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>
65 #include <net/if_var.h>
68 #include <netinet/in.h>
73 #include <netinet6/in6_var.h>
77 #include <security/mac/mac_framework.h>
79 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
81 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
82 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
84 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
87 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
89 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
93 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
95 #define _PR_IP_SADDRSEL 0
99 /* prison0 describes what is "real" about the system. */
100 struct prison prison0 = {
106 .pr_securelevel = -1,
108 .pr_childmax = JAIL_MAX,
109 .pr_hostuuid = DEFAULT_HOSTUUID,
110 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
112 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
114 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
116 .pr_allow = PR_ALLOW_ALL,
118 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
120 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
121 struct sx allprison_lock;
122 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
123 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
124 LIST_HEAD(, prison_racct) allprison_racct;
127 static int do_jail_attach(struct thread *td, struct prison *pr);
128 static void prison_complete(void *context, int pending);
129 static void prison_deref(struct prison *pr, int flags);
130 static char *prison_path(struct prison *pr1, struct prison *pr2);
131 static void prison_remove_one(struct prison *pr);
133 static void prison_racct_attach(struct prison *pr);
134 static void prison_racct_modify(struct prison *pr);
135 static void prison_racct_detach(struct prison *pr);
138 static int _prison_check_ip4(const struct prison *, const struct in_addr *);
139 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
142 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
143 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
146 /* Flags for prison_deref */
147 #define PD_DEREF 0x01
148 #define PD_DEUREF 0x02
149 #define PD_LOCKED 0x04
150 #define PD_LIST_SLOCKED 0x08
151 #define PD_LIST_XLOCKED 0x10
154 * Parameter names corresponding to PR_* flag values. Size values are for kvm
155 * as we cannot figure out the size of a sparse array, or an array without a
158 static char *pr_flag_names[] = {
161 [7] = "ip4.saddrsel",
164 [8] = "ip6.saddrsel",
167 const size_t pr_flag_names_size = sizeof(pr_flag_names);
169 static char *pr_flag_nonames[] = {
172 [7] = "ip4.nosaddrsel",
175 [8] = "ip6.nosaddrsel",
178 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
180 struct jailsys_flags {
184 } pr_flag_jailsys[] = {
185 { "host", 0, PR_HOST },
187 { "vnet", 0, PR_VNET },
190 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
193 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
196 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
198 static char *pr_allow_names[] = {
199 "allow.set_hostname",
207 "allow.mount.nullfs",
209 "allow.mount.procfs",
212 const size_t pr_allow_names_size = sizeof(pr_allow_names);
214 static char *pr_allow_nonames[] = {
215 "allow.noset_hostname",
217 "allow.noraw_sockets",
222 "allow.mount.nodevfs",
223 "allow.mount.nonullfs",
225 "allow.mount.noprocfs",
226 "allow.mount.notmpfs",
228 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
230 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
231 #define JAIL_DEFAULT_ENFORCE_STATFS 2
232 #define JAIL_DEFAULT_DEVFS_RSNUM 0
233 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
234 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
235 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
236 #if defined(INET) || defined(INET6)
237 static unsigned jail_max_af_ips = 255;
242 qcmp_v4(const void *ip1, const void *ip2)
247 * We need to compare in HBO here to get the list sorted as expected
248 * by the result of the code. Sorting NBO addresses gives you
249 * interesting results. If you do not understand, do not try.
251 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
252 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
255 * Do not simply return the difference of the two numbers, the int is
269 qcmp_v6(const void *ip1, const void *ip2)
271 const struct in6_addr *ia6a, *ia6b;
274 ia6a = (const struct in6_addr *)ip1;
275 ia6b = (const struct in6_addr *)ip2;
278 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
279 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
281 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
294 sys_jail(struct thread *td, struct jail_args *uap)
300 error = copyin(uap->jail, &version, sizeof(uint32_t));
309 /* FreeBSD single IPv4 jails. */
310 bzero(&j, sizeof(struct jail));
311 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
314 j.version = j0.version;
316 j.hostname = j0.hostname;
317 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
323 * Version 1 was used by multi-IPv4 jail implementations
324 * that never made it into the official kernel.
328 case 2: /* JAIL_API_VERSION */
329 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
330 error = copyin(uap->jail, &j, sizeof(struct jail));
336 /* Sci-Fi jails are not supported, sorry. */
339 return (kern_jail(td, &j));
343 kern_jail(struct thread *td, struct jail *j)
345 struct iovec optiov[2 * (4
346 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
355 char *u_path, *u_hostname, *u_name;
358 struct in_addr *u_ip4;
361 struct in6_addr *u_ip6;
364 int error, enforce_statfs, fi;
366 bzero(&optiov, sizeof(optiov));
367 opt.uio_iov = optiov;
371 opt.uio_segflg = UIO_SYSSPACE;
372 opt.uio_rw = UIO_READ;
375 /* Set permissions for top-level jails from sysctls. */
376 if (!jailed(td->td_ucred)) {
377 for (fi = 0; fi < sizeof(pr_allow_names) /
378 sizeof(pr_allow_names[0]); fi++) {
379 optiov[opt.uio_iovcnt].iov_base =
380 (jail_default_allow & (1 << fi))
381 ? pr_allow_names[fi] : pr_allow_nonames[fi];
382 optiov[opt.uio_iovcnt].iov_len =
383 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
386 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
387 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
389 enforce_statfs = jail_default_enforce_statfs;
390 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
391 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
395 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
397 ip4s = (j->version == 0) ? 1 : j->ip4s;
398 if (ip4s > jail_max_af_ips)
400 tmplen += ip4s * sizeof(struct in_addr);
406 if (j->ip6s > jail_max_af_ips)
408 tmplen += j->ip6s * sizeof(struct in6_addr);
413 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
414 u_hostname = u_path + MAXPATHLEN;
415 u_name = u_hostname + MAXHOSTNAMELEN;
417 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
421 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
423 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
426 optiov[opt.uio_iovcnt].iov_base = "path";
427 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
429 optiov[opt.uio_iovcnt].iov_base = u_path;
430 error = copyinstr(j->path, u_path, MAXPATHLEN,
431 &optiov[opt.uio_iovcnt].iov_len);
433 free(u_path, M_TEMP);
437 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
438 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
440 optiov[opt.uio_iovcnt].iov_base = u_hostname;
441 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
442 &optiov[opt.uio_iovcnt].iov_len);
444 free(u_path, M_TEMP);
448 if (j->jailname != NULL) {
449 optiov[opt.uio_iovcnt].iov_base = "name";
450 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
452 optiov[opt.uio_iovcnt].iov_base = u_name;
453 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
454 &optiov[opt.uio_iovcnt].iov_len);
456 free(u_path, M_TEMP);
462 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
463 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
465 optiov[opt.uio_iovcnt].iov_base = u_ip4;
466 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
468 u_ip4->s_addr = j->ip4s;
470 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
472 free(u_path, M_TEMP);
479 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
480 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
482 optiov[opt.uio_iovcnt].iov_base = u_ip6;
483 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
484 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
486 free(u_path, M_TEMP);
491 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
492 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
493 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
494 free(u_path, M_TEMP);
500 * struct jail_set_args {
501 * struct iovec *iovp;
502 * unsigned int iovcnt;
507 sys_jail_set(struct thread *td, struct jail_set_args *uap)
512 /* Check that we have an even number of iovecs. */
516 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
519 error = kern_jail_set(td, auio, uap->flags);
525 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
532 struct in6_addr *ip6;
535 struct vfsoptlist *opts;
536 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
538 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
540 #if defined(INET) || defined(INET6)
545 size_t namelen, onamelen;
546 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
547 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
548 int fi, jid, jsys, len, level;
549 int childmax, rsnum, slevel;
550 int fullpath_disabled;
551 #if defined(INET) || defined(INET6)
560 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
564 error = priv_check(td, PRIV_JAIL_SET);
565 if (!error && (flags & JAIL_ATTACH))
566 error = priv_check(td, PRIV_JAIL_ATTACH);
569 mypr = ppr = td->td_ucred->cr_prison;
570 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
572 if (flags & ~JAIL_SET_MASK)
576 * Check all the parameters before committing to anything. Not all
577 * errors can be caught early, but we may as well try. Also, this
578 * takes care of some expensive stuff (path lookup) before getting
579 * the allprison lock.
581 * XXX Jails are not filesystems, and jail parameters are not mount
582 * options. But it makes more sense to re-use the vfsopt code
583 * than duplicate it under a different name.
585 error = vfs_buildopts(optuio, &opts);
596 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
602 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
611 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
619 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
624 else if (enforce < 0 || enforce > 2) {
630 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
638 pr_flags = ch_flags = 0;
639 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
641 if (pr_flag_names[fi] == NULL)
643 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
644 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
646 ch_flags |= pr_flags;
647 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
649 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
656 case JAIL_SYS_DISABLE:
657 if (!pr_flag_jailsys[fi].disable) {
661 pr_flags |= pr_flag_jailsys[fi].disable;
664 pr_flags |= pr_flag_jailsys[fi].new;
666 case JAIL_SYS_INHERIT:
673 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
675 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
676 && !(pr_flags & PR_PERSIST)) {
678 vfs_opterror(opts, "new jail must persist or attach");
682 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
684 vfs_opterror(opts, "vnet cannot be changed after creation");
689 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
691 vfs_opterror(opts, "ip4 cannot be changed after creation");
696 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
698 vfs_opterror(opts, "ip6 cannot be changed after creation");
703 pr_allow = ch_allow = 0;
704 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
706 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
707 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
709 ch_allow |= pr_allow;
711 error = vfs_getopt(opts, "name", (void **)&name, &len);
717 if (len == 0 || name[len - 1] != '\0') {
721 if (len > MAXHOSTNAMELEN) {
722 error = ENAMETOOLONG;
727 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
735 if (len == 0 || host[len - 1] != '\0') {
739 if (len > MAXHOSTNAMELEN) {
740 error = ENAMETOOLONG;
745 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
753 if (len == 0 || domain[len - 1] != '\0') {
757 if (len > MAXHOSTNAMELEN) {
758 error = ENAMETOOLONG;
763 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
771 if (len == 0 || uuid[len - 1] != '\0') {
775 if (len > HOSTUUIDLEN) {
776 error = ENAMETOOLONG;
781 #ifdef COMPAT_FREEBSD32
782 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
785 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
789 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
801 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
803 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
806 else if (ip4s & (sizeof(*ip4) - 1)) {
810 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
812 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
814 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
815 ip4s /= sizeof(*ip4);
816 if (ip4s > jail_max_af_ips) {
818 vfs_opterror(opts, "too many IPv4 addresses");
821 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
822 bcopy(op, ip4, ip4s * sizeof(*ip4));
824 * IP addresses are all sorted but ip[0] to preserve
825 * the primary IP address as given from userland.
826 * This special IP is used for unbound outgoing
827 * connections as well for "loopback" traffic in case
828 * source address selection cannot find any more fitting
829 * address to connect from.
832 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
834 * Check for duplicate addresses and do some simple
835 * zero and broadcast checks. If users give other bogus
836 * addresses it is their problem.
