison_deref(struct prison *pr, int flags)
2658 struct prison *ppr, *tpr;
2661 if (!(flags & PD_LOCKED))
2662 mtx_lock(&pr->pr_mtx);
2664 if (flags & PD_DEUREF) {
2665 KASSERT(pr->pr_uref > 0,
2666 ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2669 lasturef = pr->pr_uref == 0;
2672 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2675 if (flags & PD_DEREF) {
2676 KASSERT(pr->pr_ref > 0,
2677 ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2682 mtx_unlock(&pr->pr_mtx);
2685 * Tell the modules if the last user reference was removed
2686 * (even it sticks around in dying state).
2689 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2690 sx_xlock(&allprison_lock);
2691 flags |= PD_LIST_XLOCKED;
2693 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2694 mtx_lock(&pr->pr_mtx);
2696 mtx_unlock(&pr->pr_mtx);
2699 /* If the prison still has references, nothing else to do. */
2701 if (flags & PD_LIST_SLOCKED)
2702 sx_sunlock(&allprison_lock);
2703 else if (flags & PD_LIST_XLOCKED)
2704 sx_xunlock(&allprison_lock);
2708 if (flags & PD_LIST_SLOCKED) {
2709 if (!sx_try_upgrade(&allprison_lock)) {
2710 sx_sunlock(&allprison_lock);
2711 sx_xlock(&allprison_lock);
2713 } else if (!(flags & PD_LIST_XLOCKED))
2714 sx_xlock(&allprison_lock);
2716 TAILQ_REMOVE(&allprison, pr, pr_list);
2717 LIST_REMOVE(pr, pr_sibling);
2718 ppr = pr->pr_parent;
2719 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2720 tpr->pr_childcount--;
2721 sx_xunlock(&allprison_lock);
2724 if (pr->pr_vnet != ppr->pr_vnet)
2725 vnet_destroy(pr->pr_vnet);
2727 if (pr->pr_root != NULL)
2729 mtx_destroy(&pr->pr_mtx);
2731 free(pr->pr_ip4, M_PRISON);
2734 free(pr->pr_ip6, M_PRISON);
2736 if (pr->pr_cpuset != NULL)
2737 cpuset_rel(pr->pr_cpuset);
2741 prison_racct_detach(pr);
2745 /* Removing a prison frees a reference on its parent. */
2747 mtx_lock(&pr->pr_mtx);
2748 flags = PD_DEREF | PD_DEUREF;
2753 prison_hold_locked(struct prison *pr)
2756 mtx_assert(&pr->pr_mtx, MA_OWNED);
2757 KASSERT(pr->pr_ref > 0,
2758 ("Trying to hold dead prison %p (jid=%d).", pr, pr->pr_id));
2763 prison_hold(struct prison *pr)
2766 mtx_lock(&pr->pr_mtx);
2767 prison_hold_locked(pr);
2768 mtx_unlock(&pr->pr_mtx);
2772 prison_proc_hold(struct prison *pr)
2775 mtx_lock(&pr->pr_mtx);
2776 KASSERT(pr->pr_uref > 0,
2777 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2779 mtx_unlock(&pr->pr_mtx);
2783 prison_proc_free(struct prison *pr)
2786 mtx_lock(&pr->pr_mtx);
2787 KASSERT(pr->pr_uref > 0,
2788 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2789 if (pr->pr_uref > 1)
2793 * Don't remove the last user reference in this context, which
2794 * is expected to be a process that is not only locked, but
2798 mtx_unlock(&pr->pr_mtx);
2799 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2802 mtx_unlock(&pr->pr_mtx);
2806 * Check if a jail supports the given address family.
2808 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
2812 prison_check_af(struct ucred *cred, int af)
2817 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2819 pr = cred->cr_prison;
2821 /* Prisons with their own network stack are not limited. */
2822 if (prison_owns_vnet(cred))
2831 if (pr->pr_flags & PR_IP4)
2833 mtx_lock(&pr->pr_mtx);
2834 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
2835 error = EAFNOSUPPORT;
2836 mtx_unlock(&pr->pr_mtx);
2842 if (pr->pr_flags & PR_IP6)
2844 mtx_lock(&pr->pr_mtx);
2845 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
2846 error = EAFNOSUPPORT;
2847 mtx_unlock(&pr->pr_mtx);
2855 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
2856 error = EAFNOSUPPORT;
2862 * Check if given address belongs to the jail referenced by cred (wrapper to
2863 * prison_check_ip[46]).
2865 * Returns 0 if jail doesn't restrict the address family or if address belongs
2866 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
2867 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
2870 prison_if(struct ucred *cred, const struct sockaddr *sa)
2873 const struct sockaddr_in *sai;
2876 const struct sockaddr_in6 *sai6;
2880 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2881 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
2884 if (prison_owns_vnet(cred))
2889 switch (sa->sa_family)
2893 sai = (const struct sockaddr_in *)sa;
2894 error = prison_check_ip4(cred, &sai->sin_addr);
2899 sai6 = (const struct sockaddr_in6 *)sa;
2900 error = prison_check_ip6(cred, &sai6->sin6_addr);
2904 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
2905 error = EAFNOSUPPORT;
2911 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
2914 prison_check(struct ucred *cred1, struct ucred *cred2)
2917 return ((cred1->cr_prison == cred2->cr_prison ||
2918 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
2922 * Return 1 if p2 is a child of p1, otherwise 0.
