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>
52 #include <sys/mutex.h>
53 #include <sys/racct.h>
55 #include <sys/refcount.h>
57 #include <sys/sysent.h>
58 #include <sys/namei.h>
59 #include <sys/mount.h>
60 #include <sys/queue.h>
61 #include <sys/socket.h>
62 #include <sys/syscallsubr.h>
63 #include <sys/sysctl.h>
64 #include <sys/vnode.h>
69 #include <netinet/in.h>
75 #include <security/mac/mac_framework.h>
77 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
79 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
80 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
82 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
85 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
87 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
91 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
93 #define _PR_IP_SADDRSEL 0
97 /* prison0 describes what is "real" about the system. */
98 struct prison prison0 = {
104 .pr_securelevel = -1,
106 .pr_childmax = JAIL_MAX,
107 .pr_hostuuid = DEFAULT_HOSTUUID,
108 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
110 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
112 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
114 .pr_allow = PR_ALLOW_ALL_STATIC,
116 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
123 struct jailsys_flags {
129 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
130 struct sx allprison_lock;
131 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
132 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
133 LIST_HEAD(, prison_racct) allprison_racct;
136 static int do_jail_attach(struct thread *td, struct prison *pr);
137 static void prison_complete(void *context, int pending);
138 static void prison_deref(struct prison *pr, int flags);
139 static char *prison_path(struct prison *pr1, struct prison *pr2);
140 static void prison_remove_one(struct prison *pr);
142 static void prison_racct_attach(struct prison *pr);
143 static void prison_racct_modify(struct prison *pr);
144 static void prison_racct_detach(struct prison *pr);
147 /* Flags for prison_deref */
148 #define PD_DEREF 0x01
149 #define PD_DEUREF 0x02
150 #define PD_LOCKED 0x04
151 #define PD_LIST_SLOCKED 0x08
152 #define PD_LIST_XLOCKED 0x10
155 * Parameter names corresponding to PR_* flag values. Size values are for kvm
156 * as we cannot figure out the size of a sparse array, or an array without a
159 static struct bool_flags pr_flag_bool[] = {
160 {"persist", "nopersist", PR_PERSIST},
162 {"ip4.saddrsel", "ip4.nosaddrsel", PR_IP4_SADDRSEL},
165 {"ip6.saddrsel", "ip6.nosaddrsel", PR_IP6_SADDRSEL},
168 const size_t pr_flag_bool_size = sizeof(pr_flag_bool);
170 static struct jailsys_flags pr_flag_jailsys[] = {
171 {"host", 0, PR_HOST},
173 {"vnet", 0, PR_VNET},
176 {"ip4", PR_IP4_USER, PR_IP4_USER},
179 {"ip6", PR_IP6_USER, PR_IP6_USER},
182 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
184 /* Make this array full-size so dynamic parameters can be added. */
185 static struct bool_flags pr_flag_allow[NBBY * NBPW] = {
186 {"allow.set_hostname", "allow.noset_hostname", PR_ALLOW_SET_HOSTNAME},
187 {"allow.sysvipc", "allow.nosysvipc", PR_ALLOW_SYSVIPC},
188 {"allow.raw_sockets", "allow.noraw_sockets", PR_ALLOW_RAW_SOCKETS},
189 {"allow.chflags", "allow.nochflags", PR_ALLOW_CHFLAGS},
190 {"allow.mount", "allow.nomount", PR_ALLOW_MOUNT},
191 {"allow.quotas", "allow.noquotas", PR_ALLOW_QUOTAS},
192 {"allow.socket_af", "allow.nosocket_af", PR_ALLOW_SOCKET_AF},
193 {"allow.mlock", "allow.nomlock", PR_ALLOW_MLOCK},
194 {"allow.reserved_ports", "allow.noreserved_ports",
195 PR_ALLOW_RESERVED_PORTS},
196 {"allow.read_msgbuf", "allow.noread_msgbuf", PR_ALLOW_READ_MSGBUF},
197 {"allow.unprivileged_proc_debug", "allow.nounprivileged_proc_debug",
198 PR_ALLOW_UNPRIV_DEBUG},
200 const size_t pr_flag_allow_size = sizeof(pr_flag_allow);
202 #define JAIL_DEFAULT_ALLOW (PR_ALLOW_SET_HOSTNAME | \
203 PR_ALLOW_RESERVED_PORTS | \
204 PR_ALLOW_UNPRIV_DEBUG)
205 #define JAIL_DEFAULT_ENFORCE_STATFS 2
206 #define JAIL_DEFAULT_DEVFS_RSNUM 0
207 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
208 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
209 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
210 #if defined(INET) || defined(INET6)
211 static unsigned jail_max_af_ips = 255;
215 * Initialize the parts of prison0 that can't be static-initialized with
216 * constants. This is called from proc0_init() after creating thread0 cpuset.
222 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
223 prison0.pr_osreldate = osreldate;
224 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
233 sys_jail(struct thread *td, struct jail_args *uap)
239 error = copyin(uap->jail, &version, sizeof(uint32_t));
248 /* FreeBSD single IPv4 jails. */
249 bzero(&j, sizeof(struct jail));
250 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
253 j.version = j0.version;
255 j.hostname = j0.hostname;
256 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
262 * Version 1 was used by multi-IPv4 jail implementations
263 * that never made it into the official kernel.
267 case 2: /* JAIL_API_VERSION */
268 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
269 error = copyin(uap->jail, &j, sizeof(struct jail));
275 /* Sci-Fi jails are not supported, sorry. */
278 return (kern_jail(td, &j));
282 kern_jail(struct thread *td, struct jail *j)
284 struct iovec optiov[2 * (4 + nitems(pr_flag_allow)
293 char *u_path, *u_hostname, *u_name;
294 struct bool_flags *bf;
297 struct in_addr *u_ip4;
300 struct in6_addr *u_ip6;
303 int error, enforce_statfs;
305 bzero(&optiov, sizeof(optiov));
306 opt.uio_iov = optiov;
310 opt.uio_segflg = UIO_SYSSPACE;
311 opt.uio_rw = UIO_READ;
314 /* Set permissions for top-level jails from sysctls. */
315 if (!jailed(td->td_ucred)) {
316 for (bf = pr_flag_allow;
317 bf < pr_flag_allow + nitems(pr_flag_allow) &&
320 optiov[opt.uio_iovcnt].iov_base = __DECONST(char *,
321 (jail_default_allow & bf->flag)
322 ? bf->name : bf->noname);
323 optiov[opt.uio_iovcnt].iov_len =
324 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
327 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
328 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
330 enforce_statfs = jail_default_enforce_statfs;
331 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
332 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
336 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
338 ip4s = (j->version == 0) ? 1 : j->ip4s;
339 if (ip4s > jail_max_af_ips)
341 tmplen += ip4s * sizeof(struct in_addr);
347 if (j->ip6s > jail_max_af_ips)
349 tmplen += j->ip6s * sizeof(struct in6_addr);
354 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
355 u_hostname = u_path + MAXPATHLEN;
356 u_name = u_hostname + MAXHOSTNAMELEN;
358 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
362 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
364 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
367 optiov[opt.uio_iovcnt].iov_base = "path";
368 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
370 optiov[opt.uio_iovcnt].iov_base = u_path;
371 error = copyinstr(j->path, u_path, MAXPATHLEN,
372 &optiov[opt.uio_iovcnt].iov_len);
374 free(u_path, M_TEMP);
378 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
379 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
381 optiov[opt.uio_iovcnt].iov_base = u_hostname;
382 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
383 &optiov[opt.uio_iovcnt].iov_len);
385 free(u_path, M_TEMP);
389 if (j->jailname != NULL) {
390 optiov[opt.uio_iovcnt].iov_base = "name";
391 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
393 optiov[opt.uio_iovcnt].iov_base = u_name;
394 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
395 &optiov[opt.uio_iovcnt].iov_len);
397 free(u_path, M_TEMP);
403 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
404 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
406 optiov[opt.uio_iovcnt].iov_base = u_ip4;
407 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
409 u_ip4->s_addr = j->ip4s;
411 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
413 free(u_path, M_TEMP);
420 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
421 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
423 optiov[opt.uio_iovcnt].iov_base = u_ip6;
424 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
425 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
427 free(u_path, M_TEMP);
432 KASSERT(opt.uio_iovcnt <= nitems(optiov),
433 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
434 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
435 free(u_path, M_TEMP);
440 * struct jail_set_args {
441 * struct iovec *iovp;
442 * unsigned int iovcnt;
447 sys_jail_set(struct thread *td, struct jail_set_args *uap)
452 /* Check that we have an even number of iovecs. */
456 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
459 error = kern_jail_set(td, auio, uap->flags);
465 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
472 struct in6_addr *ip6;
475 struct vfsoptlist *opts;
476 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
478 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
479 char *g_path, *osrelstr;
480 struct bool_flags *bf;
481 struct jailsys_flags *jsf;
482 #if defined(INET) || defined(INET6)
487 size_t namelen, onamelen, pnamelen;
488 int born, created, cuflags, descend, enforce;
489 int error, errmsg_len, errmsg_pos;
490 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
491 int jid, jsys, len, level;
492 int childmax, osreldt, rsnum, slevel;
493 #if defined(INET) || defined(INET6)
502 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
503 uint64_t pr_allow_diff;
507 error = priv_check(td, PRIV_JAIL_SET);
508 if (!error && (flags & JAIL_ATTACH))
509 error = priv_check(td, PRIV_JAIL_ATTACH);
512 mypr = td->td_ucred->cr_prison;
513 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
515 if (flags & ~JAIL_SET_MASK)
519 * Check all the parameters before committing to anything. Not all
520 * errors can be caught early, but we may as well try. Also, this
521 * takes care of some expensive stuff (path lookup) before getting
522 * the allprison lock.
524 * XXX Jails are not filesystems, and jail parameters are not mount
525 * options. But it makes more sense to re-use the vfsopt code
526 * than duplicate it under a different name.