838 * We do not have to care about byte order for these
839 * checks so we will do them in NBO.
841 for (ii = 0; ii < ip4s; ii++) {
842 if (ip4[ii].s_addr == INADDR_ANY ||
843 ip4[ii].s_addr == INADDR_BROADCAST) {
848 (ip4[0].s_addr == ip4[ii+1].s_addr ||
849 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
859 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
861 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
864 else if (ip6s & (sizeof(*ip6) - 1)) {
868 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
870 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
872 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
873 ip6s /= sizeof(*ip6);
874 if (ip6s > jail_max_af_ips) {
876 vfs_opterror(opts, "too many IPv6 addresses");
879 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
880 bcopy(op, ip6, ip6s * sizeof(*ip6));
882 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
883 for (ii = 0; ii < ip6s; ii++) {
884 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
889 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
890 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
900 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
901 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
904 "vnet jails cannot have IP address restrictions");
909 fullpath_disabled = 0;
911 error = vfs_getopt(opts, "path", (void **)&path, &len);
917 if (flags & JAIL_UPDATE) {
920 "path cannot be changed after creation");
923 if (len == 0 || path[len - 1] != '\0') {
927 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
933 NDFREE(&nd, NDF_ONLY_PNBUF);
934 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
935 strlcpy(g_path, path, MAXPATHLEN);
936 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
939 else if (error == ENODEV) {
940 /* proceed if sysctl debug.disablefullpath == 1 */
941 fullpath_disabled = 1;
942 if (len < 2 || (len == 2 && path[0] == '/'))
945 /* exit on other errors */
948 if (root->v_type != VDIR) {
954 if (fullpath_disabled) {
955 /* Leave room for a real-root full pathname. */
956 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
957 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
958 error = ENAMETOOLONG;
965 * Grab the allprison lock before letting modules check their
966 * parameters. Once we have it, do not let go so we'll have a
967 * consistent view of the OSD list.
969 sx_xlock(&allprison_lock);
970 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
972 goto done_unlock_list;
974 /* By now, all parameters should have been noted. */
975 TAILQ_FOREACH(opt, opts, link) {
976 if (!opt->seen && strcmp(opt->name, "errmsg")) {
978 vfs_opterror(opts, "unknown parameter: %s", opt->name);
979 goto done_unlock_list;
984 * See if we are creating a new record or updating an existing one.
985 * This abuses the file error codes ENOENT and EEXIST.
987 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
990 vfs_opterror(opts, "no valid operation (create or update)");
991 goto done_unlock_list;
995 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
996 namelc = strrchr(name, '.');
997 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1003 * See if a requested jid already exists. There is an
1004 * information leak here if the jid exists but is not within
1005 * the caller's jail hierarchy. Jail creators will get EEXIST
1006 * even though they cannot see the jail, and CREATE | UPDATE
1007 * will return ENOENT which is not normally a valid error.
1011 vfs_opterror(opts, "negative jid");
1012 goto done_unlock_list;
1014 pr = prison_find(jid);
1016 ppr = pr->pr_parent;
1017 /* Create: jid must not exist. */
1018 if (cuflags == JAIL_CREATE) {
1019 mtx_unlock(&pr->pr_mtx);
1021 vfs_opterror(opts, "jail %d already exists",
1023 goto done_unlock_list;
1025 if (!prison_ischild(mypr, pr)) {
1026 mtx_unlock(&pr->pr_mtx);
1028 } else if (pr->pr_uref == 0) {
1029 if (!(flags & JAIL_DYING)) {
1030 mtx_unlock(&pr->pr_mtx);
1032 vfs_opterror(opts, "jail %d is dying",
1034 goto done_unlock_list;
1035 } else if ((flags & JAIL_ATTACH) ||
1036 (pr_flags & PR_PERSIST)) {
1038 * A dying jail might be resurrected
1039 * (via attach or persist), but first
1040 * it must determine if another jail
1041 * has claimed its name. Accomplish
1042 * this by implicitly re-setting the
1046 name = prison_name(mypr, pr);
1051 /* Update: jid must exist. */
1052 if (cuflags == JAIL_UPDATE) {
1054 vfs_opterror(opts, "jail %d not found", jid);
1055 goto done_unlock_list;
1060 * If the caller provided a name, look for a jail by that name.
1061 * This has different semantics for creates and updates keyed by jid
1062 * (where the name must not already exist in a different jail),
1063 * and updates keyed by the name itself (where the name must exist
1064 * because that is the jail being updated).
1067 namelc = strrchr(name, '.');
1072 * This is a hierarchical name. Split it into the
1073 * parent and child names, and make sure the parent
1074 * exists or matches an already found jail.
1078 if (strncmp(name, ppr->pr_name, namelc - name)
1079 || ppr->pr_name[namelc - name] != '\0') {
1080 mtx_unlock(&pr->pr_mtx);
1083 "cannot change jail's parent");
1084 goto done_unlock_list;
1087 ppr = prison_find_name(mypr, name);
1091 "jail \"%s\" not found", name);
1092 goto done_unlock_list;
1094 mtx_unlock(&ppr->pr_mtx);
1098 if (name[0] != '\0') {
1100 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1103 FOREACH_PRISON_CHILD(ppr, tpr) {
1104 if (tpr != pr && tpr->pr_ref > 0 &&
1105 !strcmp(tpr->pr_name + namelen, name)) {
1107 cuflags != JAIL_CREATE) {
1108 mtx_lock(&tpr->pr_mtx);
1109 if (tpr->pr_ref > 0) {
1114 if (tpr->pr_uref > 0) {
1120 mtx_unlock(&tpr->pr_mtx);
1121 } else if (tpr->pr_uref > 0) {
1123 * Create, or update(jid):
1124 * name must not exist in an
1125 * active sibling jail.
1129 mtx_unlock(&pr->pr_mtx);
1131 "jail \"%s\" already exists",
1133 goto done_unlock_list;
1137 /* If no active jail is found, use a dying one. */
1138 if (deadpr != NULL && pr == NULL) {
1139 if (flags & JAIL_DYING) {
1140 mtx_lock(&deadpr->pr_mtx);
1141 if (deadpr->pr_ref == 0) {
1142 mtx_unlock(&deadpr->pr_mtx);
1146 } else if (cuflags == JAIL_UPDATE) {
1149 "jail \"%s\" is dying", name);
1150 goto done_unlock_list;
1153 /* Update: name must exist if no jid. */
1154 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1156 vfs_opterror(opts, "jail \"%s\" not found",
1158 goto done_unlock_list;
1162 /* Update: must provide a jid or name. */
1163 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1165 vfs_opterror(opts, "update specified no jail");
1166 goto done_unlock_list;
1169 /* If there's no prison to update, create a new one and link it in. */
1171 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1172 if (tpr->pr_childcount >= tpr->pr_childmax) {
1174 vfs_opterror(opts, "prison limit exceeded");
1175 goto done_unlock_list;
1178 mtx_lock(&ppr->pr_mtx);
1179 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1180 mtx_unlock(&ppr->pr_mtx);
1182 vfs_opterror(opts, "parent jail went away!");
1183 goto done_unlock_list;
1187 mtx_unlock(&ppr->pr_mtx);
1188 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1190 /* Find the next free jid. */
1193 if (jid == JAIL_MAX)
1195 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1196 if (tpr->pr_id < jid)
1198 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1199 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1202 if (jid == lastprid) {
1205 "no available jail IDs");
1207 prison_deref(ppr, PD_DEREF |
1208 PD_DEUREF | PD_LIST_XLOCKED);
1217 * The jail already has a jid (that did not yet exist),
1218 * so just find where to insert it.
1220 TAILQ_FOREACH(tpr, &allprison, pr_list)
1221 if (tpr->pr_id >= jid) {
1222 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1227 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1228 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1229 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1230 tpr->pr_childcount++;
1232 pr->pr_parent = ppr;
1235 /* Set some default values, and inherit some from the parent. */
1240 root = mypr->pr_root;
1243 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1244 pr->pr_flags |= PR_HOST;
1245 #if defined(INET) || defined(INET6)
1247 if (!(pr_flags & PR_VNET))
1251 if (!(ch_flags & PR_IP4_USER))
1253 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1254 else if (!(pr_flags & PR_IP4_USER)) {
1255 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1256 if (ppr->pr_ip4 != NULL) {
1257 pr->pr_ip4s = ppr->pr_ip4s;
1258 pr->pr_ip4 = malloc(pr->pr_ip4s *
1259 sizeof(struct in_addr), M_PRISON,
1261 bcopy(ppr->pr_ip4, pr->pr_ip4,
1262 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1267 if (!(ch_flags & PR_IP6_USER))
1269 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1270 else if (!(pr_flags & PR_IP6_USER)) {
1271 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1272 if (ppr->pr_ip6 != NULL) {
1273 pr->pr_ip6s = ppr->pr_ip6s;
1274 pr->pr_ip6 = malloc(pr->pr_ip6s *
1275 sizeof(struct in6_addr), M_PRISON,
1277 bcopy(ppr->pr_ip6, pr->pr_ip6,
1278 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1284 /* Source address selection is always on by default. */
1285 pr->pr_flags |= _PR_IP_SADDRSEL;
1287 pr->pr_securelevel = ppr->pr_securelevel;
1288 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1289 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1290 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1292 LIST_INIT(&pr->pr_children);
1293 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1296 /* Allocate a new vnet if specified. */
1297 pr->pr_vnet = (pr_flags & PR_VNET)
1298 ? vnet_alloc() : ppr->pr_vnet;
1301 * Allocate a dedicated cpuset for each jail.
1302 * Unlike other initial settings, this may return an erorr.
1304 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1306 prison_deref(pr, PD_LIST_XLOCKED);
1310 mtx_lock(&pr->pr_mtx);
1312 * New prisons do not yet have a reference, because we do not
1313 * want other to see the incomplete prison once the
1314 * allprison_lock is downgraded.
1319 * Grab a reference for existing prisons, to ensure they
1320 * continue to exist for the duration of the call.
1323 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1324 if ((pr->pr_flags & PR_VNET) &&
1325 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1328 "vnet jails cannot have IP address restrictions");
1329 goto done_deref_locked;
1333 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1336 "ip4 cannot be changed after creation");
1337 goto done_deref_locked;
1341 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1344 "ip6 cannot be changed after creation");
1345 goto done_deref_locked;
1350 /* Do final error checking before setting anything. */
1352 if (slevel < ppr->pr_securelevel) {
1354 goto done_deref_locked;
1358 if (childmax >= ppr->pr_childmax) {
1360 goto done_deref_locked;
1364 if (enforce < ppr->pr_enforce_statfs) {
1366 goto done_deref_locked;
1371 * devfs_rsnum is a uint16_t
1373 if (rsnum < 0 || rsnum > 65535) {
1375 goto done_deref_locked;
1378 * Nested jails always inherit parent's devfs ruleset
1380 if (jailed(td->td_ucred)) {
1381 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1383 goto done_deref_locked;
1385 rsnum = ppr->pr_devfs_rsnum;
1390 if (ppr->pr_flags & PR_IP4) {
1392 * Make sure the new set of IP addresses is a
1393 * subset of the parent's list. Don't worry
1394 * about the parent being unlocked, as any
1395 * setting is done with allprison_lock held.
