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 sx_sunlock(&allprison_lock);
1816 prison_racct_modify(pr);
1817 sx_slock(&allprison_lock);
1821 td->td_retval[0] = pr->pr_id;
1824 * Now that it is all there, drop the temporary reference from existing
1825 * prisons. Or add a reference to newly created persistent prisons
1826 * (which was not done earlier so that the prison would not be publicly
1830 prison_deref(pr, (flags & JAIL_ATTACH)
1832 : PD_DEREF | PD_LIST_SLOCKED);
1834 if (pr_flags & PR_PERSIST) {
1835 mtx_lock(&pr->pr_mtx);
1838 mtx_unlock(&pr->pr_mtx);
1840 if (!(flags & JAIL_ATTACH))
1841 sx_sunlock(&allprison_lock);
1847 prison_deref(pr, created
1848 ? PD_LOCKED | PD_LIST_XLOCKED
1849 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1852 sx_xunlock(&allprison_lock);
1858 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1859 if (errmsg_len > 0) {
1860 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1861 if (errmsg_pos > 0) {
1862 if (optuio->uio_segflg == UIO_SYSSPACE)
1864 optuio->uio_iov[errmsg_pos].iov_base,
1868 optuio->uio_iov[errmsg_pos].iov_base,
1875 free(ip4, M_PRISON);
1878 free(ip6, M_PRISON);
1881 free(g_path, M_TEMP);
1888 * struct jail_get_args {
1889 * struct iovec *iovp;
1890 * unsigned int iovcnt;
1895 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1900 /* Check that we have an even number of iovecs. */
1901 if (uap->iovcnt & 1)
1904 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1907 error = kern_jail_get(td, auio, uap->flags);
1909 error = copyout(auio->uio_iov, uap->iovp,
1910 uap->iovcnt * sizeof (struct iovec));
1916 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1918 struct prison *pr, *mypr;
1920 struct vfsoptlist *opts;
1921 char *errmsg, *name;
1922 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1924 if (flags & ~JAIL_GET_MASK)
1927 /* Get the parameter list. */
1928 error = vfs_buildopts(optuio, &opts);
1931 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1932 mypr = td->td_ucred->cr_prison;
1935 * Find the prison specified by one of: lastjid, jid, name.
1937 sx_slock(&allprison_lock);
1938 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1940 TAILQ_FOREACH(pr, &allprison, pr_list) {
1941 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1942 mtx_lock(&pr->pr_mtx);
1943 if (pr->pr_ref > 0 &&
1944 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1946 mtx_unlock(&pr->pr_mtx);
1952 vfs_opterror(opts, "no jail after %d", jid);
1953 goto done_unlock_list;
1954 } else if (error != ENOENT)
1955 goto done_unlock_list;
1957 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1960 pr = prison_find_child(mypr, jid);
1962 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1963 mtx_unlock(&pr->pr_mtx);
1965 vfs_opterror(opts, "jail %d is dying",
1967 goto done_unlock_list;
1972 vfs_opterror(opts, "jail %d not found", jid);
1973 goto done_unlock_list;
1975 } else if (error != ENOENT)
1976 goto done_unlock_list;
1978 error = vfs_getopt(opts, "name", (void **)&name, &len);
1980 if (len == 0 || name[len - 1] != '\0') {
1982 goto done_unlock_list;
1984 pr = prison_find_name(mypr, name);
1986 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1987 mtx_unlock(&pr->pr_mtx);
1989 vfs_opterror(opts, "jail \"%s\" is dying",
1991 goto done_unlock_list;
1996 vfs_opterror(opts, "jail \"%s\" not found", name);
1997 goto done_unlock_list;
1998 } else if (error != ENOENT)
1999 goto done_unlock_list;
2001 vfs_opterror(opts, "no jail specified");
2003 goto done_unlock_list;
2006 /* Get the parameters of the prison. */
2009 td->td_retval[0] = pr->pr_id;
2010 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2011 if (error != 0 && error != ENOENT)
2013 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2014 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2015 if (error != 0 && error != ENOENT)
2017 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2018 if (error != 0 && error != ENOENT)
2020 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2021 sizeof(pr->pr_cpuset->cs_id));
2022 if (error != 0 && error != ENOENT)
2024 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2025 if (error != 0 && error != ENOENT)
2028 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2029 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2030 if (error != 0 && error != ENOENT)
2034 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2035 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2036 if (error != 0 && error != ENOENT)
2039 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2040 sizeof(pr->pr_securelevel));
2041 if (error != 0 && error != ENOENT)
2043 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2044 sizeof(pr->pr_childcount));
2045 if (error != 0 && error != ENOENT)
2047 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2048 sizeof(pr->pr_childmax));
2049 if (error != 0 && error != ENOENT)
2051 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2052 if (error != 0 && error != ENOENT)
2054 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2055 if (error != 0 && error != ENOENT)
2057 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2058 if (error != 0 && error != ENOENT)
2060 #ifdef COMPAT_FREEBSD32
2061 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2062 uint32_t hid32 = pr->pr_hostid;
2064 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2067 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2068 sizeof(pr->pr_hostid));
2069 if (error != 0 && error != ENOENT)
2071 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2072 sizeof(pr->pr_enforce_statfs));
2073 if (error != 0 && error != ENOENT)
2075 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2076 sizeof(pr->pr_devfs_rsnum));
2077 if (error != 0 && error != ENOENT)
2079 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2081 if (pr_flag_names[fi] == NULL)
2083 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2084 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2085 if (error != 0 && error != ENOENT)
2088 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2089 if (error != 0 && error != ENOENT)
2092 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2095 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2096 i = pr_flag_jailsys[fi].disable &&
2097 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2098 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2101 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2102 if (error != 0 && error != ENOENT)
2105 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2107 if (pr_allow_names[fi] == NULL)
2109 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2110 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2111 if (error != 0 && error != ENOENT)
2114 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2115 if (error != 0 && error != ENOENT)
2118 i = (pr->pr_uref == 0);
2119 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2120 if (error != 0 && error != ENOENT)
2123 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2124 if (error != 0 && error != ENOENT)
2127 /* Get the module parameters. */
2128 mtx_unlock(&pr->pr_mtx);
2130 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2133 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2135 /* By now, all parameters should have been noted. */
2136 TAILQ_FOREACH(opt, opts, link) {
2137 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2139 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2144 /* Write the fetched parameters back to userspace. */
2146 TAILQ_FOREACH(opt, opts, link) {
2147 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2148 pos = 2 * opt->pos + 1;
2149 optuio->uio_iov[pos].iov_len = opt->len;
2150 if (opt->value != NULL) {
2151 if (optuio->uio_segflg == UIO_SYSSPACE) {
2153 optuio->uio_iov[pos].iov_base,
2156 error = copyout(opt->value,
2157 optuio->uio_iov[pos].iov_base,
2168 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2172 sx_sunlock(&allprison_lock);
2174 if (error && errmsg_pos >= 0) {
2175 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2176 errmsg_pos = 2 * errmsg_pos + 1;
2177 if (errmsg_len > 0) {
2178 if (optuio->uio_segflg == UIO_SYSSPACE)
2180 optuio->uio_iov[errmsg_pos].iov_base,
2184 optuio->uio_iov[errmsg_pos].iov_base,
2194 * struct jail_remove_args {
2199 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2201 struct prison *pr, *cpr, *lpr, *tpr;
2204 error = priv_check(td, PRIV_JAIL_REMOVE);
2208 sx_xlock(&allprison_lock);
2209 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2211 sx_xunlock(&allprison_lock);
2215 /* Remove all descendants of this prison, then remove this prison. */
2217 pr->pr_flags |= PR_REMOVE;
2218 if (!LIST_EMPTY(&pr->pr_children)) {
2219 mtx_unlock(&pr->pr_mtx);
2221 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2222 mtx_lock(&cpr->pr_mtx);
2223 if (cpr->pr_ref > 0) {
2226 cpr->pr_flags |= PR_REMOVE;
2228 /* Already removed - do not do it again. */
2231 mtx_unlock(&cpr->pr_mtx);
2233 mtx_lock(&lpr->pr_mtx);
2234 prison_remove_one(lpr);
2235 sx_xlock(&allprison_lock);
2240 mtx_lock(&lpr->pr_mtx);
2241 prison_remove_one(lpr);
2242 sx_xlock(&allprison_lock);
2244 mtx_lock(&pr->pr_mtx);
2246 prison_remove_one(pr);
2251 prison_remove_one(struct prison *pr)
2256 /* If the prison was persistent, it is not anymore. */
2258 if (pr->pr_flags & PR_PERSIST) {
2261 pr->pr_flags &= ~PR_PERSIST;
2265 * jail_remove added a reference. If that's the only one, remove
2268 KASSERT(pr->pr_ref > 0,
2269 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2270 if (pr->pr_ref == 1) {
2272 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2276 mtx_unlock(&pr->pr_mtx);
2277 sx_xunlock(&allprison_lock);
2279 * Kill all processes unfortunate enough to be attached to this prison.