528 error = vfs_buildopts(optuio, &opts);
539 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
542 vfs_opterror(opts, "no valid operation (create or update)");
546 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
552 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
561 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
569 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
574 else if (enforce < 0 || enforce > 2) {
580 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
588 pr_flags = ch_flags = 0;
589 for (bf = pr_flag_bool;
590 bf < pr_flag_bool + nitems(pr_flag_bool);
592 vfs_flagopt(opts, bf->name, &pr_flags, bf->flag);
593 vfs_flagopt(opts, bf->noname, &ch_flags, bf->flag);
595 ch_flags |= pr_flags;
596 for (jsf = pr_flag_jailsys;
597 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
599 error = vfs_copyopt(opts, jsf->name, &jsys, sizeof(jsys));
605 case JAIL_SYS_DISABLE:
610 pr_flags |= jsf->disable;
613 pr_flags |= jsf->new;
615 case JAIL_SYS_INHERIT:
621 ch_flags |= jsf->new | jsf->disable;
623 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
624 && !(pr_flags & PR_PERSIST)) {
626 vfs_opterror(opts, "new jail must persist or attach");
630 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
632 vfs_opterror(opts, "vnet cannot be changed after creation");
637 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
639 vfs_opterror(opts, "ip4 cannot be changed after creation");
644 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
646 vfs_opterror(opts, "ip6 cannot be changed after creation");
651 pr_allow = ch_allow = 0;
652 for (bf = pr_flag_allow;
653 bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
655 vfs_flagopt(opts, bf->name, &pr_allow, bf->flag);
656 vfs_flagopt(opts, bf->noname, &ch_allow, bf->flag);
658 ch_allow |= pr_allow;
660 error = vfs_getopt(opts, "name", (void **)&name, &len);
666 if (len == 0 || name[len - 1] != '\0') {
670 if (len > MAXHOSTNAMELEN) {
671 error = ENAMETOOLONG;
676 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
684 if (len == 0 || host[len - 1] != '\0') {
688 if (len > MAXHOSTNAMELEN) {
689 error = ENAMETOOLONG;
694 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
702 if (len == 0 || domain[len - 1] != '\0') {
706 if (len > MAXHOSTNAMELEN) {
707 error = ENAMETOOLONG;
712 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
720 if (len == 0 || uuid[len - 1] != '\0') {
724 if (len > HOSTUUIDLEN) {
725 error = ENAMETOOLONG;
730 #ifdef COMPAT_FREEBSD32
731 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
734 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
738 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
750 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
755 else if (ip4s & (sizeof(*ip4) - 1)) {
759 ch_flags |= PR_IP4_USER;
760 pr_flags |= PR_IP4_USER;
762 ip4s /= sizeof(*ip4);
763 if (ip4s > jail_max_af_ips) {
765 vfs_opterror(opts, "too many IPv4 addresses");
768 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
769 bcopy(op, ip4, ip4s * sizeof(*ip4));
771 * IP addresses are all sorted but ip[0] to preserve
772 * the primary IP address as given from userland.
773 * This special IP is used for unbound outgoing
774 * connections as well for "loopback" traffic in case
775 * source address selection cannot find any more fitting
776 * address to connect from.
779 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4),
782 * Check for duplicate addresses and do some simple
783 * zero and broadcast checks. If users give other bogus
784 * addresses it is their problem.
786 * We do not have to care about byte order for these
787 * checks so we will do them in NBO.
789 for (ii = 0; ii < ip4s; ii++) {
790 if (ip4[ii].s_addr == INADDR_ANY ||
791 ip4[ii].s_addr == INADDR_BROADCAST) {
796 (ip4[0].s_addr == ip4[ii+1].s_addr ||
797 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
807 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
812 else if (ip6s & (sizeof(*ip6) - 1)) {
816 ch_flags |= PR_IP6_USER;
817 pr_flags |= PR_IP6_USER;
819 ip6s /= sizeof(*ip6);
820 if (ip6s > jail_max_af_ips) {
822 vfs_opterror(opts, "too many IPv6 addresses");
825 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
826 bcopy(op, ip6, ip6s * sizeof(*ip6));
828 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6),
830 for (ii = 0; ii < ip6s; ii++) {
831 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
836 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
837 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
847 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
848 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
851 "vnet jails cannot have IP address restrictions");
856 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
862 if (flags & JAIL_UPDATE) {
865 "osrelease cannot be changed after creation");
868 if (len == 0 || len >= OSRELEASELEN) {
871 "osrelease string must be 1-%d bytes long",
877 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
883 if (flags & JAIL_UPDATE) {
886 "osreldate cannot be changed after creation");
891 vfs_opterror(opts, "osreldate cannot be 0");
897 error = vfs_getopt(opts, "path", (void **)&path, &len);
903 if (flags & JAIL_UPDATE) {
906 "path cannot be changed after creation");
909 if (len == 0 || path[len - 1] != '\0') {
913 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
919 NDFREE(&nd, NDF_ONLY_PNBUF);
920 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
921 strlcpy(g_path, path, MAXPATHLEN);
922 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
926 /* exit on other errors */
929 if (root->v_type != VDIR) {
938 * Find the specified jail, or at least its parent.
939 * This abuses the file error codes ENOENT and EEXIST.
943 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
944 namelc = strrchr(name, '.');
945 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
949 sx_xlock(&allprison_lock);
952 * See if a requested jid already exists. There is an
953 * information leak here if the jid exists but is not within
954 * the caller's jail hierarchy. Jail creators will get EEXIST
955 * even though they cannot see the jail, and CREATE | UPDATE
956 * will return ENOENT which is not normally a valid error.
960 vfs_opterror(opts, "negative jid");
961 goto done_unlock_list;
963 pr = prison_find(jid);
966 /* Create: jid must not exist. */
967 if (cuflags == JAIL_CREATE) {
968 mtx_unlock(&pr->pr_mtx);
970 vfs_opterror(opts, "jail %d already exists",
972 goto done_unlock_list;
974 if (!prison_ischild(mypr, pr)) {
975 mtx_unlock(&pr->pr_mtx);
977 } else if (pr->pr_uref == 0) {
978 if (!(flags & JAIL_DYING)) {
979 mtx_unlock(&pr->pr_mtx);
981 vfs_opterror(opts, "jail %d is dying",
983 goto done_unlock_list;
984 } else if ((flags & JAIL_ATTACH) ||
985 (pr_flags & PR_PERSIST)) {
987 * A dying jail might be resurrected
988 * (via attach or persist), but first
989 * it must determine if another jail
990 * has claimed its name. Accomplish
991 * this by implicitly re-setting the
995 name = prison_name(mypr, pr);
1000 /* Update: jid must exist. */
1001 if (cuflags == JAIL_UPDATE) {
1003 vfs_opterror(opts, "jail %d not found", jid);
1004 goto done_unlock_list;
1009 * If the caller provided a name, look for a jail by that name.
1010 * This has different semantics for creates and updates keyed by jid
1011 * (where the name must not already exist in a different jail),
1012 * and updates keyed by the name itself (where the name must exist
1013 * because that is the jail being updated).
1017 namelc = strrchr(name, '.');
1022 * This is a hierarchical name. Split it into the
1023 * parent and child names, and make sure the parent
1024 * exists or matches an already found jail.
1027 if (strncmp(name, ppr->pr_name, namelc - name)
1028 || ppr->pr_name[namelc - name] != '\0') {
1029 mtx_unlock(&pr->pr_mtx);
1032 "cannot change jail's parent");
1033 goto done_unlock_list;
1037 ppr = prison_find_name(mypr, name);
1041 "jail \"%s\" not found", name);
1042 goto done_unlock_list;
1044 mtx_unlock(&ppr->pr_mtx);
1049 if (namelc[0] != '\0') {
1051 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1054 FOREACH_PRISON_CHILD(ppr, tpr) {
1055 if (tpr != pr && tpr->pr_ref > 0 &&
1056 !strcmp(tpr->pr_name + pnamelen, namelc)) {
1058 cuflags != JAIL_CREATE) {
1059 mtx_lock(&tpr->pr_mtx);
1060 if (tpr->pr_ref > 0) {
1065 if (tpr->pr_uref > 0) {
1071 mtx_unlock(&tpr->pr_mtx);
1072 } else if (tpr->pr_uref > 0) {
1074 * Create, or update(jid):
1075 * name must not exist in an
1076 * active sibling jail.
1080 mtx_unlock(&pr->pr_mtx);
1082 "jail \"%s\" already exists",
1084 goto done_unlock_list;
1088 /* If no active jail is found, use a dying one. */
1089 if (deadpr != NULL && pr == NULL) {
1090 if (flags & JAIL_DYING) {
1091 mtx_lock(&deadpr->pr_mtx);
1092 if (deadpr->pr_ref == 0) {
1093 mtx_unlock(&deadpr->pr_mtx);
1097 } else if (cuflags == JAIL_UPDATE) {
1100 "jail \"%s\" is dying", name);
1101 goto done_unlock_list;
1104 /* Update: name must exist if no jid. */
1105 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1107 vfs_opterror(opts, "jail \"%s\" not found",
1109 goto done_unlock_list;
1113 /* Update: must provide a jid or name. */
1114 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1116 vfs_opterror(opts, "update specified no jail");
1117 goto done_unlock_list;
1120 /* If there's no prison to update, create a new one and link it in. */
1122 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1123 if (tpr->pr_childcount >= tpr->pr_childmax) {
1125 vfs_opterror(opts, "prison limit exceeded");
1126 goto done_unlock_list;
1129 mtx_lock(&ppr->pr_mtx);
1130 if (ppr->pr_ref == 0) {
1131 mtx_unlock(&ppr->pr_mtx);
1133 vfs_opterror(opts, "jail \"%s\" not found",
1134 prison_name(mypr, ppr));
1135 goto done_unlock_list;
1139 mtx_unlock(&ppr->pr_mtx);
1140 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1142 /* Find the next free jid. */
1145 if (jid == JAIL_MAX)
1147 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1148 if (tpr->pr_id < jid)
1150 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1151 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1154 if (jid == lastprid) {
1157 "no available jail IDs");
1159 prison_deref(ppr, PD_DEREF |
1160 PD_DEUREF | PD_LIST_XLOCKED);
1169 * The jail already has a jid (that did not yet exist),
1170 * so just find where to insert it.
1172 TAILQ_FOREACH(tpr, &allprison, pr_list)
1173 if (tpr->pr_id >= jid) {
1174 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1179 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1180 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1181 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1182 tpr->pr_childcount++;
1184 pr->pr_parent = ppr;
1187 /* Set some default values, and inherit some from the parent. */
1192 root = mypr->pr_root;
1195 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1196 pr->pr_flags |= PR_HOST;
1197 #if defined(INET) || defined(INET6)
1199 if (!(pr_flags & PR_VNET))
1203 if (!(ch_flags & PR_IP4_USER))
1204 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1205 else if (!(pr_flags & PR_IP4_USER)) {
1206 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1207 if (ppr->pr_ip4 != NULL) {
1208 pr->pr_ip4s = ppr->pr_ip4s;
1209 pr->pr_ip4 = malloc(pr->pr_ip4s *
1210 sizeof(struct in_addr), M_PRISON,
1212 bcopy(ppr->pr_ip4, pr->pr_ip4,
1213 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1218 if (!(ch_flags & PR_IP6_USER))
1219 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1220 else if (!(pr_flags & PR_IP6_USER)) {
1221 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1222 if (ppr->pr_ip6 != NULL) {
1223 pr->pr_ip6s = ppr->pr_ip6s;
1224 pr->pr_ip6 = malloc(pr->pr_ip6s *
1225 sizeof(struct in6_addr), M_PRISON,
1227 bcopy(ppr->pr_ip6, pr->pr_ip6,
1228 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1234 /* Source address selection is always on by default. */
1235 pr->pr_flags |= _PR_IP_SADDRSEL;
1237 pr->pr_securelevel = ppr->pr_securelevel;
1238 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1239 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1240 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1242 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1243 if (osrelstr == NULL)
1244 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1246 strcpy(pr->pr_osrelease, osrelstr);
1248 LIST_INIT(&pr->pr_children);
1249 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1250 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1253 /* Allocate a new vnet if specified. */
1254 pr->pr_vnet = (pr_flags & PR_VNET)
1255 ? vnet_alloc() : ppr->pr_vnet;
1258 * Allocate a dedicated cpuset for each jail.