1397 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1398 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1400 if (ij == ppr->pr_ip4s) {
1402 goto done_deref_locked;
1405 for (ii = ij = 1; ii < ip4s; ii++) {
1406 if (ip4[ii].s_addr ==
1407 ppr->pr_ip4[0].s_addr)
1409 for (; ij < ppr->pr_ip4s; ij++)
1410 if (ip4[ii].s_addr ==
1411 ppr->pr_ip4[ij].s_addr)
1413 if (ij == ppr->pr_ip4s)
1416 if (ij == ppr->pr_ip4s) {
1418 goto done_deref_locked;
1423 * Check for conflicting IP addresses. We permit them
1424 * if there is no more than one IP on each jail. If
1425 * there is a duplicate on a jail with more than one
1426 * IP stop checking and return error.
1430 for (; tppr != &prison0; tppr = tppr->pr_parent)
1431 if (tppr->pr_flags & PR_VNET)
1434 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1437 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1439 tpr->pr_uref == 0) {
1443 if (!(tpr->pr_flags & PR_IP4_USER))
1446 if (tpr->pr_ip4 == NULL ||
1447 (ip4s == 1 && tpr->pr_ip4s == 1))
1449 for (ii = 0; ii < ip4s; ii++) {
1450 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1453 "IPv4 addresses clash");
1454 goto done_deref_locked;
1462 if (ppr->pr_flags & PR_IP6) {
1464 * Make sure the new set of IP addresses is a
1465 * subset of the parent's list.
1467 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1468 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1471 if (ij == ppr->pr_ip6s) {
1473 goto done_deref_locked;
1476 for (ii = ij = 1; ii < ip6s; ii++) {
1477 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1480 for (; ij < ppr->pr_ip6s; ij++)
1481 if (IN6_ARE_ADDR_EQUAL(
1482 &ip6[ii], &ppr->pr_ip6[ij]))
1484 if (ij == ppr->pr_ip6s)
1487 if (ij == ppr->pr_ip6s) {
1489 goto done_deref_locked;
1493 /* Check for conflicting IP addresses. */
1496 for (; tppr != &prison0; tppr = tppr->pr_parent)
1497 if (tppr->pr_flags & PR_VNET)
1500 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1503 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1505 tpr->pr_uref == 0) {
1509 if (!(tpr->pr_flags & PR_IP6_USER))
1512 if (tpr->pr_ip6 == NULL ||
1513 (ip6s == 1 && tpr->pr_ip6s == 1))
1515 for (ii = 0; ii < ip6s; ii++) {
1516 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1519 "IPv6 addresses clash");
1520 goto done_deref_locked;
1526 onamelen = namelen = 0;
1528 /* Give a default name of the jid. */
1529 if (name[0] == '\0')
1530 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1531 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1535 "name cannot be numeric (unless it is the jid)");
1536 goto done_deref_locked;
1539 * Make sure the name isn't too long for the prison or its
1542 onamelen = strlen(pr->pr_name);
1543 namelen = strlen(name);
1544 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1545 error = ENAMETOOLONG;
1546 goto done_deref_locked;
1548 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1549 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1550 sizeof(pr->pr_name)) {
1551 error = ENAMETOOLONG;
1552 goto done_deref_locked;
1556 if (pr_allow & ~ppr->pr_allow) {
1558 goto done_deref_locked;
1561 /* Set the parameters of the prison. */
1564 if (pr_flags & PR_IP4_USER) {
1565 pr->pr_flags |= PR_IP4;
1566 free(pr->pr_ip4, M_PRISON);
1570 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1572 if (tpr->pr_flags & PR_VNET) {
1577 if (prison_restrict_ip4(tpr, NULL)) {
1586 if (pr_flags & PR_IP6_USER) {
1587 pr->pr_flags |= PR_IP6;
1588 free(pr->pr_ip6, M_PRISON);
1592 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1594 if (tpr->pr_flags & PR_VNET) {
1599 if (prison_restrict_ip6(tpr, NULL)) {
1607 pr->pr_securelevel = slevel;
1608 /* Set all child jails to be at least this level. */
1609 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1610 if (tpr->pr_securelevel < slevel)
1611 tpr->pr_securelevel = slevel;
1614 pr->pr_childmax = childmax;
1615 /* Set all child jails to under this limit. */
1616 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1617 if (tpr->pr_childmax > childmax - level)
1618 tpr->pr_childmax = childmax > level
1619 ? childmax - level : 0;
1622 pr->pr_enforce_statfs = enforce;
1623 /* Pass this restriction on to the children. */
1624 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1625 if (tpr->pr_enforce_statfs < enforce)
1626 tpr->pr_enforce_statfs = enforce;
1629 pr->pr_devfs_rsnum = rsnum;
1630 /* Pass this restriction on to the children. */
1631 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1632 tpr->pr_devfs_rsnum = rsnum;
1635 if (ppr == &prison0)
1636 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1638 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1639 ppr->pr_name, name);
1640 /* Change this component of child names. */
1641 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1642 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1643 strlen(tpr->pr_name + onamelen) + 1);
1644 bcopy(pr->pr_name, tpr->pr_name, namelen);
1648 /* Try to keep a real-rooted full pathname. */
1649 if (fullpath_disabled && path[0] == '/' &&
1650 strcmp(mypr->pr_path, "/"))
1651 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1652 mypr->pr_path, path);
1654 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1657 if (PR_HOST & ch_flags & ~pr_flags) {
1658 if (pr->pr_flags & PR_HOST) {
1660 * Copy the parent's host info. As with pr_ip4 above,
1661 * the lack of a lock on the parent is not a problem;
1662 * it is always set with allprison_lock at least
1663 * shared, and is held exclusively here.
1665 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1666 sizeof(pr->pr_hostname));
1667 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1668 sizeof(pr->pr_domainname));
1669 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1670 sizeof(pr->pr_hostuuid));
1671 pr->pr_hostid = pr->pr_parent->pr_hostid;
1673 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1674 /* Set this prison, and any descendants without PR_HOST. */
1676 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1678 strlcpy(pr->pr_domainname, domain,
1679 sizeof(pr->pr_domainname));
1681 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1683 pr->pr_hostid = hid;
1684 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1685 if (tpr->pr_flags & PR_HOST)
1689 strlcpy(tpr->pr_hostname,
1691 sizeof(tpr->pr_hostname));
1693 strlcpy(tpr->pr_domainname,
1695 sizeof(tpr->pr_domainname));
1697 strlcpy(tpr->pr_hostuuid,
1699 sizeof(tpr->pr_hostuuid));
1701 tpr->pr_hostid = hid;
1705 if ((tallow = ch_allow & ~pr_allow)) {
1706 /* Clear allow bits in all children. */
1707 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1708 tpr->pr_allow &= ~tallow;
1710 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1712 * Persistent prisons get an extra reference, and prisons losing their
1713 * persist flag lose that reference. Only do this for existing prisons
1714 * for now, so new ones will remain unseen until after the module
1715 * handlers have completed.
1717 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1718 if (pr_flags & PR_PERSIST) {
1726 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1727 mtx_unlock(&pr->pr_mtx);
1731 prison_racct_attach(pr);
1734 /* Locks may have prevented a complete restriction of child IP
1735 * addresses. If so, allocate some more memory and try again.
1740 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1741 mtx_lock(&pr->pr_mtx);
1743 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1745 if (tpr->pr_flags & PR_VNET) {
1750 if (prison_restrict_ip4(tpr, ip4)) {
1757 mtx_unlock(&pr->pr_mtx);
1763 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1764 mtx_lock(&pr->pr_mtx);
1766 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1768 if (tpr->pr_flags & PR_VNET) {
1773 if (prison_restrict_ip6(tpr, ip6)) {
1780 mtx_unlock(&pr->pr_mtx);
1784 /* Let the modules do their work. */
1785 sx_downgrade(&allprison_lock);
1787 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1789 prison_deref(pr, PD_LIST_SLOCKED);
1793 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1795 prison_deref(pr, created
1797 : PD_DEREF | PD_LIST_SLOCKED);
1801 /* Attach this process to the prison if requested. */
1802 if (flags & JAIL_ATTACH) {
1803 mtx_lock(&pr->pr_mtx);
1804 error = do_jail_attach(td, pr);
1806 vfs_opterror(opts, "attach failed");
1808 prison_deref(pr, PD_DEREF);
1815 if (!(flags & JAIL_ATTACH))
1816 sx_sunlock(&allprison_lock);
1817 prison_racct_modify(pr);
1818 if (!(flags & JAIL_ATTACH))
1819 sx_slock(&allprison_lock);
1823 td->td_retval[0] = pr->pr_id;
1826 * Now that it is all there, drop the temporary reference from existing
1827 * prisons. Or add a reference to newly created persistent prisons
1828 * (which was not done earlier so that the prison would not be publicly
1832 prison_deref(pr, (flags & JAIL_ATTACH)
1834 : PD_DEREF | PD_LIST_SLOCKED);
1836 if (pr_flags & PR_PERSIST) {
1837 mtx_lock(&pr->pr_mtx);
1840 mtx_unlock(&pr->pr_mtx);
1842 if (!(flags & JAIL_ATTACH))
1843 sx_sunlock(&allprison_lock);
1849 prison_deref(pr, created
1850 ? PD_LOCKED | PD_LIST_XLOCKED
1851 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1854 sx_xunlock(&allprison_lock);
1860 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1861 if (errmsg_len > 0) {
1862 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1863 if (errmsg_pos > 0) {
1864 if (optuio->uio_segflg == UIO_SYSSPACE)
1866 optuio->uio_iov[errmsg_pos].iov_base,
1870 optuio->uio_iov[errmsg_pos].iov_base,
1877 free(ip4, M_PRISON);
1880 free(ip6, M_PRISON);
1883 free(g_path, M_TEMP);
1890 * struct jail_get_args {
1891 * struct iovec *iovp;
1892 * unsigned int iovcnt;
1897 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1902 /* Check that we have an even number of iovecs. */
1903 if (uap->iovcnt & 1)
1906 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1909 error = kern_jail_get(td, auio, uap->flags);
1911 error = copyout(auio->uio_iov, uap->iovp,
1912 uap->iovcnt * sizeof (struct iovec));
1918 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1920 struct prison *pr, *mypr;
1922 struct vfsoptlist *opts;
1923 char *errmsg, *name;
1924 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1926 if (flags & ~JAIL_GET_MASK)
1929 /* Get the parameter list. */
1930 error = vfs_buildopts(optuio, &opts);
1933 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1934 mypr = td->td_ucred->cr_prison;
1937 * Find the prison specified by one of: lastjid, jid, name.
