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 void prison_set_allow_locked(struct prison *pr, unsigned flag,
145 static char *prison_path(struct prison *pr1, struct prison *pr2);
146 static void prison_remove_one(struct prison *pr);
148 static void prison_racct_attach(struct prison *pr);
149 static void prison_racct_modify(struct prison *pr);
150 static void prison_racct_detach(struct prison *pr);
153 /* Flags for prison_deref */
154 #define PD_DEREF 0x01
155 #define PD_DEUREF 0x02
156 #define PD_LOCKED 0x04
157 #define PD_LIST_SLOCKED 0x08
158 #define PD_LIST_XLOCKED 0x10
161 * Parameter names corresponding to PR_* flag values. Size values are for kvm
162 * as we cannot figure out the size of a sparse array, or an array without a
165 static struct bool_flags pr_flag_bool[] = {
166 {"persist", "nopersist", PR_PERSIST},
168 {"ip4.saddrsel", "ip4.nosaddrsel", PR_IP4_SADDRSEL},
171 {"ip6.saddrsel", "ip6.nosaddrsel", PR_IP6_SADDRSEL},
174 const size_t pr_flag_bool_size = sizeof(pr_flag_bool);
176 static struct jailsys_flags pr_flag_jailsys[] = {
177 {"host", 0, PR_HOST},
179 {"vnet", 0, PR_VNET},
182 {"ip4", PR_IP4_USER, PR_IP4_USER},
185 {"ip6", PR_IP6_USER, PR_IP6_USER},
188 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
191 * Make this array full-size so dynamic parameters can be added.
192 * It is protected by prison0.mtx, but lockless reading is allowed
193 * with an atomic check of the flag values.
195 static struct bool_flags pr_flag_allow[NBBY * NBPW] = {
196 {"allow.set_hostname", "allow.noset_hostname", PR_ALLOW_SET_HOSTNAME},
197 {"allow.sysvipc", "allow.nosysvipc", PR_ALLOW_SYSVIPC},
198 {"allow.raw_sockets", "allow.noraw_sockets", PR_ALLOW_RAW_SOCKETS},
199 {"allow.chflags", "allow.nochflags", PR_ALLOW_CHFLAGS},
200 {"allow.mount", "allow.nomount", PR_ALLOW_MOUNT},
201 {"allow.quotas", "allow.noquotas", PR_ALLOW_QUOTAS},
202 {"allow.socket_af", "allow.nosocket_af", PR_ALLOW_SOCKET_AF},
203 {"allow.mlock", "allow.nomlock", PR_ALLOW_MLOCK},
204 {"allow.reserved_ports", "allow.noreserved_ports",
205 PR_ALLOW_RESERVED_PORTS},
206 {"allow.read_msgbuf", "allow.noread_msgbuf", PR_ALLOW_READ_MSGBUF},
207 {"allow.unprivileged_proc_debug", "allow.nounprivileged_proc_debug",
208 PR_ALLOW_UNPRIV_DEBUG},
209 {"allow.suser", "allow.nosuser", PR_ALLOW_SUSER},
211 static unsigned pr_allow_all = PR_ALLOW_ALL_STATIC;
212 const size_t pr_flag_allow_size = sizeof(pr_flag_allow);
214 #define JAIL_DEFAULT_ALLOW (PR_ALLOW_SET_HOSTNAME | \
215 PR_ALLOW_RESERVED_PORTS | \
216 PR_ALLOW_UNPRIV_DEBUG | \
218 #define JAIL_DEFAULT_ENFORCE_STATFS 2
219 #define JAIL_DEFAULT_DEVFS_RSNUM 0
220 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
221 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
222 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
223 #if defined(INET) || defined(INET6)
224 static unsigned jail_max_af_ips = 255;
228 * Initialize the parts of prison0 that can't be static-initialized with
229 * constants. This is called from proc0_init() after creating thread0 cpuset.
234 uint8_t *file, *data;
237 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
238 prison0.pr_osreldate = osreldate;
239 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
241 /* If we have a preloaded hostuuid, use it. */
242 file = preload_search_by_type(PRISON0_HOSTUUID_MODULE);
244 data = preload_fetch_addr(file);
245 size = preload_fetch_size(file);
248 * The preloaded data may include trailing whitespace, almost
249 * certainly a newline; skip over any whitespace or
250 * non-printable characters to be safe.
252 while (size > 0 && data[size - 1] <= 0x20) {
255 if (validate_uuid(data, size, NULL, 0) == 0) {
256 (void)strlcpy(prison0.pr_hostuuid, data,
258 } else if (bootverbose) {
259 printf("hostuuid: preload data malformed: '%s'",
265 printf("hostuuid: using %s\n", prison0.pr_hostuuid);
274 sys_jail(struct thread *td, struct jail_args *uap)
280 error = copyin(uap->jail, &version, sizeof(uint32_t));
289 /* FreeBSD single IPv4 jails. */
290 bzero(&j, sizeof(struct jail));
291 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
294 j.version = j0.version;
296 j.hostname = j0.hostname;
297 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
303 * Version 1 was used by multi-IPv4 jail implementations
304 * that never made it into the official kernel.
308 case 2: /* JAIL_API_VERSION */
309 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
310 error = copyin(uap->jail, &j, sizeof(struct jail));
316 /* Sci-Fi jails are not supported, sorry. */
319 return (kern_jail(td, &j));
323 kern_jail(struct thread *td, struct jail *j)
325 struct iovec optiov[2 * (4 + nitems(pr_flag_allow)
334 char *u_path, *u_hostname, *u_name;
335 struct bool_flags *bf;
338 struct in_addr *u_ip4;
341 struct in6_addr *u_ip6;
344 int error, enforce_statfs;
346 bzero(&optiov, sizeof(optiov));
347 opt.uio_iov = optiov;
351 opt.uio_segflg = UIO_SYSSPACE;
352 opt.uio_rw = UIO_READ;
355 /* Set permissions for top-level jails from sysctls. */
356 if (!jailed(td->td_ucred)) {
357 for (bf = pr_flag_allow;
358 bf < pr_flag_allow + nitems(pr_flag_allow) &&
359 atomic_load_int(&bf->flag) != 0;
361 optiov[opt.uio_iovcnt].iov_base = __DECONST(char *,
362 (jail_default_allow & bf->flag)
363 ? bf->name : bf->noname);
364 optiov[opt.uio_iovcnt].iov_len =
365 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
368 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
369 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
371 enforce_statfs = jail_default_enforce_statfs;
372 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
373 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
377 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
379 ip4s = (j->version == 0) ? 1 : j->ip4s;
380 if (ip4s > jail_max_af_ips)
382 tmplen += ip4s * sizeof(struct in_addr);
388 if (j->ip6s > jail_max_af_ips)
390 tmplen += j->ip6s * sizeof(struct in6_addr);
395 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
396 u_hostname = u_path + MAXPATHLEN;
397 u_name = u_hostname + MAXHOSTNAMELEN;
399 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
403 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
405 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
408 optiov[opt.uio_iovcnt].iov_base = "path";
409 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
411 optiov[opt.uio_iovcnt].iov_base = u_path;
412 error = copyinstr(j->path, u_path, MAXPATHLEN,
413 &optiov[opt.uio_iovcnt].iov_len);
415 free(u_path, M_TEMP);
419 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
420 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
422 optiov[opt.uio_iovcnt].iov_base = u_hostname;
423 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
424 &optiov[opt.uio_iovcnt].iov_len);
426 free(u_path, M_TEMP);
430 if (j->jailname != NULL) {
431 optiov[opt.uio_iovcnt].iov_base = "name";
432 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
434 optiov[opt.uio_iovcnt].iov_base = u_name;
435 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
436 &optiov[opt.uio_iovcnt].iov_len);
438 free(u_path, M_TEMP);
444 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
445 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
447 optiov[opt.uio_iovcnt].iov_base = u_ip4;
448 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
450 u_ip4->s_addr = j->ip4s;
452 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
454 free(u_path, M_TEMP);
461 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
462 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
464 optiov[opt.uio_iovcnt].iov_base = u_ip6;
465 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
466 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
468 free(u_path, M_TEMP);
473 KASSERT(opt.uio_iovcnt <= nitems(optiov),
474 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
475 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
476 free(u_path, M_TEMP);
481 * struct jail_set_args {
482 * struct iovec *iovp;
483 * unsigned int iovcnt;
488 sys_jail_set(struct thread *td, struct jail_set_args *uap)
493 /* Check that we have an even number of iovecs. */
497 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
500 error = kern_jail_set(td, auio, uap->flags);
506 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
513 struct in6_addr *ip6;
516 struct vfsoptlist *opts;
517 struct prison *pr, *deadpr, *inspr, *mypr, *ppr, *tpr;
519 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
520 char *g_path, *osrelstr;
521 struct bool_flags *bf;
522 struct jailsys_flags *jsf;
523 #if defined(INET) || defined(INET6)
528 size_t namelen, onamelen, pnamelen;
529 int born, created, cuflags, descend, enforce, slocked;
530 int error, errmsg_len, errmsg_pos;
531 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
532 int jid, jsys, len, level;
533 int childmax, osreldt, rsnum, slevel;
534 #if defined(INET) || defined(INET6)
543 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
544 uint64_t pr_allow_diff;
548 error = priv_check(td, PRIV_JAIL_SET);
549 if (!error && (flags & JAIL_ATTACH))
550 error = priv_check(td, PRIV_JAIL_ATTACH);
553 mypr = td->td_ucred->cr_prison;
554 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
556 if (flags & ~JAIL_SET_MASK)
560 * Check all the parameters before committing to anything. Not all
561 * errors can be caught early, but we may as well try. Also, this
562 * takes care of some expensive stuff (path lookup) before getting
563 * the allprison lock.
565 * XXX Jails are not filesystems, and jail parameters are not mount
566 * options. But it makes more sense to re-use the vfsopt code
567 * than duplicate it under a different name.
