2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_compat.h"
35 #include "opt_inet6.h"
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
67 #include <netinet/in.h>
72 #include <netinet6/in6_var.h>
76 #include <security/mac/mac_framework.h>
78 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
81 MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
92 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
94 #define _PR_IP_SADDRSEL 0
98 /* prison0 describes what is "real" about the system. */
99 struct prison prison0 = {
105 .pr_securelevel = -1,
107 .pr_childmax = JAIL_MAX,
108 .pr_hostuuid = DEFAULT_HOSTUUID,
109 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
111 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
113 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
115 .pr_allow = PR_ALLOW_ALL,
117 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
119 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
120 struct sx allprison_lock;
121 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
122 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
123 LIST_HEAD(, prison_racct) allprison_racct;
126 static int do_jail_attach(struct thread *td, struct prison *pr);
127 static void prison_complete(void *context, int pending);
128 static void prison_deref(struct prison *pr, int flags);
129 static char *prison_path(struct prison *pr1, struct prison *pr2);
130 static void prison_remove_one(struct prison *pr);
132 static void prison_racct_attach(struct prison *pr);
133 static void prison_racct_detach(struct prison *pr);
136 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
137 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
140 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
141 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
144 /* Flags for prison_deref */
145 #define PD_DEREF 0x01
146 #define PD_DEUREF 0x02
147 #define PD_LOCKED 0x04
148 #define PD_LIST_SLOCKED 0x08
149 #define PD_LIST_XLOCKED 0x10
152 * Parameter names corresponding to PR_* flag values. Size values are for kvm
153 * as we cannot figure out the size of a sparse array, or an array without a
156 static char *pr_flag_names[] = {
159 [7] = "ip4.saddrsel",
162 [8] = "ip6.saddrsel",
165 const size_t pr_flag_names_size = sizeof(pr_flag_names);
167 static char *pr_flag_nonames[] = {
170 [7] = "ip4.nosaddrsel",
173 [8] = "ip6.nosaddrsel",
176 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
178 struct jailsys_flags {
182 } pr_flag_jailsys[] = {
183 { "host", 0, PR_HOST },
185 { "vnet", 0, PR_VNET },
188 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
191 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
194 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
196 static char *pr_allow_names[] = {
197 "allow.set_hostname",
205 "allow.mount.nullfs",
207 "allow.mount.procfs",
209 const size_t pr_allow_names_size = sizeof(pr_allow_names);
211 static char *pr_allow_nonames[] = {
212 "allow.noset_hostname",
214 "allow.noraw_sockets",
219 "allow.mount.nodevfs",
220 "allow.mount.nonullfs",
222 "allow.mount.noprocfs",
224 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
226 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
227 #define JAIL_DEFAULT_ENFORCE_STATFS 2
228 #define JAIL_DEFAULT_DEVFS_RSNUM 0
229 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
230 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
231 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
232 #if defined(INET) || defined(INET6)
233 static unsigned jail_max_af_ips = 255;
238 qcmp_v4(const void *ip1, const void *ip2)
243 * We need to compare in HBO here to get the list sorted as expected
244 * by the result of the code. Sorting NBO addresses gives you
245 * interesting results. If you do not understand, do not try.
247 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
248 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
251 * Do not simply return the difference of the two numbers, the int is
265 qcmp_v6(const void *ip1, const void *ip2)
267 const struct in6_addr *ia6a, *ia6b;
270 ia6a = (const struct in6_addr *)ip1;
271 ia6b = (const struct in6_addr *)ip2;
274 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
275 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
277 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
290 sys_jail(struct thread *td, struct jail_args *uap)
296 error = copyin(uap->jail, &version, sizeof(uint32_t));
305 /* FreeBSD single IPv4 jails. */
306 bzero(&j, sizeof(struct jail));
307 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
310 j.version = j0.version;
312 j.hostname = j0.hostname;
313 j.ip4s = j0.ip_number;
319 * Version 1 was used by multi-IPv4 jail implementations
320 * that never made it into the official kernel.
324 case 2: /* JAIL_API_VERSION */
325 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
326 error = copyin(uap->jail, &j, sizeof(struct jail));
332 /* Sci-Fi jails are not supported, sorry. */
335 return (kern_jail(td, &j));
339 kern_jail(struct thread *td, struct jail *j)
341 struct iovec optiov[2 * (4
342 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
351 char *u_path, *u_hostname, *u_name;
354 struct in_addr *u_ip4;
357 struct in6_addr *u_ip6;
360 int error, enforce_statfs, fi;
362 bzero(&optiov, sizeof(optiov));
363 opt.uio_iov = optiov;
367 opt.uio_segflg = UIO_SYSSPACE;
368 opt.uio_rw = UIO_READ;
371 /* Set permissions for top-level jails from sysctls. */
372 if (!jailed(td->td_ucred)) {
373 for (fi = 0; fi < sizeof(pr_allow_names) /
374 sizeof(pr_allow_names[0]); fi++) {
375 optiov[opt.uio_iovcnt].iov_base =
376 (jail_default_allow & (1 << fi))
377 ? pr_allow_names[fi] : pr_allow_nonames[fi];
378 optiov[opt.uio_iovcnt].iov_len =
379 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
382 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
383 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
385 enforce_statfs = jail_default_enforce_statfs;
386 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
387 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
391 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
393 ip4s = (j->version == 0) ? 1 : j->ip4s;
394 if (ip4s > jail_max_af_ips)
396 tmplen += ip4s * sizeof(struct in_addr);
402 if (j->ip6s > jail_max_af_ips)
404 tmplen += j->ip6s * sizeof(struct in6_addr);
409 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
410 u_hostname = u_path + MAXPATHLEN;
411 u_name = u_hostname + MAXHOSTNAMELEN;
413 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
417 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
419 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
422 optiov[opt.uio_iovcnt].iov_base = "path";
423 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
425 optiov[opt.uio_iovcnt].iov_base = u_path;
426 error = copyinstr(j->path, u_path, MAXPATHLEN,
427 &optiov[opt.uio_iovcnt].iov_len);
429 free(u_path, M_TEMP);
433 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
434 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
436 optiov[opt.uio_iovcnt].iov_base = u_hostname;
437 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
438 &optiov[opt.uio_iovcnt].iov_len);
440 free(u_path, M_TEMP);
444 if (j->jailname != NULL) {
445 optiov[opt.uio_iovcnt].iov_base = "name";
446 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
448 optiov[opt.uio_iovcnt].iov_base = u_name;
449 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
450 &optiov[opt.uio_iovcnt].iov_len);
452 free(u_path, M_TEMP);
458 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
459 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
461 optiov[opt.uio_iovcnt].iov_base = u_ip4;
462 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
464 u_ip4->s_addr = j->ip4s;
466 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
468 free(u_path, M_TEMP);
475 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
476 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
478 optiov[opt.uio_iovcnt].iov_base = u_ip6;
479 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
480 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
482 free(u_path, M_TEMP);
487 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
488 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
489 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
490 free(u_path, M_TEMP);
496 * struct jail_set_args {
497 * struct iovec *iovp;
498 * unsigned int iovcnt;
503 sys_jail_set(struct thread *td, struct jail_set_args *uap)
508 /* Check that we have an even number of iovecs. */
512 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
515 error = kern_jail_set(td, auio, uap->flags);
521 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
528 struct in6_addr *ip6;
531 struct vfsoptlist *opts;
532 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
534 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
536 #if defined(INET) || defined(INET6)
541 size_t namelen, onamelen;
542 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
543 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
544 int fi, jid, jsys, len, level;
545 int childmax, rsnum, slevel, vfslocked;
546 int fullpath_disabled;
547 #if defined(INET) || defined(INET6)
556 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
560 error = priv_check(td, PRIV_JAIL_SET);
561 if (!error && (flags & JAIL_ATTACH))
562 error = priv_check(td, PRIV_JAIL_ATTACH);
565 mypr = ppr = td->td_ucred->cr_prison;
566 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
568 if (flags & ~JAIL_SET_MASK)
572 * Check all the parameters before committing to anything. Not all
573 * errors can be caught early, but we may as well try. Also, this
574 * takes care of some expensive stuff (path lookup) before getting
575 * the allprison lock.
577 * XXX Jails are not filesystems, and jail parameters are not mount
578 * options. But it makes more sense to re-use the vfsopt code
579 * than duplicate it under a different name.
