2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_compat.h"
35 #include "opt_inet6.h"
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
67 #include <netinet/in.h>
73 #include <security/mac/mac_framework.h>
75 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
77 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
78 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
80 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
83 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
85 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
89 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
91 #define _PR_IP_SADDRSEL 0
95 /* prison0 describes what is "real" about the system. */
96 struct prison prison0 = {
102 .pr_securelevel = -1,
104 .pr_childmax = JAIL_MAX,
105 .pr_hostuuid = DEFAULT_HOSTUUID,
106 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
108 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
110 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
112 .pr_allow = PR_ALLOW_ALL,
114 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
116 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
117 struct sx allprison_lock;
118 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
119 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
120 LIST_HEAD(, prison_racct) allprison_racct;
123 static int do_jail_attach(struct thread *td, struct prison *pr);
124 static void prison_complete(void *context, int pending);
125 static void prison_deref(struct prison *pr, int flags);
126 static char *prison_path(struct prison *pr1, struct prison *pr2);
127 static void prison_remove_one(struct prison *pr);
129 static void prison_racct_attach(struct prison *pr);
130 static void prison_racct_modify(struct prison *pr);
131 static void prison_racct_detach(struct prison *pr);
134 static int _prison_check_ip4(const struct prison *, const struct in_addr *);
135 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
138 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
139 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
142 /* Flags for prison_deref */
143 #define PD_DEREF 0x01
144 #define PD_DEUREF 0x02
145 #define PD_LOCKED 0x04
146 #define PD_LIST_SLOCKED 0x08
147 #define PD_LIST_XLOCKED 0x10
150 * Parameter names corresponding to PR_* flag values. Size values are for kvm
151 * as we cannot figure out the size of a sparse array, or an array without a
154 static char *pr_flag_names[] = {
157 [7] = "ip4.saddrsel",
160 [8] = "ip6.saddrsel",
163 const size_t pr_flag_names_size = sizeof(pr_flag_names);
165 static char *pr_flag_nonames[] = {
168 [7] = "ip4.nosaddrsel",
171 [8] = "ip6.nosaddrsel",
174 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
176 struct jailsys_flags {
180 } pr_flag_jailsys[] = {
181 { "host", 0, PR_HOST },
183 { "vnet", 0, PR_VNET },
186 { "ip4", PR_IP4_USER, PR_IP4_USER },
189 { "ip6", PR_IP6_USER, PR_IP6_USER },
192 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
194 static char *pr_allow_names[] = {
195 "allow.set_hostname",
203 "allow.mount.nullfs",
205 "allow.mount.procfs",
207 "allow.mount.fdescfs",
208 "allow.mount.linprocfs",
209 "allow.mount.linsysfs",
211 const size_t pr_allow_names_size = sizeof(pr_allow_names);
213 static char *pr_allow_nonames[] = {
214 "allow.noset_hostname",
216 "allow.noraw_sockets",
221 "allow.mount.nodevfs",
222 "allow.mount.nonullfs",
224 "allow.mount.noprocfs",
225 "allow.mount.notmpfs",
226 "allow.mount.nofdescfs",
227 "allow.mount.nolinprocfs",
228 "allow.mount.nolinsysfs",
230 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
232 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
233 #define JAIL_DEFAULT_ENFORCE_STATFS 2
234 #define JAIL_DEFAULT_DEVFS_RSNUM 0
235 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
236 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
237 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
238 #if defined(INET) || defined(INET6)
239 static unsigned jail_max_af_ips = 255;
243 * Initialize the parts of prison0 that can't be static-initialized with
244 * constants. This is called from proc0_init() after creating thread0 cpuset.
250 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
251 prison0.pr_osreldate = osreldate;
252 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
257 qcmp_v4(const void *ip1, const void *ip2)
262 * We need to compare in HBO here to get the list sorted as expected
263 * by the result of the code. Sorting NBO addresses gives you
264 * interesting results. If you do not understand, do not try.
266 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
267 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
270 * Do not simply return the difference of the two numbers, the int is
284 qcmp_v6(const void *ip1, const void *ip2)
286 const struct in6_addr *ia6a, *ia6b;
289 ia6a = (const struct in6_addr *)ip1;
290 ia6b = (const struct in6_addr *)ip2;
293 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
294 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
296 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
309 sys_jail(struct thread *td, struct jail_args *uap)
315 error = copyin(uap->jail, &version, sizeof(uint32_t));
324 /* FreeBSD single IPv4 jails. */
325 bzero(&j, sizeof(struct jail));
326 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
329 j.version = j0.version;
331 j.hostname = j0.hostname;
332 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
338 * Version 1 was used by multi-IPv4 jail implementations
339 * that never made it into the official kernel.
343 case 2: /* JAIL_API_VERSION */
344 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
345 error = copyin(uap->jail, &j, sizeof(struct jail));
351 /* Sci-Fi jails are not supported, sorry. */
354 return (kern_jail(td, &j));
358 kern_jail(struct thread *td, struct jail *j)
360 struct iovec optiov[2 * (4
361 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
370 char *u_path, *u_hostname, *u_name;
373 struct in_addr *u_ip4;
376 struct in6_addr *u_ip6;
379 int error, enforce_statfs, fi;
381 bzero(&optiov, sizeof(optiov));
382 opt.uio_iov = optiov;
386 opt.uio_segflg = UIO_SYSSPACE;
387 opt.uio_rw = UIO_READ;
390 /* Set permissions for top-level jails from sysctls. */
391 if (!jailed(td->td_ucred)) {
392 for (fi = 0; fi < sizeof(pr_allow_names) /
393 sizeof(pr_allow_names[0]); fi++) {
394 optiov[opt.uio_iovcnt].iov_base =
395 (jail_default_allow & (1 << fi))
396 ? pr_allow_names[fi] : pr_allow_nonames[fi];
397 optiov[opt.uio_iovcnt].iov_len =
398 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
401 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
402 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
404 enforce_statfs = jail_default_enforce_statfs;
405 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
406 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
410 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
412 ip4s = (j->version == 0) ? 1 : j->ip4s;
413 if (ip4s > jail_max_af_ips)
415 tmplen += ip4s * sizeof(struct in_addr);
421 if (j->ip6s > jail_max_af_ips)
423 tmplen += j->ip6s * sizeof(struct in6_addr);
428 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
429 u_hostname = u_path + MAXPATHLEN;
430 u_name = u_hostname + MAXHOSTNAMELEN;
432 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
436 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
438 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
441 optiov[opt.uio_iovcnt].iov_base = "path";
442 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
444 optiov[opt.uio_iovcnt].iov_base = u_path;
445 error = copyinstr(j->path, u_path, MAXPATHLEN,
446 &optiov[opt.uio_iovcnt].iov_len);
448 free(u_path, M_TEMP);
452 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
453 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
455 optiov[opt.uio_iovcnt].iov_base = u_hostname;
456 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
457 &optiov[opt.uio_iovcnt].iov_len);
459 free(u_path, M_TEMP);
463 if (j->jailname != NULL) {
464 optiov[opt.uio_iovcnt].iov_base = "name";
465 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
467 optiov[opt.uio_iovcnt].iov_base = u_name;
468 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
469 &optiov[opt.uio_iovcnt].iov_len);
471 free(u_path, M_TEMP);
477 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
478 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
480 optiov[opt.uio_iovcnt].iov_base = u_ip4;
481 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
483 u_ip4->s_addr = j->ip4s;
485 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
487 free(u_path, M_TEMP);
494 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
495 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
497 optiov[opt.uio_iovcnt].iov_base = u_ip6;
498 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
499 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
501 free(u_path, M_TEMP);
506 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
507 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
508 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
509 free(u_path, M_TEMP);
515 * struct jail_set_args {
516 * struct iovec *iovp;
517 * unsigned int iovcnt;
522 sys_jail_set(struct thread *td, struct jail_set_args *uap)
527 /* Check that we have an even number of iovecs. */
531 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
534 error = kern_jail_set(td, auio, uap->flags);
540 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
547 struct in6_addr *ip6;
550 struct vfsoptlist *opts;
551 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
553 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
554 char *g_path, *osrelstr;
555 #if defined(INET) || defined(INET6)
560 size_t namelen, onamelen;
561 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
562 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
563 int fi, jid, jsys, len, level;
564 int childmax, osreldt, rsnum, slevel;
565 int fullpath_disabled;
566 #if defined(INET) || defined(INET6)
575 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
579 error = priv_check(td, PRIV_JAIL_SET);
580 if (!error && (flags & JAIL_ATTACH))
581 error = priv_check(td, PRIV_JAIL_ATTACH);
584 mypr = ppr = td->td_ucred->cr_prison;
585 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
587 if (flags & ~JAIL_SET_MASK)
591 * Check all the parameters before committing to anything. Not all
592 * errors can be caught early, but we may as well try. Also, this
593 * takes care of some expensive stuff (path lookup) before getting
594 * the allprison lock.
596 * XXX Jails are not filesystems, and jail parameters are not mount
597 * options. But it makes more sense to re-use the vfsopt code
598 * than duplicate it under a different name.
600 error = vfs_buildopts(optuio, &opts);
611 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
617 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
626 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
634 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
639 else if (enforce < 0 || enforce > 2) {
645 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
653 pr_flags = ch_flags = 0;
654 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
656 if (pr_flag_names[fi] == NULL)
658 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
659 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
661 ch_flags |= pr_flags;
662 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
664 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
671 case JAIL_SYS_DISABLE:
672 if (!pr_flag_jailsys[fi].disable) {
676 pr_flags |= pr_flag_jailsys[fi].disable;
679 pr_flags |= pr_flag_jailsys[fi].new;
681 case JAIL_SYS_INHERIT:
688 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
690 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
691 && !(pr_flags & PR_PERSIST)) {
693 vfs_opterror(opts, "new jail must persist or attach");
697 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
699 vfs_opterror(opts, "vnet cannot be changed after creation");
704 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
706 vfs_opterror(opts, "ip4 cannot be changed after creation");
711 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
713 vfs_opterror(opts, "ip6 cannot be changed after creation");
718 pr_allow = ch_allow = 0;
719 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
721 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
722 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
724 ch_allow |= pr_allow;
726 error = vfs_getopt(opts, "name", (void **)&name, &len);
732 if (len == 0 || name[len - 1] != '\0') {
736 if (len > MAXHOSTNAMELEN) {
737 error = ENAMETOOLONG;
742 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
750 if (len == 0 || host[len - 1] != '\0') {
754 if (len > MAXHOSTNAMELEN) {
755 error = ENAMETOOLONG;
760 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
768 if (len == 0 || domain[len - 1] != '\0') {
772 if (len > MAXHOSTNAMELEN) {
773 error = ENAMETOOLONG;
778 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
786 if (len == 0 || uuid[len - 1] != '\0') {
790 if (len > HOSTUUIDLEN) {
791 error = ENAMETOOLONG;
796 #ifdef COMPAT_FREEBSD32
797 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
800 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
804 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
816 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
821 else if (ip4s & (sizeof(*ip4) - 1)) {
825 ch_flags |= PR_IP4_USER;
826 pr_flags |= PR_IP4_USER;
828 ip4s /= sizeof(*ip4);
829 if (ip4s > jail_max_af_ips) {
831 vfs_opterror(opts, "too many IPv4 addresses");
834 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
835 bcopy(op, ip4, ip4s * sizeof(*ip4));
837 * IP addresses are all sorted but ip[0] to preserve
838 * the primary IP address as given from userland.