2281 sx_slock(&allproc_lock);
2282 LIST_FOREACH(p, &allproc, p_list) {
2284 if (p->p_state != PRS_NEW && p->p_ucred &&
2285 p->p_ucred->cr_prison == pr)
2286 kern_psignal(p, SIGKILL);
2289 sx_sunlock(&allproc_lock);
2290 /* Remove the temporary reference added by jail_remove. */
2291 prison_deref(pr, deuref | PD_DEREF);
2296 * struct jail_attach_args {
2301 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2306 error = priv_check(td, PRIV_JAIL_ATTACH);
2310 sx_slock(&allprison_lock);
2311 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2313 sx_sunlock(&allprison_lock);
2318 * Do not allow a process to attach to a prison that is not
2319 * considered to be "alive".
2321 if (pr->pr_uref == 0) {
2322 mtx_unlock(&pr->pr_mtx);
2323 sx_sunlock(&allprison_lock);
2327 return (do_jail_attach(td, pr));
2331 do_jail_attach(struct thread *td, struct prison *pr)
2335 struct ucred *newcred, *oldcred;
2339 * XXX: Note that there is a slight race here if two threads
2340 * in the same privileged process attempt to attach to two
2341 * different jails at the same time. It is important for
2342 * user processes not to do this, or they might end up with
2343 * a process root from one prison, but attached to the jail
2348 mtx_unlock(&pr->pr_mtx);
2350 /* Let modules do whatever they need to prepare for attaching. */
2351 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2353 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2356 sx_sunlock(&allprison_lock);
2359 * Reparent the newly attached process to this jail.
2361 ppr = td->td_ucred->cr_prison;
2363 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2367 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2368 if ((error = change_dir(pr->pr_root, td)) != 0)
2371 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2374 VOP_UNLOCK(pr->pr_root, 0);
2375 if ((error = change_root(pr->pr_root, td)))
2380 oldcred = p->p_ucred;
2382 crcopy(newcred, oldcred);
2383 newcred->cr_prison = pr;
2384 p->p_ucred = newcred;
2387 racct_proc_ucred_changed(p, oldcred, newcred);
2390 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2393 VOP_UNLOCK(pr->pr_root, 0);
2395 /* Tell modules this thread is still in its old jail after all. */
2396 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2397 prison_deref(pr, PD_DEREF | PD_DEUREF);
2403 * Returns a locked prison instance, or NULL on failure.
2406 prison_find(int prid)
2410 sx_assert(&allprison_lock, SX_LOCKED);
2411 TAILQ_FOREACH(pr, &allprison, pr_list) {
2412 if (pr->pr_id == prid) {
2413 mtx_lock(&pr->pr_mtx);
2416 mtx_unlock(&pr->pr_mtx);
2423 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2426 prison_find_child(struct prison *mypr, int prid)
2431 sx_assert(&allprison_lock, SX_LOCKED);
2432 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2433 if (pr->pr_id == prid) {
2434 mtx_lock(&pr->pr_mtx);
2437 mtx_unlock(&pr->pr_mtx);
2444 * Look for the name relative to mypr. Returns a locked prison or NULL.
2447 prison_find_name(struct prison *mypr, const char *name)
2449 struct prison *pr, *deadpr;
2453 sx_assert(&allprison_lock, SX_LOCKED);
2454 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2457 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2458 if (!strcmp(pr->pr_name + mylen, name)) {
2459 mtx_lock(&pr->pr_mtx);
2460 if (pr->pr_ref > 0) {
2461 if (pr->pr_uref > 0)
2465 mtx_unlock(&pr->pr_mtx);
2468 /* There was no valid prison - perhaps there was a dying one. */
2469 if (deadpr != NULL) {
2470 mtx_lock(&deadpr->pr_mtx);
2471 if (deadpr->pr_ref == 0) {
2472 mtx_unlock(&deadpr->pr_mtx);
2480 * See if a prison has the specific flag set.
2483 prison_flag(struct ucred *cred, unsigned flag)
2486 /* This is an atomic read, so no locking is necessary. */
2487 return (cred->cr_prison->pr_flags & flag);
2491 prison_allow(struct ucred *cred, unsigned flag)
2494 /* This is an atomic read, so no locking is necessary. */
2495 return (cred->cr_prison->pr_allow & flag);
2499 * Remove a prison reference. If that was the last reference, remove the
2500 * prison itself - but not in this context in case there are locks held.
2503 prison_free_locked(struct prison *pr)
2506 mtx_assert(&pr->pr_mtx, MA_OWNED);
2508 if (pr->pr_ref == 0) {
2509 mtx_unlock(&pr->pr_mtx);
2510 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2511 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2514 mtx_unlock(&pr->pr_mtx);
2518 prison_free(struct prison *pr)
2521 mtx_lock(&pr->pr_mtx);
2522 prison_free_locked(pr);
2526 prison_complete(void *context, int pending)
2529 prison_deref((struct prison *)context, 0);
2533 * Remove a prison reference (usually). This internal version assumes no
2534 * mutexes are held, except perhaps the prison itself. If there are no more
2535 * references, release and delist the prison. On completion, the prison lock
2536 * and the allprison lock are both unlocked.