581 error = vfs_buildopts(optuio, &opts);
592 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
598 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
607 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
615 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
620 else if (enforce < 0 || enforce > 2) {
626 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
634 pr_flags = ch_flags = 0;
635 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
637 if (pr_flag_names[fi] == NULL)
639 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
640 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
642 ch_flags |= pr_flags;
643 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
645 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
652 case JAIL_SYS_DISABLE:
653 if (!pr_flag_jailsys[fi].disable) {
657 pr_flags |= pr_flag_jailsys[fi].disable;
660 pr_flags |= pr_flag_jailsys[fi].new;
662 case JAIL_SYS_INHERIT:
669 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
671 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
672 && !(pr_flags & PR_PERSIST)) {
674 vfs_opterror(opts, "new jail must persist or attach");
678 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
680 vfs_opterror(opts, "vnet cannot be changed after creation");
685 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
687 vfs_opterror(opts, "ip4 cannot be changed after creation");
692 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
694 vfs_opterror(opts, "ip6 cannot be changed after creation");
699 pr_allow = ch_allow = 0;
700 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
702 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
703 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
705 ch_allow |= pr_allow;
707 error = vfs_getopt(opts, "name", (void **)&name, &len);
713 if (len == 0 || name[len - 1] != '\0') {
717 if (len > MAXHOSTNAMELEN) {
718 error = ENAMETOOLONG;
723 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
731 if (len == 0 || host[len - 1] != '\0') {
735 if (len > MAXHOSTNAMELEN) {
736 error = ENAMETOOLONG;
741 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
749 if (len == 0 || domain[len - 1] != '\0') {
753 if (len > MAXHOSTNAMELEN) {
754 error = ENAMETOOLONG;
759 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
767 if (len == 0 || uuid[len - 1] != '\0') {
771 if (len > HOSTUUIDLEN) {
772 error = ENAMETOOLONG;
777 #ifdef COMPAT_FREEBSD32
778 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
781 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
785 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
797 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
799 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
802 else if (ip4s & (sizeof(*ip4) - 1)) {
806 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
808 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
810 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
811 ip4s /= sizeof(*ip4);
812 if (ip4s > jail_max_af_ips) {
814 vfs_opterror(opts, "too many IPv4 addresses");
817 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
818 bcopy(op, ip4, ip4s * sizeof(*ip4));
820 * IP addresses are all sorted but ip[0] to preserve
821 * the primary IP address as given from userland.
822 * This special IP is used for unbound outgoing
823 * connections as well for "loopback" traffic in case
824 * source address selection cannot find any more fitting
825 * address to connect from.
828 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
830 * Check for duplicate addresses and do some simple
831 * zero and broadcast checks. If users give other bogus
832 * addresses it is their problem.
834 * We do not have to care about byte order for these
835 * checks so we will do them in NBO.
837 for (ii = 0; ii < ip4s; ii++) {
838 if (ip4[ii].s_addr == INADDR_ANY ||
839 ip4[ii].s_addr == INADDR_BROADCAST) {
844 (ip4[0].s_addr == ip4[ii+1].s_addr ||
845 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
855 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
857 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
860 else if (ip6s & (sizeof(*ip6) - 1)) {
864 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
866 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
868 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
869 ip6s /= sizeof(*ip6);
870 if (ip6s > jail_max_af_ips) {
872 vfs_opterror(opts, "too many IPv6 addresses");
875 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
876 bcopy(op, ip6, ip6s * sizeof(*ip6));
878 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
879 for (ii = 0; ii < ip6s; ii++) {
880 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
885 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
886 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
896 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
897 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
900 "vnet jails cannot have IP address restrictions");
905 fullpath_disabled = 0;
907 error = vfs_getopt(opts, "path", (void **)&path, &len);
913 if (flags & JAIL_UPDATE) {
916 "path cannot be changed after creation");
919 if (len == 0 || path[len - 1] != '\0') {
923 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_SYSSPACE,
928 vfslocked = NDHASGIANT(&nd);
930 NDFREE(&nd, NDF_ONLY_PNBUF);
931 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
932 strlcpy(g_path, path, MAXPATHLEN);
933 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
936 else if (error == ENODEV) {
937 /* proceed if sysctl debug.disablefullpath == 1 */
938 fullpath_disabled = 1;
939 if (len < 2 || (len == 2 && path[0] == '/'))
942 /* exit on other errors */
943 VFS_UNLOCK_GIANT(vfslocked);
946 if (root->v_type != VDIR) {
949 VFS_UNLOCK_GIANT(vfslocked);
953 VFS_UNLOCK_GIANT(vfslocked);
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);
1814 * Now that it is all there, drop the temporary reference from existing
1815 * prisons. Or add a reference to newly created persistent prisons
1816 * (which was not done earlier so that the prison would not be publicly
1820 prison_deref(pr, (flags & JAIL_ATTACH)
1822 : PD_DEREF | PD_LIST_SLOCKED);
1824 if (pr_flags & PR_PERSIST) {
1825 mtx_lock(&pr->pr_mtx);
1828 mtx_unlock(&pr->pr_mtx);
1830 if (!(flags & JAIL_ATTACH))
1831 sx_sunlock(&allprison_lock);
1833 td->td_retval[0] = pr->pr_id;
1837 prison_deref(pr, created
1838 ? PD_LOCKED | PD_LIST_XLOCKED
1839 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1842 sx_xunlock(&allprison_lock);
1845 vfslocked = VFS_LOCK_GIANT(root->v_mount);
1847 VFS_UNLOCK_GIANT(vfslocked);
1851 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1852 if (errmsg_len > 0) {
1853 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1854 if (errmsg_pos > 0) {
1855 if (optuio->uio_segflg == UIO_SYSSPACE)
1857 optuio->uio_iov[errmsg_pos].iov_base,
1861 optuio->uio_iov[errmsg_pos].iov_base,
1868 free(ip4, M_PRISON);
1871 free(ip6, M_PRISON);
1874 free(g_path, M_TEMP);
1881 * struct jail_get_args {
1882 * struct iovec *iovp;
1883 * unsigned int iovcnt;
1888 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1893 /* Check that we have an even number of iovecs. */
1894 if (uap->iovcnt & 1)
1897 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1900 error = kern_jail_get(td, auio, uap->flags);
1902 error = copyout(auio->uio_iov, uap->iovp,
1903 uap->iovcnt * sizeof (struct iovec));
1909 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1911 struct prison *pr, *mypr;
1913 struct vfsoptlist *opts;
1914 char *errmsg, *name;
1915 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1917 if (flags & ~JAIL_GET_MASK)
1920 /* Get the parameter list. */
1921 error = vfs_buildopts(optuio, &opts);
1924 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1925 mypr = td->td_ucred->cr_prison;
1928 * Find the prison specified by one of: lastjid, jid, name.