839 * This special IP is used for unbound outgoing
840 * connections as well for "loopback" traffic in case
841 * source address selection cannot find any more fitting
842 * address to connect from.
845 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
847 * Check for duplicate addresses and do some simple
848 * zero and broadcast checks. If users give other bogus
849 * addresses it is their problem.
851 * We do not have to care about byte order for these
852 * checks so we will do them in NBO.
854 for (ii = 0; ii < ip4s; ii++) {
855 if (ip4[ii].s_addr == INADDR_ANY ||
856 ip4[ii].s_addr == INADDR_BROADCAST) {
861 (ip4[0].s_addr == ip4[ii+1].s_addr ||
862 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
872 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
877 else if (ip6s & (sizeof(*ip6) - 1)) {
881 ch_flags |= PR_IP6_USER;
882 pr_flags |= PR_IP6_USER;
884 ip6s /= sizeof(*ip6);
885 if (ip6s > jail_max_af_ips) {
887 vfs_opterror(opts, "too many IPv6 addresses");
890 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
891 bcopy(op, ip6, ip6s * sizeof(*ip6));
893 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
894 for (ii = 0; ii < ip6s; ii++) {
895 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
900 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
901 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
911 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
912 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
915 "vnet jails cannot have IP address restrictions");
920 fullpath_disabled = 0;
922 error = vfs_getopt(opts, "path", (void **)&path, &len);
928 if (flags & JAIL_UPDATE) {
931 "path cannot be changed after creation");
934 if (len == 0 || path[len - 1] != '\0') {
938 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
944 NDFREE(&nd, NDF_ONLY_PNBUF);
945 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
946 strlcpy(g_path, path, MAXPATHLEN);
947 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
950 else if (error == ENODEV) {
951 /* proceed if sysctl debug.disablefullpath == 1 */
952 fullpath_disabled = 1;
953 if (len < 2 || (len == 2 && path[0] == '/'))
956 /* exit on other errors */
959 if (root->v_type != VDIR) {
965 if (fullpath_disabled) {
966 /* Leave room for a real-root full pathname. */
967 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
968 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
969 error = ENAMETOOLONG;
975 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
981 if (flags & JAIL_UPDATE) {
984 "osrelease cannot be changed after creation");
987 if (len == 0 || len >= OSRELEASELEN) {
990 "osrelease string must be 1-%d bytes long",
996 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
1002 if (flags & JAIL_UPDATE) {
1005 "osreldate cannot be changed after creation");
1010 vfs_opterror(opts, "osreldate cannot be 0");
1016 * Grab the allprison lock before letting modules check their
1017 * parameters. Once we have it, do not let go so we'll have a
1018 * consistent view of the OSD list.
1020 sx_xlock(&allprison_lock);
1021 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
1023 goto done_unlock_list;
1025 /* By now, all parameters should have been noted. */
1026 TAILQ_FOREACH(opt, opts, link) {
1027 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1029 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1030 goto done_unlock_list;
1035 * See if we are creating a new record or updating an existing one.
1036 * This abuses the file error codes ENOENT and EEXIST.
1038 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
1041 vfs_opterror(opts, "no valid operation (create or update)");
1042 goto done_unlock_list;
1046 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1047 namelc = strrchr(name, '.');
1048 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1054 * See if a requested jid already exists. There is an
1055 * information leak here if the jid exists but is not within
1056 * the caller's jail hierarchy. Jail creators will get EEXIST
1057 * even though they cannot see the jail, and CREATE | UPDATE
1058 * will return ENOENT which is not normally a valid error.
1062 vfs_opterror(opts, "negative jid");
1063 goto done_unlock_list;
1065 pr = prison_find(jid);
1067 ppr = pr->pr_parent;
1068 /* Create: jid must not exist. */
1069 if (cuflags == JAIL_CREATE) {
1070 mtx_unlock(&pr->pr_mtx);
1072 vfs_opterror(opts, "jail %d already exists",
1074 goto done_unlock_list;
1076 if (!prison_ischild(mypr, pr)) {
1077 mtx_unlock(&pr->pr_mtx);
1079 } else if (pr->pr_uref == 0) {
1080 if (!(flags & JAIL_DYING)) {
1081 mtx_unlock(&pr->pr_mtx);
1083 vfs_opterror(opts, "jail %d is dying",
1085 goto done_unlock_list;
1086 } else if ((flags & JAIL_ATTACH) ||
1087 (pr_flags & PR_PERSIST)) {
1089 * A dying jail might be resurrected
1090 * (via attach or persist), but first
1091 * it must determine if another jail
1092 * has claimed its name. Accomplish
1093 * this by implicitly re-setting the
1097 name = prison_name(mypr, pr);
1102 /* Update: jid must exist. */
1103 if (cuflags == JAIL_UPDATE) {
1105 vfs_opterror(opts, "jail %d not found", jid);
1106 goto done_unlock_list;
1111 * If the caller provided a name, look for a jail by that name.
1112 * This has different semantics for creates and updates keyed by jid
1113 * (where the name must not already exist in a different jail),
1114 * and updates keyed by the name itself (where the name must exist
1115 * because that is the jail being updated).
1118 namelc = strrchr(name, '.');
1123 * This is a hierarchical name. Split it into the
1124 * parent and child names, and make sure the parent
1125 * exists or matches an already found jail.
1129 if (strncmp(name, ppr->pr_name, namelc - name)
1130 || ppr->pr_name[namelc - name] != '\0') {
1131 mtx_unlock(&pr->pr_mtx);
1134 "cannot change jail's parent");
1135 goto done_unlock_list;
1138 ppr = prison_find_name(mypr, name);
1142 "jail \"%s\" not found", name);
1143 goto done_unlock_list;
1145 mtx_unlock(&ppr->pr_mtx);
1149 if (name[0] != '\0') {
1151 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1154 FOREACH_PRISON_CHILD(ppr, tpr) {
1155 if (tpr != pr && tpr->pr_ref > 0 &&
1156 !strcmp(tpr->pr_name + namelen, name)) {
1158 cuflags != JAIL_CREATE) {
1159 mtx_lock(&tpr->pr_mtx);
1160 if (tpr->pr_ref > 0) {
1165 if (tpr->pr_uref > 0) {
1171 mtx_unlock(&tpr->pr_mtx);
1172 } else if (tpr->pr_uref > 0) {
1174 * Create, or update(jid):
1175 * name must not exist in an
1176 * active sibling jail.
1180 mtx_unlock(&pr->pr_mtx);
1182 "jail \"%s\" already exists",
1184 goto done_unlock_list;
1188 /* If no active jail is found, use a dying one. */
1189 if (deadpr != NULL && pr == NULL) {
1190 if (flags & JAIL_DYING) {
1191 mtx_lock(&deadpr->pr_mtx);
1192 if (deadpr->pr_ref == 0) {
1193 mtx_unlock(&deadpr->pr_mtx);
1197 } else if (cuflags == JAIL_UPDATE) {
1200 "jail \"%s\" is dying", name);
1201 goto done_unlock_list;
1204 /* Update: name must exist if no jid. */
1205 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1207 vfs_opterror(opts, "jail \"%s\" not found",
1209 goto done_unlock_list;
1213 /* Update: must provide a jid or name. */
1214 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1216 vfs_opterror(opts, "update specified no jail");
1217 goto done_unlock_list;
1220 /* If there's no prison to update, create a new one and link it in. */
1222 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1223 if (tpr->pr_childcount >= tpr->pr_childmax) {
1225 vfs_opterror(opts, "prison limit exceeded");
1226 goto done_unlock_list;
1229 mtx_lock(&ppr->pr_mtx);
1230 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1231 mtx_unlock(&ppr->pr_mtx);
1233 vfs_opterror(opts, "parent jail went away!");
1234 goto done_unlock_list;
1238 mtx_unlock(&ppr->pr_mtx);
1239 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1241 /* Find the next free jid. */
1244 if (jid == JAIL_MAX)
1246 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1247 if (tpr->pr_id < jid)
1249 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1250 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1253 if (jid == lastprid) {
1256 "no available jail IDs");
1258 prison_deref(ppr, PD_DEREF |
1259 PD_DEUREF | PD_LIST_XLOCKED);
1268 * The jail already has a jid (that did not yet exist),
1269 * so just find where to insert it.
1271 TAILQ_FOREACH(tpr, &allprison, pr_list)
1272 if (tpr->pr_id >= jid) {
1273 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1278 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1279 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1280 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1281 tpr->pr_childcount++;
1283 pr->pr_parent = ppr;
1286 /* Set some default values, and inherit some from the parent. */
1291 root = mypr->pr_root;
1294 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1295 pr->pr_flags |= PR_HOST;
1296 #if defined(INET) || defined(INET6)
1298 if (!(pr_flags & PR_VNET))
1302 if (!(ch_flags & PR_IP4_USER))
1303 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1304 else if (!(pr_flags & PR_IP4_USER)) {
1305 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1306 if (ppr->pr_ip4 != NULL) {
1307 pr->pr_ip4s = ppr->pr_ip4s;
1308 pr->pr_ip4 = malloc(pr->pr_ip4s *
1309 sizeof(struct in_addr), M_PRISON,
1311 bcopy(ppr->pr_ip4, pr->pr_ip4,
1312 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1317 if (!(ch_flags & PR_IP6_USER))
1318 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1319 else if (!(pr_flags & PR_IP6_USER)) {
1320 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1321 if (ppr->pr_ip6 != NULL) {
1322 pr->pr_ip6s = ppr->pr_ip6s;
1323 pr->pr_ip6 = malloc(pr->pr_ip6s *
1324 sizeof(struct in6_addr), M_PRISON,
1326 bcopy(ppr->pr_ip6, pr->pr_ip6,
1327 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1333 /* Source address selection is always on by default. */
1334 pr->pr_flags |= _PR_IP_SADDRSEL;
1336 pr->pr_securelevel = ppr->pr_securelevel;
1337 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1338 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1339 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1341 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1342 if (osrelstr == NULL)
1343 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1345 strcpy(pr->pr_osrelease, osrelstr);
1347 LIST_INIT(&pr->pr_children);
1348 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1351 /* Allocate a new vnet if specified. */
1352 pr->pr_vnet = (pr_flags & PR_VNET)
1353 ? vnet_alloc() : ppr->pr_vnet;
1356 * Allocate a dedicated cpuset for each jail.
1357 * Unlike other initial settings, this may return an erorr.
1359 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1361 prison_deref(pr, PD_LIST_XLOCKED);
1365 mtx_lock(&pr->pr_mtx);
1367 * New prisons do not yet have a reference, because we do not
1368 * want other to see the incomplete prison once the
1369 * allprison_lock is downgraded.
1374 * Grab a reference for existing prisons, to ensure they
1375 * continue to exist for the duration of the call.