1259 * Unlike other initial settings, this may return an erorr.
1261 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1263 prison_deref(pr, PD_LIST_XLOCKED);
1267 mtx_lock(&pr->pr_mtx);
1269 * New prisons do not yet have a reference, because we do not
1270 * want others to see the incomplete prison once the
1271 * allprison_lock is downgraded.
1276 * Grab a reference for existing prisons, to ensure they
1277 * continue to exist for the duration of the call.
1280 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1281 if ((pr->pr_flags & PR_VNET) &&
1282 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1285 "vnet jails cannot have IP address restrictions");
1286 goto done_deref_locked;
1290 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1293 "ip4 cannot be changed after creation");
1294 goto done_deref_locked;
1298 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1301 "ip6 cannot be changed after creation");
1302 goto done_deref_locked;
1307 /* Do final error checking before setting anything. */
1309 if (slevel < ppr->pr_securelevel) {
1311 goto done_deref_locked;
1315 if (childmax >= ppr->pr_childmax) {
1317 goto done_deref_locked;
1321 if (enforce < ppr->pr_enforce_statfs) {
1323 goto done_deref_locked;
1328 * devfs_rsnum is a uint16_t
1330 if (rsnum < 0 || rsnum > 65535) {
1332 goto done_deref_locked;
1335 * Nested jails always inherit parent's devfs ruleset
1337 if (jailed(td->td_ucred)) {
1338 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1340 goto done_deref_locked;
1342 rsnum = ppr->pr_devfs_rsnum;
1347 if (ppr->pr_flags & PR_IP4) {
1349 * Make sure the new set of IP addresses is a
1350 * subset of the parent's list. Don't worry
1351 * about the parent being unlocked, as any
1352 * setting is done with allprison_lock held.
1354 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1355 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1357 if (ij == ppr->pr_ip4s) {
1359 goto done_deref_locked;
1362 for (ii = ij = 1; ii < ip4s; ii++) {
1363 if (ip4[ii].s_addr ==
1364 ppr->pr_ip4[0].s_addr)
1366 for (; ij < ppr->pr_ip4s; ij++)
1367 if (ip4[ii].s_addr ==
1368 ppr->pr_ip4[ij].s_addr)
1370 if (ij == ppr->pr_ip4s)
1373 if (ij == ppr->pr_ip4s) {
1375 goto done_deref_locked;
1380 * Check for conflicting IP addresses. We permit them
1381 * if there is no more than one IP on each jail. If
1382 * there is a duplicate on a jail with more than one
1383 * IP stop checking and return error.
1386 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1387 if (tppr->pr_flags & PR_VNET)
1392 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1395 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1397 tpr->pr_uref == 0) {
1401 if (!(tpr->pr_flags & PR_IP4_USER))
1404 if (tpr->pr_ip4 == NULL ||
1405 (ip4s == 1 && tpr->pr_ip4s == 1))
1407 for (ii = 0; ii < ip4s; ii++) {
1408 if (prison_check_ip4_locked(tpr, &ip4[ii]) ==
1412 "IPv4 addresses clash");
1413 goto done_deref_locked;
1421 if (ppr->pr_flags & PR_IP6) {
1423 * Make sure the new set of IP addresses is a
1424 * subset of the parent's list.
1426 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1427 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1430 if (ij == ppr->pr_ip6s) {
1432 goto done_deref_locked;
1435 for (ii = ij = 1; ii < ip6s; ii++) {
1436 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1439 for (; ij < ppr->pr_ip6s; ij++)
1440 if (IN6_ARE_ADDR_EQUAL(
1441 &ip6[ii], &ppr->pr_ip6[ij]))
1443 if (ij == ppr->pr_ip6s)
1446 if (ij == ppr->pr_ip6s) {
1448 goto done_deref_locked;
1452 /* Check for conflicting IP addresses. */
1454 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1455 if (tppr->pr_flags & PR_VNET)
1460 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1463 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1465 tpr->pr_uref == 0) {
1469 if (!(tpr->pr_flags & PR_IP6_USER))
1472 if (tpr->pr_ip6 == NULL ||
1473 (ip6s == 1 && tpr->pr_ip6s == 1))
1475 for (ii = 0; ii < ip6s; ii++) {
1476 if (prison_check_ip6_locked(tpr, &ip6[ii]) ==
1480 "IPv6 addresses clash");
1481 goto done_deref_locked;
1487 onamelen = namelen = 0;
1488 if (namelc != NULL) {
1489 /* Give a default name of the jid. Also allow the name to be
1490 * explicitly the jid - but not any other number, and only in
1491 * normal form (no leading zero/etc).
1493 if (namelc[0] == '\0')
1494 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1495 else if ((strtoul(namelc, &p, 10) != jid ||
1496 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1499 "name cannot be numeric (unless it is the jid)");
1500 goto done_deref_locked;
1503 * Make sure the name isn't too long for the prison or its
1506 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1507 onamelen = strlen(pr->pr_name + pnamelen);
1508 namelen = strlen(namelc);
1509 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1510 error = ENAMETOOLONG;
1511 goto done_deref_locked;
1513 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1514 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1515 sizeof(pr->pr_name)) {
1516 error = ENAMETOOLONG;
1517 goto done_deref_locked;
1521 pr_allow_diff = pr_allow & ~ppr->pr_allow;
1522 if (pr_allow_diff & ~PR_ALLOW_DIFFERENCES) {
1524 goto done_deref_locked;
1528 * Let modules check their parameters. This requires unlocking and
1529 * then re-locking the prison, but this is still a valid state as long
1530 * as allprison_lock remains xlocked.
1532 mtx_unlock(&pr->pr_mtx);
1533 error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1535 prison_deref(pr, created
1537 : PD_DEREF | PD_LIST_XLOCKED);
1540 mtx_lock(&pr->pr_mtx);
1542 /* At this point, all valid parameters should have been noted. */
1543 TAILQ_FOREACH(opt, opts, link) {
1544 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1546 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1547 goto done_deref_locked;
1551 /* Set the parameters of the prison. */
1554 if (pr_flags & PR_IP4_USER) {
1555 pr->pr_flags |= PR_IP4;
1556 free(pr->pr_ip4, M_PRISON);
1560 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1562 if (tpr->pr_flags & PR_VNET) {
1567 if (prison_restrict_ip4(tpr, NULL)) {
1576 if (pr_flags & PR_IP6_USER) {
1577 pr->pr_flags |= PR_IP6;
1578 free(pr->pr_ip6, M_PRISON);
1582 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1584 if (tpr->pr_flags & PR_VNET) {
1589 if (prison_restrict_ip6(tpr, NULL)) {
1597 pr->pr_securelevel = slevel;
1598 /* Set all child jails to be at least this level. */
1599 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1600 if (tpr->pr_securelevel < slevel)
1601 tpr->pr_securelevel = slevel;
1604 pr->pr_childmax = childmax;
1605 /* Set all child jails to under this limit. */
1606 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1607 if (tpr->pr_childmax > childmax - level)
1608 tpr->pr_childmax = childmax > level
1609 ? childmax - level : 0;
1612 pr->pr_enforce_statfs = enforce;
1613 /* Pass this restriction on to the children. */
1614 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1615 if (tpr->pr_enforce_statfs < enforce)
1616 tpr->pr_enforce_statfs = enforce;
1619 pr->pr_devfs_rsnum = rsnum;
1620 /* Pass this restriction on to the children. */
1621 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1622 tpr->pr_devfs_rsnum = rsnum;
1624 if (namelc != NULL) {
1625 if (ppr == &prison0)
1626 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1628 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1629 ppr->pr_name, namelc);
1630 /* Change this component of child names. */
1631 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1632 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1633 strlen(tpr->pr_name + onamelen) + 1);
1634 bcopy(pr->pr_name, tpr->pr_name, namelen);
1638 /* Try to keep a real-rooted full pathname. */
1639 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1642 if (PR_HOST & ch_flags & ~pr_flags) {
1643 if (pr->pr_flags & PR_HOST) {
1645 * Copy the parent's host info. As with pr_ip4 above,
1646 * the lack of a lock on the parent is not a problem;
1647 * it is always set with allprison_lock at least
1648 * shared, and is held exclusively here.
1650 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1651 sizeof(pr->pr_hostname));
1652 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1653 sizeof(pr->pr_domainname));
1654 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1655 sizeof(pr->pr_hostuuid));
1656 pr->pr_hostid = pr->pr_parent->pr_hostid;
1658 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1659 /* Set this prison, and any descendants without PR_HOST. */
1661 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1663 strlcpy(pr->pr_domainname, domain,
1664 sizeof(pr->pr_domainname));
1666 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1668 pr->pr_hostid = hid;
1669 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1670 if (tpr->pr_flags & PR_HOST)
1674 strlcpy(tpr->pr_hostname,
1676 sizeof(tpr->pr_hostname));
1678 strlcpy(tpr->pr_domainname,
1680 sizeof(tpr->pr_domainname));
1682 strlcpy(tpr->pr_hostuuid,
1684 sizeof(tpr->pr_hostuuid));
1686 tpr->pr_hostid = hid;
1690 if ((tallow = ch_allow & ~pr_allow)) {
1691 /* Clear allow bits in all children. */
1692 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1693 tpr->pr_allow &= ~tallow;
1695 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1697 * Persistent prisons get an extra reference, and prisons losing their
1698 * persist flag lose that reference. Only do this for existing prisons
1699 * for now, so new ones will remain unseen until after the module
1700 * handlers have completed.
1702 born = pr->pr_uref == 0;
1703 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1704 if (pr_flags & PR_PERSIST) {
1712 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1713 mtx_unlock(&pr->pr_mtx);
1716 if (racct_enable && created)
1717 prison_racct_attach(pr);
1720 /* Locks may have prevented a complete restriction of child IP
1721 * addresses. If so, allocate some more memory and try again.