569 error = vfs_buildopts(optuio, &opts);
580 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
583 vfs_opterror(opts, "no valid operation (create or update)");
587 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
593 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
602 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
610 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
615 else if (enforce < 0 || enforce > 2) {
621 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
629 pr_flags = ch_flags = 0;
630 for (bf = pr_flag_bool;
631 bf < pr_flag_bool + nitems(pr_flag_bool);
633 vfs_flagopt(opts, bf->name, &pr_flags, bf->flag);
634 vfs_flagopt(opts, bf->noname, &ch_flags, bf->flag);
636 ch_flags |= pr_flags;
637 for (jsf = pr_flag_jailsys;
638 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
640 error = vfs_copyopt(opts, jsf->name, &jsys, sizeof(jsys));
646 case JAIL_SYS_DISABLE:
651 pr_flags |= jsf->disable;
654 pr_flags |= jsf->new;
656 case JAIL_SYS_INHERIT:
662 ch_flags |= jsf->new | jsf->disable;
664 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
665 && !(pr_flags & PR_PERSIST)) {
667 vfs_opterror(opts, "new jail must persist or attach");
671 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
673 vfs_opterror(opts, "vnet cannot be changed after creation");
678 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
680 vfs_opterror(opts, "ip4 cannot be changed after creation");
685 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
687 vfs_opterror(opts, "ip6 cannot be changed after creation");
692 pr_allow = ch_allow = 0;
693 for (bf = pr_flag_allow;
694 bf < pr_flag_allow + nitems(pr_flag_allow) &&
695 atomic_load_int(&bf->flag) != 0;
697 vfs_flagopt(opts, bf->name, &pr_allow, bf->flag);
698 vfs_flagopt(opts, bf->noname, &ch_allow, bf->flag);
700 ch_allow |= pr_allow;
702 error = vfs_getopt(opts, "name", (void **)&name, &len);
708 if (len == 0 || name[len - 1] != '\0') {
712 if (len > MAXHOSTNAMELEN) {
713 error = ENAMETOOLONG;
718 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
726 if (len == 0 || host[len - 1] != '\0') {
730 if (len > MAXHOSTNAMELEN) {
731 error = ENAMETOOLONG;
736 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
744 if (len == 0 || domain[len - 1] != '\0') {
748 if (len > MAXHOSTNAMELEN) {
749 error = ENAMETOOLONG;
754 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
762 if (len == 0 || uuid[len - 1] != '\0') {
766 if (len > HOSTUUIDLEN) {
767 error = ENAMETOOLONG;
772 #ifdef COMPAT_FREEBSD32
773 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
776 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
780 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
792 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
797 else if (ip4s & (sizeof(*ip4) - 1)) {
801 ch_flags |= PR_IP4_USER;
802 pr_flags |= PR_IP4_USER;
804 ip4s /= sizeof(*ip4);
805 if (ip4s > jail_max_af_ips) {
807 vfs_opterror(opts, "too many IPv4 addresses");
810 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
811 bcopy(op, ip4, ip4s * sizeof(*ip4));
813 * IP addresses are all sorted but ip[0] to preserve
814 * the primary IP address as given from userland.
815 * This special IP is used for unbound outgoing
816 * connections as well for "loopback" traffic in case
817 * source address selection cannot find any more fitting
818 * address to connect from.
821 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4),
824 * Check for duplicate addresses and do some simple
825 * zero and broadcast checks. If users give other bogus
826 * addresses it is their problem.
828 * We do not have to care about byte order for these
829 * checks so we will do them in NBO.
831 for (ii = 0; ii < ip4s; ii++) {
832 if (ip4[ii].s_addr == INADDR_ANY ||
833 ip4[ii].s_addr == INADDR_BROADCAST) {
838 (ip4[0].s_addr == ip4[ii+1].s_addr ||
839 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
849 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
854 else if (ip6s & (sizeof(*ip6) - 1)) {
858 ch_flags |= PR_IP6_USER;
859 pr_flags |= PR_IP6_USER;
861 ip6s /= sizeof(*ip6);
862 if (ip6s > jail_max_af_ips) {
864 vfs_opterror(opts, "too many IPv6 addresses");
867 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
868 bcopy(op, ip6, ip6s * sizeof(*ip6));
870 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6),
872 for (ii = 0; ii < ip6s; ii++) {
873 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
878 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
879 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
889 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
890 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
893 "vnet jails cannot have IP address restrictions");
898 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
904 if (flags & JAIL_UPDATE) {
907 "osrelease cannot be changed after creation");
910 if (len == 0 || osrelstr[len - 1] != '\0') {
914 if (len >= OSRELEASELEN) {
915 error = ENAMETOOLONG;
917 "osrelease string must be 1-%d bytes long",
923 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
929 if (flags & JAIL_UPDATE) {
932 "osreldate cannot be changed after creation");
937 vfs_opterror(opts, "osreldate cannot be 0");
943 error = vfs_getopt(opts, "path", (void **)&path, &len);
949 if (flags & JAIL_UPDATE) {
952 "path cannot be changed after creation");
955 if (len == 0 || path[len - 1] != '\0') {
959 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
965 NDFREE(&nd, NDF_ONLY_PNBUF);
966 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
967 strlcpy(g_path, path, MAXPATHLEN);
968 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
972 /* exit on other errors */
975 if (root->v_type != VDIR) {
984 * Find the specified jail, or at least its parent.
985 * This abuses the file error codes ENOENT and EEXIST.
990 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
991 namelc = strrchr(name, '.');
992 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
996 sx_xlock(&allprison_lock);
1000 vfs_opterror(opts, "negative jid");
1001 goto done_unlock_list;
1004 * See if a requested jid already exists. Keep track of
1005 * where it can be inserted later.
1007 TAILQ_FOREACH(inspr, &allprison, pr_list) {
1008 if (inspr->pr_id == jid) {
1009 mtx_lock(&inspr->pr_mtx);
1010 if (inspr->pr_ref > 0) {
1014 mtx_unlock(&inspr->pr_mtx);
1017 if (inspr->pr_id > jid)
1021 ppr = pr->pr_parent;
1022 /* Create: jid must not exist. */
1023 if (cuflags == JAIL_CREATE) {
1025 * Even creators that cannot see the jail will
1028 mtx_unlock(&pr->pr_mtx);
1030 vfs_opterror(opts, "jail %d already exists",
1032 goto done_unlock_list;
1034 if (!prison_ischild(mypr, pr)) {
1036 * Updaters get ENOENT if they cannot see the
1037 * jail. This is true even for CREATE | UPDATE,
1038 * which normally cannot give this error.
1040 mtx_unlock(&pr->pr_mtx);
1042 } else if (pr->pr_uref == 0) {
1043 if (!(flags & JAIL_DYING)) {
1044 mtx_unlock(&pr->pr_mtx);
1046 vfs_opterror(opts, "jail %d is dying",
1048 goto done_unlock_list;
1049 } else if ((flags & JAIL_ATTACH) ||
1050 (pr_flags & PR_PERSIST)) {
1052 * A dying jail might be resurrected
1053 * (via attach or persist), but first
1054 * it must determine if another jail
1055 * has claimed its name. Accomplish
1056 * this by implicitly re-setting the
1060 name = prison_name(mypr, pr);
1065 /* Update: jid must exist. */
1066 if (cuflags == JAIL_UPDATE) {
1068 vfs_opterror(opts, "jail %d not found", jid);
1069 goto done_unlock_list;
1074 * If the caller provided a name, look for a jail by that name.
1075 * This has different semantics for creates and updates keyed by jid
1076 * (where the name must not already exist in a different jail),
1077 * and updates keyed by the name itself (where the name must exist
1078 * because that is the jail being updated).
1082 namelc = strrchr(name, '.');
1087 * This is a hierarchical name. Split it into the
1088 * parent and child names, and make sure the parent
1089 * exists or matches an already found jail.
1092 if (strncmp(name, ppr->pr_name, namelc - name)
1093 || ppr->pr_name[namelc - name] != '\0') {
1094 mtx_unlock(&pr->pr_mtx);
1097 "cannot change jail's parent");
1098 goto done_unlock_list;
1102 ppr = prison_find_name(mypr, name);
1106 "jail \"%s\" not found", name);
1107 goto done_unlock_list;
1109 mtx_unlock(&ppr->pr_mtx);
1114 if (namelc[0] != '\0') {
1116 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1119 FOREACH_PRISON_CHILD(ppr, tpr) {
1120 if (tpr != pr && tpr->pr_ref > 0 &&
1121 !strcmp(tpr->pr_name + pnamelen, namelc)) {
1123 cuflags != JAIL_CREATE) {
1124 mtx_lock(&tpr->pr_mtx);
1125 if (tpr->pr_ref > 0) {
1130 if (tpr->pr_uref > 0) {
1136 mtx_unlock(&tpr->pr_mtx);
1137 } else if (tpr->pr_uref > 0) {
1139 * Create, or update(jid):
1140 * name must not exist in an
1141 * active sibling jail.
1145 mtx_unlock(&pr->pr_mtx);
1147 "jail \"%s\" already exists",
1149 goto done_unlock_list;
1153 /* If no active jail is found, use a dying one. */
1154 if (deadpr != NULL && pr == NULL) {
1155 if (flags & JAIL_DYING) {
1156 mtx_lock(&deadpr->pr_mtx);
1157 if (deadpr->pr_ref == 0) {
1158 mtx_unlock(&deadpr->pr_mtx);
1162 } else if (cuflags == JAIL_UPDATE) {
1165 "jail \"%s\" is dying", name);
1166 goto done_unlock_list;
1169 /* Update: name must exist if no jid. */
1170 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1172 vfs_opterror(opts, "jail \"%s\" not found",
1174 goto done_unlock_list;
1178 /* Update: must provide a jid or name. */
1179 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1181 vfs_opterror(opts, "update specified no jail");
1182 goto done_unlock_list;
1185 /* If there's no prison to update, create a new one and link it in. */
1187 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1188 if (tpr->pr_childcount >= tpr->pr_childmax) {
1190 vfs_opterror(opts, "prison limit exceeded");
1191 goto done_unlock_list;
1194 mtx_lock(&ppr->pr_mtx);
1195 if (ppr->pr_ref == 0) {
1196 mtx_unlock(&ppr->pr_mtx);
1198 vfs_opterror(opts, "jail \"%s\" not found",
1199 prison_name(mypr, ppr));
1200 goto done_unlock_list;
1204 mtx_unlock(&ppr->pr_mtx);
1205 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1207 if (jid == 0 && (jid = get_next_prid(&inspr)) == 0) {
1209 vfs_opterror(opts, "no available jail IDs");
1212 PD_DEREF | PD_DEUREF | PD_LIST_XLOCKED);
1217 TAILQ_INSERT_BEFORE(inspr, pr, pr_list);
1219 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1221 pr->pr_parent = ppr;
1222 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1223 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1224 tpr->pr_childcount++;
1226 /* Set some default values, and inherit some from the parent. */
1231 root = mypr->pr_root;
1234 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1235 pr->pr_flags |= PR_HOST;
1236 #if defined(INET) || defined(INET6)
1238 if (!(pr_flags & PR_VNET))
1242 if (!(ch_flags & PR_IP4_USER))
1243 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1244 else if (!(pr_flags & PR_IP4_USER)) {
1245 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1246 if (ppr->pr_ip4 != NULL) {
1247 pr->pr_ip4s = ppr->pr_ip4s;
1248 pr->pr_ip4 = malloc(pr->pr_ip4s *
1249 sizeof(struct in_addr), M_PRISON,
1251 bcopy(ppr->pr_ip4, pr->pr_ip4,
1252 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1257 if (!(ch_flags & PR_IP6_USER))
1258 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1259 else if (!(pr_flags & PR_IP6_USER)) {
1260 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1261 if (ppr->pr_ip6 != NULL) {
1262 pr->pr_ip6s = ppr->pr_ip6s;
1263 pr->pr_ip6 = malloc(pr->pr_ip6s *
1264 sizeof(struct in6_addr), M_PRISON,
1266 bcopy(ppr->pr_ip6, pr->pr_ip6,
1267 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1273 /* Source address selection is always on by default. */
1274 pr->pr_flags |= _PR_IP_SADDRSEL;
1276 pr->pr_securelevel = ppr->pr_securelevel;
1277 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1278 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1279 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1281 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1282 if (osrelstr == NULL)
1283 strlcpy(pr->pr_osrelease, ppr->pr_osrelease,
1284 sizeof(pr->pr_osrelease));
1286 strlcpy(pr->pr_osrelease, osrelstr,
1287 sizeof(pr->pr_osrelease));
1289 LIST_INIT(&pr->pr_children);
1290 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1291 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1294 /* Allocate a new vnet if specified. */
1295 pr->pr_vnet = (pr_flags & PR_VNET)
1296 ? vnet_alloc() : ppr->pr_vnet;
1299 * Allocate a dedicated cpuset for each jail.