1378 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1379 if ((pr->pr_flags & PR_VNET) &&
1380 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1383 "vnet jails cannot have IP address restrictions");
1384 goto done_deref_locked;
1388 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1391 "ip4 cannot be changed after creation");
1392 goto done_deref_locked;
1396 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1399 "ip6 cannot be changed after creation");
1400 goto done_deref_locked;
1405 /* Do final error checking before setting anything. */
1407 if (slevel < ppr->pr_securelevel) {
1409 goto done_deref_locked;
1413 if (childmax >= ppr->pr_childmax) {
1415 goto done_deref_locked;
1419 if (enforce < ppr->pr_enforce_statfs) {
1421 goto done_deref_locked;
1426 * devfs_rsnum is a uint16_t
1428 if (rsnum < 0 || rsnum > 65535) {
1430 goto done_deref_locked;
1433 * Nested jails always inherit parent's devfs ruleset
1435 if (jailed(td->td_ucred)) {
1436 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1438 goto done_deref_locked;
1440 rsnum = ppr->pr_devfs_rsnum;
1445 if (ppr->pr_flags & PR_IP4) {
1447 * Make sure the new set of IP addresses is a
1448 * subset of the parent's list. Don't worry
1449 * about the parent being unlocked, as any
1450 * setting is done with allprison_lock held.
1452 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1453 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1455 if (ij == ppr->pr_ip4s) {
1457 goto done_deref_locked;
1460 for (ii = ij = 1; ii < ip4s; ii++) {
1461 if (ip4[ii].s_addr ==
1462 ppr->pr_ip4[0].s_addr)
1464 for (; ij < ppr->pr_ip4s; ij++)
1465 if (ip4[ii].s_addr ==
1466 ppr->pr_ip4[ij].s_addr)
1468 if (ij == ppr->pr_ip4s)
1471 if (ij == ppr->pr_ip4s) {
1473 goto done_deref_locked;
1478 * Check for conflicting IP addresses. We permit them
1479 * if there is no more than one IP on each jail. If
1480 * there is a duplicate on a jail with more than one
1481 * IP stop checking and return error.
1485 for (; tppr != &prison0; tppr = tppr->pr_parent)
1486 if (tppr->pr_flags & PR_VNET)
1489 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1492 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1494 tpr->pr_uref == 0) {
1498 if (!(tpr->pr_flags & PR_IP4_USER))
1501 if (tpr->pr_ip4 == NULL ||
1502 (ip4s == 1 && tpr->pr_ip4s == 1))
1504 for (ii = 0; ii < ip4s; ii++) {
1505 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1508 "IPv4 addresses clash");
1509 goto done_deref_locked;
1517 if (ppr->pr_flags & PR_IP6) {
1519 * Make sure the new set of IP addresses is a
1520 * subset of the parent's list.
1522 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1523 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1526 if (ij == ppr->pr_ip6s) {
1528 goto done_deref_locked;
1531 for (ii = ij = 1; ii < ip6s; ii++) {
1532 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1535 for (; ij < ppr->pr_ip6s; ij++)
1536 if (IN6_ARE_ADDR_EQUAL(
1537 &ip6[ii], &ppr->pr_ip6[ij]))
1539 if (ij == ppr->pr_ip6s)
1542 if (ij == ppr->pr_ip6s) {
1544 goto done_deref_locked;
1548 /* Check for conflicting IP addresses. */
1551 for (; tppr != &prison0; tppr = tppr->pr_parent)
1552 if (tppr->pr_flags & PR_VNET)
1555 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1558 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1560 tpr->pr_uref == 0) {
1564 if (!(tpr->pr_flags & PR_IP6_USER))
1567 if (tpr->pr_ip6 == NULL ||
1568 (ip6s == 1 && tpr->pr_ip6s == 1))
1570 for (ii = 0; ii < ip6s; ii++) {
1571 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1574 "IPv6 addresses clash");
1575 goto done_deref_locked;
1581 onamelen = namelen = 0;
1583 /* Give a default name of the jid. */
1584 if (name[0] == '\0')
1585 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1586 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1590 "name cannot be numeric (unless it is the jid)");
1591 goto done_deref_locked;
1594 * Make sure the name isn't too long for the prison or its
1597 onamelen = strlen(pr->pr_name);
1598 namelen = strlen(name);
1599 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1600 error = ENAMETOOLONG;
1601 goto done_deref_locked;
1603 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1604 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1605 sizeof(pr->pr_name)) {
1606 error = ENAMETOOLONG;
1607 goto done_deref_locked;
1611 if (pr_allow & ~ppr->pr_allow) {
1613 goto done_deref_locked;
1616 /* Set the parameters of the prison. */
1619 if (pr_flags & PR_IP4_USER) {
1620 pr->pr_flags |= PR_IP4;
1621 free(pr->pr_ip4, M_PRISON);
1625 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1627 if (tpr->pr_flags & PR_VNET) {
1632 if (prison_restrict_ip4(tpr, NULL)) {
1641 if (pr_flags & PR_IP6_USER) {
1642 pr->pr_flags |= PR_IP6;
1643 free(pr->pr_ip6, M_PRISON);
1647 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1649 if (tpr->pr_flags & PR_VNET) {
1654 if (prison_restrict_ip6(tpr, NULL)) {
1662 pr->pr_securelevel = slevel;
1663 /* Set all child jails to be at least this level. */
1664 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1665 if (tpr->pr_securelevel < slevel)
1666 tpr->pr_securelevel = slevel;
1669 pr->pr_childmax = childmax;
1670 /* Set all child jails to under this limit. */
1671 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1672 if (tpr->pr_childmax > childmax - level)
1673 tpr->pr_childmax = childmax > level
1674 ? childmax - level : 0;
1677 pr->pr_enforce_statfs = enforce;
1678 /* Pass this restriction on to the children. */
1679 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1680 if (tpr->pr_enforce_statfs < enforce)
1681 tpr->pr_enforce_statfs = enforce;
1684 pr->pr_devfs_rsnum = rsnum;
1685 /* Pass this restriction on to the children. */
1686 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1687 tpr->pr_devfs_rsnum = rsnum;
1690 if (ppr == &prison0)
1691 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1693 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1694 ppr->pr_name, name);
1695 /* Change this component of child names. */
1696 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1697 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1698 strlen(tpr->pr_name + onamelen) + 1);
1699 bcopy(pr->pr_name, tpr->pr_name, namelen);
1703 /* Try to keep a real-rooted full pathname. */
1704 if (fullpath_disabled && path[0] == '/' &&
1705 strcmp(mypr->pr_path, "/"))
1706 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1707 mypr->pr_path, path);
1709 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1712 if (PR_HOST & ch_flags & ~pr_flags) {
1713 if (pr->pr_flags & PR_HOST) {
1715 * Copy the parent's host info. As with pr_ip4 above,
1716 * the lack of a lock on the parent is not a problem;
1717 * it is always set with allprison_lock at least
1718 * shared, and is held exclusively here.
1720 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1721 sizeof(pr->pr_hostname));
1722 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1723 sizeof(pr->pr_domainname));
1724 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1725 sizeof(pr->pr_hostuuid));
1726 pr->pr_hostid = pr->pr_parent->pr_hostid;
1728 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1729 /* Set this prison, and any descendants without PR_HOST. */
1731 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1733 strlcpy(pr->pr_domainname, domain,
1734 sizeof(pr->pr_domainname));
1736 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1738 pr->pr_hostid = hid;
1739 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1740 if (tpr->pr_flags & PR_HOST)
1744 strlcpy(tpr->pr_hostname,
1746 sizeof(tpr->pr_hostname));
1748 strlcpy(tpr->pr_domainname,
1750 sizeof(tpr->pr_domainname));
1752 strlcpy(tpr->pr_hostuuid,
1754 sizeof(tpr->pr_hostuuid));
1756 tpr->pr_hostid = hid;
1760 if ((tallow = ch_allow & ~pr_allow)) {
1761 /* Clear allow bits in all children. */
1762 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1763 tpr->pr_allow &= ~tallow;
1765 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1767 * Persistent prisons get an extra reference, and prisons losing their
1768 * persist flag lose that reference. Only do this for existing prisons
1769 * for now, so new ones will remain unseen until after the module
1770 * handlers have completed.
1772 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1773 if (pr_flags & PR_PERSIST) {
1781 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1782 mtx_unlock(&pr->pr_mtx);
1785 if (racct_enable && created)
1786 prison_racct_attach(pr);
1789 /* Locks may have prevented a complete restriction of child IP
1790 * addresses. If so, allocate some more memory and try again.
1795 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1796 mtx_lock(&pr->pr_mtx);
1798 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1800 if (tpr->pr_flags & PR_VNET) {
1805 if (prison_restrict_ip4(tpr, ip4)) {
1812 mtx_unlock(&pr->pr_mtx);
1818 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1819 mtx_lock(&pr->pr_mtx);
1821 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1823 if (tpr->pr_flags & PR_VNET) {
1828 if (prison_restrict_ip6(tpr, ip6)) {
1835 mtx_unlock(&pr->pr_mtx);
1839 /* Let the modules do their work. */
1840 sx_downgrade(&allprison_lock);
1842 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1844 prison_deref(pr, PD_LIST_SLOCKED);
1848 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1850 prison_deref(pr, created
1852 : PD_DEREF | PD_LIST_SLOCKED);
1856 /* Attach this process to the prison if requested. */
1857 if (flags & JAIL_ATTACH) {
1858 mtx_lock(&pr->pr_mtx);
1859 error = do_jail_attach(td, pr);
1861 vfs_opterror(opts, "attach failed");
1863 prison_deref(pr, PD_DEREF);
1869 if (racct_enable && !created) {
1870 if (!(flags & JAIL_ATTACH))
1871 sx_sunlock(&allprison_lock);
1872 prison_racct_modify(pr);
1873 if (!(flags & JAIL_ATTACH))
1874 sx_slock(&allprison_lock);
1878 td->td_retval[0] = pr->pr_id;
1881 * Now that it is all there, drop the temporary reference from existing
1882 * prisons. Or add a reference to newly created persistent prisons
1883 * (which was not done earlier so that the prison would not be publicly
1887 prison_deref(pr, (flags & JAIL_ATTACH)
1889 : PD_DEREF | PD_LIST_SLOCKED);
1891 if (pr_flags & PR_PERSIST) {
1892 mtx_lock(&pr->pr_mtx);
1895 mtx_unlock(&pr->pr_mtx);
1897 if (!(flags & JAIL_ATTACH))
1898 sx_sunlock(&allprison_lock);
1904 prison_deref(pr, created
1905 ? PD_LOCKED | PD_LIST_XLOCKED
1906 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1909 sx_xunlock(&allprison_lock);
1915 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1916 if (errmsg_len > 0) {
1917 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1918 if (errmsg_pos > 0) {
1919 if (optuio->uio_segflg == UIO_SYSSPACE)
1921 optuio->uio_iov[errmsg_pos].iov_base,
1925 optuio->uio_iov[errmsg_pos].iov_base,
1932 free(ip4, M_PRISON);
1935 free(ip6, M_PRISON);
1938 free(g_path, M_TEMP);
1945 * struct jail_get_args {
1946 * struct iovec *iovp;
1947 * unsigned int iovcnt;
1952 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1957 /* Check that we have an even number of iovecs. */
1958 if (uap->iovcnt & 1)
1961 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1964 error = kern_jail_get(td, auio, uap->flags);
1966 error = copyout(auio->uio_iov, uap->iovp,
1967 uap->iovcnt * sizeof (struct iovec));
1973 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1975 struct prison *pr, *mypr;
1977 struct vfsoptlist *opts;
1978 char *errmsg, *name;
1979 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1981 if (flags & ~JAIL_GET_MASK)
1984 /* Get the parameter list. */
1985 error = vfs_buildopts(optuio, &opts);
1988 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1989 mypr = td->td_ucred->cr_prison;
1992 * Find the prison specified by one of: lastjid, jid, name.