1930 sx_slock(&allprison_lock);
1931 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1933 TAILQ_FOREACH(pr, &allprison, pr_list) {
1934 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1935 mtx_lock(&pr->pr_mtx);
1936 if (pr->pr_ref > 0 &&
1937 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1939 mtx_unlock(&pr->pr_mtx);
1945 vfs_opterror(opts, "no jail after %d", jid);
1946 goto done_unlock_list;
1947 } else if (error != ENOENT)
1948 goto done_unlock_list;
1950 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1953 pr = prison_find_child(mypr, jid);
1955 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1956 mtx_unlock(&pr->pr_mtx);
1958 vfs_opterror(opts, "jail %d is dying",
1960 goto done_unlock_list;
1965 vfs_opterror(opts, "jail %d not found", jid);
1966 goto done_unlock_list;
1968 } else if (error != ENOENT)
1969 goto done_unlock_list;
1971 error = vfs_getopt(opts, "name", (void **)&name, &len);
1973 if (len == 0 || name[len - 1] != '\0') {
1975 goto done_unlock_list;
1977 pr = prison_find_name(mypr, name);
1979 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1980 mtx_unlock(&pr->pr_mtx);
1982 vfs_opterror(opts, "jail \"%s\" is dying",
1984 goto done_unlock_list;
1989 vfs_opterror(opts, "jail \"%s\" not found", name);
1990 goto done_unlock_list;
1991 } else if (error != ENOENT)
1992 goto done_unlock_list;
1994 vfs_opterror(opts, "no jail specified");
1996 goto done_unlock_list;
1999 /* Get the parameters of the prison. */
2002 td->td_retval[0] = pr->pr_id;
2003 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2004 if (error != 0 && error != ENOENT)
2006 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2007 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2008 if (error != 0 && error != ENOENT)
2010 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2011 if (error != 0 && error != ENOENT)
2013 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2014 sizeof(pr->pr_cpuset->cs_id));
2015 if (error != 0 && error != ENOENT)
2017 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2018 if (error != 0 && error != ENOENT)
2021 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2022 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2023 if (error != 0 && error != ENOENT)
2027 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2028 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2029 if (error != 0 && error != ENOENT)
2032 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2033 sizeof(pr->pr_securelevel));
2034 if (error != 0 && error != ENOENT)
2036 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2037 sizeof(pr->pr_childcount));
2038 if (error != 0 && error != ENOENT)
2040 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2041 sizeof(pr->pr_childmax));
2042 if (error != 0 && error != ENOENT)
2044 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2045 if (error != 0 && error != ENOENT)
2047 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2048 if (error != 0 && error != ENOENT)
2050 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2051 if (error != 0 && error != ENOENT)
2053 #ifdef COMPAT_FREEBSD32
2054 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2055 uint32_t hid32 = pr->pr_hostid;
2057 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2060 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2061 sizeof(pr->pr_hostid));
2062 if (error != 0 && error != ENOENT)
2064 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2065 sizeof(pr->pr_enforce_statfs));
2066 if (error != 0 && error != ENOENT)
2068 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2069 sizeof(pr->pr_devfs_rsnum));
2070 if (error != 0 && error != ENOENT)
2072 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2074 if (pr_flag_names[fi] == NULL)
2076 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2077 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2078 if (error != 0 && error != ENOENT)
2081 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2082 if (error != 0 && error != ENOENT)
2085 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2088 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2089 i = pr_flag_jailsys[fi].disable &&
2090 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2091 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2094 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2095 if (error != 0 && error != ENOENT)
2098 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2100 if (pr_allow_names[fi] == NULL)
2102 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2103 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2104 if (error != 0 && error != ENOENT)
2107 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2108 if (error != 0 && error != ENOENT)
2111 i = (pr->pr_uref == 0);
2112 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2113 if (error != 0 && error != ENOENT)
2116 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2117 if (error != 0 && error != ENOENT)
2120 /* Get the module parameters. */
2121 mtx_unlock(&pr->pr_mtx);
2123 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2126 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2128 /* By now, all parameters should have been noted. */
2129 TAILQ_FOREACH(opt, opts, link) {
2130 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2132 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2137 /* Write the fetched parameters back to userspace. */
2139 TAILQ_FOREACH(opt, opts, link) {
2140 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2141 pos = 2 * opt->pos + 1;
2142 optuio->uio_iov[pos].iov_len = opt->len;
2143 if (opt->value != NULL) {
2144 if (optuio->uio_segflg == UIO_SYSSPACE) {
2146 optuio->uio_iov[pos].iov_base,
2149 error = copyout(opt->value,
2150 optuio->uio_iov[pos].iov_base,
2161 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2165 sx_sunlock(&allprison_lock);
2167 if (error && errmsg_pos >= 0) {
2168 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2169 errmsg_pos = 2 * errmsg_pos + 1;
2170 if (errmsg_len > 0) {
2171 if (optuio->uio_segflg == UIO_SYSSPACE)
2173 optuio->uio_iov[errmsg_pos].iov_base,
2177 optuio->uio_iov[errmsg_pos].iov_base,
2187 * struct jail_remove_args {
2192 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2194 struct prison *pr, *cpr, *lpr, *tpr;
2197 error = priv_check(td, PRIV_JAIL_REMOVE);
2201 sx_xlock(&allprison_lock);
2202 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2204 sx_xunlock(&allprison_lock);
2208 /* Remove all descendants of this prison, then remove this prison. */
2210 pr->pr_flags |= PR_REMOVE;
2211 if (!LIST_EMPTY(&pr->pr_children)) {
2212 mtx_unlock(&pr->pr_mtx);
2214 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2215 mtx_lock(&cpr->pr_mtx);
2216 if (cpr->pr_ref > 0) {
2219 cpr->pr_flags |= PR_REMOVE;
2221 /* Already removed - do not do it again. */
2224 mtx_unlock(&cpr->pr_mtx);
2226 mtx_lock(&lpr->pr_mtx);
2227 prison_remove_one(lpr);
2228 sx_xlock(&allprison_lock);
2233 mtx_lock(&lpr->pr_mtx);
2234 prison_remove_one(lpr);
2235 sx_xlock(&allprison_lock);
2237 mtx_lock(&pr->pr_mtx);
2239 prison_remove_one(pr);
2244 prison_remove_one(struct prison *pr)
2249 /* If the prison was persistent, it is not anymore. */
2251 if (pr->pr_flags & PR_PERSIST) {
2254 pr->pr_flags &= ~PR_PERSIST;
2258 * jail_remove added a reference. If that's the only one, remove
2261 KASSERT(pr->pr_ref > 0,
2262 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2263 if (pr->pr_ref == 1) {
2265 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2269 mtx_unlock(&pr->pr_mtx);
2270 sx_xunlock(&allprison_lock);
2272 * Kill all processes unfortunate enough to be attached to this prison.
2274 sx_slock(&allproc_lock);
2275 LIST_FOREACH(p, &allproc, p_list) {
2277 if (p->p_state != PRS_NEW && p->p_ucred &&
2278 p->p_ucred->cr_prison == pr)
2279 kern_psignal(p, SIGKILL);
2282 sx_sunlock(&allproc_lock);
2283 /* Remove the temporary reference added by jail_remove. */
2284 prison_deref(pr, deuref | PD_DEREF);
2289 * struct jail_attach_args {
2294 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2299 error = priv_check(td, PRIV_JAIL_ATTACH);
2303 sx_slock(&allprison_lock);
2304 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2306 sx_sunlock(&allprison_lock);
2311 * Do not allow a process to attach to a prison that is not
2312 * considered to be "alive".
2314 if (pr->pr_uref == 0) {
2315 mtx_unlock(&pr->pr_mtx);
2316 sx_sunlock(&allprison_lock);
2320 return (do_jail_attach(td, pr));
2324 do_jail_attach(struct thread *td, struct prison *pr)
2328 struct ucred *newcred, *oldcred;
2329 int vfslocked, error;
2332 * XXX: Note that there is a slight race here if two threads
2333 * in the same privileged process attempt to attach to two
2334 * different jails at the same time. It is important for
2335 * user processes not to do this, or they might end up with
2336 * a process root from one prison, but attached to the jail
2341 mtx_unlock(&pr->pr_mtx);
2343 /* Let modules do whatever they need to prepare for attaching. */
2344 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2346 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2349 sx_sunlock(&allprison_lock);
2352 * Reparent the newly attached process to this jail.
2354 ppr = td->td_ucred->cr_prison;
2356 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2360 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2361 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2362 if ((error = change_dir(pr->pr_root, td)) != 0)
2365 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2368 VOP_UNLOCK(pr->pr_root, 0);
2369 if ((error = change_root(pr->pr_root, td)))
2370 goto e_unlock_giant;
2371 VFS_UNLOCK_GIANT(vfslocked);
2375 oldcred = p->p_ucred;
2377 crcopy(newcred, oldcred);
2378 newcred->cr_prison = pr;
2379 p->p_ucred = newcred;
2382 racct_proc_ucred_changed(p, oldcred, newcred);
2385 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2388 VOP_UNLOCK(pr->pr_root, 0);
2390 VFS_UNLOCK_GIANT(vfslocked);
2392 /* Tell modules this thread is still in its old jail after all. */
2393 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2394 prison_deref(pr, PD_DEREF | PD_DEUREF);
2400 * Returns a locked prison instance, or NULL on failure.
2403 prison_find(int prid)
2407 sx_assert(&allprison_lock, SX_LOCKED);
2408 TAILQ_FOREACH(pr, &allprison, pr_list) {
2409 if (pr->pr_id == prid) {
2410 mtx_lock(&pr->pr_mtx);
2413 mtx_unlock(&pr->pr_mtx);
2420 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2423 prison_find_child(struct prison *mypr, int prid)
2428 sx_assert(&allprison_lock, SX_LOCKED);
2429 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2430 if (pr->pr_id == prid) {
2431 mtx_lock(&pr->pr_mtx);
2434 mtx_unlock(&pr->pr_mtx);
2441 * Look for the name relative to mypr. Returns a locked prison or NULL.
2444 prison_find_name(struct prison *mypr, const char *name)
2446 struct prison *pr, *deadpr;
2450 sx_assert(&allprison_lock, SX_LOCKED);
2451 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2454 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2455 if (!strcmp(pr->pr_name + mylen, name)) {
2456 mtx_lock(&pr->pr_mtx);
2457 if (pr->pr_ref > 0) {
2458 if (pr->pr_uref > 0)
2462 mtx_unlock(&pr->pr_mtx);
2465 /* There was no valid prison - perhaps there was a dying one. */
2466 if (deadpr != NULL) {
2467 mtx_lock(&deadpr->pr_mtx);
2468 if (deadpr->pr_ref == 0) {
2469 mtx_unlock(&deadpr->pr_mtx);
2477 * See if a prison has the specific flag set.
