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",
209 const size_t pr_allow_names_size = sizeof(pr_allow_names);
211 static char *pr_allow_nonames[] = {
212 "allow.noset_hostname",
214 "allow.noraw_sockets",
219 "allow.mount.nodevfs",
220 "allow.mount.nonullfs",
222 "allow.mount.noprocfs",
223 "allow.mount.notmpfs",
224 "allow.mount.nofdescfs",
226 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
228 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
229 #define JAIL_DEFAULT_ENFORCE_STATFS 2
230 #define JAIL_DEFAULT_DEVFS_RSNUM 0
231 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
232 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
233 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
234 #if defined(INET) || defined(INET6)
235 static unsigned jail_max_af_ips = 255;
239 * Initialize the parts of prison0 that can't be static-initialized with
240 * constants. This is called from proc0_init() after creating thread0 cpuset.
246 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
247 prison0.pr_osreldate = osreldate;
248 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
253 qcmp_v4(const void *ip1, const void *ip2)
258 * We need to compare in HBO here to get the list sorted as expected
259 * by the result of the code. Sorting NBO addresses gives you
260 * interesting results. If you do not understand, do not try.
262 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
263 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
266 * Do not simply return the difference of the two numbers, the int is
280 qcmp_v6(const void *ip1, const void *ip2)
282 const struct in6_addr *ia6a, *ia6b;
285 ia6a = (const struct in6_addr *)ip1;
286 ia6b = (const struct in6_addr *)ip2;
289 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
290 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
292 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
305 sys_jail(struct thread *td, struct jail_args *uap)
311 error = copyin(uap->jail, &version, sizeof(uint32_t));
320 /* FreeBSD single IPv4 jails. */
321 bzero(&j, sizeof(struct jail));
322 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
325 j.version = j0.version;
327 j.hostname = j0.hostname;
328 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
334 * Version 1 was used by multi-IPv4 jail implementations
335 * that never made it into the official kernel.
339 case 2: /* JAIL_API_VERSION */
340 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
341 error = copyin(uap->jail, &j, sizeof(struct jail));
347 /* Sci-Fi jails are not supported, sorry. */
350 return (kern_jail(td, &j));
354 kern_jail(struct thread *td, struct jail *j)
356 struct iovec optiov[2 * (4
357 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
366 char *u_path, *u_hostname, *u_name;
369 struct in_addr *u_ip4;
372 struct in6_addr *u_ip6;
375 int error, enforce_statfs, fi;
377 bzero(&optiov, sizeof(optiov));
378 opt.uio_iov = optiov;
382 opt.uio_segflg = UIO_SYSSPACE;
383 opt.uio_rw = UIO_READ;
386 /* Set permissions for top-level jails from sysctls. */
387 if (!jailed(td->td_ucred)) {
388 for (fi = 0; fi < sizeof(pr_allow_names) /
389 sizeof(pr_allow_names[0]); fi++) {
390 optiov[opt.uio_iovcnt].iov_base =
391 (jail_default_allow & (1 << fi))
392 ? pr_allow_names[fi] : pr_allow_nonames[fi];
393 optiov[opt.uio_iovcnt].iov_len =
394 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
397 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
398 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
400 enforce_statfs = jail_default_enforce_statfs;
401 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
402 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
406 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
408 ip4s = (j->version == 0) ? 1 : j->ip4s;
409 if (ip4s > jail_max_af_ips)
411 tmplen += ip4s * sizeof(struct in_addr);
417 if (j->ip6s > jail_max_af_ips)
419 tmplen += j->ip6s * sizeof(struct in6_addr);
424 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
425 u_hostname = u_path + MAXPATHLEN;
426 u_name = u_hostname + MAXHOSTNAMELEN;
428 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
432 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
434 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
437 optiov[opt.uio_iovcnt].iov_base = "path";
438 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
440 optiov[opt.uio_iovcnt].iov_base = u_path;
441 error = copyinstr(j->path, u_path, MAXPATHLEN,
442 &optiov[opt.uio_iovcnt].iov_len);
444 free(u_path, M_TEMP);
448 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
449 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
451 optiov[opt.uio_iovcnt].iov_base = u_hostname;
452 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
453 &optiov[opt.uio_iovcnt].iov_len);
455 free(u_path, M_TEMP);
459 if (j->jailname != NULL) {
460 optiov[opt.uio_iovcnt].iov_base = "name";
461 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
463 optiov[opt.uio_iovcnt].iov_base = u_name;
464 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
465 &optiov[opt.uio_iovcnt].iov_len);
467 free(u_path, M_TEMP);
473 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
474 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
476 optiov[opt.uio_iovcnt].iov_base = u_ip4;
477 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
479 u_ip4->s_addr = j->ip4s;
481 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
483 free(u_path, M_TEMP);
490 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
491 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
493 optiov[opt.uio_iovcnt].iov_base = u_ip6;
494 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
495 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
497 free(u_path, M_TEMP);
502 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
503 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
504 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
505 free(u_path, M_TEMP);
511 * struct jail_set_args {
512 * struct iovec *iovp;
513 * unsigned int iovcnt;
518 sys_jail_set(struct thread *td, struct jail_set_args *uap)
523 /* Check that we have an even number of iovecs. */
527 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
530 error = kern_jail_set(td, auio, uap->flags);
536 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
543 struct in6_addr *ip6;
546 struct vfsoptlist *opts;
547 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
549 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
550 char *g_path, *osrelstr;
551 #if defined(INET) || defined(INET6)
556 size_t namelen, onamelen;
557 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
558 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
559 int fi, jid, jsys, len, level;
560 int childmax, osreldt, rsnum, slevel;
561 int fullpath_disabled;
562 #if defined(INET) || defined(INET6)
571 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
575 error = priv_check(td, PRIV_JAIL_SET);
576 if (!error && (flags & JAIL_ATTACH))
577 error = priv_check(td, PRIV_JAIL_ATTACH);
580 mypr = ppr = td->td_ucred->cr_prison;
581 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
583 if (flags & ~JAIL_SET_MASK)
587 * Check all the parameters before committing to anything. Not all
588 * errors can be caught early, but we may as well try. Also, this
589 * takes care of some expensive stuff (path lookup) before getting
590 * the allprison lock.
592 * XXX Jails are not filesystems, and jail parameters are not mount
593 * options. But it makes more sense to re-use the vfsopt code
594 * than duplicate it under a different name.
596 error = vfs_buildopts(optuio, &opts);
607 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
613 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
622 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
630 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
635 else if (enforce < 0 || enforce > 2) {
641 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
649 pr_flags = ch_flags = 0;
650 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
652 if (pr_flag_names[fi] == NULL)
654 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
655 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
657 ch_flags |= pr_flags;
658 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
660 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
667 case JAIL_SYS_DISABLE:
668 if (!pr_flag_jailsys[fi].disable) {
672 pr_flags |= pr_flag_jailsys[fi].disable;
675 pr_flags |= pr_flag_jailsys[fi].new;
677 case JAIL_SYS_INHERIT:
684 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
686 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
687 && !(pr_flags & PR_PERSIST)) {
689 vfs_opterror(opts, "new jail must persist or attach");
693 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
695 vfs_opterror(opts, "vnet cannot be changed after creation");
700 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
702 vfs_opterror(opts, "ip4 cannot be changed after creation");
707 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
709 vfs_opterror(opts, "ip6 cannot be changed after creation");
714 pr_allow = ch_allow = 0;
715 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
717 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
718 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
720 ch_allow |= pr_allow;
722 error = vfs_getopt(opts, "name", (void **)&name, &len);
728 if (len == 0 || name[len - 1] != '\0') {
732 if (len > MAXHOSTNAMELEN) {
733 error = ENAMETOOLONG;
738 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
746 if (len == 0 || host[len - 1] != '\0') {
750 if (len > MAXHOSTNAMELEN) {
751 error = ENAMETOOLONG;
756 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
764 if (len == 0 || domain[len - 1] != '\0') {
768 if (len > MAXHOSTNAMELEN) {
769 error = ENAMETOOLONG;
774 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
782 if (len == 0 || uuid[len - 1] != '\0') {
786 if (len > HOSTUUIDLEN) {
787 error = ENAMETOOLONG;
792 #ifdef COMPAT_FREEBSD32
793 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
796 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
800 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
812 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
817 else if (ip4s & (sizeof(*ip4) - 1)) {
821 ch_flags |= PR_IP4_USER;
822 pr_flags |= PR_IP4_USER;
824 ip4s /= sizeof(*ip4);
825 if (ip4s > jail_max_af_ips) {
827 vfs_opterror(opts, "too many IPv4 addresses");
830 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
831 bcopy(op, ip4, ip4s * sizeof(*ip4));
833 * IP addresses are all sorted but ip[0] to preserve
834 * the primary IP address as given from userland.
835 * This special IP is used for unbound outgoing
836 * connections as well for "loopback" traffic in case
837 * source address selection cannot find any more fitting
838 * address to connect from.
841 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
843 * Check for duplicate addresses and do some simple
844 * zero and broadcast checks. If users give other bogus
845 * addresses it is their problem.
847 * We do not have to care about byte order for these
848 * checks so we will do them in NBO.
850 for (ii = 0; ii < ip4s; ii++) {
851 if (ip4[ii].s_addr == INADDR_ANY ||
852 ip4[ii].s_addr == INADDR_BROADCAST) {
857 (ip4[0].s_addr == ip4[ii+1].s_addr ||
858 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
868 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
873 else if (ip6s & (sizeof(*ip6) - 1)) {
877 ch_flags |= PR_IP6_USER;
878 pr_flags |= PR_IP6_USER;
880 ip6s /= sizeof(*ip6);
881 if (ip6s > jail_max_af_ips) {
883 vfs_opterror(opts, "too many IPv6 addresses");
886 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
887 bcopy(op, ip6, ip6s * sizeof(*ip6));
889 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
890 for (ii = 0; ii < ip6s; ii++) {
891 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
896 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
897 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
907 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
908 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
911 "vnet jails cannot have IP address restrictions");
916 fullpath_disabled = 0;
918 error = vfs_getopt(opts, "path", (void **)&path, &len);
924 if (flags & JAIL_UPDATE) {
927 "path cannot be changed after creation");
930 if (len == 0 || path[len - 1] != '\0') {
934 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
940 NDFREE(&nd, NDF_ONLY_PNBUF);
941 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
942 strlcpy(g_path, path, MAXPATHLEN);
943 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
946 else if (error == ENODEV) {
947 /* proceed if sysctl debug.disablefullpath == 1 */
948 fullpath_disabled = 1;
949 if (len < 2 || (len == 2 && path[0] == '/'))
952 /* exit on other errors */
955 if (root->v_type != VDIR) {
961 if (fullpath_disabled) {
962 /* Leave room for a real-root full pathname. */
963 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
964 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
965 error = ENAMETOOLONG;
971 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
977 if (flags & JAIL_UPDATE) {
980 "osrelease cannot be changed after creation");
983 if (len == 0 || len >= OSRELEASELEN) {
986 "osrelease string must be 1-%d bytes long",
992 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
998 if (flags & JAIL_UPDATE) {
1001 "osreldate cannot be changed after creation");
1006 vfs_opterror(opts, "osreldate cannot be 0");
1012 * Grab the allprison lock before letting modules check their
1013 * parameters. Once we have it, do not let go so we'll have a
1014 * consistent view of the OSD list.
