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>
54 #include <sys/refcount.h>
56 #include <sys/sysent.h>
57 #include <sys/namei.h>
58 #include <sys/mount.h>
59 #include <sys/queue.h>
60 #include <sys/socket.h>
61 #include <sys/syscallsubr.h>
62 #include <sys/sysctl.h>
63 #include <sys/vnode.h>
68 #include <netinet/in.h>
74 #include <security/mac/mac_framework.h>
76 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
78 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
79 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
81 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
84 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
90 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
92 #define _PR_IP_SADDRSEL 0
96 /* prison0 describes what is "real" about the system. */
97 struct prison prison0 = {
103 .pr_securelevel = -1,
105 .pr_childmax = JAIL_MAX,
106 .pr_hostuuid = DEFAULT_HOSTUUID,
107 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
109 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
111 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
113 .pr_allow = PR_ALLOW_ALL,
115 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
117 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
118 struct sx allprison_lock;
119 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
120 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
121 LIST_HEAD(, prison_racct) allprison_racct;
124 static int do_jail_attach(struct thread *td, struct prison *pr);
125 static void prison_complete(void *context, int pending);
126 static void prison_deref(struct prison *pr, int flags);
127 static char *prison_path(struct prison *pr1, struct prison *pr2);
128 static void prison_remove_one(struct prison *pr);
130 static void prison_racct_attach(struct prison *pr);
131 static void prison_racct_modify(struct prison *pr);
132 static void prison_racct_detach(struct prison *pr);
135 /* Flags for prison_deref */
136 #define PD_DEREF 0x01
137 #define PD_DEUREF 0x02
138 #define PD_LOCKED 0x04
139 #define PD_LIST_SLOCKED 0x08
140 #define PD_LIST_XLOCKED 0x10
143 * Parameter names corresponding to PR_* flag values. Size values are for kvm
144 * as we cannot figure out the size of a sparse array, or an array without a
147 static char *pr_flag_names[] = {
150 [7] = "ip4.saddrsel",
153 [8] = "ip6.saddrsel",
156 const size_t pr_flag_names_size = sizeof(pr_flag_names);
158 static char *pr_flag_nonames[] = {
161 [7] = "ip4.nosaddrsel",
164 [8] = "ip6.nosaddrsel",
167 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
169 struct jailsys_flags {
173 } pr_flag_jailsys[] = {
174 { "host", 0, PR_HOST },
176 { "vnet", 0, PR_VNET },
179 { "ip4", PR_IP4_USER, PR_IP4_USER },
182 { "ip6", PR_IP6_USER, PR_IP6_USER },
185 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
187 static char *pr_allow_names[] = {
188 "allow.set_hostname",
196 "allow.mount.nullfs",
198 "allow.mount.procfs",
200 "allow.mount.fdescfs",
201 "allow.mount.linprocfs",
202 "allow.mount.linsysfs",
205 const size_t pr_allow_names_size = sizeof(pr_allow_names);
207 static char *pr_allow_nonames[] = {
208 "allow.noset_hostname",
210 "allow.noraw_sockets",
215 "allow.mount.nodevfs",
216 "allow.mount.nonullfs",
218 "allow.mount.noprocfs",
219 "allow.mount.notmpfs",
220 "allow.mount.nofdescfs",
221 "allow.mount.nolinprocfs",
222 "allow.mount.nolinsysfs",
223 "allow.noread_msgbuf",
225 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
227 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
228 #define JAIL_DEFAULT_ENFORCE_STATFS 2
229 #define JAIL_DEFAULT_DEVFS_RSNUM 0
230 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
231 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
232 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
233 #if defined(INET) || defined(INET6)
234 static unsigned jail_max_af_ips = 255;
238 * Initialize the parts of prison0 that can't be static-initialized with
239 * constants. This is called from proc0_init() after creating thread0 cpuset.
245 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
246 prison0.pr_osreldate = osreldate;
247 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
256 sys_jail(struct thread *td, struct jail_args *uap)
262 error = copyin(uap->jail, &version, sizeof(uint32_t));
271 /* FreeBSD single IPv4 jails. */
272 bzero(&j, sizeof(struct jail));
273 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
276 j.version = j0.version;
278 j.hostname = j0.hostname;
279 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
285 * Version 1 was used by multi-IPv4 jail implementations
286 * that never made it into the official kernel.
290 case 2: /* JAIL_API_VERSION */
291 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
292 error = copyin(uap->jail, &j, sizeof(struct jail));
298 /* Sci-Fi jails are not supported, sorry. */
301 return (kern_jail(td, &j));
305 kern_jail(struct thread *td, struct jail *j)
307 struct iovec optiov[2 * (4 + nitems(pr_allow_names)
316 char *u_path, *u_hostname, *u_name;
319 struct in_addr *u_ip4;
322 struct in6_addr *u_ip6;
325 int error, enforce_statfs, fi;
327 bzero(&optiov, sizeof(optiov));
328 opt.uio_iov = optiov;
332 opt.uio_segflg = UIO_SYSSPACE;
333 opt.uio_rw = UIO_READ;
336 /* Set permissions for top-level jails from sysctls. */
337 if (!jailed(td->td_ucred)) {
338 for (fi = 0; fi < nitems(pr_allow_names); fi++) {
339 optiov[opt.uio_iovcnt].iov_base =
340 (jail_default_allow & (1 << fi))
341 ? pr_allow_names[fi] : pr_allow_nonames[fi];
342 optiov[opt.uio_iovcnt].iov_len =
343 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
346 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
347 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
349 enforce_statfs = jail_default_enforce_statfs;
350 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
351 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
355 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
357 ip4s = (j->version == 0) ? 1 : j->ip4s;
358 if (ip4s > jail_max_af_ips)
360 tmplen += ip4s * sizeof(struct in_addr);
366 if (j->ip6s > jail_max_af_ips)
368 tmplen += j->ip6s * sizeof(struct in6_addr);
373 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
374 u_hostname = u_path + MAXPATHLEN;
375 u_name = u_hostname + MAXHOSTNAMELEN;
377 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
381 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
383 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
386 optiov[opt.uio_iovcnt].iov_base = "path";
387 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
389 optiov[opt.uio_iovcnt].iov_base = u_path;
390 error = copyinstr(j->path, u_path, MAXPATHLEN,
391 &optiov[opt.uio_iovcnt].iov_len);
393 free(u_path, M_TEMP);
397 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
398 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
400 optiov[opt.uio_iovcnt].iov_base = u_hostname;
401 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
402 &optiov[opt.uio_iovcnt].iov_len);
404 free(u_path, M_TEMP);
408 if (j->jailname != NULL) {
409 optiov[opt.uio_iovcnt].iov_base = "name";
410 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
412 optiov[opt.uio_iovcnt].iov_base = u_name;
413 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
414 &optiov[opt.uio_iovcnt].iov_len);
416 free(u_path, M_TEMP);
422 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
423 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
425 optiov[opt.uio_iovcnt].iov_base = u_ip4;
426 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
428 u_ip4->s_addr = j->ip4s;
430 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
432 free(u_path, M_TEMP);
439 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
440 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
442 optiov[opt.uio_iovcnt].iov_base = u_ip6;
443 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
444 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
446 free(u_path, M_TEMP);
451 KASSERT(opt.uio_iovcnt <= nitems(optiov),
452 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
453 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
454 free(u_path, M_TEMP);
460 * struct jail_set_args {
461 * struct iovec *iovp;
462 * unsigned int iovcnt;
467 sys_jail_set(struct thread *td, struct jail_set_args *uap)
472 /* Check that we have an even number of iovecs. */
476 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
479 error = kern_jail_set(td, auio, uap->flags);
485 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
492 struct in6_addr *ip6;
495 struct vfsoptlist *opts;
496 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
498 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
499 char *g_path, *osrelstr;
500 #if defined(INET) || defined(INET6)
505 size_t namelen, onamelen, pnamelen;
506 int born, created, cuflags, descend, enforce;
507 int error, errmsg_len, errmsg_pos;
508 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
509 int fi, jid, jsys, len, level;
510 int childmax, osreldt, rsnum, slevel;
511 int fullpath_disabled;
512 #if defined(INET) || defined(INET6)
521 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
525 error = priv_check(td, PRIV_JAIL_SET);
526 if (!error && (flags & JAIL_ATTACH))
527 error = priv_check(td, PRIV_JAIL_ATTACH);
530 mypr = td->td_ucred->cr_prison;
531 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
533 if (flags & ~JAIL_SET_MASK)
537 * Check all the parameters before committing to anything. Not all
538 * errors can be caught early, but we may as well try. Also, this
539 * takes care of some expensive stuff (path lookup) before getting
540 * the allprison lock.
542 * XXX Jails are not filesystems, and jail parameters are not mount
543 * options. But it makes more sense to re-use the vfsopt code
544 * than duplicate it under a different name.
546 error = vfs_buildopts(optuio, &opts);
557 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
560 vfs_opterror(opts, "no valid operation (create or update)");
564 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
570 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
579 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
587 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
592 else if (enforce < 0 || enforce > 2) {
598 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
606 pr_flags = ch_flags = 0;
607 for (fi = 0; fi < nitems(pr_flag_names); fi++) {
608 if (pr_flag_names[fi] == NULL)
610 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
611 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
613 ch_flags |= pr_flags;
614 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
615 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
622 case JAIL_SYS_DISABLE:
623 if (!pr_flag_jailsys[fi].disable) {
627 pr_flags |= pr_flag_jailsys[fi].disable;
630 pr_flags |= pr_flag_jailsys[fi].new;
632 case JAIL_SYS_INHERIT:
639 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
641 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
642 && !(pr_flags & PR_PERSIST)) {
644 vfs_opterror(opts, "new jail must persist or attach");
648 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
650 vfs_opterror(opts, "vnet cannot be changed after creation");
655 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
657 vfs_opterror(opts, "ip4 cannot be changed after creation");
662 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
664 vfs_opterror(opts, "ip6 cannot be changed after creation");
669 pr_allow = ch_allow = 0;
670 for (fi = 0; fi < nitems(pr_allow_names); fi++) {
671 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
672 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
674 ch_allow |= pr_allow;
676 error = vfs_getopt(opts, "name", (void **)&name, &len);
682 if (len == 0 || name[len - 1] != '\0') {
686 if (len > MAXHOSTNAMELEN) {
687 error = ENAMETOOLONG;
692 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
700 if (len == 0 || host[len - 1] != '\0') {
704 if (len > MAXHOSTNAMELEN) {
705 error = ENAMETOOLONG;
710 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
718 if (len == 0 || domain[len - 1] != '\0') {
722 if (len > MAXHOSTNAMELEN) {
723 error = ENAMETOOLONG;
728 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
736 if (len == 0 || uuid[len - 1] != '\0') {
740 if (len > HOSTUUIDLEN) {
741 error = ENAMETOOLONG;
746 #ifdef COMPAT_FREEBSD32
747 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
750 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
754 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
766 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
771 else if (ip4s & (sizeof(*ip4) - 1)) {
775 ch_flags |= PR_IP4_USER;
776 pr_flags |= PR_IP4_USER;
778 ip4s /= sizeof(*ip4);
779 if (ip4s > jail_max_af_ips) {
781 vfs_opterror(opts, "too many IPv4 addresses");
784 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
785 bcopy(op, ip4, ip4s * sizeof(*ip4));
787 * IP addresses are all sorted but ip[0] to preserve
788 * the primary IP address as given from userland.
