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 static 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_modify(struct prison *pr);
134 static void prison_racct_detach(struct prison *pr);
137 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
138 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
141 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
142 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
145 /* Flags for prison_deref */
146 #define PD_DEREF 0x01
147 #define PD_DEUREF 0x02
148 #define PD_LOCKED 0x04
149 #define PD_LIST_SLOCKED 0x08
150 #define PD_LIST_XLOCKED 0x10
153 * Parameter names corresponding to PR_* flag values. Size values are for kvm
154 * as we cannot figure out the size of a sparse array, or an array without a
157 static char *pr_flag_names[] = {
160 [7] = "ip4.saddrsel",
163 [8] = "ip6.saddrsel",
166 const size_t pr_flag_names_size = sizeof(pr_flag_names);
168 static char *pr_flag_nonames[] = {
171 [7] = "ip4.nosaddrsel",
174 [8] = "ip6.nosaddrsel",
177 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
179 struct jailsys_flags {
183 } pr_flag_jailsys[] = {
184 { "host", 0, PR_HOST },
186 { "vnet", 0, PR_VNET },
189 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
192 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
195 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
197 static char *pr_allow_names[] = {
198 "allow.set_hostname",
206 "allow.mount.nullfs",
208 "allow.mount.procfs",
210 "allow.mount.fdescfs",
211 "allow.mount.linprocfs",
212 "allow.mount.linsysfs",
214 const size_t pr_allow_names_size = sizeof(pr_allow_names);
216 static char *pr_allow_nonames[] = {
217 "allow.noset_hostname",
219 "allow.noraw_sockets",
224 "allow.mount.nodevfs",
225 "allow.mount.nonullfs",
227 "allow.mount.noprocfs",
228 "allow.mount.notmpfs",
229 "allow.mount.nofdescfs",
230 "allow.mount.nolinprocfs",
231 "allow.mount.nolinsysfs",
233 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
235 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
236 #define JAIL_DEFAULT_ENFORCE_STATFS 2
237 #define JAIL_DEFAULT_DEVFS_RSNUM 0
238 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
239 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
240 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
241 #if defined(INET) || defined(INET6)
242 static unsigned jail_max_af_ips = 255;
246 * Initialize the parts of prison0 that can't be static-initialized with
247 * constants. This is called from proc0_init() after creating thread0 cpuset.
253 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
254 prison0.pr_osreldate = osreldate;
255 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
260 qcmp_v4(const void *ip1, const void *ip2)
265 * We need to compare in HBO here to get the list sorted as expected
266 * by the result of the code. Sorting NBO addresses gives you
267 * interesting results. If you do not understand, do not try.
269 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
270 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
273 * Do not simply return the difference of the two numbers, the int is
287 qcmp_v6(const void *ip1, const void *ip2)
289 const struct in6_addr *ia6a, *ia6b;
292 ia6a = (const struct in6_addr *)ip1;
293 ia6b = (const struct in6_addr *)ip2;
296 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
297 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
299 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
312 sys_jail(struct thread *td, struct jail_args *uap)
318 error = copyin(uap->jail, &version, sizeof(uint32_t));
327 /* FreeBSD single IPv4 jails. */
328 bzero(&j, sizeof(struct jail));
329 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
332 j.version = j0.version;
334 j.hostname = j0.hostname;
335 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
341 * Version 1 was used by multi-IPv4 jail implementations
342 * that never made it into the official kernel.