1016 sx_xlock(&allprison_lock);
1017 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
1019 goto done_unlock_list;
1021 /* By now, all parameters should have been noted. */
1022 TAILQ_FOREACH(opt, opts, link) {
1023 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1025 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1026 goto done_unlock_list;
1031 * See if we are creating a new record or updating an existing one.
1032 * This abuses the file error codes ENOENT and EEXIST.
1034 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
1037 vfs_opterror(opts, "no valid operation (create or update)");
1038 goto done_unlock_list;
1042 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1043 namelc = strrchr(name, '.');
1044 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1050 * See if a requested jid already exists. There is an
1051 * information leak here if the jid exists but is not within
1052 * the caller's jail hierarchy. Jail creators will get EEXIST
1053 * even though they cannot see the jail, and CREATE | UPDATE
1054 * will return ENOENT which is not normally a valid error.
1058 vfs_opterror(opts, "negative jid");
1059 goto done_unlock_list;
1061 pr = prison_find(jid);
1063 ppr = pr->pr_parent;
1064 /* Create: jid must not exist. */
1065 if (cuflags == JAIL_CREATE) {
1066 mtx_unlock(&pr->pr_mtx);
1068 vfs_opterror(opts, "jail %d already exists",
1070 goto done_unlock_list;
1072 if (!prison_ischild(mypr, pr)) {
1073 mtx_unlock(&pr->pr_mtx);
1075 } else if (pr->pr_uref == 0) {
1076 if (!(flags & JAIL_DYING)) {
1077 mtx_unlock(&pr->pr_mtx);
1079 vfs_opterror(opts, "jail %d is dying",
1081 goto done_unlock_list;
1082 } else if ((flags & JAIL_ATTACH) ||
1083 (pr_flags & PR_PERSIST)) {
1085 * A dying jail might be resurrected
1086 * (via attach or persist), but first
1087 * it must determine if another jail
1088 * has claimed its name. Accomplish
1089 * this by implicitly re-setting the
1093 name = prison_name(mypr, pr);
1098 /* Update: jid must exist. */
1099 if (cuflags == JAIL_UPDATE) {
1101 vfs_opterror(opts, "jail %d not found", jid);
1102 goto done_unlock_list;
1107 * If the caller provided a name, look for a jail by that name.
1108 * This has different semantics for creates and updates keyed by jid
1109 * (where the name must not already exist in a different jail),
1110 * and updates keyed by the name itself (where the name must exist
1111 * because that is the jail being updated).
1114 namelc = strrchr(name, '.');
1119 * This is a hierarchical name. Split it into the
1120 * parent and child names, and make sure the parent
1121 * exists or matches an already found jail.
1125 if (strncmp(name, ppr->pr_name, namelc - name)
1126 || ppr->pr_name[namelc - name] != '\0') {
1127 mtx_unlock(&pr->pr_mtx);
1130 "cannot change jail's parent");
1131 goto done_unlock_list;
1134 ppr = prison_find_name(mypr, name);
1138 "jail \"%s\" not found", name);
1139 goto done_unlock_list;
1141 mtx_unlock(&ppr->pr_mtx);
1145 if (name[0] != '\0') {
1147 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1150 FOREACH_PRISON_CHILD(ppr, tpr) {
1151 if (tpr != pr && tpr->pr_ref > 0 &&
1152 !strcmp(tpr->pr_name + namelen, name)) {
1154 cuflags != JAIL_CREATE) {
1155 mtx_lock(&tpr->pr_mtx);
1156 if (tpr->pr_ref > 0) {
1161 if (tpr->pr_uref > 0) {
1167 mtx_unlock(&tpr->pr_mtx);
1168 } else if (tpr->pr_uref > 0) {
1170 * Create, or update(jid):
1171 * name must not exist in an
1172 * active sibling jail.
1176 mtx_unlock(&pr->pr_mtx);
1178 "jail \"%s\" already exists",
1180 goto done_unlock_list;
1184 /* If no active jail is found, use a dying one. */
1185 if (deadpr != NULL && pr == NULL) {
1186 if (flags & JAIL_DYING) {
1187 mtx_lock(&deadpr->pr_mtx);
1188 if (deadpr->pr_ref == 0) {
1189 mtx_unlock(&deadpr->pr_mtx);
1193 } else if (cuflags == JAIL_UPDATE) {
1196 "jail \"%s\" is dying", name);
1197 goto done_unlock_list;
1200 /* Update: name must exist if no jid. */
1201 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1203 vfs_opterror(opts, "jail \"%s\" not found",
1205 goto done_unlock_list;
1209 /* Update: must provide a jid or name. */
1210 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1212 vfs_opterror(opts, "update specified no jail");
1213 goto done_unlock_list;
1216 /* If there's no prison to update, create a new one and link it in. */
1218 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1219 if (tpr->pr_childcount >= tpr->pr_childmax) {
1221 vfs_opterror(opts, "prison limit exceeded");
1222 goto done_unlock_list;
1225 mtx_lock(&ppr->pr_mtx);
1226 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1227 mtx_unlock(&ppr->pr_mtx);
1229 vfs_opterror(opts, "parent jail went away!");
1230 goto done_unlock_list;
1234 mtx_unlock(&ppr->pr_mtx);
1235 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1237 /* Find the next free jid. */
1240 if (jid == JAIL_MAX)
1242 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1243 if (tpr->pr_id < jid)
1245 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1246 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1249 if (jid == lastprid) {
1252 "no available jail IDs");
1254 prison_deref(ppr, PD_DEREF |
1255 PD_DEUREF | PD_LIST_XLOCKED);
1264 * The jail already has a jid (that did not yet exist),
1265 * so just find where to insert it.
1267 TAILQ_FOREACH(tpr, &allprison, pr_list)
1268 if (tpr->pr_id >= jid) {
1269 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1274 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1275 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1276 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1277 tpr->pr_childcount++;
1279 pr->pr_parent = ppr;
1282 /* Set some default values, and inherit some from the parent. */
1287 root = mypr->pr_root;
1290 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1291 pr->pr_flags |= PR_HOST;
1292 #if defined(INET) || defined(INET6)
1294 if (!(pr_flags & PR_VNET))
1298 if (!(ch_flags & PR_IP4_USER))
1299 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1300 else if (!(pr_flags & PR_IP4_USER)) {
1301 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1302 if (ppr->pr_ip4 != NULL) {
1303 pr->pr_ip4s = ppr->pr_ip4s;
1304 pr->pr_ip4 = malloc(pr->pr_ip4s *
1305 sizeof(struct in_addr), M_PRISON,
1307 bcopy(ppr->pr_ip4, pr->pr_ip4,
1308 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1313 if (!(ch_flags & PR_IP6_USER))
1314 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1315 else if (!(pr_flags & PR_IP6_USER)) {
1316 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1317 if (ppr->pr_ip6 != NULL) {
1318 pr->pr_ip6s = ppr->pr_ip6s;
1319 pr->pr_ip6 = malloc(pr->pr_ip6s *
1320 sizeof(struct in6_addr), M_PRISON,
1322 bcopy(ppr->pr_ip6, pr->pr_ip6,
1323 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1329 /* Source address selection is always on by default. */
1330 pr->pr_flags |= _PR_IP_SADDRSEL;
1332 pr->pr_securelevel = ppr->pr_securelevel;
1333 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1334 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1335 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1337 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1338 if (osrelstr == NULL)
1339 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1341 strcpy(pr->pr_osrelease, osrelstr);
1343 LIST_INIT(&pr->pr_children);
1344 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1347 /* Allocate a new vnet if specified. */
1348 pr->pr_vnet = (pr_flags & PR_VNET)
1349 ? vnet_alloc() : ppr->pr_vnet;
1352 * Allocate a dedicated cpuset for each jail.
1353 * Unlike other initial settings, this may return an erorr.
1355 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1357 prison_deref(pr, PD_LIST_XLOCKED);
1361 mtx_lock(&pr->pr_mtx);
1363 * New prisons do not yet have a reference, because we do not
1364 * want other to see the incomplete prison once the
1365 * allprison_lock is downgraded.
1370 * Grab a reference for existing prisons, to ensure they
1371 * continue to exist for the duration of the call.
1374 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1375 if ((pr->pr_flags & PR_VNET) &&
1376 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1379 "vnet jails cannot have IP address restrictions");
1380 goto done_deref_locked;
1384 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1387 "ip4 cannot be changed after creation");
1388 goto done_deref_locked;
1392 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1395 "ip6 cannot be changed after creation");
1396 goto done_deref_locked;
1401 /* Do final error checking before setting anything. */
1403 if (slevel < ppr->pr_securelevel) {
1405 goto done_deref_locked;
1409 if (childmax >= ppr->pr_childmax) {
1411 goto done_deref_locked;
1415 if (enforce < ppr->pr_enforce_statfs) {
1417 goto done_deref_locked;
1422 * devfs_rsnum is a uint16_t
1424 if (rsnum < 0 || rsnum > 65535) {
1426 goto done_deref_locked;
1429 * Nested jails always inherit parent's devfs ruleset
1431 if (jailed(td->td_ucred)) {
1432 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1434 goto done_deref_locked;
1436 rsnum = ppr->pr_devfs_rsnum;
1441 if (ppr->pr_flags & PR_IP4) {
1443 * Make sure the new set of IP addresses is a
1444 * subset of the parent's list. Don't worry
1445 * about the parent being unlocked, as any
1446 * setting is done with allprison_lock held.
1448 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1449 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1451 if (ij == ppr->pr_ip4s) {
1453 goto done_deref_locked;
1456 for (ii = ij = 1; ii < ip4s; ii++) {
1457 if (ip4[ii].s_addr ==
1458 ppr->pr_ip4[0].s_addr)
1460 for (; ij < ppr->pr_ip4s; ij++)
1461 if (ip4[ii].s_addr ==
1462 ppr->pr_ip4[ij].s_addr)
1464 if (ij == ppr->pr_ip4s)
1467 if (ij == ppr->pr_ip4s) {
1469 goto done_deref_locked;
1474 * Check for conflicting IP addresses. We permit them
1475 * if there is no more than one IP on each jail. If
1476 * there is a duplicate on a jail with more than one
1477 * IP stop checking and return error.
1481 for (; tppr != &prison0; tppr = tppr->pr_parent)
1482 if (tppr->pr_flags & PR_VNET)
1485 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1488 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1490 tpr->pr_uref == 0) {
1494 if (!(tpr->pr_flags & PR_IP4_USER))
1497 if (tpr->pr_ip4 == NULL ||
1498 (ip4s == 1 && tpr->pr_ip4s == 1))
1500 for (ii = 0; ii < ip4s; ii++) {
1501 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1504 "IPv4 addresses clash");
1505 goto done_deref_locked;
1513 if (ppr->pr_flags & PR_IP6) {
1515 * Make sure the new set of IP addresses is a
1516 * subset of the parent's list.