1994 sx_slock(&allprison_lock);
1995 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1997 TAILQ_FOREACH(pr, &allprison, pr_list) {
1998 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1999 mtx_lock(&pr->pr_mtx);
2000 if (pr->pr_ref > 0 &&
2001 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2003 mtx_unlock(&pr->pr_mtx);
2009 vfs_opterror(opts, "no jail after %d", jid);
2010 goto done_unlock_list;
2011 } else if (error != ENOENT)
2012 goto done_unlock_list;
2014 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2017 pr = prison_find_child(mypr, jid);
2019 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2020 mtx_unlock(&pr->pr_mtx);
2022 vfs_opterror(opts, "jail %d is dying",
2024 goto done_unlock_list;
2029 vfs_opterror(opts, "jail %d not found", jid);
2030 goto done_unlock_list;
2032 } else if (error != ENOENT)
2033 goto done_unlock_list;
2035 error = vfs_getopt(opts, "name", (void **)&name, &len);
2037 if (len == 0 || name[len - 1] != '\0') {
2039 goto done_unlock_list;
2041 pr = prison_find_name(mypr, name);
2043 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2044 mtx_unlock(&pr->pr_mtx);
2046 vfs_opterror(opts, "jail \"%s\" is dying",
2048 goto done_unlock_list;
2053 vfs_opterror(opts, "jail \"%s\" not found", name);
2054 goto done_unlock_list;
2055 } else if (error != ENOENT)
2056 goto done_unlock_list;
2058 vfs_opterror(opts, "no jail specified");
2060 goto done_unlock_list;
2063 /* Get the parameters of the prison. */
2066 td->td_retval[0] = pr->pr_id;
2067 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2068 if (error != 0 && error != ENOENT)
2070 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2071 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2072 if (error != 0 && error != ENOENT)
2074 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2075 if (error != 0 && error != ENOENT)
2077 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2078 sizeof(pr->pr_cpuset->cs_id));
2079 if (error != 0 && error != ENOENT)
2081 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2082 if (error != 0 && error != ENOENT)
2085 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2086 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2087 if (error != 0 && error != ENOENT)
2091 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2092 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2093 if (error != 0 && error != ENOENT)
2096 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2097 sizeof(pr->pr_securelevel));
2098 if (error != 0 && error != ENOENT)
2100 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2101 sizeof(pr->pr_childcount));
2102 if (error != 0 && error != ENOENT)
2104 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2105 sizeof(pr->pr_childmax));
2106 if (error != 0 && error != ENOENT)
2108 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2109 if (error != 0 && error != ENOENT)
2111 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2112 if (error != 0 && error != ENOENT)
2114 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2115 if (error != 0 && error != ENOENT)
2117 #ifdef COMPAT_FREEBSD32
2118 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2119 uint32_t hid32 = pr->pr_hostid;
2121 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2124 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2125 sizeof(pr->pr_hostid));
2126 if (error != 0 && error != ENOENT)
2128 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2129 sizeof(pr->pr_enforce_statfs));
2130 if (error != 0 && error != ENOENT)
2132 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2133 sizeof(pr->pr_devfs_rsnum));
2134 if (error != 0 && error != ENOENT)
2136 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2138 if (pr_flag_names[fi] == NULL)
2140 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2141 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2142 if (error != 0 && error != ENOENT)
2145 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2146 if (error != 0 && error != ENOENT)
2149 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2152 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2153 i = pr_flag_jailsys[fi].disable &&
2154 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2155 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2158 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2159 if (error != 0 && error != ENOENT)
2162 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2164 if (pr_allow_names[fi] == NULL)
2166 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2167 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2168 if (error != 0 && error != ENOENT)
2171 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2172 if (error != 0 && error != ENOENT)
2175 i = (pr->pr_uref == 0);
2176 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2177 if (error != 0 && error != ENOENT)
2180 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2181 if (error != 0 && error != ENOENT)
2183 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2184 sizeof(pr->pr_osreldate));
2185 if (error != 0 && error != ENOENT)
2187 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2188 if (error != 0 && error != ENOENT)
2191 /* Get the module parameters. */
2192 mtx_unlock(&pr->pr_mtx);
2194 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2197 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2199 /* By now, all parameters should have been noted. */
2200 TAILQ_FOREACH(opt, opts, link) {
2201 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2203 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2208 /* Write the fetched parameters back to userspace. */
2210 TAILQ_FOREACH(opt, opts, link) {
2211 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2212 pos = 2 * opt->pos + 1;
2213 optuio->uio_iov[pos].iov_len = opt->len;
2214 if (opt->value != NULL) {
2215 if (optuio->uio_segflg == UIO_SYSSPACE) {
2217 optuio->uio_iov[pos].iov_base,
2220 error = copyout(opt->value,
2221 optuio->uio_iov[pos].iov_base,
2232 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2236 sx_sunlock(&allprison_lock);
2238 if (error && errmsg_pos >= 0) {
2239 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2240 errmsg_pos = 2 * errmsg_pos + 1;
2241 if (errmsg_len > 0) {
2242 if (optuio->uio_segflg == UIO_SYSSPACE)
2244 optuio->uio_iov[errmsg_pos].iov_base,
2248 optuio->uio_iov[errmsg_pos].iov_base,
2258 * struct jail_remove_args {
2263 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2265 struct prison *pr, *cpr, *lpr, *tpr;
2268 error = priv_check(td, PRIV_JAIL_REMOVE);
2272 sx_xlock(&allprison_lock);
2273 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2275 sx_xunlock(&allprison_lock);
2279 /* Remove all descendants of this prison, then remove this prison. */
2281 pr->pr_flags |= PR_REMOVE;
2282 if (!LIST_EMPTY(&pr->pr_children)) {
2283 mtx_unlock(&pr->pr_mtx);
2285 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2286 mtx_lock(&cpr->pr_mtx);
2287 if (cpr->pr_ref > 0) {
2290 cpr->pr_flags |= PR_REMOVE;
2292 /* Already removed - do not do it again. */
2295 mtx_unlock(&cpr->pr_mtx);
2297 mtx_lock(&lpr->pr_mtx);
2298 prison_remove_one(lpr);
2299 sx_xlock(&allprison_lock);
2304 mtx_lock(&lpr->pr_mtx);
2305 prison_remove_one(lpr);
2306 sx_xlock(&allprison_lock);
2308 mtx_lock(&pr->pr_mtx);
2310 prison_remove_one(pr);
2315 prison_remove_one(struct prison *pr)
2320 /* If the prison was persistent, it is not anymore. */
2322 if (pr->pr_flags & PR_PERSIST) {
2325 pr->pr_flags &= ~PR_PERSIST;
2329 * jail_remove added a reference. If that's the only one, remove
2332 KASSERT(pr->pr_ref > 0,
2333 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2334 if (pr->pr_ref == 1) {
2336 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2340 mtx_unlock(&pr->pr_mtx);
2341 sx_xunlock(&allprison_lock);
2343 * Kill all processes unfortunate enough to be attached to this prison.
2345 sx_slock(&allproc_lock);
2346 LIST_FOREACH(p, &allproc, p_list) {
2348 if (p->p_state != PRS_NEW && p->p_ucred &&
2349 p->p_ucred->cr_prison == pr)
2350 kern_psignal(p, SIGKILL);
2353 sx_sunlock(&allproc_lock);
2354 /* Remove the temporary reference added by jail_remove. */
2355 prison_deref(pr, deuref | PD_DEREF);
2360 * struct jail_attach_args {
2365 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2370 error = priv_check(td, PRIV_JAIL_ATTACH);
2374 sx_slock(&allprison_lock);
2375 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2377 sx_sunlock(&allprison_lock);
2382 * Do not allow a process to attach to a prison that is not
2383 * considered to be "alive".
2385 if (pr->pr_uref == 0) {
2386 mtx_unlock(&pr->pr_mtx);
2387 sx_sunlock(&allprison_lock);
2391 return (do_jail_attach(td, pr));
2395 do_jail_attach(struct thread *td, struct prison *pr)
2399 struct ucred *newcred, *oldcred;
2403 * XXX: Note that there is a slight race here if two threads
2404 * in the same privileged process attempt to attach to two
2405 * different jails at the same time. It is important for
2406 * user processes not to do this, or they might end up with
2407 * a process root from one prison, but attached to the jail
2412 mtx_unlock(&pr->pr_mtx);
2414 /* Let modules do whatever they need to prepare for attaching. */
2415 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2417 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2420 sx_sunlock(&allprison_lock);
2423 * Reparent the newly attached process to this jail.
2425 ppr = td->td_ucred->cr_prison;
2427 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2431 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2432 if ((error = change_dir(pr->pr_root, td)) != 0)
2435 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2438 VOP_UNLOCK(pr->pr_root, 0);
2439 if ((error = pwd_chroot(td, pr->pr_root)))
2444 oldcred = p->p_ucred;
2446 crcopy(newcred, oldcred);
2447 newcred->cr_prison = pr;
2448 proc_set_cred(p, newcred);
2451 racct_proc_ucred_changed(p, oldcred, newcred);
2454 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2457 VOP_UNLOCK(pr->pr_root, 0);
2459 /* Tell modules this thread is still in its old jail after all. */
2460 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2461 prison_deref(pr, PD_DEREF | PD_DEUREF);
2467 * Returns a locked prison instance, or NULL on failure.
2470 prison_find(int prid)
2474 sx_assert(&allprison_lock, SX_LOCKED);
2475 TAILQ_FOREACH(pr, &allprison, pr_list) {
2476 if (pr->pr_id == prid) {
2477 mtx_lock(&pr->pr_mtx);
2480 mtx_unlock(&pr->pr_mtx);
2487 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2490 prison_find_child(struct prison *mypr, int prid)
2495 sx_assert(&allprison_lock, SX_LOCKED);
2496 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2497 if (pr->pr_id == prid) {
2498 mtx_lock(&pr->pr_mtx);
2501 mtx_unlock(&pr->pr_mtx);
2508 * Look for the name relative to mypr. Returns a locked prison or NULL.
2511 prison_find_name(struct prison *mypr, const char *name)
2513 struct prison *pr, *deadpr;
2517 sx_assert(&allprison_lock, SX_LOCKED);
2518 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2521 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2522 if (!strcmp(pr->pr_name + mylen, name)) {
2523 mtx_lock(&pr->pr_mtx);
2524 if (pr->pr_ref > 0) {
2525 if (pr->pr_uref > 0)
2529 mtx_unlock(&pr->pr_mtx);
2532 /* There was no valid prison - perhaps there was a dying one. */
2533 if (deadpr != NULL) {
2534 mtx_lock(&deadpr->pr_mtx);
2535 if (deadpr->pr_ref == 0) {
2536 mtx_unlock(&deadpr->pr_mtx);
2544 * See if a prison has the specific flag set.
2547 prison_flag(struct ucred *cred, unsigned flag)
2550 /* This is an atomic read, so no locking is necessary. */
2551 return (cred->cr_prison->pr_flags & flag);
2555 prison_allow(struct ucred *cred, unsigned flag)
2558 /* This is an atomic read, so no locking is necessary. */
2559 return (cred->cr_prison->pr_allow & flag);
2563 * Remove a prison reference. If that was the last reference, remove the
2564 * prison itself - but not in this context in case there are locks held.