2925 prison_ischild(struct prison *pr1, struct prison *pr2)
2928 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
2935 * Return 1 if the passed credential is in a jail and that jail does not
2936 * have its own virtual network stack, otherwise 0.
2939 jailed_without_vnet(struct ucred *cred)
2945 if (prison_owns_vnet(cred))
2953 * Return the correct hostname (domainname, et al) for the passed credential.
2956 getcredhostname(struct ucred *cred, char *buf, size_t size)
2961 * A NULL credential can be used to shortcut to the physical
2962 * system's hostname.
2964 pr = (cred != NULL) ? cred->cr_prison : &prison0;
2965 mtx_lock(&pr->pr_mtx);
2966 strlcpy(buf, pr->pr_hostname, size);
2967 mtx_unlock(&pr->pr_mtx);
2971 getcreddomainname(struct ucred *cred, char *buf, size_t size)
2974 mtx_lock(&cred->cr_prison->pr_mtx);
2975 strlcpy(buf, cred->cr_prison->pr_domainname, size);
2976 mtx_unlock(&cred->cr_prison->pr_mtx);
2980 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
2983 mtx_lock(&cred->cr_prison->pr_mtx);
2984 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
2985 mtx_unlock(&cred->cr_prison->pr_mtx);
2989 getcredhostid(struct ucred *cred, unsigned long *hostid)
2992 mtx_lock(&cred->cr_prison->pr_mtx);
2993 *hostid = cred->cr_prison->pr_hostid;
2994 mtx_unlock(&cred->cr_prison->pr_mtx);
2998 getjailname(struct ucred *cred, char *name, size_t len)
3001 mtx_lock(&cred->cr_prison->pr_mtx);
3002 strlcpy(name, cred->cr_prison->pr_name, len);
3003 mtx_unlock(&cred->cr_prison->pr_mtx);
3008 * Determine whether the prison represented by cred owns
3009 * its vnet rather than having it inherited.
3011 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3014 prison_owns_vnet(struct ucred *cred)
3018 * vnets cannot be added/removed after jail creation,
3019 * so no need to lock here.
3021 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3026 * Determine whether the subject represented by cred can "see"
3027 * status of a mount point.
3028 * Returns: 0 for permitted, ENOENT otherwise.
3029 * XXX: This function should be called cr_canseemount() and should be
3030 * placed in kern_prot.c.
3033 prison_canseemount(struct ucred *cred, struct mount *mp)
3039 pr = cred->cr_prison;
3040 if (pr->pr_enforce_statfs == 0)
3042 if (pr->pr_root->v_mount == mp)
3044 if (pr->pr_enforce_statfs == 2)
3047 * If jail's chroot directory is set to "/" we should be able to see
3048 * all mount-points from inside a jail.
3049 * This is ugly check, but this is the only situation when jail's
3050 * directory ends with '/'.
3052 if (strcmp(pr->pr_path, "/") == 0)
3054 len = strlen(pr->pr_path);
3056 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3059 * Be sure that we don't have situation where jail's root directory
3060 * is "/some/path" and mount point is "/some/pathpath".
3062 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3068 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3070 char jpath[MAXPATHLEN];
3074 pr = cred->cr_prison;
3075 if (pr->pr_enforce_statfs == 0)
3077 if (prison_canseemount(cred, mp) != 0) {
3078 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3079 strlcpy(sp->f_mntonname, "[restricted]",
3080 sizeof(sp->f_mntonname));
3083 if (pr->pr_root->v_mount == mp) {
3085 * Clear current buffer data, so we are sure nothing from
3086 * the valid path left there.
3088 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3089 *sp->f_mntonname = '/';
3093 * If jail's chroot directory is set to "/" we should be able to see
3094 * all mount-points from inside a jail.
3096 if (strcmp(pr->pr_path, "/") == 0)
3098 len = strlen(pr->pr_path);
3099 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3101 * Clear current buffer data, so we are sure nothing from
3102 * the valid path left there.
3104 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3105 if (*jpath == '\0') {
3106 /* Should never happen. */
3107 *sp->f_mntonname = '/';
3109 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3114 * Check with permission for a specific privilege is granted within jail. We
3115 * have a specific list of accepted privileges; the rest are denied.
3118 prison_priv_check(struct ucred *cred, int priv)
3122 * Some policies have custom handlers. This routine should not be
3123 * called for them. See priv_check_cred().
3126 case PRIV_VFS_LOOKUP:
3127 case PRIV_VFS_GENERATION:
3128 KASSERT(0, ("prison_priv_check instead of a custom handler "
3129 "called for %d\n", priv));
3137 * Privileges specific to prisons with a virtual network stack.
