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>
53 #include <sys/sysent.h>
54 #include <sys/namei.h>
55 #include <sys/mount.h>
56 #include <sys/queue.h>
57 #include <sys/socket.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysctl.h>
60 #include <sys/vnode.h>
65 #include <netinet/in.h>
70 #include <netinet6/in6_var.h>
74 #include <security/mac/mac_framework.h>
76 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
78 MALLOC_DEFINE(M_PRISON, "prison", "Prison 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,
103 .pr_childmax = JAIL_MAX,
104 .pr_hostuuid = DEFAULT_HOSTUUID,
105 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
107 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
109 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
111 .pr_allow = PR_ALLOW_ALL,
113 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
115 /* allprison and lastprid are protected by allprison_lock. */
116 struct sx allprison_lock;
117 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
118 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
121 static int do_jail_attach(struct thread *td, struct prison *pr);
122 static void prison_complete(void *context, int pending);
123 static void prison_deref(struct prison *pr, int flags);
124 static char *prison_path(struct prison *pr1, struct prison *pr2);
125 static void prison_remove_one(struct prison *pr);
127 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
128 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
131 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
132 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
135 /* Flags for prison_deref */
136 #define PD_DEREF 0x01
137 #define PD_DEUREF 0x02
138 #define PD_LOCKED 0x04
139 #define PD_LIST_SLOCKED 0x08
140 #define PD_LIST_XLOCKED 0x10
143 * Parameter names corresponding to PR_* flag values. Size values are for kvm
144 * as we cannot figure out the size of a sparse array, or an array without a
147 static char *pr_flag_names[] = {
150 [7] = "ip4.saddrsel",
153 [8] = "ip6.saddrsel",
156 const size_t pr_flag_names_size = sizeof(pr_flag_names);
158 static char *pr_flag_nonames[] = {
161 [7] = "ip4.nosaddrsel",
164 [8] = "ip6.nosaddrsel",
167 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
169 struct jailsys_flags {
173 } pr_flag_jailsys[] = {
174 { "host", 0, PR_HOST },
176 { "vnet", 0, PR_VNET },
179 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
182 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
185 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
187 static char *pr_allow_names[] = {
188 "allow.set_hostname",
196 const size_t pr_allow_names_size = sizeof(pr_allow_names);
198 static char *pr_allow_nonames[] = {
199 "allow.noset_hostname",
201 "allow.noraw_sockets",
207 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
209 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
210 #define JAIL_DEFAULT_ENFORCE_STATFS 2
211 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
212 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
213 #if defined(INET) || defined(INET6)
214 static unsigned jail_max_af_ips = 255;
219 qcmp_v4(const void *ip1, const void *ip2)
224 * We need to compare in HBO here to get the list sorted as expected
225 * by the result of the code. Sorting NBO addresses gives you
226 * interesting results. If you do not understand, do not try.
228 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
229 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
232 * Do not simply return the difference of the two numbers, the int is
246 qcmp_v6(const void *ip1, const void *ip2)
248 const struct in6_addr *ia6a, *ia6b;
251 ia6a = (const struct in6_addr *)ip1;
252 ia6b = (const struct in6_addr *)ip2;
255 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
256 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
258 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
271 jail(struct thread *td, struct jail_args *uap)
277 error = copyin(uap->jail, &version, sizeof(uint32_t));
286 /* FreeBSD single IPv4 jails. */
287 bzero(&j, sizeof(struct jail));
288 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
291 j.version = j0.version;
293 j.hostname = j0.hostname;
294 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
300 * Version 1 was used by multi-IPv4 jail implementations
301 * that never made it into the official kernel.
305 case 2: /* JAIL_API_VERSION */
306 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
307 error = copyin(uap->jail, &j, sizeof(struct jail));
313 /* Sci-Fi jails are not supported, sorry. */
316 return (kern_jail(td, &j));
320 kern_jail(struct thread *td, struct jail *j)
322 struct iovec optiov[2 * (4
323 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
332 char *u_path, *u_hostname, *u_name;
335 struct in_addr *u_ip4;
338 struct in6_addr *u_ip6;
341 int error, enforce_statfs, fi;
343 bzero(&optiov, sizeof(optiov));
344 opt.uio_iov = optiov;
348 opt.uio_segflg = UIO_SYSSPACE;
349 opt.uio_rw = UIO_READ;
352 /* Set permissions for top-level jails from sysctls. */
353 if (!jailed(td->td_ucred)) {
354 for (fi = 0; fi < sizeof(pr_allow_names) /
355 sizeof(pr_allow_names[0]); fi++) {
356 optiov[opt.uio_iovcnt].iov_base =
357 (jail_default_allow & (1 << fi))
358 ? pr_allow_names[fi] : pr_allow_nonames[fi];
359 optiov[opt.uio_iovcnt].iov_len =
360 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
363 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
364 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
366 enforce_statfs = jail_default_enforce_statfs;
367 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
368 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
372 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
374 ip4s = (j->version == 0) ? 1 : j->ip4s;
375 if (ip4s > jail_max_af_ips)
377 tmplen += ip4s * sizeof(struct in_addr);
383 if (j->ip6s > jail_max_af_ips)
385 tmplen += j->ip6s * sizeof(struct in6_addr);
390 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
391 u_hostname = u_path + MAXPATHLEN;
392 u_name = u_hostname + MAXHOSTNAMELEN;
394 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
398 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
400 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
403 optiov[opt.uio_iovcnt].iov_base = "path";
404 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
406 optiov[opt.uio_iovcnt].iov_base = u_path;
407 error = copyinstr(j->path, u_path, MAXPATHLEN,
408 &optiov[opt.uio_iovcnt].iov_len);
410 free(u_path, M_TEMP);
414 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
415 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
417 optiov[opt.uio_iovcnt].iov_base = u_hostname;
418 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
419 &optiov[opt.uio_iovcnt].iov_len);
421 free(u_path, M_TEMP);
425 if (j->jailname != NULL) {
426 optiov[opt.uio_iovcnt].iov_base = "name";
427 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
429 optiov[opt.uio_iovcnt].iov_base = u_name;
430 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
431 &optiov[opt.uio_iovcnt].iov_len);
433 free(u_path, M_TEMP);
439 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
440 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
442 optiov[opt.uio_iovcnt].iov_base = u_ip4;
443 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
445 u_ip4->s_addr = j->ip4s;
447 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
449 free(u_path, M_TEMP);
456 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
457 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
459 optiov[opt.uio_iovcnt].iov_base = u_ip6;
460 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
461 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
463 free(u_path, M_TEMP);
468 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
469 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
470 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
471 free(u_path, M_TEMP);
477 * struct jail_set_args {
478 * struct iovec *iovp;
479 * unsigned int iovcnt;
484 jail_set(struct thread *td, struct jail_set_args *uap)
489 /* Check that we have an even number of iovecs. */
493 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
496 error = kern_jail_set(td, auio, uap->flags);
502 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
509 struct in6_addr *ip6;
512 struct vfsoptlist *opts;
513 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
515 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
516 #if defined(INET) || defined(INET6)
521 size_t namelen, onamelen;
522 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
523 int gotchildmax, gotenforce, gothid, gotslevel;
524 int fi, jid, jsys, len, level;
525 int childmax, slevel, vfslocked;
526 #if defined(INET) || defined(INET6)
535 unsigned pr_flags, ch_flags;
536 unsigned pr_allow, ch_allow, tallow;
539 error = priv_check(td, PRIV_JAIL_SET);
540 if (!error && (flags & JAIL_ATTACH))
541 error = priv_check(td, PRIV_JAIL_ATTACH);
544 mypr = ppr = td->td_ucred->cr_prison;
545 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
547 if (flags & ~JAIL_SET_MASK)
551 * Check all the parameters before committing to anything. Not all
552 * errors can be caught early, but we may as well try. Also, this
553 * takes care of some expensive stuff (path lookup) before getting
554 * the allprison lock.
556 * XXX Jails are not filesystems, and jail parameters are not mount
557 * options. But it makes more sense to re-use the vfsopt code
558 * than duplicate it under a different name.
560 error = vfs_buildopts(optuio, &opts);
570 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
576 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
585 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
593 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
598 else if (enforce < 0 || enforce > 2) {
604 pr_flags = ch_flags = 0;
605 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
607 if (pr_flag_names[fi] == NULL)
609 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
610 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
612 ch_flags |= pr_flags;
613 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
615 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
622 case JAIL_SYS_DISABLE:
623 if (!pr_flag_jailsys[fi].disable) {
627 pr_flags |= pr_flag_jailsys[fi].disable;
630 pr_flags |= pr_flag_jailsys[fi].new;
632 case JAIL_SYS_INHERIT:
639 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
641 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
642 && !(pr_flags & PR_PERSIST)) {
644 vfs_opterror(opts, "new jail must persist or attach");
648 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
650 vfs_opterror(opts, "vnet cannot be changed after creation");
655 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
657 vfs_opterror(opts, "ip4 cannot be changed after creation");
662 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
664 vfs_opterror(opts, "ip6 cannot be changed after creation");
669 pr_allow = ch_allow = 0;
670 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
672 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
673 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
675 ch_allow |= pr_allow;
677 error = vfs_getopt(opts, "name", (void **)&name, &len);
683 if (len == 0 || name[len - 1] != '\0') {
687 if (len > MAXHOSTNAMELEN) {
688 error = ENAMETOOLONG;
693 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
701 if (len == 0 || host[len - 1] != '\0') {
705 if (len > MAXHOSTNAMELEN) {
706 error = ENAMETOOLONG;
711 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
719 if (len == 0 || domain[len - 1] != '\0') {
723 if (len > MAXHOSTNAMELEN) {
724 error = ENAMETOOLONG;
729 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
737 if (len == 0 || uuid[len - 1] != '\0') {
741 if (len > HOSTUUIDLEN) {
742 error = ENAMETOOLONG;
747 #ifdef COMPAT_FREEBSD32
748 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
751 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
755 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
767 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
769 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
772 else if (ip4s & (sizeof(*ip4) - 1)) {
776 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
778 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
780 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
781 ip4s /= sizeof(*ip4);
782 if (ip4s > jail_max_af_ips) {
784 vfs_opterror(opts, "too many IPv4 addresses");
787 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
788 bcopy(op, ip4, ip4s * sizeof(*ip4));
790 * IP addresses are all sorted but ip[0] to preserve
791 * the primary IP address as given from userland.