1726 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1727 mtx_lock(&pr->pr_mtx);
1729 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1731 if (tpr->pr_flags & PR_VNET) {
1736 if (prison_restrict_ip4(tpr, ip4)) {
1743 mtx_unlock(&pr->pr_mtx);
1749 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1750 mtx_lock(&pr->pr_mtx);
1752 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1754 if (tpr->pr_flags & PR_VNET) {
1759 if (prison_restrict_ip6(tpr, ip6)) {
1766 mtx_unlock(&pr->pr_mtx);
1770 /* Let the modules do their work. */
1771 sx_downgrade(&allprison_lock);
1773 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1775 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1776 prison_deref(pr, created
1778 : PD_DEREF | PD_LIST_SLOCKED);
1782 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1785 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1786 prison_deref(pr, created
1788 : PD_DEREF | PD_LIST_SLOCKED);
1792 /* Attach this process to the prison if requested. */
1793 if (flags & JAIL_ATTACH) {
1794 mtx_lock(&pr->pr_mtx);
1795 error = do_jail_attach(td, pr);
1797 vfs_opterror(opts, "attach failed");
1799 prison_deref(pr, PD_DEREF);
1805 if (racct_enable && !created) {
1806 if (!(flags & JAIL_ATTACH))
1807 sx_sunlock(&allprison_lock);
1808 prison_racct_modify(pr);
1809 if (!(flags & JAIL_ATTACH))
1810 sx_slock(&allprison_lock);
1814 td->td_retval[0] = pr->pr_id;
1817 * Now that it is all there, drop the temporary reference from existing
1818 * prisons. Or add a reference to newly created persistent prisons
1819 * (which was not done earlier so that the prison would not be publicly
1823 prison_deref(pr, (flags & JAIL_ATTACH)
1825 : PD_DEREF | PD_LIST_SLOCKED);
1827 if (pr_flags & PR_PERSIST) {
1828 mtx_lock(&pr->pr_mtx);
1831 mtx_unlock(&pr->pr_mtx);
1833 if (!(flags & JAIL_ATTACH))
1834 sx_sunlock(&allprison_lock);
1840 prison_deref(pr, created
1841 ? PD_LOCKED | PD_LIST_XLOCKED
1842 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1845 sx_xunlock(&allprison_lock);
1851 if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1852 &errmsg_len) == 0 && errmsg_len > 0) {
1853 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1854 if (optuio->uio_segflg == UIO_SYSSPACE)
1856 optuio->uio_iov[errmsg_pos].iov_base,
1860 optuio->uio_iov[errmsg_pos].iov_base,
1866 free(ip4, M_PRISON);
1869 free(ip6, M_PRISON);
1872 free(g_path, M_TEMP);
1878 * struct jail_get_args {
1879 * struct iovec *iovp;
1880 * unsigned int iovcnt;
1885 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1890 /* Check that we have an even number of iovecs. */
1891 if (uap->iovcnt & 1)
1894 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1897 error = kern_jail_get(td, auio, uap->flags);
1899 error = copyout(auio->uio_iov, uap->iovp,
1900 uap->iovcnt * sizeof (struct iovec));
1906 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1908 struct bool_flags *bf;
1909 struct jailsys_flags *jsf;
1910 struct prison *pr, *mypr;
1912 struct vfsoptlist *opts;
1913 char *errmsg, *name;
1914 int error, errmsg_len, errmsg_pos, i, jid, len, locked, pos;
1917 if (flags & ~JAIL_GET_MASK)
1920 /* Get the parameter list. */
1921 error = vfs_buildopts(optuio, &opts);
1924 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1925 mypr = td->td_ucred->cr_prison;
1928 * Find the prison specified by one of: lastjid, jid, name.
1930 sx_slock(&allprison_lock);
1931 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1933 TAILQ_FOREACH(pr, &allprison, pr_list) {
1934 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1935 mtx_lock(&pr->pr_mtx);
1936 if (pr->pr_ref > 0 &&
1937 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1939 mtx_unlock(&pr->pr_mtx);
1945 vfs_opterror(opts, "no jail after %d", jid);
1946 goto done_unlock_list;
1947 } else if (error != ENOENT)
1948 goto done_unlock_list;
1950 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1953 pr = prison_find_child(mypr, jid);
1955 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1956 mtx_unlock(&pr->pr_mtx);
1958 vfs_opterror(opts, "jail %d is dying",
1960 goto done_unlock_list;
1965 vfs_opterror(opts, "jail %d not found", jid);
1966 goto done_unlock_list;
1968 } else if (error != ENOENT)
1969 goto done_unlock_list;
1971 error = vfs_getopt(opts, "name", (void **)&name, &len);
1973 if (len == 0 || name[len - 1] != '\0') {
1975 goto done_unlock_list;
1977 pr = prison_find_name(mypr, name);
1979 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1980 mtx_unlock(&pr->pr_mtx);
1982 vfs_opterror(opts, "jail \"%s\" is dying",
1984 goto done_unlock_list;
1989 vfs_opterror(opts, "jail \"%s\" not found", name);
1990 goto done_unlock_list;
1991 } else if (error != ENOENT)
1992 goto done_unlock_list;
1994 vfs_opterror(opts, "no jail specified");
1996 goto done_unlock_list;
1999 /* Get the parameters of the prison. */
2002 td->td_retval[0] = pr->pr_id;
2003 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2004 if (error != 0 && error != ENOENT)
2006 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2007 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2008 if (error != 0 && error != ENOENT)
2010 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2011 if (error != 0 && error != ENOENT)
2013 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2014 sizeof(pr->pr_cpuset->cs_id));
2015 if (error != 0 && error != ENOENT)
2017 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2018 if (error != 0 && error != ENOENT)
2021 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2022 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2023 if (error != 0 && error != ENOENT)
2027 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2028 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2029 if (error != 0 && error != ENOENT)
2032 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2033 sizeof(pr->pr_securelevel));
2034 if (error != 0 && error != ENOENT)
2036 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2037 sizeof(pr->pr_childcount));
2038 if (error != 0 && error != ENOENT)
2040 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2041 sizeof(pr->pr_childmax));
2042 if (error != 0 && error != ENOENT)
2044 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2045 if (error != 0 && error != ENOENT)
2047 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2048 if (error != 0 && error != ENOENT)
2050 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2051 if (error != 0 && error != ENOENT)
2053 #ifdef COMPAT_FREEBSD32
2054 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2055 uint32_t hid32 = pr->pr_hostid;
2057 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2060 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2061 sizeof(pr->pr_hostid));
2062 if (error != 0 && error != ENOENT)
2064 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2065 sizeof(pr->pr_enforce_statfs));
2066 if (error != 0 && error != ENOENT)
2068 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2069 sizeof(pr->pr_devfs_rsnum));
2070 if (error != 0 && error != ENOENT)
2072 for (bf = pr_flag_bool;
2073 bf < pr_flag_bool + nitems(pr_flag_bool);
2075 i = (pr->pr_flags & bf->flag) ? 1 : 0;
2076 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2077 if (error != 0 && error != ENOENT)
2080 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2081 if (error != 0 && error != ENOENT)
2084 for (jsf = pr_flag_jailsys;
2085 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
2087 f = pr->pr_flags & (jsf->disable | jsf->new);
2088 i = (f != 0 && f == jsf->disable) ? JAIL_SYS_DISABLE
2089 : (f == jsf->new) ? JAIL_SYS_NEW
2091 error = vfs_setopt(opts, jsf->name, &i, sizeof(i));
2092 if (error != 0 && error != ENOENT)
2095 for (bf = pr_flag_allow;
2096 bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
2098 i = (pr->pr_allow & bf->flag) ? 1 : 0;
2099 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2100 if (error != 0 && error != ENOENT)
2103 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2104 if (error != 0 && error != ENOENT)
2107 i = (pr->pr_uref == 0);
2108 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2109 if (error != 0 && error != ENOENT)
2112 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2113 if (error != 0 && error != ENOENT)
2115 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2116 sizeof(pr->pr_osreldate));
2117 if (error != 0 && error != ENOENT)
2119 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2120 if (error != 0 && error != ENOENT)
2123 /* Get the module parameters. */
2124 mtx_unlock(&pr->pr_mtx);
2126 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2129 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2131 /* By now, all parameters should have been noted. */
2132 TAILQ_FOREACH(opt, opts, link) {
2133 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2135 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2140 /* Write the fetched parameters back to userspace. */
2142 TAILQ_FOREACH(opt, opts, link) {
2143 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2144 pos = 2 * opt->pos + 1;
2145 optuio->uio_iov[pos].iov_len = opt->len;
2146 if (opt->value != NULL) {
2147 if (optuio->uio_segflg == UIO_SYSSPACE) {
2149 optuio->uio_iov[pos].iov_base,
2152 error = copyout(opt->value,
2153 optuio->uio_iov[pos].iov_base,
2164 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2168 sx_sunlock(&allprison_lock);
2170 if (error && errmsg_pos >= 0) {
2171 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2172 errmsg_pos = 2 * errmsg_pos + 1;
2173 if (errmsg_len > 0) {
2174 if (optuio->uio_segflg == UIO_SYSSPACE)
2176 optuio->uio_iov[errmsg_pos].iov_base,
2180 optuio->uio_iov[errmsg_pos].iov_base,
2189 * struct jail_remove_args {
2194 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2196 struct prison *pr, *cpr, *lpr, *tpr;
2199 error = priv_check(td, PRIV_JAIL_REMOVE);
2203 sx_xlock(&allprison_lock);
2204 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2206 sx_xunlock(&allprison_lock);
2210 /* Remove all descendants of this prison, then remove this prison. */
2212 if (!LIST_EMPTY(&pr->pr_children)) {
2213 mtx_unlock(&pr->pr_mtx);
2215 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2216 mtx_lock(&cpr->pr_mtx);
2217 if (cpr->pr_ref > 0) {
2221 /* Already removed - do not do it again. */
2224 mtx_unlock(&cpr->pr_mtx);
2226 mtx_lock(&lpr->pr_mtx);
2227 prison_remove_one(lpr);
2228 sx_xlock(&allprison_lock);
2233 mtx_lock(&lpr->pr_mtx);
2234 prison_remove_one(lpr);
2235 sx_xlock(&allprison_lock);
2237 mtx_lock(&pr->pr_mtx);
2239 prison_remove_one(pr);
2244 prison_remove_one(struct prison *pr)
2249 /* If the prison was persistent, it is not anymore. */
2251 if (pr->pr_flags & PR_PERSIST) {
2254 pr->pr_flags &= ~PR_PERSIST;
2258 * jail_remove added a reference. If that's the only one, remove
2261 KASSERT(pr->pr_ref > 0,
2262 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2263 if (pr->pr_ref == 1) {
2265 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2269 mtx_unlock(&pr->pr_mtx);
2270 sx_xunlock(&allprison_lock);
2272 * Kill all processes unfortunate enough to be attached to this prison.