2480 prison_flag(struct ucred *cred, unsigned flag)
2483 /* This is an atomic read, so no locking is necessary. */
2484 return (cred->cr_prison->pr_flags & flag);
2488 prison_allow(struct ucred *cred, unsigned flag)
2491 /* This is an atomic read, so no locking is necessary. */
2492 return (cred->cr_prison->pr_allow & flag);
2496 * Remove a prison reference. If that was the last reference, remove the
2497 * prison itself - but not in this context in case there are locks held.
2500 prison_free_locked(struct prison *pr)
2503 mtx_assert(&pr->pr_mtx, MA_OWNED);
2505 if (pr->pr_ref == 0) {
2506 mtx_unlock(&pr->pr_mtx);
2507 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2508 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2511 mtx_unlock(&pr->pr_mtx);
2515 prison_free(struct prison *pr)
2518 mtx_lock(&pr->pr_mtx);
2519 prison_free_locked(pr);
2523 prison_complete(void *context, int pending)
2526 prison_deref((struct prison *)context, 0);
2530 * Remove a prison reference (usually). This internal version assumes no
2531 * mutexes are held, except perhaps the prison itself. If there are no more
2532 * references, release and delist the prison. On completion, the prison lock
2533 * and the allprison lock are both unlocked.
2536 prison_deref(struct prison *pr, int flags)
2538 struct prison *ppr, *tpr;
2541 if (!(flags & PD_LOCKED))
2542 mtx_lock(&pr->pr_mtx);
2544 if (flags & PD_DEUREF) {
2546 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2548 if (flags & PD_DEREF)
2550 /* If the prison still has references, nothing else to do. */
2551 if (pr->pr_ref > 0) {
2552 mtx_unlock(&pr->pr_mtx);
2553 if (flags & PD_LIST_SLOCKED)
2554 sx_sunlock(&allprison_lock);
2555 else if (flags & PD_LIST_XLOCKED)
2556 sx_xunlock(&allprison_lock);
2560 mtx_unlock(&pr->pr_mtx);
2561 if (flags & PD_LIST_SLOCKED) {
2562 if (!sx_try_upgrade(&allprison_lock)) {
2563 sx_sunlock(&allprison_lock);
2564 sx_xlock(&allprison_lock);
2566 } else if (!(flags & PD_LIST_XLOCKED))
2567 sx_xlock(&allprison_lock);
2569 TAILQ_REMOVE(&allprison, pr, pr_list);
2570 LIST_REMOVE(pr, pr_sibling);
2571 ppr = pr->pr_parent;
2572 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2573 tpr->pr_childcount--;
2574 sx_xunlock(&allprison_lock);
2577 if (pr->pr_vnet != ppr->pr_vnet)
2578 vnet_destroy(pr->pr_vnet);
2580 if (pr->pr_root != NULL) {
2581 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2583 VFS_UNLOCK_GIANT(vfslocked);
2585 mtx_destroy(&pr->pr_mtx);
2587 free(pr->pr_ip4, M_PRISON);
2590 free(pr->pr_ip6, M_PRISON);
2592 if (pr->pr_cpuset != NULL)
2593 cpuset_rel(pr->pr_cpuset);
2596 prison_racct_detach(pr);
2600 /* Removing a prison frees a reference on its parent. */
2602 mtx_lock(&pr->pr_mtx);
2603 flags = PD_DEREF | PD_DEUREF;
2608 prison_hold_locked(struct prison *pr)
2611 mtx_assert(&pr->pr_mtx, MA_OWNED);
2612 KASSERT(pr->pr_ref > 0,
2613 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2618 prison_hold(struct prison *pr)
2621 mtx_lock(&pr->pr_mtx);
2622 prison_hold_locked(pr);
2623 mtx_unlock(&pr->pr_mtx);
2627 prison_proc_hold(struct prison *pr)
2630 mtx_lock(&pr->pr_mtx);
2631 KASSERT(pr->pr_uref > 0,
2632 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2634 mtx_unlock(&pr->pr_mtx);
2638 prison_proc_free(struct prison *pr)
2641 mtx_lock(&pr->pr_mtx);
2642 KASSERT(pr->pr_uref > 0,
2643 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2644 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2650 * Restrict a prison's IP address list with its parent's, possibly replacing
2651 * it. Return true if the replacement buffer was used (or would have been).
2654 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2659 ppr = pr->pr_parent;
2660 if (!(pr->pr_flags & PR_IP4_USER)) {
2661 /* This has no user settings, so just copy the parent's list. */
2662 if (pr->pr_ip4s < ppr->pr_ip4s) {
2664 * There's no room for the parent's list. Use the
2665 * new list buffer, which is assumed to be big enough
2666 * (if it was passed). If there's no buffer, try to
2670 if (newip4 == NULL) {
2671 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2672 M_PRISON, M_NOWAIT);
2676 if (newip4 != NULL) {
2677 bcopy(ppr->pr_ip4, newip4,
2678 ppr->pr_ip4s * sizeof(*newip4));
2679 free(pr->pr_ip4, M_PRISON);
2680 pr->pr_ip4 = newip4;
2681 pr->pr_ip4s = ppr->pr_ip4s;
2685 pr->pr_ip4s = ppr->pr_ip4s;
2686 if (pr->pr_ip4s > 0)
2687 bcopy(ppr->pr_ip4, pr->pr_ip4,
2688 pr->pr_ip4s * sizeof(*newip4));
2689 else if (pr->pr_ip4 != NULL) {
2690 free(pr->pr_ip4, M_PRISON);
2693 } else if (pr->pr_ip4s > 0) {
2694 /* Remove addresses that aren't in the parent. */
2695 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2696 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2698 if (ij < ppr->pr_ip4s)
2701 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2702 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2705 for (ij = 1; ii < pr->pr_ip4s; ) {
2706 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2710 switch (ij >= ppr->pr_ip4s ? -1 :
2711 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2713 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2714 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2725 if (pr->pr_ip4s == 0) {
2726 pr->pr_flags |= PR_IP4_DISABLE;
2727 free(pr->pr_ip4, M_PRISON);
2735 * Pass back primary IPv4 address of this jail.
2737 * If not restricted return success but do not alter the address. Caller has
2738 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2740 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2741 * Address returned in NBO.
2744 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2748 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2749 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2751 pr = cred->cr_prison;
2752 if (!(pr->pr_flags & PR_IP4))
2754 mtx_lock(&pr->pr_mtx);
2755 if (!(pr->pr_flags & PR_IP4)) {
2756 mtx_unlock(&pr->pr_mtx);
2759 if (pr->pr_ip4 == NULL) {
2760 mtx_unlock(&pr->pr_mtx);
2761 return (EAFNOSUPPORT);
2764 ia->s_addr = pr->pr_ip4[0].s_addr;
2765 mtx_unlock(&pr->pr_mtx);
2770 * Return 1 if we should do proper source address selection or are not jailed.
2771 * We will return 0 if we should bypass source address selection in favour
2772 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2774 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2777 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2783 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2784 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2789 pr = cred->cr_prison;
2790 if (pr->pr_flags & PR_IP4_SADDRSEL)
2793 lia.s_addr = INADDR_ANY;
2794 error = prison_get_ip4(cred, &lia);
2797 if (lia.s_addr == INADDR_ANY)
2800 ia->s_addr = lia.s_addr;
2805 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2808 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2815 * No need to lock since the PR_IP4_USER flag can't be altered for
2818 while (pr1 != &prison0 &&
2820 !(pr1->pr_flags & PR_VNET) &&
2822 !(pr1->pr_flags & PR_IP4_USER))
2823 pr1 = pr1->pr_parent;
2824 while (pr2 != &prison0 &&
2826 !(pr2->pr_flags & PR_VNET) &&
2828 !(pr2->pr_flags & PR_IP4_USER))
2829 pr2 = pr2->pr_parent;
2830 return (pr1 == pr2);
2834 * Make sure our (source) address is set to something meaningful to this
2837 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2838 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2839 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2842 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2848 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2849 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2851 pr = cred->cr_prison;
2852 if (!(pr->pr_flags & PR_IP4))
2854 mtx_lock(&pr->pr_mtx);
2855 if (!(pr->pr_flags & PR_IP4)) {
2856 mtx_unlock(&pr->pr_mtx);
2859 if (pr->pr_ip4 == NULL) {
2860 mtx_unlock(&pr->pr_mtx);
2861 return (EAFNOSUPPORT);
2864 ia0.s_addr = ntohl(ia->s_addr);
2865 if (ia0.s_addr == INADDR_LOOPBACK) {
2866 ia->s_addr = pr->pr_ip4[0].s_addr;
2867 mtx_unlock(&pr->pr_mtx);
2871 if (ia0.s_addr == INADDR_ANY) {
2873 * In case there is only 1 IPv4 address, bind directly.