789 * This special IP is used for unbound outgoing
790 * connections as well for "loopback" traffic in case
791 * source address selection cannot find any more fitting
792 * address to connect from.
795 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4),
798 * Check for duplicate addresses and do some simple
799 * zero and broadcast checks. If users give other bogus
800 * addresses it is their problem.
802 * We do not have to care about byte order for these
803 * checks so we will do them in NBO.
805 for (ii = 0; ii < ip4s; ii++) {
806 if (ip4[ii].s_addr == INADDR_ANY ||
807 ip4[ii].s_addr == INADDR_BROADCAST) {
812 (ip4[0].s_addr == ip4[ii+1].s_addr ||
813 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
823 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
828 else if (ip6s & (sizeof(*ip6) - 1)) {
832 ch_flags |= PR_IP6_USER;
833 pr_flags |= PR_IP6_USER;
835 ip6s /= sizeof(*ip6);
836 if (ip6s > jail_max_af_ips) {
838 vfs_opterror(opts, "too many IPv6 addresses");
841 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
842 bcopy(op, ip6, ip6s * sizeof(*ip6));
844 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6),
846 for (ii = 0; ii < ip6s; ii++) {
847 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
852 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
853 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
863 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
864 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
867 "vnet jails cannot have IP address restrictions");
872 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
878 if (flags & JAIL_UPDATE) {
881 "osrelease cannot be changed after creation");
884 if (len == 0 || osrelstr[len - 1] != '\0') {
888 if (len >= OSRELEASELEN) {
889 error = ENAMETOOLONG;
891 "osrelease string must be 1-%d bytes long",
897 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
903 if (flags & JAIL_UPDATE) {
906 "osreldate cannot be changed after creation");
911 vfs_opterror(opts, "osreldate cannot be 0");
916 fullpath_disabled = 0;
918 error = vfs_getopt(opts, "path", (void **)&path, &len);
924 if (flags & JAIL_UPDATE) {
927 "path cannot be changed after creation");
930 if (len == 0 || path[len - 1] != '\0') {
934 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
940 NDFREE(&nd, NDF_ONLY_PNBUF);
941 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
942 strlcpy(g_path, path, MAXPATHLEN);
943 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
946 else if (error == ENODEV) {
947 /* proceed if sysctl debug.disablefullpath == 1 */
948 fullpath_disabled = 1;
949 if (len < 2 || (len == 2 && path[0] == '/'))
952 /* exit on other errors */
955 if (root->v_type != VDIR) {
961 if (fullpath_disabled) {
962 /* Leave room for a real-root full pathname. */
963 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
964 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
965 error = ENAMETOOLONG;
973 * Find the specified jail, or at least its parent.
974 * This abuses the file error codes ENOENT and EEXIST.
978 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
979 namelc = strrchr(name, '.');
980 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
984 sx_xlock(&allprison_lock);
987 * See if a requested jid already exists. There is an
988 * information leak here if the jid exists but is not within
989 * the caller's jail hierarchy. Jail creators will get EEXIST
990 * even though they cannot see the jail, and CREATE | UPDATE
991 * will return ENOENT which is not normally a valid error.
995 vfs_opterror(opts, "negative jid");
996 goto done_unlock_list;
998 pr = prison_find(jid);
1000 ppr = pr->pr_parent;
1001 /* Create: jid must not exist. */
1002 if (cuflags == JAIL_CREATE) {
1003 mtx_unlock(&pr->pr_mtx);
1005 vfs_opterror(opts, "jail %d already exists",
1007 goto done_unlock_list;
1009 if (!prison_ischild(mypr, pr)) {
1010 mtx_unlock(&pr->pr_mtx);
1012 } else if (pr->pr_uref == 0) {
1013 if (!(flags & JAIL_DYING)) {
1014 mtx_unlock(&pr->pr_mtx);
1016 vfs_opterror(opts, "jail %d is dying",
1018 goto done_unlock_list;
1019 } else if ((flags & JAIL_ATTACH) ||
1020 (pr_flags & PR_PERSIST)) {
1022 * A dying jail might be resurrected
1023 * (via attach or persist), but first
1024 * it must determine if another jail
1025 * has claimed its name. Accomplish
1026 * this by implicitly re-setting the
1030 name = prison_name(mypr, pr);
1035 /* Update: jid must exist. */
1036 if (cuflags == JAIL_UPDATE) {
1038 vfs_opterror(opts, "jail %d not found", jid);
1039 goto done_unlock_list;
1044 * If the caller provided a name, look for a jail by that name.
1045 * This has different semantics for creates and updates keyed by jid
1046 * (where the name must not already exist in a different jail),
1047 * and updates keyed by the name itself (where the name must exist
1048 * because that is the jail being updated).
1052 namelc = strrchr(name, '.');
1057 * This is a hierarchical name. Split it into the
1058 * parent and child names, and make sure the parent
1059 * exists or matches an already found jail.
1062 if (strncmp(name, ppr->pr_name, namelc - name)
1063 || ppr->pr_name[namelc - name] != '\0') {
1064 mtx_unlock(&pr->pr_mtx);
1067 "cannot change jail's parent");
1068 goto done_unlock_list;
1072 ppr = prison_find_name(mypr, name);
1076 "jail \"%s\" not found", name);
1077 goto done_unlock_list;
1079 mtx_unlock(&ppr->pr_mtx);
1084 if (namelc[0] != '\0') {
1086 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1089 FOREACH_PRISON_CHILD(ppr, tpr) {
1090 if (tpr != pr && tpr->pr_ref > 0 &&
1091 !strcmp(tpr->pr_name + pnamelen, namelc)) {
1093 cuflags != JAIL_CREATE) {
1094 mtx_lock(&tpr->pr_mtx);
1095 if (tpr->pr_ref > 0) {
1100 if (tpr->pr_uref > 0) {
1106 mtx_unlock(&tpr->pr_mtx);
1107 } else if (tpr->pr_uref > 0) {
1109 * Create, or update(jid):
1110 * name must not exist in an
1111 * active sibling jail.
1115 mtx_unlock(&pr->pr_mtx);
1117 "jail \"%s\" already exists",
1119 goto done_unlock_list;
1123 /* If no active jail is found, use a dying one. */
1124 if (deadpr != NULL && pr == NULL) {
1125 if (flags & JAIL_DYING) {
1126 mtx_lock(&deadpr->pr_mtx);
1127 if (deadpr->pr_ref == 0) {
1128 mtx_unlock(&deadpr->pr_mtx);
1132 } else if (cuflags == JAIL_UPDATE) {
1135 "jail \"%s\" is dying", name);
1136 goto done_unlock_list;
1139 /* Update: name must exist if no jid. */
1140 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1142 vfs_opterror(opts, "jail \"%s\" not found",
1144 goto done_unlock_list;
1148 /* Update: must provide a jid or name. */
1149 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1151 vfs_opterror(opts, "update specified no jail");
1152 goto done_unlock_list;
1155 /* If there's no prison to update, create a new one and link it in. */
1157 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1158 if (tpr->pr_childcount >= tpr->pr_childmax) {
1160 vfs_opterror(opts, "prison limit exceeded");
1161 goto done_unlock_list;
1164 mtx_lock(&ppr->pr_mtx);
1165 if (ppr->pr_ref == 0) {
1166 mtx_unlock(&ppr->pr_mtx);
1168 vfs_opterror(opts, "jail \"%s\" not found",
1169 prison_name(mypr, ppr));
1170 goto done_unlock_list;
1174 mtx_unlock(&ppr->pr_mtx);
1175 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1177 /* Find the next free jid. */
1180 if (jid == JAIL_MAX)
1182 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1183 if (tpr->pr_id < jid)
1185 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1186 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1189 if (jid == lastprid) {
1192 "no available jail IDs");
1194 prison_deref(ppr, PD_DEREF |
1195 PD_DEUREF | PD_LIST_XLOCKED);
1204 * The jail already has a jid (that did not yet exist),
1205 * so just find where to insert it.