2567 prison_free_locked(struct prison *pr)
2570 mtx_assert(&pr->pr_mtx, MA_OWNED);
2572 if (pr->pr_ref == 0) {
2573 mtx_unlock(&pr->pr_mtx);
2574 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2575 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2578 mtx_unlock(&pr->pr_mtx);
2582 prison_free(struct prison *pr)
2585 mtx_lock(&pr->pr_mtx);
2586 prison_free_locked(pr);
2590 prison_complete(void *context, int pending)
2593 prison_deref((struct prison *)context, 0);
2597 * Remove a prison reference (usually). This internal version assumes no
2598 * mutexes are held, except perhaps the prison itself. If there are no more
2599 * references, release and delist the prison. On completion, the prison lock
2600 * and the allprison lock are both unlocked.
2603 prison_deref(struct prison *pr, int flags)
2605 struct prison *ppr, *tpr;
2607 if (!(flags & PD_LOCKED))
2608 mtx_lock(&pr->pr_mtx);
2610 if (flags & PD_DEUREF) {
2612 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2614 if (flags & PD_DEREF)
2616 /* If the prison still has references, nothing else to do. */
2617 if (pr->pr_ref > 0) {
2618 mtx_unlock(&pr->pr_mtx);
2619 if (flags & PD_LIST_SLOCKED)
2620 sx_sunlock(&allprison_lock);
2621 else if (flags & PD_LIST_XLOCKED)
2622 sx_xunlock(&allprison_lock);
2626 mtx_unlock(&pr->pr_mtx);
2627 if (flags & PD_LIST_SLOCKED) {
2628 if (!sx_try_upgrade(&allprison_lock)) {
2629 sx_sunlock(&allprison_lock);
2630 sx_xlock(&allprison_lock);
2632 } else if (!(flags & PD_LIST_XLOCKED))
2633 sx_xlock(&allprison_lock);
2635 TAILQ_REMOVE(&allprison, pr, pr_list);
2636 LIST_REMOVE(pr, pr_sibling);
2637 ppr = pr->pr_parent;
2638 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2639 tpr->pr_childcount--;
2640 sx_xunlock(&allprison_lock);
2643 if (pr->pr_vnet != ppr->pr_vnet)
2644 vnet_destroy(pr->pr_vnet);
2646 if (pr->pr_root != NULL)
2648 mtx_destroy(&pr->pr_mtx);
2650 free(pr->pr_ip4, M_PRISON);
2653 free(pr->pr_ip6, M_PRISON);
2655 if (pr->pr_cpuset != NULL)
2656 cpuset_rel(pr->pr_cpuset);
2660 prison_racct_detach(pr);
2664 /* Removing a prison frees a reference on its parent. */
2666 mtx_lock(&pr->pr_mtx);
2667 flags = PD_DEREF | PD_DEUREF;
2672 prison_hold_locked(struct prison *pr)
2675 mtx_assert(&pr->pr_mtx, MA_OWNED);
2676 KASSERT(pr->pr_ref > 0,
2677 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2682 prison_hold(struct prison *pr)
2685 mtx_lock(&pr->pr_mtx);
2686 prison_hold_locked(pr);
2687 mtx_unlock(&pr->pr_mtx);
2691 prison_proc_hold(struct prison *pr)
2694 mtx_lock(&pr->pr_mtx);
2695 KASSERT(pr->pr_uref > 0,
2696 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2698 mtx_unlock(&pr->pr_mtx);
2702 prison_proc_free(struct prison *pr)
2705 mtx_lock(&pr->pr_mtx);
2706 KASSERT(pr->pr_uref > 0,
2707 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2708 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2714 * Restrict a prison's IP address list with its parent's, possibly replacing
2715 * it. Return true if the replacement buffer was used (or would have been).
2718 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2723 ppr = pr->pr_parent;
2724 if (!(pr->pr_flags & PR_IP4_USER)) {
2725 /* This has no user settings, so just copy the parent's list. */
2726 if (pr->pr_ip4s < ppr->pr_ip4s) {
2728 * There's no room for the parent's list. Use the
2729 * new list buffer, which is assumed to be big enough
2730 * (if it was passed). If there's no buffer, try to
2734 if (newip4 == NULL) {
2735 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2736 M_PRISON, M_NOWAIT);
2740 if (newip4 != NULL) {
2741 bcopy(ppr->pr_ip4, newip4,
2742 ppr->pr_ip4s * sizeof(*newip4));
2743 free(pr->pr_ip4, M_PRISON);
2744 pr->pr_ip4 = newip4;
2745 pr->pr_ip4s = ppr->pr_ip4s;
2749 pr->pr_ip4s = ppr->pr_ip4s;
2750 if (pr->pr_ip4s > 0)
2751 bcopy(ppr->pr_ip4, pr->pr_ip4,
2752 pr->pr_ip4s * sizeof(*newip4));
2753 else if (pr->pr_ip4 != NULL) {
2754 free(pr->pr_ip4, M_PRISON);
2757 } else if (pr->pr_ip4s > 0) {
2758 /* Remove addresses that aren't in the parent. */
2759 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2760 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2762 if (ij < ppr->pr_ip4s)
2765 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2766 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2769 for (ij = 1; ii < pr->pr_ip4s; ) {
2770 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2774 switch (ij >= ppr->pr_ip4s ? -1 :
2775 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2777 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2778 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2789 if (pr->pr_ip4s == 0) {
2790 free(pr->pr_ip4, M_PRISON);
2798 * Pass back primary IPv4 address of this jail.
2800 * If not restricted return success but do not alter the address. Caller has
2801 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2803 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2804 * Address returned in NBO.
2807 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2811 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2812 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2814 pr = cred->cr_prison;
2815 if (!(pr->pr_flags & PR_IP4))
2817 mtx_lock(&pr->pr_mtx);
2818 if (!(pr->pr_flags & PR_IP4)) {
2819 mtx_unlock(&pr->pr_mtx);
2822 if (pr->pr_ip4 == NULL) {
2823 mtx_unlock(&pr->pr_mtx);
2824 return (EAFNOSUPPORT);
2827 ia->s_addr = pr->pr_ip4[0].s_addr;
2828 mtx_unlock(&pr->pr_mtx);
2833 * Return 1 if we should do proper source address selection or are not jailed.
2834 * We will return 0 if we should bypass source address selection in favour
2835 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2837 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2840 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2846 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2847 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2852 pr = cred->cr_prison;
2853 if (pr->pr_flags & PR_IP4_SADDRSEL)
2856 lia.s_addr = INADDR_ANY;
2857 error = prison_get_ip4(cred, &lia);
2860 if (lia.s_addr == INADDR_ANY)
2863 ia->s_addr = lia.s_addr;
2868 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2871 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2878 * No need to lock since the PR_IP4_USER flag can't be altered for
2881 while (pr1 != &prison0 &&
2883 !(pr1->pr_flags & PR_VNET) &&
2885 !(pr1->pr_flags & PR_IP4_USER))
2886 pr1 = pr1->pr_parent;
2887 while (pr2 != &prison0 &&
2889 !(pr2->pr_flags & PR_VNET) &&
2891 !(pr2->pr_flags & PR_IP4_USER))
2892 pr2 = pr2->pr_parent;
2893 return (pr1 == pr2);
2897 * Make sure our (source) address is set to something meaningful to this
2900 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2901 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2902 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2905 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2911 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2912 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2914 pr = cred->cr_prison;
2915 if (!(pr->pr_flags & PR_IP4))
2917 mtx_lock(&pr->pr_mtx);
2918 if (!(pr->pr_flags & PR_IP4)) {
2919 mtx_unlock(&pr->pr_mtx);
2922 if (pr->pr_ip4 == NULL) {
2923 mtx_unlock(&pr->pr_mtx);
2924 return (EAFNOSUPPORT);
2927 ia0.s_addr = ntohl(ia->s_addr);
2928 if (ia0.s_addr == INADDR_LOOPBACK) {
2929 ia->s_addr = pr->pr_ip4[0].s_addr;
2930 mtx_unlock(&pr->pr_mtx);
2934 if (ia0.s_addr == INADDR_ANY) {
2936 * In case there is only 1 IPv4 address, bind directly.
2938 if (pr->pr_ip4s == 1)
2939 ia->s_addr = pr->pr_ip4[0].s_addr;
2940 mtx_unlock(&pr->pr_mtx);
2944 error = _prison_check_ip4(pr, ia);
2945 mtx_unlock(&pr->pr_mtx);
2950 * Rewrite destination address in case we will connect to loopback address.
2952 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2953 * Address passed in in NBO and returned in NBO.
2956 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2960 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2961 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2963 pr = cred->cr_prison;
2964 if (!(pr->pr_flags & PR_IP4))
2966 mtx_lock(&pr->pr_mtx);
2967 if (!(pr->pr_flags & PR_IP4)) {
2968 mtx_unlock(&pr->pr_mtx);
2971 if (pr->pr_ip4 == NULL) {
2972 mtx_unlock(&pr->pr_mtx);
2973 return (EAFNOSUPPORT);
2976 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2977 ia->s_addr = pr->pr_ip4[0].s_addr;
2978 mtx_unlock(&pr->pr_mtx);
2983 * Return success because nothing had to be changed.
2985 mtx_unlock(&pr->pr_mtx);
2990 * Check if given address belongs to the jail referenced by cred/prison.
2992 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2993 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2994 * doesn't allow IPv4. Address passed in in NBO.
2997 _prison_check_ip4(const struct prison *pr, const struct in_addr *ia)
3002 * Check the primary IP.
3004 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3008 * All the other IPs are sorted so we can do a binary search.
3011 z = pr->pr_ip4s - 2;
3014 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3023 return (EADDRNOTAVAIL);
3027 prison_check_ip4(const struct ucred *cred, const struct in_addr *ia)
3032 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3033 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3035 pr = cred->cr_prison;
3036 if (!(pr->pr_flags & PR_IP4))
3038 mtx_lock(&pr->pr_mtx);
3039 if (!(pr->pr_flags & PR_IP4)) {
3040 mtx_unlock(&pr->pr_mtx);
3043 if (pr->pr_ip4 == NULL) {
3044 mtx_unlock(&pr->pr_mtx);
3045 return (EAFNOSUPPORT);
3048 error = _prison_check_ip4(pr, ia);
3049 mtx_unlock(&pr->pr_mtx);
3056 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3061 ppr = pr->pr_parent;
3062 if (!(pr->pr_flags & PR_IP6_USER)) {
3063 /* This has no user settings, so just copy the parent's list. */
3064 if (pr->pr_ip6s < ppr->pr_ip6s) {
3066 * There's no room for the parent's list. Use the
3067 * new list buffer, which is assumed to be big enough
3068 * (if it was passed). If there's no buffer, try to
3072 if (newip6 == NULL) {
3073 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3074 M_PRISON, M_NOWAIT);
3078 if (newip6 != NULL) {
3079 bcopy(ppr->pr_ip6, newip6,
3080 ppr->pr_ip6s * sizeof(*newip6));
3081 free(pr->pr_ip6, M_PRISON);
3082 pr->pr_ip6 = newip6;
3083 pr->pr_ip6s = ppr->pr_ip6s;
3087 pr->pr_ip6s = ppr->pr_ip6s;
3088 if (pr->pr_ip6s > 0)
3089 bcopy(ppr->pr_ip6, pr->pr_ip6,
3090 pr->pr_ip6s * sizeof(*newip6));
3091 else if (pr->pr_ip6 != NULL) {
3092 free(pr->pr_ip6, M_PRISON);
3095 } else if (pr->pr_ip6s > 0) {
3096 /* Remove addresses that aren't in the parent. */
3097 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3098 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3101 if (ij < ppr->pr_ip6s)
3104 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3105 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3108 for (ij = 1; ii < pr->pr_ip6s; ) {
3109 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3114 switch (ij >= ppr->pr_ip6s ? -1 :
3115 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3117 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3118 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3129 if (pr->pr_ip6s == 0) {
3130 free(pr->pr_ip6, M_PRISON);
3138 * Pass back primary IPv6 address for this jail.