2875 if (pr->pr_ip4s == 1)
2876 ia->s_addr = pr->pr_ip4[0].s_addr;
2877 mtx_unlock(&pr->pr_mtx);
2881 error = _prison_check_ip4(pr, ia);
2882 mtx_unlock(&pr->pr_mtx);
2887 * Rewrite destination address in case we will connect to loopback address.
2889 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2890 * Address passed in in NBO and returned in NBO.
2893 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2897 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2898 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2900 pr = cred->cr_prison;
2901 if (!(pr->pr_flags & PR_IP4))
2903 mtx_lock(&pr->pr_mtx);
2904 if (!(pr->pr_flags & PR_IP4)) {
2905 mtx_unlock(&pr->pr_mtx);
2908 if (pr->pr_ip4 == NULL) {
2909 mtx_unlock(&pr->pr_mtx);
2910 return (EAFNOSUPPORT);
2913 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2914 ia->s_addr = pr->pr_ip4[0].s_addr;
2915 mtx_unlock(&pr->pr_mtx);
2920 * Return success because nothing had to be changed.
2922 mtx_unlock(&pr->pr_mtx);
2927 * Check if given address belongs to the jail referenced by cred/prison.
2929 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2930 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2931 * doesn't allow IPv4. Address passed in in NBO.
2934 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2939 * Check the primary IP.
2941 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2945 * All the other IPs are sorted so we can do a binary search.
2948 z = pr->pr_ip4s - 2;
2951 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2960 return (EADDRNOTAVAIL);
2964 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2969 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2970 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2972 pr = cred->cr_prison;
2973 if (!(pr->pr_flags & PR_IP4))
2975 mtx_lock(&pr->pr_mtx);
2976 if (!(pr->pr_flags & PR_IP4)) {
2977 mtx_unlock(&pr->pr_mtx);
2980 if (pr->pr_ip4 == NULL) {
2981 mtx_unlock(&pr->pr_mtx);
2982 return (EAFNOSUPPORT);
2985 error = _prison_check_ip4(pr, ia);
2986 mtx_unlock(&pr->pr_mtx);
2993 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2998 ppr = pr->pr_parent;
2999 if (!(pr->pr_flags & PR_IP6_USER)) {
3000 /* This has no user settings, so just copy the parent's list. */
3001 if (pr->pr_ip6s < ppr->pr_ip6s) {
3003 * There's no room for the parent's list. Use the
3004 * new list buffer, which is assumed to be big enough
3005 * (if it was passed). If there's no buffer, try to
3009 if (newip6 == NULL) {
3010 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3011 M_PRISON, M_NOWAIT);
3015 if (newip6 != NULL) {
3016 bcopy(ppr->pr_ip6, newip6,
3017 ppr->pr_ip6s * sizeof(*newip6));
3018 free(pr->pr_ip6, M_PRISON);
3019 pr->pr_ip6 = newip6;
3020 pr->pr_ip6s = ppr->pr_ip6s;
3024 pr->pr_ip6s = ppr->pr_ip6s;
3025 if (pr->pr_ip6s > 0)
3026 bcopy(ppr->pr_ip6, pr->pr_ip6,
3027 pr->pr_ip6s * sizeof(*newip6));
3028 else if (pr->pr_ip6 != NULL) {
3029 free(pr->pr_ip6, M_PRISON);
3032 } else if (pr->pr_ip6s > 0) {
3033 /* Remove addresses that aren't in the parent. */
3034 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3035 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3038 if (ij < ppr->pr_ip6s)
3041 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3042 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3045 for (ij = 1; ii < pr->pr_ip6s; ) {
3046 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3051 switch (ij >= ppr->pr_ip4s ? -1 :
3052 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3054 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3055 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3066 if (pr->pr_ip6s == 0) {
3067 pr->pr_flags |= PR_IP6_DISABLE;
3068 free(pr->pr_ip6, M_PRISON);
3076 * Pass back primary IPv6 address for this jail.
3078 * If not restricted return success but do not alter the address. Caller has
3079 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3081 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3084 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3088 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3089 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3091 pr = cred->cr_prison;
3092 if (!(pr->pr_flags & PR_IP6))
3094 mtx_lock(&pr->pr_mtx);
3095 if (!(pr->pr_flags & PR_IP6)) {
3096 mtx_unlock(&pr->pr_mtx);
3099 if (pr->pr_ip6 == NULL) {
3100 mtx_unlock(&pr->pr_mtx);
3101 return (EAFNOSUPPORT);
3104 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3105 mtx_unlock(&pr->pr_mtx);
3110 * Return 1 if we should do proper source address selection or are not jailed.
3111 * We will return 0 if we should bypass source address selection in favour
3112 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3114 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3117 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3120 struct in6_addr lia6;
3123 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3124 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3129 pr = cred->cr_prison;
3130 if (pr->pr_flags & PR_IP6_SADDRSEL)
3134 error = prison_get_ip6(cred, &lia6);
3137 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3140 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3145 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3148 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3154 while (pr1 != &prison0 &&
3156 !(pr1->pr_flags & PR_VNET) &&
3158 !(pr1->pr_flags & PR_IP6_USER))
3159 pr1 = pr1->pr_parent;
3160 while (pr2 != &prison0 &&
3162 !(pr2->pr_flags & PR_VNET) &&
3164 !(pr2->pr_flags & PR_IP6_USER))
3165 pr2 = pr2->pr_parent;
3166 return (pr1 == pr2);
3170 * Make sure our (source) address is set to something meaningful to this jail.
3172 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3173 * when needed while binding.
3175 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3176 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3177 * doesn't allow IPv6.
3180 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3185 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3186 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3188 pr = cred->cr_prison;
3189 if (!(pr->pr_flags & PR_IP6))
3191 mtx_lock(&pr->pr_mtx);
3192 if (!(pr->pr_flags & PR_IP6)) {
3193 mtx_unlock(&pr->pr_mtx);
3196 if (pr->pr_ip6 == NULL) {
3197 mtx_unlock(&pr->pr_mtx);
3198 return (EAFNOSUPPORT);
3201 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3202 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3203 mtx_unlock(&pr->pr_mtx);
3207 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3209 * In case there is only 1 IPv6 address, and v6only is true,
3210 * then bind directly.
3212 if (v6only != 0 && pr->pr_ip6s == 1)
3213 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3214 mtx_unlock(&pr->pr_mtx);
3218 error = _prison_check_ip6(pr, ia6);
3219 mtx_unlock(&pr->pr_mtx);
3224 * Rewrite destination address in case we will connect to loopback address.
3226 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3229 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3233 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3234 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3236 pr = cred->cr_prison;
3237 if (!(pr->pr_flags & PR_IP6))
3239 mtx_lock(&pr->pr_mtx);
3240 if (!(pr->pr_flags & PR_IP6)) {
3241 mtx_unlock(&pr->pr_mtx);
3244 if (pr->pr_ip6 == NULL) {
3245 mtx_unlock(&pr->pr_mtx);
3246 return (EAFNOSUPPORT);
3249 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3250 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3251 mtx_unlock(&pr->pr_mtx);
3256 * Return success because nothing had to be changed.
3258 mtx_unlock(&pr->pr_mtx);
3263 * Check if given address belongs to the jail referenced by cred/prison.
3265 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3266 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3267 * doesn't allow IPv6.
3270 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3275 * Check the primary IP.
3277 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3281 * All the other IPs are sorted so we can do a binary search.
3284 z = pr->pr_ip6s - 2;
3287 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3296 return (EADDRNOTAVAIL);
3300 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3305 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3306 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3308 pr = cred->cr_prison;
3309 if (!(pr->pr_flags & PR_IP6))
3311 mtx_lock(&pr->pr_mtx);
3312 if (!(pr->pr_flags & PR_IP6)) {
3313 mtx_unlock(&pr->pr_mtx);
3316 if (pr->pr_ip6 == NULL) {
3317 mtx_unlock(&pr->pr_mtx);
3318 return (EAFNOSUPPORT);
3321 error = _prison_check_ip6(pr, ia6);
3322 mtx_unlock(&pr->pr_mtx);
3328 * Check if a jail supports the given address family.
3330 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3334 prison_check_af(struct ucred *cred, int af)
3339 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3341 pr = cred->cr_prison;
3343 /* Prisons with their own network stack are not limited. */
3344 if (prison_owns_vnet(cred))
3353 if (pr->pr_flags & PR_IP4)
3355 mtx_lock(&pr->pr_mtx);
3356 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3357 error = EAFNOSUPPORT;
3358 mtx_unlock(&pr->pr_mtx);
3364 if (pr->pr_flags & PR_IP6)
3366 mtx_lock(&pr->pr_mtx);
3367 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3368 error = EAFNOSUPPORT;
3369 mtx_unlock(&pr->pr_mtx);
3377 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3378 error = EAFNOSUPPORT;
3384 * Check if given address belongs to the jail referenced by cred (wrapper to
3385 * prison_check_ip[46]).