1207 TAILQ_FOREACH(tpr, &allprison, pr_list)
1208 if (tpr->pr_id >= jid) {
1209 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1214 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1215 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1216 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1217 tpr->pr_childcount++;
1219 pr->pr_parent = ppr;
1222 /* Set some default values, and inherit some from the parent. */
1227 root = mypr->pr_root;
1230 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1231 pr->pr_flags |= PR_HOST;
1232 #if defined(INET) || defined(INET6)
1234 if (!(pr_flags & PR_VNET))
1238 if (!(ch_flags & PR_IP4_USER))
1239 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1240 else if (!(pr_flags & PR_IP4_USER)) {
1241 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1242 if (ppr->pr_ip4 != NULL) {
1243 pr->pr_ip4s = ppr->pr_ip4s;
1244 pr->pr_ip4 = malloc(pr->pr_ip4s *
1245 sizeof(struct in_addr), M_PRISON,
1247 bcopy(ppr->pr_ip4, pr->pr_ip4,
1248 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1253 if (!(ch_flags & PR_IP6_USER))
1254 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1255 else if (!(pr_flags & PR_IP6_USER)) {
1256 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1257 if (ppr->pr_ip6 != NULL) {
1258 pr->pr_ip6s = ppr->pr_ip6s;
1259 pr->pr_ip6 = malloc(pr->pr_ip6s *
1260 sizeof(struct in6_addr), M_PRISON,
1262 bcopy(ppr->pr_ip6, pr->pr_ip6,
1263 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1269 /* Source address selection is always on by default. */
1270 pr->pr_flags |= _PR_IP_SADDRSEL;
1272 pr->pr_securelevel = ppr->pr_securelevel;
1273 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1274 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1275 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1277 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1278 if (osrelstr == NULL)
1279 strlcpy(pr->pr_osrelease, ppr->pr_osrelease,
1280 sizeof(pr->pr_osrelease));
1282 strlcpy(pr->pr_osrelease, osrelstr,
1283 sizeof(pr->pr_osrelease));
1285 LIST_INIT(&pr->pr_children);
1286 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1287 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1290 /* Allocate a new vnet if specified. */
1291 pr->pr_vnet = (pr_flags & PR_VNET)
1292 ? vnet_alloc() : ppr->pr_vnet;
1295 * Allocate a dedicated cpuset for each jail.
1296 * Unlike other initial settings, this may return an erorr.
1298 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1300 prison_deref(pr, PD_LIST_XLOCKED);
1304 mtx_lock(&pr->pr_mtx);
1306 * New prisons do not yet have a reference, because we do not
1307 * want others to see the incomplete prison once the
1308 * allprison_lock is downgraded.
1313 * Grab a reference for existing prisons, to ensure they
1314 * continue to exist for the duration of the call.
1317 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1318 if ((pr->pr_flags & PR_VNET) &&
1319 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1322 "vnet jails cannot have IP address restrictions");
1323 goto done_deref_locked;
1327 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1330 "ip4 cannot be changed after creation");
1331 goto done_deref_locked;
1335 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1338 "ip6 cannot be changed after creation");
1339 goto done_deref_locked;
1344 /* Do final error checking before setting anything. */
1346 if (slevel < ppr->pr_securelevel) {
1348 goto done_deref_locked;
1352 if (childmax >= ppr->pr_childmax) {
1354 goto done_deref_locked;
1358 if (enforce < ppr->pr_enforce_statfs) {
1360 goto done_deref_locked;
1365 * devfs_rsnum is a uint16_t
1367 if (rsnum < 0 || rsnum > 65535) {
1369 goto done_deref_locked;
1372 * Nested jails always inherit parent's devfs ruleset
1374 if (jailed(td->td_ucred)) {
1375 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1377 goto done_deref_locked;
1379 rsnum = ppr->pr_devfs_rsnum;
1384 if (ppr->pr_flags & PR_IP4) {
1386 * Make sure the new set of IP addresses is a
1387 * subset of the parent's list. Don't worry
1388 * about the parent being unlocked, as any
1389 * setting is done with allprison_lock held.
1391 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1392 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1394 if (ij == ppr->pr_ip4s) {
1396 goto done_deref_locked;
1399 for (ii = ij = 1; ii < ip4s; ii++) {
1400 if (ip4[ii].s_addr ==
1401 ppr->pr_ip4[0].s_addr)
1403 for (; ij < ppr->pr_ip4s; ij++)
1404 if (ip4[ii].s_addr ==
1405 ppr->pr_ip4[ij].s_addr)
1407 if (ij == ppr->pr_ip4s)
1410 if (ij == ppr->pr_ip4s) {
1412 goto done_deref_locked;
1417 * Check for conflicting IP addresses. We permit them
1418 * if there is no more than one IP on each jail. If
1419 * there is a duplicate on a jail with more than one
1420 * IP stop checking and return error.
1423 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1424 if (tppr->pr_flags & PR_VNET)
1429 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1432 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1434 tpr->pr_uref == 0) {
1438 if (!(tpr->pr_flags & PR_IP4_USER))
1441 if (tpr->pr_ip4 == NULL ||
1442 (ip4s == 1 && tpr->pr_ip4s == 1))
1444 for (ii = 0; ii < ip4s; ii++) {
1445 if (prison_check_ip4_locked(tpr, &ip4[ii]) ==
1449 "IPv4 addresses clash");
1450 goto done_deref_locked;
1458 if (ppr->pr_flags & PR_IP6) {
1460 * Make sure the new set of IP addresses is a
1461 * subset of the parent's list.
1463 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1464 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1467 if (ij == ppr->pr_ip6s) {
1469 goto done_deref_locked;
1472 for (ii = ij = 1; ii < ip6s; ii++) {
1473 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1476 for (; ij < ppr->pr_ip6s; ij++)
1477 if (IN6_ARE_ADDR_EQUAL(
1478 &ip6[ii], &ppr->pr_ip6[ij]))
1480 if (ij == ppr->pr_ip6s)
1483 if (ij == ppr->pr_ip6s) {
1485 goto done_deref_locked;
1489 /* Check for conflicting IP addresses. */
1491 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent)
1492 if (tppr->pr_flags & PR_VNET)
1497 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1500 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1502 tpr->pr_uref == 0) {
1506 if (!(tpr->pr_flags & PR_IP6_USER))
1509 if (tpr->pr_ip6 == NULL ||
1510 (ip6s == 1 && tpr->pr_ip6s == 1))
1512 for (ii = 0; ii < ip6s; ii++) {
1513 if (prison_check_ip6_locked(tpr, &ip6[ii]) ==
1517 "IPv6 addresses clash");
1518 goto done_deref_locked;
1524 onamelen = namelen = 0;
1525 if (namelc != NULL) {
1526 /* Give a default name of the jid. Also allow the name to be
1527 * explicitly the jid - but not any other number, and only in
1528 * normal form (no leading zero/etc).
1530 if (namelc[0] == '\0')
1531 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1532 else if ((strtoul(namelc, &p, 10) != jid ||
1533 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1536 "name cannot be numeric (unless it is the jid)");
1537 goto done_deref_locked;
1540 * Make sure the name isn't too long for the prison or its
1543 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1544 onamelen = strlen(pr->pr_name + pnamelen);
1545 namelen = strlen(namelc);
1546 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1547 error = ENAMETOOLONG;
1548 goto done_deref_locked;
1550 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1551 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1552 sizeof(pr->pr_name)) {
1553 error = ENAMETOOLONG;
1554 goto done_deref_locked;
1558 if (pr_allow & ~ppr->pr_allow) {
1560 goto done_deref_locked;
1564 * Let modules check their parameters. This requires unlocking and
1565 * then re-locking the prison, but this is still a valid state as long
1566 * as allprison_lock remains xlocked.
1568 mtx_unlock(&pr->pr_mtx);
1569 error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1571 prison_deref(pr, created
1573 : PD_DEREF | PD_LIST_XLOCKED);
1576 mtx_lock(&pr->pr_mtx);
1578 /* At this point, all valid parameters should have been noted. */
1579 TAILQ_FOREACH(opt, opts, link) {
1580 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1582 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1583 goto done_deref_locked;
1587 /* Set the parameters of the prison. */
1590 if (pr_flags & PR_IP4_USER) {
1591 pr->pr_flags |= PR_IP4;
1592 free(pr->pr_ip4, M_PRISON);
1596 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1598 if (tpr->pr_flags & PR_VNET) {
1603 if (prison_restrict_ip4(tpr, NULL)) {
1612 if (pr_flags & PR_IP6_USER) {
1613 pr->pr_flags |= PR_IP6;
1614 free(pr->pr_ip6, M_PRISON);
1618 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1620 if (tpr->pr_flags & PR_VNET) {
1625 if (prison_restrict_ip6(tpr, NULL)) {
1633 pr->pr_securelevel = slevel;
1634 /* Set all child jails to be at least this level. */
1635 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1636 if (tpr->pr_securelevel < slevel)
1637 tpr->pr_securelevel = slevel;
1640 pr->pr_childmax = childmax;
1641 /* Set all child jails to under this limit. */
1642 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1643 if (tpr->pr_childmax > childmax - level)
1644 tpr->pr_childmax = childmax > level
1645 ? childmax - level : 0;
1648 pr->pr_enforce_statfs = enforce;
1649 /* Pass this restriction on to the children. */
1650 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1651 if (tpr->pr_enforce_statfs < enforce)
1652 tpr->pr_enforce_statfs = enforce;
1655 pr->pr_devfs_rsnum = rsnum;
1656 /* Pass this restriction on to the children. */
1657 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1658 tpr->pr_devfs_rsnum = rsnum;
1660 if (namelc != NULL) {
1661 if (ppr == &prison0)
1662 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1664 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1665 ppr->pr_name, namelc);
1666 /* Change this component of child names. */
1667 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1668 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1669 strlen(tpr->pr_name + onamelen) + 1);
1670 bcopy(pr->pr_name, tpr->pr_name, namelen);
1674 /* Try to keep a real-rooted full pathname. */
1675 if (fullpath_disabled && path[0] == '/' &&
1676 strcmp(mypr->pr_path, "/"))
1677 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1678 mypr->pr_path, path);
1680 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1683 if (PR_HOST & ch_flags & ~pr_flags) {
1684 if (pr->pr_flags & PR_HOST) {
1686 * Copy the parent's host info. As with pr_ip4 above,
1687 * the lack of a lock on the parent is not a problem;
1688 * it is always set with allprison_lock at least
1689 * shared, and is held exclusively here.
1691 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1692 sizeof(pr->pr_hostname));
1693 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1694 sizeof(pr->pr_domainname));
1695 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1696 sizeof(pr->pr_hostuuid));
1697 pr->pr_hostid = pr->pr_parent->pr_hostid;
1699 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1700 /* Set this prison, and any descendants without PR_HOST. */
1702 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1704 strlcpy(pr->pr_domainname, domain,
1705 sizeof(pr->pr_domainname));
1707 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1709 pr->pr_hostid = hid;
1710 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1711 if (tpr->pr_flags & PR_HOST)
1715 strlcpy(tpr->pr_hostname,
1717 sizeof(tpr->pr_hostname));
1719 strlcpy(tpr->pr_domainname,
1721 sizeof(tpr->pr_domainname));
1723 strlcpy(tpr->pr_hostuuid,
1725 sizeof(tpr->pr_hostuuid));
1727 tpr->pr_hostid = hid;
1731 if ((tallow = ch_allow & ~pr_allow)) {
1732 /* Clear allow bits in all children. */
1733 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1734 tpr->pr_allow &= ~tallow;
1736 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1738 * Persistent prisons get an extra reference, and prisons losing their
1739 * persist flag lose that reference. Only do this for existing prisons
1740 * for now, so new ones will remain unseen until after the module
1741 * handlers have completed.