2539 prison_deref(struct prison *pr, int flags)
2541 struct prison *ppr, *tpr;
2543 if (!(flags & PD_LOCKED))
2544 mtx_lock(&pr->pr_mtx);
2546 if (flags & PD_DEUREF) {
2548 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2550 if (flags & PD_DEREF)
2552 /* If the prison still has references, nothing else to do. */
2553 if (pr->pr_ref > 0) {
2554 mtx_unlock(&pr->pr_mtx);
2555 if (flags & PD_LIST_SLOCKED)
2556 sx_sunlock(&allprison_lock);
2557 else if (flags & PD_LIST_XLOCKED)
2558 sx_xunlock(&allprison_lock);
2562 mtx_unlock(&pr->pr_mtx);
2563 if (flags & PD_LIST_SLOCKED) {
2564 if (!sx_try_upgrade(&allprison_lock)) {
2565 sx_sunlock(&allprison_lock);
2566 sx_xlock(&allprison_lock);
2568 } else if (!(flags & PD_LIST_XLOCKED))
2569 sx_xlock(&allprison_lock);
2571 TAILQ_REMOVE(&allprison, pr, pr_list);
2572 LIST_REMOVE(pr, pr_sibling);
2573 ppr = pr->pr_parent;
2574 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2575 tpr->pr_childcount--;
2576 sx_xunlock(&allprison_lock);
2579 if (pr->pr_vnet != ppr->pr_vnet)
2580 vnet_destroy(pr->pr_vnet);
2582 if (pr->pr_root != NULL)
2584 mtx_destroy(&pr->pr_mtx);
2586 free(pr->pr_ip4, M_PRISON);
2589 free(pr->pr_ip6, M_PRISON);
2591 if (pr->pr_cpuset != NULL)
2592 cpuset_rel(pr->pr_cpuset);
2595 prison_racct_detach(pr);
2599 /* Removing a prison frees a reference on its parent. */
2601 mtx_lock(&pr->pr_mtx);
2602 flags = PD_DEREF | PD_DEUREF;
2607 prison_hold_locked(struct prison *pr)
2610 mtx_assert(&pr->pr_mtx, MA_OWNED);
2611 KASSERT(pr->pr_ref > 0,
2612 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2617 prison_hold(struct prison *pr)
2620 mtx_lock(&pr->pr_mtx);
2621 prison_hold_locked(pr);
2622 mtx_unlock(&pr->pr_mtx);
2626 prison_proc_hold(struct prison *pr)
2629 mtx_lock(&pr->pr_mtx);
2630 KASSERT(pr->pr_uref > 0,
2631 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2633 mtx_unlock(&pr->pr_mtx);
2637 prison_proc_free(struct prison *pr)
2640 mtx_lock(&pr->pr_mtx);
2641 KASSERT(pr->pr_uref > 0,
2642 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2643 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2649 * Restrict a prison's IP address list with its parent's, possibly replacing
2650 * it. Return true if the replacement buffer was used (or would have been).
2653 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2658 ppr = pr->pr_parent;
2659 if (!(pr->pr_flags & PR_IP4_USER)) {
2660 /* This has no user settings, so just copy the parent's list. */
2661 if (pr->pr_ip4s < ppr->pr_ip4s) {
2663 * There's no room for the parent's list. Use the
2664 * new list buffer, which is assumed to be big enough
2665 * (if it was passed). If there's no buffer, try to
2669 if (newip4 == NULL) {
2670 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2671 M_PRISON, M_NOWAIT);
2675 if (newip4 != NULL) {
2676 bcopy(ppr->pr_ip4, newip4,
2677 ppr->pr_ip4s * sizeof(*newip4));
2678 free(pr->pr_ip4, M_PRISON);
2679 pr->pr_ip4 = newip4;
2680 pr->pr_ip4s = ppr->pr_ip4s;
2684 pr->pr_ip4s = ppr->pr_ip4s;
2685 if (pr->pr_ip4s > 0)
2686 bcopy(ppr->pr_ip4, pr->pr_ip4,
2687 pr->pr_ip4s * sizeof(*newip4));
2688 else if (pr->pr_ip4 != NULL) {
2689 free(pr->pr_ip4, M_PRISON);
2692 } else if (pr->pr_ip4s > 0) {
2693 /* Remove addresses that aren't in the parent. */
2694 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2695 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2697 if (ij < ppr->pr_ip4s)
2700 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2701 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2704 for (ij = 1; ii < pr->pr_ip4s; ) {
2705 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2709 switch (ij >= ppr->pr_ip4s ? -1 :
2710 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2712 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2713 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2724 if (pr->pr_ip4s == 0) {
2725 pr->pr_flags |= PR_IP4_DISABLE;
2726 free(pr->pr_ip4, M_PRISON);
2734 * Pass back primary IPv4 address of this jail.
2736 * If not restricted return success but do not alter the address. Caller has
2737 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2739 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2740 * Address returned in NBO.
2743 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2747 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2748 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2750 pr = cred->cr_prison;
2751 if (!(pr->pr_flags & PR_IP4))
2753 mtx_lock(&pr->pr_mtx);
2754 if (!(pr->pr_flags & PR_IP4)) {
2755 mtx_unlock(&pr->pr_mtx);
2758 if (pr->pr_ip4 == NULL) {
2759 mtx_unlock(&pr->pr_mtx);
2760 return (EAFNOSUPPORT);
2763 ia->s_addr = pr->pr_ip4[0].s_addr;
2764 mtx_unlock(&pr->pr_mtx);
2769 * Return 1 if we should do proper source address selection or are not jailed.
2770 * We will return 0 if we should bypass source address selection in favour
2771 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2773 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2776 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2782 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2783 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2788 pr = cred->cr_prison;
2789 if (pr->pr_flags & PR_IP4_SADDRSEL)
2792 lia.s_addr = INADDR_ANY;
2793 error = prison_get_ip4(cred, &lia);
2796 if (lia.s_addr == INADDR_ANY)
2799 ia->s_addr = lia.s_addr;
2804 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2807 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2814 * No need to lock since the PR_IP4_USER flag can't be altered for
2817 while (pr1 != &prison0 &&
2819 !(pr1->pr_flags & PR_VNET) &&
2821 !(pr1->pr_flags & PR_IP4_USER))
2822 pr1 = pr1->pr_parent;
2823 while (pr2 != &prison0 &&
2825 !(pr2->pr_flags & PR_VNET) &&
2827 !(pr2->pr_flags & PR_IP4_USER))
2828 pr2 = pr2->pr_parent;
2829 return (pr1 == pr2);
2833 * Make sure our (source) address is set to something meaningful to this
2836 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2837 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2838 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2841 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2847 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2848 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2850 pr = cred->cr_prison;
2851 if (!(pr->pr_flags & PR_IP4))
2853 mtx_lock(&pr->pr_mtx);
2854 if (!(pr->pr_flags & PR_IP4)) {
2855 mtx_unlock(&pr->pr_mtx);
2858 if (pr->pr_ip4 == NULL) {
2859 mtx_unlock(&pr->pr_mtx);
2860 return (EAFNOSUPPORT);
2863 ia0.s_addr = ntohl(ia->s_addr);
2864 if (ia0.s_addr == INADDR_LOOPBACK) {
2865 ia->s_addr = pr->pr_ip4[0].s_addr;
2866 mtx_unlock(&pr->pr_mtx);
2870 if (ia0.s_addr == INADDR_ANY) {
2872 * In case there is only 1 IPv4 address, bind directly.
2874 if (pr->pr_ip4s == 1)
2875 ia->s_addr = pr->pr_ip4[0].s_addr;
2876 mtx_unlock(&pr->pr_mtx);
2880 error = _prison_check_ip4(pr, ia);
2881 mtx_unlock(&pr->pr_mtx);
2886 * Rewrite destination address in case we will connect to loopback address.
2888 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2889 * Address passed in in NBO and returned in NBO.