2274 sx_slock(&allproc_lock);
2275 FOREACH_PROC_IN_SYSTEM(p) {
2277 if (p->p_state != PRS_NEW && p->p_ucred &&
2278 p->p_ucred->cr_prison == pr)
2279 kern_psignal(p, SIGKILL);
2282 sx_sunlock(&allproc_lock);
2283 /* Remove the temporary reference added by jail_remove. */
2284 prison_deref(pr, deuref | PD_DEREF);
2288 * struct jail_attach_args {
2293 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2298 error = priv_check(td, PRIV_JAIL_ATTACH);
2303 * Start with exclusive hold on allprison_lock to ensure that a possible
2304 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2305 * But then immediately downgrade it since we don't need to stop
2308 sx_xlock(&allprison_lock);
2309 sx_downgrade(&allprison_lock);
2310 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2312 sx_sunlock(&allprison_lock);
2317 * Do not allow a process to attach to a prison that is not
2318 * considered to be "alive".
2320 if (pr->pr_uref == 0) {
2321 mtx_unlock(&pr->pr_mtx);
2322 sx_sunlock(&allprison_lock);
2326 return (do_jail_attach(td, pr));
2330 do_jail_attach(struct thread *td, struct prison *pr)
2333 struct ucred *newcred, *oldcred;
2337 * XXX: Note that there is a slight race here if two threads
2338 * in the same privileged process attempt to attach to two
2339 * different jails at the same time. It is important for
2340 * user processes not to do this, or they might end up with
2341 * a process root from one prison, but attached to the jail
2346 mtx_unlock(&pr->pr_mtx);
2348 /* Let modules do whatever they need to prepare for attaching. */
2349 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2351 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2354 sx_sunlock(&allprison_lock);
2357 * Reparent the newly attached process to this jail.
2360 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2364 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2365 if ((error = change_dir(pr->pr_root, td)) != 0)
2368 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2371 VOP_UNLOCK(pr->pr_root);
2372 if ((error = pwd_chroot(td, pr->pr_root)))
2377 oldcred = crcopysafe(p, newcred);
2378 newcred->cr_prison = pr;
2379 proc_set_cred(p, newcred);
2382 racct_proc_ucred_changed(p, oldcred, newcred);
2387 rctl_proc_ucred_changed(p, newcred);
2390 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2395 VOP_UNLOCK(pr->pr_root);
2397 /* Tell modules this thread is still in its old jail after all. */
2398 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2399 prison_deref(pr, PD_DEREF | PD_DEUREF);
2404 * Returns a locked prison instance, or NULL on failure.
2407 prison_find(int prid)
2411 sx_assert(&allprison_lock, SX_LOCKED);
2412 TAILQ_FOREACH(pr, &allprison, pr_list) {
2413 if (pr->pr_id == prid) {
2414 mtx_lock(&pr->pr_mtx);
2417 mtx_unlock(&pr->pr_mtx);
2424 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2427 prison_find_child(struct prison *mypr, int prid)
2432 sx_assert(&allprison_lock, SX_LOCKED);
2433 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2434 if (pr->pr_id == prid) {
2435 mtx_lock(&pr->pr_mtx);
2438 mtx_unlock(&pr->pr_mtx);
2445 * Look for the name relative to mypr. Returns a locked prison or NULL.
2448 prison_find_name(struct prison *mypr, const char *name)
2450 struct prison *pr, *deadpr;
2454 sx_assert(&allprison_lock, SX_LOCKED);
2455 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2458 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2459 if (!strcmp(pr->pr_name + mylen, name)) {
2460 mtx_lock(&pr->pr_mtx);
2461 if (pr->pr_ref > 0) {
2462 if (pr->pr_uref > 0)
2466 mtx_unlock(&pr->pr_mtx);
2469 /* There was no valid prison - perhaps there was a dying one. */
2470 if (deadpr != NULL) {
2471 mtx_lock(&deadpr->pr_mtx);
2472 if (deadpr->pr_ref == 0) {
2473 mtx_unlock(&deadpr->pr_mtx);
2481 * See if a prison has the specific flag set.
2484 prison_flag(struct ucred *cred, unsigned flag)
2487 /* This is an atomic read, so no locking is necessary. */
2488 return (cred->cr_prison->pr_flags & flag);
2492 prison_allow(struct ucred *cred, unsigned flag)
2495 /* This is an atomic read, so no locking is necessary. */
2496 return (cred->cr_prison->pr_allow & flag);
2500 * Remove a prison reference. If that was the last reference, remove the
2501 * prison itself - but not in this context in case there are locks held.
2504 prison_free_locked(struct prison *pr)
2508 mtx_assert(&pr->pr_mtx, MA_OWNED);
2510 mtx_unlock(&pr->pr_mtx);
2512 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2516 prison_free(struct prison *pr)
2519 mtx_lock(&pr->pr_mtx);
2520 prison_free_locked(pr);
2524 * Complete a call to either prison_free or prison_proc_free.
2527 prison_complete(void *context, int pending)
2529 struct prison *pr = context;
2531 sx_xlock(&allprison_lock);
2532 mtx_lock(&pr->pr_mtx);
2533 prison_deref(pr, pr->pr_uref
2534 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2535 : PD_LOCKED | PD_LIST_XLOCKED);
2539 * Remove a prison reference (usually). This internal version assumes no
2540 * mutexes are held, except perhaps the prison itself. If there are no more
2541 * references, release and delist the prison. On completion, the prison lock
2542 * and the allprison lock are both unlocked.
2545 prison_deref(struct prison *pr, int flags)
2547 struct prison *ppr, *tpr;
2550 if (!(flags & PD_LOCKED))
2551 mtx_lock(&pr->pr_mtx);
2553 if (flags & PD_DEUREF) {
2554 KASSERT(pr->pr_uref > 0,
2555 ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2558 lasturef = pr->pr_uref == 0;
2561 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2564 if (flags & PD_DEREF) {
2565 KASSERT(pr->pr_ref > 0,
2566 ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2571 mtx_unlock(&pr->pr_mtx);
2574 * Tell the modules if the last user reference was removed
2575 * (even it sticks around in dying state).
2578 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2579 sx_xlock(&allprison_lock);
2580 flags |= PD_LIST_XLOCKED;
2582 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2583 mtx_lock(&pr->pr_mtx);
2585 mtx_unlock(&pr->pr_mtx);
2588 /* If the prison still has references, nothing else to do. */
2590 if (flags & PD_LIST_SLOCKED)
2591 sx_sunlock(&allprison_lock);
2592 else if (flags & PD_LIST_XLOCKED)
2593 sx_xunlock(&allprison_lock);
2597 if (flags & PD_LIST_SLOCKED) {
2598 if (!sx_try_upgrade(&allprison_lock)) {
2599 sx_sunlock(&allprison_lock);
2600 sx_xlock(&allprison_lock);
2602 } else if (!(flags & PD_LIST_XLOCKED))
2603 sx_xlock(&allprison_lock);
2605 TAILQ_REMOVE(&allprison, pr, pr_list);
2606 LIST_REMOVE(pr, pr_sibling);
2607 ppr = pr->pr_parent;
2608 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2609 tpr->pr_childcount--;
2610 sx_xunlock(&allprison_lock);
2613 if (pr->pr_vnet != ppr->pr_vnet)
2614 vnet_destroy(pr->pr_vnet);
2616 if (pr->pr_root != NULL)
2618 mtx_destroy(&pr->pr_mtx);
2620 free(pr->pr_ip4, M_PRISON);
2623 free(pr->pr_ip6, M_PRISON);
2625 if (pr->pr_cpuset != NULL)
2626 cpuset_rel(pr->pr_cpuset);
2630 prison_racct_detach(pr);
2634 /* Removing a prison frees a reference on its parent. */
2636 mtx_lock(&pr->pr_mtx);
2637 flags = PD_DEREF | PD_DEUREF;
2642 prison_hold_locked(struct prison *pr)
2645 mtx_assert(&pr->pr_mtx, MA_OWNED);
2646 KASSERT(pr->pr_ref > 0,
2647 ("Trying to hold dead prison %p (jid=%d).", pr, pr->pr_id));
2652 prison_hold(struct prison *pr)
2655 mtx_lock(&pr->pr_mtx);
2656 prison_hold_locked(pr);
2657 mtx_unlock(&pr->pr_mtx);
2661 prison_proc_hold(struct prison *pr)
2664 mtx_lock(&pr->pr_mtx);
2665 KASSERT(pr->pr_uref > 0,
2666 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2668 mtx_unlock(&pr->pr_mtx);
2672 prison_proc_free(struct prison *pr)
2675 mtx_lock(&pr->pr_mtx);
2676 KASSERT(pr->pr_uref > 0,
2677 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2678 if (pr->pr_uref > 1)
2682 * Don't remove the last user reference in this context, which
2683 * is expected to be a process that is not only locked, but
2687 mtx_unlock(&pr->pr_mtx);
2688 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2691 mtx_unlock(&pr->pr_mtx);
2695 * Check if a jail supports the given address family.
2697 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
2701 prison_check_af(struct ucred *cred, int af)
2706 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2708 pr = cred->cr_prison;
2710 /* Prisons with their own network stack are not limited. */
2711 if (prison_owns_vnet(cred))
2720 if (pr->pr_flags & PR_IP4)
2722 mtx_lock(&pr->pr_mtx);
2723 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
2724 error = EAFNOSUPPORT;
2725 mtx_unlock(&pr->pr_mtx);
2731 if (pr->pr_flags & PR_IP6)
2733 mtx_lock(&pr->pr_mtx);
2734 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
2735 error = EAFNOSUPPORT;
2736 mtx_unlock(&pr->pr_mtx);
2744 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
2745 error = EAFNOSUPPORT;
2751 * Check if given address belongs to the jail referenced by cred (wrapper to
2752 * prison_check_ip[46]).
2754 * Returns 0 if jail doesn't restrict the address family or if address belongs
2755 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
2756 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
2759 prison_if(struct ucred *cred, const struct sockaddr *sa)
2762 const struct sockaddr_in *sai;
2765 const struct sockaddr_in6 *sai6;
2769 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2770 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
2773 if (prison_owns_vnet(cred))
2778 switch (sa->sa_family)
2782 sai = (const struct sockaddr_in *)sa;
2783 error = prison_check_ip4(cred, &sai->sin_addr);
2788 sai6 = (const struct sockaddr_in6 *)sa;
2789 error = prison_check_ip6(cred, &sai6->sin6_addr);
2793 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
2794 error = EAFNOSUPPORT;
2800 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
2803 prison_check(struct ucred *cred1, struct ucred *cred2)
2806 return ((cred1->cr_prison == cred2->cr_prison ||
2807 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
2811 * Return 1 if p2 is a child of p1, otherwise 0.
2814 prison_ischild(struct prison *pr1, struct prison *pr2)
2817 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
2824 * Return 1 if the passed credential is in a jail and that jail does not
2825 * have its own virtual network stack, otherwise 0.
2828 jailed_without_vnet(struct ucred *cred)
2834 if (prison_owns_vnet(cred))
2842 * Return the correct hostname (domainname, et al) for the passed credential.
2845 getcredhostname(struct ucred *cred, char *buf, size_t size)
2850 * A NULL credential can be used to shortcut to the physical
2851 * system's hostname.