3140 * If not restricted return success but do not alter the address. Caller has
3141 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3143 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3146 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3150 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3151 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3153 pr = cred->cr_prison;
3154 if (!(pr->pr_flags & PR_IP6))
3156 mtx_lock(&pr->pr_mtx);
3157 if (!(pr->pr_flags & PR_IP6)) {
3158 mtx_unlock(&pr->pr_mtx);
3161 if (pr->pr_ip6 == NULL) {
3162 mtx_unlock(&pr->pr_mtx);
3163 return (EAFNOSUPPORT);
3166 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3167 mtx_unlock(&pr->pr_mtx);
3172 * Return 1 if we should do proper source address selection or are not jailed.
3173 * We will return 0 if we should bypass source address selection in favour
3174 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3176 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3179 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3182 struct in6_addr lia6;
3185 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3186 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3191 pr = cred->cr_prison;
3192 if (pr->pr_flags & PR_IP6_SADDRSEL)
3196 error = prison_get_ip6(cred, &lia6);
3199 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3202 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3207 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3210 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3216 while (pr1 != &prison0 &&
3218 !(pr1->pr_flags & PR_VNET) &&
3220 !(pr1->pr_flags & PR_IP6_USER))
3221 pr1 = pr1->pr_parent;
3222 while (pr2 != &prison0 &&
3224 !(pr2->pr_flags & PR_VNET) &&
3226 !(pr2->pr_flags & PR_IP6_USER))
3227 pr2 = pr2->pr_parent;
3228 return (pr1 == pr2);
3232 * Make sure our (source) address is set to something meaningful to this jail.
3234 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3235 * when needed while binding.
3237 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3238 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3239 * doesn't allow IPv6.
3242 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3247 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3248 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3250 pr = cred->cr_prison;
3251 if (!(pr->pr_flags & PR_IP6))
3253 mtx_lock(&pr->pr_mtx);
3254 if (!(pr->pr_flags & PR_IP6)) {
3255 mtx_unlock(&pr->pr_mtx);
3258 if (pr->pr_ip6 == NULL) {
3259 mtx_unlock(&pr->pr_mtx);
3260 return (EAFNOSUPPORT);
3263 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3264 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3265 mtx_unlock(&pr->pr_mtx);
3269 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3271 * In case there is only 1 IPv6 address, and v6only is true,
3272 * then bind directly.
3274 if (v6only != 0 && pr->pr_ip6s == 1)
3275 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3276 mtx_unlock(&pr->pr_mtx);
3280 error = _prison_check_ip6(pr, ia6);
3281 mtx_unlock(&pr->pr_mtx);
3286 * Rewrite destination address in case we will connect to loopback address.
3288 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3291 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3295 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3296 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3298 pr = cred->cr_prison;
3299 if (!(pr->pr_flags & PR_IP6))
3301 mtx_lock(&pr->pr_mtx);
3302 if (!(pr->pr_flags & PR_IP6)) {
3303 mtx_unlock(&pr->pr_mtx);
3306 if (pr->pr_ip6 == NULL) {
3307 mtx_unlock(&pr->pr_mtx);
3308 return (EAFNOSUPPORT);
3311 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3312 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3313 mtx_unlock(&pr->pr_mtx);
3318 * Return success because nothing had to be changed.
3320 mtx_unlock(&pr->pr_mtx);
3325 * Check if given address belongs to the jail referenced by cred/prison.
3327 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3328 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3329 * doesn't allow IPv6.
3332 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3337 * Check the primary IP.
3339 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3343 * All the other IPs are sorted so we can do a binary search.
3346 z = pr->pr_ip6s - 2;
3349 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3358 return (EADDRNOTAVAIL);
3362 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3367 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3368 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3370 pr = cred->cr_prison;
3371 if (!(pr->pr_flags & PR_IP6))
3373 mtx_lock(&pr->pr_mtx);
3374 if (!(pr->pr_flags & PR_IP6)) {
3375 mtx_unlock(&pr->pr_mtx);
3378 if (pr->pr_ip6 == NULL) {
3379 mtx_unlock(&pr->pr_mtx);
3380 return (EAFNOSUPPORT);
3383 error = _prison_check_ip6(pr, ia6);
3384 mtx_unlock(&pr->pr_mtx);
3390 * Check if a jail supports the given address family.
3392 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3396 prison_check_af(struct ucred *cred, int af)
3401 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3403 pr = cred->cr_prison;
3405 /* Prisons with their own network stack are not limited. */
3406 if (prison_owns_vnet(cred))
3415 if (pr->pr_flags & PR_IP4)
3417 mtx_lock(&pr->pr_mtx);
3418 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3419 error = EAFNOSUPPORT;
3420 mtx_unlock(&pr->pr_mtx);
3426 if (pr->pr_flags & PR_IP6)
3428 mtx_lock(&pr->pr_mtx);
3429 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3430 error = EAFNOSUPPORT;
3431 mtx_unlock(&pr->pr_mtx);
3439 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3440 error = EAFNOSUPPORT;
3446 * Check if given address belongs to the jail referenced by cred (wrapper to
3447 * prison_check_ip[46]).
3449 * Returns 0 if jail doesn't restrict the address family or if address belongs
3450 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3451 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3454 prison_if(struct ucred *cred, struct sockaddr *sa)
3457 struct sockaddr_in *sai;
3460 struct sockaddr_in6 *sai6;
3464 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3465 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3468 if (prison_owns_vnet(cred))
3473 switch (sa->sa_family)
3477 sai = (struct sockaddr_in *)sa;
3478 error = prison_check_ip4(cred, &sai->sin_addr);
3483 sai6 = (struct sockaddr_in6 *)sa;
3484 error = prison_check_ip6(cred, &sai6->sin6_addr);
3488 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3489 error = EAFNOSUPPORT;
3495 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3498 prison_check(struct ucred *cred1, struct ucred *cred2)
3501 return ((cred1->cr_prison == cred2->cr_prison ||
3502 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3506 * Return 1 if p2 is a child of p1, otherwise 0.
3509 prison_ischild(struct prison *pr1, struct prison *pr2)
3512 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3519 * Return 1 if the passed credential is in a jail, otherwise 0.
3522 jailed(struct ucred *cred)
3525 return (cred->cr_prison != &prison0);
3529 * Return 1 if the passed credential is in a jail and that jail does not
3530 * have its own virtual network stack, otherwise 0.
3533 jailed_without_vnet(struct ucred *cred)
3539 if (prison_owns_vnet(cred))
3547 * Return the correct hostname (domainname, et al) for the passed credential.
3550 getcredhostname(struct ucred *cred, char *buf, size_t size)
3555 * A NULL credential can be used to shortcut to the physical
3556 * system's hostname.
3558 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3559 mtx_lock(&pr->pr_mtx);
3560 strlcpy(buf, pr->pr_hostname, size);
3561 mtx_unlock(&pr->pr_mtx);
3565 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3568 mtx_lock(&cred->cr_prison->pr_mtx);
3569 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3570 mtx_unlock(&cred->cr_prison->pr_mtx);
3574 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3577 mtx_lock(&cred->cr_prison->pr_mtx);
3578 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3579 mtx_unlock(&cred->cr_prison->pr_mtx);
3583 getcredhostid(struct ucred *cred, unsigned long *hostid)
3586 mtx_lock(&cred->cr_prison->pr_mtx);
3587 *hostid = cred->cr_prison->pr_hostid;
3588 mtx_unlock(&cred->cr_prison->pr_mtx);
3593 * Determine whether the prison represented by cred owns
3594 * its vnet rather than having it inherited.
3596 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3599 prison_owns_vnet(struct ucred *cred)
3603 * vnets cannot be added/removed after jail creation,
3604 * so no need to lock here.
3606 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3611 * Determine whether the subject represented by cred can "see"
3612 * status of a mount point.
3613 * Returns: 0 for permitted, ENOENT otherwise.
3614 * XXX: This function should be called cr_canseemount() and should be
3615 * placed in kern_prot.c.
3618 prison_canseemount(struct ucred *cred, struct mount *mp)
3624 pr = cred->cr_prison;
3625 if (pr->pr_enforce_statfs == 0)
3627 if (pr->pr_root->v_mount == mp)
3629 if (pr->pr_enforce_statfs == 2)
3632 * If jail's chroot directory is set to "/" we should be able to see
3633 * all mount-points from inside a jail.
3634 * This is ugly check, but this is the only situation when jail's
3635 * directory ends with '/'.
3637 if (strcmp(pr->pr_path, "/") == 0)
3639 len = strlen(pr->pr_path);
3641 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3644 * Be sure that we don't have situation where jail's root directory
3645 * is "/some/path" and mount point is "/some/pathpath".
3647 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3653 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3655 char jpath[MAXPATHLEN];
3659 pr = cred->cr_prison;
3660 if (pr->pr_enforce_statfs == 0)
3662 if (prison_canseemount(cred, mp) != 0) {
3663 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3664 strlcpy(sp->f_mntonname, "[restricted]",
3665 sizeof(sp->f_mntonname));
3668 if (pr->pr_root->v_mount == mp) {
3670 * Clear current buffer data, so we are sure nothing from
3671 * the valid path left there.
3673 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3674 *sp->f_mntonname = '/';
3678 * If jail's chroot directory is set to "/" we should be able to see
3679 * all mount-points from inside a jail.
3681 if (strcmp(pr->pr_path, "/") == 0)
3683 len = strlen(pr->pr_path);
3684 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3686 * Clear current buffer data, so we are sure nothing from
3687 * the valid path left there.
3689 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3690 if (*jpath == '\0') {
3691 /* Should never happen. */
3692 *sp->f_mntonname = '/';
3694 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3699 * Check with permission for a specific privilege is granted within jail. We
3700 * have a specific list of accepted privileges; the rest are denied.
3703 prison_priv_check(struct ucred *cred, int priv)
3711 * Privileges specific to prisons with a virtual network stack.
3712 * There might be a duplicate entry here in case the privilege
3713 * is only granted conditionally in the legacy jail case.
3718 * NFS-specific privileges.
3720 case PRIV_NFS_DAEMON:
3721 case PRIV_NFS_LOCKD:
3724 * Network stack privileges.
3726 case PRIV_NET_BRIDGE:
3729 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3730 case PRIV_NET_ROUTE:
3732 case PRIV_NET_SETIFMTU:
3733 case PRIV_NET_SETIFFLAGS:
3734 case PRIV_NET_SETIFCAP:
3735 case PRIV_NET_SETIFDESCR:
3736 case PRIV_NET_SETIFNAME :
3737 case PRIV_NET_SETIFMETRIC:
3738 case PRIV_NET_SETIFPHYS:
3739 case PRIV_NET_SETIFMAC:
3740 case PRIV_NET_ADDMULTI:
3741 case PRIV_NET_DELMULTI:
3742 case PRIV_NET_HWIOCTL:
3743 case PRIV_NET_SETLLADDR:
3744 case PRIV_NET_ADDIFGROUP:
3745 case PRIV_NET_DELIFGROUP:
3746 case PRIV_NET_IFCREATE:
3747 case PRIV_NET_IFDESTROY:
3748 case PRIV_NET_ADDIFADDR:
3749 case PRIV_NET_DELIFADDR:
3752 case PRIV_NET_SETIFVNET:
3753 case PRIV_NET_SETIFFIB:
3756 * 802.11-related privileges.