1518 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1519 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1522 if (ij == ppr->pr_ip6s) {
1524 goto done_deref_locked;
1527 for (ii = ij = 1; ii < ip6s; ii++) {
1528 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1531 for (; ij < ppr->pr_ip6s; ij++)
1532 if (IN6_ARE_ADDR_EQUAL(
1533 &ip6[ii], &ppr->pr_ip6[ij]))
1535 if (ij == ppr->pr_ip6s)
1538 if (ij == ppr->pr_ip6s) {
1540 goto done_deref_locked;
1544 /* Check for conflicting IP addresses. */
1547 for (; tppr != &prison0; tppr = tppr->pr_parent)
1548 if (tppr->pr_flags & PR_VNET)
1551 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1554 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1556 tpr->pr_uref == 0) {
1560 if (!(tpr->pr_flags & PR_IP6_USER))
1563 if (tpr->pr_ip6 == NULL ||
1564 (ip6s == 1 && tpr->pr_ip6s == 1))
1566 for (ii = 0; ii < ip6s; ii++) {
1567 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1570 "IPv6 addresses clash");
1571 goto done_deref_locked;
1577 onamelen = namelen = 0;
1579 /* Give a default name of the jid. */
1580 if (name[0] == '\0')
1581 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1582 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1586 "name cannot be numeric (unless it is the jid)");
1587 goto done_deref_locked;
1590 * Make sure the name isn't too long for the prison or its
1593 onamelen = strlen(pr->pr_name);
1594 namelen = strlen(name);
1595 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1596 error = ENAMETOOLONG;
1597 goto done_deref_locked;
1599 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1600 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1601 sizeof(pr->pr_name)) {
1602 error = ENAMETOOLONG;
1603 goto done_deref_locked;
1607 if (pr_allow & ~ppr->pr_allow) {
1609 goto done_deref_locked;
1612 /* Set the parameters of the prison. */
1615 if (pr_flags & PR_IP4_USER) {
1616 pr->pr_flags |= PR_IP4;
1617 free(pr->pr_ip4, M_PRISON);
1621 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1623 if (tpr->pr_flags & PR_VNET) {
1628 if (prison_restrict_ip4(tpr, NULL)) {
1637 if (pr_flags & PR_IP6_USER) {
1638 pr->pr_flags |= PR_IP6;
1639 free(pr->pr_ip6, M_PRISON);
1643 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1645 if (tpr->pr_flags & PR_VNET) {
1650 if (prison_restrict_ip6(tpr, NULL)) {
1658 pr->pr_securelevel = slevel;
1659 /* Set all child jails to be at least this level. */
1660 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1661 if (tpr->pr_securelevel < slevel)
1662 tpr->pr_securelevel = slevel;
1665 pr->pr_childmax = childmax;
1666 /* Set all child jails to under this limit. */
1667 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1668 if (tpr->pr_childmax > childmax - level)
1669 tpr->pr_childmax = childmax > level
1670 ? childmax - level : 0;
1673 pr->pr_enforce_statfs = enforce;
1674 /* Pass this restriction on to the children. */
1675 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1676 if (tpr->pr_enforce_statfs < enforce)
1677 tpr->pr_enforce_statfs = enforce;
1680 pr->pr_devfs_rsnum = rsnum;
1681 /* Pass this restriction on to the children. */
1682 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1683 tpr->pr_devfs_rsnum = rsnum;
1686 if (ppr == &prison0)
1687 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1689 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1690 ppr->pr_name, name);
1691 /* Change this component of child names. */
1692 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1693 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1694 strlen(tpr->pr_name + onamelen) + 1);
1695 bcopy(pr->pr_name, tpr->pr_name, namelen);
1699 /* Try to keep a real-rooted full pathname. */
1700 if (fullpath_disabled && path[0] == '/' &&
1701 strcmp(mypr->pr_path, "/"))
1702 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1703 mypr->pr_path, path);
1705 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1708 if (PR_HOST & ch_flags & ~pr_flags) {
1709 if (pr->pr_flags & PR_HOST) {
1711 * Copy the parent's host info. As with pr_ip4 above,
1712 * the lack of a lock on the parent is not a problem;
1713 * it is always set with allprison_lock at least
1714 * shared, and is held exclusively here.
1716 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1717 sizeof(pr->pr_hostname));
1718 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1719 sizeof(pr->pr_domainname));
1720 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1721 sizeof(pr->pr_hostuuid));
1722 pr->pr_hostid = pr->pr_parent->pr_hostid;
1724 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1725 /* Set this prison, and any descendants without PR_HOST. */
1727 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1729 strlcpy(pr->pr_domainname, domain,
1730 sizeof(pr->pr_domainname));
1732 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1734 pr->pr_hostid = hid;
1735 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1736 if (tpr->pr_flags & PR_HOST)
1740 strlcpy(tpr->pr_hostname,
1742 sizeof(tpr->pr_hostname));
1744 strlcpy(tpr->pr_domainname,
1746 sizeof(tpr->pr_domainname));
1748 strlcpy(tpr->pr_hostuuid,
1750 sizeof(tpr->pr_hostuuid));
1752 tpr->pr_hostid = hid;
1756 if ((tallow = ch_allow & ~pr_allow)) {
1757 /* Clear allow bits in all children. */
1758 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1759 tpr->pr_allow &= ~tallow;
1761 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1763 * Persistent prisons get an extra reference, and prisons losing their
1764 * persist flag lose that reference. Only do this for existing prisons
1765 * for now, so new ones will remain unseen until after the module
1766 * handlers have completed.
1768 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1769 if (pr_flags & PR_PERSIST) {
1777 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1778 mtx_unlock(&pr->pr_mtx);
1781 if (racct_enable && created)
1782 prison_racct_attach(pr);
1785 /* Locks may have prevented a complete restriction of child IP
1786 * addresses. If so, allocate some more memory and try again.
1791 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1792 mtx_lock(&pr->pr_mtx);
1794 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1796 if (tpr->pr_flags & PR_VNET) {
1801 if (prison_restrict_ip4(tpr, ip4)) {
1808 mtx_unlock(&pr->pr_mtx);
1814 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1815 mtx_lock(&pr->pr_mtx);
1817 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1819 if (tpr->pr_flags & PR_VNET) {
1824 if (prison_restrict_ip6(tpr, ip6)) {
1831 mtx_unlock(&pr->pr_mtx);
1835 /* Let the modules do their work. */
1836 sx_downgrade(&allprison_lock);
1838 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1840 prison_deref(pr, PD_LIST_SLOCKED);
1844 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1846 prison_deref(pr, created
1848 : PD_DEREF | PD_LIST_SLOCKED);
1852 /* Attach this process to the prison if requested. */
1853 if (flags & JAIL_ATTACH) {
1854 mtx_lock(&pr->pr_mtx);
1855 error = do_jail_attach(td, pr);
1857 vfs_opterror(opts, "attach failed");
1859 prison_deref(pr, PD_DEREF);
1865 if (racct_enable && !created) {
1866 if (!(flags & JAIL_ATTACH))
1867 sx_sunlock(&allprison_lock);
1868 prison_racct_modify(pr);
1869 if (!(flags & JAIL_ATTACH))
1870 sx_slock(&allprison_lock);
1874 td->td_retval[0] = pr->pr_id;
1877 * Now that it is all there, drop the temporary reference from existing
1878 * prisons. Or add a reference to newly created persistent prisons
1879 * (which was not done earlier so that the prison would not be publicly
1883 prison_deref(pr, (flags & JAIL_ATTACH)
1885 : PD_DEREF | PD_LIST_SLOCKED);
1887 if (pr_flags & PR_PERSIST) {
1888 mtx_lock(&pr->pr_mtx);
1891 mtx_unlock(&pr->pr_mtx);
1893 if (!(flags & JAIL_ATTACH))
1894 sx_sunlock(&allprison_lock);
1900 prison_deref(pr, created
1901 ? PD_LOCKED | PD_LIST_XLOCKED
1902 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1905 sx_xunlock(&allprison_lock);
1911 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1912 if (errmsg_len > 0) {
1913 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1914 if (errmsg_pos > 0) {
1915 if (optuio->uio_segflg == UIO_SYSSPACE)
1917 optuio->uio_iov[errmsg_pos].iov_base,
1921 optuio->uio_iov[errmsg_pos].iov_base,
1928 free(ip4, M_PRISON);
1931 free(ip6, M_PRISON);
1934 free(g_path, M_TEMP);
1941 * struct jail_get_args {
1942 * struct iovec *iovp;
1943 * unsigned int iovcnt;
1948 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1953 /* Check that we have an even number of iovecs. */
1954 if (uap->iovcnt & 1)
1957 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1960 error = kern_jail_get(td, auio, uap->flags);
1962 error = copyout(auio->uio_iov, uap->iovp,
1963 uap->iovcnt * sizeof (struct iovec));
1969 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1971 struct prison *pr, *mypr;
1973 struct vfsoptlist *opts;
1974 char *errmsg, *name;
1975 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1977 if (flags & ~JAIL_GET_MASK)
1980 /* Get the parameter list. */
1981 error = vfs_buildopts(optuio, &opts);
1984 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1985 mypr = td->td_ucred->cr_prison;
1988 * Find the prison specified by one of: lastjid, jid, name.