3138 * There might be a duplicate entry here in case the privilege
3139 * is only granted conditionally in the legacy jail case.
3144 * NFS-specific privileges.
3146 case PRIV_NFS_DAEMON:
3147 case PRIV_NFS_LOCKD:
3150 * Network stack privileges.
3152 case PRIV_NET_BRIDGE:
3155 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3156 case PRIV_NET_ROUTE:
3158 case PRIV_NET_SETIFMTU:
3159 case PRIV_NET_SETIFFLAGS:
3160 case PRIV_NET_SETIFCAP:
3161 case PRIV_NET_SETIFDESCR:
3162 case PRIV_NET_SETIFNAME :
3163 case PRIV_NET_SETIFMETRIC:
3164 case PRIV_NET_SETIFPHYS:
3165 case PRIV_NET_SETIFMAC:
3166 case PRIV_NET_SETLANPCP:
3167 case PRIV_NET_ADDMULTI:
3168 case PRIV_NET_DELMULTI:
3169 case PRIV_NET_HWIOCTL:
3170 case PRIV_NET_SETLLADDR:
3171 case PRIV_NET_ADDIFGROUP:
3172 case PRIV_NET_DELIFGROUP:
3173 case PRIV_NET_IFCREATE:
3174 case PRIV_NET_IFDESTROY:
3175 case PRIV_NET_ADDIFADDR:
3176 case PRIV_NET_DELIFADDR:
3179 case PRIV_NET_SETIFVNET:
3180 case PRIV_NET_SETIFFIB:
3183 * 802.11-related privileges.
3185 case PRIV_NET80211_VAP_GETKEY:
3186 case PRIV_NET80211_VAP_MANAGE:
3192 case PRIV_NETATM_CFG:
3193 case PRIV_NETATM_ADD:
3194 case PRIV_NETATM_DEL:
3195 case PRIV_NETATM_SET:
3198 * Bluetooth privileges.
3200 case PRIV_NETBLUETOOTH_RAW:
3204 * Netgraph and netgraph module privileges.
3206 case PRIV_NETGRAPH_CONTROL:
3208 case PRIV_NETGRAPH_TTY:
3212 * IPv4 and IPv6 privileges.
3214 case PRIV_NETINET_IPFW:
3215 case PRIV_NETINET_DIVERT:
3216 case PRIV_NETINET_PF:
3217 case PRIV_NETINET_DUMMYNET:
3218 case PRIV_NETINET_CARP:
3219 case PRIV_NETINET_MROUTE:
3220 case PRIV_NETINET_RAW:
3221 case PRIV_NETINET_ADDRCTRL6:
3222 case PRIV_NETINET_ND6:
3223 case PRIV_NETINET_SCOPE6:
3224 case PRIV_NETINET_ALIFETIME6:
3225 case PRIV_NETINET_IPSEC:
3226 case PRIV_NETINET_BINDANY:
3241 * No default: or deny here.
3242 * In case of no permit fall through to next switch().
3244 if (cred->cr_prison->pr_flags & PR_VNET)
3251 * Allow ktrace privileges for root in jail.
3257 * Allow jailed processes to configure audit identity and
3258 * submit audit records (login, etc). In the future we may
3259 * want to further refine the relationship between audit and
3262 case PRIV_AUDIT_GETAUDIT:
3263 case PRIV_AUDIT_SETAUDIT:
3264 case PRIV_AUDIT_SUBMIT:
3268 * Allow jailed processes to manipulate process UNIX
3269 * credentials in any way they see fit.
3271 case PRIV_CRED_SETUID:
3272 case PRIV_CRED_SETEUID:
3273 case PRIV_CRED_SETGID:
3274 case PRIV_CRED_SETEGID:
3275 case PRIV_CRED_SETGROUPS:
3276 case PRIV_CRED_SETREUID:
3277 case PRIV_CRED_SETREGID:
3278 case PRIV_CRED_SETRESUID:
3279 case PRIV_CRED_SETRESGID:
3282 * Jail implements visibility constraints already, so allow
3283 * jailed root to override uid/gid-based constraints.
3285 case PRIV_SEEOTHERGIDS:
3286 case PRIV_SEEOTHERUIDS:
3289 * Jail implements inter-process debugging limits already, so
3290 * allow jailed root various debugging privileges.
3292 case PRIV_DEBUG_DIFFCRED:
3293 case PRIV_DEBUG_SUGID:
3294 case PRIV_DEBUG_UNPRIV:
3297 * Allow jail to set various resource limits and login
3298 * properties, and for now, exceed process resource limits.
3300 case PRIV_PROC_LIMIT:
3301 case PRIV_PROC_SETLOGIN:
3302 case PRIV_PROC_SETRLIMIT:
3305 * System V and POSIX IPC privileges are granted in jail.
3308 case PRIV_IPC_WRITE:
3309 case PRIV_IPC_ADMIN:
3310 case PRIV_IPC_MSGSIZE:
3314 * Jail operations within a jail work on child jails.