1300 * Unlike other initial settings, this may return an erorr.
1302 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1304 prison_deref(pr, PD_LIST_XLOCKED);
1308 mtx_lock(&pr->pr_mtx);
1310 * New prisons do not yet have a reference, because we do not
1311 * want others to see the incomplete prison once the
1312 * allprison_lock is downgraded.
1317 * Grab a reference for existing prisons, to ensure they
1318 * continue to exist for the duration of the call.
1321 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1322 if ((pr->pr_flags & PR_VNET) &&
1323 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1326 "vnet jails cannot have IP address restrictions");
1327 goto done_deref_locked;
1331 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1334 "ip4 cannot be changed after creation");
1335 goto done_deref_locked;
1339 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1342 "ip6 cannot be changed after creation");
1343 goto done_deref_locked;
1348 /* Do final error checking before setting anything. */
1350 if (slevel < ppr->pr_securelevel) {
1352 goto done_deref_locked;
1356 if (childmax >= ppr->pr_childmax) {
1358 goto done_deref_locked;
1362 if (enforce < ppr->pr_enforce_statfs) {
1364 goto done_deref_locked;
1369 * devfs_rsnum is a uint16_t
1371 if (rsnum < 0 || rsnum > 65535) {
1373 goto done_deref_locked;
1376 * Nested jails always inherit parent's devfs ruleset
1378 if (jailed(td->td_ucred)) {
1379 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1381 goto done_deref_locked;
1383 rsnum = ppr->pr_devfs_rsnum;
1388 if (ppr->pr_flags & PR_IP4) {
1390 * Make sure the new set of IP addresses is a
1391 * subset of the parent's list. Don't worry
1392 * about the parent being unlocked, as any
1393 * setting is done with allprison_lock held.
1395 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1396 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1398 if (ij == ppr->pr_ip4s) {
1400 goto done_deref_locked;
1403 for (ii = ij = 1; ii < ip4s; ii++) {
1404 if (ip4[ii].s_addr ==
1405 ppr->pr_ip4[0].s_addr)
1407 for (; ij < ppr->pr_ip4s; ij++)
1408 if (ip4[ii].s_addr ==
1409 ppr->pr_ip4[ij].s_addr)
1411 if (ij == ppr->pr_ip4s)
1414 if (ij == ppr->pr_ip4s) {
1416 goto done_deref_locked;
1421 * Check for conflicting IP addresses. We permit them
1422 * if there is no more than one IP on each jail. If
1423 * there is a duplicate on a jail with more than one
1424 * IP stop checking and return error.
1427 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1428 if (tppr->pr_flags & PR_VNET)
1433 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1436 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1438 tpr->pr_uref == 0) {
1442 if (!(tpr->pr_flags & PR_IP4_USER))
1445 if (tpr->pr_ip4 == NULL ||
1446 (ip4s == 1 && tpr->pr_ip4s == 1))
1448 for (ii = 0; ii < ip4s; ii++) {
1449 if (prison_check_ip4_locked(tpr, &ip4[ii]) ==
1453 "IPv4 addresses clash");
1454 goto done_deref_locked;
1462 if (ppr->pr_flags & PR_IP6) {
1464 * Make sure the new set of IP addresses is a
1465 * subset of the parent's list.
1467 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1468 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1471 if (ij == ppr->pr_ip6s) {
1473 goto done_deref_locked;
1476 for (ii = ij = 1; ii < ip6s; ii++) {
1477 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1480 for (; ij < ppr->pr_ip6s; ij++)
1481 if (IN6_ARE_ADDR_EQUAL(
1482 &ip6[ii], &ppr->pr_ip6[ij]))
1484 if (ij == ppr->pr_ip6s)
1487 if (ij == ppr->pr_ip6s) {
1489 goto done_deref_locked;
1493 /* Check for conflicting IP addresses. */
1495 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1496 if (tppr->pr_flags & PR_VNET)
1501 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1504 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1506 tpr->pr_uref == 0) {
1510 if (!(tpr->pr_flags & PR_IP6_USER))
1513 if (tpr->pr_ip6 == NULL ||
1514 (ip6s == 1 && tpr->pr_ip6s == 1))
1516 for (ii = 0; ii < ip6s; ii++) {
1517 if (prison_check_ip6_locked(tpr, &ip6[ii]) ==
1521 "IPv6 addresses clash");
1522 goto done_deref_locked;
1528 onamelen = namelen = 0;
1529 if (namelc != NULL) {
1530 /* Give a default name of the jid. Also allow the name to be
1531 * explicitly the jid - but not any other number, and only in
1532 * normal form (no leading zero/etc).
1534 if (namelc[0] == '\0')
1535 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1536 else if ((strtoul(namelc, &p, 10) != jid ||
1537 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1540 "name cannot be numeric (unless it is the jid)");
1541 goto done_deref_locked;
1544 * Make sure the name isn't too long for the prison or its
1547 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1548 onamelen = strlen(pr->pr_name + pnamelen);
1549 namelen = strlen(namelc);
1550 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1551 error = ENAMETOOLONG;
1552 goto done_deref_locked;
1554 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1555 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1556 sizeof(pr->pr_name)) {
1557 error = ENAMETOOLONG;
1558 goto done_deref_locked;
1562 pr_allow_diff = pr_allow & ~ppr->pr_allow;
1563 if (pr_allow_diff & ~PR_ALLOW_DIFFERENCES) {
1565 goto done_deref_locked;
1569 * Let modules check their parameters. This requires unlocking and
1570 * then re-locking the prison, but this is still a valid state as long
1571 * as allprison_lock remains xlocked.
1573 mtx_unlock(&pr->pr_mtx);
1574 error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1576 prison_deref(pr, created
1578 : PD_DEREF | PD_LIST_XLOCKED);
1581 mtx_lock(&pr->pr_mtx);
1583 /* At this point, all valid parameters should have been noted. */
1584 TAILQ_FOREACH(opt, opts, link) {
1585 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1587 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1588 goto done_deref_locked;
1592 /* Set the parameters of the prison. */
1595 if (pr_flags & PR_IP4_USER) {
1596 pr->pr_flags |= PR_IP4;
1597 free(pr->pr_ip4, M_PRISON);
1601 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1603 if (tpr->pr_flags & PR_VNET) {
1608 if (prison_restrict_ip4(tpr, NULL)) {
1617 if (pr_flags & PR_IP6_USER) {
1618 pr->pr_flags |= PR_IP6;
1619 free(pr->pr_ip6, M_PRISON);
1623 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1625 if (tpr->pr_flags & PR_VNET) {
1630 if (prison_restrict_ip6(tpr, NULL)) {
1638 pr->pr_securelevel = slevel;
1639 /* Set all child jails to be at least this level. */
1640 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1641 if (tpr->pr_securelevel < slevel)
1642 tpr->pr_securelevel = slevel;
1645 pr->pr_childmax = childmax;
1646 /* Set all child jails to under this limit. */
1647 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1648 if (tpr->pr_childmax > childmax - level)
1649 tpr->pr_childmax = childmax > level
1650 ? childmax - level : 0;
1653 pr->pr_enforce_statfs = enforce;
1654 /* Pass this restriction on to the children. */
1655 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1656 if (tpr->pr_enforce_statfs < enforce)
1657 tpr->pr_enforce_statfs = enforce;
1660 pr->pr_devfs_rsnum = rsnum;
1661 /* Pass this restriction on to the children. */
1662 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1663 tpr->pr_devfs_rsnum = rsnum;
1665 if (namelc != NULL) {
1666 if (ppr == &prison0)
1667 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1669 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1670 ppr->pr_name, namelc);
1671 /* Change this component of child names. */
1672 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1673 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1674 strlen(tpr->pr_name + onamelen) + 1);
1675 bcopy(pr->pr_name, tpr->pr_name, namelen);
1679 /* Try to keep a real-rooted full pathname. */
1680 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1683 if (PR_HOST & ch_flags & ~pr_flags) {
1684 if (pr->pr_flags & PR_HOST) {
1686 * Copy the parent's host info. As with pr_ip4 above,
1687 * the lack of a lock on the parent is not a problem;
1688 * it is always set with allprison_lock at least
1689 * shared, and is held exclusively here.
1691 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1692 sizeof(pr->pr_hostname));
1693 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1694 sizeof(pr->pr_domainname));
1695 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1696 sizeof(pr->pr_hostuuid));
1697 pr->pr_hostid = pr->pr_parent->pr_hostid;
1699 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1700 /* Set this prison, and any descendants without PR_HOST. */
1702 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1704 strlcpy(pr->pr_domainname, domain,
1705 sizeof(pr->pr_domainname));
1707 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1709 pr->pr_hostid = hid;
1710 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1711 if (tpr->pr_flags & PR_HOST)
1715 strlcpy(tpr->pr_hostname,
1717 sizeof(tpr->pr_hostname));
1719 strlcpy(tpr->pr_domainname,
1721 sizeof(tpr->pr_domainname));
1723 strlcpy(tpr->pr_hostuuid,
1725 sizeof(tpr->pr_hostuuid));
1727 tpr->pr_hostid = hid;
1731 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1732 if ((tallow = ch_allow & ~pr_allow))
1733 prison_set_allow_locked(pr, tallow, 0);
1735 * Persistent prisons get an extra reference, and prisons losing their
1736 * persist flag lose that reference. Only do this for existing prisons
1737 * for now, so new ones will remain unseen until after the module
1738 * handlers have completed.
1740 born = pr->pr_uref == 0;
1741 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1742 if (pr_flags & PR_PERSIST) {
1750 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1751 mtx_unlock(&pr->pr_mtx);
1754 if (racct_enable && created)
1755 prison_racct_attach(pr);
1758 /* Locks may have prevented a complete restriction of child IP
1759 * addresses. If so, allocate some more memory and try again.