1990 sx_slock(&allprison_lock);
1991 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1993 TAILQ_FOREACH(pr, &allprison, pr_list) {
1994 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1995 mtx_lock(&pr->pr_mtx);
1996 if (pr->pr_ref > 0 &&
1997 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1999 mtx_unlock(&pr->pr_mtx);
2005 vfs_opterror(opts, "no jail after %d", jid);
2006 goto done_unlock_list;
2007 } else if (error != ENOENT)
2008 goto done_unlock_list;
2010 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2013 pr = prison_find_child(mypr, jid);
2015 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2016 mtx_unlock(&pr->pr_mtx);
2018 vfs_opterror(opts, "jail %d is dying",
2020 goto done_unlock_list;
2025 vfs_opterror(opts, "jail %d not found", jid);
2026 goto done_unlock_list;
2028 } else if (error != ENOENT)
2029 goto done_unlock_list;
2031 error = vfs_getopt(opts, "name", (void **)&name, &len);
2033 if (len == 0 || name[len - 1] != '\0') {
2035 goto done_unlock_list;
2037 pr = prison_find_name(mypr, name);
2039 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2040 mtx_unlock(&pr->pr_mtx);
2042 vfs_opterror(opts, "jail \"%s\" is dying",
2044 goto done_unlock_list;
2049 vfs_opterror(opts, "jail \"%s\" not found", name);
2050 goto done_unlock_list;
2051 } else if (error != ENOENT)
2052 goto done_unlock_list;
2054 vfs_opterror(opts, "no jail specified");
2056 goto done_unlock_list;
2059 /* Get the parameters of the prison. */
2062 td->td_retval[0] = pr->pr_id;
2063 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2064 if (error != 0 && error != ENOENT)
2066 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2067 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2068 if (error != 0 && error != ENOENT)
2070 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2071 if (error != 0 && error != ENOENT)
2073 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2074 sizeof(pr->pr_cpuset->cs_id));
2075 if (error != 0 && error != ENOENT)
2077 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2078 if (error != 0 && error != ENOENT)
2081 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2082 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2083 if (error != 0 && error != ENOENT)
2087 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2088 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2089 if (error != 0 && error != ENOENT)
2092 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2093 sizeof(pr->pr_securelevel));
2094 if (error != 0 && error != ENOENT)
2096 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2097 sizeof(pr->pr_childcount));
2098 if (error != 0 && error != ENOENT)
2100 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2101 sizeof(pr->pr_childmax));
2102 if (error != 0 && error != ENOENT)
2104 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2105 if (error != 0 && error != ENOENT)
2107 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2108 if (error != 0 && error != ENOENT)
2110 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2111 if (error != 0 && error != ENOENT)
2113 #ifdef COMPAT_FREEBSD32
2114 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2115 uint32_t hid32 = pr->pr_hostid;
2117 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2120 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2121 sizeof(pr->pr_hostid));
2122 if (error != 0 && error != ENOENT)
2124 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2125 sizeof(pr->pr_enforce_statfs));
2126 if (error != 0 && error != ENOENT)
2128 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2129 sizeof(pr->pr_devfs_rsnum));
2130 if (error != 0 && error != ENOENT)
2132 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2134 if (pr_flag_names[fi] == NULL)
2136 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2137 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2138 if (error != 0 && error != ENOENT)
2141 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2142 if (error != 0 && error != ENOENT)
2145 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2148 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2149 i = pr_flag_jailsys[fi].disable &&
2150 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2151 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2154 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2155 if (error != 0 && error != ENOENT)
2158 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2160 if (pr_allow_names[fi] == NULL)
2162 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2163 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2164 if (error != 0 && error != ENOENT)
2167 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2168 if (error != 0 && error != ENOENT)
2171 i = (pr->pr_uref == 0);
2172 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2173 if (error != 0 && error != ENOENT)
2176 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2177 if (error != 0 && error != ENOENT)
2179 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2180 sizeof(pr->pr_osreldate));
2181 if (error != 0 && error != ENOENT)
2183 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2184 if (error != 0 && error != ENOENT)
2187 /* Get the module parameters. */
2188 mtx_unlock(&pr->pr_mtx);
2190 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2193 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2195 /* By now, all parameters should have been noted. */
2196 TAILQ_FOREACH(opt, opts, link) {
2197 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2199 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2204 /* Write the fetched parameters back to userspace. */
2206 TAILQ_FOREACH(opt, opts, link) {
2207 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2208 pos = 2 * opt->pos + 1;
2209 optuio->uio_iov[pos].iov_len = opt->len;
2210 if (opt->value != NULL) {
2211 if (optuio->uio_segflg == UIO_SYSSPACE) {
2213 optuio->uio_iov[pos].iov_base,
2216 error = copyout(opt->value,
2217 optuio->uio_iov[pos].iov_base,
2228 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2232 sx_sunlock(&allprison_lock);
2234 if (error && errmsg_pos >= 0) {
2235 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2236 errmsg_pos = 2 * errmsg_pos + 1;
2237 if (errmsg_len > 0) {
2238 if (optuio->uio_segflg == UIO_SYSSPACE)
2240 optuio->uio_iov[errmsg_pos].iov_base,
2244 optuio->uio_iov[errmsg_pos].iov_base,
2254 * struct jail_remove_args {
2259 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2261 struct prison *pr, *cpr, *lpr, *tpr;
2264 error = priv_check(td, PRIV_JAIL_REMOVE);
2268 sx_xlock(&allprison_lock);
2269 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2271 sx_xunlock(&allprison_lock);
2275 /* Remove all descendants of this prison, then remove this prison. */
2277 pr->pr_flags |= PR_REMOVE;
2278 if (!LIST_EMPTY(&pr->pr_children)) {
2279 mtx_unlock(&pr->pr_mtx);
2281 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2282 mtx_lock(&cpr->pr_mtx);
2283 if (cpr->pr_ref > 0) {
2286 cpr->pr_flags |= PR_REMOVE;
2288 /* Already removed - do not do it again. */
2291 mtx_unlock(&cpr->pr_mtx);
2293 mtx_lock(&lpr->pr_mtx);
2294 prison_remove_one(lpr);
2295 sx_xlock(&allprison_lock);
2300 mtx_lock(&lpr->pr_mtx);
2301 prison_remove_one(lpr);
2302 sx_xlock(&allprison_lock);
2304 mtx_lock(&pr->pr_mtx);
2306 prison_remove_one(pr);
2311 prison_remove_one(struct prison *pr)
2316 /* If the prison was persistent, it is not anymore. */
2318 if (pr->pr_flags & PR_PERSIST) {
2321 pr->pr_flags &= ~PR_PERSIST;
2325 * jail_remove added a reference. If that's the only one, remove
2328 KASSERT(pr->pr_ref > 0,
2329 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2330 if (pr->pr_ref == 1) {
2332 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2336 mtx_unlock(&pr->pr_mtx);
2337 sx_xunlock(&allprison_lock);
2339 * Kill all processes unfortunate enough to be attached to this prison.
2341 sx_slock(&allproc_lock);
2342 LIST_FOREACH(p, &allproc, p_list) {
2344 if (p->p_state != PRS_NEW && p->p_ucred &&
2345 p->p_ucred->cr_prison == pr)
2346 kern_psignal(p, SIGKILL);
2349 sx_sunlock(&allproc_lock);
2350 /* Remove the temporary reference added by jail_remove. */
2351 prison_deref(pr, deuref | PD_DEREF);
2356 * struct jail_attach_args {
2361 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2366 error = priv_check(td, PRIV_JAIL_ATTACH);
2370 sx_slock(&allprison_lock);
2371 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2373 sx_sunlock(&allprison_lock);
2378 * Do not allow a process to attach to a prison that is not
2379 * considered to be "alive".
2381 if (pr->pr_uref == 0) {
2382 mtx_unlock(&pr->pr_mtx);
2383 sx_sunlock(&allprison_lock);
2387 return (do_jail_attach(td, pr));
2391 do_jail_attach(struct thread *td, struct prison *pr)
2395 struct ucred *newcred, *oldcred;
2399 * XXX: Note that there is a slight race here if two threads
2400 * in the same privileged process attempt to attach to two
2401 * different jails at the same time. It is important for
2402 * user processes not to do this, or they might end up with
2403 * a process root from one prison, but attached to the jail
2408 mtx_unlock(&pr->pr_mtx);
2410 /* Let modules do whatever they need to prepare for attaching. */
2411 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2413 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2416 sx_sunlock(&allprison_lock);
2419 * Reparent the newly attached process to this jail.
2421 ppr = td->td_ucred->cr_prison;
2423 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2427 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2428 if ((error = change_dir(pr->pr_root, td)) != 0)
2431 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2434 VOP_UNLOCK(pr->pr_root, 0);
2435 if ((error = change_root(pr->pr_root, td)))
2440 oldcred = p->p_ucred;
2442 crcopy(newcred, oldcred);
2443 newcred->cr_prison = pr;
2444 proc_set_cred(p, newcred);
2447 racct_proc_ucred_changed(p, oldcred, newcred);
2450 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2453 VOP_UNLOCK(pr->pr_root, 0);
2455 /* Tell modules this thread is still in its old jail after all. */
2456 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2457 prison_deref(pr, PD_DEREF | PD_DEUREF);
2463 * Returns a locked prison instance, or NULL on failure.
2466 prison_find(int prid)
2470 sx_assert(&allprison_lock, SX_LOCKED);
2471 TAILQ_FOREACH(pr, &allprison, pr_list) {
2472 if (pr->pr_id == prid) {
2473 mtx_lock(&pr->pr_mtx);
2476 mtx_unlock(&pr->pr_mtx);
2483 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2486 prison_find_child(struct prison *mypr, int prid)
2491 sx_assert(&allprison_lock, SX_LOCKED);
2492 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2493 if (pr->pr_id == prid) {
2494 mtx_lock(&pr->pr_mtx);
2497 mtx_unlock(&pr->pr_mtx);
2504 * Look for the name relative to mypr. Returns a locked prison or NULL.
2507 prison_find_name(struct prison *mypr, const char *name)
2509 struct prison *pr, *deadpr;
2513 sx_assert(&allprison_lock, SX_LOCKED);
2514 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2517 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2518 if (!strcmp(pr->pr_name + mylen, name)) {
2519 mtx_lock(&pr->pr_mtx);
2520 if (pr->pr_ref > 0) {
2521 if (pr->pr_uref > 0)
2525 mtx_unlock(&pr->pr_mtx);
2528 /* There was no valid prison - perhaps there was a dying one. */
2529 if (deadpr != NULL) {
2530 mtx_lock(&deadpr->pr_mtx);
2531 if (deadpr->pr_ref == 0) {
2532 mtx_unlock(&deadpr->pr_mtx);
2540 * See if a prison has the specific flag set.
2543 prison_flag(struct ucred *cred, unsigned flag)
2546 /* This is an atomic read, so no locking is necessary. */
2547 return (cred->cr_prison->pr_flags & flag);
2551 prison_allow(struct ucred *cred, unsigned flag)
2554 /* This is an atomic read, so no locking is necessary. */
2555 return (cred->cr_prison->pr_allow & flag);
2559 * Remove a prison reference. If that was the last reference, remove the
2560 * prison itself - but not in this context in case there are locks held.
2563 prison_free_locked(struct prison *pr)
2566 mtx_assert(&pr->pr_mtx, MA_OWNED);
2568 if (pr->pr_ref == 0) {
2569 mtx_unlock(&pr->pr_mtx);
2570 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2571 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2574 mtx_unlock(&pr->pr_mtx);
2578 prison_free(struct prison *pr)
2581 mtx_lock(&pr->pr_mtx);
2582 prison_free_locked(pr);
2586 prison_complete(void *context, int pending)
2589 prison_deref((struct prison *)context, 0);
2593 * Remove a prison reference (usually). This internal version assumes no
2594 * mutexes are held, except perhaps the prison itself. If there are no more
2595 * references, release and delist the prison. On completion, the prison lock
2596 * and the allprison lock are both unlocked.