2892 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2896 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2897 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2899 pr = cred->cr_prison;
2900 if (!(pr->pr_flags & PR_IP4))
2902 mtx_lock(&pr->pr_mtx);
2903 if (!(pr->pr_flags & PR_IP4)) {
2904 mtx_unlock(&pr->pr_mtx);
2907 if (pr->pr_ip4 == NULL) {
2908 mtx_unlock(&pr->pr_mtx);
2909 return (EAFNOSUPPORT);
2912 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2913 ia->s_addr = pr->pr_ip4[0].s_addr;
2914 mtx_unlock(&pr->pr_mtx);
2919 * Return success because nothing had to be changed.
2921 mtx_unlock(&pr->pr_mtx);
2926 * Check if given address belongs to the jail referenced by cred/prison.
2928 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2929 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2930 * doesn't allow IPv4. Address passed in in NBO.
2933 _prison_check_ip4(const struct prison *pr, const struct in_addr *ia)
2938 * Check the primary IP.
2940 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2944 * All the other IPs are sorted so we can do a binary search.
2947 z = pr->pr_ip4s - 2;
2950 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2959 return (EADDRNOTAVAIL);
2963 prison_check_ip4(const struct ucred *cred, const struct in_addr *ia)
2968 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2969 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2971 pr = cred->cr_prison;
2972 if (!(pr->pr_flags & PR_IP4))
2974 mtx_lock(&pr->pr_mtx);
2975 if (!(pr->pr_flags & PR_IP4)) {
2976 mtx_unlock(&pr->pr_mtx);
2979 if (pr->pr_ip4 == NULL) {
2980 mtx_unlock(&pr->pr_mtx);
2981 return (EAFNOSUPPORT);
2984 error = _prison_check_ip4(pr, ia);
2985 mtx_unlock(&pr->pr_mtx);
2992 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2997 ppr = pr->pr_parent;
2998 if (!(pr->pr_flags & PR_IP6_USER)) {
2999 /* This has no user settings, so just copy the parent's list. */
3000 if (pr->pr_ip6s < ppr->pr_ip6s) {
3002 * There's no room for the parent's list. Use the
3003 * new list buffer, which is assumed to be big enough
3004 * (if it was passed). If there's no buffer, try to
3008 if (newip6 == NULL) {
3009 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3010 M_PRISON, M_NOWAIT);
3014 if (newip6 != NULL) {
3015 bcopy(ppr->pr_ip6, newip6,
3016 ppr->pr_ip6s * sizeof(*newip6));
3017 free(pr->pr_ip6, M_PRISON);
3018 pr->pr_ip6 = newip6;
3019 pr->pr_ip6s = ppr->pr_ip6s;
3023 pr->pr_ip6s = ppr->pr_ip6s;
3024 if (pr->pr_ip6s > 0)
3025 bcopy(ppr->pr_ip6, pr->pr_ip6,
3026 pr->pr_ip6s * sizeof(*newip6));
3027 else if (pr->pr_ip6 != NULL) {
3028 free(pr->pr_ip6, M_PRISON);
3031 } else if (pr->pr_ip6s > 0) {
3032 /* Remove addresses that aren't in the parent. */
3033 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3034 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3037 if (ij < ppr->pr_ip6s)
3040 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3041 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3044 for (ij = 1; ii < pr->pr_ip6s; ) {
3045 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3050 switch (ij >= ppr->pr_ip6s ? -1 :
3051 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3053 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3054 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3065 if (pr->pr_ip6s == 0) {
3066 pr->pr_flags |= PR_IP6_DISABLE;
3067 free(pr->pr_ip6, M_PRISON);
3075 * Pass back primary IPv6 address for this jail.
3077 * If not restricted return success but do not alter the address. Caller has
3078 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3080 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3083 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3087 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3088 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3090 pr = cred->cr_prison;
3091 if (!(pr->pr_flags & PR_IP6))
3093 mtx_lock(&pr->pr_mtx);
3094 if (!(pr->pr_flags & PR_IP6)) {
3095 mtx_unlock(&pr->pr_mtx);
3098 if (pr->pr_ip6 == NULL) {
3099 mtx_unlock(&pr->pr_mtx);
3100 return (EAFNOSUPPORT);
3103 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3104 mtx_unlock(&pr->pr_mtx);
3109 * Return 1 if we should do proper source address selection or are not jailed.
3110 * We will return 0 if we should bypass source address selection in favour
3111 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3113 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3116 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3119 struct in6_addr lia6;
3122 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3123 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3128 pr = cred->cr_prison;
3129 if (pr->pr_flags & PR_IP6_SADDRSEL)
3133 error = prison_get_ip6(cred, &lia6);
3136 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3139 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3144 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3147 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3153 while (pr1 != &prison0 &&
3155 !(pr1->pr_flags & PR_VNET) &&
3157 !(pr1->pr_flags & PR_IP6_USER))
3158 pr1 = pr1->pr_parent;
3159 while (pr2 != &prison0 &&
3161 !(pr2->pr_flags & PR_VNET) &&
3163 !(pr2->pr_flags & PR_IP6_USER))
3164 pr2 = pr2->pr_parent;
3165 return (pr1 == pr2);
3169 * Make sure our (source) address is set to something meaningful to this jail.
3171 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3172 * when needed while binding.
3174 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3175 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3176 * doesn't allow IPv6.
3179 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3184 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3185 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3187 pr = cred->cr_prison;
3188 if (!(pr->pr_flags & PR_IP6))
3190 mtx_lock(&pr->pr_mtx);
3191 if (!(pr->pr_flags & PR_IP6)) {
3192 mtx_unlock(&pr->pr_mtx);
3195 if (pr->pr_ip6 == NULL) {
3196 mtx_unlock(&pr->pr_mtx);
3197 return (EAFNOSUPPORT);
3200 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3201 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3202 mtx_unlock(&pr->pr_mtx);
3206 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3208 * In case there is only 1 IPv6 address, and v6only is true,
3209 * then bind directly.
3211 if (v6only != 0 && pr->pr_ip6s == 1)
3212 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3213 mtx_unlock(&pr->pr_mtx);
3217 error = _prison_check_ip6(pr, ia6);
3218 mtx_unlock(&pr->pr_mtx);
3223 * Rewrite destination address in case we will connect to loopback address.
3225 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3228 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3232 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3233 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3235 pr = cred->cr_prison;
3236 if (!(pr->pr_flags & PR_IP6))
3238 mtx_lock(&pr->pr_mtx);
3239 if (!(pr->pr_flags & PR_IP6)) {
3240 mtx_unlock(&pr->pr_mtx);
3243 if (pr->pr_ip6 == NULL) {
3244 mtx_unlock(&pr->pr_mtx);
3245 return (EAFNOSUPPORT);
3248 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3249 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3250 mtx_unlock(&pr->pr_mtx);
3255 * Return success because nothing had to be changed.
3257 mtx_unlock(&pr->pr_mtx);
3262 * Check if given address belongs to the jail referenced by cred/prison.
3264 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3265 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3266 * doesn't allow IPv6.
3269 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3274 * Check the primary IP.
3276 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3280 * All the other IPs are sorted so we can do a binary search.