1743 born = pr->pr_uref == 0;
1744 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1745 if (pr_flags & PR_PERSIST) {
1753 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1754 mtx_unlock(&pr->pr_mtx);
1757 if (racct_enable && created)
1758 prison_racct_attach(pr);
1761 /* Locks may have prevented a complete restriction of child IP
1762 * addresses. If so, allocate some more memory and try again.
1767 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1768 mtx_lock(&pr->pr_mtx);
1770 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1772 if (tpr->pr_flags & PR_VNET) {
1777 if (prison_restrict_ip4(tpr, ip4)) {
1784 mtx_unlock(&pr->pr_mtx);
1790 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1791 mtx_lock(&pr->pr_mtx);
1793 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1795 if (tpr->pr_flags & PR_VNET) {
1800 if (prison_restrict_ip6(tpr, ip6)) {
1807 mtx_unlock(&pr->pr_mtx);
1811 /* Let the modules do their work. */
1812 sx_downgrade(&allprison_lock);
1814 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1816 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1817 prison_deref(pr, created
1819 : PD_DEREF | PD_LIST_SLOCKED);
1823 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1826 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1827 prison_deref(pr, created
1829 : PD_DEREF | PD_LIST_SLOCKED);
1833 /* Attach this process to the prison if requested. */
1834 if (flags & JAIL_ATTACH) {
1835 mtx_lock(&pr->pr_mtx);
1836 error = do_jail_attach(td, pr);
1838 vfs_opterror(opts, "attach failed");
1840 prison_deref(pr, PD_DEREF);
1846 if (racct_enable && !created) {
1847 if (!(flags & JAIL_ATTACH))
1848 sx_sunlock(&allprison_lock);
1849 prison_racct_modify(pr);
1850 if (!(flags & JAIL_ATTACH))
1851 sx_slock(&allprison_lock);
1855 td->td_retval[0] = pr->pr_id;
1858 * Now that it is all there, drop the temporary reference from existing
1859 * prisons. Or add a reference to newly created persistent prisons
1860 * (which was not done earlier so that the prison would not be publicly
1864 prison_deref(pr, (flags & JAIL_ATTACH)
1866 : PD_DEREF | PD_LIST_SLOCKED);
1868 if (pr_flags & PR_PERSIST) {
1869 mtx_lock(&pr->pr_mtx);
1872 mtx_unlock(&pr->pr_mtx);
1874 if (!(flags & JAIL_ATTACH))
1875 sx_sunlock(&allprison_lock);
1881 prison_deref(pr, created
1882 ? PD_LOCKED | PD_LIST_XLOCKED
1883 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1886 sx_xunlock(&allprison_lock);
1892 if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1893 &errmsg_len) == 0 && errmsg_len > 0) {
1894 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1895 if (optuio->uio_segflg == UIO_SYSSPACE)
1897 optuio->uio_iov[errmsg_pos].iov_base,
1901 optuio->uio_iov[errmsg_pos].iov_base,
1907 free(ip4, M_PRISON);
1910 free(ip6, M_PRISON);
1913 free(g_path, M_TEMP);
1920 * struct jail_get_args {
1921 * struct iovec *iovp;
1922 * unsigned int iovcnt;
1927 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1932 /* Check that we have an even number of iovecs. */
1933 if (uap->iovcnt & 1)
1936 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1939 error = kern_jail_get(td, auio, uap->flags);
1941 error = copyout(auio->uio_iov, uap->iovp,
1942 uap->iovcnt * sizeof (struct iovec));
1948 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1950 struct prison *pr, *mypr;
1952 struct vfsoptlist *opts;
1953 char *errmsg, *name;
1954 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1956 if (flags & ~JAIL_GET_MASK)
1959 /* Get the parameter list. */
1960 error = vfs_buildopts(optuio, &opts);
1963 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1964 mypr = td->td_ucred->cr_prison;
1967 * Find the prison specified by one of: lastjid, jid, name.
1969 sx_slock(&allprison_lock);
1970 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1972 TAILQ_FOREACH(pr, &allprison, pr_list) {
1973 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1974 mtx_lock(&pr->pr_mtx);
1975 if (pr->pr_ref > 0 &&
1976 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1978 mtx_unlock(&pr->pr_mtx);
1984 vfs_opterror(opts, "no jail after %d", jid);
1985 goto done_unlock_list;
1986 } else if (error != ENOENT)
1987 goto done_unlock_list;
1989 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1992 pr = prison_find_child(mypr, jid);
1994 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1995 mtx_unlock(&pr->pr_mtx);
1997 vfs_opterror(opts, "jail %d is dying",
1999 goto done_unlock_list;
2004 vfs_opterror(opts, "jail %d not found", jid);
2005 goto done_unlock_list;
2007 } else if (error != ENOENT)
2008 goto done_unlock_list;
2010 error = vfs_getopt(opts, "name", (void **)&name, &len);
2012 if (len == 0 || name[len - 1] != '\0') {
2014 goto done_unlock_list;
2016 pr = prison_find_name(mypr, name);
2018 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2019 mtx_unlock(&pr->pr_mtx);
2021 vfs_opterror(opts, "jail \"%s\" is dying",
2023 goto done_unlock_list;
2028 vfs_opterror(opts, "jail \"%s\" not found", name);
2029 goto done_unlock_list;
2030 } else if (error != ENOENT)
2031 goto done_unlock_list;
2033 vfs_opterror(opts, "no jail specified");
2035 goto done_unlock_list;
2038 /* Get the parameters of the prison. */
2041 td->td_retval[0] = pr->pr_id;
2042 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2043 if (error != 0 && error != ENOENT)
2045 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2046 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2047 if (error != 0 && error != ENOENT)
2049 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2050 if (error != 0 && error != ENOENT)
2052 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2053 sizeof(pr->pr_cpuset->cs_id));
2054 if (error != 0 && error != ENOENT)
2056 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2057 if (error != 0 && error != ENOENT)
2060 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2061 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2062 if (error != 0 && error != ENOENT)
2066 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2067 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2068 if (error != 0 && error != ENOENT)
2071 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2072 sizeof(pr->pr_securelevel));
2073 if (error != 0 && error != ENOENT)
2075 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2076 sizeof(pr->pr_childcount));
2077 if (error != 0 && error != ENOENT)
2079 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2080 sizeof(pr->pr_childmax));
2081 if (error != 0 && error != ENOENT)
2083 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2084 if (error != 0 && error != ENOENT)
2086 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2087 if (error != 0 && error != ENOENT)
2089 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2090 if (error != 0 && error != ENOENT)
2092 #ifdef COMPAT_FREEBSD32
2093 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2094 uint32_t hid32 = pr->pr_hostid;
2096 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2099 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2100 sizeof(pr->pr_hostid));
2101 if (error != 0 && error != ENOENT)
2103 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2104 sizeof(pr->pr_enforce_statfs));
2105 if (error != 0 && error != ENOENT)
2107 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2108 sizeof(pr->pr_devfs_rsnum));
2109 if (error != 0 && error != ENOENT)
2111 for (fi = 0; fi < nitems(pr_flag_names); fi++) {
2112 if (pr_flag_names[fi] == NULL)
2114 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2115 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2116 if (error != 0 && error != ENOENT)
2119 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2120 if (error != 0 && error != ENOENT)
2123 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
2125 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2126 i = pr_flag_jailsys[fi].disable &&
2127 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2128 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2131 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2132 if (error != 0 && error != ENOENT)
2135 for (fi = 0; fi < nitems(pr_allow_names); fi++) {
2136 if (pr_allow_names[fi] == NULL)
2138 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2139 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2140 if (error != 0 && error != ENOENT)
2143 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2144 if (error != 0 && error != ENOENT)
2147 i = (pr->pr_uref == 0);
2148 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2149 if (error != 0 && error != ENOENT)
2152 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2153 if (error != 0 && error != ENOENT)
2155 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2156 sizeof(pr->pr_osreldate));
2157 if (error != 0 && error != ENOENT)
2159 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2160 if (error != 0 && error != ENOENT)
2163 /* Get the module parameters. */
2164 mtx_unlock(&pr->pr_mtx);
2166 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2169 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2171 /* By now, all parameters should have been noted. */
2172 TAILQ_FOREACH(opt, opts, link) {
2173 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2175 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2180 /* Write the fetched parameters back to userspace. */
2182 TAILQ_FOREACH(opt, opts, link) {
2183 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2184 pos = 2 * opt->pos + 1;
2185 optuio->uio_iov[pos].iov_len = opt->len;
2186 if (opt->value != NULL) {
2187 if (optuio->uio_segflg == UIO_SYSSPACE) {
2189 optuio->uio_iov[pos].iov_base,
2192 error = copyout(opt->value,
2193 optuio->uio_iov[pos].iov_base,
2204 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2208 sx_sunlock(&allprison_lock);
2210 if (error && errmsg_pos >= 0) {
2211 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2212 errmsg_pos = 2 * errmsg_pos + 1;
2213 if (errmsg_len > 0) {
2214 if (optuio->uio_segflg == UIO_SYSSPACE)
2216 optuio->uio_iov[errmsg_pos].iov_base,
2220 optuio->uio_iov[errmsg_pos].iov_base,
2230 * struct jail_remove_args {
2235 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2237 struct prison *pr, *cpr, *lpr, *tpr;
2240 error = priv_check(td, PRIV_JAIL_REMOVE);
2244 sx_xlock(&allprison_lock);
2245 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2247 sx_xunlock(&allprison_lock);
2251 /* Remove all descendants of this prison, then remove this prison. */
2253 if (!LIST_EMPTY(&pr->pr_children)) {
2254 mtx_unlock(&pr->pr_mtx);
2256 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2257 mtx_lock(&cpr->pr_mtx);
2258 if (cpr->pr_ref > 0) {
2262 /* Already removed - do not do it again. */
2265 mtx_unlock(&cpr->pr_mtx);
2267 mtx_lock(&lpr->pr_mtx);
2268 prison_remove_one(lpr);
2269 sx_xlock(&allprison_lock);
2274 mtx_lock(&lpr->pr_mtx);
2275 prison_remove_one(lpr);
2276 sx_xlock(&allprison_lock);
2278 mtx_lock(&pr->pr_mtx);
2280 prison_remove_one(pr);
2285 prison_remove_one(struct prison *pr)
2290 /* If the prison was persistent, it is not anymore. */
2292 if (pr->pr_flags & PR_PERSIST) {
2295 pr->pr_flags &= ~PR_PERSIST;
2299 * jail_remove added a reference. If that's the only one, remove
2302 KASSERT(pr->pr_ref > 0,
2303 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2304 if (pr->pr_ref == 1) {
2306 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2310 mtx_unlock(&pr->pr_mtx);
2311 sx_xunlock(&allprison_lock);
2313 * Kill all processes unfortunate enough to be attached to this prison.