792 * This special IP is used for unbound outgoing
793 * connections as well for "loopback" traffic in case
794 * source address selection cannot find any more fitting
795 * address to connect from.
798 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
800 * Check for duplicate addresses and do some simple
801 * zero and broadcast checks. If users give other bogus
802 * addresses it is their problem.
804 * We do not have to care about byte order for these
805 * checks so we will do them in NBO.
807 for (ii = 0; ii < ip4s; ii++) {
808 if (ip4[ii].s_addr == INADDR_ANY ||
809 ip4[ii].s_addr == INADDR_BROADCAST) {
814 (ip4[0].s_addr == ip4[ii+1].s_addr ||
815 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
825 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
827 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
830 else if (ip6s & (sizeof(*ip6) - 1)) {
834 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
836 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
838 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
839 ip6s /= sizeof(*ip6);
840 if (ip6s > jail_max_af_ips) {
842 vfs_opterror(opts, "too many IPv6 addresses");
845 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
846 bcopy(op, ip6, ip6s * sizeof(*ip6));
848 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
849 for (ii = 0; ii < ip6s; ii++) {
850 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
855 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
856 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
866 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
867 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
870 "vnet jails cannot have IP address restrictions");
876 error = vfs_getopt(opts, "path", (void **)&path, &len);
882 if (flags & JAIL_UPDATE) {
885 "path cannot be changed after creation");
888 if (len == 0 || path[len - 1] != '\0') {
892 if (len < 2 || (len == 2 && path[0] == '/'))
895 /* Leave room for a real-root full pathname. */
896 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
897 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
898 error = ENAMETOOLONG;
901 NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
906 vfslocked = NDHASGIANT(&nd);
908 NDFREE(&nd, NDF_ONLY_PNBUF);
909 if (root->v_type != VDIR) {
912 VFS_UNLOCK_GIANT(vfslocked);
915 VFS_UNLOCK_GIANT(vfslocked);
920 * Grab the allprison lock before letting modules check their
921 * parameters. Once we have it, do not let go so we'll have a
922 * consistent view of the OSD list.
924 sx_xlock(&allprison_lock);
925 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
927 goto done_unlock_list;
929 /* By now, all parameters should have been noted. */
930 TAILQ_FOREACH(opt, opts, link) {
931 if (!opt->seen && strcmp(opt->name, "errmsg")) {
933 vfs_opterror(opts, "unknown parameter: %s", opt->name);
934 goto done_unlock_list;
939 * See if we are creating a new record or updating an existing one.
940 * This abuses the file error codes ENOENT and EEXIST.
942 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
945 vfs_opterror(opts, "no valid operation (create or update)");
946 goto done_unlock_list;
950 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
951 namelc = strrchr(name, '.');
952 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
958 * See if a requested jid already exists. There is an
959 * information leak here if the jid exists but is not within
960 * the caller's jail hierarchy. Jail creators will get EEXIST
961 * even though they cannot see the jail, and CREATE | UPDATE
962 * will return ENOENT which is not normally a valid error.
966 vfs_opterror(opts, "negative jid");
967 goto done_unlock_list;
969 pr = prison_find(jid);
972 /* Create: jid must not exist. */
973 if (cuflags == JAIL_CREATE) {
974 mtx_unlock(&pr->pr_mtx);
976 vfs_opterror(opts, "jail %d already exists",
978 goto done_unlock_list;
980 if (!prison_ischild(mypr, pr)) {
981 mtx_unlock(&pr->pr_mtx);
983 } else if (pr->pr_uref == 0) {
984 if (!(flags & JAIL_DYING)) {
985 mtx_unlock(&pr->pr_mtx);
987 vfs_opterror(opts, "jail %d is dying",
989 goto done_unlock_list;
990 } else if ((flags & JAIL_ATTACH) ||
991 (pr_flags & PR_PERSIST)) {
993 * A dying jail might be resurrected
994 * (via attach or persist), but first
995 * it must determine if another jail
996 * has claimed its name. Accomplish
997 * this by implicitly re-setting the
1001 name = prison_name(mypr, pr);
1006 /* Update: jid must exist. */
1007 if (cuflags == JAIL_UPDATE) {
1009 vfs_opterror(opts, "jail %d not found", jid);
1010 goto done_unlock_list;
1015 * If the caller provided a name, look for a jail by that name.
1016 * This has different semantics for creates and updates keyed by jid
1017 * (where the name must not already exist in a different jail),
1018 * and updates keyed by the name itself (where the name must exist
1019 * because that is the jail being updated).
1022 namelc = strrchr(name, '.');
1027 * This is a hierarchical name. Split it into the
1028 * parent and child names, and make sure the parent
1029 * exists or matches an already found jail.
1033 if (strncmp(name, ppr->pr_name, namelc - name)
1034 || ppr->pr_name[namelc - name] != '\0') {
1035 mtx_unlock(&pr->pr_mtx);
1038 "cannot change jail's parent");
1039 goto done_unlock_list;
1042 ppr = prison_find_name(mypr, name);
1046 "jail \"%s\" not found", name);
1047 goto done_unlock_list;
1049 mtx_unlock(&ppr->pr_mtx);
1053 if (name[0] != '\0') {
1055 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1058 FOREACH_PRISON_CHILD(ppr, tpr) {
1059 if (tpr != pr && tpr->pr_ref > 0 &&
1060 !strcmp(tpr->pr_name + namelen, name)) {
1062 cuflags != JAIL_CREATE) {
1063 mtx_lock(&tpr->pr_mtx);
1064 if (tpr->pr_ref > 0) {
1069 if (tpr->pr_uref > 0) {
1075 mtx_unlock(&tpr->pr_mtx);
1076 } else if (tpr->pr_uref > 0) {
1078 * Create, or update(jid):
1079 * name must not exist in an
1080 * active sibling jail.
1084 mtx_unlock(&pr->pr_mtx);
1086 "jail \"%s\" already exists",
1088 goto done_unlock_list;
1092 /* If no active jail is found, use a dying one. */
1093 if (deadpr != NULL && pr == NULL) {
1094 if (flags & JAIL_DYING) {
1095 mtx_lock(&deadpr->pr_mtx);
1096 if (deadpr->pr_ref == 0) {
1097 mtx_unlock(&deadpr->pr_mtx);
1101 } else if (cuflags == JAIL_UPDATE) {
1104 "jail \"%s\" is dying", name);
1105 goto done_unlock_list;
1108 /* Update: name must exist if no jid. */
1109 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1111 vfs_opterror(opts, "jail \"%s\" not found",
1113 goto done_unlock_list;
1117 /* Update: must provide a jid or name. */
1118 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1120 vfs_opterror(opts, "update specified no jail");
1121 goto done_unlock_list;
1124 /* If there's no prison to update, create a new one and link it in. */
1126 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1127 if (tpr->pr_childcount >= tpr->pr_childmax) {
1129 vfs_opterror(opts, "prison limit exceeded");
1130 goto done_unlock_list;
1133 mtx_lock(&ppr->pr_mtx);
1134 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1135 mtx_unlock(&ppr->pr_mtx);
1137 vfs_opterror(opts, "parent jail went away!");
1138 goto done_unlock_list;
1142 mtx_unlock(&ppr->pr_mtx);
1143 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1145 /* Find the next free jid. */
1148 if (jid == JAIL_MAX)
1150 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1151 if (tpr->pr_id < jid)
1153 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1154 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1157 if (jid == lastprid) {
1160 "no available jail IDs");
1162 prison_deref(ppr, PD_DEREF |
1163 PD_DEUREF | PD_LIST_XLOCKED);
1172 * The jail already has a jid (that did not yet exist),
1173 * so just find where to insert it.
1175 TAILQ_FOREACH(tpr, &allprison, pr_list)
1176 if (tpr->pr_id >= jid) {
1177 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1182 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1183 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1184 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1185 tpr->pr_childcount++;
1187 pr->pr_parent = ppr;
1190 /* Set some default values, and inherit some from the parent. */
1195 root = mypr->pr_root;
1198 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1199 pr->pr_flags |= PR_HOST;
1200 #if defined(INET) || defined(INET6)
1202 if (!(pr_flags & PR_VNET))
1206 if (!(ch_flags & PR_IP4_USER))
1208 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1209 else if (!(pr_flags & PR_IP4_USER)) {
1210 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1211 if (ppr->pr_ip4 != NULL) {
1212 pr->pr_ip4s = ppr->pr_ip4s;
1213 pr->pr_ip4 = malloc(pr->pr_ip4s *
1214 sizeof(struct in_addr), M_PRISON,
1216 bcopy(ppr->pr_ip4, pr->pr_ip4,
1217 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1222 if (!(ch_flags & PR_IP6_USER))
1224 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1225 else if (!(pr_flags & PR_IP6_USER)) {
1226 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1227 if (ppr->pr_ip6 != NULL) {
1228 pr->pr_ip6s = ppr->pr_ip6s;
1229 pr->pr_ip6 = malloc(pr->pr_ip6s *
1230 sizeof(struct in6_addr), M_PRISON,
1232 bcopy(ppr->pr_ip6, pr->pr_ip6,
1233 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1239 /* Source address selection is always on by default. */
1240 pr->pr_flags |= _PR_IP_SADDRSEL;
1242 pr->pr_securelevel = ppr->pr_securelevel;
1243 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1244 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1246 LIST_INIT(&pr->pr_children);
1247 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1250 /* Allocate a new vnet if specified. */
1251 pr->pr_vnet = (pr_flags & PR_VNET)
1252 ? vnet_alloc() : ppr->pr_vnet;
1255 * Allocate a dedicated cpuset for each jail.
1256 * Unlike other initial settings, this may return an erorr.
1258 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1260 prison_deref(pr, PD_LIST_XLOCKED);
1264 mtx_lock(&pr->pr_mtx);
1266 * New prisons do not yet have a reference, because we do not
1267 * want other to see the incomplete prison once the
1268 * allprison_lock is downgraded.
1273 * Grab a reference for existing prisons, to ensure they
1274 * continue to exist for the duration of the call.