2853 pr = (cred != NULL) ? cred->cr_prison : &prison0;
2854 mtx_lock(&pr->pr_mtx);
2855 strlcpy(buf, pr->pr_hostname, size);
2856 mtx_unlock(&pr->pr_mtx);
2860 getcreddomainname(struct ucred *cred, char *buf, size_t size)
2863 mtx_lock(&cred->cr_prison->pr_mtx);
2864 strlcpy(buf, cred->cr_prison->pr_domainname, size);
2865 mtx_unlock(&cred->cr_prison->pr_mtx);
2869 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
2872 mtx_lock(&cred->cr_prison->pr_mtx);
2873 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
2874 mtx_unlock(&cred->cr_prison->pr_mtx);
2878 getcredhostid(struct ucred *cred, unsigned long *hostid)
2881 mtx_lock(&cred->cr_prison->pr_mtx);
2882 *hostid = cred->cr_prison->pr_hostid;
2883 mtx_unlock(&cred->cr_prison->pr_mtx);
2888 * Determine whether the prison represented by cred owns
2889 * its vnet rather than having it inherited.
2891 * Returns 1 in case the prison owns the vnet, 0 otherwise.
2894 prison_owns_vnet(struct ucred *cred)
2898 * vnets cannot be added/removed after jail creation,
2899 * so no need to lock here.
2901 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
2906 * Determine whether the subject represented by cred can "see"
2907 * status of a mount point.
2908 * Returns: 0 for permitted, ENOENT otherwise.
2909 * XXX: This function should be called cr_canseemount() and should be
2910 * placed in kern_prot.c.
2913 prison_canseemount(struct ucred *cred, struct mount *mp)
2919 pr = cred->cr_prison;
2920 if (pr->pr_enforce_statfs == 0)
2922 if (pr->pr_root->v_mount == mp)
2924 if (pr->pr_enforce_statfs == 2)
2927 * If jail's chroot directory is set to "/" we should be able to see
2928 * all mount-points from inside a jail.
2929 * This is ugly check, but this is the only situation when jail's
2930 * directory ends with '/'.
2932 if (strcmp(pr->pr_path, "/") == 0)
2934 len = strlen(pr->pr_path);
2936 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
2939 * Be sure that we don't have situation where jail's root directory
2940 * is "/some/path" and mount point is "/some/pathpath".
2942 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
2948 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
2950 char jpath[MAXPATHLEN];
2954 pr = cred->cr_prison;
2955 if (pr->pr_enforce_statfs == 0)
2957 if (prison_canseemount(cred, mp) != 0) {
2958 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
2959 strlcpy(sp->f_mntonname, "[restricted]",
2960 sizeof(sp->f_mntonname));
2963 if (pr->pr_root->v_mount == mp) {
2965 * Clear current buffer data, so we are sure nothing from
2966 * the valid path left there.
2968 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
2969 *sp->f_mntonname = '/';
2973 * If jail's chroot directory is set to "/" we should be able to see
2974 * all mount-points from inside a jail.
2976 if (strcmp(pr->pr_path, "/") == 0)
2978 len = strlen(pr->pr_path);
2979 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
2981 * Clear current buffer data, so we are sure nothing from
2982 * the valid path left there.
2984 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
2985 if (*jpath == '\0') {
2986 /* Should never happen. */
2987 *sp->f_mntonname = '/';
2989 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
2994 * Check with permission for a specific privilege is granted within jail. We
2995 * have a specific list of accepted privileges; the rest are denied.
2998 prison_priv_check(struct ucred *cred, int priv)
3002 * Some policies have custom handlers. This routine should not be
3003 * called for them. See priv_check_cred().
3006 case PRIV_VFS_GENERATION:
3007 KASSERT(0, ("prison_priv_check instead of a custom handler "
3008 "called for %d\n", priv));
3016 * Privileges specific to prisons with a virtual network stack.
3017 * There might be a duplicate entry here in case the privilege
3018 * is only granted conditionally in the legacy jail case.
3023 * NFS-specific privileges.
3025 case PRIV_NFS_DAEMON:
3026 case PRIV_NFS_LOCKD:
3029 * Network stack privileges.
3031 case PRIV_NET_BRIDGE:
3034 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3035 case PRIV_NET_ROUTE:
3037 case PRIV_NET_SETIFMTU:
3038 case PRIV_NET_SETIFFLAGS:
3039 case PRIV_NET_SETIFCAP:
3040 case PRIV_NET_SETIFDESCR:
3041 case PRIV_NET_SETIFNAME :
3042 case PRIV_NET_SETIFMETRIC:
3043 case PRIV_NET_SETIFPHYS:
3044 case PRIV_NET_SETIFMAC:
3045 case PRIV_NET_SETLANPCP:
3046 case PRIV_NET_ADDMULTI:
3047 case PRIV_NET_DELMULTI:
3048 case PRIV_NET_HWIOCTL:
3049 case PRIV_NET_SETLLADDR:
3050 case PRIV_NET_ADDIFGROUP:
3051 case PRIV_NET_DELIFGROUP:
3052 case PRIV_NET_IFCREATE:
3053 case PRIV_NET_IFDESTROY:
3054 case PRIV_NET_ADDIFADDR:
3055 case PRIV_NET_DELIFADDR:
3058 case PRIV_NET_SETIFVNET:
3059 case PRIV_NET_SETIFFIB:
3062 * 802.11-related privileges.
3064 case PRIV_NET80211_GETKEY:
3066 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3073 case PRIV_NETATM_CFG:
3074 case PRIV_NETATM_ADD:
3075 case PRIV_NETATM_DEL:
3076 case PRIV_NETATM_SET:
3079 * Bluetooth privileges.
3081 case PRIV_NETBLUETOOTH_RAW:
3085 * Netgraph and netgraph module privileges.
3087 case PRIV_NETGRAPH_CONTROL:
3089 case PRIV_NETGRAPH_TTY:
3093 * IPv4 and IPv6 privileges.
3095 case PRIV_NETINET_IPFW:
3096 case PRIV_NETINET_DIVERT:
3097 case PRIV_NETINET_PF:
3098 case PRIV_NETINET_DUMMYNET:
3099 case PRIV_NETINET_CARP:
3100 case PRIV_NETINET_MROUTE:
3101 case PRIV_NETINET_RAW:
3102 case PRIV_NETINET_ADDRCTRL6:
3103 case PRIV_NETINET_ND6:
3104 case PRIV_NETINET_SCOPE6:
3105 case PRIV_NETINET_ALIFETIME6:
3106 case PRIV_NETINET_IPSEC:
3107 case PRIV_NETINET_BINDANY:
3122 * No default: or deny here.
3123 * In case of no permit fall through to next switch().
3125 if (cred->cr_prison->pr_flags & PR_VNET)
3133 * Allow ktrace privileges for root in jail.
3139 * Allow jailed processes to configure audit identity and
3140 * submit audit records (login, etc). In the future we may
3141 * want to further refine the relationship between audit and
3144 case PRIV_AUDIT_GETAUDIT:
3145 case PRIV_AUDIT_SETAUDIT:
3146 case PRIV_AUDIT_SUBMIT:
3150 * Allow jailed processes to manipulate process UNIX
3151 * credentials in any way they see fit.
3153 case PRIV_CRED_SETUID:
3154 case PRIV_CRED_SETEUID:
3155 case PRIV_CRED_SETGID:
3156 case PRIV_CRED_SETEGID:
3157 case PRIV_CRED_SETGROUPS:
3158 case PRIV_CRED_SETREUID:
3159 case PRIV_CRED_SETREGID:
3160 case PRIV_CRED_SETRESUID:
3161 case PRIV_CRED_SETRESGID:
3164 * Jail implements visibility constraints already, so allow
3165 * jailed root to override uid/gid-based constraints.
3167 case PRIV_SEEOTHERGIDS:
3168 case PRIV_SEEOTHERUIDS:
3171 * Jail implements inter-process debugging limits already, so
3172 * allow jailed root various debugging privileges.
3174 case PRIV_DEBUG_DIFFCRED:
3175 case PRIV_DEBUG_SUGID:
3176 case PRIV_DEBUG_UNPRIV:
3179 * Allow jail to set various resource limits and login
3180 * properties, and for now, exceed process resource limits.
3182 case PRIV_PROC_LIMIT:
3183 case PRIV_PROC_SETLOGIN:
3184 case PRIV_PROC_SETRLIMIT:
3187 * System V and POSIX IPC privileges are granted in jail.
3190 case PRIV_IPC_WRITE:
3191 case PRIV_IPC_ADMIN:
3192 case PRIV_IPC_MSGSIZE:
3196 * Jail operations within a jail work on child jails.
3198 case PRIV_JAIL_ATTACH:
3200 case PRIV_JAIL_REMOVE:
3203 * Jail implements its own inter-process limits, so allow
3204 * root processes in jail to change scheduling on other
3205 * processes in the same jail. Likewise for signalling.
3207 case PRIV_SCHED_DIFFCRED:
3208 case PRIV_SCHED_CPUSET:
3209 case PRIV_SIGNAL_DIFFCRED:
3210 case PRIV_SIGNAL_SUGID:
3213 * Allow jailed processes to write to sysctls marked as jail
3216 case PRIV_SYSCTL_WRITEJAIL:
3219 * Allow root in jail to manage a variety of quota
3220 * properties. These should likely be conditional on a
3221 * configuration option.
3223 case PRIV_VFS_GETQUOTA:
3224 case PRIV_VFS_SETQUOTA:
3227 * Since Jail relies on chroot() to implement file system
3228 * protections, grant many VFS privileges to root in jail.
3229 * Be careful to exclude mount-related and NFS-related
3233 case PRIV_VFS_WRITE:
3234 case PRIV_VFS_ADMIN:
3236 case PRIV_VFS_LOOKUP:
3237 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3238 case PRIV_VFS_CHFLAGS_DEV:
3239 case PRIV_VFS_CHOWN:
3240 case PRIV_VFS_CHROOT:
3241 case PRIV_VFS_RETAINSUGID:
3242 case PRIV_VFS_FCHROOT:
3244 case PRIV_VFS_SETGID:
3246 case PRIV_VFS_STICKYFILE:
3249 * As in the non-jail case, non-root users are expected to be
3250 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3251 * exists in the jail and they have permission to access it).
3253 case PRIV_KMEM_READ:
3257 * Depending on the global setting, allow privilege of
3258 * setting system flags.
3260 case PRIV_VFS_SYSFLAGS:
3261 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3267 * Depending on the global setting, allow privilege of
3268 * mounting/unmounting file systems.
3270 case PRIV_VFS_MOUNT:
3271 case PRIV_VFS_UNMOUNT:
3272 case PRIV_VFS_MOUNT_NONUSER:
3273 case PRIV_VFS_MOUNT_OWNER:
3274 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3275 cred->cr_prison->pr_enforce_statfs < 2)
3281 * Conditionnaly allow locking (unlocking) physical pages
3285 case PRIV_VM_MUNLOCK:
3286 if (cred->cr_prison->pr_allow & PR_ALLOW_MLOCK)
3292 * Conditionally allow jailed root to bind reserved ports.