1939 sx_slock(&allprison_lock);
1940 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1942 TAILQ_FOREACH(pr, &allprison, pr_list) {
1943 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1944 mtx_lock(&pr->pr_mtx);
1945 if (pr->pr_ref > 0 &&
1946 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1948 mtx_unlock(&pr->pr_mtx);
1954 vfs_opterror(opts, "no jail after %d", jid);
1955 goto done_unlock_list;
1956 } else if (error != ENOENT)
1957 goto done_unlock_list;
1959 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1962 pr = prison_find_child(mypr, jid);
1964 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1965 mtx_unlock(&pr->pr_mtx);
1967 vfs_opterror(opts, "jail %d is dying",
1969 goto done_unlock_list;
1974 vfs_opterror(opts, "jail %d not found", jid);
1975 goto done_unlock_list;
1977 } else if (error != ENOENT)
1978 goto done_unlock_list;
1980 error = vfs_getopt(opts, "name", (void **)&name, &len);
1982 if (len == 0 || name[len - 1] != '\0') {
1984 goto done_unlock_list;
1986 pr = prison_find_name(mypr, name);
1988 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1989 mtx_unlock(&pr->pr_mtx);
1991 vfs_opterror(opts, "jail \"%s\" is dying",
1993 goto done_unlock_list;
1998 vfs_opterror(opts, "jail \"%s\" not found", name);
1999 goto done_unlock_list;
2000 } else if (error != ENOENT)
2001 goto done_unlock_list;
2003 vfs_opterror(opts, "no jail specified");
2005 goto done_unlock_list;
2008 /* Get the parameters of the prison. */
2011 td->td_retval[0] = pr->pr_id;
2012 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2013 if (error != 0 && error != ENOENT)
2015 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2016 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2017 if (error != 0 && error != ENOENT)
2019 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2020 if (error != 0 && error != ENOENT)
2022 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2023 sizeof(pr->pr_cpuset->cs_id));
2024 if (error != 0 && error != ENOENT)
2026 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2027 if (error != 0 && error != ENOENT)
2030 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2031 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2032 if (error != 0 && error != ENOENT)
2036 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2037 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2038 if (error != 0 && error != ENOENT)
2041 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2042 sizeof(pr->pr_securelevel));
2043 if (error != 0 && error != ENOENT)
2045 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2046 sizeof(pr->pr_childcount));
2047 if (error != 0 && error != ENOENT)
2049 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2050 sizeof(pr->pr_childmax));
2051 if (error != 0 && error != ENOENT)
2053 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2054 if (error != 0 && error != ENOENT)
2056 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2057 if (error != 0 && error != ENOENT)
2059 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2060 if (error != 0 && error != ENOENT)
2062 #ifdef COMPAT_FREEBSD32
2063 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2064 uint32_t hid32 = pr->pr_hostid;
2066 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2069 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2070 sizeof(pr->pr_hostid));
2071 if (error != 0 && error != ENOENT)
2073 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2074 sizeof(pr->pr_enforce_statfs));
2075 if (error != 0 && error != ENOENT)
2077 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2078 sizeof(pr->pr_devfs_rsnum));
2079 if (error != 0 && error != ENOENT)
2081 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2083 if (pr_flag_names[fi] == NULL)
2085 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2086 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2087 if (error != 0 && error != ENOENT)
2090 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2091 if (error != 0 && error != ENOENT)
2094 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2097 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2098 i = pr_flag_jailsys[fi].disable &&
2099 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2100 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2103 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2104 if (error != 0 && error != ENOENT)
2107 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2109 if (pr_allow_names[fi] == NULL)
2111 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2112 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2113 if (error != 0 && error != ENOENT)
2116 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2117 if (error != 0 && error != ENOENT)
2120 i = (pr->pr_uref == 0);
2121 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2122 if (error != 0 && error != ENOENT)
2125 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2126 if (error != 0 && error != ENOENT)
2129 /* Get the module parameters. */
2130 mtx_unlock(&pr->pr_mtx);
2132 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2135 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2137 /* By now, all parameters should have been noted. */
2138 TAILQ_FOREACH(opt, opts, link) {
2139 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2141 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2146 /* Write the fetched parameters back to userspace. */
2148 TAILQ_FOREACH(opt, opts, link) {
2149 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2150 pos = 2 * opt->pos + 1;
2151 optuio->uio_iov[pos].iov_len = opt->len;
2152 if (opt->value != NULL) {
2153 if (optuio->uio_segflg == UIO_SYSSPACE) {
2155 optuio->uio_iov[pos].iov_base,
2158 error = copyout(opt->value,
2159 optuio->uio_iov[pos].iov_base,
2170 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2174 sx_sunlock(&allprison_lock);
2176 if (error && errmsg_pos >= 0) {
2177 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2178 errmsg_pos = 2 * errmsg_pos + 1;
2179 if (errmsg_len > 0) {
2180 if (optuio->uio_segflg == UIO_SYSSPACE)
2182 optuio->uio_iov[errmsg_pos].iov_base,
2186 optuio->uio_iov[errmsg_pos].iov_base,
2196 * struct jail_remove_args {
2201 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2203 struct prison *pr, *cpr, *lpr, *tpr;
2206 error = priv_check(td, PRIV_JAIL_REMOVE);
2210 sx_xlock(&allprison_lock);
2211 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2213 sx_xunlock(&allprison_lock);
2217 /* Remove all descendants of this prison, then remove this prison. */
2219 pr->pr_flags |= PR_REMOVE;
2220 if (!LIST_EMPTY(&pr->pr_children)) {
2221 mtx_unlock(&pr->pr_mtx);
2223 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2224 mtx_lock(&cpr->pr_mtx);
2225 if (cpr->pr_ref > 0) {
2228 cpr->pr_flags |= PR_REMOVE;
2230 /* Already removed - do not do it again. */
2233 mtx_unlock(&cpr->pr_mtx);
2235 mtx_lock(&lpr->pr_mtx);
2236 prison_remove_one(lpr);
2237 sx_xlock(&allprison_lock);
2242 mtx_lock(&lpr->pr_mtx);
2243 prison_remove_one(lpr);
2244 sx_xlock(&allprison_lock);
2246 mtx_lock(&pr->pr_mtx);
2248 prison_remove_one(pr);
2253 prison_remove_one(struct prison *pr)
2258 /* If the prison was persistent, it is not anymore. */
2260 if (pr->pr_flags & PR_PERSIST) {
2263 pr->pr_flags &= ~PR_PERSIST;
2267 * jail_remove added a reference. If that's the only one, remove
2270 KASSERT(pr->pr_ref > 0,
2271 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2272 if (pr->pr_ref == 1) {
2274 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2278 mtx_unlock(&pr->pr_mtx);
2279 sx_xunlock(&allprison_lock);
2281 * Kill all processes unfortunate enough to be attached to this prison.
2283 sx_slock(&allproc_lock);
2284 LIST_FOREACH(p, &allproc, p_list) {
2286 if (p->p_state != PRS_NEW && p->p_ucred &&
2287 p->p_ucred->cr_prison == pr)
2288 kern_psignal(p, SIGKILL);
2291 sx_sunlock(&allproc_lock);
2292 /* Remove the temporary reference added by jail_remove. */
2293 prison_deref(pr, deuref | PD_DEREF);
2298 * struct jail_attach_args {
2303 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2308 error = priv_check(td, PRIV_JAIL_ATTACH);
2312 sx_slock(&allprison_lock);
2313 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2315 sx_sunlock(&allprison_lock);
2320 * Do not allow a process to attach to a prison that is not
2321 * considered to be "alive".
2323 if (pr->pr_uref == 0) {
2324 mtx_unlock(&pr->pr_mtx);
2325 sx_sunlock(&allprison_lock);
2329 return (do_jail_attach(td, pr));
2333 do_jail_attach(struct thread *td, struct prison *pr)
2337 struct ucred *newcred, *oldcred;
2341 * XXX: Note that there is a slight race here if two threads
2342 * in the same privileged process attempt to attach to two
2343 * different jails at the same time. It is important for
2344 * user processes not to do this, or they might end up with
2345 * a process root from one prison, but attached to the jail
2350 mtx_unlock(&pr->pr_mtx);
2352 /* Let modules do whatever they need to prepare for attaching. */
2353 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2355 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2358 sx_sunlock(&allprison_lock);
2361 * Reparent the newly attached process to this jail.
2363 ppr = td->td_ucred->cr_prison;
2365 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2369 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2370 if ((error = change_dir(pr->pr_root, td)) != 0)
2373 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2376 VOP_UNLOCK(pr->pr_root, 0);
2377 if ((error = change_root(pr->pr_root, td)))
2382 oldcred = p->p_ucred;
2384 crcopy(newcred, oldcred);
2385 newcred->cr_prison = pr;
2386 p->p_ucred = newcred;
2389 racct_proc_ucred_changed(p, oldcred, newcred);
2392 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2395 VOP_UNLOCK(pr->pr_root, 0);
2397 /* Tell modules this thread is still in its old jail after all. */
2398 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2399 prison_deref(pr, PD_DEREF | PD_DEUREF);
2405 * Returns a locked prison instance, or NULL on failure.
2408 prison_find(int prid)
2412 sx_assert(&allprison_lock, SX_LOCKED);
2413 TAILQ_FOREACH(pr, &allprison, pr_list) {
2414 if (pr->pr_id == prid) {
2415 mtx_lock(&pr->pr_mtx);
2418 mtx_unlock(&pr->pr_mtx);
2425 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2428 prison_find_child(struct prison *mypr, int prid)
2433 sx_assert(&allprison_lock, SX_LOCKED);
2434 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2435 if (pr->pr_id == prid) {
2436 mtx_lock(&pr->pr_mtx);
2439 mtx_unlock(&pr->pr_mtx);
2446 * Look for the name relative to mypr. Returns a locked prison or NULL.
2449 prison_find_name(struct prison *mypr, const char *name)
2451 struct prison *pr, *deadpr;
2455 sx_assert(&allprison_lock, SX_LOCKED);
2456 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2459 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2460 if (!strcmp(pr->pr_name + mylen, name)) {
2461 mtx_lock(&pr->pr_mtx);
2462 if (pr->pr_ref > 0) {
2463 if (pr->pr_uref > 0)
2467 mtx_unlock(&pr->pr_mtx);
2470 /* There was no valid prison - perhaps there was a dying one. */
2471 if (deadpr != NULL) {
2472 mtx_lock(&deadpr->pr_mtx);
2473 if (deadpr->pr_ref == 0) {
2474 mtx_unlock(&deadpr->pr_mtx);
2482 * See if a prison has the specific flag set.
2485 prison_flag(struct ucred *cred, unsigned flag)
2488 /* This is an atomic read, so no locking is necessary. */
2489 return (cred->cr_prison->pr_flags & flag);
2493 prison_allow(struct ucred *cred, unsigned flag)
2496 /* This is an atomic read, so no locking is necessary. */
2497 return (cred->cr_prison->pr_allow & flag);
2501 * Remove a prison reference. If that was the last reference, remove the
2502 * prison itself - but not in this context in case there are locks held.