2315 sx_slock(&allproc_lock);
2316 LIST_FOREACH(p, &allproc, p_list) {
2318 if (p->p_state != PRS_NEW && p->p_ucred &&
2319 p->p_ucred->cr_prison == pr)
2320 kern_psignal(p, SIGKILL);
2323 sx_sunlock(&allproc_lock);
2324 /* Remove the temporary reference added by jail_remove. */
2325 prison_deref(pr, deuref | PD_DEREF);
2330 * struct jail_attach_args {
2335 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2340 error = priv_check(td, PRIV_JAIL_ATTACH);
2345 * Start with exclusive hold on allprison_lock to ensure that a possible
2346 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2347 * But then immediately downgrade it since we don't need to stop
2350 sx_xlock(&allprison_lock);
2351 sx_downgrade(&allprison_lock);
2352 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2354 sx_sunlock(&allprison_lock);
2359 * Do not allow a process to attach to a prison that is not
2360 * considered to be "alive".
2362 if (pr->pr_uref == 0) {
2363 mtx_unlock(&pr->pr_mtx);
2364 sx_sunlock(&allprison_lock);
2368 return (do_jail_attach(td, pr));
2372 do_jail_attach(struct thread *td, struct prison *pr)
2375 struct ucred *newcred, *oldcred;
2379 * XXX: Note that there is a slight race here if two threads
2380 * in the same privileged process attempt to attach to two
2381 * different jails at the same time. It is important for
2382 * user processes not to do this, or they might end up with
2383 * a process root from one prison, but attached to the jail
2388 mtx_unlock(&pr->pr_mtx);
2390 /* Let modules do whatever they need to prepare for attaching. */
2391 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2393 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2396 sx_sunlock(&allprison_lock);
2399 * Reparent the newly attached process to this jail.
2402 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2406 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2407 if ((error = change_dir(pr->pr_root, td)) != 0)
2410 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2413 VOP_UNLOCK(pr->pr_root, 0);
2414 if ((error = pwd_chroot(td, pr->pr_root)))
2419 oldcred = crcopysafe(p, newcred);
2420 newcred->cr_prison = pr;
2421 proc_set_cred(p, newcred);
2424 racct_proc_ucred_changed(p, oldcred, newcred);
2429 rctl_proc_ucred_changed(p, newcred);
2432 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2437 VOP_UNLOCK(pr->pr_root, 0);
2439 /* Tell modules this thread is still in its old jail after all. */
2440 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2441 prison_deref(pr, PD_DEREF | PD_DEUREF);
2447 * Returns a locked prison instance, or NULL on failure.
2450 prison_find(int prid)
2454 sx_assert(&allprison_lock, SX_LOCKED);
2455 TAILQ_FOREACH(pr, &allprison, pr_list) {
2456 if (pr->pr_id == prid) {
2457 mtx_lock(&pr->pr_mtx);
2460 mtx_unlock(&pr->pr_mtx);
2467 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2470 prison_find_child(struct prison *mypr, int prid)
2475 sx_assert(&allprison_lock, SX_LOCKED);
2476 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2477 if (pr->pr_id == prid) {
2478 mtx_lock(&pr->pr_mtx);
2481 mtx_unlock(&pr->pr_mtx);
2488 * Look for the name relative to mypr. Returns a locked prison or NULL.
2491 prison_find_name(struct prison *mypr, const char *name)
2493 struct prison *pr, *deadpr;
2497 sx_assert(&allprison_lock, SX_LOCKED);
2498 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2501 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2502 if (!strcmp(pr->pr_name + mylen, name)) {
2503 mtx_lock(&pr->pr_mtx);
2504 if (pr->pr_ref > 0) {
2505 if (pr->pr_uref > 0)
2509 mtx_unlock(&pr->pr_mtx);
2512 /* There was no valid prison - perhaps there was a dying one. */
2513 if (deadpr != NULL) {
2514 mtx_lock(&deadpr->pr_mtx);
2515 if (deadpr->pr_ref == 0) {
2516 mtx_unlock(&deadpr->pr_mtx);
2524 * See if a prison has the specific flag set.
2527 prison_flag(struct ucred *cred, unsigned flag)
2530 /* This is an atomic read, so no locking is necessary. */
2531 return (cred->cr_prison->pr_flags & flag);
2535 prison_allow(struct ucred *cred, unsigned flag)
2538 /* This is an atomic read, so no locking is necessary. */
2539 return (cred->cr_prison->pr_allow & flag);
2543 * Remove a prison reference. If that was the last reference, remove the
2544 * prison itself - but not in this context in case there are locks held.
2547 prison_free_locked(struct prison *pr)
2551 mtx_assert(&pr->pr_mtx, MA_OWNED);
2553 mtx_unlock(&pr->pr_mtx);
2555 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2559 prison_free(struct prison *pr)
2562 mtx_lock(&pr->pr_mtx);
2563 prison_free_locked(pr);
2567 * Complete a call to either prison_free or prison_proc_free.
2570 prison_complete(void *context, int pending)
2572 struct prison *pr = context;
2574 sx_xlock(&allprison_lock);
2575 mtx_lock(&pr->pr_mtx);
2576 prison_deref(pr, pr->pr_uref
2577 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2578 : PD_LOCKED | PD_LIST_XLOCKED);
2582 * Remove a prison reference (usually). This internal version assumes no
2583 * mutexes are held, except perhaps the prison itself. If there are no more
2584 * references, release and delist the prison. On completion, the prison lock
2585 * and the allprison lock are both unlocked.
2588 prison_deref(struct prison *pr, int flags)
2590 struct prison *ppr, *tpr;
2593 if (!(flags & PD_LOCKED))
2594 mtx_lock(&pr->pr_mtx);
2596 if (flags & PD_DEUREF) {
2597 KASSERT(pr->pr_uref > 0,
2598 ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2601 lasturef = pr->pr_uref == 0;
2604 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2607 if (flags & PD_DEREF) {
2608 KASSERT(pr->pr_ref > 0,
2609 ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2614 mtx_unlock(&pr->pr_mtx);
2617 * Tell the modules if the last user reference was removed
2618 * (even it sticks around in dying state).
2621 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2622 sx_xlock(&allprison_lock);
2623 flags |= PD_LIST_XLOCKED;
2625 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2626 mtx_lock(&pr->pr_mtx);
2628 mtx_unlock(&pr->pr_mtx);
2631 /* If the prison still has references, nothing else to do. */
2633 if (flags & PD_LIST_SLOCKED)
2634 sx_sunlock(&allprison_lock);
2635 else if (flags & PD_LIST_XLOCKED)
2636 sx_xunlock(&allprison_lock);
2640 if (flags & PD_LIST_SLOCKED) {
2641 if (!sx_try_upgrade(&allprison_lock)) {
2642 sx_sunlock(&allprison_lock);
2643 sx_xlock(&allprison_lock);
2645 } else if (!(flags & PD_LIST_XLOCKED))
2646 sx_xlock(&allprison_lock);
2648 TAILQ_REMOVE(&allprison, pr, pr_list);
2649 LIST_REMOVE(pr, pr_sibling);
2650 ppr = pr->pr_parent;
2651 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2652 tpr->pr_childcount--;
2653 sx_xunlock(&allprison_lock);
2656 if (pr->pr_vnet != ppr->pr_vnet)
2657 vnet_destroy(pr->pr_vnet);
2659 if (pr->pr_root != NULL)
2661 mtx_destroy(&pr->pr_mtx);
2663 free(pr->pr_ip4, M_PRISON);
2666 free(pr->pr_ip6, M_PRISON);
2668 if (pr->pr_cpuset != NULL)
2669 cpuset_rel(pr->pr_cpuset);
2673 prison_racct_detach(pr);
2677 /* Removing a prison frees a reference on its parent. */
2679 mtx_lock(&pr->pr_mtx);
2680 flags = PD_DEREF | PD_DEUREF;
2685 prison_hold_locked(struct prison *pr)
2688 mtx_assert(&pr->pr_mtx, MA_OWNED);
2689 KASSERT(pr->pr_ref > 0,
2690 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2695 prison_hold(struct prison *pr)
2698 mtx_lock(&pr->pr_mtx);
2699 prison_hold_locked(pr);
2700 mtx_unlock(&pr->pr_mtx);
2704 prison_proc_hold(struct prison *pr)
2707 mtx_lock(&pr->pr_mtx);
2708 KASSERT(pr->pr_uref > 0,
2709 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2711 mtx_unlock(&pr->pr_mtx);
2715 prison_proc_free(struct prison *pr)
2718 mtx_lock(&pr->pr_mtx);
2719 KASSERT(pr->pr_uref > 0,
2720 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2721 if (pr->pr_uref > 1)
2725 * Don't remove the last user reference in this context, which
2726 * is expected to be a process that is not only locked, but
2730 mtx_unlock(&pr->pr_mtx);
2731 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2734 mtx_unlock(&pr->pr_mtx);
2738 * Check if a jail supports the given address family.