3316 case PRIV_JAIL_ATTACH:
3318 case PRIV_JAIL_REMOVE:
3321 * Jail implements its own inter-process limits, so allow
3322 * root processes in jail to change scheduling on other
3323 * processes in the same jail. Likewise for signalling.
3325 case PRIV_SCHED_DIFFCRED:
3326 case PRIV_SCHED_CPUSET:
3327 case PRIV_SIGNAL_DIFFCRED:
3328 case PRIV_SIGNAL_SUGID:
3331 * Allow jailed processes to write to sysctls marked as jail
3334 case PRIV_SYSCTL_WRITEJAIL:
3337 * Allow root in jail to manage a variety of quota
3338 * properties. These should likely be conditional on a
3339 * configuration option.
3341 case PRIV_VFS_GETQUOTA:
3342 case PRIV_VFS_SETQUOTA:
3345 * Since Jail relies on chroot() to implement file system
3346 * protections, grant many VFS privileges to root in jail.
3347 * Be careful to exclude mount-related and NFS-related
3351 case PRIV_VFS_WRITE:
3352 case PRIV_VFS_ADMIN:
3354 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3355 case PRIV_VFS_CHFLAGS_DEV:
3356 case PRIV_VFS_CHOWN:
3357 case PRIV_VFS_CHROOT:
3358 case PRIV_VFS_RETAINSUGID:
3359 case PRIV_VFS_FCHROOT:
3361 case PRIV_VFS_SETGID:
3363 case PRIV_VFS_STICKYFILE:
3366 * As in the non-jail case, non-root users are expected to be
3367 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3368 * exists in the jail and they have permission to access it).
3370 case PRIV_KMEM_READ:
3374 * Depending on the global setting, allow privilege of
3375 * setting system flags.
3377 case PRIV_VFS_SYSFLAGS:
3378 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3384 * Depending on the global setting, allow privilege of
3385 * mounting/unmounting file systems.
3387 case PRIV_VFS_MOUNT:
3388 case PRIV_VFS_UNMOUNT:
3389 case PRIV_VFS_MOUNT_NONUSER:
3390 case PRIV_VFS_MOUNT_OWNER:
3391 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3392 cred->cr_prison->pr_enforce_statfs < 2)
3398 * Jails should hold no disposition on the PRIV_VFS_READ_DIR
3399 * policy. priv_check_cred will not specifically allow it, and
3400 * we may want a MAC policy to allow it.
3402 case PRIV_VFS_READ_DIR:
3406 * Conditionnaly allow locking (unlocking) physical pages
3410 case PRIV_VM_MUNLOCK:
3411 if (cred->cr_prison->pr_allow & PR_ALLOW_MLOCK)
3417 * Conditionally allow jailed root to bind reserved ports.
3419 case PRIV_NETINET_RESERVEDPORT:
3420 if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS)
3426 * Allow jailed root to reuse in-use ports.
3428 case PRIV_NETINET_REUSEPORT:
3432 * Allow jailed root to set certain IPv4/6 (option) headers.
3434 case PRIV_NETINET_SETHDROPTS:
3438 * Conditionally allow creating raw sockets in jail.
3440 case PRIV_NETINET_RAW:
3441 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3447 * Since jail implements its own visibility limits on netstat
3448 * sysctls, allow getcred. This allows identd to work in
3451 case PRIV_NETINET_GETCRED:
3455 * Allow jailed root to set loginclass.
3457 case PRIV_PROC_SETLOGINCLASS:
3461 * Do not allow a process inside a jail to read the kernel
3462 * message buffer unless explicitly permitted.
3465 if (cred->cr_prison->pr_allow & PR_ALLOW_READ_MSGBUF)
3471 * In all remaining cases, deny the privilege request. This
3472 * includes almost all network privileges, many system
3473 * configuration privileges.
3480 * Return the part of pr2's name that is relative to pr1, or the whole name
3481 * if it does not directly follow.
3485 prison_name(struct prison *pr1, struct prison *pr2)
3489 /* Jails see themselves as "0" (if they see themselves at all). */
3492 name = pr2->pr_name;
3493 if (prison_ischild(pr1, pr2)) {
3495 * pr1 isn't locked (and allprison_lock may not be either)
3496 * so its length can't be counted on. But the number of dots
3497 * can be counted on - and counted.
3499 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3500 name = strchr(name, '.') + 1;
3506 * Return the part of pr2's path that is relative to pr1, or the whole path
3507 * if it does not directly follow.
3510 prison_path(struct prison *pr1, struct prison *pr2)
3512 char *path1, *path2;
3515 path1 = pr1->pr_path;
3516 path2 = pr2->pr_path;
3517 if (!strcmp(path1, "/"))
3519 len1 = strlen(path1);
3520 if (strncmp(path1, path2, len1))
3522 if (path2[len1] == '\0')
3524 if (path2[len1] == '/')
3525 return (path2 + len1);
3530 * Jail-related sysctls.