1764 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1765 mtx_lock(&pr->pr_mtx);
1767 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1769 if (tpr->pr_flags & PR_VNET) {
1774 if (prison_restrict_ip4(tpr, ip4)) {
1781 mtx_unlock(&pr->pr_mtx);
1787 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1788 mtx_lock(&pr->pr_mtx);
1790 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1792 if (tpr->pr_flags & PR_VNET) {
1797 if (prison_restrict_ip6(tpr, ip6)) {
1804 mtx_unlock(&pr->pr_mtx);
1808 /* Let the modules do their work. */
1809 sx_downgrade(&allprison_lock);
1811 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1813 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1814 prison_deref(pr, created
1816 : PD_DEREF | PD_LIST_SLOCKED);
1820 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1823 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1824 prison_deref(pr, created
1826 : PD_DEREF | PD_LIST_SLOCKED);
1830 /* Attach this process to the prison if requested. */
1831 slocked = PD_LIST_SLOCKED;
1832 if (flags & JAIL_ATTACH) {
1833 mtx_lock(&pr->pr_mtx);
1834 error = do_jail_attach(td, pr);
1837 vfs_opterror(opts, "attach failed");
1839 prison_deref(pr, PD_DEREF);
1845 if (racct_enable && !created) {
1847 sx_sunlock(&allprison_lock);
1850 prison_racct_modify(pr);
1854 td->td_retval[0] = pr->pr_id;
1857 * Now that it is all there, drop the temporary reference from existing
1858 * prisons. Or add a reference to newly created persistent prisons
1859 * (which was not done earlier so that the prison would not be publicly
1863 prison_deref(pr, PD_DEREF | slocked);
1865 if (pr_flags & PR_PERSIST) {
1866 mtx_lock(&pr->pr_mtx);
1869 mtx_unlock(&pr->pr_mtx);
1872 sx_sunlock(&allprison_lock);
1878 prison_deref(pr, created
1879 ? PD_LOCKED | PD_LIST_XLOCKED
1880 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1883 sx_xunlock(&allprison_lock);
1889 if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1890 &errmsg_len) == 0 && errmsg_len > 0) {
1891 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1892 if (optuio->uio_segflg == UIO_SYSSPACE)
1894 optuio->uio_iov[errmsg_pos].iov_base,
1898 optuio->uio_iov[errmsg_pos].iov_base,
1904 free(ip4, M_PRISON);
1907 free(ip6, M_PRISON);
1910 free(g_path, M_TEMP);
1916 * Find the next available prison ID. Return the ID on success, or zero
1917 * on failure. Also set a pointer to the allprison list entry the prison
1918 * should be inserted before.
1921 get_next_prid(struct prison **insprp)
1923 struct prison *inspr;
1926 jid = lastprid % JAIL_MAX + 1;
1927 if (TAILQ_EMPTY(&allprison) ||
1928 TAILQ_LAST(&allprison, prisonlist)->pr_id < jid) {
1930 * A common case is for all jails to be implicitly numbered,
1931 * which means they'll go on the end of the list, at least
1932 * for the first JAIL_MAX times.
1937 * Take two passes through the allprison list: first starting
1938 * with the proposed jid, then ending with it.
1940 for (maxid = JAIL_MAX; maxid != 0; ) {
1941 TAILQ_FOREACH(inspr, &allprison, pr_list) {
1942 if (inspr->pr_id < jid)
1944 if (inspr->pr_id > jid || inspr->pr_ref == 0) {
1946 * Found an opening. This may be a gap
1947 * in the list, or a dead jail with the
1953 if (++jid > maxid) {
1954 if (lastprid == maxid || lastprid == 0)
1957 * The entire legal range
1958 * has been traversed
1962 /* Try again from the start. */
1968 if (inspr == NULL) {
1969 /* Found room at the end of the list. */
1980 * struct jail_get_args {
1981 * struct iovec *iovp;
1982 * unsigned int iovcnt;
1987 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1992 /* Check that we have an even number of iovecs. */
1993 if (uap->iovcnt & 1)
1996 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1999 error = kern_jail_get(td, auio, uap->flags);
2001 error = copyout(auio->uio_iov, uap->iovp,
2002 uap->iovcnt * sizeof (struct iovec));
2008 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
2010 struct bool_flags *bf;
2011 struct jailsys_flags *jsf;
2012 struct prison *pr, *mypr;
2014 struct vfsoptlist *opts;
2015 char *errmsg, *name;
2016 int error, errmsg_len, errmsg_pos, i, jid, len, locked, pos;
2019 if (flags & ~JAIL_GET_MASK)
2022 /* Get the parameter list. */
2023 error = vfs_buildopts(optuio, &opts);
2026 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
2027 mypr = td->td_ucred->cr_prison;
2030 * Find the prison specified by one of: lastjid, jid, name.
2032 sx_slock(&allprison_lock);
2033 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2035 TAILQ_FOREACH(pr, &allprison, pr_list) {
2036 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2037 mtx_lock(&pr->pr_mtx);
2038 if (pr->pr_ref > 0 &&
2039 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2041 mtx_unlock(&pr->pr_mtx);
2047 vfs_opterror(opts, "no jail after %d", jid);
2048 goto done_unlock_list;
2049 } else if (error != ENOENT)
2050 goto done_unlock_list;
2052 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2055 pr = prison_find_child(mypr, jid);
2057 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2058 mtx_unlock(&pr->pr_mtx);
2060 vfs_opterror(opts, "jail %d is dying",
2062 goto done_unlock_list;
2067 vfs_opterror(opts, "jail %d not found", jid);
2068 goto done_unlock_list;
2070 } else if (error != ENOENT)
2071 goto done_unlock_list;
2073 error = vfs_getopt(opts, "name", (void **)&name, &len);
2075 if (len == 0 || name[len - 1] != '\0') {
2077 goto done_unlock_list;
2079 pr = prison_find_name(mypr, name);
2081 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2082 mtx_unlock(&pr->pr_mtx);
2084 vfs_opterror(opts, "jail \"%s\" is dying",
2086 goto done_unlock_list;
2091 vfs_opterror(opts, "jail \"%s\" not found", name);
2092 goto done_unlock_list;
2093 } else if (error != ENOENT)
2094 goto done_unlock_list;
2096 vfs_opterror(opts, "no jail specified");
2098 goto done_unlock_list;
2101 /* Get the parameters of the prison. */
2104 td->td_retval[0] = pr->pr_id;
2105 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2106 if (error != 0 && error != ENOENT)
2108 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2109 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2110 if (error != 0 && error != ENOENT)
2112 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2113 if (error != 0 && error != ENOENT)
2115 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2116 sizeof(pr->pr_cpuset->cs_id));
2117 if (error != 0 && error != ENOENT)
2119 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2120 if (error != 0 && error != ENOENT)
2123 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2124 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2125 if (error != 0 && error != ENOENT)
2129 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2130 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2131 if (error != 0 && error != ENOENT)
2134 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2135 sizeof(pr->pr_securelevel));
2136 if (error != 0 && error != ENOENT)
2138 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2139 sizeof(pr->pr_childcount));
2140 if (error != 0 && error != ENOENT)
2142 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2143 sizeof(pr->pr_childmax));
2144 if (error != 0 && error != ENOENT)
2146 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2147 if (error != 0 && error != ENOENT)
2149 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2150 if (error != 0 && error != ENOENT)
2152 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2153 if (error != 0 && error != ENOENT)
2155 #ifdef COMPAT_FREEBSD32
2156 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2157 uint32_t hid32 = pr->pr_hostid;
2159 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2162 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2163 sizeof(pr->pr_hostid));
2164 if (error != 0 && error != ENOENT)
2166 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2167 sizeof(pr->pr_enforce_statfs));
2168 if (error != 0 && error != ENOENT)
2170 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2171 sizeof(pr->pr_devfs_rsnum));
2172 if (error != 0 && error != ENOENT)
2174 for (bf = pr_flag_bool;
2175 bf < pr_flag_bool + nitems(pr_flag_bool);
2177 i = (pr->pr_flags & bf->flag) ? 1 : 0;
2178 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2179 if (error != 0 && error != ENOENT)
2182 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2183 if (error != 0 && error != ENOENT)
2186 for (jsf = pr_flag_jailsys;
2187 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
2189 f = pr->pr_flags & (jsf->disable | jsf->new);
2190 i = (f != 0 && f == jsf->disable) ? JAIL_SYS_DISABLE
2191 : (f == jsf->new) ? JAIL_SYS_NEW
2193 error = vfs_setopt(opts, jsf->name, &i, sizeof(i));
2194 if (error != 0 && error != ENOENT)
2197 for (bf = pr_flag_allow;
2198 bf < pr_flag_allow + nitems(pr_flag_allow) &&
2199 atomic_load_int(&bf->flag) != 0;
2201 i = (pr->pr_allow & bf->flag) ? 1 : 0;
2202 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2203 if (error != 0 && error != ENOENT)
2206 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2207 if (error != 0 && error != ENOENT)
2210 i = (pr->pr_uref == 0);
2211 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2212 if (error != 0 && error != ENOENT)
2215 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2216 if (error != 0 && error != ENOENT)
2218 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2219 sizeof(pr->pr_osreldate));
2220 if (error != 0 && error != ENOENT)
2222 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2223 if (error != 0 && error != ENOENT)
2226 /* Get the module parameters. */
2227 mtx_unlock(&pr->pr_mtx);
2229 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2232 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2234 /* By now, all parameters should have been noted. */
2235 TAILQ_FOREACH(opt, opts, link) {
2236 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2238 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2243 /* Write the fetched parameters back to userspace. */
2245 TAILQ_FOREACH(opt, opts, link) {
2246 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2247 pos = 2 * opt->pos + 1;
2248 optuio->uio_iov[pos].iov_len = opt->len;
2249 if (opt->value != NULL) {
2250 if (optuio->uio_segflg == UIO_SYSSPACE) {
2252 optuio->uio_iov[pos].iov_base,
2255 error = copyout(opt->value,
2256 optuio->uio_iov[pos].iov_base,
2267 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2271 sx_sunlock(&allprison_lock);
2273 if (error && errmsg_pos >= 0) {
2274 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2275 errmsg_pos = 2 * errmsg_pos + 1;
2276 if (errmsg_len > 0) {
2277 if (optuio->uio_segflg == UIO_SYSSPACE)
2279 optuio->uio_iov[errmsg_pos].iov_base,
2283 optuio->uio_iov[errmsg_pos].iov_base,
2292 * struct jail_remove_args {
2297 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2299 struct prison *pr, *cpr, *lpr, *tpr;
2302 error = priv_check(td, PRIV_JAIL_REMOVE);
2306 sx_xlock(&allprison_lock);
2307 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2309 sx_xunlock(&allprison_lock);
2313 /* Remove all descendants of this prison, then remove this prison. */
2315 if (!LIST_EMPTY(&pr->pr_children)) {
2316 mtx_unlock(&pr->pr_mtx);
2318 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2319 mtx_lock(&cpr->pr_mtx);
2320 if (cpr->pr_ref > 0) {
2324 /* Already removed - do not do it again. */
2327 mtx_unlock(&cpr->pr_mtx);
2329 mtx_lock(&lpr->pr_mtx);
2330 prison_remove_one(lpr);
2331 sx_xlock(&allprison_lock);
2336 mtx_lock(&lpr->pr_mtx);
2337 prison_remove_one(lpr);
2338 sx_xlock(&allprison_lock);
2340 mtx_lock(&pr->pr_mtx);
2342 prison_remove_one(pr);
2347 prison_remove_one(struct prison *pr)
2352 /* If the prison was persistent, it is not anymore. */
2354 if (pr->pr_flags & PR_PERSIST) {
2357 pr->pr_flags &= ~PR_PERSIST;
2361 * jail_remove added a reference. If that's the only one, remove
2364 KASSERT(pr->pr_ref > 0,
2365 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2366 if (pr->pr_ref == 1) {
2368 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2372 mtx_unlock(&pr->pr_mtx);
2373 sx_xunlock(&allprison_lock);
2375 * Kill all processes unfortunate enough to be attached to this prison.