2599 prison_deref(struct prison *pr, int flags)
2601 struct prison *ppr, *tpr;
2603 if (!(flags & PD_LOCKED))
2604 mtx_lock(&pr->pr_mtx);
2606 if (flags & PD_DEUREF) {
2608 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2610 if (flags & PD_DEREF)
2612 /* If the prison still has references, nothing else to do. */
2613 if (pr->pr_ref > 0) {
2614 mtx_unlock(&pr->pr_mtx);
2615 if (flags & PD_LIST_SLOCKED)
2616 sx_sunlock(&allprison_lock);
2617 else if (flags & PD_LIST_XLOCKED)
2618 sx_xunlock(&allprison_lock);
2622 mtx_unlock(&pr->pr_mtx);
2623 if (flags & PD_LIST_SLOCKED) {
2624 if (!sx_try_upgrade(&allprison_lock)) {
2625 sx_sunlock(&allprison_lock);
2626 sx_xlock(&allprison_lock);
2628 } else if (!(flags & PD_LIST_XLOCKED))
2629 sx_xlock(&allprison_lock);
2631 TAILQ_REMOVE(&allprison, pr, pr_list);
2632 LIST_REMOVE(pr, pr_sibling);
2633 ppr = pr->pr_parent;
2634 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2635 tpr->pr_childcount--;
2636 sx_xunlock(&allprison_lock);
2639 if (pr->pr_vnet != ppr->pr_vnet)
2640 vnet_destroy(pr->pr_vnet);
2642 if (pr->pr_root != NULL)
2644 mtx_destroy(&pr->pr_mtx);
2646 free(pr->pr_ip4, M_PRISON);
2649 free(pr->pr_ip6, M_PRISON);
2651 if (pr->pr_cpuset != NULL)
2652 cpuset_rel(pr->pr_cpuset);
2656 prison_racct_detach(pr);
2660 /* Removing a prison frees a reference on its parent. */
2662 mtx_lock(&pr->pr_mtx);
2663 flags = PD_DEREF | PD_DEUREF;
2668 prison_hold_locked(struct prison *pr)
2671 mtx_assert(&pr->pr_mtx, MA_OWNED);
2672 KASSERT(pr->pr_ref > 0,
2673 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2678 prison_hold(struct prison *pr)
2681 mtx_lock(&pr->pr_mtx);
2682 prison_hold_locked(pr);
2683 mtx_unlock(&pr->pr_mtx);
2687 prison_proc_hold(struct prison *pr)
2690 mtx_lock(&pr->pr_mtx);
2691 KASSERT(pr->pr_uref > 0,
2692 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2694 mtx_unlock(&pr->pr_mtx);
2698 prison_proc_free(struct prison *pr)
2701 mtx_lock(&pr->pr_mtx);
2702 KASSERT(pr->pr_uref > 0,
2703 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2704 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2710 * Restrict a prison's IP address list with its parent's, possibly replacing
2711 * it. Return true if the replacement buffer was used (or would have been).
2714 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2719 ppr = pr->pr_parent;
2720 if (!(pr->pr_flags & PR_IP4_USER)) {
2721 /* This has no user settings, so just copy the parent's list. */
2722 if (pr->pr_ip4s < ppr->pr_ip4s) {
2724 * There's no room for the parent's list. Use the
2725 * new list buffer, which is assumed to be big enough
2726 * (if it was passed). If there's no buffer, try to
2730 if (newip4 == NULL) {
2731 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2732 M_PRISON, M_NOWAIT);
2736 if (newip4 != NULL) {
2737 bcopy(ppr->pr_ip4, newip4,
2738 ppr->pr_ip4s * sizeof(*newip4));
2739 free(pr->pr_ip4, M_PRISON);
2740 pr->pr_ip4 = newip4;
2741 pr->pr_ip4s = ppr->pr_ip4s;
2745 pr->pr_ip4s = ppr->pr_ip4s;
2746 if (pr->pr_ip4s > 0)
2747 bcopy(ppr->pr_ip4, pr->pr_ip4,
2748 pr->pr_ip4s * sizeof(*newip4));
2749 else if (pr->pr_ip4 != NULL) {
2750 free(pr->pr_ip4, M_PRISON);
2753 } else if (pr->pr_ip4s > 0) {
2754 /* Remove addresses that aren't in the parent. */
2755 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2756 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2758 if (ij < ppr->pr_ip4s)
2761 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2762 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2765 for (ij = 1; ii < pr->pr_ip4s; ) {
2766 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2770 switch (ij >= ppr->pr_ip4s ? -1 :
2771 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2773 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2774 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2785 if (pr->pr_ip4s == 0) {
2786 free(pr->pr_ip4, M_PRISON);
2794 * Pass back primary IPv4 address of this jail.
2796 * If not restricted return success but do not alter the address. Caller has
2797 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2799 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2800 * Address returned in NBO.
2803 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2807 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2808 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2810 pr = cred->cr_prison;
2811 if (!(pr->pr_flags & PR_IP4))
2813 mtx_lock(&pr->pr_mtx);
2814 if (!(pr->pr_flags & PR_IP4)) {
2815 mtx_unlock(&pr->pr_mtx);
2818 if (pr->pr_ip4 == NULL) {
2819 mtx_unlock(&pr->pr_mtx);
2820 return (EAFNOSUPPORT);
2823 ia->s_addr = pr->pr_ip4[0].s_addr;
2824 mtx_unlock(&pr->pr_mtx);
2829 * Return 1 if we should do proper source address selection or are not jailed.
2830 * We will return 0 if we should bypass source address selection in favour
2831 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2833 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2836 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2842 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2843 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2848 pr = cred->cr_prison;
2849 if (pr->pr_flags & PR_IP4_SADDRSEL)
2852 lia.s_addr = INADDR_ANY;
2853 error = prison_get_ip4(cred, &lia);
2856 if (lia.s_addr == INADDR_ANY)
2859 ia->s_addr = lia.s_addr;
2864 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2867 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2874 * No need to lock since the PR_IP4_USER flag can't be altered for
2877 while (pr1 != &prison0 &&
2879 !(pr1->pr_flags & PR_VNET) &&
2881 !(pr1->pr_flags & PR_IP4_USER))
2882 pr1 = pr1->pr_parent;
2883 while (pr2 != &prison0 &&
2885 !(pr2->pr_flags & PR_VNET) &&
2887 !(pr2->pr_flags & PR_IP4_USER))
2888 pr2 = pr2->pr_parent;
2889 return (pr1 == pr2);
2893 * Make sure our (source) address is set to something meaningful to this
2896 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2897 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2898 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2901 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2907 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2908 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2910 pr = cred->cr_prison;
2911 if (!(pr->pr_flags & PR_IP4))
2913 mtx_lock(&pr->pr_mtx);
2914 if (!(pr->pr_flags & PR_IP4)) {
2915 mtx_unlock(&pr->pr_mtx);
2918 if (pr->pr_ip4 == NULL) {
2919 mtx_unlock(&pr->pr_mtx);
2920 return (EAFNOSUPPORT);
2923 ia0.s_addr = ntohl(ia->s_addr);
2924 if (ia0.s_addr == INADDR_LOOPBACK) {
2925 ia->s_addr = pr->pr_ip4[0].s_addr;
2926 mtx_unlock(&pr->pr_mtx);
2930 if (ia0.s_addr == INADDR_ANY) {
2932 * In case there is only 1 IPv4 address, bind directly.
2934 if (pr->pr_ip4s == 1)
2935 ia->s_addr = pr->pr_ip4[0].s_addr;
2936 mtx_unlock(&pr->pr_mtx);
2940 error = _prison_check_ip4(pr, ia);
2941 mtx_unlock(&pr->pr_mtx);
2946 * Rewrite destination address in case we will connect to loopback address.
2948 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2949 * Address passed in in NBO and returned in NBO.
2952 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2956 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2957 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2959 pr = cred->cr_prison;
2960 if (!(pr->pr_flags & PR_IP4))
2962 mtx_lock(&pr->pr_mtx);
2963 if (!(pr->pr_flags & PR_IP4)) {
2964 mtx_unlock(&pr->pr_mtx);
2967 if (pr->pr_ip4 == NULL) {
2968 mtx_unlock(&pr->pr_mtx);
2969 return (EAFNOSUPPORT);
2972 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2973 ia->s_addr = pr->pr_ip4[0].s_addr;
2974 mtx_unlock(&pr->pr_mtx);
2979 * Return success because nothing had to be changed.
2981 mtx_unlock(&pr->pr_mtx);
2986 * Check if given address belongs to the jail referenced by cred/prison.
2988 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2989 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2990 * doesn't allow IPv4. Address passed in in NBO.
2993 _prison_check_ip4(const struct prison *pr, const struct in_addr *ia)
2998 * Check the primary IP.
3000 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3004 * All the other IPs are sorted so we can do a binary search.
3007 z = pr->pr_ip4s - 2;
3010 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3019 return (EADDRNOTAVAIL);
3023 prison_check_ip4(const struct ucred *cred, const struct in_addr *ia)
3028 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3029 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3031 pr = cred->cr_prison;
3032 if (!(pr->pr_flags & PR_IP4))
3034 mtx_lock(&pr->pr_mtx);
3035 if (!(pr->pr_flags & PR_IP4)) {
3036 mtx_unlock(&pr->pr_mtx);
3039 if (pr->pr_ip4 == NULL) {
3040 mtx_unlock(&pr->pr_mtx);
3041 return (EAFNOSUPPORT);
3044 error = _prison_check_ip4(pr, ia);
3045 mtx_unlock(&pr->pr_mtx);
3052 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3057 ppr = pr->pr_parent;
3058 if (!(pr->pr_flags & PR_IP6_USER)) {
3059 /* This has no user settings, so just copy the parent's list. */
3060 if (pr->pr_ip6s < ppr->pr_ip6s) {
3062 * There's no room for the parent's list. Use the
3063 * new list buffer, which is assumed to be big enough
3064 * (if it was passed). If there's no buffer, try to
3068 if (newip6 == NULL) {
3069 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3070 M_PRISON, M_NOWAIT);
3074 if (newip6 != NULL) {
3075 bcopy(ppr->pr_ip6, newip6,
3076 ppr->pr_ip6s * sizeof(*newip6));
3077 free(pr->pr_ip6, M_PRISON);
3078 pr->pr_ip6 = newip6;
3079 pr->pr_ip6s = ppr->pr_ip6s;
3083 pr->pr_ip6s = ppr->pr_ip6s;
3084 if (pr->pr_ip6s > 0)
3085 bcopy(ppr->pr_ip6, pr->pr_ip6,
3086 pr->pr_ip6s * sizeof(*newip6));
3087 else if (pr->pr_ip6 != NULL) {
3088 free(pr->pr_ip6, M_PRISON);
3091 } else if (pr->pr_ip6s > 0) {
3092 /* Remove addresses that aren't in the parent. */
3093 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3094 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3097 if (ij < ppr->pr_ip6s)
3100 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3101 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3104 for (ij = 1; ii < pr->pr_ip6s; ) {
3105 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3110 switch (ij >= ppr->pr_ip6s ? -1 :
3111 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3113 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3114 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3125 if (pr->pr_ip6s == 0) {
3126 free(pr->pr_ip6, M_PRISON);
3134 * Pass back primary IPv6 address for this jail.
3136 * If not restricted return success but do not alter the address. Caller has
3137 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3139 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3142 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3146 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3147 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3149 pr = cred->cr_prison;
3150 if (!(pr->pr_flags & PR_IP6))
3152 mtx_lock(&pr->pr_mtx);
3153 if (!(pr->pr_flags & PR_IP6)) {
3154 mtx_unlock(&pr->pr_mtx);
3157 if (pr->pr_ip6 == NULL) {
3158 mtx_unlock(&pr->pr_mtx);
3159 return (EAFNOSUPPORT);
3162 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3163 mtx_unlock(&pr->pr_mtx);
3168 * Return 1 if we should do proper source address selection or are not jailed.