3283 z = pr->pr_ip6s - 2;
3286 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3295 return (EADDRNOTAVAIL);
3299 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3304 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3305 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3307 pr = cred->cr_prison;
3308 if (!(pr->pr_flags & PR_IP6))
3310 mtx_lock(&pr->pr_mtx);
3311 if (!(pr->pr_flags & PR_IP6)) {
3312 mtx_unlock(&pr->pr_mtx);
3315 if (pr->pr_ip6 == NULL) {
3316 mtx_unlock(&pr->pr_mtx);
3317 return (EAFNOSUPPORT);
3320 error = _prison_check_ip6(pr, ia6);
3321 mtx_unlock(&pr->pr_mtx);
3327 * Check if a jail supports the given address family.
3329 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3333 prison_check_af(struct ucred *cred, int af)
3338 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3340 pr = cred->cr_prison;
3342 /* Prisons with their own network stack are not limited. */
3343 if (prison_owns_vnet(cred))
3352 if (pr->pr_flags & PR_IP4)
3354 mtx_lock(&pr->pr_mtx);
3355 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3356 error = EAFNOSUPPORT;
3357 mtx_unlock(&pr->pr_mtx);
3363 if (pr->pr_flags & PR_IP6)
3365 mtx_lock(&pr->pr_mtx);
3366 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3367 error = EAFNOSUPPORT;
3368 mtx_unlock(&pr->pr_mtx);
3376 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3377 error = EAFNOSUPPORT;
3383 * Check if given address belongs to the jail referenced by cred (wrapper to
3384 * prison_check_ip[46]).
3386 * Returns 0 if jail doesn't restrict the address family or if address belongs
3387 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3388 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3391 prison_if(struct ucred *cred, struct sockaddr *sa)
3394 struct sockaddr_in *sai;
3397 struct sockaddr_in6 *sai6;
3401 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3402 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3405 if (prison_owns_vnet(cred))
3410 switch (sa->sa_family)
3414 sai = (struct sockaddr_in *)sa;
3415 error = prison_check_ip4(cred, &sai->sin_addr);
3420 sai6 = (struct sockaddr_in6 *)sa;
3421 error = prison_check_ip6(cred, &sai6->sin6_addr);
3425 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3426 error = EAFNOSUPPORT;
3432 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3435 prison_check(struct ucred *cred1, struct ucred *cred2)
3438 return ((cred1->cr_prison == cred2->cr_prison ||
3439 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3443 * Return 1 if p2 is a child of p1, otherwise 0.
3446 prison_ischild(struct prison *pr1, struct prison *pr2)
3449 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3456 * Return 1 if the passed credential is in a jail, otherwise 0.
3459 jailed(struct ucred *cred)
3462 return (cred->cr_prison != &prison0);
3466 * Return 1 if the passed credential is in a jail and that jail does not
3467 * have its own virtual network stack, otherwise 0.
3470 jailed_without_vnet(struct ucred *cred)
3476 if (prison_owns_vnet(cred))
3484 * Return the correct hostname (domainname, et al) for the passed credential.
3487 getcredhostname(struct ucred *cred, char *buf, size_t size)
3492 * A NULL credential can be used to shortcut to the physical
3493 * system's hostname.
3495 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3496 mtx_lock(&pr->pr_mtx);
3497 strlcpy(buf, pr->pr_hostname, size);
3498 mtx_unlock(&pr->pr_mtx);
3502 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3505 mtx_lock(&cred->cr_prison->pr_mtx);
3506 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3507 mtx_unlock(&cred->cr_prison->pr_mtx);
3511 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3514 mtx_lock(&cred->cr_prison->pr_mtx);
3515 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3516 mtx_unlock(&cred->cr_prison->pr_mtx);
3520 getcredhostid(struct ucred *cred, unsigned long *hostid)
3523 mtx_lock(&cred->cr_prison->pr_mtx);
3524 *hostid = cred->cr_prison->pr_hostid;
3525 mtx_unlock(&cred->cr_prison->pr_mtx);
3530 * Determine whether the prison represented by cred owns
3531 * its vnet rather than having it inherited.
3533 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3536 prison_owns_vnet(struct ucred *cred)
3540 * vnets cannot be added/removed after jail creation,
3541 * so no need to lock here.
3543 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3548 * Determine whether the subject represented by cred can "see"
3549 * status of a mount point.
3550 * Returns: 0 for permitted, ENOENT otherwise.
3551 * XXX: This function should be called cr_canseemount() and should be
3552 * placed in kern_prot.c.
3555 prison_canseemount(struct ucred *cred, struct mount *mp)
3561 pr = cred->cr_prison;
3562 if (pr->pr_enforce_statfs == 0)
3564 if (pr->pr_root->v_mount == mp)
3566 if (pr->pr_enforce_statfs == 2)
3569 * If jail's chroot directory is set to "/" we should be able to see
3570 * all mount-points from inside a jail.
3571 * This is ugly check, but this is the only situation when jail's
3572 * directory ends with '/'.
3574 if (strcmp(pr->pr_path, "/") == 0)
3576 len = strlen(pr->pr_path);
3578 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3581 * Be sure that we don't have situation where jail's root directory
3582 * is "/some/path" and mount point is "/some/pathpath".
3584 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3590 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3592 char jpath[MAXPATHLEN];
3596 pr = cred->cr_prison;
3597 if (pr->pr_enforce_statfs == 0)
3599 if (prison_canseemount(cred, mp) != 0) {
3600 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3601 strlcpy(sp->f_mntonname, "[restricted]",
3602 sizeof(sp->f_mntonname));
3605 if (pr->pr_root->v_mount == mp) {
3607 * Clear current buffer data, so we are sure nothing from
3608 * the valid path left there.
3610 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3611 *sp->f_mntonname = '/';
3615 * If jail's chroot directory is set to "/" we should be able to see
3616 * all mount-points from inside a jail.
3618 if (strcmp(pr->pr_path, "/") == 0)
3620 len = strlen(pr->pr_path);
3621 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3623 * Clear current buffer data, so we are sure nothing from
3624 * the valid path left there.
3626 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3627 if (*jpath == '\0') {
3628 /* Should never happen. */
3629 *sp->f_mntonname = '/';
3631 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3636 * Check with permission for a specific privilege is granted within jail. We
3637 * have a specific list of accepted privileges; the rest are denied.
3640 prison_priv_check(struct ucred *cred, int priv)
3648 * Privileges specific to prisons with a virtual network stack.
3649 * There might be a duplicate entry here in case the privilege
3650 * is only granted conditionally in the legacy jail case.
3655 * NFS-specific privileges.
3657 case PRIV_NFS_DAEMON:
3658 case PRIV_NFS_LOCKD:
3661 * Network stack privileges.
3663 case PRIV_NET_BRIDGE:
3666 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3667 case PRIV_NET_ROUTE:
3669 case PRIV_NET_SETIFMTU:
3670 case PRIV_NET_SETIFFLAGS:
3671 case PRIV_NET_SETIFCAP:
3672 case PRIV_NET_SETIFDESCR:
3673 case PRIV_NET_SETIFNAME :
3674 case PRIV_NET_SETIFMETRIC:
3675 case PRIV_NET_SETIFPHYS:
3676 case PRIV_NET_SETIFMAC:
3677 case PRIV_NET_ADDMULTI:
3678 case PRIV_NET_DELMULTI:
3679 case PRIV_NET_HWIOCTL:
3680 case PRIV_NET_SETLLADDR:
3681 case PRIV_NET_ADDIFGROUP:
3682 case PRIV_NET_DELIFGROUP:
3683 case PRIV_NET_IFCREATE:
3684 case PRIV_NET_IFDESTROY:
3685 case PRIV_NET_ADDIFADDR:
3686 case PRIV_NET_DELIFADDR:
3689 case PRIV_NET_SETIFVNET:
3690 case PRIV_NET_SETIFFIB:
3693 * 802.11-related privileges.