2377 sx_slock(&allproc_lock);
2378 FOREACH_PROC_IN_SYSTEM(p) {
2380 if (p->p_state != PRS_NEW && p->p_ucred &&
2381 p->p_ucred->cr_prison == pr)
2382 kern_psignal(p, SIGKILL);
2385 sx_sunlock(&allproc_lock);
2386 /* Remove the temporary reference added by jail_remove. */
2387 prison_deref(pr, deuref | PD_DEREF);
2391 * struct jail_attach_args {
2396 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2401 error = priv_check(td, PRIV_JAIL_ATTACH);
2406 * Start with exclusive hold on allprison_lock to ensure that a possible
2407 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2408 * But then immediately downgrade it since we don't need to stop
2411 sx_xlock(&allprison_lock);
2412 sx_downgrade(&allprison_lock);
2413 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2415 sx_sunlock(&allprison_lock);
2420 * Do not allow a process to attach to a prison that is not
2421 * considered to be "alive".
2423 if (pr->pr_uref == 0) {
2424 mtx_unlock(&pr->pr_mtx);
2425 sx_sunlock(&allprison_lock);
2429 return (do_jail_attach(td, pr));
2433 do_jail_attach(struct thread *td, struct prison *pr)
2436 struct ucred *newcred, *oldcred;
2440 * XXX: Note that there is a slight race here if two threads
2441 * in the same privileged process attempt to attach to two
2442 * different jails at the same time. It is important for
2443 * user processes not to do this, or they might end up with
2444 * a process root from one prison, but attached to the jail
2449 mtx_unlock(&pr->pr_mtx);
2451 /* Let modules do whatever they need to prepare for attaching. */
2452 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2454 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2457 sx_sunlock(&allprison_lock);
2460 * Reparent the newly attached process to this jail.
2463 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2467 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2468 if ((error = change_dir(pr->pr_root, td)) != 0)
2471 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2474 VOP_UNLOCK(pr->pr_root);
2475 if ((error = pwd_chroot(td, pr->pr_root)))
2480 oldcred = crcopysafe(p, newcred);
2481 newcred->cr_prison = pr;
2482 proc_set_cred(p, newcred);
2485 racct_proc_ucred_changed(p, oldcred, newcred);
2490 rctl_proc_ucred_changed(p, newcred);
2493 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2498 VOP_UNLOCK(pr->pr_root);
2500 /* Tell modules this thread is still in its old jail after all. */
2501 sx_slock(&allprison_lock);
2502 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2503 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
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. The prison should be locked,
2593 * unless checking for flags that are only set at jail creation (such as
2594 * PR_IP4 and PR_IP6), or only the single bit is examined, without regard
2595 * to any other prison data.
2598 prison_flag(struct ucred *cred, unsigned flag)
2601 return (cred->cr_prison->pr_flags & flag);
2605 prison_allow(struct ucred *cred, unsigned flag)
2608 return ((cred->cr_prison->pr_allow & flag) != 0);
2612 * Remove a prison reference. If that was the last reference, remove the
2613 * prison itself - but not in this context in case there are locks held.
2616 prison_free_locked(struct prison *pr)
2620 mtx_assert(&pr->pr_mtx, MA_OWNED);
2622 mtx_unlock(&pr->pr_mtx);
2624 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2628 prison_free(struct prison *pr)
2631 mtx_lock(&pr->pr_mtx);
2632 prison_free_locked(pr);
2636 * Complete a call to either prison_free or prison_proc_free.
2639 prison_complete(void *context, int pending)
2641 struct prison *pr = context;
2643 sx_xlock(&allprison_lock);
2644 mtx_lock(&pr->pr_mtx);
2645 prison_deref(pr, pr->pr_uref
2646 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2647 : PD_LOCKED | PD_LIST_XLOCKED);
2651 * Remove a prison reference (usually). This internal version assumes no
2652 * mutexes are held, except perhaps the prison itself. If there are no more
2653 * references, release and delist the prison. On completion, the prison lock
2654 * and the allprison lock are both unlocked.
2657 prison_deref(struct prison *pr, int flags)
2659 struct prison *ppr, *tpr;
2662 if (!(flags & PD_LOCKED))
2663 mtx_lock(&pr->pr_mtx);
2665 if (flags & PD_DEUREF) {
2666 KASSERT(pr->pr_uref > 0,
2667 ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2670 lasturef = pr->pr_uref == 0;
2673 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2676 if (flags & PD_DEREF) {
2677 KASSERT(pr->pr_ref > 0,
2678 ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2683 mtx_unlock(&pr->pr_mtx);
2686 * Tell the modules if the last user reference was removed
2687 * (even it sticks around in dying state).
2690 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2692 sx_slock(&allprison_lock);
2693 flags |= PD_LIST_SLOCKED;
2695 sx_xlock(&allprison_lock);
2696 flags |= PD_LIST_XLOCKED;
2699 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2700 mtx_lock(&pr->pr_mtx);
2702 mtx_unlock(&pr->pr_mtx);
2705 /* If the prison still has references, nothing else to do. */
2707 if (flags & PD_LIST_SLOCKED)
2708 sx_sunlock(&allprison_lock);
2709 else if (flags & PD_LIST_XLOCKED)
2710 sx_xunlock(&allprison_lock);
2714 if (flags & PD_LIST_SLOCKED) {
2715 if (!sx_try_upgrade(&allprison_lock)) {
2716 sx_sunlock(&allprison_lock);
2717 sx_xlock(&allprison_lock);
2719 } else if (!(flags & PD_LIST_XLOCKED))
2720 sx_xlock(&allprison_lock);
2722 TAILQ_REMOVE(&allprison, pr, pr_list);
2723 LIST_REMOVE(pr, pr_sibling);
2724 ppr = pr->pr_parent;
2725 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2726 tpr->pr_childcount--;
2727 sx_xunlock(&allprison_lock);
2730 if (pr->pr_vnet != ppr->pr_vnet)
2731 vnet_destroy(pr->pr_vnet);
2733 if (pr->pr_root != NULL)
2735 mtx_destroy(&pr->pr_mtx);
2737 free(pr->pr_ip4, M_PRISON);
2740 free(pr->pr_ip6, M_PRISON);
2742 if (pr->pr_cpuset != NULL)
2743 cpuset_rel(pr->pr_cpuset);
2747 prison_racct_detach(pr);
2751 /* Removing a prison frees a reference on its parent. */
2753 mtx_lock(&pr->pr_mtx);
2754 flags = PD_DEREF | PD_DEUREF;
2759 prison_hold_locked(struct prison *pr)
2762 mtx_assert(&pr->pr_mtx, MA_OWNED);
2763 KASSERT(pr->pr_ref > 0,
2764 ("Trying to hold dead prison %p (jid=%d).", pr, pr->pr_id));
2769 prison_hold(struct prison *pr)
2772 mtx_lock(&pr->pr_mtx);
2773 prison_hold_locked(pr);
2774 mtx_unlock(&pr->pr_mtx);
2778 prison_proc_hold(struct prison *pr)
2781 mtx_lock(&pr->pr_mtx);
2782 KASSERT(pr->pr_uref > 0,
2783 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2785 mtx_unlock(&pr->pr_mtx);
2789 prison_proc_free(struct prison *pr)
2792 mtx_lock(&pr->pr_mtx);
2793 KASSERT(pr->pr_uref > 0,
2794 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2795 if (pr->pr_uref > 1)
2799 * Don't remove the last user reference in this context, which
2800 * is expected to be a process that is not only locked, but
2804 mtx_unlock(&pr->pr_mtx);
2805 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2808 mtx_unlock(&pr->pr_mtx);
2812 * Set or clear a permission bit in the pr_allow field, passing restrictions
2813 * (cleared permission) down to child jails.
2816 prison_set_allow(struct ucred *cred, unsigned flag, int enable)
2820 pr = cred->cr_prison;
2821 sx_slock(&allprison_lock);
2822 mtx_lock(&pr->pr_mtx);
2823 prison_set_allow_locked(pr, flag, enable);
2824 mtx_unlock(&pr->pr_mtx);
2825 sx_sunlock(&allprison_lock);
2829 prison_set_allow_locked(struct prison *pr, unsigned flag, int enable)
2835 pr->pr_allow |= flag;
2837 pr->pr_allow &= ~flag;
2838 FOREACH_PRISON_DESCENDANT_LOCKED(pr, cpr, descend)
2839 cpr->pr_allow &= ~flag;
2844 * Check if a jail supports the given address family.
2846 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
2850 prison_check_af(struct ucred *cred, int af)
2855 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2857 pr = cred->cr_prison;
2859 /* Prisons with their own network stack are not limited. */
2860 if (prison_owns_vnet(cred))
2869 if (pr->pr_flags & PR_IP4)
2871 mtx_lock(&pr->pr_mtx);
2872 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
2873 error = EAFNOSUPPORT;
2874 mtx_unlock(&pr->pr_mtx);
2880 if (pr->pr_flags & PR_IP6)
2882 mtx_lock(&pr->pr_mtx);
2883 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
2884 error = EAFNOSUPPORT;
2885 mtx_unlock(&pr->pr_mtx);
2893 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
2894 error = EAFNOSUPPORT;
2900 * Check if given address belongs to the jail referenced by cred (wrapper to
2901 * prison_check_ip[46]).
2903 * Returns 0 if jail doesn't restrict the address family or if address belongs
2904 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
2905 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
2908 prison_if(struct ucred *cred, const struct sockaddr *sa)
2911 const struct sockaddr_in *sai;
2914 const struct sockaddr_in6 *sai6;
2918 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2919 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
2922 if (prison_owns_vnet(cred))
2927 switch (sa->sa_family)
2931 sai = (const struct sockaddr_in *)sa;
2932 error = prison_check_ip4(cred, &sai->sin_addr);
2937 sai6 = (const struct sockaddr_in6 *)sa;
2938 error = prison_check_ip6(cred, &sai6->sin6_addr);
2942 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
2943 error = EAFNOSUPPORT;
2949 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
2952 prison_check(struct ucred *cred1, struct ucred *cred2)
2955 return ((cred1->cr_prison == cred2->cr_prison ||
2956 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
2960 * Return 1 if p2 is a child of p1, otherwise 0.
2963 prison_ischild(struct prison *pr1, struct prison *pr2)
2966 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
2973 * Return 1 if the passed credential is in a jail and that jail does not
2974 * have its own virtual network stack, otherwise 0.
2977 jailed_without_vnet(struct ucred *cred)
2983 if (prison_owns_vnet(cred))
2991 * Return the correct hostname (domainname, et al) for the passed credential.
2994 getcredhostname(struct ucred *cred, char *buf, size_t size)
2999 * A NULL credential can be used to shortcut to the physical
3000 * system's hostname.