2505 prison_free_locked(struct prison *pr)
2508 mtx_assert(&pr->pr_mtx, MA_OWNED);
2510 if (pr->pr_ref == 0) {
2511 mtx_unlock(&pr->pr_mtx);
2512 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2513 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2516 mtx_unlock(&pr->pr_mtx);
2520 prison_free(struct prison *pr)
2523 mtx_lock(&pr->pr_mtx);
2524 prison_free_locked(pr);
2528 prison_complete(void *context, int pending)
2531 prison_deref((struct prison *)context, 0);
2535 * Remove a prison reference (usually). This internal version assumes no
2536 * mutexes are held, except perhaps the prison itself. If there are no more
2537 * references, release and delist the prison. On completion, the prison lock
2538 * and the allprison lock are both unlocked.
2541 prison_deref(struct prison *pr, int flags)
2543 struct prison *ppr, *tpr;
2545 if (!(flags & PD_LOCKED))
2546 mtx_lock(&pr->pr_mtx);
2548 if (flags & PD_DEUREF) {
2550 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2552 if (flags & PD_DEREF)
2554 /* If the prison still has references, nothing else to do. */
2555 if (pr->pr_ref > 0) {
2556 mtx_unlock(&pr->pr_mtx);
2557 if (flags & PD_LIST_SLOCKED)
2558 sx_sunlock(&allprison_lock);
2559 else if (flags & PD_LIST_XLOCKED)
2560 sx_xunlock(&allprison_lock);
2564 mtx_unlock(&pr->pr_mtx);
2565 if (flags & PD_LIST_SLOCKED) {
2566 if (!sx_try_upgrade(&allprison_lock)) {
2567 sx_sunlock(&allprison_lock);
2568 sx_xlock(&allprison_lock);
2570 } else if (!(flags & PD_LIST_XLOCKED))
2571 sx_xlock(&allprison_lock);
2573 TAILQ_REMOVE(&allprison, pr, pr_list);
2574 LIST_REMOVE(pr, pr_sibling);
2575 ppr = pr->pr_parent;
2576 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2577 tpr->pr_childcount--;
2578 sx_xunlock(&allprison_lock);
2581 if (pr->pr_vnet != ppr->pr_vnet)
2582 vnet_destroy(pr->pr_vnet);
2584 if (pr->pr_root != NULL)
2586 mtx_destroy(&pr->pr_mtx);
2588 free(pr->pr_ip4, M_PRISON);
2591 free(pr->pr_ip6, M_PRISON);
2593 if (pr->pr_cpuset != NULL)
2594 cpuset_rel(pr->pr_cpuset);
2597 prison_racct_detach(pr);
2601 /* Removing a prison frees a reference on its parent. */
2603 mtx_lock(&pr->pr_mtx);
2604 flags = PD_DEREF | PD_DEUREF;
2609 prison_hold_locked(struct prison *pr)
2612 mtx_assert(&pr->pr_mtx, MA_OWNED);
2613 KASSERT(pr->pr_ref > 0,
2614 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2619 prison_hold(struct prison *pr)
2622 mtx_lock(&pr->pr_mtx);
2623 prison_hold_locked(pr);
2624 mtx_unlock(&pr->pr_mtx);
2628 prison_proc_hold(struct prison *pr)
2631 mtx_lock(&pr->pr_mtx);
2632 KASSERT(pr->pr_uref > 0,
2633 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2635 mtx_unlock(&pr->pr_mtx);
2639 prison_proc_free(struct prison *pr)
2642 mtx_lock(&pr->pr_mtx);
2643 KASSERT(pr->pr_uref > 0,
2644 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2645 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2651 * Restrict a prison's IP address list with its parent's, possibly replacing
2652 * it. Return true if the replacement buffer was used (or would have been).
2655 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2660 ppr = pr->pr_parent;
2661 if (!(pr->pr_flags & PR_IP4_USER)) {
2662 /* This has no user settings, so just copy the parent's list. */
2663 if (pr->pr_ip4s < ppr->pr_ip4s) {
2665 * There's no room for the parent's list. Use the
2666 * new list buffer, which is assumed to be big enough
2667 * (if it was passed). If there's no buffer, try to
2671 if (newip4 == NULL) {
2672 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2673 M_PRISON, M_NOWAIT);
2677 if (newip4 != NULL) {
2678 bcopy(ppr->pr_ip4, newip4,
2679 ppr->pr_ip4s * sizeof(*newip4));
2680 free(pr->pr_ip4, M_PRISON);
2681 pr->pr_ip4 = newip4;
2682 pr->pr_ip4s = ppr->pr_ip4s;
2686 pr->pr_ip4s = ppr->pr_ip4s;
2687 if (pr->pr_ip4s > 0)
2688 bcopy(ppr->pr_ip4, pr->pr_ip4,
2689 pr->pr_ip4s * sizeof(*newip4));
2690 else if (pr->pr_ip4 != NULL) {
2691 free(pr->pr_ip4, M_PRISON);
2694 } else if (pr->pr_ip4s > 0) {
2695 /* Remove addresses that aren't in the parent. */
2696 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2697 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2699 if (ij < ppr->pr_ip4s)
2702 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2703 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2706 for (ij = 1; ii < pr->pr_ip4s; ) {
2707 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2711 switch (ij >= ppr->pr_ip4s ? -1 :
2712 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2714 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2715 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2726 if (pr->pr_ip4s == 0) {
2727 pr->pr_flags |= PR_IP4_DISABLE;
2728 free(pr->pr_ip4, M_PRISON);
2736 * Pass back primary IPv4 address of this jail.
2738 * If not restricted return success but do not alter the address. Caller has
2739 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2741 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2742 * Address returned in NBO.
2745 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2749 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2750 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2752 pr = cred->cr_prison;
2753 if (!(pr->pr_flags & PR_IP4))
2755 mtx_lock(&pr->pr_mtx);
2756 if (!(pr->pr_flags & PR_IP4)) {
2757 mtx_unlock(&pr->pr_mtx);
2760 if (pr->pr_ip4 == NULL) {
2761 mtx_unlock(&pr->pr_mtx);
2762 return (EAFNOSUPPORT);
2765 ia->s_addr = pr->pr_ip4[0].s_addr;
2766 mtx_unlock(&pr->pr_mtx);
2771 * Return 1 if we should do proper source address selection or are not jailed.
2772 * We will return 0 if we should bypass source address selection in favour
2773 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2775 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2778 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2784 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2785 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2790 pr = cred->cr_prison;
2791 if (pr->pr_flags & PR_IP4_SADDRSEL)
2794 lia.s_addr = INADDR_ANY;
2795 error = prison_get_ip4(cred, &lia);
2798 if (lia.s_addr == INADDR_ANY)
2801 ia->s_addr = lia.s_addr;
2806 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2809 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2816 * No need to lock since the PR_IP4_USER flag can't be altered for
2819 while (pr1 != &prison0 &&
2821 !(pr1->pr_flags & PR_VNET) &&
2823 !(pr1->pr_flags & PR_IP4_USER))
2824 pr1 = pr1->pr_parent;
2825 while (pr2 != &prison0 &&
2827 !(pr2->pr_flags & PR_VNET) &&
2829 !(pr2->pr_flags & PR_IP4_USER))
2830 pr2 = pr2->pr_parent;
2831 return (pr1 == pr2);
2835 * Make sure our (source) address is set to something meaningful to this
2838 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2839 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2840 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2843 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2849 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2850 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2852 pr = cred->cr_prison;
2853 if (!(pr->pr_flags & PR_IP4))
2855 mtx_lock(&pr->pr_mtx);
2856 if (!(pr->pr_flags & PR_IP4)) {
2857 mtx_unlock(&pr->pr_mtx);
2860 if (pr->pr_ip4 == NULL) {
2861 mtx_unlock(&pr->pr_mtx);
2862 return (EAFNOSUPPORT);
2865 ia0.s_addr = ntohl(ia->s_addr);
2866 if (ia0.s_addr == INADDR_LOOPBACK) {
2867 ia->s_addr = pr->pr_ip4[0].s_addr;
2868 mtx_unlock(&pr->pr_mtx);
2872 if (ia0.s_addr == INADDR_ANY) {
2874 * In case there is only 1 IPv4 address, bind directly.
2876 if (pr->pr_ip4s == 1)
2877 ia->s_addr = pr->pr_ip4[0].s_addr;
2878 mtx_unlock(&pr->pr_mtx);
2882 error = _prison_check_ip4(pr, ia);
2883 mtx_unlock(&pr->pr_mtx);
2888 * Rewrite destination address in case we will connect to loopback address.
2890 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2891 * Address passed in in NBO and returned in NBO.
2894 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2898 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2899 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2901 pr = cred->cr_prison;
2902 if (!(pr->pr_flags & PR_IP4))
2904 mtx_lock(&pr->pr_mtx);
2905 if (!(pr->pr_flags & PR_IP4)) {
2906 mtx_unlock(&pr->pr_mtx);
2909 if (pr->pr_ip4 == NULL) {
2910 mtx_unlock(&pr->pr_mtx);
2911 return (EAFNOSUPPORT);
2914 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2915 ia->s_addr = pr->pr_ip4[0].s_addr;
2916 mtx_unlock(&pr->pr_mtx);
2921 * Return success because nothing had to be changed.
2923 mtx_unlock(&pr->pr_mtx);
2928 * Check if given address belongs to the jail referenced by cred/prison.
2930 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2931 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2932 * doesn't allow IPv4. Address passed in in NBO.
2935 _prison_check_ip4(const struct prison *pr, const struct in_addr *ia)
2940 * Check the primary IP.
2942 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2946 * All the other IPs are sorted so we can do a binary search.
2949 z = pr->pr_ip4s - 2;
2952 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2961 return (EADDRNOTAVAIL);
2965 prison_check_ip4(const struct ucred *cred, const struct in_addr *ia)
2970 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2971 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2973 pr = cred->cr_prison;
2974 if (!(pr->pr_flags & PR_IP4))
2976 mtx_lock(&pr->pr_mtx);
2977 if (!(pr->pr_flags & PR_IP4)) {
2978 mtx_unlock(&pr->pr_mtx);
2981 if (pr->pr_ip4 == NULL) {
2982 mtx_unlock(&pr->pr_mtx);
2983 return (EAFNOSUPPORT);
2986 error = _prison_check_ip4(pr, ia);
2987 mtx_unlock(&pr->pr_mtx);
2994 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2999 ppr = pr->pr_parent;
3000 if (!(pr->pr_flags & PR_IP6_USER)) {
3001 /* This has no user settings, so just copy the parent's list. */
3002 if (pr->pr_ip6s < ppr->pr_ip6s) {
3004 * There's no room for the parent's list. Use the
3005 * new list buffer, which is assumed to be big enough
3006 * (if it was passed). If there's no buffer, try to
3010 if (newip6 == NULL) {
3011 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3012 M_PRISON, M_NOWAIT);
3016 if (newip6 != NULL) {
3017 bcopy(ppr->pr_ip6, newip6,
3018 ppr->pr_ip6s * sizeof(*newip6));
3019 free(pr->pr_ip6, M_PRISON);
3020 pr->pr_ip6 = newip6;
3021 pr->pr_ip6s = ppr->pr_ip6s;
3025 pr->pr_ip6s = ppr->pr_ip6s;
3026 if (pr->pr_ip6s > 0)
3027 bcopy(ppr->pr_ip6, pr->pr_ip6,
3028 pr->pr_ip6s * sizeof(*newip6));
3029 else if (pr->pr_ip6 != NULL) {
3030 free(pr->pr_ip6, M_PRISON);
3033 } else if (pr->pr_ip6s > 0) {
3034 /* Remove addresses that aren't in the parent. */
3035 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3036 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3039 if (ij < ppr->pr_ip6s)
3042 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3043 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3046 for (ij = 1; ii < pr->pr_ip6s; ) {
3047 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3052 switch (ij >= ppr->pr_ip6s ? -1 :
3053 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3055 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3056 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3067 if (pr->pr_ip6s == 0) {
3068 pr->pr_flags |= PR_IP6_DISABLE;
3069 free(pr->pr_ip6, M_PRISON);
3077 * Pass back primary IPv6 address for this jail.