3695 case PRIV_NET80211_GETKEY:
3697 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3702 * AppleTalk privileges.
3704 case PRIV_NETATALK_RESERVEDPORT:
3709 case PRIV_NETATM_CFG:
3710 case PRIV_NETATM_ADD:
3711 case PRIV_NETATM_DEL:
3712 case PRIV_NETATM_SET:
3715 * Bluetooth privileges.
3717 case PRIV_NETBLUETOOTH_RAW:
3721 * Netgraph and netgraph module privileges.
3723 case PRIV_NETGRAPH_CONTROL:
3725 case PRIV_NETGRAPH_TTY:
3729 * IPv4 and IPv6 privileges.
3731 case PRIV_NETINET_IPFW:
3732 case PRIV_NETINET_DIVERT:
3733 case PRIV_NETINET_PF:
3734 case PRIV_NETINET_DUMMYNET:
3735 case PRIV_NETINET_CARP:
3736 case PRIV_NETINET_MROUTE:
3737 case PRIV_NETINET_RAW:
3738 case PRIV_NETINET_ADDRCTRL6:
3739 case PRIV_NETINET_ND6:
3740 case PRIV_NETINET_SCOPE6:
3741 case PRIV_NETINET_ALIFETIME6:
3742 case PRIV_NETINET_IPSEC:
3743 case PRIV_NETINET_BINDANY:
3747 * IPX/SPX privileges.
3749 case PRIV_NETIPX_RESERVEDPORT:
3750 case PRIV_NETIPX_RAW:
3764 * No default: or deny here.
3765 * In case of no permit fall through to next switch().
3767 if (cred->cr_prison->pr_flags & PR_VNET)
3775 * Allow ktrace privileges for root in jail.
3781 * Allow jailed processes to configure audit identity and
3782 * submit audit records (login, etc). In the future we may
3783 * want to further refine the relationship between audit and
3786 case PRIV_AUDIT_GETAUDIT:
3787 case PRIV_AUDIT_SETAUDIT:
3788 case PRIV_AUDIT_SUBMIT:
3792 * Allow jailed processes to manipulate process UNIX
3793 * credentials in any way they see fit.
3795 case PRIV_CRED_SETUID:
3796 case PRIV_CRED_SETEUID:
3797 case PRIV_CRED_SETGID:
3798 case PRIV_CRED_SETEGID:
3799 case PRIV_CRED_SETGROUPS:
3800 case PRIV_CRED_SETREUID:
3801 case PRIV_CRED_SETREGID:
3802 case PRIV_CRED_SETRESUID:
3803 case PRIV_CRED_SETRESGID:
3806 * Jail implements visibility constraints already, so allow
3807 * jailed root to override uid/gid-based constraints.
3809 case PRIV_SEEOTHERGIDS:
3810 case PRIV_SEEOTHERUIDS:
3813 * Jail implements inter-process debugging limits already, so
3814 * allow jailed root various debugging privileges.
3816 case PRIV_DEBUG_DIFFCRED:
3817 case PRIV_DEBUG_SUGID:
3818 case PRIV_DEBUG_UNPRIV:
3821 * Allow jail to set various resource limits and login
3822 * properties, and for now, exceed process resource limits.
3824 case PRIV_PROC_LIMIT:
3825 case PRIV_PROC_SETLOGIN:
3826 case PRIV_PROC_SETRLIMIT:
3829 * System V and POSIX IPC privileges are granted in jail.
3832 case PRIV_IPC_WRITE:
3833 case PRIV_IPC_ADMIN:
3834 case PRIV_IPC_MSGSIZE:
3838 * Jail operations within a jail work on child jails.
3840 case PRIV_JAIL_ATTACH:
3842 case PRIV_JAIL_REMOVE:
3845 * Jail implements its own inter-process limits, so allow
3846 * root processes in jail to change scheduling on other
3847 * processes in the same jail. Likewise for signalling.
3849 case PRIV_SCHED_DIFFCRED:
3850 case PRIV_SCHED_CPUSET:
3851 case PRIV_SIGNAL_DIFFCRED:
3852 case PRIV_SIGNAL_SUGID:
3855 * Allow jailed processes to write to sysctls marked as jail
3858 case PRIV_SYSCTL_WRITEJAIL:
3861 * Allow root in jail to manage a variety of quota
3862 * properties. These should likely be conditional on a
3863 * configuration option.
3865 case PRIV_VFS_GETQUOTA:
3866 case PRIV_VFS_SETQUOTA:
3869 * Since Jail relies on chroot() to implement file system
3870 * protections, grant many VFS privileges to root in jail.
3871 * Be careful to exclude mount-related and NFS-related
3875 case PRIV_VFS_WRITE:
3876 case PRIV_VFS_ADMIN:
3878 case PRIV_VFS_LOOKUP:
3879 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3880 case PRIV_VFS_CHFLAGS_DEV:
3881 case PRIV_VFS_CHOWN:
3882 case PRIV_VFS_CHROOT:
3883 case PRIV_VFS_RETAINSUGID:
3884 case PRIV_VFS_FCHROOT:
3886 case PRIV_VFS_SETGID:
3888 case PRIV_VFS_STICKYFILE:
3891 * As in the non-jail case, non-root users are expected to be
3892 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3893 * exists in the jail and they have permission to access it).
3895 case PRIV_KMEM_READ:
3899 * Depending on the global setting, allow privilege of
3900 * setting system flags.
3902 case PRIV_VFS_SYSFLAGS:
3903 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3909 * Depending on the global setting, allow privilege of
3910 * mounting/unmounting file systems.
3912 case PRIV_VFS_MOUNT:
3913 case PRIV_VFS_UNMOUNT:
3914 case PRIV_VFS_MOUNT_NONUSER:
3915 case PRIV_VFS_MOUNT_OWNER:
3916 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3917 cred->cr_prison->pr_enforce_statfs < 2)
3923 * Allow jailed root to bind reserved ports and reuse in-use
3926 case PRIV_NETINET_RESERVEDPORT:
3927 case PRIV_NETINET_REUSEPORT:
3931 * Allow jailed root to set certian IPv4/6 (option) headers.
3933 case PRIV_NETINET_SETHDROPTS:
3937 * Conditionally allow creating raw sockets in jail.
3939 case PRIV_NETINET_RAW:
3940 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3946 * Since jail implements its own visibility limits on netstat
3947 * sysctls, allow getcred. This allows identd to work in
3950 case PRIV_NETINET_GETCRED:
3954 * Allow jailed root to set loginclass.
3956 case PRIV_PROC_SETLOGINCLASS:
3961 * In all remaining cases, deny the privilege request. This
3962 * includes almost all network privileges, many system
3963 * configuration privileges.
3970 * Return the part of pr2's name that is relative to pr1, or the whole name
3971 * if it does not directly follow.
3975 prison_name(struct prison *pr1, struct prison *pr2)
3979 /* Jails see themselves as "0" (if they see themselves at all). */
3982 name = pr2->pr_name;
3983 if (prison_ischild(pr1, pr2)) {
3985 * pr1 isn't locked (and allprison_lock may not be either)
3986 * so its length can't be counted on. But the number of dots
3987 * can be counted on - and counted.