1277 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1278 if ((pr->pr_flags & PR_VNET) &&
1279 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1282 "vnet jails cannot have IP address restrictions");
1283 goto done_deref_locked;
1287 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1290 "ip4 cannot be changed after creation");
1291 goto done_deref_locked;
1295 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1298 "ip6 cannot be changed after creation");
1299 goto done_deref_locked;
1304 /* Do final error checking before setting anything. */
1306 if (slevel < ppr->pr_securelevel) {
1308 goto done_deref_locked;
1312 if (childmax >= ppr->pr_childmax) {
1314 goto done_deref_locked;
1318 if (enforce < ppr->pr_enforce_statfs) {
1320 goto done_deref_locked;
1325 if (ppr->pr_flags & PR_IP4) {
1327 * Make sure the new set of IP addresses is a
1328 * subset of the parent's list. Don't worry
1329 * about the parent being unlocked, as any
1330 * setting is done with allprison_lock held.
1332 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1333 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1335 if (ij == ppr->pr_ip4s) {
1337 goto done_deref_locked;
1340 for (ii = ij = 1; ii < ip4s; ii++) {
1341 if (ip4[ii].s_addr ==
1342 ppr->pr_ip4[0].s_addr)
1344 for (; ij < ppr->pr_ip4s; ij++)
1345 if (ip4[ii].s_addr ==
1346 ppr->pr_ip4[ij].s_addr)
1348 if (ij == ppr->pr_ip4s)
1351 if (ij == ppr->pr_ip4s) {
1353 goto done_deref_locked;
1358 * Check for conflicting IP addresses. We permit them
1359 * if there is no more than one IP on each jail. If
1360 * there is a duplicate on a jail with more than one
1361 * IP stop checking and return error.
1365 for (; tppr != &prison0; tppr = tppr->pr_parent)
1366 if (tppr->pr_flags & PR_VNET)
1369 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1372 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1374 tpr->pr_uref == 0) {
1378 if (!(tpr->pr_flags & PR_IP4_USER))
1381 if (tpr->pr_ip4 == NULL ||
1382 (ip4s == 1 && tpr->pr_ip4s == 1))
1384 for (ii = 0; ii < ip4s; ii++) {
1385 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1388 "IPv4 addresses clash");
1389 goto done_deref_locked;
1397 if (ppr->pr_flags & PR_IP6) {
1399 * Make sure the new set of IP addresses is a
1400 * subset of the parent's list.
1402 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1403 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1406 if (ij == ppr->pr_ip6s) {
1408 goto done_deref_locked;
1411 for (ii = ij = 1; ii < ip6s; ii++) {
1412 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1415 for (; ij < ppr->pr_ip6s; ij++)
1416 if (IN6_ARE_ADDR_EQUAL(
1417 &ip6[ii], &ppr->pr_ip6[ij]))
1419 if (ij == ppr->pr_ip6s)
1422 if (ij == ppr->pr_ip6s) {
1424 goto done_deref_locked;
1428 /* Check for conflicting IP addresses. */
1431 for (; tppr != &prison0; tppr = tppr->pr_parent)
1432 if (tppr->pr_flags & PR_VNET)
1435 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1438 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1440 tpr->pr_uref == 0) {
1444 if (!(tpr->pr_flags & PR_IP6_USER))
1447 if (tpr->pr_ip6 == NULL ||
1448 (ip6s == 1 && tpr->pr_ip6s == 1))
1450 for (ii = 0; ii < ip6s; ii++) {
1451 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1454 "IPv6 addresses clash");
1455 goto done_deref_locked;
1461 onamelen = namelen = 0;
1463 /* Give a default name of the jid. */
1464 if (name[0] == '\0')
1465 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1466 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1470 "name cannot be numeric (unless it is the jid)");
1471 goto done_deref_locked;
1474 * Make sure the name isn't too long for the prison or its
1477 onamelen = strlen(pr->pr_name);
1478 namelen = strlen(name);
1479 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1480 error = ENAMETOOLONG;
1481 goto done_deref_locked;
1483 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1484 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1485 sizeof(pr->pr_name)) {
1486 error = ENAMETOOLONG;
1487 goto done_deref_locked;
1491 if (pr_allow & ~ppr->pr_allow) {
1493 goto done_deref_locked;
1496 /* Set the parameters of the prison. */
1499 if (pr_flags & PR_IP4_USER) {
1500 pr->pr_flags |= PR_IP4;
1501 free(pr->pr_ip4, M_PRISON);
1505 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1507 if (tpr->pr_flags & PR_VNET) {
1512 if (prison_restrict_ip4(tpr, NULL)) {
1521 if (pr_flags & PR_IP6_USER) {
1522 pr->pr_flags |= PR_IP6;
1523 free(pr->pr_ip6, M_PRISON);
1527 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1529 if (tpr->pr_flags & PR_VNET) {
1534 if (prison_restrict_ip6(tpr, NULL)) {
1542 pr->pr_securelevel = slevel;
1543 /* Set all child jails to be at least this level. */
1544 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1545 if (tpr->pr_securelevel < slevel)
1546 tpr->pr_securelevel = slevel;
1549 pr->pr_childmax = childmax;
1550 /* Set all child jails to under this limit. */
1551 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1552 if (tpr->pr_childmax > childmax - level)
1553 tpr->pr_childmax = childmax > level
1554 ? childmax - level : 0;
1557 pr->pr_enforce_statfs = enforce;
1558 /* Pass this restriction on to the children. */
1559 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1560 if (tpr->pr_enforce_statfs < enforce)
1561 tpr->pr_enforce_statfs = enforce;
1564 if (ppr == &prison0)
1565 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1567 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1568 ppr->pr_name, name);
1569 /* Change this component of child names. */
1570 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1571 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1572 strlen(tpr->pr_name + onamelen) + 1);
1573 bcopy(pr->pr_name, tpr->pr_name, namelen);
1577 /* Try to keep a real-rooted full pathname. */
1578 if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
1579 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1580 mypr->pr_path, path);
1582 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1585 if (PR_HOST & ch_flags & ~pr_flags) {
1586 if (pr->pr_flags & PR_HOST) {
1588 * Copy the parent's host info. As with pr_ip4 above,
1589 * the lack of a lock on the parent is not a problem;
1590 * it is always set with allprison_lock at least
1591 * shared, and is held exclusively here.
1593 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1594 sizeof(pr->pr_hostname));
1595 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1596 sizeof(pr->pr_domainname));
1597 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1598 sizeof(pr->pr_hostuuid));
1599 pr->pr_hostid = pr->pr_parent->pr_hostid;
1601 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1602 /* Set this prison, and any descendants without PR_HOST. */
1604 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1606 strlcpy(pr->pr_domainname, domain,
1607 sizeof(pr->pr_domainname));
1609 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1611 pr->pr_hostid = hid;
1612 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1613 if (tpr->pr_flags & PR_HOST)
1617 strlcpy(tpr->pr_hostname,
1619 sizeof(tpr->pr_hostname));
1621 strlcpy(tpr->pr_domainname,
1623 sizeof(tpr->pr_domainname));
1625 strlcpy(tpr->pr_hostuuid,
1627 sizeof(tpr->pr_hostuuid));
1629 tpr->pr_hostid = hid;
1633 if ((tallow = ch_allow & ~pr_allow)) {
1634 /* Clear allow bits in all children. */
1635 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1636 tpr->pr_allow &= ~tallow;
1638 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1640 * Persistent prisons get an extra reference, and prisons losing their
1641 * persist flag lose that reference. Only do this for existing prisons
1642 * for now, so new ones will remain unseen until after the module
1643 * handlers have completed.
1645 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1646 if (pr_flags & PR_PERSIST) {
1654 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1655 mtx_unlock(&pr->pr_mtx);
1657 /* Locks may have prevented a complete restriction of child IP
1658 * addresses. If so, allocate some more memory and try again.
1663 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1664 mtx_lock(&pr->pr_mtx);
1666 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1668 if (tpr->pr_flags & PR_VNET) {
1673 if (prison_restrict_ip4(tpr, ip4)) {
1680 mtx_unlock(&pr->pr_mtx);
1686 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1687 mtx_lock(&pr->pr_mtx);
1689 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1691 if (tpr->pr_flags & PR_VNET) {
1696 if (prison_restrict_ip6(tpr, ip6)) {
1703 mtx_unlock(&pr->pr_mtx);
1707 /* Let the modules do their work. */
1708 sx_downgrade(&allprison_lock);
1710 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1712 prison_deref(pr, PD_LIST_SLOCKED);
1716 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1718 prison_deref(pr, created
1720 : PD_DEREF | PD_LIST_SLOCKED);
1724 /* Attach this process to the prison if requested. */
1725 if (flags & JAIL_ATTACH) {
1726 mtx_lock(&pr->pr_mtx);
1727 error = do_jail_attach(td, pr);
1729 vfs_opterror(opts, "attach failed");
1731 prison_deref(pr, PD_DEREF);
1737 * Now that it is all there, drop the temporary reference from existing
1738 * prisons. Or add a reference to newly created persistent prisons
1739 * (which was not done earlier so that the prison would not be publicly
1743 prison_deref(pr, (flags & JAIL_ATTACH)
1745 : PD_DEREF | PD_LIST_SLOCKED);
1747 if (pr_flags & PR_PERSIST) {
1748 mtx_lock(&pr->pr_mtx);
1751 mtx_unlock(&pr->pr_mtx);
1753 if (!(flags & JAIL_ATTACH))
1754 sx_sunlock(&allprison_lock);
1756 td->td_retval[0] = pr->pr_id;
1760 prison_deref(pr, created
1761 ? PD_LOCKED | PD_LIST_XLOCKED
1762 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1765 sx_xunlock(&allprison_lock);
1768 vfslocked = VFS_LOCK_GIANT(root->v_mount);
1770 VFS_UNLOCK_GIANT(vfslocked);
1774 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1775 if (errmsg_len > 0) {
1776 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1777 if (errmsg_pos > 0) {
1778 if (optuio->uio_segflg == UIO_SYSSPACE)
1780 optuio->uio_iov[errmsg_pos].iov_base,
1784 optuio->uio_iov[errmsg_pos].iov_base,
1791 free(ip4, M_PRISON);
1794 free(ip6, M_PRISON);
1802 * struct jail_get_args {
1803 * struct iovec *iovp;
1804 * unsigned int iovcnt;
1809 jail_get(struct thread *td, struct jail_get_args *uap)
1814 /* Check that we have an even number of iovecs. */
1815 if (uap->iovcnt & 1)
1818 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1821 error = kern_jail_get(td, auio, uap->flags);
1823 error = copyout(auio->uio_iov, uap->iovp,
1824 uap->iovcnt * sizeof (struct iovec));
1830 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1832 struct prison *pr, *mypr;
1834 struct vfsoptlist *opts;
1835 char *errmsg, *name;
1836 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1838 if (flags & ~JAIL_GET_MASK)
1841 /* Get the parameter list. */
1842 error = vfs_buildopts(optuio, &opts);
1845 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1846 mypr = td->td_ucred->cr_prison;
1849 * Find the prison specified by one of: lastjid, jid, name.