3758 case PRIV_NET80211_GETKEY:
3760 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3767 case PRIV_NETATM_CFG:
3768 case PRIV_NETATM_ADD:
3769 case PRIV_NETATM_DEL:
3770 case PRIV_NETATM_SET:
3773 * Bluetooth privileges.
3775 case PRIV_NETBLUETOOTH_RAW:
3779 * Netgraph and netgraph module privileges.
3781 case PRIV_NETGRAPH_CONTROL:
3783 case PRIV_NETGRAPH_TTY:
3787 * IPv4 and IPv6 privileges.
3789 case PRIV_NETINET_IPFW:
3790 case PRIV_NETINET_DIVERT:
3791 case PRIV_NETINET_PF:
3792 case PRIV_NETINET_DUMMYNET:
3793 case PRIV_NETINET_CARP:
3794 case PRIV_NETINET_MROUTE:
3795 case PRIV_NETINET_RAW:
3796 case PRIV_NETINET_ADDRCTRL6:
3797 case PRIV_NETINET_ND6:
3798 case PRIV_NETINET_SCOPE6:
3799 case PRIV_NETINET_ALIFETIME6:
3800 case PRIV_NETINET_IPSEC:
3801 case PRIV_NETINET_BINDANY:
3816 * No default: or deny here.
3817 * In case of no permit fall through to next switch().
3819 if (cred->cr_prison->pr_flags & PR_VNET)
3827 * Allow ktrace privileges for root in jail.
3833 * Allow jailed processes to configure audit identity and
3834 * submit audit records (login, etc). In the future we may
3835 * want to further refine the relationship between audit and
3838 case PRIV_AUDIT_GETAUDIT:
3839 case PRIV_AUDIT_SETAUDIT:
3840 case PRIV_AUDIT_SUBMIT:
3844 * Allow jailed processes to manipulate process UNIX
3845 * credentials in any way they see fit.
3847 case PRIV_CRED_SETUID:
3848 case PRIV_CRED_SETEUID:
3849 case PRIV_CRED_SETGID:
3850 case PRIV_CRED_SETEGID:
3851 case PRIV_CRED_SETGROUPS:
3852 case PRIV_CRED_SETREUID:
3853 case PRIV_CRED_SETREGID:
3854 case PRIV_CRED_SETRESUID:
3855 case PRIV_CRED_SETRESGID:
3858 * Jail implements visibility constraints already, so allow
3859 * jailed root to override uid/gid-based constraints.
3861 case PRIV_SEEOTHERGIDS:
3862 case PRIV_SEEOTHERUIDS:
3865 * Jail implements inter-process debugging limits already, so
3866 * allow jailed root various debugging privileges.
3868 case PRIV_DEBUG_DIFFCRED:
3869 case PRIV_DEBUG_SUGID:
3870 case PRIV_DEBUG_UNPRIV:
3873 * Allow jail to set various resource limits and login
3874 * properties, and for now, exceed process resource limits.
3876 case PRIV_PROC_LIMIT:
3877 case PRIV_PROC_SETLOGIN:
3878 case PRIV_PROC_SETRLIMIT:
3881 * System V and POSIX IPC privileges are granted in jail.
3884 case PRIV_IPC_WRITE:
3885 case PRIV_IPC_ADMIN:
3886 case PRIV_IPC_MSGSIZE:
3890 * Jail operations within a jail work on child jails.
3892 case PRIV_JAIL_ATTACH:
3894 case PRIV_JAIL_REMOVE:
3897 * Jail implements its own inter-process limits, so allow
3898 * root processes in jail to change scheduling on other
3899 * processes in the same jail. Likewise for signalling.
3901 case PRIV_SCHED_DIFFCRED:
3902 case PRIV_SCHED_CPUSET:
3903 case PRIV_SIGNAL_DIFFCRED:
3904 case PRIV_SIGNAL_SUGID:
3907 * Allow jailed processes to write to sysctls marked as jail
3910 case PRIV_SYSCTL_WRITEJAIL:
3913 * Allow root in jail to manage a variety of quota
3914 * properties. These should likely be conditional on a
3915 * configuration option.
3917 case PRIV_VFS_GETQUOTA:
3918 case PRIV_VFS_SETQUOTA:
3921 * Since Jail relies on chroot() to implement file system
3922 * protections, grant many VFS privileges to root in jail.
3923 * Be careful to exclude mount-related and NFS-related
3927 case PRIV_VFS_WRITE:
3928 case PRIV_VFS_ADMIN:
3930 case PRIV_VFS_LOOKUP:
3931 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3932 case PRIV_VFS_CHFLAGS_DEV:
3933 case PRIV_VFS_CHOWN:
3934 case PRIV_VFS_CHROOT:
3935 case PRIV_VFS_RETAINSUGID:
3936 case PRIV_VFS_FCHROOT:
3938 case PRIV_VFS_SETGID:
3940 case PRIV_VFS_STICKYFILE:
3943 * As in the non-jail case, non-root users are expected to be
3944 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3945 * exists in the jail and they have permission to access it).
3947 case PRIV_KMEM_READ:
3951 * Depending on the global setting, allow privilege of
3952 * setting system flags.
3954 case PRIV_VFS_SYSFLAGS:
3955 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3961 * Depending on the global setting, allow privilege of
3962 * mounting/unmounting file systems.
3964 case PRIV_VFS_MOUNT:
3965 case PRIV_VFS_UNMOUNT:
3966 case PRIV_VFS_MOUNT_NONUSER:
3967 case PRIV_VFS_MOUNT_OWNER:
3968 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3969 cred->cr_prison->pr_enforce_statfs < 2)
3975 * Allow jailed root to bind reserved ports and reuse in-use
3978 case PRIV_NETINET_RESERVEDPORT:
3979 case PRIV_NETINET_REUSEPORT:
3983 * Allow jailed root to set certian IPv4/6 (option) headers.
3985 case PRIV_NETINET_SETHDROPTS:
3989 * Conditionally allow creating raw sockets in jail.
3991 case PRIV_NETINET_RAW:
3992 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3998 * Since jail implements its own visibility limits on netstat
3999 * sysctls, allow getcred. This allows identd to work in
4002 case PRIV_NETINET_GETCRED:
4006 * Allow jailed root to set loginclass.
4008 case PRIV_PROC_SETLOGINCLASS:
4013 * In all remaining cases, deny the privilege request. This
4014 * includes almost all network privileges, many system
4015 * configuration privileges.
4022 * Return the part of pr2's name that is relative to pr1, or the whole name
4023 * if it does not directly follow.
4027 prison_name(struct prison *pr1, struct prison *pr2)
4031 /* Jails see themselves as "0" (if they see themselves at all). */
4034 name = pr2->pr_name;
4035 if (prison_ischild(pr1, pr2)) {
4037 * pr1 isn't locked (and allprison_lock may not be either)
4038 * so its length can't be counted on. But the number of dots
4039 * can be counted on - and counted.
4041 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4042 name = strchr(name, '.') + 1;
4048 * Return the part of pr2's path that is relative to pr1, or the whole path
4049 * if it does not directly follow.
4052 prison_path(struct prison *pr1, struct prison *pr2)
4054 char *path1, *path2;
4057 path1 = pr1->pr_path;
4058 path2 = pr2->pr_path;
4059 if (!strcmp(path1, "/"))
4061 len1 = strlen(path1);
4062 if (strncmp(path1, path2, len1))
4064 if (path2[len1] == '\0')
4066 if (path2[len1] == '/')
4067 return (path2 + len1);
4073 * Jail-related sysctls.
4075 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4079 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4082 struct prison *pr, *cpr;
4084 struct in_addr *ip4 = NULL;
4088 struct in6_addr *ip6 = NULL;
4093 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4094 pr = req->td->td_ucred->cr_prison;
4096 sx_slock(&allprison_lock);
4097 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4098 #if defined(INET) || defined(INET6)
4101 mtx_lock(&cpr->pr_mtx);
4103 if (cpr->pr_ip4s > 0) {
4104 if (ip4s < cpr->pr_ip4s) {
4105 ip4s = cpr->pr_ip4s;
4106 mtx_unlock(&cpr->pr_mtx);
4107 ip4 = realloc(ip4, ip4s *
4108 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4111 bcopy(cpr->pr_ip4, ip4,
4112 cpr->pr_ip4s * sizeof(struct in_addr));
4116 if (cpr->pr_ip6s > 0) {
4117 if (ip6s < cpr->pr_ip6s) {
4118 ip6s = cpr->pr_ip6s;
4119 mtx_unlock(&cpr->pr_mtx);
4120 ip6 = realloc(ip6, ip6s *
4121 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4124 bcopy(cpr->pr_ip6, ip6,
4125 cpr->pr_ip6s * sizeof(struct in6_addr));
4128 if (cpr->pr_ref == 0) {
4129 mtx_unlock(&cpr->pr_mtx);
4132 bzero(xp, sizeof(*xp));
4133 xp->pr_version = XPRISON_VERSION;
4134 xp->pr_id = cpr->pr_id;
4135 xp->pr_state = cpr->pr_uref > 0
4136 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4137 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4138 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4139 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4141 xp->pr_ip4s = cpr->pr_ip4s;
4144 xp->pr_ip6s = cpr->pr_ip6s;
4146 mtx_unlock(&cpr->pr_mtx);
4147 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4151 if (xp->pr_ip4s > 0) {
4152 error = SYSCTL_OUT(req, ip4,
4153 xp->pr_ip4s * sizeof(struct in_addr));
4159 if (xp->pr_ip6s > 0) {
4160 error = SYSCTL_OUT(req, ip6,
4161 xp->pr_ip6s * sizeof(struct in6_addr));
4167 sx_sunlock(&allprison_lock);
4178 SYSCTL_OID(_security_jail, OID_AUTO, list,
4179 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4180 sysctl_jail_list, "S", "List of active jails");
4183 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4187 injail = jailed(req->td->td_ucred);
4188 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4193 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4194 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4195 sysctl_jail_jailed, "I", "Process in jail?");
4198 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4200 int error, havevnet;
4202 struct ucred *cred = req->td->td_ucred;
4204 havevnet = jailed(cred) && prison_owns_vnet(cred);
4208 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4213 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4214 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4215 sysctl_jail_vnet, "I", "Jail owns VNET?");
4217 #if defined(INET) || defined(INET6)
4218 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4219 &jail_max_af_ips, 0,
4220 "Number of IP addresses a jail may have at most per address family");
4224 * Default parameters for jail(2) compatability. For historical reasons,
4225 * the sysctl names have varying similarity to the parameter names. Prisons
4226 * just see their own parameters, and can't change them.