3387 * Returns 0 if jail doesn't restrict the address family or if address belongs
3388 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3389 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3392 prison_if(struct ucred *cred, struct sockaddr *sa)
3395 struct sockaddr_in *sai;
3398 struct sockaddr_in6 *sai6;
3402 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3403 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3406 if (prison_owns_vnet(cred))
3411 switch (sa->sa_family)
3415 sai = (struct sockaddr_in *)sa;
3416 error = prison_check_ip4(cred, &sai->sin_addr);
3421 sai6 = (struct sockaddr_in6 *)sa;
3422 error = prison_check_ip6(cred, &sai6->sin6_addr);
3426 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3427 error = EAFNOSUPPORT;
3433 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3436 prison_check(struct ucred *cred1, struct ucred *cred2)
3439 return ((cred1->cr_prison == cred2->cr_prison ||
3440 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3444 * Return 1 if p2 is a child of p1, otherwise 0.
3447 prison_ischild(struct prison *pr1, struct prison *pr2)
3450 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3457 * Return 1 if the passed credential is in a jail, otherwise 0.
3460 jailed(struct ucred *cred)
3463 return (cred->cr_prison != &prison0);
3467 * Return 1 if the passed credential is in a jail and that jail does not
3468 * have its own virtual network stack, otherwise 0.
3471 jailed_without_vnet(struct ucred *cred)
3477 if (prison_owns_vnet(cred))
3485 * Return the correct hostname (domainname, et al) for the passed credential.
3488 getcredhostname(struct ucred *cred, char *buf, size_t size)
3493 * A NULL credential can be used to shortcut to the physical
3494 * system's hostname.
3496 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3497 mtx_lock(&pr->pr_mtx);
3498 strlcpy(buf, pr->pr_hostname, size);
3499 mtx_unlock(&pr->pr_mtx);
3503 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3506 mtx_lock(&cred->cr_prison->pr_mtx);
3507 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3508 mtx_unlock(&cred->cr_prison->pr_mtx);
3512 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3515 mtx_lock(&cred->cr_prison->pr_mtx);
3516 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3517 mtx_unlock(&cred->cr_prison->pr_mtx);
3521 getcredhostid(struct ucred *cred, unsigned long *hostid)
3524 mtx_lock(&cred->cr_prison->pr_mtx);
3525 *hostid = cred->cr_prison->pr_hostid;
3526 mtx_unlock(&cred->cr_prison->pr_mtx);
3531 * Determine whether the prison represented by cred owns
3532 * its vnet rather than having it inherited.
3534 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3537 prison_owns_vnet(struct ucred *cred)
3541 * vnets cannot be added/removed after jail creation,
3542 * so no need to lock here.
3544 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3549 * Determine whether the subject represented by cred can "see"
3550 * status of a mount point.
3551 * Returns: 0 for permitted, ENOENT otherwise.
3552 * XXX: This function should be called cr_canseemount() and should be
3553 * placed in kern_prot.c.
3556 prison_canseemount(struct ucred *cred, struct mount *mp)
3562 pr = cred->cr_prison;
3563 if (pr->pr_enforce_statfs == 0)
3565 if (pr->pr_root->v_mount == mp)
3567 if (pr->pr_enforce_statfs == 2)
3570 * If jail's chroot directory is set to "/" we should be able to see
3571 * all mount-points from inside a jail.
3572 * This is ugly check, but this is the only situation when jail's
3573 * directory ends with '/'.
3575 if (strcmp(pr->pr_path, "/") == 0)
3577 len = strlen(pr->pr_path);
3579 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3582 * Be sure that we don't have situation where jail's root directory
3583 * is "/some/path" and mount point is "/some/pathpath".
3585 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3591 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3593 char jpath[MAXPATHLEN];
3597 pr = cred->cr_prison;
3598 if (pr->pr_enforce_statfs == 0)
3600 if (prison_canseemount(cred, mp) != 0) {
3601 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3602 strlcpy(sp->f_mntonname, "[restricted]",
3603 sizeof(sp->f_mntonname));
3606 if (pr->pr_root->v_mount == mp) {
3608 * Clear current buffer data, so we are sure nothing from
3609 * the valid path left there.
3611 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3612 *sp->f_mntonname = '/';
3616 * If jail's chroot directory is set to "/" we should be able to see
3617 * all mount-points from inside a jail.
3619 if (strcmp(pr->pr_path, "/") == 0)
3621 len = strlen(pr->pr_path);
3622 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3624 * Clear current buffer data, so we are sure nothing from
3625 * the valid path left there.
3627 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3628 if (*jpath == '\0') {
3629 /* Should never happen. */
3630 *sp->f_mntonname = '/';
3632 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3637 * Check with permission for a specific privilege is granted within jail. We
3638 * have a specific list of accepted privileges; the rest are denied.
3641 prison_priv_check(struct ucred *cred, int priv)
3649 * Privileges specific to prisons with a virtual network stack.
3650 * There might be a duplicate entry here in case the privilege
3651 * is only granted conditionally in the legacy jail case.
3656 * NFS-specific privileges.
3658 case PRIV_NFS_DAEMON:
3659 case PRIV_NFS_LOCKD:
3662 * Network stack privileges.
3664 case PRIV_NET_BRIDGE:
3667 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3668 case PRIV_NET_ROUTE:
3670 case PRIV_NET_SETIFMTU:
3671 case PRIV_NET_SETIFFLAGS:
3672 case PRIV_NET_SETIFCAP:
3673 case PRIV_NET_SETIFDESCR:
3674 case PRIV_NET_SETIFNAME :
3675 case PRIV_NET_SETIFMETRIC:
3676 case PRIV_NET_SETIFPHYS:
3677 case PRIV_NET_SETIFMAC:
3678 case PRIV_NET_ADDMULTI:
3679 case PRIV_NET_DELMULTI:
3680 case PRIV_NET_HWIOCTL:
3681 case PRIV_NET_SETLLADDR:
3682 case PRIV_NET_ADDIFGROUP:
3683 case PRIV_NET_DELIFGROUP:
3684 case PRIV_NET_IFCREATE:
3685 case PRIV_NET_IFDESTROY:
3686 case PRIV_NET_ADDIFADDR:
3687 case PRIV_NET_DELIFADDR:
3690 case PRIV_NET_SETIFVNET:
3691 case PRIV_NET_SETIFFIB:
3694 * 802.11-related privileges.
3696 case PRIV_NET80211_GETKEY:
3698 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3703 * AppleTalk privileges.
3705 case PRIV_NETATALK_RESERVEDPORT:
3710 case PRIV_NETATM_CFG:
3711 case PRIV_NETATM_ADD:
3712 case PRIV_NETATM_DEL:
3713 case PRIV_NETATM_SET:
3716 * Bluetooth privileges.
3718 case PRIV_NETBLUETOOTH_RAW:
3722 * Netgraph and netgraph module privileges.
3724 case PRIV_NETGRAPH_CONTROL:
3726 case PRIV_NETGRAPH_TTY:
3730 * IPv4 and IPv6 privileges.
3732 case PRIV_NETINET_IPFW:
3733 case PRIV_NETINET_DIVERT:
3734 case PRIV_NETINET_PF:
3735 case PRIV_NETINET_DUMMYNET:
3736 case PRIV_NETINET_CARP:
3737 case PRIV_NETINET_MROUTE:
3738 case PRIV_NETINET_RAW:
3739 case PRIV_NETINET_ADDRCTRL6:
3740 case PRIV_NETINET_ND6:
3741 case PRIV_NETINET_SCOPE6:
3742 case PRIV_NETINET_ALIFETIME6:
3743 case PRIV_NETINET_IPSEC:
3744 case PRIV_NETINET_BINDANY:
3748 * IPX/SPX privileges.
3750 case PRIV_NETIPX_RESERVEDPORT:
3751 case PRIV_NETIPX_RAW:
3765 * No default: or deny here.
3766 * In case of no permit fall through to next switch().
3768 if (cred->cr_prison->pr_flags & PR_VNET)
3776 * Allow ktrace privileges for root in jail.
3782 * Allow jailed processes to configure audit identity and
3783 * submit audit records (login, etc). In the future we may
3784 * want to further refine the relationship between audit and
3787 case PRIV_AUDIT_GETAUDIT:
3788 case PRIV_AUDIT_SETAUDIT:
3789 case PRIV_AUDIT_SUBMIT:
3793 * Allow jailed processes to manipulate process UNIX
3794 * credentials in any way they see fit.