1851 sx_slock(&allprison_lock);
1852 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1854 TAILQ_FOREACH(pr, &allprison, pr_list) {
1855 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1856 mtx_lock(&pr->pr_mtx);
1857 if (pr->pr_ref > 0 &&
1858 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1860 mtx_unlock(&pr->pr_mtx);
1866 vfs_opterror(opts, "no jail after %d", jid);
1867 goto done_unlock_list;
1868 } else if (error != ENOENT)
1869 goto done_unlock_list;
1871 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1874 pr = prison_find_child(mypr, jid);
1876 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1877 mtx_unlock(&pr->pr_mtx);
1879 vfs_opterror(opts, "jail %d is dying",
1881 goto done_unlock_list;
1886 vfs_opterror(opts, "jail %d not found", jid);
1887 goto done_unlock_list;
1889 } else if (error != ENOENT)
1890 goto done_unlock_list;
1892 error = vfs_getopt(opts, "name", (void **)&name, &len);
1894 if (len == 0 || name[len - 1] != '\0') {
1896 goto done_unlock_list;
1898 pr = prison_find_name(mypr, name);
1900 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1901 mtx_unlock(&pr->pr_mtx);
1903 vfs_opterror(opts, "jail \"%s\" is dying",
1905 goto done_unlock_list;
1910 vfs_opterror(opts, "jail \"%s\" not found", name);
1911 goto done_unlock_list;
1912 } else if (error != ENOENT)
1913 goto done_unlock_list;
1915 vfs_opterror(opts, "no jail specified");
1917 goto done_unlock_list;
1920 /* Get the parameters of the prison. */
1923 td->td_retval[0] = pr->pr_id;
1924 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
1925 if (error != 0 && error != ENOENT)
1927 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
1928 error = vfs_setopt(opts, "parent", &i, sizeof(i));
1929 if (error != 0 && error != ENOENT)
1931 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
1932 if (error != 0 && error != ENOENT)
1934 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
1935 sizeof(pr->pr_cpuset->cs_id));
1936 if (error != 0 && error != ENOENT)
1938 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
1939 if (error != 0 && error != ENOENT)
1942 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
1943 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1944 if (error != 0 && error != ENOENT)
1948 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
1949 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1950 if (error != 0 && error != ENOENT)
1953 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
1954 sizeof(pr->pr_securelevel));
1955 if (error != 0 && error != ENOENT)
1957 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
1958 sizeof(pr->pr_childcount));
1959 if (error != 0 && error != ENOENT)
1961 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
1962 sizeof(pr->pr_childmax));
1963 if (error != 0 && error != ENOENT)
1965 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
1966 if (error != 0 && error != ENOENT)
1968 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
1969 if (error != 0 && error != ENOENT)
1971 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
1972 if (error != 0 && error != ENOENT)
1974 #ifdef COMPAT_FREEBSD32
1975 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
1976 uint32_t hid32 = pr->pr_hostid;
1978 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
1981 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
1982 sizeof(pr->pr_hostid));
1983 if (error != 0 && error != ENOENT)
1985 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
1986 sizeof(pr->pr_enforce_statfs));
1987 if (error != 0 && error != ENOENT)
1989 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
1991 if (pr_flag_names[fi] == NULL)
1993 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
1994 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
1995 if (error != 0 && error != ENOENT)
1998 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
1999 if (error != 0 && error != ENOENT)
2002 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2005 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2006 i = pr_flag_jailsys[fi].disable &&
2007 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2008 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2011 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2012 if (error != 0 && error != ENOENT)
2015 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2017 if (pr_allow_names[fi] == NULL)
2019 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2020 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2021 if (error != 0 && error != ENOENT)
2024 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2025 if (error != 0 && error != ENOENT)
2028 i = (pr->pr_uref == 0);
2029 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2030 if (error != 0 && error != ENOENT)
2033 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2034 if (error != 0 && error != ENOENT)
2037 /* Get the module parameters. */
2038 mtx_unlock(&pr->pr_mtx);
2040 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2043 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2045 /* By now, all parameters should have been noted. */
2046 TAILQ_FOREACH(opt, opts, link) {
2047 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2049 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2054 /* Write the fetched parameters back to userspace. */
2056 TAILQ_FOREACH(opt, opts, link) {
2057 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2058 pos = 2 * opt->pos + 1;
2059 optuio->uio_iov[pos].iov_len = opt->len;
2060 if (opt->value != NULL) {
2061 if (optuio->uio_segflg == UIO_SYSSPACE) {
2063 optuio->uio_iov[pos].iov_base,
2066 error = copyout(opt->value,
2067 optuio->uio_iov[pos].iov_base,
2078 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2082 sx_sunlock(&allprison_lock);
2084 if (error && errmsg_pos >= 0) {
2085 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2086 errmsg_pos = 2 * errmsg_pos + 1;
2087 if (errmsg_len > 0) {
2088 if (optuio->uio_segflg == UIO_SYSSPACE)
2090 optuio->uio_iov[errmsg_pos].iov_base,
2094 optuio->uio_iov[errmsg_pos].iov_base,
2104 * struct jail_remove_args {
2109 jail_remove(struct thread *td, struct jail_remove_args *uap)
2111 struct prison *pr, *cpr, *lpr, *tpr;
2114 error = priv_check(td, PRIV_JAIL_REMOVE);
2118 sx_xlock(&allprison_lock);
2119 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2121 sx_xunlock(&allprison_lock);
2125 /* Remove all descendants of this prison, then remove this prison. */
2127 pr->pr_flags |= PR_REMOVE;
2128 if (!LIST_EMPTY(&pr->pr_children)) {
2129 mtx_unlock(&pr->pr_mtx);
2131 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2132 mtx_lock(&cpr->pr_mtx);
2133 if (cpr->pr_ref > 0) {
2136 cpr->pr_flags |= PR_REMOVE;
2138 /* Already removed - do not do it again. */
2141 mtx_unlock(&cpr->pr_mtx);
2143 mtx_lock(&lpr->pr_mtx);
2144 prison_remove_one(lpr);
2145 sx_xlock(&allprison_lock);
2150 mtx_lock(&lpr->pr_mtx);
2151 prison_remove_one(lpr);
2152 sx_xlock(&allprison_lock);
2154 mtx_lock(&pr->pr_mtx);
2156 prison_remove_one(pr);
2161 prison_remove_one(struct prison *pr)
2166 /* If the prison was persistent, it is not anymore. */
2168 if (pr->pr_flags & PR_PERSIST) {
2171 pr->pr_flags &= ~PR_PERSIST;
2175 * jail_remove added a reference. If that's the only one, remove
2178 KASSERT(pr->pr_ref > 0,
2179 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2180 if (pr->pr_ref == 1) {
2182 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2186 mtx_unlock(&pr->pr_mtx);
2187 sx_xunlock(&allprison_lock);
2189 * Kill all processes unfortunate enough to be attached to this prison.
2191 sx_slock(&allproc_lock);
2192 LIST_FOREACH(p, &allproc, p_list) {
2194 if (p->p_state != PRS_NEW && p->p_ucred &&
2195 p->p_ucred->cr_prison == pr)
2196 psignal(p, SIGKILL);
2199 sx_sunlock(&allproc_lock);
2200 /* Remove the temporary reference added by jail_remove. */
2201 prison_deref(pr, deuref | PD_DEREF);
2206 * struct jail_attach_args {
2211 jail_attach(struct thread *td, struct jail_attach_args *uap)
2216 error = priv_check(td, PRIV_JAIL_ATTACH);
2220 sx_slock(&allprison_lock);
2221 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2223 sx_sunlock(&allprison_lock);
2228 * Do not allow a process to attach to a prison that is not
2229 * considered to be "alive".
2231 if (pr->pr_uref == 0) {
2232 mtx_unlock(&pr->pr_mtx);
2233 sx_sunlock(&allprison_lock);
2237 return (do_jail_attach(td, pr));
2241 do_jail_attach(struct thread *td, struct prison *pr)
2245 struct ucred *newcred, *oldcred;
2246 int vfslocked, error;
2249 * XXX: Note that there is a slight race here if two threads
2250 * in the same privileged process attempt to attach to two
2251 * different jails at the same time. It is important for
2252 * user processes not to do this, or they might end up with
2253 * a process root from one prison, but attached to the jail
2258 mtx_unlock(&pr->pr_mtx);
2260 /* Let modules do whatever they need to prepare for attaching. */
2261 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2263 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2266 sx_sunlock(&allprison_lock);
2269 * Reparent the newly attached process to this jail.
2271 ppr = td->td_ucred->cr_prison;
2273 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2277 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2278 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2279 if ((error = change_dir(pr->pr_root, td)) != 0)
2282 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2285 VOP_UNLOCK(pr->pr_root, 0);
2286 if ((error = change_root(pr->pr_root, td)))
2287 goto e_unlock_giant;
2288 VFS_UNLOCK_GIANT(vfslocked);
2292 oldcred = p->p_ucred;
2294 crcopy(newcred, oldcred);
2295 newcred->cr_prison = pr;
2296 p->p_ucred = newcred;
2299 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2302 VOP_UNLOCK(pr->pr_root, 0);
2304 VFS_UNLOCK_GIANT(vfslocked);
2306 /* Tell modules this thread is still in its old jail after all. */
2307 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2308 prison_deref(pr, PD_DEREF | PD_DEUREF);
2314 * Returns a locked prison instance, or NULL on failure.
2317 prison_find(int prid)
2321 sx_assert(&allprison_lock, SX_LOCKED);
2322 TAILQ_FOREACH(pr, &allprison, pr_list) {
2323 if (pr->pr_id == prid) {
2324 mtx_lock(&pr->pr_mtx);
2327 mtx_unlock(&pr->pr_mtx);
2334 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2337 prison_find_child(struct prison *mypr, int prid)
2342 sx_assert(&allprison_lock, SX_LOCKED);
2343 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2344 if (pr->pr_id == prid) {
2345 mtx_lock(&pr->pr_mtx);
2348 mtx_unlock(&pr->pr_mtx);
2355 * Look for the name relative to mypr. Returns a locked prison or NULL.