3002 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3003 mtx_lock(&pr->pr_mtx);
3004 strlcpy(buf, pr->pr_hostname, size);
3005 mtx_unlock(&pr->pr_mtx);
3009 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3012 mtx_lock(&cred->cr_prison->pr_mtx);
3013 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3014 mtx_unlock(&cred->cr_prison->pr_mtx);
3018 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3021 mtx_lock(&cred->cr_prison->pr_mtx);
3022 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3023 mtx_unlock(&cred->cr_prison->pr_mtx);
3027 getcredhostid(struct ucred *cred, unsigned long *hostid)
3030 mtx_lock(&cred->cr_prison->pr_mtx);
3031 *hostid = cred->cr_prison->pr_hostid;
3032 mtx_unlock(&cred->cr_prison->pr_mtx);
3036 getjailname(struct ucred *cred, char *name, size_t len)
3039 mtx_lock(&cred->cr_prison->pr_mtx);
3040 strlcpy(name, cred->cr_prison->pr_name, len);
3041 mtx_unlock(&cred->cr_prison->pr_mtx);
3046 * Determine whether the prison represented by cred owns
3047 * its vnet rather than having it inherited.
3049 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3052 prison_owns_vnet(struct ucred *cred)
3056 * vnets cannot be added/removed after jail creation,
3057 * so no need to lock here.
3059 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3064 * Determine whether the subject represented by cred can "see"
3065 * status of a mount point.
3066 * Returns: 0 for permitted, ENOENT otherwise.
3067 * XXX: This function should be called cr_canseemount() and should be
3068 * placed in kern_prot.c.
3071 prison_canseemount(struct ucred *cred, struct mount *mp)
3077 pr = cred->cr_prison;
3078 if (pr->pr_enforce_statfs == 0)
3080 if (pr->pr_root->v_mount == mp)
3082 if (pr->pr_enforce_statfs == 2)
3085 * If jail's chroot directory is set to "/" we should be able to see
3086 * all mount-points from inside a jail.
3087 * This is ugly check, but this is the only situation when jail's
3088 * directory ends with '/'.
3090 if (strcmp(pr->pr_path, "/") == 0)
3092 len = strlen(pr->pr_path);
3094 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3097 * Be sure that we don't have situation where jail's root directory
3098 * is "/some/path" and mount point is "/some/pathpath".
3100 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3106 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3108 char jpath[MAXPATHLEN];
3112 pr = cred->cr_prison;
3113 if (pr->pr_enforce_statfs == 0)
3115 if (prison_canseemount(cred, mp) != 0) {
3116 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3117 strlcpy(sp->f_mntonname, "[restricted]",
3118 sizeof(sp->f_mntonname));
3121 if (pr->pr_root->v_mount == mp) {
3123 * Clear current buffer data, so we are sure nothing from
3124 * the valid path left there.
3126 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3127 *sp->f_mntonname = '/';
3131 * If jail's chroot directory is set to "/" we should be able to see
3132 * all mount-points from inside a jail.
3134 if (strcmp(pr->pr_path, "/") == 0)
3136 len = strlen(pr->pr_path);
3137 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3139 * Clear current buffer data, so we are sure nothing from
3140 * the valid path left there.
3142 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3143 if (*jpath == '\0') {
3144 /* Should never happen. */
3145 *sp->f_mntonname = '/';
3147 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3152 * Check with permission for a specific privilege is granted within jail. We
3153 * have a specific list of accepted privileges; the rest are denied.
3156 prison_priv_check(struct ucred *cred, int priv)
3162 * Some policies have custom handlers. This routine should not be
3163 * called for them. See priv_check_cred().
3166 case PRIV_VFS_LOOKUP:
3167 case PRIV_VFS_GENERATION:
3168 KASSERT(0, ("prison_priv_check instead of a custom handler "
3169 "called for %d\n", priv));
3177 * Privileges specific to prisons with a virtual network stack.
3178 * There might be a duplicate entry here in case the privilege
3179 * is only granted conditionally in the legacy jail case.
3184 * NFS-specific privileges.
3186 case PRIV_NFS_DAEMON:
3187 case PRIV_NFS_LOCKD:
3190 * Network stack privileges.
3192 case PRIV_NET_BRIDGE:
3195 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3196 case PRIV_NET_ROUTE:
3198 case PRIV_NET_SETIFMTU:
3199 case PRIV_NET_SETIFFLAGS:
3200 case PRIV_NET_SETIFCAP:
3201 case PRIV_NET_SETIFDESCR:
3202 case PRIV_NET_SETIFNAME :
3203 case PRIV_NET_SETIFMETRIC:
3204 case PRIV_NET_SETIFPHYS:
3205 case PRIV_NET_SETIFMAC:
3206 case PRIV_NET_SETLANPCP:
3207 case PRIV_NET_ADDMULTI:
3208 case PRIV_NET_DELMULTI:
3209 case PRIV_NET_HWIOCTL:
3210 case PRIV_NET_SETLLADDR:
3211 case PRIV_NET_ADDIFGROUP:
3212 case PRIV_NET_DELIFGROUP:
3213 case PRIV_NET_IFCREATE:
3214 case PRIV_NET_IFDESTROY:
3215 case PRIV_NET_ADDIFADDR:
3216 case PRIV_NET_DELIFADDR:
3219 case PRIV_NET_SETIFVNET:
3220 case PRIV_NET_SETIFFIB:
3223 * 802.11-related privileges.
3225 case PRIV_NET80211_VAP_GETKEY:
3226 case PRIV_NET80211_VAP_MANAGE:
3232 case PRIV_NETATM_CFG:
3233 case PRIV_NETATM_ADD:
3234 case PRIV_NETATM_DEL:
3235 case PRIV_NETATM_SET:
3238 * Bluetooth privileges.
3240 case PRIV_NETBLUETOOTH_RAW:
3244 * Netgraph and netgraph module privileges.
3246 case PRIV_NETGRAPH_CONTROL:
3248 case PRIV_NETGRAPH_TTY:
3252 * IPv4 and IPv6 privileges.
3254 case PRIV_NETINET_IPFW:
3255 case PRIV_NETINET_DIVERT:
3256 case PRIV_NETINET_PF:
3257 case PRIV_NETINET_DUMMYNET:
3258 case PRIV_NETINET_CARP:
3259 case PRIV_NETINET_MROUTE:
3260 case PRIV_NETINET_RAW:
3261 case PRIV_NETINET_ADDRCTRL6:
3262 case PRIV_NETINET_ND6:
3263 case PRIV_NETINET_SCOPE6:
3264 case PRIV_NETINET_ALIFETIME6:
3265 case PRIV_NETINET_IPSEC:
3266 case PRIV_NETINET_BINDANY:
3281 * No default: or deny here.
3282 * In case of no permit fall through to next switch().
3284 if (cred->cr_prison->pr_flags & PR_VNET)
3291 * Allow ktrace privileges for root in jail.
3297 * Allow jailed processes to configure audit identity and
3298 * submit audit records (login, etc). In the future we may
3299 * want to further refine the relationship between audit and
3302 case PRIV_AUDIT_GETAUDIT:
3303 case PRIV_AUDIT_SETAUDIT:
3304 case PRIV_AUDIT_SUBMIT:
3308 * Allow jailed processes to manipulate process UNIX
3309 * credentials in any way they see fit.
3311 case PRIV_CRED_SETUID:
3312 case PRIV_CRED_SETEUID:
3313 case PRIV_CRED_SETGID:
3314 case PRIV_CRED_SETEGID:
3315 case PRIV_CRED_SETGROUPS:
3316 case PRIV_CRED_SETREUID:
3317 case PRIV_CRED_SETREGID:
3318 case PRIV_CRED_SETRESUID:
3319 case PRIV_CRED_SETRESGID:
3322 * Jail implements visibility constraints already, so allow
3323 * jailed root to override uid/gid-based constraints.
3325 case PRIV_SEEOTHERGIDS:
3326 case PRIV_SEEOTHERUIDS:
3329 * Jail implements inter-process debugging limits already, so
3330 * allow jailed root various debugging privileges.
3332 case PRIV_DEBUG_DIFFCRED:
3333 case PRIV_DEBUG_SUGID:
3334 case PRIV_DEBUG_UNPRIV:
3337 * Allow jail to set various resource limits and login
3338 * properties, and for now, exceed process resource limits.
3340 case PRIV_PROC_LIMIT:
3341 case PRIV_PROC_SETLOGIN:
3342 case PRIV_PROC_SETRLIMIT:
3345 * System V and POSIX IPC privileges are granted in jail.
3348 case PRIV_IPC_WRITE:
3349 case PRIV_IPC_ADMIN:
3350 case PRIV_IPC_MSGSIZE:
3354 * Jail operations within a jail work on child jails.
3356 case PRIV_JAIL_ATTACH:
3358 case PRIV_JAIL_REMOVE:
3361 * Jail implements its own inter-process limits, so allow
3362 * root processes in jail to change scheduling on other
3363 * processes in the same jail. Likewise for signalling.
3365 case PRIV_SCHED_DIFFCRED:
3366 case PRIV_SCHED_CPUSET:
3367 case PRIV_SIGNAL_DIFFCRED:
3368 case PRIV_SIGNAL_SUGID:
3371 * Allow jailed processes to write to sysctls marked as jail
3374 case PRIV_SYSCTL_WRITEJAIL:
3377 * Allow root in jail to manage a variety of quota
3378 * properties. These should likely be conditional on a
3379 * configuration option.
3381 case PRIV_VFS_GETQUOTA:
3382 case PRIV_VFS_SETQUOTA:
3385 * Since Jail relies on chroot() to implement file system
3386 * protections, grant many VFS privileges to root in jail.
3387 * Be careful to exclude mount-related and NFS-related
3391 case PRIV_VFS_WRITE:
3392 case PRIV_VFS_ADMIN:
3394 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3395 case PRIV_VFS_CHFLAGS_DEV:
3396 case PRIV_VFS_CHOWN:
3397 case PRIV_VFS_CHROOT:
3398 case PRIV_VFS_RETAINSUGID:
3399 case PRIV_VFS_FCHROOT:
3401 case PRIV_VFS_SETGID:
3403 case PRIV_VFS_STICKYFILE:
3406 * As in the non-jail case, non-root users are expected to be
3407 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3408 * exists in the jail and they have permission to access it).
3410 case PRIV_KMEM_READ:
3414 * Depending on the global setting, allow privilege of
3415 * setting system flags.
3417 case PRIV_VFS_SYSFLAGS:
3418 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3424 * Depending on the global setting, allow privilege of
3425 * mounting/unmounting file systems.
3427 case PRIV_VFS_MOUNT:
3428 case PRIV_VFS_UNMOUNT:
3429 case PRIV_VFS_MOUNT_NONUSER:
3430 case PRIV_VFS_MOUNT_OWNER:
3431 pr = cred->cr_prison;
3433 if (pr->pr_allow & PR_ALLOW_MOUNT && pr->pr_enforce_statfs < 2)
3441 * Jails should hold no disposition on the PRIV_VFS_READ_DIR
3442 * policy. priv_check_cred will not specifically allow it, and
3443 * we may want a MAC policy to allow it.