3294 case PRIV_NETINET_RESERVEDPORT:
3295 if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS)
3301 * Allow jailed root to reuse in-use ports.
3303 case PRIV_NETINET_REUSEPORT:
3307 * Allow jailed root to set certain IPv4/6 (option) headers.
3309 case PRIV_NETINET_SETHDROPTS:
3313 * Conditionally allow creating raw sockets in jail.
3315 case PRIV_NETINET_RAW:
3316 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3322 * Since jail implements its own visibility limits on netstat
3323 * sysctls, allow getcred. This allows identd to work in
3326 case PRIV_NETINET_GETCRED:
3330 * Allow jailed root to set loginclass.
3332 case PRIV_PROC_SETLOGINCLASS:
3336 * Do not allow a process inside a jail to read the kernel
3337 * message buffer unless explicitly permitted.
3340 if (cred->cr_prison->pr_allow & PR_ALLOW_READ_MSGBUF)
3346 * In all remaining cases, deny the privilege request. This
3347 * includes almost all network privileges, many system
3348 * configuration privileges.
3355 * Return the part of pr2's name that is relative to pr1, or the whole name
3356 * if it does not directly follow.
3360 prison_name(struct prison *pr1, struct prison *pr2)
3364 /* Jails see themselves as "0" (if they see themselves at all). */
3367 name = pr2->pr_name;
3368 if (prison_ischild(pr1, pr2)) {
3370 * pr1 isn't locked (and allprison_lock may not be either)
3371 * so its length can't be counted on. But the number of dots
3372 * can be counted on - and counted.
3374 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3375 name = strchr(name, '.') + 1;
3381 * Return the part of pr2's path that is relative to pr1, or the whole path
3382 * if it does not directly follow.
3385 prison_path(struct prison *pr1, struct prison *pr2)
3387 char *path1, *path2;
3390 path1 = pr1->pr_path;
3391 path2 = pr2->pr_path;
3392 if (!strcmp(path1, "/"))
3394 len1 = strlen(path1);
3395 if (strncmp(path1, path2, len1))
3397 if (path2[len1] == '\0')
3399 if (path2[len1] == '/')
3400 return (path2 + len1);
3405 * Jail-related sysctls.
3407 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3411 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3414 struct prison *pr, *cpr;
3416 struct in_addr *ip4 = NULL;
3420 struct in6_addr *ip6 = NULL;
3425 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3426 pr = req->td->td_ucred->cr_prison;
3428 sx_slock(&allprison_lock);
3429 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3430 #if defined(INET) || defined(INET6)
3433 mtx_lock(&cpr->pr_mtx);
3435 if (cpr->pr_ip4s > 0) {
3436 if (ip4s < cpr->pr_ip4s) {
3437 ip4s = cpr->pr_ip4s;
3438 mtx_unlock(&cpr->pr_mtx);
3439 ip4 = realloc(ip4, ip4s *
3440 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3443 bcopy(cpr->pr_ip4, ip4,
3444 cpr->pr_ip4s * sizeof(struct in_addr));
3448 if (cpr->pr_ip6s > 0) {
3449 if (ip6s < cpr->pr_ip6s) {
3450 ip6s = cpr->pr_ip6s;
3451 mtx_unlock(&cpr->pr_mtx);
3452 ip6 = realloc(ip6, ip6s *
3453 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3456 bcopy(cpr->pr_ip6, ip6,
3457 cpr->pr_ip6s * sizeof(struct in6_addr));
3460 if (cpr->pr_ref == 0) {
3461 mtx_unlock(&cpr->pr_mtx);
3464 bzero(xp, sizeof(*xp));
3465 xp->pr_version = XPRISON_VERSION;
3466 xp->pr_id = cpr->pr_id;
3467 xp->pr_state = cpr->pr_uref > 0
3468 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3469 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3470 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3471 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3473 xp->pr_ip4s = cpr->pr_ip4s;
3476 xp->pr_ip6s = cpr->pr_ip6s;
3478 mtx_unlock(&cpr->pr_mtx);
3479 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3483 if (xp->pr_ip4s > 0) {
3484 error = SYSCTL_OUT(req, ip4,
3485 xp->pr_ip4s * sizeof(struct in_addr));
3491 if (xp->pr_ip6s > 0) {
3492 error = SYSCTL_OUT(req, ip6,
3493 xp->pr_ip6s * sizeof(struct in6_addr));
3499 sx_sunlock(&allprison_lock);
3510 SYSCTL_OID(_security_jail, OID_AUTO, list,
3511 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3512 sysctl_jail_list, "S", "List of active jails");
3515 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3519 injail = jailed(req->td->td_ucred);
3520 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3525 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3526 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3527 sysctl_jail_jailed, "I", "Process in jail?");
3530 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
3532 int error, havevnet;
3534 struct ucred *cred = req->td->td_ucred;
3536 havevnet = jailed(cred) && prison_owns_vnet(cred);
3540 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
3545 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
3546 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3547 sysctl_jail_vnet, "I", "Jail owns vnet?");
3549 #if defined(INET) || defined(INET6)
3550 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3551 &jail_max_af_ips, 0,
3552 "Number of IP addresses a jail may have at most per address family (deprecated)");
3556 * Default parameters for jail(2) compatibility. For historical reasons,
3557 * the sysctl names have varying similarity to the parameter names. Prisons
3558 * just see their own parameters, and can't change them.
3561 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3564 int allow, error, i;
3566 pr = req->td->td_ucred->cr_prison;
3567 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3569 /* Get the current flag value, and convert it to a boolean. */
3570 i = (allow & arg2) ? 1 : 0;
3573 error = sysctl_handle_int(oidp, &i, 0, req);
3574 if (error || !req->newptr)
3580 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3583 mtx_lock(&prison0.pr_mtx);
3584 jail_default_allow = (jail_default_allow & ~arg2) | i;
3585 mtx_unlock(&prison0.pr_mtx);
3589 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3590 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3591 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3592 "Processes in jail can set their hostnames (deprecated)");
3593 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3594 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3595 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3596 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
3597 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3598 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3599 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3600 "Processes in jail can use System V IPC primitives (deprecated)");
3601 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3602 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3603 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3604 "Prison root can create raw sockets (deprecated)");
3605 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3606 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3607 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3608 "Processes in jail can alter system file flags (deprecated)");
3609 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3610 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3611 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3612 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
3615 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3620 pr = req->td->td_ucred->cr_prison;
3621 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3622 error = sysctl_handle_int(oidp, &level, 0, req);
3623 if (error || !req->newptr)
3625 *(int *)arg1 = level;
3629 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3630 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3631 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3632 sysctl_jail_default_level, "I",
3633 "Processes in jail cannot see all mounted file systems (deprecated)");
3635 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
3636 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
3637 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
3638 sysctl_jail_default_level, "I",
3639 "Ruleset for the devfs filesystem in jail (deprecated)");
3642 * Nodes to describe jail parameters. Maximum length of string parameters
3643 * is returned in the string itself, and the other parameters exist merely
3644 * to make themselves and their types known.
3646 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
3650 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3657 switch (oidp->oid_kind & CTLTYPE)
3663 if (!(req->flags & SCTL_MASK32))
3665 return (SYSCTL_OUT(req, &l, sizeof(l)));
3669 return (SYSCTL_OUT(req, &i, sizeof(i)));
3670 case CTLTYPE_STRING:
3671 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
3673 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3674 case CTLTYPE_STRUCT:
3676 return (SYSCTL_OUT(req, &s, sizeof(s)));
3682 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
3683 * jail creation time but cannot be changed in an existing jail.
3685 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3686 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3687 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3688 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3689 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3690 "I", "Jail secure level");
3691 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
3692 "Jail value for kern.osreldate and uname -K");
3693 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
3694 "Jail value for kern.osrelease and uname -r");
3695 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3696 "I", "Jail cannot see all mounted file systems");
3697 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
3698 "I", "Ruleset for in-jail devfs mounts");
3699 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3700 "B", "Jail persistence");
3702 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3703 "E,jailsys", "Virtual network stack");
3705 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3706 "B", "Jail is in the process of shutting down");
3708 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
3709 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
3710 "I", "Current number of child jails");
3711 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
3712 "I", "Maximum number of child jails");
3714 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
3715 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3717 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3718 "Jail NIS domainname");
3719 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3721 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3722 "LU", "Jail host ID");
3724 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3725 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3728 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
3729 "Jail IPv4 address virtualization");
3730 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3731 "S,in_addr,a", "Jail IPv4 addresses");
3732 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3733 "B", "Do (not) use IPv4 source address selection rather than the "
3734 "primary jail IPv4 address.");
3737 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
3738 "Jail IPv6 address virtualization");
3739 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3740 "S,in6_addr,a", "Jail IPv6 addresses");
3741 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3742 "B", "Do (not) use IPv6 source address selection rather than the "
3743 "primary jail IPv6 address.");
3746 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3747 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3748 "B", "Jail may set hostname");
3749 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3750 "B", "Jail may use SYSV IPC");
3751 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3752 "B", "Jail may create raw sockets");
3753 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3754 "B", "Jail may alter system file flags");
3755 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3756 "B", "Jail may set file quotas");
3757 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3758 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3759 SYSCTL_JAIL_PARAM(_allow, mlock, CTLTYPE_INT | CTLFLAG_RW,
3760 "B", "Jail may lock (unlock) physical pages in memory");
3761 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW,
3762 "B", "Jail may bind sockets to reserved ports");
3763 SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW,
3764 "B", "Jail may read the kernel message buffer");
3765 SYSCTL_JAIL_PARAM(_allow, unprivileged_proc_debug, CTLTYPE_INT | CTLFLAG_RW,
3766 "B", "Unprivileged processes may use process debugging facilities");
3768 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
3769 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
3770 "B", "Jail may mount/unmount jail-friendly file systems in general");
3773 * Add a dynamic parameter allow.<name>, or allow.<prefix>.<name>. Return
3774 * its associated bit in the pr_allow bitmask, or zero if the parameter was
3778 prison_add_allow(const char *prefix, const char *name, const char *prefix_descr,
3781 struct bool_flags *bf;
3782 struct sysctl_oid *parent;
3783 char *allow_name, *allow_noname, *allowed;
3784 #ifndef NO_SYSCTL_DESCR
3785 char *descr_deprecated;
3787 unsigned allow_flag;
3790 ? asprintf(&allow_name, M_PRISON, "allow.%s.%s", prefix, name)
3792 asprintf(&allow_noname, M_PRISON, "allow.%s.no%s", prefix, name)
3794 : asprintf(&allow_name, M_PRISON, "allow.%s", name) < 0 ||
3795 asprintf(&allow_noname, M_PRISON, "allow.no%s", name) < 0) {
3796 free(allow_name, M_PRISON);
3801 * See if this parameter has already beed added, i.e. a module was
3802 * previously loaded/unloaded.