2358 prison_find_name(struct prison *mypr, const char *name)
2360 struct prison *pr, *deadpr;
2364 sx_assert(&allprison_lock, SX_LOCKED);
2365 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2368 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2369 if (!strcmp(pr->pr_name + mylen, name)) {
2370 mtx_lock(&pr->pr_mtx);
2371 if (pr->pr_ref > 0) {
2372 if (pr->pr_uref > 0)
2376 mtx_unlock(&pr->pr_mtx);
2379 /* There was no valid prison - perhaps there was a dying one. */
2380 if (deadpr != NULL) {
2381 mtx_lock(&deadpr->pr_mtx);
2382 if (deadpr->pr_ref == 0) {
2383 mtx_unlock(&deadpr->pr_mtx);
2391 * See if a prison has the specific flag set.
2394 prison_flag(struct ucred *cred, unsigned flag)
2397 /* This is an atomic read, so no locking is necessary. */
2398 return (cred->cr_prison->pr_flags & flag);
2402 prison_allow(struct ucred *cred, unsigned flag)
2405 /* This is an atomic read, so no locking is necessary. */
2406 return (cred->cr_prison->pr_allow & flag);
2410 * Remove a prison reference. If that was the last reference, remove the
2411 * prison itself - but not in this context in case there are locks held.
2414 prison_free_locked(struct prison *pr)
2417 mtx_assert(&pr->pr_mtx, MA_OWNED);
2419 if (pr->pr_ref == 0) {
2420 mtx_unlock(&pr->pr_mtx);
2421 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2422 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2425 mtx_unlock(&pr->pr_mtx);
2429 prison_free(struct prison *pr)
2432 mtx_lock(&pr->pr_mtx);
2433 prison_free_locked(pr);
2437 prison_complete(void *context, int pending)
2440 prison_deref((struct prison *)context, 0);
2444 * Remove a prison reference (usually). This internal version assumes no
2445 * mutexes are held, except perhaps the prison itself. If there are no more
2446 * references, release and delist the prison. On completion, the prison lock
2447 * and the allprison lock are both unlocked.
2450 prison_deref(struct prison *pr, int flags)
2452 struct prison *ppr, *tpr;
2455 if (!(flags & PD_LOCKED))
2456 mtx_lock(&pr->pr_mtx);
2458 if (flags & PD_DEUREF) {
2460 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2462 if (flags & PD_DEREF)
2464 /* If the prison still has references, nothing else to do. */
2465 if (pr->pr_ref > 0) {
2466 mtx_unlock(&pr->pr_mtx);
2467 if (flags & PD_LIST_SLOCKED)
2468 sx_sunlock(&allprison_lock);
2469 else if (flags & PD_LIST_XLOCKED)
2470 sx_xunlock(&allprison_lock);
2474 mtx_unlock(&pr->pr_mtx);
2475 if (flags & PD_LIST_SLOCKED) {
2476 if (!sx_try_upgrade(&allprison_lock)) {
2477 sx_sunlock(&allprison_lock);
2478 sx_xlock(&allprison_lock);
2480 } else if (!(flags & PD_LIST_XLOCKED))
2481 sx_xlock(&allprison_lock);
2483 TAILQ_REMOVE(&allprison, pr, pr_list);
2484 LIST_REMOVE(pr, pr_sibling);
2485 ppr = pr->pr_parent;
2486 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2487 tpr->pr_childcount--;
2488 sx_xunlock(&allprison_lock);
2491 if (pr->pr_vnet != ppr->pr_vnet)
2492 vnet_destroy(pr->pr_vnet);
2494 if (pr->pr_root != NULL) {
2495 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2497 VFS_UNLOCK_GIANT(vfslocked);
2499 mtx_destroy(&pr->pr_mtx);
2501 free(pr->pr_ip4, M_PRISON);
2504 free(pr->pr_ip6, M_PRISON);
2506 if (pr->pr_cpuset != NULL)
2507 cpuset_rel(pr->pr_cpuset);
2511 /* Removing a prison frees a reference on its parent. */
2513 mtx_lock(&pr->pr_mtx);
2514 flags = PD_DEREF | PD_DEUREF;
2519 prison_hold_locked(struct prison *pr)
2522 mtx_assert(&pr->pr_mtx, MA_OWNED);
2523 KASSERT(pr->pr_ref > 0,
2524 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2529 prison_hold(struct prison *pr)
2532 mtx_lock(&pr->pr_mtx);
2533 prison_hold_locked(pr);
2534 mtx_unlock(&pr->pr_mtx);
2538 prison_proc_hold(struct prison *pr)
2541 mtx_lock(&pr->pr_mtx);
2542 KASSERT(pr->pr_uref > 0,
2543 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2545 mtx_unlock(&pr->pr_mtx);
2549 prison_proc_free(struct prison *pr)
2552 mtx_lock(&pr->pr_mtx);
2553 KASSERT(pr->pr_uref > 0,
2554 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2555 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2561 * Restrict a prison's IP address list with its parent's, possibly replacing
2562 * it. Return true if the replacement buffer was used (or would have been).
2565 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2570 ppr = pr->pr_parent;
2571 if (!(pr->pr_flags & PR_IP4_USER)) {
2572 /* This has no user settings, so just copy the parent's list. */
2573 if (pr->pr_ip4s < ppr->pr_ip4s) {
2575 * There's no room for the parent's list. Use the
2576 * new list buffer, which is assumed to be big enough
2577 * (if it was passed). If there's no buffer, try to
2581 if (newip4 == NULL) {
2582 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2583 M_PRISON, M_NOWAIT);
2587 if (newip4 != NULL) {
2588 bcopy(ppr->pr_ip4, newip4,
2589 ppr->pr_ip4s * sizeof(*newip4));
2590 free(pr->pr_ip4, M_PRISON);
2591 pr->pr_ip4 = newip4;
2592 pr->pr_ip4s = ppr->pr_ip4s;
2596 pr->pr_ip4s = ppr->pr_ip4s;
2597 if (pr->pr_ip4s > 0)
2598 bcopy(ppr->pr_ip4, pr->pr_ip4,
2599 pr->pr_ip4s * sizeof(*newip4));
2600 else if (pr->pr_ip4 != NULL) {
2601 free(pr->pr_ip4, M_PRISON);
2604 } else if (pr->pr_ip4s > 0) {
2605 /* Remove addresses that aren't in the parent. */
2606 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2607 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2609 if (ij < ppr->pr_ip4s)
2612 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2613 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2616 for (ij = 1; ii < pr->pr_ip4s; ) {
2617 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2621 switch (ij >= ppr->pr_ip4s ? -1 :
2622 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2624 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2625 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2636 if (pr->pr_ip4s == 0) {
2637 pr->pr_flags |= PR_IP4_DISABLE;
2638 free(pr->pr_ip4, M_PRISON);
2646 * Pass back primary IPv4 address of this jail.
2648 * If not restricted return success but do not alter the address. Caller has
2649 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2651 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2652 * Address returned in NBO.
2655 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2659 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2660 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2662 pr = cred->cr_prison;
2663 if (!(pr->pr_flags & PR_IP4))
2665 mtx_lock(&pr->pr_mtx);
2666 if (!(pr->pr_flags & PR_IP4)) {
2667 mtx_unlock(&pr->pr_mtx);
2670 if (pr->pr_ip4 == NULL) {
2671 mtx_unlock(&pr->pr_mtx);
2672 return (EAFNOSUPPORT);
2675 ia->s_addr = pr->pr_ip4[0].s_addr;
2676 mtx_unlock(&pr->pr_mtx);
2681 * Return 1 if we should do proper source address selection or are not jailed.
2682 * We will return 0 if we should bypass source address selection in favour
2683 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2685 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2688 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2694 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2695 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2700 pr = cred->cr_prison;
2701 if (pr->pr_flags & PR_IP4_SADDRSEL)
2704 lia.s_addr = INADDR_ANY;
2705 error = prison_get_ip4(cred, &lia);
2708 if (lia.s_addr == INADDR_ANY)
2711 ia->s_addr = lia.s_addr;
2716 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2719 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2726 * No need to lock since the PR_IP4_USER flag can't be altered for
2729 while (pr1 != &prison0 &&
2731 !(pr1->pr_flags & PR_VNET) &&
2733 !(pr1->pr_flags & PR_IP4_USER))
2734 pr1 = pr1->pr_parent;
2735 while (pr2 != &prison0 &&
2737 !(pr2->pr_flags & PR_VNET) &&
2739 !(pr2->pr_flags & PR_IP4_USER))
2740 pr2 = pr2->pr_parent;
2741 return (pr1 == pr2);
2745 * Make sure our (source) address is set to something meaningful to this
2748 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2749 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2750 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2753 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2759 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2760 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2762 pr = cred->cr_prison;
2763 if (!(pr->pr_flags & PR_IP4))
2765 mtx_lock(&pr->pr_mtx);
2766 if (!(pr->pr_flags & PR_IP4)) {
2767 mtx_unlock(&pr->pr_mtx);
2770 if (pr->pr_ip4 == NULL) {
2771 mtx_unlock(&pr->pr_mtx);
2772 return (EAFNOSUPPORT);
2775 ia0.s_addr = ntohl(ia->s_addr);
2776 if (ia0.s_addr == INADDR_LOOPBACK) {
2777 ia->s_addr = pr->pr_ip4[0].s_addr;
2778 mtx_unlock(&pr->pr_mtx);
2782 if (ia0.s_addr == INADDR_ANY) {
2784 * In case there is only 1 IPv4 address, bind directly.
2786 if (pr->pr_ip4s == 1)
2787 ia->s_addr = pr->pr_ip4[0].s_addr;
2788 mtx_unlock(&pr->pr_mtx);
2792 error = _prison_check_ip4(pr, ia);
2793 mtx_unlock(&pr->pr_mtx);
2798 * Rewrite destination address in case we will connect to loopback address.
2800 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2801 * Address passed in in NBO and returned in NBO.