2740 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
2744 prison_check_af(struct ucred *cred, int af)
2749 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2751 pr = cred->cr_prison;
2753 /* Prisons with their own network stack are not limited. */
2754 if (prison_owns_vnet(cred))
2763 if (pr->pr_flags & PR_IP4)
2765 mtx_lock(&pr->pr_mtx);
2766 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
2767 error = EAFNOSUPPORT;
2768 mtx_unlock(&pr->pr_mtx);
2774 if (pr->pr_flags & PR_IP6)
2776 mtx_lock(&pr->pr_mtx);
2777 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
2778 error = EAFNOSUPPORT;
2779 mtx_unlock(&pr->pr_mtx);
2787 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
2788 error = EAFNOSUPPORT;
2794 * Check if given address belongs to the jail referenced by cred (wrapper to
2795 * prison_check_ip[46]).
2797 * Returns 0 if jail doesn't restrict the address family or if address belongs
2798 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
2799 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
2802 prison_if(struct ucred *cred, struct sockaddr *sa)
2805 struct sockaddr_in *sai;
2808 struct sockaddr_in6 *sai6;
2812 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2813 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
2816 if (prison_owns_vnet(cred))
2821 switch (sa->sa_family)
2825 sai = (struct sockaddr_in *)sa;
2826 error = prison_check_ip4(cred, &sai->sin_addr);
2831 sai6 = (struct sockaddr_in6 *)sa;
2832 error = prison_check_ip6(cred, &sai6->sin6_addr);
2836 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
2837 error = EAFNOSUPPORT;
2843 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
2846 prison_check(struct ucred *cred1, struct ucred *cred2)
2849 return ((cred1->cr_prison == cred2->cr_prison ||
2850 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
2854 * Return 1 if p2 is a child of p1, otherwise 0.
2857 prison_ischild(struct prison *pr1, struct prison *pr2)
2860 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
2867 * Return 1 if the passed credential is in a jail, otherwise 0.
2870 jailed(struct ucred *cred)
2873 return (cred->cr_prison != &prison0);
2877 * Return 1 if the passed credential is in a jail and that jail does not
2878 * have its own virtual network stack, otherwise 0.
2881 jailed_without_vnet(struct ucred *cred)
2887 if (prison_owns_vnet(cred))
2895 * Return the correct hostname (domainname, et al) for the passed credential.
2898 getcredhostname(struct ucred *cred, char *buf, size_t size)
2903 * A NULL credential can be used to shortcut to the physical
2904 * system's hostname.
2906 pr = (cred != NULL) ? cred->cr_prison : &prison0;
2907 mtx_lock(&pr->pr_mtx);
2908 strlcpy(buf, pr->pr_hostname, size);
2909 mtx_unlock(&pr->pr_mtx);
2913 getcreddomainname(struct ucred *cred, char *buf, size_t size)
2916 mtx_lock(&cred->cr_prison->pr_mtx);
2917 strlcpy(buf, cred->cr_prison->pr_domainname, size);
2918 mtx_unlock(&cred->cr_prison->pr_mtx);
2922 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
2925 mtx_lock(&cred->cr_prison->pr_mtx);
2926 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
2927 mtx_unlock(&cred->cr_prison->pr_mtx);
2931 getcredhostid(struct ucred *cred, unsigned long *hostid)
2934 mtx_lock(&cred->cr_prison->pr_mtx);
2935 *hostid = cred->cr_prison->pr_hostid;
2936 mtx_unlock(&cred->cr_prison->pr_mtx);
2941 * Determine whether the prison represented by cred owns
2942 * its vnet rather than having it inherited.
2944 * Returns 1 in case the prison owns the vnet, 0 otherwise.
2947 prison_owns_vnet(struct ucred *cred)
2951 * vnets cannot be added/removed after jail creation,
2952 * so no need to lock here.
2954 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
2959 * Determine whether the subject represented by cred can "see"
2960 * status of a mount point.
2961 * Returns: 0 for permitted, ENOENT otherwise.
2962 * XXX: This function should be called cr_canseemount() and should be
2963 * placed in kern_prot.c.
2966 prison_canseemount(struct ucred *cred, struct mount *mp)
2972 pr = cred->cr_prison;
2973 if (pr->pr_enforce_statfs == 0)
2975 if (pr->pr_root->v_mount == mp)
2977 if (pr->pr_enforce_statfs == 2)
2980 * If jail's chroot directory is set to "/" we should be able to see
2981 * all mount-points from inside a jail.
2982 * This is ugly check, but this is the only situation when jail's
2983 * directory ends with '/'.
2985 if (strcmp(pr->pr_path, "/") == 0)
2987 len = strlen(pr->pr_path);
2989 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
2992 * Be sure that we don't have situation where jail's root directory
2993 * is "/some/path" and mount point is "/some/pathpath".
2995 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3001 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3003 char jpath[MAXPATHLEN];
3007 pr = cred->cr_prison;
3008 if (pr->pr_enforce_statfs == 0)
3010 if (prison_canseemount(cred, mp) != 0) {
3011 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3012 strlcpy(sp->f_mntonname, "[restricted]",
3013 sizeof(sp->f_mntonname));
3016 if (pr->pr_root->v_mount == mp) {
3018 * Clear current buffer data, so we are sure nothing from
3019 * the valid path left there.
3021 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3022 *sp->f_mntonname = '/';
3026 * If jail's chroot directory is set to "/" we should be able to see
3027 * all mount-points from inside a jail.
3029 if (strcmp(pr->pr_path, "/") == 0)
3031 len = strlen(pr->pr_path);
3032 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3034 * Clear current buffer data, so we are sure nothing from
3035 * the valid path left there.
3037 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3038 if (*jpath == '\0') {
3039 /* Should never happen. */
3040 *sp->f_mntonname = '/';
3042 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3047 * Check with permission for a specific privilege is granted within jail. We
3048 * have a specific list of accepted privileges; the rest are denied.
3051 prison_priv_check(struct ucred *cred, int priv)
3059 * Privileges specific to prisons with a virtual network stack.
3060 * There might be a duplicate entry here in case the privilege
3061 * is only granted conditionally in the legacy jail case.
3066 * NFS-specific privileges.
3068 case PRIV_NFS_DAEMON:
3069 case PRIV_NFS_LOCKD:
3072 * Network stack privileges.
3074 case PRIV_NET_BRIDGE:
3077 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3078 case PRIV_NET_ROUTE:
3080 case PRIV_NET_SETIFMTU:
3081 case PRIV_NET_SETIFFLAGS:
3082 case PRIV_NET_SETIFCAP:
3083 case PRIV_NET_SETIFDESCR:
3084 case PRIV_NET_SETIFNAME :
3085 case PRIV_NET_SETIFMETRIC:
3086 case PRIV_NET_SETIFPHYS:
3087 case PRIV_NET_SETIFMAC:
3088 case PRIV_NET_ADDMULTI:
3089 case PRIV_NET_DELMULTI:
3090 case PRIV_NET_HWIOCTL:
3091 case PRIV_NET_SETLLADDR:
3092 case PRIV_NET_ADDIFGROUP:
3093 case PRIV_NET_DELIFGROUP:
3094 case PRIV_NET_IFCREATE:
3095 case PRIV_NET_IFDESTROY:
3096 case PRIV_NET_ADDIFADDR:
3097 case PRIV_NET_DELIFADDR:
3100 case PRIV_NET_SETIFVNET:
3101 case PRIV_NET_SETIFFIB:
3104 * 802.11-related privileges.
3106 case PRIV_NET80211_GETKEY:
3108 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3115 case PRIV_NETATM_CFG:
3116 case PRIV_NETATM_ADD:
3117 case PRIV_NETATM_DEL:
3118 case PRIV_NETATM_SET:
3121 * Bluetooth privileges.
3123 case PRIV_NETBLUETOOTH_RAW:
3127 * Netgraph and netgraph module privileges.
3129 case PRIV_NETGRAPH_CONTROL:
3131 case PRIV_NETGRAPH_TTY:
3135 * IPv4 and IPv6 privileges.
3137 case PRIV_NETINET_IPFW:
3138 case PRIV_NETINET_DIVERT:
3139 case PRIV_NETINET_PF:
3140 case PRIV_NETINET_DUMMYNET:
3141 case PRIV_NETINET_CARP:
3142 case PRIV_NETINET_MROUTE:
3143 case PRIV_NETINET_RAW:
3144 case PRIV_NETINET_ADDRCTRL6:
3145 case PRIV_NETINET_ND6:
3146 case PRIV_NETINET_SCOPE6:
3147 case PRIV_NETINET_ALIFETIME6:
3148 case PRIV_NETINET_IPSEC:
3149 case PRIV_NETINET_BINDANY:
3164 * No default: or deny here.
3165 * In case of no permit fall through to next switch().
3167 if (cred->cr_prison->pr_flags & PR_VNET)
3175 * Allow ktrace privileges for root in jail.
3181 * Allow jailed processes to configure audit identity and
3182 * submit audit records (login, etc). In the future we may
3183 * want to further refine the relationship between audit and
3186 case PRIV_AUDIT_GETAUDIT:
3187 case PRIV_AUDIT_SETAUDIT:
3188 case PRIV_AUDIT_SUBMIT:
3192 * Allow jailed processes to manipulate process UNIX
3193 * credentials in any way they see fit.
3195 case PRIV_CRED_SETUID:
3196 case PRIV_CRED_SETEUID:
3197 case PRIV_CRED_SETGID:
3198 case PRIV_CRED_SETEGID:
3199 case PRIV_CRED_SETGROUPS:
3200 case PRIV_CRED_SETREUID:
3201 case PRIV_CRED_SETREGID:
3202 case PRIV_CRED_SETRESUID:
3203 case PRIV_CRED_SETRESGID:
3206 * Jail implements visibility constraints already, so allow
3207 * jailed root to override uid/gid-based constraints.
3209 case PRIV_SEEOTHERGIDS:
3210 case PRIV_SEEOTHERUIDS:
3213 * Jail implements inter-process debugging limits already, so
3214 * allow jailed root various debugging privileges.
3216 case PRIV_DEBUG_DIFFCRED:
3217 case PRIV_DEBUG_SUGID:
3218 case PRIV_DEBUG_UNPRIV:
3221 * Allow jail to set various resource limits and login
3222 * properties, and for now, exceed process resource limits.