3445 case PRIV_VFS_READ_DIR:
3449 * Conditionnaly allow locking (unlocking) physical pages
3453 case PRIV_VM_MUNLOCK:
3454 if (cred->cr_prison->pr_allow & PR_ALLOW_MLOCK)
3460 * Conditionally allow jailed root to bind reserved ports.
3462 case PRIV_NETINET_RESERVEDPORT:
3463 if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS)
3469 * Allow jailed root to reuse in-use ports.
3471 case PRIV_NETINET_REUSEPORT:
3475 * Allow jailed root to set certain IPv4/6 (option) headers.
3477 case PRIV_NETINET_SETHDROPTS:
3481 * Conditionally allow creating raw sockets in jail.
3483 case PRIV_NETINET_RAW:
3484 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3490 * Since jail implements its own visibility limits on netstat
3491 * sysctls, allow getcred. This allows identd to work in
3494 case PRIV_NETINET_GETCRED:
3498 * Allow jailed root to set loginclass.
3500 case PRIV_PROC_SETLOGINCLASS:
3504 * Do not allow a process inside a jail to read the kernel
3505 * message buffer unless explicitly permitted.
3508 if (cred->cr_prison->pr_allow & PR_ALLOW_READ_MSGBUF)
3514 * In all remaining cases, deny the privilege request. This
3515 * includes almost all network privileges, many system
3516 * configuration privileges.
3523 * Return the part of pr2's name that is relative to pr1, or the whole name
3524 * if it does not directly follow.
3528 prison_name(struct prison *pr1, struct prison *pr2)
3532 /* Jails see themselves as "0" (if they see themselves at all). */
3535 name = pr2->pr_name;
3536 if (prison_ischild(pr1, pr2)) {
3538 * pr1 isn't locked (and allprison_lock may not be either)
3539 * so its length can't be counted on. But the number of dots
3540 * can be counted on - and counted.
3542 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3543 name = strchr(name, '.') + 1;
3549 * Return the part of pr2's path that is relative to pr1, or the whole path
3550 * if it does not directly follow.
3553 prison_path(struct prison *pr1, struct prison *pr2)
3555 char *path1, *path2;
3558 path1 = pr1->pr_path;
3559 path2 = pr2->pr_path;
3560 if (!strcmp(path1, "/"))
3562 len1 = strlen(path1);
3563 if (strncmp(path1, path2, len1))
3565 if (path2[len1] == '\0')
3567 if (path2[len1] == '/')
3568 return (path2 + len1);
3573 * Jail-related sysctls.
3575 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
3579 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3582 struct prison *pr, *cpr;
3584 struct in_addr *ip4 = NULL;
3588 struct in6_addr *ip6 = NULL;
3593 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3594 pr = req->td->td_ucred->cr_prison;
3596 sx_slock(&allprison_lock);
3597 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3598 #if defined(INET) || defined(INET6)
3601 mtx_lock(&cpr->pr_mtx);
3603 if (cpr->pr_ip4s > 0) {
3604 if (ip4s < cpr->pr_ip4s) {
3605 ip4s = cpr->pr_ip4s;
3606 mtx_unlock(&cpr->pr_mtx);
3607 ip4 = realloc(ip4, ip4s *
3608 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3611 bcopy(cpr->pr_ip4, ip4,
3612 cpr->pr_ip4s * sizeof(struct in_addr));
3616 if (cpr->pr_ip6s > 0) {
3617 if (ip6s < cpr->pr_ip6s) {
3618 ip6s = cpr->pr_ip6s;
3619 mtx_unlock(&cpr->pr_mtx);
3620 ip6 = realloc(ip6, ip6s *
3621 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3624 bcopy(cpr->pr_ip6, ip6,
3625 cpr->pr_ip6s * sizeof(struct in6_addr));
3628 if (cpr->pr_ref == 0) {
3629 mtx_unlock(&cpr->pr_mtx);
3632 bzero(xp, sizeof(*xp));
3633 xp->pr_version = XPRISON_VERSION;
3634 xp->pr_id = cpr->pr_id;
3635 xp->pr_state = cpr->pr_uref > 0
3636 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3637 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3638 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3639 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3641 xp->pr_ip4s = cpr->pr_ip4s;
3644 xp->pr_ip6s = cpr->pr_ip6s;
3646 mtx_unlock(&cpr->pr_mtx);
3647 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3651 if (xp->pr_ip4s > 0) {
3652 error = SYSCTL_OUT(req, ip4,
3653 xp->pr_ip4s * sizeof(struct in_addr));
3659 if (xp->pr_ip6s > 0) {
3660 error = SYSCTL_OUT(req, ip6,
3661 xp->pr_ip6s * sizeof(struct in6_addr));
3667 sx_sunlock(&allprison_lock);
3678 SYSCTL_OID(_security_jail, OID_AUTO, list,
3679 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3680 sysctl_jail_list, "S", "List of active jails");
3683 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3687 injail = jailed(req->td->td_ucred);
3688 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3693 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3694 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3695 sysctl_jail_jailed, "I", "Process in jail?");
3698 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
3700 int error, havevnet;
3702 struct ucred *cred = req->td->td_ucred;
3704 havevnet = jailed(cred) && prison_owns_vnet(cred);
3708 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
3713 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
3714 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3715 sysctl_jail_vnet, "I", "Jail owns vnet?");
3717 #if defined(INET) || defined(INET6)
3718 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3719 &jail_max_af_ips, 0,
3720 "Number of IP addresses a jail may have at most per address family (deprecated)");
3724 * Default parameters for jail(2) compatibility. For historical reasons,
3725 * the sysctl names have varying similarity to the parameter names. Prisons
3726 * just see their own parameters, and can't change them.
3729 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3733 /* Get the current flag value, and convert it to a boolean. */
3734 if (req->td->td_ucred->cr_prison == &prison0) {
3735 mtx_lock(&prison0.pr_mtx);
3736 i = (jail_default_allow & arg2) != 0;
3737 mtx_unlock(&prison0.pr_mtx);
3739 i = prison_allow(req->td->td_ucred, arg2);
3743 error = sysctl_handle_int(oidp, &i, 0, req);
3744 if (error || !req->newptr)
3750 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3753 mtx_lock(&prison0.pr_mtx);
3754 jail_default_allow = (jail_default_allow & ~arg2) | i;
3755 mtx_unlock(&prison0.pr_mtx);
3759 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3760 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3761 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3762 "Processes in jail can set their hostnames (deprecated)");
3763 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3764 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3765 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3766 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
3767 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3768 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3769 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3770 "Processes in jail can use System V IPC primitives (deprecated)");
3771 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3772 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3773 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3774 "Prison root can create raw sockets (deprecated)");
3775 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3776 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3777 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3778 "Processes in jail can alter system file flags (deprecated)");
3779 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3780 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3781 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3782 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
3785 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3790 pr = req->td->td_ucred->cr_prison;
3791 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3792 error = sysctl_handle_int(oidp, &level, 0, req);
3793 if (error || !req->newptr)
3795 *(int *)arg1 = level;
3799 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3800 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3801 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3802 sysctl_jail_default_level, "I",
3803 "Processes in jail cannot see all mounted file systems (deprecated)");
3805 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
3806 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
3807 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
3808 sysctl_jail_default_level, "I",
3809 "Ruleset for the devfs filesystem in jail (deprecated)");
3812 * Nodes to describe jail parameters. Maximum length of string parameters
3813 * is returned in the string itself, and the other parameters exist merely
3814 * to make themselves and their types known.
3816 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
3820 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3827 switch (oidp->oid_kind & CTLTYPE)
3833 if (!(req->flags & SCTL_MASK32))
3835 return (SYSCTL_OUT(req, &l, sizeof(l)));
3839 return (SYSCTL_OUT(req, &i, sizeof(i)));
3840 case CTLTYPE_STRING:
3841 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
3843 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3844 case CTLTYPE_STRUCT:
3846 return (SYSCTL_OUT(req, &s, sizeof(s)));
3852 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
3853 * jail creation time but cannot be changed in an existing jail.
3855 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3856 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3857 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3858 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3859 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3860 "I", "Jail secure level");
3861 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
3862 "Jail value for kern.osreldate and uname -K");
3863 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
3864 "Jail value for kern.osrelease and uname -r");
3865 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3866 "I", "Jail cannot see all mounted file systems");
3867 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
3868 "I", "Ruleset for in-jail devfs mounts");
3869 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3870 "B", "Jail persistence");
3872 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3873 "E,jailsys", "Virtual network stack");
3875 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3876 "B", "Jail is in the process of shutting down");
3878 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
3879 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
3880 "I", "Current number of child jails");
3881 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
3882 "I", "Maximum number of child jails");
3884 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
3885 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3887 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3888 "Jail NIS domainname");
3889 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3891 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3892 "LU", "Jail host ID");
3894 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3895 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3898 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
3899 "Jail IPv4 address virtualization");
3900 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3901 "S,in_addr,a", "Jail IPv4 addresses");
3902 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3903 "B", "Do (not) use IPv4 source address selection rather than the "
3904 "primary jail IPv4 address.");
3907 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
3908 "Jail IPv6 address virtualization");
3909 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3910 "S,in6_addr,a", "Jail IPv6 addresses");
3911 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3912 "B", "Do (not) use IPv6 source address selection rather than the "
3913 "primary jail IPv6 address.");
3916 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3917 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3918 "B", "Jail may set hostname");
3919 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3920 "B", "Jail may use SYSV IPC");
3921 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3922 "B", "Jail may create raw sockets");
3923 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3924 "B", "Jail may alter system file flags");
3925 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3926 "B", "Jail may set file quotas");
3927 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3928 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3929 SYSCTL_JAIL_PARAM(_allow, mlock, CTLTYPE_INT | CTLFLAG_RW,
3930 "B", "Jail may lock (unlock) physical pages in memory");
3931 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW,
3932 "B", "Jail may bind sockets to reserved ports");
3933 SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW,
3934 "B", "Jail may read the kernel message buffer");
3935 SYSCTL_JAIL_PARAM(_allow, unprivileged_proc_debug, CTLTYPE_INT | CTLFLAG_RW,
3936 "B", "Unprivileged processes may use process debugging facilities");
3937 SYSCTL_JAIL_PARAM(_allow, suser, CTLTYPE_INT | CTLFLAG_RW,
3938 "B", "Processes in jail with uid 0 have privilege");
3940 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
3941 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
3942 "B", "Jail may mount/unmount jail-friendly file systems in general");
3945 * Add a dynamic parameter allow.<name>, or allow.<prefix>.<name>. Return
3946 * its associated bit in the pr_allow bitmask, or zero if the parameter was
3950 prison_add_allow(const char *prefix, const char *name, const char *prefix_descr,
3953 struct bool_flags *bf;
3954 struct sysctl_oid *parent;
3955 char *allow_name, *allow_noname, *allowed;
3956 #ifndef NO_SYSCTL_DESCR
3957 char *descr_deprecated;
3962 ? asprintf(&allow_name, M_PRISON, "allow.%s.%s", prefix, name)
3964 asprintf(&allow_noname, M_PRISON, "allow.%s.no%s", prefix, name)
3966 : asprintf(&allow_name, M_PRISON, "allow.%s", name) < 0 ||
3967 asprintf(&allow_noname, M_PRISON, "allow.no%s", name) < 0) {
3968 free(allow_name, M_PRISON);
3973 * See if this parameter has already beed added, i.e. a module was
3974 * previously loaded/unloaded.