3532 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
3536 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3539 struct prison *pr, *cpr;
3541 struct in_addr *ip4 = NULL;
3545 struct in6_addr *ip6 = NULL;
3550 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3551 pr = req->td->td_ucred->cr_prison;
3553 sx_slock(&allprison_lock);
3554 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3555 #if defined(INET) || defined(INET6)
3558 mtx_lock(&cpr->pr_mtx);
3560 if (cpr->pr_ip4s > 0) {
3561 if (ip4s < cpr->pr_ip4s) {
3562 ip4s = cpr->pr_ip4s;
3563 mtx_unlock(&cpr->pr_mtx);
3564 ip4 = realloc(ip4, ip4s *
3565 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3568 bcopy(cpr->pr_ip4, ip4,
3569 cpr->pr_ip4s * sizeof(struct in_addr));
3573 if (cpr->pr_ip6s > 0) {
3574 if (ip6s < cpr->pr_ip6s) {
3575 ip6s = cpr->pr_ip6s;
3576 mtx_unlock(&cpr->pr_mtx);
3577 ip6 = realloc(ip6, ip6s *
3578 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3581 bcopy(cpr->pr_ip6, ip6,
3582 cpr->pr_ip6s * sizeof(struct in6_addr));
3585 if (cpr->pr_ref == 0) {
3586 mtx_unlock(&cpr->pr_mtx);
3589 bzero(xp, sizeof(*xp));
3590 xp->pr_version = XPRISON_VERSION;
3591 xp->pr_id = cpr->pr_id;
3592 xp->pr_state = cpr->pr_uref > 0
3593 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3594 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3595 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3596 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3598 xp->pr_ip4s = cpr->pr_ip4s;
3601 xp->pr_ip6s = cpr->pr_ip6s;
3603 mtx_unlock(&cpr->pr_mtx);
3604 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3608 if (xp->pr_ip4s > 0) {
3609 error = SYSCTL_OUT(req, ip4,
3610 xp->pr_ip4s * sizeof(struct in_addr));
3616 if (xp->pr_ip6s > 0) {
3617 error = SYSCTL_OUT(req, ip6,
3618 xp->pr_ip6s * sizeof(struct in6_addr));
3624 sx_sunlock(&allprison_lock);
3635 SYSCTL_OID(_security_jail, OID_AUTO, list,
3636 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3637 sysctl_jail_list, "S", "List of active jails");
3640 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3644 injail = jailed(req->td->td_ucred);
3645 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3650 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3651 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3652 sysctl_jail_jailed, "I", "Process in jail?");
3655 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
3657 int error, havevnet;
3659 struct ucred *cred = req->td->td_ucred;
3661 havevnet = jailed(cred) && prison_owns_vnet(cred);
3665 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
3670 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
3671 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3672 sysctl_jail_vnet, "I", "Jail owns vnet?");
3674 #if defined(INET) || defined(INET6)
3675 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3676 &jail_max_af_ips, 0,
3677 "Number of IP addresses a jail may have at most per address family (deprecated)");
3681 * Default parameters for jail(2) compatibility. For historical reasons,
3682 * the sysctl names have varying similarity to the parameter names. Prisons
3683 * just see their own parameters, and can't change them.
3686 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3689 int allow, error, i;
3691 pr = req->td->td_ucred->cr_prison;
3692 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3694 /* Get the current flag value, and convert it to a boolean. */
3695 i = (allow & arg2) ? 1 : 0;
3698 error = sysctl_handle_int(oidp, &i, 0, req);
3699 if (error || !req->newptr)
3705 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3708 mtx_lock(&prison0.pr_mtx);
3709 jail_default_allow = (jail_default_allow & ~arg2) | i;
3710 mtx_unlock(&prison0.pr_mtx);
3714 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3715 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3716 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3717 "Processes in jail can set their hostnames (deprecated)");
3718 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3719 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3720 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3721 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
3722 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3723 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3724 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3725 "Processes in jail can use System V IPC primitives (deprecated)");
3726 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3727 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3728 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3729 "Prison root can create raw sockets (deprecated)");
3730 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3731 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3732 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3733 "Processes in jail can alter system file flags (deprecated)");
3734 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3735 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3736 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3737 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
3740 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3745 pr = req->td->td_ucred->cr_prison;
3746 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3747 error = sysctl_handle_int(oidp, &level, 0, req);
3748 if (error || !req->newptr)
3750 *(int *)arg1 = level;
3754 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3755 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3756 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3757 sysctl_jail_default_level, "I",
3758 "Processes in jail cannot see all mounted file systems (deprecated)");
3760 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
3761 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
3762 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
3763 sysctl_jail_default_level, "I",
3764 "Ruleset for the devfs filesystem in jail (deprecated)");
3767 * Nodes to describe jail parameters. Maximum length of string parameters
3768 * is returned in the string itself, and the other parameters exist merely
3769 * to make themselves and their types known.