3224 case PRIV_PROC_LIMIT:
3225 case PRIV_PROC_SETLOGIN:
3226 case PRIV_PROC_SETRLIMIT:
3229 * System V and POSIX IPC privileges are granted in jail.
3232 case PRIV_IPC_WRITE:
3233 case PRIV_IPC_ADMIN:
3234 case PRIV_IPC_MSGSIZE:
3238 * Jail operations within a jail work on child jails.
3240 case PRIV_JAIL_ATTACH:
3242 case PRIV_JAIL_REMOVE:
3245 * Jail implements its own inter-process limits, so allow
3246 * root processes in jail to change scheduling on other
3247 * processes in the same jail. Likewise for signalling.
3249 case PRIV_SCHED_DIFFCRED:
3250 case PRIV_SCHED_CPUSET:
3251 case PRIV_SIGNAL_DIFFCRED:
3252 case PRIV_SIGNAL_SUGID:
3255 * Allow jailed processes to write to sysctls marked as jail
3258 case PRIV_SYSCTL_WRITEJAIL:
3261 * Allow root in jail to manage a variety of quota
3262 * properties. These should likely be conditional on a
3263 * configuration option.
3265 case PRIV_VFS_GETQUOTA:
3266 case PRIV_VFS_SETQUOTA:
3269 * Since Jail relies on chroot() to implement file system
3270 * protections, grant many VFS privileges to root in jail.
3271 * Be careful to exclude mount-related and NFS-related
3275 case PRIV_VFS_WRITE:
3276 case PRIV_VFS_ADMIN:
3278 case PRIV_VFS_LOOKUP:
3279 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3280 case PRIV_VFS_CHFLAGS_DEV:
3281 case PRIV_VFS_CHOWN:
3282 case PRIV_VFS_CHROOT:
3283 case PRIV_VFS_RETAINSUGID:
3284 case PRIV_VFS_FCHROOT:
3286 case PRIV_VFS_SETGID:
3288 case PRIV_VFS_STICKYFILE:
3291 * As in the non-jail case, non-root users are expected to be
3292 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3293 * exists in the jail and they have permission to access it).
3295 case PRIV_KMEM_READ:
3299 * Depending on the global setting, allow privilege of
3300 * setting system flags.
3302 case PRIV_VFS_SYSFLAGS:
3303 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3309 * Depending on the global setting, allow privilege of
3310 * mounting/unmounting file systems.
3312 case PRIV_VFS_MOUNT:
3313 case PRIV_VFS_UNMOUNT:
3314 case PRIV_VFS_MOUNT_NONUSER:
3315 case PRIV_VFS_MOUNT_OWNER:
3316 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3317 cred->cr_prison->pr_enforce_statfs < 2)
3323 * Allow jailed root to bind reserved ports and reuse in-use
3326 case PRIV_NETINET_RESERVEDPORT:
3327 case PRIV_NETINET_REUSEPORT:
3331 * Allow jailed root to set certain IPv4/6 (option) headers.
3333 case PRIV_NETINET_SETHDROPTS:
3337 * Conditionally allow creating raw sockets in jail.
3339 case PRIV_NETINET_RAW:
3340 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3346 * Since jail implements its own visibility limits on netstat
3347 * sysctls, allow getcred. This allows identd to work in
3350 case PRIV_NETINET_GETCRED:
3354 * Allow jailed root to set loginclass.
3356 case PRIV_PROC_SETLOGINCLASS:
3360 * Do not allow a process inside a jail to read the kernel
3361 * message buffer unless explicitly permitted.
3364 if (cred->cr_prison->pr_allow & PR_ALLOW_READ_MSGBUF)
3370 * In all remaining cases, deny the privilege request. This
3371 * includes almost all network privileges, many system
3372 * configuration privileges.
3379 * Return the part of pr2's name that is relative to pr1, or the whole name
3380 * if it does not directly follow.
3384 prison_name(struct prison *pr1, struct prison *pr2)
3388 /* Jails see themselves as "0" (if they see themselves at all). */
3391 name = pr2->pr_name;
3392 if (prison_ischild(pr1, pr2)) {
3394 * pr1 isn't locked (and allprison_lock may not be either)
3395 * so its length can't be counted on. But the number of dots
3396 * can be counted on - and counted.
3398 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3399 name = strchr(name, '.') + 1;
3405 * Return the part of pr2's path that is relative to pr1, or the whole path
3406 * if it does not directly follow.
3409 prison_path(struct prison *pr1, struct prison *pr2)
3411 char *path1, *path2;
3414 path1 = pr1->pr_path;
3415 path2 = pr2->pr_path;
3416 if (!strcmp(path1, "/"))
3418 len1 = strlen(path1);
3419 if (strncmp(path1, path2, len1))
3421 if (path2[len1] == '\0')
3423 if (path2[len1] == '/')
3424 return (path2 + len1);
3430 * Jail-related sysctls.
3432 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3436 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3439 struct prison *pr, *cpr;
3441 struct in_addr *ip4 = NULL;
3445 struct in6_addr *ip6 = NULL;
3450 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3451 pr = req->td->td_ucred->cr_prison;
3453 sx_slock(&allprison_lock);
3454 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3455 #if defined(INET) || defined(INET6)
3458 mtx_lock(&cpr->pr_mtx);
3460 if (cpr->pr_ip4s > 0) {
3461 if (ip4s < cpr->pr_ip4s) {
3462 ip4s = cpr->pr_ip4s;
3463 mtx_unlock(&cpr->pr_mtx);
3464 ip4 = realloc(ip4, ip4s *
3465 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3468 bcopy(cpr->pr_ip4, ip4,
3469 cpr->pr_ip4s * sizeof(struct in_addr));
3473 if (cpr->pr_ip6s > 0) {
3474 if (ip6s < cpr->pr_ip6s) {
3475 ip6s = cpr->pr_ip6s;
3476 mtx_unlock(&cpr->pr_mtx);
3477 ip6 = realloc(ip6, ip6s *
3478 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3481 bcopy(cpr->pr_ip6, ip6,
3482 cpr->pr_ip6s * sizeof(struct in6_addr));
3485 if (cpr->pr_ref == 0) {
3486 mtx_unlock(&cpr->pr_mtx);
3489 bzero(xp, sizeof(*xp));
3490 xp->pr_version = XPRISON_VERSION;
3491 xp->pr_id = cpr->pr_id;
3492 xp->pr_state = cpr->pr_uref > 0
3493 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3494 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3495 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3496 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3498 xp->pr_ip4s = cpr->pr_ip4s;
3501 xp->pr_ip6s = cpr->pr_ip6s;
3503 mtx_unlock(&cpr->pr_mtx);
3504 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3508 if (xp->pr_ip4s > 0) {
3509 error = SYSCTL_OUT(req, ip4,
3510 xp->pr_ip4s * sizeof(struct in_addr));
3516 if (xp->pr_ip6s > 0) {
3517 error = SYSCTL_OUT(req, ip6,
3518 xp->pr_ip6s * sizeof(struct in6_addr));
3524 sx_sunlock(&allprison_lock);
3535 SYSCTL_OID(_security_jail, OID_AUTO, list,
3536 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3537 sysctl_jail_list, "S", "List of active jails");
3540 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3544 injail = jailed(req->td->td_ucred);
3545 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3550 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3551 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3552 sysctl_jail_jailed, "I", "Process in jail?");
3555 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
3557 int error, havevnet;
3559 struct ucred *cred = req->td->td_ucred;
3561 havevnet = jailed(cred) && prison_owns_vnet(cred);
3565 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
3570 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
3571 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3572 sysctl_jail_vnet, "I", "Jail owns VNET?");
3574 #if defined(INET) || defined(INET6)
3575 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3576 &jail_max_af_ips, 0,
3577 "Number of IP addresses a jail may have at most per address family (deprecated)");
3581 * Default parameters for jail(2) compatibility. For historical reasons,
3582 * the sysctl names have varying similarity to the parameter names. Prisons
3583 * just see their own parameters, and can't change them.
3586 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3589 int allow, error, i;
3591 pr = req->td->td_ucred->cr_prison;
3592 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3594 /* Get the current flag value, and convert it to a boolean. */
3595 i = (allow & arg2) ? 1 : 0;
3598 error = sysctl_handle_int(oidp, &i, 0, req);
3599 if (error || !req->newptr)
3605 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3608 mtx_lock(&prison0.pr_mtx);
3609 jail_default_allow = (jail_default_allow & ~arg2) | i;
3610 mtx_unlock(&prison0.pr_mtx);
3614 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3615 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3616 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3617 "Processes in jail can set their hostnames (deprecated)");
3618 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3619 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3620 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3621 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
3622 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3623 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3624 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3625 "Processes in jail can use System V IPC primitives (deprecated)");
3626 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3627 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3628 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3629 "Prison root can create raw sockets (deprecated)");
3630 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3631 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3632 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3633 "Processes in jail can alter system file flags (deprecated)");
3634 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3635 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3636 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3637 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
3638 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
3639 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3640 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
3641 "Processes in jail can mount the devfs file system (deprecated)");
3642 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
3643 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3644 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
3645 "Processes in jail can mount the fdescfs file system (deprecated)");
3646 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
3647 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3648 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
3649 "Processes in jail can mount the nullfs file system (deprecated)");
3650 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
3651 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3652 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
3653 "Processes in jail can mount the procfs file system (deprecated)");
3654 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
3655 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3656 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
3657 "Processes in jail can mount the linprocfs file system (deprecated)");
3658 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
3659 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3660 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
3661 "Processes in jail can mount the linsysfs file system (deprecated)");
3662 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
3663 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3664 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
3665 "Processes in jail can mount the tmpfs file system (deprecated)");
3666 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
3667 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3668 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
3669 "Processes in jail can mount the zfs file system (deprecated)");
3672 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3677 pr = req->td->td_ucred->cr_prison;
3678 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3679 error = sysctl_handle_int(oidp, &level, 0, req);
3680 if (error || !req->newptr)
3682 *(int *)arg1 = level;
3686 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3687 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3688 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3689 sysctl_jail_default_level, "I",
3690 "Processes in jail cannot see all mounted file systems (deprecated)");
3692 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
3693 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
3694 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
3695 sysctl_jail_default_level, "I",
3696 "Ruleset for the devfs filesystem in jail (deprecated)");
3699 * Nodes to describe jail parameters. Maximum length of string parameters
3700 * is returned in the string itself, and the other parameters exist merely
3701 * to make themselves and their types known.