2804 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2808 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2809 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2811 pr = cred->cr_prison;
2812 if (!(pr->pr_flags & PR_IP4))
2814 mtx_lock(&pr->pr_mtx);
2815 if (!(pr->pr_flags & PR_IP4)) {
2816 mtx_unlock(&pr->pr_mtx);
2819 if (pr->pr_ip4 == NULL) {
2820 mtx_unlock(&pr->pr_mtx);
2821 return (EAFNOSUPPORT);
2824 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2825 ia->s_addr = pr->pr_ip4[0].s_addr;
2826 mtx_unlock(&pr->pr_mtx);
2831 * Return success because nothing had to be changed.
2833 mtx_unlock(&pr->pr_mtx);
2838 * Check if given address belongs to the jail referenced by cred/prison.
2840 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2841 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2842 * doesn't allow IPv4. Address passed in in NBO.
2845 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2850 * Check the primary IP.
2852 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2856 * All the other IPs are sorted so we can do a binary search.
2859 z = pr->pr_ip4s - 2;
2862 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2871 return (EADDRNOTAVAIL);
2875 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2880 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2881 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2883 pr = cred->cr_prison;
2884 if (!(pr->pr_flags & PR_IP4))
2886 mtx_lock(&pr->pr_mtx);
2887 if (!(pr->pr_flags & PR_IP4)) {
2888 mtx_unlock(&pr->pr_mtx);
2891 if (pr->pr_ip4 == NULL) {
2892 mtx_unlock(&pr->pr_mtx);
2893 return (EAFNOSUPPORT);
2896 error = _prison_check_ip4(pr, ia);
2897 mtx_unlock(&pr->pr_mtx);
2904 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2909 ppr = pr->pr_parent;
2910 if (!(pr->pr_flags & PR_IP6_USER)) {
2911 /* This has no user settings, so just copy the parent's list. */
2912 if (pr->pr_ip6s < ppr->pr_ip6s) {
2914 * There's no room for the parent's list. Use the
2915 * new list buffer, which is assumed to be big enough
2916 * (if it was passed). If there's no buffer, try to
2920 if (newip6 == NULL) {
2921 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
2922 M_PRISON, M_NOWAIT);
2926 if (newip6 != NULL) {
2927 bcopy(ppr->pr_ip6, newip6,
2928 ppr->pr_ip6s * sizeof(*newip6));
2929 free(pr->pr_ip6, M_PRISON);
2930 pr->pr_ip6 = newip6;
2931 pr->pr_ip6s = ppr->pr_ip6s;
2935 pr->pr_ip6s = ppr->pr_ip6s;
2936 if (pr->pr_ip6s > 0)
2937 bcopy(ppr->pr_ip6, pr->pr_ip6,
2938 pr->pr_ip6s * sizeof(*newip6));
2939 else if (pr->pr_ip6 != NULL) {
2940 free(pr->pr_ip6, M_PRISON);
2943 } else if (pr->pr_ip6s > 0) {
2944 /* Remove addresses that aren't in the parent. */
2945 for (ij = 0; ij < ppr->pr_ip6s; ij++)
2946 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
2949 if (ij < ppr->pr_ip6s)
2952 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
2953 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
2956 for (ij = 1; ii < pr->pr_ip6s; ) {
2957 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
2962 switch (ij >= ppr->pr_ip6s ? -1 :
2963 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
2965 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
2966 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
2977 if (pr->pr_ip6s == 0) {
2978 pr->pr_flags |= PR_IP6_DISABLE;
2979 free(pr->pr_ip6, M_PRISON);
2987 * Pass back primary IPv6 address for this jail.
2989 * If not restricted return success but do not alter the address. Caller has
2990 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
2992 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
2995 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
2999 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3000 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3002 pr = cred->cr_prison;
3003 if (!(pr->pr_flags & PR_IP6))
3005 mtx_lock(&pr->pr_mtx);
3006 if (!(pr->pr_flags & PR_IP6)) {
3007 mtx_unlock(&pr->pr_mtx);
3010 if (pr->pr_ip6 == NULL) {
3011 mtx_unlock(&pr->pr_mtx);
3012 return (EAFNOSUPPORT);
3015 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3016 mtx_unlock(&pr->pr_mtx);
3021 * Return 1 if we should do proper source address selection or are not jailed.
3022 * We will return 0 if we should bypass source address selection in favour
3023 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3025 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3028 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3031 struct in6_addr lia6;
3034 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3035 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3040 pr = cred->cr_prison;
3041 if (pr->pr_flags & PR_IP6_SADDRSEL)
3045 error = prison_get_ip6(cred, &lia6);
3048 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3051 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3056 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3059 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3065 while (pr1 != &prison0 &&
3067 !(pr1->pr_flags & PR_VNET) &&
3069 !(pr1->pr_flags & PR_IP6_USER))
3070 pr1 = pr1->pr_parent;
3071 while (pr2 != &prison0 &&
3073 !(pr2->pr_flags & PR_VNET) &&
3075 !(pr2->pr_flags & PR_IP6_USER))
3076 pr2 = pr2->pr_parent;
3077 return (pr1 == pr2);
3081 * Make sure our (source) address is set to something meaningful to this jail.
3083 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3084 * when needed while binding.
3086 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3087 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3088 * doesn't allow IPv6.
3091 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3096 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3097 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3099 pr = cred->cr_prison;
3100 if (!(pr->pr_flags & PR_IP6))
3102 mtx_lock(&pr->pr_mtx);
3103 if (!(pr->pr_flags & PR_IP6)) {
3104 mtx_unlock(&pr->pr_mtx);
3107 if (pr->pr_ip6 == NULL) {
3108 mtx_unlock(&pr->pr_mtx);
3109 return (EAFNOSUPPORT);
3112 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3113 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3114 mtx_unlock(&pr->pr_mtx);
3118 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3120 * In case there is only 1 IPv6 address, and v6only is true,
3121 * then bind directly.
3123 if (v6only != 0 && pr->pr_ip6s == 1)
3124 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3125 mtx_unlock(&pr->pr_mtx);
3129 error = _prison_check_ip6(pr, ia6);
3130 mtx_unlock(&pr->pr_mtx);
3135 * Rewrite destination address in case we will connect to loopback address.
3137 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3140 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3144 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3145 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3147 pr = cred->cr_prison;
3148 if (!(pr->pr_flags & PR_IP6))
3150 mtx_lock(&pr->pr_mtx);
3151 if (!(pr->pr_flags & PR_IP6)) {
3152 mtx_unlock(&pr->pr_mtx);
3155 if (pr->pr_ip6 == NULL) {
3156 mtx_unlock(&pr->pr_mtx);
3157 return (EAFNOSUPPORT);
3160 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3161 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3162 mtx_unlock(&pr->pr_mtx);
3167 * Return success because nothing had to be changed.
3169 mtx_unlock(&pr->pr_mtx);
3174 * Check if given address belongs to the jail referenced by cred/prison.
3176 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3177 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3178 * doesn't allow IPv6.
3181 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3186 * Check the primary IP.
3188 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3192 * All the other IPs are sorted so we can do a binary search.
3195 z = pr->pr_ip6s - 2;
3198 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3207 return (EADDRNOTAVAIL);
3211 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3216 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3217 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3219 pr = cred->cr_prison;
3220 if (!(pr->pr_flags & PR_IP6))
3222 mtx_lock(&pr->pr_mtx);
3223 if (!(pr->pr_flags & PR_IP6)) {
3224 mtx_unlock(&pr->pr_mtx);
3227 if (pr->pr_ip6 == NULL) {
3228 mtx_unlock(&pr->pr_mtx);
3229 return (EAFNOSUPPORT);
3232 error = _prison_check_ip6(pr, ia6);
3233 mtx_unlock(&pr->pr_mtx);
3239 * Check if a jail supports the given address family.
3241 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3245 prison_check_af(struct ucred *cred, int af)
3250 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3252 pr = cred->cr_prison;
3254 /* Prisons with their own network stack are not limited. */
3255 if (prison_owns_vnet(cred))
3264 if (pr->pr_flags & PR_IP4)
3266 mtx_lock(&pr->pr_mtx);
3267 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3268 error = EAFNOSUPPORT;
3269 mtx_unlock(&pr->pr_mtx);
3275 if (pr->pr_flags & PR_IP6)
3277 mtx_lock(&pr->pr_mtx);
3278 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3279 error = EAFNOSUPPORT;
3280 mtx_unlock(&pr->pr_mtx);
3288 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3289 error = EAFNOSUPPORT;
3295 * Check if given address belongs to the jail referenced by cred (wrapper to
3296 * prison_check_ip[46]).
3298 * Returns 0 if jail doesn't restrict the address family or if address belongs
3299 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3300 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3303 prison_if(struct ucred *cred, struct sockaddr *sa)
3306 struct sockaddr_in *sai;
3309 struct sockaddr_in6 *sai6;
3313 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3314 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3317 if (prison_owns_vnet(cred))
3322 switch (sa->sa_family)
3326 sai = (struct sockaddr_in *)sa;
3327 error = prison_check_ip4(cred, &sai->sin_addr);
3332 sai6 = (struct sockaddr_in6 *)sa;
3333 error = prison_check_ip6(cred, &sai6->sin6_addr);
3337 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3338 error = EAFNOSUPPORT;
3344 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3347 prison_check(struct ucred *cred1, struct ucred *cred2)
3350 return ((cred1->cr_prison == cred2->cr_prison ||
3351 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3355 * Return 1 if p2 is a child of p1, otherwise 0.
3358 prison_ischild(struct prison *pr1, struct prison *pr2)
3361 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3368 * Return 1 if the passed credential is in a jail, otherwise 0.
3371 jailed(struct ucred *cred)
3374 return (cred->cr_prison != &prison0);
3378 * Return 1 if the passed credential is in a jail and that jail does not
3379 * have its own virtual network stack, otherwise 0.
3382 jailed_without_vnet(struct ucred *cred)
3388 if (prison_owns_vnet(cred))
3396 * Return the correct hostname (domainname, et al) for the passed credential.
3399 getcredhostname(struct ucred *cred, char *buf, size_t size)
3404 * A NULL credential can be used to shortcut to the physical
3405 * system's hostname.
3407 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3408 mtx_lock(&pr->pr_mtx);
3409 strlcpy(buf, pr->pr_hostname, size);
3410 mtx_unlock(&pr->pr_mtx);
3414 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3417 mtx_lock(&cred->cr_prison->pr_mtx);
3418 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3419 mtx_unlock(&cred->cr_prison->pr_mtx);
3423 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3426 mtx_lock(&cred->cr_prison->pr_mtx);
3427 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3428 mtx_unlock(&cred->cr_prison->pr_mtx);
3432 getcredhostid(struct ucred *cred, unsigned long *hostid)
3435 mtx_lock(&cred->cr_prison->pr_mtx);
3436 *hostid = cred->cr_prison->pr_hostid;
3437 mtx_unlock(&cred->cr_prison->pr_mtx);
3442 * Determine whether the prison represented by cred owns
3443 * its vnet rather than having it inherited.