3771 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
3775 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3782 switch (oidp->oid_kind & CTLTYPE)
3788 if (!(req->flags & SCTL_MASK32))
3790 return (SYSCTL_OUT(req, &l, sizeof(l)));
3794 return (SYSCTL_OUT(req, &i, sizeof(i)));
3795 case CTLTYPE_STRING:
3796 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
3798 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3799 case CTLTYPE_STRUCT:
3801 return (SYSCTL_OUT(req, &s, sizeof(s)));
3807 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
3808 * jail creation time but cannot be changed in an existing jail.
3810 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3811 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3812 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3813 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3814 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3815 "I", "Jail secure level");
3816 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
3817 "Jail value for kern.osreldate and uname -K");
3818 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
3819 "Jail value for kern.osrelease and uname -r");
3820 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3821 "I", "Jail cannot see all mounted file systems");
3822 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
3823 "I", "Ruleset for in-jail devfs mounts");
3824 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3825 "B", "Jail persistence");
3827 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3828 "E,jailsys", "Virtual network stack");
3830 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3831 "B", "Jail is in the process of shutting down");
3833 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
3834 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
3835 "I", "Current number of child jails");
3836 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
3837 "I", "Maximum number of child jails");
3839 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
3840 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3842 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3843 "Jail NIS domainname");
3844 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3846 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3847 "LU", "Jail host ID");
3849 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3850 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3853 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
3854 "Jail IPv4 address virtualization");
3855 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3856 "S,in_addr,a", "Jail IPv4 addresses");
3857 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3858 "B", "Do (not) use IPv4 source address selection rather than the "
3859 "primary jail IPv4 address.");
3862 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
3863 "Jail IPv6 address virtualization");
3864 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3865 "S,in6_addr,a", "Jail IPv6 addresses");
3866 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3867 "B", "Do (not) use IPv6 source address selection rather than the "
3868 "primary jail IPv6 address.");
3871 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3872 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3873 "B", "Jail may set hostname");
3874 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3875 "B", "Jail may use SYSV IPC");
3876 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3877 "B", "Jail may create raw sockets");
3878 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3879 "B", "Jail may alter system file flags");
3880 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3881 "B", "Jail may set file quotas");
3882 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3883 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3884 SYSCTL_JAIL_PARAM(_allow, mlock, CTLTYPE_INT | CTLFLAG_RW,
3885 "B", "Jail may lock (unlock) physical pages in memory");
3886 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW,
3887 "B", "Jail may bind sockets to reserved ports");
3888 SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW,
3889 "B", "Jail may read the kernel message buffer");
3890 SYSCTL_JAIL_PARAM(_allow, unprivileged_proc_debug, CTLTYPE_INT | CTLFLAG_RW,
3891 "B", "Unprivileged processes may use process debugging facilities");
3892 SYSCTL_JAIL_PARAM(_allow, suser, CTLTYPE_INT | CTLFLAG_RW,
3893 "B", "Processes in jail with uid 0 have privilege");
3895 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
3896 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
3897 "B", "Jail may mount/unmount jail-friendly file systems in general");
3900 * Add a dynamic parameter allow.<name>, or allow.<prefix>.<name>. Return
3901 * its associated bit in the pr_allow bitmask, or zero if the parameter was
3905 prison_add_allow(const char *prefix, const char *name, const char *prefix_descr,
3908 struct bool_flags *bf;
3909 struct sysctl_oid *parent;
3910 char *allow_name, *allow_noname, *allowed;
3911 #ifndef NO_SYSCTL_DESCR
3912 char *descr_deprecated;
3917 ? asprintf(&allow_name, M_PRISON, "allow.%s.%s", prefix, name)
3919 asprintf(&allow_noname, M_PRISON, "allow.%s.no%s", prefix, name)
3921 : asprintf(&allow_name, M_PRISON, "allow.%s", name) < 0 ||
3922 asprintf(&allow_noname, M_PRISON, "allow.no%s", name) < 0) {
3923 free(allow_name, M_PRISON);
3928 * See if this parameter has already beed added, i.e. a module was
3929 * previously loaded/unloaded.
3931 mtx_lock(&prison0.pr_mtx);
3932 for (bf = pr_flag_allow;
3933 bf < pr_flag_allow + nitems(pr_flag_allow) &&
3934 atomic_load_int(&bf->flag) != 0;
3936 if (strcmp(bf->name, allow_name) == 0) {
3937 allow_flag = bf->flag;
3943 * Find a free bit in pr_allow_all, failing if there are none
3944 * (which shouldn't happen as long as we keep track of how many
3945 * potential dynamic flags exist).
3947 for (allow_flag = 1;; allow_flag <<= 1) {
3948 if (allow_flag == 0)
3950 if ((pr_allow_all & allow_flag) == 0)
3954 /* Note the parameter in the next open slot in pr_flag_allow. */
3955 for (bf = pr_flag_allow; ; bf++) {
3956 if (bf == pr_flag_allow + nitems(pr_flag_allow)) {
3957 /* This should never happen, but is not fatal. */
3961 if (atomic_load_int(&bf->flag) == 0)
3964 bf->name = allow_name;
3965 bf->noname = allow_noname;
3966 pr_allow_all |= allow_flag;
3968 * prison0 always has permission for the new parameter.