3989 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3990 name = strchr(name, '.') + 1;
3996 * Return the part of pr2's path that is relative to pr1, or the whole path
3997 * if it does not directly follow.
4000 prison_path(struct prison *pr1, struct prison *pr2)
4002 char *path1, *path2;
4005 path1 = pr1->pr_path;
4006 path2 = pr2->pr_path;
4007 if (!strcmp(path1, "/"))
4009 len1 = strlen(path1);
4010 if (strncmp(path1, path2, len1))
4012 if (path2[len1] == '\0')
4014 if (path2[len1] == '/')
4015 return (path2 + len1);
4021 * Jail-related sysctls.
4023 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4027 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4030 struct prison *pr, *cpr;
4032 struct in_addr *ip4 = NULL;
4036 struct in6_addr *ip6 = NULL;
4041 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4042 pr = req->td->td_ucred->cr_prison;
4044 sx_slock(&allprison_lock);
4045 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4046 #if defined(INET) || defined(INET6)
4049 mtx_lock(&cpr->pr_mtx);
4051 if (cpr->pr_ip4s > 0) {
4052 if (ip4s < cpr->pr_ip4s) {
4053 ip4s = cpr->pr_ip4s;
4054 mtx_unlock(&cpr->pr_mtx);
4055 ip4 = realloc(ip4, ip4s *
4056 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4059 bcopy(cpr->pr_ip4, ip4,
4060 cpr->pr_ip4s * sizeof(struct in_addr));
4064 if (cpr->pr_ip6s > 0) {
4065 if (ip6s < cpr->pr_ip6s) {
4066 ip6s = cpr->pr_ip6s;
4067 mtx_unlock(&cpr->pr_mtx);
4068 ip6 = realloc(ip6, ip6s *
4069 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4072 bcopy(cpr->pr_ip6, ip6,
4073 cpr->pr_ip6s * sizeof(struct in6_addr));
4076 if (cpr->pr_ref == 0) {
4077 mtx_unlock(&cpr->pr_mtx);
4080 bzero(xp, sizeof(*xp));
4081 xp->pr_version = XPRISON_VERSION;
4082 xp->pr_id = cpr->pr_id;
4083 xp->pr_state = cpr->pr_uref > 0
4084 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4085 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4086 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4087 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4089 xp->pr_ip4s = cpr->pr_ip4s;
4092 xp->pr_ip6s = cpr->pr_ip6s;
4094 mtx_unlock(&cpr->pr_mtx);
4095 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4099 if (xp->pr_ip4s > 0) {
4100 error = SYSCTL_OUT(req, ip4,
4101 xp->pr_ip4s * sizeof(struct in_addr));
4107 if (xp->pr_ip6s > 0) {
4108 error = SYSCTL_OUT(req, ip6,
4109 xp->pr_ip6s * sizeof(struct in6_addr));
4115 sx_sunlock(&allprison_lock);
4126 SYSCTL_OID(_security_jail, OID_AUTO, list,
4127 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4128 sysctl_jail_list, "S", "List of active jails");
4131 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4135 injail = jailed(req->td->td_ucred);
4136 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4141 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4142 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4143 sysctl_jail_jailed, "I", "Process in jail?");
4146 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4148 int error, havevnet;
4150 struct ucred *cred = req->td->td_ucred;
4152 havevnet = jailed(cred) && prison_owns_vnet(cred);
4156 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4161 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4162 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4163 sysctl_jail_vnet, "I", "Jail owns VNET?");
4165 #if defined(INET) || defined(INET6)
4166 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4167 &jail_max_af_ips, 0,
4168 "Number of IP addresses a jail may have at most per address family");
4172 * Default parameters for jail(2) compatability. For historical reasons,
4173 * the sysctl names have varying similarity to the parameter names. Prisons
4174 * just see their own parameters, and can't change them.
4177 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4180 int allow, error, i;
4182 pr = req->td->td_ucred->cr_prison;
4183 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4185 /* Get the current flag value, and convert it to a boolean. */
4186 i = (allow & arg2) ? 1 : 0;
4189 error = sysctl_handle_int(oidp, &i, 0, req);
4190 if (error || !req->newptr)
4196 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4199 mtx_lock(&prison0.pr_mtx);
4200 jail_default_allow = (jail_default_allow & ~arg2) | i;
4201 mtx_unlock(&prison0.pr_mtx);
4205 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4206 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4207 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4208 "Processes in jail can set their hostnames");
4209 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4210 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4211 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4212 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4213 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4214 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4215 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4216 "Processes in jail can use System V IPC primitives");
4217 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4218 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4219 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4220 "Prison root can create raw sockets");
4221 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4222 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4223 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4224 "Processes in jail can alter system file flags");
4225 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4226 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4227 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4228 "Processes in jail can mount/unmount jail-friendly file systems");
4229 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4230 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4231 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4232 "Processes in jail can mount the devfs file system");
4233 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4234 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4235 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4236 "Processes in jail can mount the nullfs file system");
4237 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4238 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4239 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4240 "Processes in jail can mount the procfs file system");
4241 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4242 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4243 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4244 "Processes in jail can mount the tmpfs file system");
4245 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4246 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4247 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4248 "Processes in jail can mount the zfs file system");
4251 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4256 pr = req->td->td_ucred->cr_prison;
4257 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4258 error = sysctl_handle_int(oidp, &level, 0, req);
4259 if (error || !req->newptr)
4261 *(int *)arg1 = level;
4265 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4266 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4267 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4268 sysctl_jail_default_level, "I",
4269 "Processes in jail cannot see all mounted file systems");
4271 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4272 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4273 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4274 sysctl_jail_default_level, "I",
4275 "Ruleset for the devfs filesystem in jail");
4278 * Nodes to describe jail parameters. Maximum length of string parameters
4279 * is returned in the string itself, and the other parameters exist merely
4280 * to make themselves and their types known.