3079 * If not restricted return success but do not alter the address. Caller has
3080 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3082 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3085 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3089 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3090 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3092 pr = cred->cr_prison;
3093 if (!(pr->pr_flags & PR_IP6))
3095 mtx_lock(&pr->pr_mtx);
3096 if (!(pr->pr_flags & PR_IP6)) {
3097 mtx_unlock(&pr->pr_mtx);
3100 if (pr->pr_ip6 == NULL) {
3101 mtx_unlock(&pr->pr_mtx);
3102 return (EAFNOSUPPORT);
3105 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3106 mtx_unlock(&pr->pr_mtx);
3111 * Return 1 if we should do proper source address selection or are not jailed.
3112 * We will return 0 if we should bypass source address selection in favour
3113 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3115 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3118 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3121 struct in6_addr lia6;
3124 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3125 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3130 pr = cred->cr_prison;
3131 if (pr->pr_flags & PR_IP6_SADDRSEL)
3135 error = prison_get_ip6(cred, &lia6);
3138 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3141 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3146 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3149 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3155 while (pr1 != &prison0 &&
3157 !(pr1->pr_flags & PR_VNET) &&
3159 !(pr1->pr_flags & PR_IP6_USER))
3160 pr1 = pr1->pr_parent;
3161 while (pr2 != &prison0 &&
3163 !(pr2->pr_flags & PR_VNET) &&
3165 !(pr2->pr_flags & PR_IP6_USER))
3166 pr2 = pr2->pr_parent;
3167 return (pr1 == pr2);
3171 * Make sure our (source) address is set to something meaningful to this jail.
3173 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3174 * when needed while binding.
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_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3186 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3187 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3189 pr = cred->cr_prison;
3190 if (!(pr->pr_flags & PR_IP6))
3192 mtx_lock(&pr->pr_mtx);
3193 if (!(pr->pr_flags & PR_IP6)) {
3194 mtx_unlock(&pr->pr_mtx);
3197 if (pr->pr_ip6 == NULL) {
3198 mtx_unlock(&pr->pr_mtx);
3199 return (EAFNOSUPPORT);
3202 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3203 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3204 mtx_unlock(&pr->pr_mtx);
3208 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3210 * In case there is only 1 IPv6 address, and v6only is true,
3211 * then bind directly.
3213 if (v6only != 0 && pr->pr_ip6s == 1)
3214 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3215 mtx_unlock(&pr->pr_mtx);
3219 error = _prison_check_ip6(pr, ia6);
3220 mtx_unlock(&pr->pr_mtx);
3225 * Rewrite destination address in case we will connect to loopback address.
3227 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3230 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3234 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3235 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3237 pr = cred->cr_prison;
3238 if (!(pr->pr_flags & PR_IP6))
3240 mtx_lock(&pr->pr_mtx);
3241 if (!(pr->pr_flags & PR_IP6)) {
3242 mtx_unlock(&pr->pr_mtx);
3245 if (pr->pr_ip6 == NULL) {
3246 mtx_unlock(&pr->pr_mtx);
3247 return (EAFNOSUPPORT);
3250 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3251 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3252 mtx_unlock(&pr->pr_mtx);
3257 * Return success because nothing had to be changed.
3259 mtx_unlock(&pr->pr_mtx);
3264 * Check if given address belongs to the jail referenced by cred/prison.
3266 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3267 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3268 * doesn't allow IPv6.
3271 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3276 * Check the primary IP.
3278 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3282 * All the other IPs are sorted so we can do a binary search.
3285 z = pr->pr_ip6s - 2;
3288 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3297 return (EADDRNOTAVAIL);
3301 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3306 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3307 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3309 pr = cred->cr_prison;
3310 if (!(pr->pr_flags & PR_IP6))
3312 mtx_lock(&pr->pr_mtx);
3313 if (!(pr->pr_flags & PR_IP6)) {
3314 mtx_unlock(&pr->pr_mtx);
3317 if (pr->pr_ip6 == NULL) {
3318 mtx_unlock(&pr->pr_mtx);
3319 return (EAFNOSUPPORT);
3322 error = _prison_check_ip6(pr, ia6);
3323 mtx_unlock(&pr->pr_mtx);
3329 * Check if a jail supports the given address family.
3331 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3335 prison_check_af(struct ucred *cred, int af)
3340 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3342 pr = cred->cr_prison;
3344 /* Prisons with their own network stack are not limited. */
3345 if (prison_owns_vnet(cred))
3354 if (pr->pr_flags & PR_IP4)
3356 mtx_lock(&pr->pr_mtx);
3357 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3358 error = EAFNOSUPPORT;
3359 mtx_unlock(&pr->pr_mtx);
3365 if (pr->pr_flags & PR_IP6)
3367 mtx_lock(&pr->pr_mtx);
3368 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3369 error = EAFNOSUPPORT;
3370 mtx_unlock(&pr->pr_mtx);
3378 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3379 error = EAFNOSUPPORT;
3385 * Check if given address belongs to the jail referenced by cred (wrapper to
3386 * prison_check_ip[46]).
3388 * Returns 0 if jail doesn't restrict the address family or if address belongs
3389 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3390 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3393 prison_if(struct ucred *cred, struct sockaddr *sa)
3396 struct sockaddr_in *sai;
3399 struct sockaddr_in6 *sai6;
3403 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3404 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3407 if (prison_owns_vnet(cred))
3412 switch (sa->sa_family)
3416 sai = (struct sockaddr_in *)sa;
3417 error = prison_check_ip4(cred, &sai->sin_addr);
3422 sai6 = (struct sockaddr_in6 *)sa;
3423 error = prison_check_ip6(cred, &sai6->sin6_addr);
3427 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3428 error = EAFNOSUPPORT;
3434 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3437 prison_check(struct ucred *cred1, struct ucred *cred2)
3440 return ((cred1->cr_prison == cred2->cr_prison ||
3441 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3445 * Return 1 if p2 is a child of p1, otherwise 0.
3448 prison_ischild(struct prison *pr1, struct prison *pr2)
3451 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3458 * Return 1 if the passed credential is in a jail, otherwise 0.
3461 jailed(struct ucred *cred)
3464 return (cred->cr_prison != &prison0);
3468 * Return 1 if the passed credential is in a jail and that jail does not
3469 * have its own virtual network stack, otherwise 0.
3472 jailed_without_vnet(struct ucred *cred)
3478 if (prison_owns_vnet(cred))
3486 * Return the correct hostname (domainname, et al) for the passed credential.
3489 getcredhostname(struct ucred *cred, char *buf, size_t size)
3494 * A NULL credential can be used to shortcut to the physical
3495 * system's hostname.
3497 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3498 mtx_lock(&pr->pr_mtx);
3499 strlcpy(buf, pr->pr_hostname, size);
3500 mtx_unlock(&pr->pr_mtx);
3504 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3507 mtx_lock(&cred->cr_prison->pr_mtx);
3508 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3509 mtx_unlock(&cred->cr_prison->pr_mtx);
3513 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3516 mtx_lock(&cred->cr_prison->pr_mtx);
3517 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3518 mtx_unlock(&cred->cr_prison->pr_mtx);
3522 getcredhostid(struct ucred *cred, unsigned long *hostid)
3525 mtx_lock(&cred->cr_prison->pr_mtx);
3526 *hostid = cred->cr_prison->pr_hostid;
3527 mtx_unlock(&cred->cr_prison->pr_mtx);
3532 * Determine whether the prison represented by cred owns
3533 * its vnet rather than having it inherited.
3535 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3538 prison_owns_vnet(struct ucred *cred)
3542 * vnets cannot be added/removed after jail creation,
3543 * so no need to lock here.
3545 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3550 * Determine whether the subject represented by cred can "see"
3551 * status of a mount point.
3552 * Returns: 0 for permitted, ENOENT otherwise.
3553 * XXX: This function should be called cr_canseemount() and should be
3554 * placed in kern_prot.c.
3557 prison_canseemount(struct ucred *cred, struct mount *mp)
3563 pr = cred->cr_prison;
3564 if (pr->pr_enforce_statfs == 0)
3566 if (pr->pr_root->v_mount == mp)
3568 if (pr->pr_enforce_statfs == 2)
3571 * If jail's chroot directory is set to "/" we should be able to see
3572 * all mount-points from inside a jail.
3573 * This is ugly check, but this is the only situation when jail's
3574 * directory ends with '/'.
3576 if (strcmp(pr->pr_path, "/") == 0)
3578 len = strlen(pr->pr_path);
3580 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3583 * Be sure that we don't have situation where jail's root directory
3584 * is "/some/path" and mount point is "/some/pathpath".
3586 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3592 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3594 char jpath[MAXPATHLEN];
3598 pr = cred->cr_prison;
3599 if (pr->pr_enforce_statfs == 0)
3601 if (prison_canseemount(cred, mp) != 0) {
3602 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3603 strlcpy(sp->f_mntonname, "[restricted]",
3604 sizeof(sp->f_mntonname));
3607 if (pr->pr_root->v_mount == mp) {
3609 * Clear current buffer data, so we are sure nothing from
3610 * the valid path left there.
3612 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3613 *sp->f_mntonname = '/';
3617 * If jail's chroot directory is set to "/" we should be able to see
3618 * all mount-points from inside a jail.
3620 if (strcmp(pr->pr_path, "/") == 0)
3622 len = strlen(pr->pr_path);
3623 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3625 * Clear current buffer data, so we are sure nothing from
3626 * the valid path left there.
3628 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3629 if (*jpath == '\0') {
3630 /* Should never happen. */
3631 *sp->f_mntonname = '/';
3633 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3638 * Check with permission for a specific privilege is granted within jail. We
3639 * have a specific list of accepted privileges; the rest are denied.
3642 prison_priv_check(struct ucred *cred, int priv)
3650 * Privileges specific to prisons with a virtual network stack.
3651 * There might be a duplicate entry here in case the privilege
3652 * is only granted conditionally in the legacy jail case.
3657 * NFS-specific privileges.
3659 case PRIV_NFS_DAEMON:
3660 case PRIV_NFS_LOCKD:
3663 * Network stack privileges.