3969 * Other jails must have it granted to them.
3971 prison0.pr_allow |= allow_flag;
3972 /* The flag indicates a valid entry, so make sure it is set last. */
3973 atomic_store_rel_int(&bf->flag, allow_flag);
3974 mtx_unlock(&prison0.pr_mtx);
3977 * Create sysctls for the paramter, and the back-compat global
3981 ? SYSCTL_ADD_NODE(NULL,
3982 SYSCTL_CHILDREN(&sysctl___security_jail_param_allow),
3983 OID_AUTO, prefix, CTLFLAG_MPSAFE, 0, prefix_descr)
3984 : &sysctl___security_jail_param_allow;
3985 (void)SYSCTL_ADD_PROC(NULL, SYSCTL_CHILDREN(parent), OID_AUTO,
3986 name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3987 NULL, 0, sysctl_jail_param, "B", descr);
3989 ? asprintf(&allowed, M_TEMP, "%s_%s_allowed", prefix, name)
3990 : asprintf(&allowed, M_TEMP, "%s_allowed", name)) >= 0) {
3991 #ifndef NO_SYSCTL_DESCR
3992 (void)asprintf(&descr_deprecated, M_TEMP, "%s (deprecated)",
3995 (void)SYSCTL_ADD_PROC(NULL,
3996 SYSCTL_CHILDREN(&sysctl___security_jail), OID_AUTO, allowed,
3997 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, allow_flag,
3998 sysctl_jail_default_allow, "I", descr_deprecated);
3999 #ifndef NO_SYSCTL_DESCR
4000 free(descr_deprecated, M_TEMP);
4002 free(allowed, M_TEMP);
4007 mtx_unlock(&prison0.pr_mtx);
4008 free(allow_name, M_PRISON);
4009 free(allow_noname, M_PRISON);
4014 * The VFS system will register jail-aware filesystems here. They each get
4015 * a parameter allow.mount.xxxfs and a flag to check when a jailed user
4016 * attempts to mount.
4019 prison_add_vfs(struct vfsconf *vfsp)
4021 #ifdef NO_SYSCTL_DESCR
4023 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
4028 (void)asprintf(&descr, M_TEMP, "Jail may mount the %s file system",
4030 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
4032 free(descr, M_TEMP);
4038 prison_racct_foreach(void (*callback)(struct racct *racct,
4039 void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
4040 void *arg2, void *arg3)
4042 struct prison_racct *prr;
4044 ASSERT_RACCT_ENABLED();
4046 sx_slock(&allprison_lock);
4049 LIST_FOREACH(prr, &allprison_racct, prr_next)
4050 (callback)(prr->prr_racct, arg2, arg3);
4053 sx_sunlock(&allprison_lock);
4056 static struct prison_racct *
4057 prison_racct_find_locked(const char *name)
4059 struct prison_racct *prr;
4061 ASSERT_RACCT_ENABLED();
4062 sx_assert(&allprison_lock, SA_XLOCKED);
4064 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4067 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4068 if (strcmp(name, prr->prr_name) != 0)
4071 /* Found prison_racct with a matching name? */
4072 prison_racct_hold(prr);
4076 /* Add new prison_racct. */
4077 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4078 racct_create(&prr->prr_racct);
4080 strcpy(prr->prr_name, name);
4081 refcount_init(&prr->prr_refcount, 1);
4082 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4087 struct prison_racct *
4088 prison_racct_find(const char *name)
4090 struct prison_racct *prr;
4092 ASSERT_RACCT_ENABLED();
4094 sx_xlock(&allprison_lock);
4095 prr = prison_racct_find_locked(name);
4096 sx_xunlock(&allprison_lock);
4101 prison_racct_hold(struct prison_racct *prr)
4104 ASSERT_RACCT_ENABLED();
4106 refcount_acquire(&prr->prr_refcount);
4110 prison_racct_free_locked(struct prison_racct *prr)
4113 ASSERT_RACCT_ENABLED();
4114 sx_assert(&allprison_lock, SA_XLOCKED);
4116 if (refcount_release(&prr->prr_refcount)) {
4117 racct_destroy(&prr->prr_racct);
4118 LIST_REMOVE(prr, prr_next);
4119 free(prr, M_PRISON_RACCT);
4124 prison_racct_free(struct prison_racct *prr)
4127 ASSERT_RACCT_ENABLED();
4128 sx_assert(&allprison_lock, SA_UNLOCKED);
4130 if (refcount_release_if_not_last(&prr->prr_refcount))
4133 sx_xlock(&allprison_lock);
4134 prison_racct_free_locked(prr);
4135 sx_xunlock(&allprison_lock);
4139 prison_racct_attach(struct prison *pr)
4141 struct prison_racct *prr;
4143 ASSERT_RACCT_ENABLED();
4144 sx_assert(&allprison_lock, SA_XLOCKED);
4146 prr = prison_racct_find_locked(pr->pr_name);
4147 KASSERT(prr != NULL, ("cannot find prison_racct"));
4149 pr->pr_prison_racct = prr;
4153 * Handle jail renaming. From the racct point of view, renaming means
4154 * moving from one prison_racct to another.