3169 * We will return 0 if we should bypass source address selection in favour
3170 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3172 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3175 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3178 struct in6_addr lia6;
3181 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3182 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3187 pr = cred->cr_prison;
3188 if (pr->pr_flags & PR_IP6_SADDRSEL)
3192 error = prison_get_ip6(cred, &lia6);
3195 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3198 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3203 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3206 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3212 while (pr1 != &prison0 &&
3214 !(pr1->pr_flags & PR_VNET) &&
3216 !(pr1->pr_flags & PR_IP6_USER))
3217 pr1 = pr1->pr_parent;
3218 while (pr2 != &prison0 &&
3220 !(pr2->pr_flags & PR_VNET) &&
3222 !(pr2->pr_flags & PR_IP6_USER))
3223 pr2 = pr2->pr_parent;
3224 return (pr1 == pr2);
3228 * Make sure our (source) address is set to something meaningful to this jail.
3230 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3231 * when needed while binding.
3233 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3234 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3235 * doesn't allow IPv6.
3238 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3243 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3244 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3246 pr = cred->cr_prison;
3247 if (!(pr->pr_flags & PR_IP6))
3249 mtx_lock(&pr->pr_mtx);
3250 if (!(pr->pr_flags & PR_IP6)) {
3251 mtx_unlock(&pr->pr_mtx);
3254 if (pr->pr_ip6 == NULL) {
3255 mtx_unlock(&pr->pr_mtx);
3256 return (EAFNOSUPPORT);
3259 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3260 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3261 mtx_unlock(&pr->pr_mtx);
3265 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3267 * In case there is only 1 IPv6 address, and v6only is true,
3268 * then bind directly.
3270 if (v6only != 0 && pr->pr_ip6s == 1)
3271 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3272 mtx_unlock(&pr->pr_mtx);
3276 error = _prison_check_ip6(pr, ia6);
3277 mtx_unlock(&pr->pr_mtx);
3282 * Rewrite destination address in case we will connect to loopback address.
3284 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3287 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3291 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3292 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3294 pr = cred->cr_prison;
3295 if (!(pr->pr_flags & PR_IP6))
3297 mtx_lock(&pr->pr_mtx);
3298 if (!(pr->pr_flags & PR_IP6)) {
3299 mtx_unlock(&pr->pr_mtx);
3302 if (pr->pr_ip6 == NULL) {
3303 mtx_unlock(&pr->pr_mtx);
3304 return (EAFNOSUPPORT);
3307 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3308 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3309 mtx_unlock(&pr->pr_mtx);
3314 * Return success because nothing had to be changed.
3316 mtx_unlock(&pr->pr_mtx);
3321 * Check if given address belongs to the jail referenced by cred/prison.
3323 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3324 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3325 * doesn't allow IPv6.
3328 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3333 * Check the primary IP.
3335 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3339 * All the other IPs are sorted so we can do a binary search.
3342 z = pr->pr_ip6s - 2;
3345 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3354 return (EADDRNOTAVAIL);
3358 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3363 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3364 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3366 pr = cred->cr_prison;
3367 if (!(pr->pr_flags & PR_IP6))
3369 mtx_lock(&pr->pr_mtx);
3370 if (!(pr->pr_flags & PR_IP6)) {
3371 mtx_unlock(&pr->pr_mtx);
3374 if (pr->pr_ip6 == NULL) {
3375 mtx_unlock(&pr->pr_mtx);
3376 return (EAFNOSUPPORT);
3379 error = _prison_check_ip6(pr, ia6);
3380 mtx_unlock(&pr->pr_mtx);
3386 * Check if a jail supports the given address family.
3388 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3392 prison_check_af(struct ucred *cred, int af)
3397 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3399 pr = cred->cr_prison;
3401 /* Prisons with their own network stack are not limited. */
3402 if (prison_owns_vnet(cred))
3411 if (pr->pr_flags & PR_IP4)
3413 mtx_lock(&pr->pr_mtx);
3414 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3415 error = EAFNOSUPPORT;
3416 mtx_unlock(&pr->pr_mtx);
3422 if (pr->pr_flags & PR_IP6)
3424 mtx_lock(&pr->pr_mtx);
3425 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3426 error = EAFNOSUPPORT;
3427 mtx_unlock(&pr->pr_mtx);
3435 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3436 error = EAFNOSUPPORT;
3442 * Check if given address belongs to the jail referenced by cred (wrapper to
3443 * prison_check_ip[46]).
3445 * Returns 0 if jail doesn't restrict the address family or if address belongs
3446 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3447 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3450 prison_if(struct ucred *cred, struct sockaddr *sa)
3453 struct sockaddr_in *sai;
3456 struct sockaddr_in6 *sai6;
3460 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3461 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3464 if (prison_owns_vnet(cred))
3469 switch (sa->sa_family)
3473 sai = (struct sockaddr_in *)sa;
3474 error = prison_check_ip4(cred, &sai->sin_addr);
3479 sai6 = (struct sockaddr_in6 *)sa;
3480 error = prison_check_ip6(cred, &sai6->sin6_addr);
3484 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3485 error = EAFNOSUPPORT;
3491 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3494 prison_check(struct ucred *cred1, struct ucred *cred2)
3497 return ((cred1->cr_prison == cred2->cr_prison ||
3498 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3502 * Return 1 if p2 is a child of p1, otherwise 0.
3505 prison_ischild(struct prison *pr1, struct prison *pr2)
3508 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3515 * Return 1 if the passed credential is in a jail, otherwise 0.
3518 jailed(struct ucred *cred)
3521 return (cred->cr_prison != &prison0);
3525 * Return 1 if the passed credential is in a jail and that jail does not
3526 * have its own virtual network stack, otherwise 0.
3529 jailed_without_vnet(struct ucred *cred)
3535 if (prison_owns_vnet(cred))
3543 * Return the correct hostname (domainname, et al) for the passed credential.
3546 getcredhostname(struct ucred *cred, char *buf, size_t size)
3551 * A NULL credential can be used to shortcut to the physical
3552 * system's hostname.
3554 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3555 mtx_lock(&pr->pr_mtx);
3556 strlcpy(buf, pr->pr_hostname, size);
3557 mtx_unlock(&pr->pr_mtx);
3561 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3564 mtx_lock(&cred->cr_prison->pr_mtx);
3565 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3566 mtx_unlock(&cred->cr_prison->pr_mtx);
3570 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3573 mtx_lock(&cred->cr_prison->pr_mtx);
3574 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3575 mtx_unlock(&cred->cr_prison->pr_mtx);
3579 getcredhostid(struct ucred *cred, unsigned long *hostid)
3582 mtx_lock(&cred->cr_prison->pr_mtx);
3583 *hostid = cred->cr_prison->pr_hostid;
3584 mtx_unlock(&cred->cr_prison->pr_mtx);
3589 * Determine whether the prison represented by cred owns
3590 * its vnet rather than having it inherited.
3592 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3595 prison_owns_vnet(struct ucred *cred)
3599 * vnets cannot be added/removed after jail creation,
3600 * so no need to lock here.
3602 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3607 * Determine whether the subject represented by cred can "see"
3608 * status of a mount point.
3609 * Returns: 0 for permitted, ENOENT otherwise.
3610 * XXX: This function should be called cr_canseemount() and should be
3611 * placed in kern_prot.c.
3614 prison_canseemount(struct ucred *cred, struct mount *mp)
3620 pr = cred->cr_prison;
3621 if (pr->pr_enforce_statfs == 0)
3623 if (pr->pr_root->v_mount == mp)
3625 if (pr->pr_enforce_statfs == 2)
3628 * If jail's chroot directory is set to "/" we should be able to see
3629 * all mount-points from inside a jail.
3630 * This is ugly check, but this is the only situation when jail's
3631 * directory ends with '/'.
3633 if (strcmp(pr->pr_path, "/") == 0)
3635 len = strlen(pr->pr_path);
3637 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3640 * Be sure that we don't have situation where jail's root directory
3641 * is "/some/path" and mount point is "/some/pathpath".
3643 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3649 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3651 char jpath[MAXPATHLEN];
3655 pr = cred->cr_prison;
3656 if (pr->pr_enforce_statfs == 0)
3658 if (prison_canseemount(cred, mp) != 0) {
3659 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3660 strlcpy(sp->f_mntonname, "[restricted]",
3661 sizeof(sp->f_mntonname));
3664 if (pr->pr_root->v_mount == mp) {
3666 * Clear current buffer data, so we are sure nothing from
3667 * the valid path left there.
3669 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3670 *sp->f_mntonname = '/';
3674 * If jail's chroot directory is set to "/" we should be able to see
3675 * all mount-points from inside a jail.
3677 if (strcmp(pr->pr_path, "/") == 0)
3679 len = strlen(pr->pr_path);
3680 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3682 * Clear current buffer data, so we are sure nothing from
3683 * the valid path left there.
3685 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3686 if (*jpath == '\0') {
3687 /* Should never happen. */
3688 *sp->f_mntonname = '/';
3690 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3695 * Check with permission for a specific privilege is granted within jail. We
3696 * have a specific list of accepted privileges; the rest are denied.
3699 prison_priv_check(struct ucred *cred, int priv)
3707 * Privileges specific to prisons with a virtual network stack.
3708 * There might be a duplicate entry here in case the privilege
3709 * is only granted conditionally in the legacy jail case.
3714 * NFS-specific privileges.
3716 case PRIV_NFS_DAEMON:
3717 case PRIV_NFS_LOCKD:
3720 * Network stack privileges.
3722 case PRIV_NET_BRIDGE:
3725 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3726 case PRIV_NET_ROUTE:
3728 case PRIV_NET_SETIFMTU:
3729 case PRIV_NET_SETIFFLAGS:
3730 case PRIV_NET_SETIFCAP:
3731 case PRIV_NET_SETIFDESCR:
3732 case PRIV_NET_SETIFNAME :
3733 case PRIV_NET_SETIFMETRIC:
3734 case PRIV_NET_SETIFPHYS:
3735 case PRIV_NET_SETIFMAC:
3736 case PRIV_NET_ADDMULTI:
3737 case PRIV_NET_DELMULTI:
3738 case PRIV_NET_HWIOCTL:
3739 case PRIV_NET_SETLLADDR:
3740 case PRIV_NET_ADDIFGROUP:
3741 case PRIV_NET_DELIFGROUP:
3742 case PRIV_NET_IFCREATE:
3743 case PRIV_NET_IFDESTROY:
3744 case PRIV_NET_ADDIFADDR:
3745 case PRIV_NET_DELIFADDR:
3748 case PRIV_NET_SETIFVNET:
3749 case PRIV_NET_SETIFFIB:
3752 * 802.11-related privileges.