3796 case PRIV_CRED_SETUID:
3797 case PRIV_CRED_SETEUID:
3798 case PRIV_CRED_SETGID:
3799 case PRIV_CRED_SETEGID:
3800 case PRIV_CRED_SETGROUPS:
3801 case PRIV_CRED_SETREUID:
3802 case PRIV_CRED_SETREGID:
3803 case PRIV_CRED_SETRESUID:
3804 case PRIV_CRED_SETRESGID:
3807 * Jail implements visibility constraints already, so allow
3808 * jailed root to override uid/gid-based constraints.
3810 case PRIV_SEEOTHERGIDS:
3811 case PRIV_SEEOTHERUIDS:
3814 * Jail implements inter-process debugging limits already, so
3815 * allow jailed root various debugging privileges.
3817 case PRIV_DEBUG_DIFFCRED:
3818 case PRIV_DEBUG_SUGID:
3819 case PRIV_DEBUG_UNPRIV:
3822 * Allow jail to set various resource limits and login
3823 * properties, and for now, exceed process resource limits.
3825 case PRIV_PROC_LIMIT:
3826 case PRIV_PROC_SETLOGIN:
3827 case PRIV_PROC_SETRLIMIT:
3830 * System V and POSIX IPC privileges are granted in jail.
3833 case PRIV_IPC_WRITE:
3834 case PRIV_IPC_ADMIN:
3835 case PRIV_IPC_MSGSIZE:
3839 * Jail operations within a jail work on child jails.
3841 case PRIV_JAIL_ATTACH:
3843 case PRIV_JAIL_REMOVE:
3846 * Jail implements its own inter-process limits, so allow
3847 * root processes in jail to change scheduling on other
3848 * processes in the same jail. Likewise for signalling.
3850 case PRIV_SCHED_DIFFCRED:
3851 case PRIV_SCHED_CPUSET:
3852 case PRIV_SIGNAL_DIFFCRED:
3853 case PRIV_SIGNAL_SUGID:
3856 * Allow jailed processes to write to sysctls marked as jail
3859 case PRIV_SYSCTL_WRITEJAIL:
3862 * Allow root in jail to manage a variety of quota
3863 * properties. These should likely be conditional on a
3864 * configuration option.
3866 case PRIV_VFS_GETQUOTA:
3867 case PRIV_VFS_SETQUOTA:
3870 * Since Jail relies on chroot() to implement file system
3871 * protections, grant many VFS privileges to root in jail.
3872 * Be careful to exclude mount-related and NFS-related
3876 case PRIV_VFS_WRITE:
3877 case PRIV_VFS_ADMIN:
3879 case PRIV_VFS_LOOKUP:
3880 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3881 case PRIV_VFS_CHFLAGS_DEV:
3882 case PRIV_VFS_CHOWN:
3883 case PRIV_VFS_CHROOT:
3884 case PRIV_VFS_RETAINSUGID:
3885 case PRIV_VFS_FCHROOT:
3887 case PRIV_VFS_SETGID:
3889 case PRIV_VFS_STICKYFILE:
3893 * Depending on the global setting, allow privilege of
3894 * setting system flags.
3896 case PRIV_VFS_SYSFLAGS:
3897 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3903 * Depending on the global setting, allow privilege of
3904 * mounting/unmounting file systems.
3906 case PRIV_VFS_MOUNT:
3907 case PRIV_VFS_UNMOUNT:
3908 case PRIV_VFS_MOUNT_NONUSER:
3909 case PRIV_VFS_MOUNT_OWNER:
3910 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3911 cred->cr_prison->pr_enforce_statfs < 2)
3917 * Allow jailed root to bind reserved ports and reuse in-use
3920 case PRIV_NETINET_RESERVEDPORT:
3921 case PRIV_NETINET_REUSEPORT:
3925 * Allow jailed root to set certian IPv4/6 (option) headers.
3927 case PRIV_NETINET_SETHDROPTS:
3931 * Conditionally allow creating raw sockets in jail.
3933 case PRIV_NETINET_RAW:
3934 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3940 * Since jail implements its own visibility limits on netstat
3941 * sysctls, allow getcred. This allows identd to work in
3944 case PRIV_NETINET_GETCRED:
3948 * Allow jailed root to set loginclass.
3950 case PRIV_PROC_SETLOGINCLASS:
3955 * In all remaining cases, deny the privilege request. This
3956 * includes almost all network privileges, many system
3957 * configuration privileges.
3964 * Return the part of pr2's name that is relative to pr1, or the whole name
3965 * if it does not directly follow.
3969 prison_name(struct prison *pr1, struct prison *pr2)
3973 /* Jails see themselves as "0" (if they see themselves at all). */
3976 name = pr2->pr_name;
3977 if (prison_ischild(pr1, pr2)) {
3979 * pr1 isn't locked (and allprison_lock may not be either)
3980 * so its length can't be counted on. But the number of dots
3981 * can be counted on - and counted.
3983 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3984 name = strchr(name, '.') + 1;
3990 * Return the part of pr2's path that is relative to pr1, or the whole path
3991 * if it does not directly follow.
3994 prison_path(struct prison *pr1, struct prison *pr2)
3996 char *path1, *path2;
3999 path1 = pr1->pr_path;
4000 path2 = pr2->pr_path;
4001 if (!strcmp(path1, "/"))
4003 len1 = strlen(path1);
4004 if (strncmp(path1, path2, len1))
4006 if (path2[len1] == '\0')
4008 if (path2[len1] == '/')
4009 return (path2 + len1);
4015 * Jail-related sysctls.
4017 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4021 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4024 struct prison *pr, *cpr;
4026 struct in_addr *ip4 = NULL;
4030 struct in6_addr *ip6 = NULL;
4035 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4036 pr = req->td->td_ucred->cr_prison;
4038 sx_slock(&allprison_lock);
4039 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4040 #if defined(INET) || defined(INET6)
4043 mtx_lock(&cpr->pr_mtx);
4045 if (cpr->pr_ip4s > 0) {
4046 if (ip4s < cpr->pr_ip4s) {
4047 ip4s = cpr->pr_ip4s;
4048 mtx_unlock(&cpr->pr_mtx);
4049 ip4 = realloc(ip4, ip4s *
4050 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4053 bcopy(cpr->pr_ip4, ip4,
4054 cpr->pr_ip4s * sizeof(struct in_addr));
4058 if (cpr->pr_ip6s > 0) {
4059 if (ip6s < cpr->pr_ip6s) {
4060 ip6s = cpr->pr_ip6s;
4061 mtx_unlock(&cpr->pr_mtx);
4062 ip6 = realloc(ip6, ip6s *
4063 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4066 bcopy(cpr->pr_ip6, ip6,
4067 cpr->pr_ip6s * sizeof(struct in6_addr));
4070 if (cpr->pr_ref == 0) {
4071 mtx_unlock(&cpr->pr_mtx);
4074 bzero(xp, sizeof(*xp));
4075 xp->pr_version = XPRISON_VERSION;
4076 xp->pr_id = cpr->pr_id;
4077 xp->pr_state = cpr->pr_uref > 0
4078 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4079 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4080 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4081 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4083 xp->pr_ip4s = cpr->pr_ip4s;
4086 xp->pr_ip6s = cpr->pr_ip6s;
4088 mtx_unlock(&cpr->pr_mtx);
4089 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4093 if (xp->pr_ip4s > 0) {
4094 error = SYSCTL_OUT(req, ip4,
4095 xp->pr_ip4s * sizeof(struct in_addr));
4101 if (xp->pr_ip6s > 0) {
4102 error = SYSCTL_OUT(req, ip6,
4103 xp->pr_ip6s * sizeof(struct in6_addr));
4109 sx_sunlock(&allprison_lock);
4120 SYSCTL_OID(_security_jail, OID_AUTO, list,
4121 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4122 sysctl_jail_list, "S", "List of active jails");
4125 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4129 injail = jailed(req->td->td_ucred);
4130 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4135 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4136 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4137 sysctl_jail_jailed, "I", "Process in jail?");
4139 #if defined(INET) || defined(INET6)
4140 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4141 &jail_max_af_ips, 0,
4142 "Number of IP addresses a jail may have at most per address family");
4146 * Default parameters for jail(2) compatability. For historical reasons,
4147 * the sysctl names have varying similarity to the parameter names. Prisons
4148 * just see their own parameters, and can't change them.