4157 prison_racct_modify(struct prison *pr)
4163 struct prison_racct *oldprr;
4165 ASSERT_RACCT_ENABLED();
4167 sx_slock(&allproc_lock);
4168 sx_xlock(&allprison_lock);
4170 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4171 sx_xunlock(&allprison_lock);
4172 sx_sunlock(&allproc_lock);
4176 oldprr = pr->pr_prison_racct;
4177 pr->pr_prison_racct = NULL;
4179 prison_racct_attach(pr);
4182 * Move resource utilisation records.
4184 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4188 * Force rctl to reattach rules to processes.
4190 FOREACH_PROC_IN_SYSTEM(p) {
4192 cred = crhold(p->p_ucred);
4194 rctl_proc_ucred_changed(p, cred);
4199 sx_sunlock(&allproc_lock);
4200 prison_racct_free_locked(oldprr);
4201 sx_xunlock(&allprison_lock);
4205 prison_racct_detach(struct prison *pr)
4208 ASSERT_RACCT_ENABLED();
4209 sx_assert(&allprison_lock, SA_UNLOCKED);
4211 if (pr->pr_prison_racct == NULL)
4213 prison_racct_free(pr->pr_prison_racct);
4214 pr->pr_prison_racct = NULL;
4221 db_show_prison(struct prison *pr)
4223 struct bool_flags *bf;
4224 struct jailsys_flags *jsf;
4225 #if defined(INET) || defined(INET6)
4230 char ip4buf[INET_ADDRSTRLEN];
4233 char ip6buf[INET6_ADDRSTRLEN];
4236 db_printf("prison %p:\n", pr);
4237 db_printf(" jid = %d\n", pr->pr_id);
4238 db_printf(" name = %s\n", pr->pr_name);
4239 db_printf(" parent = %p\n", pr->pr_parent);
4240 db_printf(" ref = %d\n", pr->pr_ref);
4241 db_printf(" uref = %d\n", pr->pr_uref);
4242 db_printf(" path = %s\n", pr->pr_path);
4243 db_printf(" cpuset = %d\n", pr->pr_cpuset
4244 ? pr->pr_cpuset->cs_id : -1);
4246 db_printf(" vnet = %p\n", pr->pr_vnet);
4248 db_printf(" root = %p\n", pr->pr_root);
4249 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4250 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4251 db_printf(" children.max = %d\n", pr->pr_childmax);
4252 db_printf(" children.cur = %d\n", pr->pr_childcount);
4253 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4254 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4255 db_printf(" flags = 0x%x", pr->pr_flags);
4256 for (bf = pr_flag_bool; bf < pr_flag_bool + nitems(pr_flag_bool); bf++)
4257 if (pr->pr_flags & bf->flag)
4258 db_printf(" %s", bf->name);
4259 for (jsf = pr_flag_jailsys;
4260 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
4262 f = pr->pr_flags & (jsf->disable | jsf->new);
4263 db_printf(" %-16s= %s\n", jsf->name,
4264 (f != 0 && f == jsf->disable) ? "disable"
4265 : (f == jsf->new) ? "new"
4268 db_printf(" allow = 0x%x", pr->pr_allow);
4269 for (bf = pr_flag_allow;
4270 bf < pr_flag_allow + nitems(pr_flag_allow) &&
4271 atomic_load_int(&bf->flag) != 0;
4273 if (pr->pr_allow & bf->flag)
4274 db_printf(" %s", bf->name);
4276 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4277 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4278 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4279 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4280 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4282 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4283 for (ii = 0; ii < pr->pr_ip4s; ii++)
4284 db_printf(" %s %s\n",
4285 ii == 0 ? "ip4.addr =" : " ",
4286 inet_ntoa_r(pr->pr_ip4[ii], ip4buf));
4289 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4290 for (ii = 0; ii < pr->pr_ip6s; ii++)
4291 db_printf(" %s %s\n",
4292 ii == 0 ? "ip6.addr =" : " ",
4293 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4297 DB_SHOW_COMMAND(prison, db_show_prison_command)
4303 * Show all prisons in the list, and prison0 which is not
4306 db_show_prison(&prison0);
4307 if (!db_pager_quit) {
4308 TAILQ_FOREACH(pr, &allprison, pr_list) {
4320 /* Look for a prison with the ID and with references. */
4321 TAILQ_FOREACH(pr, &allprison, pr_list)
4322 if (pr->pr_id == addr && pr->pr_ref > 0)
4325 /* Look again, without requiring a reference. */
4326 TAILQ_FOREACH(pr, &allprison, pr_list)
4327 if (pr->pr_id == addr)
4330 /* Assume address points to a valid prison. */
4331 pr = (struct prison *)addr;