3665 case PRIV_NET_BRIDGE:
3668 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3669 case PRIV_NET_ROUTE:
3671 case PRIV_NET_SETIFMTU:
3672 case PRIV_NET_SETIFFLAGS:
3673 case PRIV_NET_SETIFCAP:
3674 case PRIV_NET_SETIFDESCR:
3675 case PRIV_NET_SETIFNAME :
3676 case PRIV_NET_SETIFMETRIC:
3677 case PRIV_NET_SETIFPHYS:
3678 case PRIV_NET_SETIFMAC:
3679 case PRIV_NET_ADDMULTI:
3680 case PRIV_NET_DELMULTI:
3681 case PRIV_NET_HWIOCTL:
3682 case PRIV_NET_SETLLADDR:
3683 case PRIV_NET_ADDIFGROUP:
3684 case PRIV_NET_DELIFGROUP:
3685 case PRIV_NET_IFCREATE:
3686 case PRIV_NET_IFDESTROY:
3687 case PRIV_NET_ADDIFADDR:
3688 case PRIV_NET_DELIFADDR:
3691 case PRIV_NET_SETIFVNET:
3692 case PRIV_NET_SETIFFIB:
3695 * 802.11-related privileges.
3697 case PRIV_NET80211_GETKEY:
3699 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3706 case PRIV_NETATM_CFG:
3707 case PRIV_NETATM_ADD:
3708 case PRIV_NETATM_DEL:
3709 case PRIV_NETATM_SET:
3712 * Bluetooth privileges.
3714 case PRIV_NETBLUETOOTH_RAW:
3718 * Netgraph and netgraph module privileges.
3720 case PRIV_NETGRAPH_CONTROL:
3722 case PRIV_NETGRAPH_TTY:
3726 * IPv4 and IPv6 privileges.
3728 case PRIV_NETINET_IPFW:
3729 case PRIV_NETINET_DIVERT:
3730 case PRIV_NETINET_PF:
3731 case PRIV_NETINET_DUMMYNET:
3732 case PRIV_NETINET_CARP:
3733 case PRIV_NETINET_MROUTE:
3734 case PRIV_NETINET_RAW:
3735 case PRIV_NETINET_ADDRCTRL6:
3736 case PRIV_NETINET_ND6:
3737 case PRIV_NETINET_SCOPE6:
3738 case PRIV_NETINET_ALIFETIME6:
3739 case PRIV_NETINET_IPSEC:
3740 case PRIV_NETINET_BINDANY:
3755 * No default: or deny here.
3756 * In case of no permit fall through to next switch().
3758 if (cred->cr_prison->pr_flags & PR_VNET)
3766 * Allow ktrace privileges for root in jail.
3772 * Allow jailed processes to configure audit identity and
3773 * submit audit records (login, etc). In the future we may
3774 * want to further refine the relationship between audit and
3777 case PRIV_AUDIT_GETAUDIT:
3778 case PRIV_AUDIT_SETAUDIT:
3779 case PRIV_AUDIT_SUBMIT:
3783 * Allow jailed processes to manipulate process UNIX
3784 * credentials in any way they see fit.
3786 case PRIV_CRED_SETUID:
3787 case PRIV_CRED_SETEUID:
3788 case PRIV_CRED_SETGID:
3789 case PRIV_CRED_SETEGID:
3790 case PRIV_CRED_SETGROUPS:
3791 case PRIV_CRED_SETREUID:
3792 case PRIV_CRED_SETREGID:
3793 case PRIV_CRED_SETRESUID:
3794 case PRIV_CRED_SETRESGID:
3797 * Jail implements visibility constraints already, so allow
3798 * jailed root to override uid/gid-based constraints.
3800 case PRIV_SEEOTHERGIDS:
3801 case PRIV_SEEOTHERUIDS:
3804 * Jail implements inter-process debugging limits already, so
3805 * allow jailed root various debugging privileges.
3807 case PRIV_DEBUG_DIFFCRED:
3808 case PRIV_DEBUG_SUGID:
3809 case PRIV_DEBUG_UNPRIV:
3812 * Allow jail to set various resource limits and login
3813 * properties, and for now, exceed process resource limits.
3815 case PRIV_PROC_LIMIT:
3816 case PRIV_PROC_SETLOGIN:
3817 case PRIV_PROC_SETRLIMIT:
3820 * System V and POSIX IPC privileges are granted in jail.
3823 case PRIV_IPC_WRITE:
3824 case PRIV_IPC_ADMIN:
3825 case PRIV_IPC_MSGSIZE:
3829 * Jail operations within a jail work on child jails.
3831 case PRIV_JAIL_ATTACH:
3833 case PRIV_JAIL_REMOVE:
3836 * Jail implements its own inter-process limits, so allow
3837 * root processes in jail to change scheduling on other
3838 * processes in the same jail. Likewise for signalling.
3840 case PRIV_SCHED_DIFFCRED:
3841 case PRIV_SCHED_CPUSET:
3842 case PRIV_SIGNAL_DIFFCRED:
3843 case PRIV_SIGNAL_SUGID:
3846 * Allow jailed processes to write to sysctls marked as jail
3849 case PRIV_SYSCTL_WRITEJAIL:
3852 * Allow root in jail to manage a variety of quota
3853 * properties. These should likely be conditional on a
3854 * configuration option.
3856 case PRIV_VFS_GETQUOTA:
3857 case PRIV_VFS_SETQUOTA:
3860 * Since Jail relies on chroot() to implement file system
3861 * protections, grant many VFS privileges to root in jail.
3862 * Be careful to exclude mount-related and NFS-related
3866 case PRIV_VFS_WRITE:
3867 case PRIV_VFS_ADMIN:
3869 case PRIV_VFS_LOOKUP:
3870 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3871 case PRIV_VFS_CHFLAGS_DEV:
3872 case PRIV_VFS_CHOWN:
3873 case PRIV_VFS_CHROOT:
3874 case PRIV_VFS_RETAINSUGID:
3875 case PRIV_VFS_FCHROOT:
3877 case PRIV_VFS_SETGID:
3879 case PRIV_VFS_STICKYFILE:
3882 * As in the non-jail case, non-root users are expected to be
3883 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3884 * exists in the jail and they have permission to access it).
3886 case PRIV_KMEM_READ:
3890 * Depending on the global setting, allow privilege of
3891 * setting system flags.
3893 case PRIV_VFS_SYSFLAGS:
3894 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3900 * Depending on the global setting, allow privilege of
3901 * mounting/unmounting file systems.
3903 case PRIV_VFS_MOUNT:
3904 case PRIV_VFS_UNMOUNT:
3905 case PRIV_VFS_MOUNT_NONUSER:
3906 case PRIV_VFS_MOUNT_OWNER:
3907 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3908 cred->cr_prison->pr_enforce_statfs < 2)
3914 * Allow jailed root to bind reserved ports and reuse in-use
3917 case PRIV_NETINET_RESERVEDPORT:
3918 case PRIV_NETINET_REUSEPORT:
3922 * Allow jailed root to set certian IPv4/6 (option) headers.
3924 case PRIV_NETINET_SETHDROPTS:
3928 * Conditionally allow creating raw sockets in jail.
3930 case PRIV_NETINET_RAW:
3931 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3937 * Since jail implements its own visibility limits on netstat
3938 * sysctls, allow getcred. This allows identd to work in
3941 case PRIV_NETINET_GETCRED:
3945 * Allow jailed root to set loginclass.
3947 case PRIV_PROC_SETLOGINCLASS:
3952 * In all remaining cases, deny the privilege request. This
3953 * includes almost all network privileges, many system
3954 * configuration privileges.
3961 * Return the part of pr2's name that is relative to pr1, or the whole name
3962 * if it does not directly follow.
3966 prison_name(struct prison *pr1, struct prison *pr2)
3970 /* Jails see themselves as "0" (if they see themselves at all). */
3973 name = pr2->pr_name;
3974 if (prison_ischild(pr1, pr2)) {
3976 * pr1 isn't locked (and allprison_lock may not be either)
3977 * so its length can't be counted on. But the number of dots
3978 * can be counted on - and counted.
3980 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3981 name = strchr(name, '.') + 1;
3987 * Return the part of pr2's path that is relative to pr1, or the whole path
3988 * if it does not directly follow.
3991 prison_path(struct prison *pr1, struct prison *pr2)
3993 char *path1, *path2;
3996 path1 = pr1->pr_path;
3997 path2 = pr2->pr_path;
3998 if (!strcmp(path1, "/"))
4000 len1 = strlen(path1);
4001 if (strncmp(path1, path2, len1))
4003 if (path2[len1] == '\0')
4005 if (path2[len1] == '/')
4006 return (path2 + len1);
4012 * Jail-related sysctls.
4014 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4018 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4021 struct prison *pr, *cpr;
4023 struct in_addr *ip4 = NULL;
4027 struct in6_addr *ip6 = NULL;
4032 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4033 pr = req->td->td_ucred->cr_prison;
4035 sx_slock(&allprison_lock);
4036 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4037 #if defined(INET) || defined(INET6)
4040 mtx_lock(&cpr->pr_mtx);
4042 if (cpr->pr_ip4s > 0) {
4043 if (ip4s < cpr->pr_ip4s) {
4044 ip4s = cpr->pr_ip4s;
4045 mtx_unlock(&cpr->pr_mtx);
4046 ip4 = realloc(ip4, ip4s *
4047 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4050 bcopy(cpr->pr_ip4, ip4,
4051 cpr->pr_ip4s * sizeof(struct in_addr));
4055 if (cpr->pr_ip6s > 0) {
4056 if (ip6s < cpr->pr_ip6s) {
4057 ip6s = cpr->pr_ip6s;
4058 mtx_unlock(&cpr->pr_mtx);
4059 ip6 = realloc(ip6, ip6s *
4060 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4063 bcopy(cpr->pr_ip6, ip6,
4064 cpr->pr_ip6s * sizeof(struct in6_addr));
4067 if (cpr->pr_ref == 0) {
4068 mtx_unlock(&cpr->pr_mtx);
4071 bzero(xp, sizeof(*xp));
4072 xp->pr_version = XPRISON_VERSION;
4073 xp->pr_id = cpr->pr_id;
4074 xp->pr_state = cpr->pr_uref > 0
4075 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4076 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4077 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4078 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4080 xp->pr_ip4s = cpr->pr_ip4s;
4083 xp->pr_ip6s = cpr->pr_ip6s;
4085 mtx_unlock(&cpr->pr_mtx);
4086 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4090 if (xp->pr_ip4s > 0) {
4091 error = SYSCTL_OUT(req, ip4,
4092 xp->pr_ip4s * sizeof(struct in_addr));
4098 if (xp->pr_ip6s > 0) {
4099 error = SYSCTL_OUT(req, ip6,
4100 xp->pr_ip6s * sizeof(struct in6_addr));
4106 sx_sunlock(&allprison_lock);
4117 SYSCTL_OID(_security_jail, OID_AUTO, list,
4118 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4119 sysctl_jail_list, "S", "List of active jails");
4122 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4126 injail = jailed(req->td->td_ucred);
4127 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4132 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4133 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4134 sysctl_jail_jailed, "I", "Process in jail?");
4137 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4139 int error, havevnet;
4141 struct ucred *cred = req->td->td_ucred;
4143 havevnet = jailed(cred) && prison_owns_vnet(cred);
4147 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4152 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4153 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4154 sysctl_jail_vnet, "I", "Jail owns VNET?");
4156 #if defined(INET) || defined(INET6)
4157 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4158 &jail_max_af_ips, 0,
4159 "Number of IP addresses a jail may have at most per address family");
4163 * Default parameters for jail(2) compatability. For historical reasons,
4164 * the sysctl names have varying similarity to the parameter names. Prisons
4165 * just see their own parameters, and can't change them.