4282 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4286 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4293 switch (oidp->oid_kind & CTLTYPE)
4299 if (!(req->flags & SCTL_MASK32))
4301 return (SYSCTL_OUT(req, &l, sizeof(l)));
4305 return (SYSCTL_OUT(req, &i, sizeof(i)));
4306 case CTLTYPE_STRING:
4307 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4309 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4310 case CTLTYPE_STRUCT:
4312 return (SYSCTL_OUT(req, &s, sizeof(s)));
4317 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4318 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4319 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4320 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4321 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4322 "I", "Jail secure level");
4323 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4324 "I", "Jail cannot see all mounted file systems");
4325 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4326 "I", "Ruleset for in-jail devfs mounts");
4327 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4328 "B", "Jail persistence");
4330 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4331 "E,jailsys", "Virtual network stack");
4333 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4334 "B", "Jail is in the process of shutting down");
4336 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4337 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4338 "I", "Current number of child jails");
4339 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4340 "I", "Maximum number of child jails");
4342 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4343 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4345 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4346 "Jail NIS domainname");
4347 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4349 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4350 "LU", "Jail host ID");
4352 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4353 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4356 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4357 "Jail IPv4 address virtualization");
4358 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4359 "S,in_addr,a", "Jail IPv4 addresses");
4360 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4361 "B", "Do (not) use IPv4 source address selection rather than the "
4362 "primary jail IPv4 address.");
4365 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4366 "Jail IPv6 address virtualization");
4367 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4368 "S,in6_addr,a", "Jail IPv6 addresses");
4369 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4370 "B", "Do (not) use IPv6 source address selection rather than the "
4371 "primary jail IPv6 address.");
4374 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4375 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4376 "B", "Jail may set hostname");
4377 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4378 "B", "Jail may use SYSV IPC");
4379 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4380 "B", "Jail may create raw sockets");
4381 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4382 "B", "Jail may alter system file flags");
4383 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4384 "B", "Jail may set file quotas");
4385 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4386 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4388 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4389 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4390 "B", "Jail may mount/unmount jail-friendly file systems in general");
4391 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4392 "B", "Jail may mount the devfs file system");
4393 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4394 "B", "Jail may mount the nullfs file system");
4395 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4396 "B", "Jail may mount the procfs file system");
4397 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4398 "B", "Jail may mount the tmpfs file system");
4399 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4400 "B", "Jail may mount the zfs file system");
4403 prison_racct_foreach(void (*callback)(struct racct *racct,
4404 void *arg2, void *arg3), void *arg2, void *arg3)
4406 struct prison_racct *prr;
4408 sx_slock(&allprison_lock);
4409 LIST_FOREACH(prr, &allprison_racct, prr_next)
4410 (callback)(prr->prr_racct, arg2, arg3);
4411 sx_sunlock(&allprison_lock);
4414 static struct prison_racct *
4415 prison_racct_find_locked(const char *name)
4417 struct prison_racct *prr;
4419 sx_assert(&allprison_lock, SA_XLOCKED);
4421 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4424 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4425 if (strcmp(name, prr->prr_name) != 0)
4428 /* Found prison_racct with a matching name? */
4429 prison_racct_hold(prr);
4433 /* Add new prison_racct. */
4434 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4435 racct_create(&prr->prr_racct);
4437 strcpy(prr->prr_name, name);
4438 refcount_init(&prr->prr_refcount, 1);
4439 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4444 struct prison_racct *
4445 prison_racct_find(const char *name)
4447 struct prison_racct *prr;
4449 sx_xlock(&allprison_lock);
4450 prr = prison_racct_find_locked(name);
4451 sx_xunlock(&allprison_lock);
4456 prison_racct_hold(struct prison_racct *prr)
4459 refcount_acquire(&prr->prr_refcount);
4463 prison_racct_free_locked(struct prison_racct *prr)
4466 sx_assert(&allprison_lock, SA_XLOCKED);
4468 if (refcount_release(&prr->prr_refcount)) {
4469 racct_destroy(&prr->prr_racct);
4470 LIST_REMOVE(prr, prr_next);
4471 free(prr, M_PRISON_RACCT);
4476 prison_racct_free(struct prison_racct *prr)
4480 sx_assert(&allprison_lock, SA_UNLOCKED);
4482 old = prr->prr_refcount;
4483 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4486 sx_xlock(&allprison_lock);
4487 prison_racct_free_locked(prr);
4488 sx_xunlock(&allprison_lock);
4493 prison_racct_attach(struct prison *pr)
4495 struct prison_racct *prr;
4497 sx_assert(&allprison_lock, SA_XLOCKED);
4499 prr = prison_racct_find_locked(pr->pr_name);
4500 KASSERT(prr != NULL, ("cannot find prison_racct"));
4502 pr->pr_prison_racct = prr;
4506 * Handle jail renaming. From the racct point of view, renaming means
4507 * moving from one prison_racct to another.
4510 prison_racct_modify(struct prison *pr)
4514 struct prison_racct *oldprr;
4516 sx_slock(&allproc_lock);
4517 sx_xlock(&allprison_lock);
4519 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4520 sx_xunlock(&allprison_lock);
4521 sx_sunlock(&allproc_lock);
4525 oldprr = pr->pr_prison_racct;
4526 pr->pr_prison_racct = NULL;
4528 prison_racct_attach(pr);
4531 * Move resource utilisation records.
4533 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4536 * Force rctl to reattach rules to processes.
4538 FOREACH_PROC_IN_SYSTEM(p) {
4540 cred = crhold(p->p_ucred);
4542 racct_proc_ucred_changed(p, cred, cred);
4546 sx_sunlock(&allproc_lock);
4547 prison_racct_free_locked(oldprr);
4548 sx_xunlock(&allprison_lock);
4552 prison_racct_detach(struct prison *pr)
4555 sx_assert(&allprison_lock, SA_UNLOCKED);
4557 if (pr->pr_prison_racct == NULL)
4559 prison_racct_free(pr->pr_prison_racct);
4560 pr->pr_prison_racct = NULL;
4567 db_show_prison(struct prison *pr)
4570 #if defined(INET) || defined(INET6)
4575 char ip6buf[INET6_ADDRSTRLEN];
4578 db_printf("prison %p:\n", pr);
4579 db_printf(" jid = %d\n", pr->pr_id);
4580 db_printf(" name = %s\n", pr->pr_name);
4581 db_printf(" parent = %p\n", pr->pr_parent);
4582 db_printf(" ref = %d\n", pr->pr_ref);
4583 db_printf(" uref = %d\n", pr->pr_uref);
4584 db_printf(" path = %s\n", pr->pr_path);
4585 db_printf(" cpuset = %d\n", pr->pr_cpuset
4586 ? pr->pr_cpuset->cs_id : -1);
4588 db_printf(" vnet = %p\n", pr->pr_vnet);
4590 db_printf(" root = %p\n", pr->pr_root);
4591 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4592 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4593 db_printf(" children.max = %d\n", pr->pr_childmax);
4594 db_printf(" children.cur = %d\n", pr->pr_childcount);
4595 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4596 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4597 db_printf(" flags = 0x%x", pr->pr_flags);
4598 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4600 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4601 db_printf(" %s", pr_flag_names[fi]);
4602 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4604 jsf = pr->pr_flags &
4605 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4606 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4607 pr_flag_jailsys[fi].disable &&
4608 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4609 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4612 db_printf(" allow = 0x%x", pr->pr_allow);
4613 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4615 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4616 db_printf(" %s", pr_allow_names[fi]);
4618 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4619 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4620 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4621 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4622 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4624 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4625 for (ii = 0; ii < pr->pr_ip4s; ii++)
4626 db_printf(" %s %s\n",
4627 ii == 0 ? "ip4.addr =" : " ",
4628 inet_ntoa(pr->pr_ip4[ii]));
4631 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4632 for (ii = 0; ii < pr->pr_ip6s; ii++)
4633 db_printf(" %s %s\n",
4634 ii == 0 ? "ip6.addr =" : " ",
4635 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4639 DB_SHOW_COMMAND(prison, db_show_prison_command)
4645 * Show all prisons in the list, and prison0 which is not
4648 db_show_prison(&prison0);
4649 if (!db_pager_quit) {
4650 TAILQ_FOREACH(pr, &allprison, pr_list) {
4662 /* Look for a prison with the ID and with references. */
4663 TAILQ_FOREACH(pr, &allprison, pr_list)
4664 if (pr->pr_id == addr && pr->pr_ref > 0)
4667 /* Look again, without requiring a reference. */
4668 TAILQ_FOREACH(pr, &allprison, pr_list)
4669 if (pr->pr_id == addr)
4672 /* Assume address points to a valid prison. */
4673 pr = (struct prison *)addr;