4151 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4154 int allow, error, i;
4156 pr = req->td->td_ucred->cr_prison;
4157 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4159 /* Get the current flag value, and convert it to a boolean. */
4160 i = (allow & arg2) ? 1 : 0;
4163 error = sysctl_handle_int(oidp, &i, 0, req);
4164 if (error || !req->newptr)
4170 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4173 mtx_lock(&prison0.pr_mtx);
4174 jail_default_allow = (jail_default_allow & ~arg2) | i;
4175 mtx_unlock(&prison0.pr_mtx);
4179 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4180 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4181 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4182 "Processes in jail can set their hostnames");
4183 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4184 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4185 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4186 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4187 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4188 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4189 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4190 "Processes in jail can use System V IPC primitives");
4191 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4192 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4193 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4194 "Prison root can create raw sockets");
4195 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4196 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4197 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4198 "Processes in jail can alter system file flags");
4199 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4200 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4201 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4202 "Processes in jail can mount/unmount jail-friendly file systems");
4203 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4204 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4205 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4206 "Processes in jail can mount the devfs file system");
4207 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4208 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4209 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4210 "Processes in jail can mount the nullfs file system");
4211 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4212 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4213 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4214 "Processes in jail can mount the procfs file system");
4215 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4216 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4217 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4218 "Processes in jail can mount the zfs file system");
4221 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4226 pr = req->td->td_ucred->cr_prison;
4227 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4228 error = sysctl_handle_int(oidp, &level, 0, req);
4229 if (error || !req->newptr)
4231 *(int *)arg1 = level;
4235 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4236 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4237 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4238 sysctl_jail_default_level, "I",
4239 "Processes in jail cannot see all mounted file systems");
4241 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4242 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4243 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4244 sysctl_jail_default_level, "I",
4245 "Ruleset for the devfs filesystem in jail");
4248 * Nodes to describe jail parameters. Maximum length of string parameters
4249 * is returned in the string itself, and the other parameters exist merely
4250 * to make themselves and their types known.
4252 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4256 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4263 switch (oidp->oid_kind & CTLTYPE)
4269 if (!(req->flags & SCTL_MASK32))
4271 return (SYSCTL_OUT(req, &l, sizeof(l)));
4275 return (SYSCTL_OUT(req, &i, sizeof(i)));
4276 case CTLTYPE_STRING:
4277 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4279 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4280 case CTLTYPE_STRUCT:
4282 return (SYSCTL_OUT(req, &s, sizeof(s)));
4287 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4288 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4289 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4290 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4291 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4292 "I", "Jail secure level");
4293 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4294 "I", "Jail cannot see all mounted file systems");
4295 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4296 "I", "Ruleset for in-jail devfs mounts");
4297 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4298 "B", "Jail persistence");
4300 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4301 "E,jailsys", "Virtual network stack");
4303 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4304 "B", "Jail is in the process of shutting down");
4306 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4307 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4308 "I", "Current number of child jails");
4309 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4310 "I", "Maximum number of child jails");
4312 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4313 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4315 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4316 "Jail NIS domainname");
4317 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4319 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4320 "LU", "Jail host ID");
4322 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4323 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4326 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4327 "Jail IPv4 address virtualization");
4328 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4329 "S,in_addr,a", "Jail IPv4 addresses");
4330 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4331 "B", "Do (not) use IPv4 source address selection rather than the "
4332 "primary jail IPv4 address.");
4335 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4336 "Jail IPv6 address virtualization");
4337 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4338 "S,in6_addr,a", "Jail IPv6 addresses");
4339 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4340 "B", "Do (not) use IPv6 source address selection rather than the "
4341 "primary jail IPv6 address.");
4344 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4345 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4346 "B", "Jail may set hostname");
4347 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4348 "B", "Jail may use SYSV IPC");
4349 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4350 "B", "Jail may create raw sockets");
4351 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4352 "B", "Jail may alter system file flags");
4353 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4354 "B", "Jail may set file quotas");
4355 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4356 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4358 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4359 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4360 "B", "Jail may mount/unmount jail-friendly file systems in general");
4361 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4362 "B", "Jail may mount the devfs file system");
4363 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4364 "B", "Jail may mount the nullfs file system");
4365 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4366 "B", "Jail may mount the procfs file system");
4367 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4368 "B", "Jail may mount the zfs file system");
4371 prison_racct_foreach(void (*callback)(struct racct *racct,
4372 void *arg2, void *arg3), void *arg2, void *arg3)
4374 struct prison_racct *prr;
4376 sx_slock(&allprison_lock);
4377 LIST_FOREACH(prr, &allprison_racct, prr_next)
4378 (callback)(prr->prr_racct, arg2, arg3);
4379 sx_sunlock(&allprison_lock);
4382 static struct prison_racct *
4383 prison_racct_find_locked(const char *name)
4385 struct prison_racct *prr;
4387 sx_assert(&allprison_lock, SA_XLOCKED);
4389 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4392 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4393 if (strcmp(name, prr->prr_name) != 0)
4396 /* Found prison_racct with a matching name? */
4397 prison_racct_hold(prr);
4401 /* Add new prison_racct. */
4402 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4403 racct_create(&prr->prr_racct);
4405 strcpy(prr->prr_name, name);
4406 refcount_init(&prr->prr_refcount, 1);
4407 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4412 struct prison_racct *
4413 prison_racct_find(const char *name)
4415 struct prison_racct *prr;
4417 sx_xlock(&allprison_lock);
4418 prr = prison_racct_find_locked(name);
4419 sx_xunlock(&allprison_lock);
4424 prison_racct_hold(struct prison_racct *prr)
4427 refcount_acquire(&prr->prr_refcount);
4431 prison_racct_free(struct prison_racct *prr)
4435 old = prr->prr_refcount;
4436 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4439 sx_xlock(&allprison_lock);
4440 if (refcount_release(&prr->prr_refcount)) {
4441 racct_destroy(&prr->prr_racct);
4442 LIST_REMOVE(prr, prr_next);
4443 sx_xunlock(&allprison_lock);
4444 free(prr, M_PRISON_RACCT);
4448 sx_xunlock(&allprison_lock);
4453 prison_racct_attach(struct prison *pr)
4455 struct prison_racct *prr;
4457 prr = prison_racct_find_locked(pr->pr_name);
4458 KASSERT(prr != NULL, ("cannot find prison_racct"));
4460 pr->pr_prison_racct = prr;
4464 prison_racct_detach(struct prison *pr)
4466 prison_racct_free(pr->pr_prison_racct);
4467 pr->pr_prison_racct = NULL;
4474 db_show_prison(struct prison *pr)
4477 #if defined(INET) || defined(INET6)
4482 char ip6buf[INET6_ADDRSTRLEN];
4485 db_printf("prison %p:\n", pr);
4486 db_printf(" jid = %d\n", pr->pr_id);
4487 db_printf(" name = %s\n", pr->pr_name);
4488 db_printf(" parent = %p\n", pr->pr_parent);
4489 db_printf(" ref = %d\n", pr->pr_ref);
4490 db_printf(" uref = %d\n", pr->pr_uref);
4491 db_printf(" path = %s\n", pr->pr_path);
4492 db_printf(" cpuset = %d\n", pr->pr_cpuset
4493 ? pr->pr_cpuset->cs_id : -1);
4495 db_printf(" vnet = %p\n", pr->pr_vnet);
4497 db_printf(" root = %p\n", pr->pr_root);
4498 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4499 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4500 db_printf(" children.max = %d\n", pr->pr_childmax);
4501 db_printf(" children.cur = %d\n", pr->pr_childcount);
4502 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4503 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4504 db_printf(" flags = 0x%x", pr->pr_flags);
4505 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4507 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4508 db_printf(" %s", pr_flag_names[fi]);
4509 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4511 jsf = pr->pr_flags &
4512 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4513 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4514 pr_flag_jailsys[fi].disable &&
4515 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4516 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4519 db_printf(" allow = 0x%x", pr->pr_allow);
4520 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4522 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4523 db_printf(" %s", pr_allow_names[fi]);
4525 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4526 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4527 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4528 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4529 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4531 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4532 for (ii = 0; ii < pr->pr_ip4s; ii++)
4533 db_printf(" %s %s\n",
4534 ii == 0 ? "ip4.addr =" : " ",
4535 inet_ntoa(pr->pr_ip4[ii]));
4538 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4539 for (ii = 0; ii < pr->pr_ip6s; ii++)
4540 db_printf(" %s %s\n",
4541 ii == 0 ? "ip6.addr =" : " ",
4542 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4546 DB_SHOW_COMMAND(prison, db_show_prison_command)
4552 * Show all prisons in the list, and prison0 which is not
4555 db_show_prison(&prison0);
4556 if (!db_pager_quit) {
4557 TAILQ_FOREACH(pr, &allprison, pr_list) {
4569 /* Look for a prison with the ID and with references. */
4570 TAILQ_FOREACH(pr, &allprison, pr_list)
4571 if (pr->pr_id == addr && pr->pr_ref > 0)
4574 /* Look again, without requiring a reference. */
4575 TAILQ_FOREACH(pr, &allprison, pr_list)
4576 if (pr->pr_id == addr)
4579 /* Assume address points to a valid prison. */
4580 pr = (struct prison *)addr;