3976 mtx_lock(&prison0.pr_mtx);
3977 for (bf = pr_flag_allow;
3978 bf < pr_flag_allow + nitems(pr_flag_allow) &&
3979 atomic_load_int(&bf->flag) != 0;
3981 if (strcmp(bf->name, allow_name) == 0) {
3982 allow_flag = bf->flag;
3988 * Find a free bit in pr_allow_all, failing if there are none
3989 * (which shouldn't happen as long as we keep track of how many
3990 * potential dynamic flags exist).
3992 for (allow_flag = 1;; allow_flag <<= 1) {
3993 if (allow_flag == 0)
3995 if ((pr_allow_all & allow_flag) == 0)
3999 /* Note the parameter in the next open slot in pr_flag_allow. */
4000 for (bf = pr_flag_allow; ; bf++) {
4001 if (bf == pr_flag_allow + nitems(pr_flag_allow)) {
4002 /* This should never happen, but is not fatal. */
4006 if (atomic_load_int(&bf->flag) == 0)
4009 bf->name = allow_name;
4010 bf->noname = allow_noname;
4011 pr_allow_all |= allow_flag;
4013 * prison0 always has permission for the new parameter.
4014 * Other jails must have it granted to them.
4016 prison0.pr_allow |= allow_flag;
4017 /* The flag indicates a valid entry, so make sure it is set last. */
4018 atomic_store_rel_int(&bf->flag, allow_flag);
4019 mtx_unlock(&prison0.pr_mtx);
4022 * Create sysctls for the paramter, and the back-compat global
4026 ? SYSCTL_ADD_NODE(NULL,
4027 SYSCTL_CHILDREN(&sysctl___security_jail_param_allow),
4028 OID_AUTO, prefix, CTLFLAG_MPSAFE, 0, prefix_descr)
4029 : &sysctl___security_jail_param_allow;
4030 (void)SYSCTL_ADD_PROC(NULL, SYSCTL_CHILDREN(parent), OID_AUTO,
4031 name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4032 NULL, 0, sysctl_jail_param, "B", descr);
4034 ? asprintf(&allowed, M_TEMP, "%s_%s_allowed", prefix, name)
4035 : asprintf(&allowed, M_TEMP, "%s_allowed", name)) >= 0) {
4036 #ifndef NO_SYSCTL_DESCR
4037 (void)asprintf(&descr_deprecated, M_TEMP, "%s (deprecated)",
4040 (void)SYSCTL_ADD_PROC(NULL,
4041 SYSCTL_CHILDREN(&sysctl___security_jail), OID_AUTO, allowed,
4042 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, allow_flag,
4043 sysctl_jail_default_allow, "I", descr_deprecated);
4044 #ifndef NO_SYSCTL_DESCR
4045 free(descr_deprecated, M_TEMP);
4047 free(allowed, M_TEMP);
4052 mtx_unlock(&prison0.pr_mtx);
4053 free(allow_name, M_PRISON);
4054 free(allow_noname, M_PRISON);
4059 * The VFS system will register jail-aware filesystems here. They each get
4060 * a parameter allow.mount.xxxfs and a flag to check when a jailed user
4061 * attempts to mount.
4064 prison_add_vfs(struct vfsconf *vfsp)
4066 #ifdef NO_SYSCTL_DESCR
4068 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
4073 (void)asprintf(&descr, M_TEMP, "Jail may mount the %s file system",
4075 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
4077 free(descr, M_TEMP);
4083 prison_racct_foreach(void (*callback)(struct racct *racct,
4084 void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
4085 void *arg2, void *arg3)
4087 struct prison_racct *prr;
4089 ASSERT_RACCT_ENABLED();
4091 sx_slock(&allprison_lock);
4094 LIST_FOREACH(prr, &allprison_racct, prr_next)
4095 (callback)(prr->prr_racct, arg2, arg3);
4098 sx_sunlock(&allprison_lock);
4101 static struct prison_racct *
4102 prison_racct_find_locked(const char *name)
4104 struct prison_racct *prr;
4106 ASSERT_RACCT_ENABLED();
4107 sx_assert(&allprison_lock, SA_XLOCKED);
4109 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4112 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4113 if (strcmp(name, prr->prr_name) != 0)
4116 /* Found prison_racct with a matching name? */
4117 prison_racct_hold(prr);
4121 /* Add new prison_racct. */
4122 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4123 racct_create(&prr->prr_racct);
4125 strcpy(prr->prr_name, name);
4126 refcount_init(&prr->prr_refcount, 1);
4127 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4132 struct prison_racct *
4133 prison_racct_find(const char *name)
4135 struct prison_racct *prr;
4137 ASSERT_RACCT_ENABLED();
4139 sx_xlock(&allprison_lock);
4140 prr = prison_racct_find_locked(name);
4141 sx_xunlock(&allprison_lock);
4146 prison_racct_hold(struct prison_racct *prr)
4149 ASSERT_RACCT_ENABLED();
4151 refcount_acquire(&prr->prr_refcount);
4155 prison_racct_free_locked(struct prison_racct *prr)
4158 ASSERT_RACCT_ENABLED();
4159 sx_assert(&allprison_lock, SA_XLOCKED);
4161 if (refcount_release(&prr->prr_refcount)) {
4162 racct_destroy(&prr->prr_racct);
4163 LIST_REMOVE(prr, prr_next);
4164 free(prr, M_PRISON_RACCT);
4169 prison_racct_free(struct prison_racct *prr)
4172 ASSERT_RACCT_ENABLED();
4173 sx_assert(&allprison_lock, SA_UNLOCKED);
4175 if (refcount_release_if_not_last(&prr->prr_refcount))
4178 sx_xlock(&allprison_lock);
4179 prison_racct_free_locked(prr);
4180 sx_xunlock(&allprison_lock);
4184 prison_racct_attach(struct prison *pr)
4186 struct prison_racct *prr;
4188 ASSERT_RACCT_ENABLED();
4189 sx_assert(&allprison_lock, SA_XLOCKED);
4191 prr = prison_racct_find_locked(pr->pr_name);
4192 KASSERT(prr != NULL, ("cannot find prison_racct"));
4194 pr->pr_prison_racct = prr;
4198 * Handle jail renaming. From the racct point of view, renaming means
4199 * moving from one prison_racct to another.
4202 prison_racct_modify(struct prison *pr)
4208 struct prison_racct *oldprr;
4210 ASSERT_RACCT_ENABLED();
4212 sx_slock(&allproc_lock);
4213 sx_xlock(&allprison_lock);
4215 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4216 sx_xunlock(&allprison_lock);
4217 sx_sunlock(&allproc_lock);
4221 oldprr = pr->pr_prison_racct;
4222 pr->pr_prison_racct = NULL;
4224 prison_racct_attach(pr);
4227 * Move resource utilisation records.
4229 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4233 * Force rctl to reattach rules to processes.
4235 FOREACH_PROC_IN_SYSTEM(p) {
4237 cred = crhold(p->p_ucred);
4239 rctl_proc_ucred_changed(p, cred);
4244 sx_sunlock(&allproc_lock);
4245 prison_racct_free_locked(oldprr);
4246 sx_xunlock(&allprison_lock);
4250 prison_racct_detach(struct prison *pr)
4253 ASSERT_RACCT_ENABLED();
4254 sx_assert(&allprison_lock, SA_UNLOCKED);
4256 if (pr->pr_prison_racct == NULL)
4258 prison_racct_free(pr->pr_prison_racct);
4259 pr->pr_prison_racct = NULL;
4266 db_show_prison(struct prison *pr)
4268 struct bool_flags *bf;
4269 struct jailsys_flags *jsf;
4270 #if defined(INET) || defined(INET6)
4275 char ip4buf[INET_ADDRSTRLEN];
4278 char ip6buf[INET6_ADDRSTRLEN];
4281 db_printf("prison %p:\n", pr);
4282 db_printf(" jid = %d\n", pr->pr_id);
4283 db_printf(" name = %s\n", pr->pr_name);
4284 db_printf(" parent = %p\n", pr->pr_parent);
4285 db_printf(" ref = %d\n", pr->pr_ref);
4286 db_printf(" uref = %d\n", pr->pr_uref);
4287 db_printf(" path = %s\n", pr->pr_path);
4288 db_printf(" cpuset = %d\n", pr->pr_cpuset
4289 ? pr->pr_cpuset->cs_id : -1);
4291 db_printf(" vnet = %p\n", pr->pr_vnet);
4293 db_printf(" root = %p\n", pr->pr_root);
4294 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4295 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4296 db_printf(" children.max = %d\n", pr->pr_childmax);
4297 db_printf(" children.cur = %d\n", pr->pr_childcount);
4298 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4299 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4300 db_printf(" flags = 0x%x", pr->pr_flags);
4301 for (bf = pr_flag_bool; bf < pr_flag_bool + nitems(pr_flag_bool); bf++)
4302 if (pr->pr_flags & bf->flag)
4303 db_printf(" %s", bf->name);
4304 for (jsf = pr_flag_jailsys;
4305 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
4307 f = pr->pr_flags & (jsf->disable | jsf->new);
4308 db_printf(" %-16s= %s\n", jsf->name,
4309 (f != 0 && f == jsf->disable) ? "disable"
4310 : (f == jsf->new) ? "new"
4313 db_printf(" allow = 0x%x", pr->pr_allow);
4314 for (bf = pr_flag_allow;
4315 bf < pr_flag_allow + nitems(pr_flag_allow) &&
4316 atomic_load_int(&bf->flag) != 0;
4318 if (pr->pr_allow & bf->flag)
4319 db_printf(" %s", bf->name);
4321 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4322 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4323 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4324 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4325 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4327 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4328 for (ii = 0; ii < pr->pr_ip4s; ii++)
4329 db_printf(" %s %s\n",
4330 ii == 0 ? "ip4.addr =" : " ",
4331 inet_ntoa_r(pr->pr_ip4[ii], ip4buf));
4334 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4335 for (ii = 0; ii < pr->pr_ip6s; ii++)
4336 db_printf(" %s %s\n",
4337 ii == 0 ? "ip6.addr =" : " ",
4338 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4342 DB_SHOW_COMMAND(prison, db_show_prison_command)
4348 * Show all prisons in the list, and prison0 which is not
4351 db_show_prison(&prison0);
4352 if (!db_pager_quit) {
4353 TAILQ_FOREACH(pr, &allprison, pr_list) {
4365 /* Look for a prison with the ID and with references. */
4366 TAILQ_FOREACH(pr, &allprison, pr_list)
4367 if (pr->pr_id == addr && pr->pr_ref > 0)
4370 /* Look again, without requiring a reference. */
4371 TAILQ_FOREACH(pr, &allprison, pr_list)
4372 if (pr->pr_id == addr)
4375 /* Assume address points to a valid prison. */
4376 pr = (struct prison *)addr;