3754 case PRIV_NET80211_GETKEY:
3756 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3763 case PRIV_NETATM_CFG:
3764 case PRIV_NETATM_ADD:
3765 case PRIV_NETATM_DEL:
3766 case PRIV_NETATM_SET:
3769 * Bluetooth privileges.
3771 case PRIV_NETBLUETOOTH_RAW:
3775 * Netgraph and netgraph module privileges.
3777 case PRIV_NETGRAPH_CONTROL:
3779 case PRIV_NETGRAPH_TTY:
3783 * IPv4 and IPv6 privileges.
3785 case PRIV_NETINET_IPFW:
3786 case PRIV_NETINET_DIVERT:
3787 case PRIV_NETINET_PF:
3788 case PRIV_NETINET_DUMMYNET:
3789 case PRIV_NETINET_CARP:
3790 case PRIV_NETINET_MROUTE:
3791 case PRIV_NETINET_RAW:
3792 case PRIV_NETINET_ADDRCTRL6:
3793 case PRIV_NETINET_ND6:
3794 case PRIV_NETINET_SCOPE6:
3795 case PRIV_NETINET_ALIFETIME6:
3796 case PRIV_NETINET_IPSEC:
3797 case PRIV_NETINET_BINDANY:
3812 * No default: or deny here.
3813 * In case of no permit fall through to next switch().
3815 if (cred->cr_prison->pr_flags & PR_VNET)
3823 * Allow ktrace privileges for root in jail.
3829 * Allow jailed processes to configure audit identity and
3830 * submit audit records (login, etc). In the future we may
3831 * want to further refine the relationship between audit and
3834 case PRIV_AUDIT_GETAUDIT:
3835 case PRIV_AUDIT_SETAUDIT:
3836 case PRIV_AUDIT_SUBMIT:
3840 * Allow jailed processes to manipulate process UNIX
3841 * credentials in any way they see fit.
3843 case PRIV_CRED_SETUID:
3844 case PRIV_CRED_SETEUID:
3845 case PRIV_CRED_SETGID:
3846 case PRIV_CRED_SETEGID:
3847 case PRIV_CRED_SETGROUPS:
3848 case PRIV_CRED_SETREUID:
3849 case PRIV_CRED_SETREGID:
3850 case PRIV_CRED_SETRESUID:
3851 case PRIV_CRED_SETRESGID:
3854 * Jail implements visibility constraints already, so allow
3855 * jailed root to override uid/gid-based constraints.
3857 case PRIV_SEEOTHERGIDS:
3858 case PRIV_SEEOTHERUIDS:
3861 * Jail implements inter-process debugging limits already, so
3862 * allow jailed root various debugging privileges.
3864 case PRIV_DEBUG_DIFFCRED:
3865 case PRIV_DEBUG_SUGID:
3866 case PRIV_DEBUG_UNPRIV:
3869 * Allow jail to set various resource limits and login
3870 * properties, and for now, exceed process resource limits.
3872 case PRIV_PROC_LIMIT:
3873 case PRIV_PROC_SETLOGIN:
3874 case PRIV_PROC_SETRLIMIT:
3877 * System V and POSIX IPC privileges are granted in jail.
3880 case PRIV_IPC_WRITE:
3881 case PRIV_IPC_ADMIN:
3882 case PRIV_IPC_MSGSIZE:
3886 * Jail operations within a jail work on child jails.
3888 case PRIV_JAIL_ATTACH:
3890 case PRIV_JAIL_REMOVE:
3893 * Jail implements its own inter-process limits, so allow
3894 * root processes in jail to change scheduling on other
3895 * processes in the same jail. Likewise for signalling.
3897 case PRIV_SCHED_DIFFCRED:
3898 case PRIV_SCHED_CPUSET:
3899 case PRIV_SIGNAL_DIFFCRED:
3900 case PRIV_SIGNAL_SUGID:
3903 * Allow jailed processes to write to sysctls marked as jail
3906 case PRIV_SYSCTL_WRITEJAIL:
3909 * Allow root in jail to manage a variety of quota
3910 * properties. These should likely be conditional on a
3911 * configuration option.
3913 case PRIV_VFS_GETQUOTA:
3914 case PRIV_VFS_SETQUOTA:
3917 * Since Jail relies on chroot() to implement file system
3918 * protections, grant many VFS privileges to root in jail.
3919 * Be careful to exclude mount-related and NFS-related
3923 case PRIV_VFS_WRITE:
3924 case PRIV_VFS_ADMIN:
3926 case PRIV_VFS_LOOKUP:
3927 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3928 case PRIV_VFS_CHFLAGS_DEV:
3929 case PRIV_VFS_CHOWN:
3930 case PRIV_VFS_CHROOT:
3931 case PRIV_VFS_RETAINSUGID:
3932 case PRIV_VFS_FCHROOT:
3934 case PRIV_VFS_SETGID:
3936 case PRIV_VFS_STICKYFILE:
3939 * As in the non-jail case, non-root users are expected to be
3940 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3941 * exists in the jail and they have permission to access it).
3943 case PRIV_KMEM_READ:
3947 * Depending on the global setting, allow privilege of
3948 * setting system flags.
3950 case PRIV_VFS_SYSFLAGS:
3951 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3957 * Depending on the global setting, allow privilege of
3958 * mounting/unmounting file systems.
3960 case PRIV_VFS_MOUNT:
3961 case PRIV_VFS_UNMOUNT:
3962 case PRIV_VFS_MOUNT_NONUSER:
3963 case PRIV_VFS_MOUNT_OWNER:
3964 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3965 cred->cr_prison->pr_enforce_statfs < 2)
3971 * Allow jailed root to bind reserved ports and reuse in-use
3974 case PRIV_NETINET_RESERVEDPORT:
3975 case PRIV_NETINET_REUSEPORT:
3979 * Allow jailed root to set certian IPv4/6 (option) headers.
3981 case PRIV_NETINET_SETHDROPTS:
3985 * Conditionally allow creating raw sockets in jail.
3987 case PRIV_NETINET_RAW:
3988 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3994 * Since jail implements its own visibility limits on netstat
3995 * sysctls, allow getcred. This allows identd to work in
3998 case PRIV_NETINET_GETCRED:
4002 * Allow jailed root to set loginclass.
4004 case PRIV_PROC_SETLOGINCLASS:
4009 * In all remaining cases, deny the privilege request. This
4010 * includes almost all network privileges, many system
4011 * configuration privileges.
4018 * Return the part of pr2's name that is relative to pr1, or the whole name
4019 * if it does not directly follow.
4023 prison_name(struct prison *pr1, struct prison *pr2)
4027 /* Jails see themselves as "0" (if they see themselves at all). */
4030 name = pr2->pr_name;
4031 if (prison_ischild(pr1, pr2)) {
4033 * pr1 isn't locked (and allprison_lock may not be either)
4034 * so its length can't be counted on. But the number of dots
4035 * can be counted on - and counted.
4037 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4038 name = strchr(name, '.') + 1;
4044 * Return the part of pr2's path that is relative to pr1, or the whole path
4045 * if it does not directly follow.
4048 prison_path(struct prison *pr1, struct prison *pr2)
4050 char *path1, *path2;
4053 path1 = pr1->pr_path;
4054 path2 = pr2->pr_path;
4055 if (!strcmp(path1, "/"))
4057 len1 = strlen(path1);
4058 if (strncmp(path1, path2, len1))
4060 if (path2[len1] == '\0')
4062 if (path2[len1] == '/')
4063 return (path2 + len1);
4069 * Jail-related sysctls.
4071 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4075 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4078 struct prison *pr, *cpr;
4080 struct in_addr *ip4 = NULL;
4084 struct in6_addr *ip6 = NULL;
4089 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4090 pr = req->td->td_ucred->cr_prison;
4092 sx_slock(&allprison_lock);
4093 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4094 #if defined(INET) || defined(INET6)
4097 mtx_lock(&cpr->pr_mtx);
4099 if (cpr->pr_ip4s > 0) {
4100 if (ip4s < cpr->pr_ip4s) {
4101 ip4s = cpr->pr_ip4s;
4102 mtx_unlock(&cpr->pr_mtx);
4103 ip4 = realloc(ip4, ip4s *
4104 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4107 bcopy(cpr->pr_ip4, ip4,
4108 cpr->pr_ip4s * sizeof(struct in_addr));
4112 if (cpr->pr_ip6s > 0) {
4113 if (ip6s < cpr->pr_ip6s) {
4114 ip6s = cpr->pr_ip6s;
4115 mtx_unlock(&cpr->pr_mtx);
4116 ip6 = realloc(ip6, ip6s *
4117 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4120 bcopy(cpr->pr_ip6, ip6,
4121 cpr->pr_ip6s * sizeof(struct in6_addr));
4124 if (cpr->pr_ref == 0) {
4125 mtx_unlock(&cpr->pr_mtx);
4128 bzero(xp, sizeof(*xp));
4129 xp->pr_version = XPRISON_VERSION;
4130 xp->pr_id = cpr->pr_id;
4131 xp->pr_state = cpr->pr_uref > 0
4132 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4133 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4134 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4135 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4137 xp->pr_ip4s = cpr->pr_ip4s;
4140 xp->pr_ip6s = cpr->pr_ip6s;
4142 mtx_unlock(&cpr->pr_mtx);
4143 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4147 if (xp->pr_ip4s > 0) {
4148 error = SYSCTL_OUT(req, ip4,
4149 xp->pr_ip4s * sizeof(struct in_addr));
4155 if (xp->pr_ip6s > 0) {
4156 error = SYSCTL_OUT(req, ip6,
4157 xp->pr_ip6s * sizeof(struct in6_addr));
4163 sx_sunlock(&allprison_lock);
4174 SYSCTL_OID(_security_jail, OID_AUTO, list,
4175 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4176 sysctl_jail_list, "S", "List of active jails");
4179 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4183 injail = jailed(req->td->td_ucred);
4184 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4189 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4190 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4191 sysctl_jail_jailed, "I", "Process in jail?");
4194 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4196 int error, havevnet;
4198 struct ucred *cred = req->td->td_ucred;
4200 havevnet = jailed(cred) && prison_owns_vnet(cred);
4204 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4209 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4210 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4211 sysctl_jail_vnet, "I", "Jail owns VNET?");
4213 #if defined(INET) || defined(INET6)
4214 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4215 &jail_max_af_ips, 0,
4216 "Number of IP addresses a jail may have at most per address family");
4220 * Default parameters for jail(2) compatability. For historical reasons,
4221 * the sysctl names have varying similarity to the parameter names. Prisons
4222 * just see their own parameters, and can't change them.