3804 mtx_lock(&prison0.pr_mtx);
3805 for (bf = pr_flag_allow;
3806 bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
3808 if (strcmp(bf->name, allow_name) == 0) {
3809 allow_flag = bf->flag;
3815 * Find a free bit in prison0's pr_allow, failing if there are none
3816 * (which shouldn't happen as long as we keep track of how many
3817 * potential dynamic flags exist).
3819 * Due to per-jail unprivileged process debugging support
3820 * using pr_allow, also verify against PR_ALLOW_ALL_STATIC.
3821 * prison0 may have unprivileged process debugging unset.
3823 for (allow_flag = 1;; allow_flag <<= 1) {
3824 if (allow_flag == 0)
3826 if (allow_flag & PR_ALLOW_ALL_STATIC)
3828 if ((prison0.pr_allow & allow_flag) == 0)
3833 * Note the parameter in the next open slot in pr_flag_allow.
3834 * Set the flag last so code that checks pr_flag_allow can do so
3837 for (bf = pr_flag_allow; bf->flag != 0; bf++)
3838 if (bf == pr_flag_allow + nitems(pr_flag_allow)) {
3839 /* This should never happen, but is not fatal. */
3843 prison0.pr_allow |= allow_flag;
3844 bf->name = allow_name;
3845 bf->noname = allow_noname;
3846 bf->flag = allow_flag;
3847 mtx_unlock(&prison0.pr_mtx);
3850 * Create sysctls for the paramter, and the back-compat global
3854 ? SYSCTL_ADD_NODE(NULL,
3855 SYSCTL_CHILDREN(&sysctl___security_jail_param_allow),
3856 OID_AUTO, prefix, 0, 0, prefix_descr)
3857 : &sysctl___security_jail_param_allow;
3858 (void)SYSCTL_ADD_PROC(NULL, SYSCTL_CHILDREN(parent), OID_AUTO,
3859 name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3860 NULL, 0, sysctl_jail_param, "B", descr);
3862 ? asprintf(&allowed, M_TEMP, "%s_%s_allowed", prefix, name)
3863 : asprintf(&allowed, M_TEMP, "%s_allowed", name)) >= 0) {
3864 #ifndef NO_SYSCTL_DESCR
3865 (void)asprintf(&descr_deprecated, M_TEMP, "%s (deprecated)",
3868 (void)SYSCTL_ADD_PROC(NULL,
3869 SYSCTL_CHILDREN(&sysctl___security_jail), OID_AUTO, allowed,
3870 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, allow_flag,
3871 sysctl_jail_default_allow, "I", descr_deprecated);
3872 #ifndef NO_SYSCTL_DESCR
3873 free(descr_deprecated, M_TEMP);
3875 free(allowed, M_TEMP);
3880 mtx_unlock(&prison0.pr_mtx);
3881 free(allow_name, M_PRISON);
3882 free(allow_noname, M_PRISON);
3887 * The VFS system will register jail-aware filesystems here. They each get
3888 * a parameter allow.mount.xxxfs and a flag to check when a jailed user
3889 * attempts to mount.
3892 prison_add_vfs(struct vfsconf *vfsp)
3894 #ifdef NO_SYSCTL_DESCR
3896 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
3901 (void)asprintf(&descr, M_TEMP, "Jail may mount the %s file system",
3903 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name,
3905 free(descr, M_TEMP);
3911 prison_racct_foreach(void (*callback)(struct racct *racct,
3912 void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
3913 void *arg2, void *arg3)
3915 struct prison_racct *prr;
3917 ASSERT_RACCT_ENABLED();
3919 sx_slock(&allprison_lock);
3922 LIST_FOREACH(prr, &allprison_racct, prr_next)
3923 (callback)(prr->prr_racct, arg2, arg3);
3926 sx_sunlock(&allprison_lock);
3929 static struct prison_racct *
3930 prison_racct_find_locked(const char *name)
3932 struct prison_racct *prr;
3934 ASSERT_RACCT_ENABLED();
3935 sx_assert(&allprison_lock, SA_XLOCKED);
3937 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
3940 LIST_FOREACH(prr, &allprison_racct, prr_next) {
3941 if (strcmp(name, prr->prr_name) != 0)
3944 /* Found prison_racct with a matching name? */
3945 prison_racct_hold(prr);
3949 /* Add new prison_racct. */
3950 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
3951 racct_create(&prr->prr_racct);
3953 strcpy(prr->prr_name, name);
3954 refcount_init(&prr->prr_refcount, 1);
3955 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
3960 struct prison_racct *
3961 prison_racct_find(const char *name)
3963 struct prison_racct *prr;
3965 ASSERT_RACCT_ENABLED();
3967 sx_xlock(&allprison_lock);
3968 prr = prison_racct_find_locked(name);
3969 sx_xunlock(&allprison_lock);
3974 prison_racct_hold(struct prison_racct *prr)
3977 ASSERT_RACCT_ENABLED();
3979 refcount_acquire(&prr->prr_refcount);
3983 prison_racct_free_locked(struct prison_racct *prr)
3986 ASSERT_RACCT_ENABLED();
3987 sx_assert(&allprison_lock, SA_XLOCKED);
3989 if (refcount_release(&prr->prr_refcount)) {
3990 racct_destroy(&prr->prr_racct);
3991 LIST_REMOVE(prr, prr_next);
3992 free(prr, M_PRISON_RACCT);
3997 prison_racct_free(struct prison_racct *prr)
4000 ASSERT_RACCT_ENABLED();
4001 sx_assert(&allprison_lock, SA_UNLOCKED);
4003 if (refcount_release_if_not_last(&prr->prr_refcount))
4006 sx_xlock(&allprison_lock);
4007 prison_racct_free_locked(prr);
4008 sx_xunlock(&allprison_lock);
4012 prison_racct_attach(struct prison *pr)
4014 struct prison_racct *prr;
4016 ASSERT_RACCT_ENABLED();
4017 sx_assert(&allprison_lock, SA_XLOCKED);
4019 prr = prison_racct_find_locked(pr->pr_name);
4020 KASSERT(prr != NULL, ("cannot find prison_racct"));
4022 pr->pr_prison_racct = prr;
4026 * Handle jail renaming. From the racct point of view, renaming means
4027 * moving from one prison_racct to another.
4030 prison_racct_modify(struct prison *pr)
4036 struct prison_racct *oldprr;
4038 ASSERT_RACCT_ENABLED();
4040 sx_slock(&allproc_lock);
4041 sx_xlock(&allprison_lock);
4043 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4044 sx_xunlock(&allprison_lock);
4045 sx_sunlock(&allproc_lock);
4049 oldprr = pr->pr_prison_racct;
4050 pr->pr_prison_racct = NULL;
4052 prison_racct_attach(pr);
4055 * Move resource utilisation records.
4057 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4061 * Force rctl to reattach rules to processes.
4063 FOREACH_PROC_IN_SYSTEM(p) {
4065 cred = crhold(p->p_ucred);
4067 rctl_proc_ucred_changed(p, cred);
4072 sx_sunlock(&allproc_lock);
4073 prison_racct_free_locked(oldprr);
4074 sx_xunlock(&allprison_lock);
4078 prison_racct_detach(struct prison *pr)
4081 ASSERT_RACCT_ENABLED();
4082 sx_assert(&allprison_lock, SA_UNLOCKED);
4084 if (pr->pr_prison_racct == NULL)
4086 prison_racct_free(pr->pr_prison_racct);
4087 pr->pr_prison_racct = NULL;
4094 db_show_prison(struct prison *pr)
4096 struct bool_flags *bf;
4097 struct jailsys_flags *jsf;
4098 #if defined(INET) || defined(INET6)
4103 char ip4buf[INET_ADDRSTRLEN];
4106 char ip6buf[INET6_ADDRSTRLEN];
4109 db_printf("prison %p:\n", pr);
4110 db_printf(" jid = %d\n", pr->pr_id);
4111 db_printf(" name = %s\n", pr->pr_name);
4112 db_printf(" parent = %p\n", pr->pr_parent);
4113 db_printf(" ref = %d\n", pr->pr_ref);
4114 db_printf(" uref = %d\n", pr->pr_uref);
4115 db_printf(" path = %s\n", pr->pr_path);
4116 db_printf(" cpuset = %d\n", pr->pr_cpuset
4117 ? pr->pr_cpuset->cs_id : -1);
4119 db_printf(" vnet = %p\n", pr->pr_vnet);
4121 db_printf(" root = %p\n", pr->pr_root);
4122 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4123 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4124 db_printf(" children.max = %d\n", pr->pr_childmax);
4125 db_printf(" children.cur = %d\n", pr->pr_childcount);
4126 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4127 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4128 db_printf(" flags = 0x%x", pr->pr_flags);
4129 for (bf = pr_flag_bool; bf < pr_flag_bool + nitems(pr_flag_bool); bf++)
4130 if (pr->pr_flags & bf->flag)
4131 db_printf(" %s", bf->name);
4132 for (jsf = pr_flag_jailsys;
4133 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
4135 f = pr->pr_flags & (jsf->disable | jsf->new);
4136 db_printf(" %-16s= %s\n", jsf->name,
4137 (f != 0 && f == jsf->disable) ? "disable"
4138 : (f == jsf->new) ? "new"
4141 db_printf(" allow = 0x%x", pr->pr_allow);
4142 for (bf = pr_flag_allow;
4143 bf < pr_flag_allow + nitems(pr_flag_allow) && bf->flag != 0;
4145 if (pr->pr_allow & bf->flag)
4146 db_printf(" %s", bf->name);
4148 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4149 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4150 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4151 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4152 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4154 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4155 for (ii = 0; ii < pr->pr_ip4s; ii++)
4156 db_printf(" %s %s\n",
4157 ii == 0 ? "ip4.addr =" : " ",
4158 inet_ntoa_r(pr->pr_ip4[ii], ip4buf));
4161 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4162 for (ii = 0; ii < pr->pr_ip6s; ii++)
4163 db_printf(" %s %s\n",
4164 ii == 0 ? "ip6.addr =" : " ",
4165 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4169 DB_SHOW_COMMAND(prison, db_show_prison_command)
4175 * Show all prisons in the list, and prison0 which is not
4178 db_show_prison(&prison0);
4179 if (!db_pager_quit) {
4180 TAILQ_FOREACH(pr, &allprison, pr_list) {
4192 /* Look for a prison with the ID and with references. */
4193 TAILQ_FOREACH(pr, &allprison, pr_list)
4194 if (pr->pr_id == addr && pr->pr_ref > 0)
4197 /* Look again, without requiring a reference. */
4198 TAILQ_FOREACH(pr, &allprison, pr_list)
4199 if (pr->pr_id == addr)
4202 /* Assume address points to a valid prison. */
4203 pr = (struct prison *)addr;