4229 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4232 int allow, error, i;
4234 pr = req->td->td_ucred->cr_prison;
4235 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4237 /* Get the current flag value, and convert it to a boolean. */
4238 i = (allow & arg2) ? 1 : 0;
4241 error = sysctl_handle_int(oidp, &i, 0, req);
4242 if (error || !req->newptr)
4248 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4251 mtx_lock(&prison0.pr_mtx);
4252 jail_default_allow = (jail_default_allow & ~arg2) | i;
4253 mtx_unlock(&prison0.pr_mtx);
4257 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4258 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4259 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4260 "Processes in jail can set their hostnames");
4261 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4262 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4263 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4264 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4265 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4266 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4267 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4268 "Processes in jail can use System V IPC primitives");
4269 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4270 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4271 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4272 "Prison root can create raw sockets");
4273 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4274 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4275 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4276 "Processes in jail can alter system file flags");
4277 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4278 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4279 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4280 "Processes in jail can mount/unmount jail-friendly file systems");
4281 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4282 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4283 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4284 "Processes in jail can mount the devfs file system");
4285 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4286 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4287 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4288 "Processes in jail can mount the fdescfs file system");
4289 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4290 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4291 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4292 "Processes in jail can mount the nullfs file system");
4293 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4294 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4295 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4296 "Processes in jail can mount the procfs file system");
4297 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
4298 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4299 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
4300 "Processes in jail can mount the linprocfs file system");
4301 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
4302 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4303 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
4304 "Processes in jail can mount the linsysfs file system");
4305 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4306 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4307 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4308 "Processes in jail can mount the tmpfs file system");
4309 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4310 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4311 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4312 "Processes in jail can mount the zfs file system");
4315 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4320 pr = req->td->td_ucred->cr_prison;
4321 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4322 error = sysctl_handle_int(oidp, &level, 0, req);
4323 if (error || !req->newptr)
4325 *(int *)arg1 = level;
4329 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4330 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4331 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4332 sysctl_jail_default_level, "I",
4333 "Processes in jail cannot see all mounted file systems");
4335 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4336 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4337 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4338 sysctl_jail_default_level, "I",
4339 "Ruleset for the devfs filesystem in jail");
4342 * Nodes to describe jail parameters. Maximum length of string parameters
4343 * is returned in the string itself, and the other parameters exist merely
4344 * to make themselves and their types known.
4346 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4350 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4357 switch (oidp->oid_kind & CTLTYPE)
4363 if (!(req->flags & SCTL_MASK32))
4365 return (SYSCTL_OUT(req, &l, sizeof(l)));
4369 return (SYSCTL_OUT(req, &i, sizeof(i)));
4370 case CTLTYPE_STRING:
4371 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4373 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4374 case CTLTYPE_STRUCT:
4376 return (SYSCTL_OUT(req, &s, sizeof(s)));
4382 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4383 * jail creation time but cannot be changed in an existing jail.
4385 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4386 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4387 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4388 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4389 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4390 "I", "Jail secure level");
4391 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4392 "Jail value for kern.osreldate and uname -K");
4393 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4394 "Jail value for kern.osrelease and uname -r");
4395 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4396 "I", "Jail cannot see all mounted file systems");
4397 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4398 "I", "Ruleset for in-jail devfs mounts");
4399 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4400 "B", "Jail persistence");
4402 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4403 "E,jailsys", "Virtual network stack");
4405 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4406 "B", "Jail is in the process of shutting down");
4408 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4409 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4410 "I", "Current number of child jails");
4411 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4412 "I", "Maximum number of child jails");
4414 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4415 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4417 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4418 "Jail NIS domainname");
4419 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4421 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4422 "LU", "Jail host ID");
4424 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4425 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4428 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4429 "Jail IPv4 address virtualization");
4430 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4431 "S,in_addr,a", "Jail IPv4 addresses");
4432 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4433 "B", "Do (not) use IPv4 source address selection rather than the "
4434 "primary jail IPv4 address.");
4437 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4438 "Jail IPv6 address virtualization");
4439 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4440 "S,in6_addr,a", "Jail IPv6 addresses");
4441 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4442 "B", "Do (not) use IPv6 source address selection rather than the "
4443 "primary jail IPv6 address.");
4446 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4447 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4448 "B", "Jail may set hostname");
4449 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4450 "B", "Jail may use SYSV IPC");
4451 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4452 "B", "Jail may create raw sockets");
4453 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4454 "B", "Jail may alter system file flags");
4455 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4456 "B", "Jail may set file quotas");
4457 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4458 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4460 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4461 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4462 "B", "Jail may mount/unmount jail-friendly file systems in general");
4463 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4464 "B", "Jail may mount the devfs file system");
4465 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4466 "B", "Jail may mount the fdescfs file system");
4467 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4468 "B", "Jail may mount the nullfs file system");
4469 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4470 "B", "Jail may mount the procfs file system");
4471 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
4472 "B", "Jail may mount the linprocfs file system");
4473 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
4474 "B", "Jail may mount the linsysfs file system");
4475 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4476 "B", "Jail may mount the tmpfs file system");
4477 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4478 "B", "Jail may mount the zfs file system");
4482 prison_racct_foreach(void (*callback)(struct racct *racct,
4483 void *arg2, void *arg3), void *arg2, void *arg3)
4485 struct prison_racct *prr;
4487 ASSERT_RACCT_ENABLED();
4489 sx_slock(&allprison_lock);
4490 LIST_FOREACH(prr, &allprison_racct, prr_next)
4491 (callback)(prr->prr_racct, arg2, arg3);
4492 sx_sunlock(&allprison_lock);
4495 static struct prison_racct *
4496 prison_racct_find_locked(const char *name)
4498 struct prison_racct *prr;
4500 ASSERT_RACCT_ENABLED();
4501 sx_assert(&allprison_lock, SA_XLOCKED);
4503 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4506 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4507 if (strcmp(name, prr->prr_name) != 0)
4510 /* Found prison_racct with a matching name? */
4511 prison_racct_hold(prr);
4515 /* Add new prison_racct. */
4516 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4517 racct_create(&prr->prr_racct);
4519 strcpy(prr->prr_name, name);
4520 refcount_init(&prr->prr_refcount, 1);
4521 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4526 struct prison_racct *
4527 prison_racct_find(const char *name)
4529 struct prison_racct *prr;
4531 ASSERT_RACCT_ENABLED();
4533 sx_xlock(&allprison_lock);
4534 prr = prison_racct_find_locked(name);
4535 sx_xunlock(&allprison_lock);
4540 prison_racct_hold(struct prison_racct *prr)
4543 ASSERT_RACCT_ENABLED();
4545 refcount_acquire(&prr->prr_refcount);
4549 prison_racct_free_locked(struct prison_racct *prr)
4552 ASSERT_RACCT_ENABLED();
4553 sx_assert(&allprison_lock, SA_XLOCKED);
4555 if (refcount_release(&prr->prr_refcount)) {
4556 racct_destroy(&prr->prr_racct);
4557 LIST_REMOVE(prr, prr_next);
4558 free(prr, M_PRISON_RACCT);
4563 prison_racct_free(struct prison_racct *prr)
4567 ASSERT_RACCT_ENABLED();
4568 sx_assert(&allprison_lock, SA_UNLOCKED);
4570 old = prr->prr_refcount;
4571 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4574 sx_xlock(&allprison_lock);
4575 prison_racct_free_locked(prr);
4576 sx_xunlock(&allprison_lock);
4580 prison_racct_attach(struct prison *pr)
4582 struct prison_racct *prr;
4584 ASSERT_RACCT_ENABLED();
4585 sx_assert(&allprison_lock, SA_XLOCKED);
4587 prr = prison_racct_find_locked(pr->pr_name);
4588 KASSERT(prr != NULL, ("cannot find prison_racct"));
4590 pr->pr_prison_racct = prr;
4594 * Handle jail renaming. From the racct point of view, renaming means
4595 * moving from one prison_racct to another.
4598 prison_racct_modify(struct prison *pr)
4602 struct prison_racct *oldprr;
4604 ASSERT_RACCT_ENABLED();
4606 sx_slock(&allproc_lock);
4607 sx_xlock(&allprison_lock);
4609 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4610 sx_xunlock(&allprison_lock);
4611 sx_sunlock(&allproc_lock);
4615 oldprr = pr->pr_prison_racct;
4616 pr->pr_prison_racct = NULL;
4618 prison_racct_attach(pr);
4621 * Move resource utilisation records.
4623 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4626 * Force rctl to reattach rules to processes.
4628 FOREACH_PROC_IN_SYSTEM(p) {
4630 cred = crhold(p->p_ucred);
4632 racct_proc_ucred_changed(p, cred, cred);
4636 sx_sunlock(&allproc_lock);
4637 prison_racct_free_locked(oldprr);
4638 sx_xunlock(&allprison_lock);
4642 prison_racct_detach(struct prison *pr)
4645 ASSERT_RACCT_ENABLED();
4646 sx_assert(&allprison_lock, SA_UNLOCKED);
4648 if (pr->pr_prison_racct == NULL)
4650 prison_racct_free(pr->pr_prison_racct);
4651 pr->pr_prison_racct = NULL;
4658 db_show_prison(struct prison *pr)
4661 #if defined(INET) || defined(INET6)
4666 char ip6buf[INET6_ADDRSTRLEN];
4669 db_printf("prison %p:\n", pr);
4670 db_printf(" jid = %d\n", pr->pr_id);
4671 db_printf(" name = %s\n", pr->pr_name);
4672 db_printf(" parent = %p\n", pr->pr_parent);
4673 db_printf(" ref = %d\n", pr->pr_ref);
4674 db_printf(" uref = %d\n", pr->pr_uref);
4675 db_printf(" path = %s\n", pr->pr_path);
4676 db_printf(" cpuset = %d\n", pr->pr_cpuset
4677 ? pr->pr_cpuset->cs_id : -1);
4679 db_printf(" vnet = %p\n", pr->pr_vnet);
4681 db_printf(" root = %p\n", pr->pr_root);
4682 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4683 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4684 db_printf(" children.max = %d\n", pr->pr_childmax);
4685 db_printf(" children.cur = %d\n", pr->pr_childcount);
4686 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4687 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4688 db_printf(" flags = 0x%x", pr->pr_flags);
4689 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4691 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4692 db_printf(" %s", pr_flag_names[fi]);
4693 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4695 jsf = pr->pr_flags &
4696 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4697 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4698 pr_flag_jailsys[fi].disable &&
4699 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4700 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4703 db_printf(" allow = 0x%x", pr->pr_allow);
4704 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4706 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4707 db_printf(" %s", pr_allow_names[fi]);
4709 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4710 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4711 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4712 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4713 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4715 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4716 for (ii = 0; ii < pr->pr_ip4s; ii++)
4717 db_printf(" %s %s\n",
4718 ii == 0 ? "ip4.addr =" : " ",
4719 inet_ntoa(pr->pr_ip4[ii]));
4722 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4723 for (ii = 0; ii < pr->pr_ip6s; ii++)
4724 db_printf(" %s %s\n",
4725 ii == 0 ? "ip6.addr =" : " ",
4726 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4730 DB_SHOW_COMMAND(prison, db_show_prison_command)
4736 * Show all prisons in the list, and prison0 which is not
4739 db_show_prison(&prison0);
4740 if (!db_pager_quit) {
4741 TAILQ_FOREACH(pr, &allprison, pr_list) {
4753 /* Look for a prison with the ID and with references. */
4754 TAILQ_FOREACH(pr, &allprison, pr_list)
4755 if (pr->pr_id == addr && pr->pr_ref > 0)
4758 /* Look again, without requiring a reference. */
4759 TAILQ_FOREACH(pr, &allprison, pr_list)
4760 if (pr->pr_id == addr)
4763 /* Assume address points to a valid prison. */
4764 pr = (struct prison *)addr;