4168 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4171 int allow, error, i;
4173 pr = req->td->td_ucred->cr_prison;
4174 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4176 /* Get the current flag value, and convert it to a boolean. */
4177 i = (allow & arg2) ? 1 : 0;
4180 error = sysctl_handle_int(oidp, &i, 0, req);
4181 if (error || !req->newptr)
4187 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4190 mtx_lock(&prison0.pr_mtx);
4191 jail_default_allow = (jail_default_allow & ~arg2) | i;
4192 mtx_unlock(&prison0.pr_mtx);
4196 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4197 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4198 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4199 "Processes in jail can set their hostnames");
4200 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4201 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4202 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4203 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4204 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4205 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4206 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4207 "Processes in jail can use System V IPC primitives");
4208 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4209 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4210 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4211 "Prison root can create raw sockets");
4212 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4213 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4214 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4215 "Processes in jail can alter system file flags");
4216 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4218 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4219 "Processes in jail can mount/unmount jail-friendly file systems");
4220 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4221 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4222 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4223 "Processes in jail can mount the devfs file system");
4224 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4225 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4226 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4227 "Processes in jail can mount the nullfs file system");
4228 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4229 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4230 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4231 "Processes in jail can mount the procfs file system");
4232 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4233 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4234 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4235 "Processes in jail can mount the tmpfs file system");
4236 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4237 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4238 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4239 "Processes in jail can mount the zfs file system");
4242 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4247 pr = req->td->td_ucred->cr_prison;
4248 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4249 error = sysctl_handle_int(oidp, &level, 0, req);
4250 if (error || !req->newptr)
4252 *(int *)arg1 = level;
4256 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4257 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4258 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4259 sysctl_jail_default_level, "I",
4260 "Processes in jail cannot see all mounted file systems");
4262 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4263 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4264 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4265 sysctl_jail_default_level, "I",
4266 "Ruleset for the devfs filesystem in jail");
4269 * Nodes to describe jail parameters. Maximum length of string parameters
4270 * is returned in the string itself, and the other parameters exist merely
4271 * to make themselves and their types known.
4273 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4277 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4284 switch (oidp->oid_kind & CTLTYPE)
4290 if (!(req->flags & SCTL_MASK32))
4292 return (SYSCTL_OUT(req, &l, sizeof(l)));
4296 return (SYSCTL_OUT(req, &i, sizeof(i)));
4297 case CTLTYPE_STRING:
4298 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4300 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4301 case CTLTYPE_STRUCT:
4303 return (SYSCTL_OUT(req, &s, sizeof(s)));
4308 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4309 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4310 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4311 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4312 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4313 "I", "Jail secure level");
4314 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4315 "I", "Jail cannot see all mounted file systems");
4316 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4317 "I", "Ruleset for in-jail devfs mounts");
4318 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4319 "B", "Jail persistence");
4321 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4322 "E,jailsys", "Virtual network stack");
4324 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4325 "B", "Jail is in the process of shutting down");
4327 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4328 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4329 "I", "Current number of child jails");
4330 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4331 "I", "Maximum number of child jails");
4333 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4334 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4336 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4337 "Jail NIS domainname");
4338 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4340 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4341 "LU", "Jail host ID");
4343 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4344 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4347 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4348 "Jail IPv4 address virtualization");
4349 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4350 "S,in_addr,a", "Jail IPv4 addresses");
4351 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4352 "B", "Do (not) use IPv4 source address selection rather than the "
4353 "primary jail IPv4 address.");
4356 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4357 "Jail IPv6 address virtualization");
4358 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4359 "S,in6_addr,a", "Jail IPv6 addresses");
4360 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4361 "B", "Do (not) use IPv6 source address selection rather than the "
4362 "primary jail IPv6 address.");
4365 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4366 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4367 "B", "Jail may set hostname");
4368 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4369 "B", "Jail may use SYSV IPC");
4370 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4371 "B", "Jail may create raw sockets");
4372 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4373 "B", "Jail may alter system file flags");
4374 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4375 "B", "Jail may set file quotas");
4376 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4377 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4379 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4380 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4381 "B", "Jail may mount/unmount jail-friendly file systems in general");
4382 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4383 "B", "Jail may mount the devfs file system");
4384 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4385 "B", "Jail may mount the nullfs file system");
4386 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4387 "B", "Jail may mount the procfs file system");
4388 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4389 "B", "Jail may mount the tmpfs file system");
4390 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4391 "B", "Jail may mount the zfs file system");
4394 prison_racct_foreach(void (*callback)(struct racct *racct,
4395 void *arg2, void *arg3), void *arg2, void *arg3)
4397 struct prison_racct *prr;
4399 sx_slock(&allprison_lock);
4400 LIST_FOREACH(prr, &allprison_racct, prr_next)
4401 (callback)(prr->prr_racct, arg2, arg3);
4402 sx_sunlock(&allprison_lock);
4405 static struct prison_racct *
4406 prison_racct_find_locked(const char *name)
4408 struct prison_racct *prr;
4410 sx_assert(&allprison_lock, SA_XLOCKED);
4412 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4415 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4416 if (strcmp(name, prr->prr_name) != 0)
4419 /* Found prison_racct with a matching name? */
4420 prison_racct_hold(prr);
4424 /* Add new prison_racct. */
4425 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4426 racct_create(&prr->prr_racct);
4428 strcpy(prr->prr_name, name);
4429 refcount_init(&prr->prr_refcount, 1);
4430 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4435 struct prison_racct *
4436 prison_racct_find(const char *name)
4438 struct prison_racct *prr;
4440 sx_xlock(&allprison_lock);
4441 prr = prison_racct_find_locked(name);
4442 sx_xunlock(&allprison_lock);
4447 prison_racct_hold(struct prison_racct *prr)
4450 refcount_acquire(&prr->prr_refcount);
4454 prison_racct_free_locked(struct prison_racct *prr)
4457 sx_assert(&allprison_lock, SA_XLOCKED);
4459 if (refcount_release(&prr->prr_refcount)) {
4460 racct_destroy(&prr->prr_racct);
4461 LIST_REMOVE(prr, prr_next);
4462 free(prr, M_PRISON_RACCT);
4467 prison_racct_free(struct prison_racct *prr)
4471 sx_assert(&allprison_lock, SA_UNLOCKED);
4473 old = prr->prr_refcount;
4474 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4477 sx_xlock(&allprison_lock);
4478 prison_racct_free_locked(prr);
4479 sx_xunlock(&allprison_lock);
4484 prison_racct_attach(struct prison *pr)
4486 struct prison_racct *prr;
4488 sx_assert(&allprison_lock, SA_XLOCKED);
4490 prr = prison_racct_find_locked(pr->pr_name);
4491 KASSERT(prr != NULL, ("cannot find prison_racct"));
4493 pr->pr_prison_racct = prr;
4497 * Handle jail renaming. From the racct point of view, renaming means
4498 * moving from one prison_racct to another.
4501 prison_racct_modify(struct prison *pr)
4505 struct prison_racct *oldprr;
4507 sx_slock(&allproc_lock);
4508 sx_xlock(&allprison_lock);
4510 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4511 sx_xunlock(&allprison_lock);
4512 sx_sunlock(&allproc_lock);
4516 oldprr = pr->pr_prison_racct;
4517 pr->pr_prison_racct = NULL;
4519 prison_racct_attach(pr);
4522 * Move resource utilisation records.
4524 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4527 * Force rctl to reattach rules to processes.
4529 FOREACH_PROC_IN_SYSTEM(p) {
4531 cred = crhold(p->p_ucred);
4533 racct_proc_ucred_changed(p, cred, cred);
4537 sx_sunlock(&allproc_lock);
4538 prison_racct_free_locked(oldprr);
4539 sx_xunlock(&allprison_lock);
4543 prison_racct_detach(struct prison *pr)
4546 sx_assert(&allprison_lock, SA_UNLOCKED);
4548 if (pr->pr_prison_racct == NULL)
4550 prison_racct_free(pr->pr_prison_racct);
4551 pr->pr_prison_racct = NULL;
4558 db_show_prison(struct prison *pr)
4561 #if defined(INET) || defined(INET6)
4566 char ip6buf[INET6_ADDRSTRLEN];
4569 db_printf("prison %p:\n", pr);
4570 db_printf(" jid = %d\n", pr->pr_id);
4571 db_printf(" name = %s\n", pr->pr_name);
4572 db_printf(" parent = %p\n", pr->pr_parent);
4573 db_printf(" ref = %d\n", pr->pr_ref);
4574 db_printf(" uref = %d\n", pr->pr_uref);
4575 db_printf(" path = %s\n", pr->pr_path);
4576 db_printf(" cpuset = %d\n", pr->pr_cpuset
4577 ? pr->pr_cpuset->cs_id : -1);
4579 db_printf(" vnet = %p\n", pr->pr_vnet);
4581 db_printf(" root = %p\n", pr->pr_root);
4582 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4583 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4584 db_printf(" children.max = %d\n", pr->pr_childmax);
4585 db_printf(" children.cur = %d\n", pr->pr_childcount);
4586 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4587 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4588 db_printf(" flags = 0x%x", pr->pr_flags);
4589 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4591 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4592 db_printf(" %s", pr_flag_names[fi]);
4593 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4595 jsf = pr->pr_flags &
4596 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4597 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4598 pr_flag_jailsys[fi].disable &&
4599 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4600 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4603 db_printf(" allow = 0x%x", pr->pr_allow);
4604 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4606 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4607 db_printf(" %s", pr_allow_names[fi]);
4609 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4610 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4611 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4612 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4613 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4615 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4616 for (ii = 0; ii < pr->pr_ip4s; ii++)
4617 db_printf(" %s %s\n",
4618 ii == 0 ? "ip4.addr =" : " ",
4619 inet_ntoa(pr->pr_ip4[ii]));
4622 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4623 for (ii = 0; ii < pr->pr_ip6s; ii++)
4624 db_printf(" %s %s\n",
4625 ii == 0 ? "ip6.addr =" : " ",
4626 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4630 DB_SHOW_COMMAND(prison, db_show_prison_command)
4636 * Show all prisons in the list, and prison0 which is not
4639 db_show_prison(&prison0);
4640 if (!db_pager_quit) {
4641 TAILQ_FOREACH(pr, &allprison, pr_list) {
4653 /* Look for a prison with the ID and with references. */
4654 TAILQ_FOREACH(pr, &allprison, pr_list)
4655 if (pr->pr_id == addr && pr->pr_ref > 0)
4658 /* Look again, without requiring a reference. */
4659 TAILQ_FOREACH(pr, &allprison, pr_list)
4660 if (pr->pr_id == addr)
4663 /* Assume address points to a valid prison. */
4664 pr = (struct prison *)addr;