3445 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3448 prison_owns_vnet(struct ucred *cred)
3452 * vnets cannot be added/removed after jail creation,
3453 * so no need to lock here.
3455 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3460 * Determine whether the subject represented by cred can "see"
3461 * status of a mount point.
3462 * Returns: 0 for permitted, ENOENT otherwise.
3463 * XXX: This function should be called cr_canseemount() and should be
3464 * placed in kern_prot.c.
3467 prison_canseemount(struct ucred *cred, struct mount *mp)
3473 pr = cred->cr_prison;
3474 if (pr->pr_enforce_statfs == 0)
3476 if (pr->pr_root->v_mount == mp)
3478 if (pr->pr_enforce_statfs == 2)
3481 * If jail's chroot directory is set to "/" we should be able to see
3482 * all mount-points from inside a jail.
3483 * This is ugly check, but this is the only situation when jail's
3484 * directory ends with '/'.
3486 if (strcmp(pr->pr_path, "/") == 0)
3488 len = strlen(pr->pr_path);
3490 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3493 * Be sure that we don't have situation where jail's root directory
3494 * is "/some/path" and mount point is "/some/pathpath".
3496 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3502 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3504 char jpath[MAXPATHLEN];
3508 pr = cred->cr_prison;
3509 if (pr->pr_enforce_statfs == 0)
3511 if (prison_canseemount(cred, mp) != 0) {
3512 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3513 strlcpy(sp->f_mntonname, "[restricted]",
3514 sizeof(sp->f_mntonname));
3517 if (pr->pr_root->v_mount == mp) {
3519 * Clear current buffer data, so we are sure nothing from
3520 * the valid path left there.
3522 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3523 *sp->f_mntonname = '/';
3527 * If jail's chroot directory is set to "/" we should be able to see
3528 * all mount-points from inside a jail.
3530 if (strcmp(pr->pr_path, "/") == 0)
3532 len = strlen(pr->pr_path);
3533 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3535 * Clear current buffer data, so we are sure nothing from
3536 * the valid path left there.
3538 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3539 if (*jpath == '\0') {
3540 /* Should never happen. */
3541 *sp->f_mntonname = '/';
3543 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3548 * Check with permission for a specific privilege is granted within jail. We
3549 * have a specific list of accepted privileges; the rest are denied.
3552 prison_priv_check(struct ucred *cred, int priv)
3560 * Privileges specific to prisons with a virtual network stack.
3561 * There might be a duplicate entry here in case the privilege
3562 * is only granted conditionally in the legacy jail case.
3567 * NFS-specific privileges.
3569 case PRIV_NFS_DAEMON:
3570 case PRIV_NFS_LOCKD:
3573 * Network stack privileges.
3575 case PRIV_NET_BRIDGE:
3578 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3579 case PRIV_NET_ROUTE:
3581 case PRIV_NET_SETIFMTU:
3582 case PRIV_NET_SETIFFLAGS:
3583 case PRIV_NET_SETIFCAP:
3584 case PRIV_NET_SETIFDESCR:
3585 case PRIV_NET_SETIFNAME :
3586 case PRIV_NET_SETIFMETRIC:
3587 case PRIV_NET_SETIFPHYS:
3588 case PRIV_NET_SETIFMAC:
3589 case PRIV_NET_ADDMULTI:
3590 case PRIV_NET_DELMULTI:
3591 case PRIV_NET_HWIOCTL:
3592 case PRIV_NET_SETLLADDR:
3593 case PRIV_NET_ADDIFGROUP:
3594 case PRIV_NET_DELIFGROUP:
3595 case PRIV_NET_IFCREATE:
3596 case PRIV_NET_IFDESTROY:
3597 case PRIV_NET_ADDIFADDR:
3598 case PRIV_NET_DELIFADDR:
3601 case PRIV_NET_SETIFVNET:
3602 case PRIV_NET_SETIFFIB:
3605 * 802.11-related privileges.
3607 case PRIV_NET80211_GETKEY:
3609 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3614 * AppleTalk privileges.
3616 case PRIV_NETATALK_RESERVEDPORT:
3621 case PRIV_NETATM_CFG:
3622 case PRIV_NETATM_ADD:
3623 case PRIV_NETATM_DEL:
3624 case PRIV_NETATM_SET:
3627 * Bluetooth privileges.
3629 case PRIV_NETBLUETOOTH_RAW:
3633 * Netgraph and netgraph module privileges.
3635 case PRIV_NETGRAPH_CONTROL:
3637 case PRIV_NETGRAPH_TTY:
3641 * IPv4 and IPv6 privileges.
3643 case PRIV_NETINET_IPFW:
3644 case PRIV_NETINET_DIVERT:
3645 case PRIV_NETINET_PF:
3646 case PRIV_NETINET_DUMMYNET:
3647 case PRIV_NETINET_CARP:
3648 case PRIV_NETINET_MROUTE:
3649 case PRIV_NETINET_RAW:
3650 case PRIV_NETINET_ADDRCTRL6:
3651 case PRIV_NETINET_ND6:
3652 case PRIV_NETINET_SCOPE6:
3653 case PRIV_NETINET_ALIFETIME6:
3654 case PRIV_NETINET_IPSEC:
3655 case PRIV_NETINET_BINDANY:
3659 * IPX/SPX privileges.
3661 case PRIV_NETIPX_RESERVEDPORT:
3662 case PRIV_NETIPX_RAW:
3676 * No default: or deny here.
3677 * In case of no permit fall through to next switch().
3679 if (cred->cr_prison->pr_flags & PR_VNET)
3687 * Allow ktrace privileges for root in jail.
3693 * Allow jailed processes to configure audit identity and
3694 * submit audit records (login, etc). In the future we may
3695 * want to further refine the relationship between audit and
3698 case PRIV_AUDIT_GETAUDIT:
3699 case PRIV_AUDIT_SETAUDIT:
3700 case PRIV_AUDIT_SUBMIT:
3704 * Allow jailed processes to manipulate process UNIX
3705 * credentials in any way they see fit.
3707 case PRIV_CRED_SETUID:
3708 case PRIV_CRED_SETEUID:
3709 case PRIV_CRED_SETGID:
3710 case PRIV_CRED_SETEGID:
3711 case PRIV_CRED_SETGROUPS:
3712 case PRIV_CRED_SETREUID:
3713 case PRIV_CRED_SETREGID:
3714 case PRIV_CRED_SETRESUID:
3715 case PRIV_CRED_SETRESGID:
3718 * Jail implements visibility constraints already, so allow
3719 * jailed root to override uid/gid-based constraints.
3721 case PRIV_SEEOTHERGIDS:
3722 case PRIV_SEEOTHERUIDS:
3725 * Jail implements inter-process debugging limits already, so
3726 * allow jailed root various debugging privileges.
3728 case PRIV_DEBUG_DIFFCRED:
3729 case PRIV_DEBUG_SUGID:
3730 case PRIV_DEBUG_UNPRIV:
3733 * Allow jail to set various resource limits and login
3734 * properties, and for now, exceed process resource limits.
3736 case PRIV_PROC_LIMIT:
3737 case PRIV_PROC_SETLOGIN:
3738 case PRIV_PROC_SETRLIMIT:
3741 * System V and POSIX IPC privileges are granted in jail.
3744 case PRIV_IPC_WRITE:
3745 case PRIV_IPC_ADMIN:
3746 case PRIV_IPC_MSGSIZE:
3750 * Jail operations within a jail work on child jails.
3752 case PRIV_JAIL_ATTACH:
3754 case PRIV_JAIL_REMOVE:
3757 * Jail implements its own inter-process limits, so allow
3758 * root processes in jail to change scheduling on other
3759 * processes in the same jail. Likewise for signalling.
3761 case PRIV_SCHED_DIFFCRED:
3762 case PRIV_SCHED_CPUSET:
3763 case PRIV_SIGNAL_DIFFCRED:
3764 case PRIV_SIGNAL_SUGID:
3767 * Allow jailed processes to write to sysctls marked as jail
3770 case PRIV_SYSCTL_WRITEJAIL:
3773 * Allow root in jail to manage a variety of quota
3774 * properties. These should likely be conditional on a
3775 * configuration option.
3777 case PRIV_VFS_GETQUOTA:
3778 case PRIV_VFS_SETQUOTA:
3781 * Since Jail relies on chroot() to implement file system
3782 * protections, grant many VFS privileges to root in jail.
3783 * Be careful to exclude mount-related and NFS-related
3787 case PRIV_VFS_WRITE:
3788 case PRIV_VFS_ADMIN:
3790 case PRIV_VFS_LOOKUP:
3791 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3792 case PRIV_VFS_CHFLAGS_DEV:
3793 case PRIV_VFS_CHOWN:
3794 case PRIV_VFS_CHROOT:
3795 case PRIV_VFS_RETAINSUGID:
3796 case PRIV_VFS_FCHROOT:
3798 case PRIV_VFS_SETGID:
3800 case PRIV_VFS_STICKYFILE:
3804 * Depending on the global setting, allow privilege of
3805 * setting system flags.
3807 case PRIV_VFS_SYSFLAGS:
3808 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3814 * Depending on the global setting, allow privilege of
3815 * mounting/unmounting file systems.
3817 case PRIV_VFS_MOUNT:
3818 case PRIV_VFS_UNMOUNT:
3819 case PRIV_VFS_MOUNT_NONUSER:
3820 case PRIV_VFS_MOUNT_OWNER:
3821 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT)
3827 * Allow jailed root to bind reserved ports and reuse in-use
3830 case PRIV_NETINET_RESERVEDPORT:
3831 case PRIV_NETINET_REUSEPORT:
3835 * Allow jailed root to set certian IPv4/6 (option) headers.
3837 case PRIV_NETINET_SETHDROPTS:
3841 * Conditionally allow creating raw sockets in jail.