3703 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
3707 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3714 switch (oidp->oid_kind & CTLTYPE)
3720 if (!(req->flags & SCTL_MASK32))
3722 return (SYSCTL_OUT(req, &l, sizeof(l)));
3726 return (SYSCTL_OUT(req, &i, sizeof(i)));
3727 case CTLTYPE_STRING:
3728 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
3730 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3731 case CTLTYPE_STRUCT:
3733 return (SYSCTL_OUT(req, &s, sizeof(s)));
3739 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
3740 * jail creation time but cannot be changed in an existing jail.
3742 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3743 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3744 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3745 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3746 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3747 "I", "Jail secure level");
3748 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
3749 "Jail value for kern.osreldate and uname -K");
3750 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
3751 "Jail value for kern.osrelease and uname -r");
3752 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3753 "I", "Jail cannot see all mounted file systems");
3754 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
3755 "I", "Ruleset for in-jail devfs mounts");
3756 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3757 "B", "Jail persistence");
3759 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3760 "E,jailsys", "Virtual network stack");
3762 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3763 "B", "Jail is in the process of shutting down");
3765 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
3766 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
3767 "I", "Current number of child jails");
3768 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
3769 "I", "Maximum number of child jails");
3771 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
3772 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3774 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3775 "Jail NIS domainname");
3776 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3778 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3779 "LU", "Jail host ID");
3781 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3782 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3785 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
3786 "Jail IPv4 address virtualization");
3787 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3788 "S,in_addr,a", "Jail IPv4 addresses");
3789 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3790 "B", "Do (not) use IPv4 source address selection rather than the "
3791 "primary jail IPv4 address.");
3794 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
3795 "Jail IPv6 address virtualization");
3796 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3797 "S,in6_addr,a", "Jail IPv6 addresses");
3798 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3799 "B", "Do (not) use IPv6 source address selection rather than the "
3800 "primary jail IPv6 address.");
3803 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3804 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3805 "B", "Jail may set hostname");
3806 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3807 "B", "Jail may use SYSV IPC");
3808 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3809 "B", "Jail may create raw sockets");
3810 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3811 "B", "Jail may alter system file flags");
3812 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3813 "B", "Jail may set file quotas");
3814 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3815 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3816 SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW,
3817 "B", "Jail may read the kernel message buffer");
3819 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
3820 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
3821 "B", "Jail may mount/unmount jail-friendly file systems in general");
3822 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
3823 "B", "Jail may mount the devfs file system");
3824 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
3825 "B", "Jail may mount the fdescfs file system");
3826 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
3827 "B", "Jail may mount the nullfs file system");
3828 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
3829 "B", "Jail may mount the procfs file system");
3830 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
3831 "B", "Jail may mount the linprocfs file system");
3832 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
3833 "B", "Jail may mount the linsysfs file system");
3834 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
3835 "B", "Jail may mount the tmpfs file system");
3836 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
3837 "B", "Jail may mount the zfs file system");
3841 prison_racct_foreach(void (*callback)(struct racct *racct,
3842 void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
3843 void *arg2, void *arg3)
3845 struct prison_racct *prr;
3847 ASSERT_RACCT_ENABLED();
3849 sx_slock(&allprison_lock);
3852 LIST_FOREACH(prr, &allprison_racct, prr_next)
3853 (callback)(prr->prr_racct, arg2, arg3);
3856 sx_sunlock(&allprison_lock);
3859 static struct prison_racct *
3860 prison_racct_find_locked(const char *name)
3862 struct prison_racct *prr;
3864 ASSERT_RACCT_ENABLED();
3865 sx_assert(&allprison_lock, SA_XLOCKED);
3867 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
3870 LIST_FOREACH(prr, &allprison_racct, prr_next) {
3871 if (strcmp(name, prr->prr_name) != 0)
3874 /* Found prison_racct with a matching name? */
3875 prison_racct_hold(prr);
3879 /* Add new prison_racct. */
3880 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
3881 racct_create(&prr->prr_racct);
3883 strcpy(prr->prr_name, name);
3884 refcount_init(&prr->prr_refcount, 1);
3885 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
3890 struct prison_racct *
3891 prison_racct_find(const char *name)
3893 struct prison_racct *prr;
3895 ASSERT_RACCT_ENABLED();
3897 sx_xlock(&allprison_lock);
3898 prr = prison_racct_find_locked(name);
3899 sx_xunlock(&allprison_lock);
3904 prison_racct_hold(struct prison_racct *prr)
3907 ASSERT_RACCT_ENABLED();
3909 refcount_acquire(&prr->prr_refcount);
3913 prison_racct_free_locked(struct prison_racct *prr)
3916 ASSERT_RACCT_ENABLED();
3917 sx_assert(&allprison_lock, SA_XLOCKED);
3919 if (refcount_release(&prr->prr_refcount)) {
3920 racct_destroy(&prr->prr_racct);
3921 LIST_REMOVE(prr, prr_next);
3922 free(prr, M_PRISON_RACCT);
3927 prison_racct_free(struct prison_racct *prr)
3931 ASSERT_RACCT_ENABLED();
3932 sx_assert(&allprison_lock, SA_UNLOCKED);
3934 old = prr->prr_refcount;
3935 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
3938 sx_xlock(&allprison_lock);
3939 prison_racct_free_locked(prr);
3940 sx_xunlock(&allprison_lock);
3944 prison_racct_attach(struct prison *pr)
3946 struct prison_racct *prr;
3948 ASSERT_RACCT_ENABLED();
3949 sx_assert(&allprison_lock, SA_XLOCKED);
3951 prr = prison_racct_find_locked(pr->pr_name);
3952 KASSERT(prr != NULL, ("cannot find prison_racct"));
3954 pr->pr_prison_racct = prr;
3958 * Handle jail renaming. From the racct point of view, renaming means
3959 * moving from one prison_racct to another.
3962 prison_racct_modify(struct prison *pr)
3968 struct prison_racct *oldprr;
3970 ASSERT_RACCT_ENABLED();
3972 sx_slock(&allproc_lock);
3973 sx_xlock(&allprison_lock);
3975 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
3976 sx_xunlock(&allprison_lock);
3977 sx_sunlock(&allproc_lock);
3981 oldprr = pr->pr_prison_racct;
3982 pr->pr_prison_racct = NULL;
3984 prison_racct_attach(pr);
3987 * Move resource utilisation records.
3989 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
3993 * Force rctl to reattach rules to processes.
3995 FOREACH_PROC_IN_SYSTEM(p) {
3997 cred = crhold(p->p_ucred);
3999 rctl_proc_ucred_changed(p, cred);
4004 sx_sunlock(&allproc_lock);
4005 prison_racct_free_locked(oldprr);
4006 sx_xunlock(&allprison_lock);
4010 prison_racct_detach(struct prison *pr)
4013 ASSERT_RACCT_ENABLED();
4014 sx_assert(&allprison_lock, SA_UNLOCKED);
4016 if (pr->pr_prison_racct == NULL)
4018 prison_racct_free(pr->pr_prison_racct);
4019 pr->pr_prison_racct = NULL;
4026 db_show_prison(struct prison *pr)
4029 #if defined(INET) || defined(INET6)
4034 char ip4buf[INET_ADDRSTRLEN];
4037 char ip6buf[INET6_ADDRSTRLEN];
4040 db_printf("prison %p:\n", pr);
4041 db_printf(" jid = %d\n", pr->pr_id);
4042 db_printf(" name = %s\n", pr->pr_name);
4043 db_printf(" parent = %p\n", pr->pr_parent);
4044 db_printf(" ref = %d\n", pr->pr_ref);
4045 db_printf(" uref = %d\n", pr->pr_uref);
4046 db_printf(" path = %s\n", pr->pr_path);
4047 db_printf(" cpuset = %d\n", pr->pr_cpuset
4048 ? pr->pr_cpuset->cs_id : -1);
4050 db_printf(" vnet = %p\n", pr->pr_vnet);
4052 db_printf(" root = %p\n", pr->pr_root);
4053 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4054 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4055 db_printf(" children.max = %d\n", pr->pr_childmax);
4056 db_printf(" children.cur = %d\n", pr->pr_childcount);
4057 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4058 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4059 db_printf(" flags = 0x%x", pr->pr_flags);
4060 for (fi = 0; fi < nitems(pr_flag_names); fi++)
4061 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4062 db_printf(" %s", pr_flag_names[fi]);
4063 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
4064 jsf = pr->pr_flags &
4065 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4066 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4067 pr_flag_jailsys[fi].disable &&
4068 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4069 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4072 db_printf(" allow = 0x%x", pr->pr_allow);
4073 for (fi = 0; fi < nitems(pr_allow_names); fi++)
4074 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4075 db_printf(" %s", pr_allow_names[fi]);
4077 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4078 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4079 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4080 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4081 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4083 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4084 for (ii = 0; ii < pr->pr_ip4s; ii++)
4085 db_printf(" %s %s\n",
4086 ii == 0 ? "ip4.addr =" : " ",
4087 inet_ntoa_r(pr->pr_ip4[ii], ip4buf));
4090 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4091 for (ii = 0; ii < pr->pr_ip6s; ii++)
4092 db_printf(" %s %s\n",
4093 ii == 0 ? "ip6.addr =" : " ",
4094 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4098 DB_SHOW_COMMAND(prison, db_show_prison_command)
4104 * Show all prisons in the list, and prison0 which is not
4107 db_show_prison(&prison0);
4108 if (!db_pager_quit) {
4109 TAILQ_FOREACH(pr, &allprison, pr_list) {
4121 /* Look for a prison with the ID and with references. */
4122 TAILQ_FOREACH(pr, &allprison, pr_list)
4123 if (pr->pr_id == addr && pr->pr_ref > 0)
4126 /* Look again, without requiring a reference. */
4127 TAILQ_FOREACH(pr, &allprison, pr_list)
4128 if (pr->pr_id == addr)
4131 /* Assume address points to a valid prison. */
4132 pr = (struct prison *)addr;