4225 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4228 int allow, error, i;
4230 pr = req->td->td_ucred->cr_prison;
4231 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4233 /* Get the current flag value, and convert it to a boolean. */
4234 i = (allow & arg2) ? 1 : 0;
4237 error = sysctl_handle_int(oidp, &i, 0, req);
4238 if (error || !req->newptr)
4244 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4247 mtx_lock(&prison0.pr_mtx);
4248 jail_default_allow = (jail_default_allow & ~arg2) | i;
4249 mtx_unlock(&prison0.pr_mtx);
4253 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4254 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4255 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4256 "Processes in jail can set their hostnames");
4257 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4258 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4259 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4260 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4261 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4262 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4263 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4264 "Processes in jail can use System V IPC primitives");
4265 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4266 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4267 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4268 "Prison root can create raw sockets");
4269 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4270 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4271 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4272 "Processes in jail can alter system file flags");
4273 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4274 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4275 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4276 "Processes in jail can mount/unmount jail-friendly file systems");
4277 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4278 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4279 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4280 "Processes in jail can mount the devfs file system");
4281 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4282 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4283 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4284 "Processes in jail can mount the fdescfs file system");
4285 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4286 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4287 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4288 "Processes in jail can mount the nullfs file system");
4289 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4290 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4291 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4292 "Processes in jail can mount the procfs file system");
4293 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4294 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4295 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4296 "Processes in jail can mount the tmpfs file system");
4297 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4298 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4299 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4300 "Processes in jail can mount the zfs file system");
4303 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4308 pr = req->td->td_ucred->cr_prison;
4309 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4310 error = sysctl_handle_int(oidp, &level, 0, req);
4311 if (error || !req->newptr)
4313 *(int *)arg1 = level;
4317 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4318 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4319 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4320 sysctl_jail_default_level, "I",
4321 "Processes in jail cannot see all mounted file systems");
4323 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4324 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4325 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4326 sysctl_jail_default_level, "I",
4327 "Ruleset for the devfs filesystem in jail");
4330 * Nodes to describe jail parameters. Maximum length of string parameters
4331 * is returned in the string itself, and the other parameters exist merely
4332 * to make themselves and their types known.
4334 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4338 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4345 switch (oidp->oid_kind & CTLTYPE)
4351 if (!(req->flags & SCTL_MASK32))
4353 return (SYSCTL_OUT(req, &l, sizeof(l)));
4357 return (SYSCTL_OUT(req, &i, sizeof(i)));
4358 case CTLTYPE_STRING:
4359 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4361 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4362 case CTLTYPE_STRUCT:
4364 return (SYSCTL_OUT(req, &s, sizeof(s)));
4370 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4371 * jail creation time but cannot be changed in an existing jail.
4373 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4374 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4375 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4376 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4377 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4378 "I", "Jail secure level");
4379 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4380 "Jail value for kern.osreldate and uname -K");
4381 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4382 "Jail value for kern.osrelease and uname -r");
4383 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4384 "I", "Jail cannot see all mounted file systems");
4385 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4386 "I", "Ruleset for in-jail devfs mounts");
4387 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4388 "B", "Jail persistence");
4390 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4391 "E,jailsys", "Virtual network stack");
4393 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4394 "B", "Jail is in the process of shutting down");
4396 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4397 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4398 "I", "Current number of child jails");
4399 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4400 "I", "Maximum number of child jails");
4402 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4403 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4405 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4406 "Jail NIS domainname");
4407 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4409 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4410 "LU", "Jail host ID");
4412 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4413 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4416 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4417 "Jail IPv4 address virtualization");
4418 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4419 "S,in_addr,a", "Jail IPv4 addresses");
4420 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4421 "B", "Do (not) use IPv4 source address selection rather than the "
4422 "primary jail IPv4 address.");
4425 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4426 "Jail IPv6 address virtualization");
4427 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4428 "S,in6_addr,a", "Jail IPv6 addresses");
4429 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4430 "B", "Do (not) use IPv6 source address selection rather than the "
4431 "primary jail IPv6 address.");
4434 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4435 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4436 "B", "Jail may set hostname");
4437 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4438 "B", "Jail may use SYSV IPC");
4439 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4440 "B", "Jail may create raw sockets");
4441 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4442 "B", "Jail may alter system file flags");
4443 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4444 "B", "Jail may set file quotas");
4445 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4446 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4448 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4449 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4450 "B", "Jail may mount/unmount jail-friendly file systems in general");
4451 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4452 "B", "Jail may mount the devfs file system");
4453 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4454 "B", "Jail may mount the fdescfs file system");
4455 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4456 "B", "Jail may mount the nullfs file system");
4457 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4458 "B", "Jail may mount the procfs file system");
4459 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4460 "B", "Jail may mount the tmpfs file system");
4461 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4462 "B", "Jail may mount the zfs file system");
4466 prison_racct_foreach(void (*callback)(struct racct *racct,
4467 void *arg2, void *arg3), void *arg2, void *arg3)
4469 struct prison_racct *prr;
4471 ASSERT_RACCT_ENABLED();
4473 sx_slock(&allprison_lock);
4474 LIST_FOREACH(prr, &allprison_racct, prr_next)
4475 (callback)(prr->prr_racct, arg2, arg3);
4476 sx_sunlock(&allprison_lock);
4479 static struct prison_racct *
4480 prison_racct_find_locked(const char *name)
4482 struct prison_racct *prr;
4484 ASSERT_RACCT_ENABLED();
4485 sx_assert(&allprison_lock, SA_XLOCKED);
4487 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4490 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4491 if (strcmp(name, prr->prr_name) != 0)
4494 /* Found prison_racct with a matching name? */
4495 prison_racct_hold(prr);
4499 /* Add new prison_racct. */
4500 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4501 racct_create(&prr->prr_racct);
4503 strcpy(prr->prr_name, name);
4504 refcount_init(&prr->prr_refcount, 1);
4505 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4510 struct prison_racct *
4511 prison_racct_find(const char *name)
4513 struct prison_racct *prr;
4515 ASSERT_RACCT_ENABLED();
4517 sx_xlock(&allprison_lock);
4518 prr = prison_racct_find_locked(name);
4519 sx_xunlock(&allprison_lock);
4524 prison_racct_hold(struct prison_racct *prr)
4527 ASSERT_RACCT_ENABLED();
4529 refcount_acquire(&prr->prr_refcount);
4533 prison_racct_free_locked(struct prison_racct *prr)
4536 ASSERT_RACCT_ENABLED();
4537 sx_assert(&allprison_lock, SA_XLOCKED);
4539 if (refcount_release(&prr->prr_refcount)) {
4540 racct_destroy(&prr->prr_racct);
4541 LIST_REMOVE(prr, prr_next);
4542 free(prr, M_PRISON_RACCT);
4547 prison_racct_free(struct prison_racct *prr)
4551 ASSERT_RACCT_ENABLED();
4552 sx_assert(&allprison_lock, SA_UNLOCKED);
4554 old = prr->prr_refcount;
4555 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4558 sx_xlock(&allprison_lock);
4559 prison_racct_free_locked(prr);
4560 sx_xunlock(&allprison_lock);
4564 prison_racct_attach(struct prison *pr)
4566 struct prison_racct *prr;
4568 ASSERT_RACCT_ENABLED();
4569 sx_assert(&allprison_lock, SA_XLOCKED);
4571 prr = prison_racct_find_locked(pr->pr_name);
4572 KASSERT(prr != NULL, ("cannot find prison_racct"));
4574 pr->pr_prison_racct = prr;
4578 * Handle jail renaming. From the racct point of view, renaming means
4579 * moving from one prison_racct to another.
4582 prison_racct_modify(struct prison *pr)
4586 struct prison_racct *oldprr;
4588 ASSERT_RACCT_ENABLED();
4590 sx_slock(&allproc_lock);
4591 sx_xlock(&allprison_lock);
4593 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4594 sx_xunlock(&allprison_lock);
4595 sx_sunlock(&allproc_lock);
4599 oldprr = pr->pr_prison_racct;
4600 pr->pr_prison_racct = NULL;
4602 prison_racct_attach(pr);
4605 * Move resource utilisation records.
4607 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4610 * Force rctl to reattach rules to processes.
4612 FOREACH_PROC_IN_SYSTEM(p) {
4614 cred = crhold(p->p_ucred);
4616 racct_proc_ucred_changed(p, cred, cred);
4620 sx_sunlock(&allproc_lock);
4621 prison_racct_free_locked(oldprr);
4622 sx_xunlock(&allprison_lock);
4626 prison_racct_detach(struct prison *pr)
4629 ASSERT_RACCT_ENABLED();
4630 sx_assert(&allprison_lock, SA_UNLOCKED);
4632 if (pr->pr_prison_racct == NULL)
4634 prison_racct_free(pr->pr_prison_racct);
4635 pr->pr_prison_racct = NULL;
4642 db_show_prison(struct prison *pr)
4645 #if defined(INET) || defined(INET6)
4650 char ip6buf[INET6_ADDRSTRLEN];
4653 db_printf("prison %p:\n", pr);
4654 db_printf(" jid = %d\n", pr->pr_id);
4655 db_printf(" name = %s\n", pr->pr_name);
4656 db_printf(" parent = %p\n", pr->pr_parent);
4657 db_printf(" ref = %d\n", pr->pr_ref);
4658 db_printf(" uref = %d\n", pr->pr_uref);
4659 db_printf(" path = %s\n", pr->pr_path);
4660 db_printf(" cpuset = %d\n", pr->pr_cpuset
4661 ? pr->pr_cpuset->cs_id : -1);
4663 db_printf(" vnet = %p\n", pr->pr_vnet);
4665 db_printf(" root = %p\n", pr->pr_root);
4666 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4667 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4668 db_printf(" children.max = %d\n", pr->pr_childmax);
4669 db_printf(" children.cur = %d\n", pr->pr_childcount);
4670 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4671 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4672 db_printf(" flags = 0x%x", pr->pr_flags);
4673 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4675 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4676 db_printf(" %s", pr_flag_names[fi]);
4677 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4679 jsf = pr->pr_flags &
4680 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4681 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4682 pr_flag_jailsys[fi].disable &&
4683 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4684 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4687 db_printf(" allow = 0x%x", pr->pr_allow);
4688 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4690 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4691 db_printf(" %s", pr_allow_names[fi]);
4693 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4694 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4695 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4696 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4697 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4699 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4700 for (ii = 0; ii < pr->pr_ip4s; ii++)
4701 db_printf(" %s %s\n",
4702 ii == 0 ? "ip4.addr =" : " ",
4703 inet_ntoa(pr->pr_ip4[ii]));
4706 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4707 for (ii = 0; ii < pr->pr_ip6s; ii++)
4708 db_printf(" %s %s\n",
4709 ii == 0 ? "ip6.addr =" : " ",
4710 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4714 DB_SHOW_COMMAND(prison, db_show_prison_command)
4720 * Show all prisons in the list, and prison0 which is not
4723 db_show_prison(&prison0);
4724 if (!db_pager_quit) {
4725 TAILQ_FOREACH(pr, &allprison, pr_list) {
4737 /* Look for a prison with the ID and with references. */
4738 TAILQ_FOREACH(pr, &allprison, pr_list)
4739 if (pr->pr_id == addr && pr->pr_ref > 0)
4742 /* Look again, without requiring a reference. */
4743 TAILQ_FOREACH(pr, &allprison, pr_list)
4744 if (pr->pr_id == addr)
4747 /* Assume address points to a valid prison. */
4748 pr = (struct prison *)addr;