3843 case PRIV_NETINET_RAW:
3844 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3850 * Since jail implements its own visibility limits on netstat
3851 * sysctls, allow getcred. This allows identd to work in
3854 case PRIV_NETINET_GETCRED:
3859 * In all remaining cases, deny the privilege request. This
3860 * includes almost all network privileges, many system
3861 * configuration privileges.
3868 * Return the part of pr2's name that is relative to pr1, or the whole name
3869 * if it does not directly follow.
3873 prison_name(struct prison *pr1, struct prison *pr2)
3877 /* Jails see themselves as "0" (if they see themselves at all). */
3880 name = pr2->pr_name;
3881 if (prison_ischild(pr1, pr2)) {
3883 * pr1 isn't locked (and allprison_lock may not be either)
3884 * so its length can't be counted on. But the number of dots
3885 * can be counted on - and counted.
3887 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3888 name = strchr(name, '.') + 1;
3894 * Return the part of pr2's path that is relative to pr1, or the whole path
3895 * if it does not directly follow.
3898 prison_path(struct prison *pr1, struct prison *pr2)
3900 char *path1, *path2;
3903 path1 = pr1->pr_path;
3904 path2 = pr2->pr_path;
3905 if (!strcmp(path1, "/"))
3907 len1 = strlen(path1);
3908 if (strncmp(path1, path2, len1))
3910 if (path2[len1] == '\0')
3912 if (path2[len1] == '/')
3913 return (path2 + len1);
3919 * Jail-related sysctls.
3921 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3925 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3928 struct prison *pr, *cpr;
3930 struct in_addr *ip4 = NULL;
3934 struct in_addr *ip6 = NULL;
3939 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3940 pr = req->td->td_ucred->cr_prison;
3942 sx_slock(&allprison_lock);
3943 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3944 #if defined(INET) || defined(INET6)
3947 mtx_lock(&cpr->pr_mtx);
3949 if (cpr->pr_ip4s > 0) {
3950 if (ip4s < cpr->pr_ip4s) {
3951 ip4s = cpr->pr_ip4s;
3952 mtx_unlock(&cpr->pr_mtx);
3953 ip4 = realloc(ip4, ip4s *
3954 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3957 bcopy(cpr->pr_ip4, ip4,
3958 cpr->pr_ip4s * sizeof(struct in_addr));
3962 if (cpr->pr_ip6s > 0) {
3963 if (ip6s < cpr->pr_ip6s) {
3964 ip6s = cpr->pr_ip6s;
3965 mtx_unlock(&cpr->pr_mtx);
3966 ip6 = realloc(ip6, ip6s *
3967 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3970 bcopy(cpr->pr_ip6, ip6,
3971 cpr->pr_ip6s * sizeof(struct in6_addr));
3974 if (cpr->pr_ref == 0) {
3975 mtx_unlock(&cpr->pr_mtx);
3978 bzero(xp, sizeof(*xp));
3979 xp->pr_version = XPRISON_VERSION;
3980 xp->pr_id = cpr->pr_id;
3981 xp->pr_state = cpr->pr_uref > 0
3982 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3983 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3984 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3985 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3987 xp->pr_ip4s = cpr->pr_ip4s;
3990 xp->pr_ip6s = cpr->pr_ip6s;
3992 mtx_unlock(&cpr->pr_mtx);
3993 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3997 if (xp->pr_ip4s > 0) {
3998 error = SYSCTL_OUT(req, ip4,
3999 xp->pr_ip4s * sizeof(struct in_addr));
4005 if (xp->pr_ip6s > 0) {
4006 error = SYSCTL_OUT(req, ip6,
4007 xp->pr_ip6s * sizeof(struct in6_addr));
4013 sx_sunlock(&allprison_lock);
4024 SYSCTL_OID(_security_jail, OID_AUTO, list,
4025 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4026 sysctl_jail_list, "S", "List of active jails");
4029 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4033 injail = jailed(req->td->td_ucred);
4034 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4039 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4040 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4041 sysctl_jail_jailed, "I", "Process in jail?");
4043 #if defined(INET) || defined(INET6)
4044 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4045 &jail_max_af_ips, 0,
4046 "Number of IP addresses a jail may have at most per address family");
4050 * Default parameters for jail(2) compatability. For historical reasons,
4051 * the sysctl names have varying similarity to the parameter names. Prisons
4052 * just see their own parameters, and can't change them.
4055 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4058 int allow, error, i;
4060 pr = req->td->td_ucred->cr_prison;
4061 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4063 /* Get the current flag value, and convert it to a boolean. */
4064 i = (allow & arg2) ? 1 : 0;
4067 error = sysctl_handle_int(oidp, &i, 0, req);
4068 if (error || !req->newptr)
4074 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4077 mtx_lock(&prison0.pr_mtx);
4078 jail_default_allow = (jail_default_allow & ~arg2) | i;
4079 mtx_unlock(&prison0.pr_mtx);
4083 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4084 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4085 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4086 "Processes in jail can set their hostnames");
4087 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4088 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4089 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4090 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4091 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4092 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4093 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4094 "Processes in jail can use System V IPC primitives");
4095 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4096 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4097 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4098 "Prison root can create raw sockets");
4099 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4100 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4101 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4102 "Processes in jail can alter system file flags");
4103 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4104 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4105 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4106 "Processes in jail can mount/unmount jail-friendly file systems");
4109 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4114 pr = req->td->td_ucred->cr_prison;
4115 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4116 error = sysctl_handle_int(oidp, &level, 0, req);
4117 if (error || !req->newptr)
4119 *(int *)arg1 = level;
4123 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4124 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4125 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4126 sysctl_jail_default_level, "I",
4127 "Processes in jail cannot see all mounted file systems");
4130 * Nodes to describe jail parameters. Maximum length of string parameters
4131 * is returned in the string itself, and the other parameters exist merely
4132 * to make themselves and their types known.
4134 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4138 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4145 switch (oidp->oid_kind & CTLTYPE)
4151 if (!(req->flags & SCTL_MASK32))
4153 return (SYSCTL_OUT(req, &l, sizeof(l)));
4157 return (SYSCTL_OUT(req, &i, sizeof(i)));
4158 case CTLTYPE_STRING:
4159 snprintf(numbuf, sizeof(numbuf), "%d", arg2);
4161 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4162 case CTLTYPE_STRUCT:
4164 return (SYSCTL_OUT(req, &s, sizeof(s)));
4169 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4170 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4171 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4172 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4173 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4174 "I", "Jail secure level");
4175 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4176 "I", "Jail cannot see all mounted file systems");
4177 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4178 "B", "Jail persistence");
4180 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4181 "E,jailsys", "Virtual network stack");
4183 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4184 "B", "Jail is in the process of shutting down");
4186 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4187 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4188 "I", "Current number of child jails");
4189 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4190 "I", "Maximum number of child jails");
4192 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4193 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4195 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4196 "Jail NIS domainname");
4197 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4199 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4200 "LU", "Jail host ID");
4202 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4203 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4206 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4207 "Jail IPv4 address virtualization");
4208 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4209 "S,in_addr,a", "Jail IPv4 addresses");
4210 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4211 "B", "Do (not) use IPv4 source address selection rather than the "
4212 "primary jail IPv4 address.");
4215 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4216 "Jail IPv6 address virtualization");
4217 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4218 "S,in6_addr,a", "Jail IPv6 addresses");
4219 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4220 "B", "Do (not) use IPv6 source address selection rather than the "
4221 "primary jail IPv6 address.");
4224 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4225 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4226 "B", "Jail may set hostname");
4227 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4228 "B", "Jail may use SYSV IPC");
4229 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4230 "B", "Jail may create raw sockets");
4231 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4232 "B", "Jail may alter system file flags");
4233 SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
4234 "B", "Jail may mount/unmount jail-friendly file systems");
4235 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4236 "B", "Jail may set file quotas");
4237 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4238 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4244 db_show_prison(struct prison *pr)
4247 #if defined(INET) || defined(INET6)
4252 char ip6buf[INET6_ADDRSTRLEN];
4255 db_printf("prison %p:\n", pr);
4256 db_printf(" jid = %d\n", pr->pr_id);
4257 db_printf(" name = %s\n", pr->pr_name);
4258 db_printf(" parent = %p\n", pr->pr_parent);
4259 db_printf(" ref = %d\n", pr->pr_ref);
4260 db_printf(" uref = %d\n", pr->pr_uref);
4261 db_printf(" path = %s\n", pr->pr_path);
4262 db_printf(" cpuset = %d\n", pr->pr_cpuset
4263 ? pr->pr_cpuset->cs_id : -1);
4265 db_printf(" vnet = %p\n", pr->pr_vnet);
4267 db_printf(" root = %p\n", pr->pr_root);
4268 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4269 db_printf(" children.max = %d\n", pr->pr_childmax);
4270 db_printf(" children.cur = %d\n", pr->pr_childcount);
4271 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4272 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4273 db_printf(" flags = 0x%x", pr->pr_flags);
4274 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4276 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4277 db_printf(" %s", pr_flag_names[fi]);
4278 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4280 jsf = pr->pr_flags &
4281 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4282 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4283 pr_flag_jailsys[fi].disable &&
4284 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4285 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4288 db_printf(" allow = 0x%x", pr->pr_allow);
4289 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4291 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4292 db_printf(" %s", pr_allow_names[fi]);
4294 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4295 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4296 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4297 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4298 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4300 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4301 for (ii = 0; ii < pr->pr_ip4s; ii++)
4302 db_printf(" %s %s\n",
4303 ii == 0 ? "ip4.addr =" : " ",
4304 inet_ntoa(pr->pr_ip4[ii]));
4307 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4308 for (ii = 0; ii < pr->pr_ip6s; ii++)
4309 db_printf(" %s %s\n",
4310 ii == 0 ? "ip6.addr =" : " ",
4311 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4315 DB_SHOW_COMMAND(prison, db_show_prison_command)
4321 * Show all prisons in the list, and prison0 which is not
4324 db_show_prison(&prison0);
4325 if (!db_pager_quit) {
4326 TAILQ_FOREACH(pr, &allprison, pr_list) {
4338 /* Look for a prison with the ID and with references. */
4339 TAILQ_FOREACH(pr, &allprison, pr_list)
4340 if (pr->pr_id == addr && pr->pr_ref > 0)
4343 /* Look again, without requiring a reference. */
4344 TAILQ_FOREACH(pr, &allprison, pr_list)
4345 if (pr->pr_id == addr)
4348 /* Assume address points to a valid prison. */
4349 pr = (struct prison *)addr;