346 case 2: /* JAIL_API_VERSION */
347 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
348 error = copyin(uap->jail, &j, sizeof(struct jail));
354 /* Sci-Fi jails are not supported, sorry. */
357 return (kern_jail(td, &j));
361 kern_jail(struct thread *td, struct jail *j)
363 struct iovec optiov[2 * (4
364 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
373 char *u_path, *u_hostname, *u_name;
376 struct in_addr *u_ip4;
379 struct in6_addr *u_ip6;
382 int error, enforce_statfs, fi;
384 bzero(&optiov, sizeof(optiov));
385 opt.uio_iov = optiov;
389 opt.uio_segflg = UIO_SYSSPACE;
390 opt.uio_rw = UIO_READ;
393 /* Set permissions for top-level jails from sysctls. */
394 if (!jailed(td->td_ucred)) {
395 for (fi = 0; fi < sizeof(pr_allow_names) /
396 sizeof(pr_allow_names[0]); fi++) {
397 optiov[opt.uio_iovcnt].iov_base =
398 (jail_default_allow & (1 << fi))
399 ? pr_allow_names[fi] : pr_allow_nonames[fi];
400 optiov[opt.uio_iovcnt].iov_len =
401 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
404 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
405 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
407 enforce_statfs = jail_default_enforce_statfs;
408 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
409 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
413 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
415 ip4s = (j->version == 0) ? 1 : j->ip4s;
416 if (ip4s > jail_max_af_ips)
418 tmplen += ip4s * sizeof(struct in_addr);
424 if (j->ip6s > jail_max_af_ips)
426 tmplen += j->ip6s * sizeof(struct in6_addr);
431 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
432 u_hostname = u_path + MAXPATHLEN;
433 u_name = u_hostname + MAXHOSTNAMELEN;
435 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
439 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
441 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
444 optiov[opt.uio_iovcnt].iov_base = "path";
445 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
447 optiov[opt.uio_iovcnt].iov_base = u_path;
448 error = copyinstr(j->path, u_path, MAXPATHLEN,
449 &optiov[opt.uio_iovcnt].iov_len);
451 free(u_path, M_TEMP);
455 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
456 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
458 optiov[opt.uio_iovcnt].iov_base = u_hostname;
459 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
460 &optiov[opt.uio_iovcnt].iov_len);
462 free(u_path, M_TEMP);
466 if (j->jailname != NULL) {
467 optiov[opt.uio_iovcnt].iov_base = "name";
468 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
470 optiov[opt.uio_iovcnt].iov_base = u_name;
471 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
472 &optiov[opt.uio_iovcnt].iov_len);
474 free(u_path, M_TEMP);
480 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
481 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
483 optiov[opt.uio_iovcnt].iov_base = u_ip4;
484 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
486 u_ip4->s_addr = j->ip4s;
488 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
490 free(u_path, M_TEMP);
497 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
498 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
500 optiov[opt.uio_iovcnt].iov_base = u_ip6;
501 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
502 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
504 free(u_path, M_TEMP);
509 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
510 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
511 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
512 free(u_path, M_TEMP);
518 * struct jail_set_args {
519 * struct iovec *iovp;
520 * unsigned int iovcnt;
525 sys_jail_set(struct thread *td, struct jail_set_args *uap)
530 /* Check that we have an even number of iovecs. */
534 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
537 error = kern_jail_set(td, auio, uap->flags);
543 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
550 struct in6_addr *ip6;
553 struct vfsoptlist *opts;
554 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
556 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
557 char *g_path, *osrelstr;
558 #if defined(INET) || defined(INET6)
563 size_t namelen, onamelen;
564 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
565 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
566 int fi, jid, jsys, len, level;
567 int childmax, osreldt, rsnum, slevel;
568 int fullpath_disabled;
569 #if defined(INET) || defined(INET6)
578 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
582 error = priv_check(td, PRIV_JAIL_SET);
583 if (!error && (flags & JAIL_ATTACH))
584 error = priv_check(td, PRIV_JAIL_ATTACH);
587 mypr = ppr = td->td_ucred->cr_prison;
588 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
590 if (flags & ~JAIL_SET_MASK)
594 * Check all the parameters before committing to anything. Not all
595 * errors can be caught early, but we may as well try. Also, this
596 * takes care of some expensive stuff (path lookup) before getting
597 * the allprison lock.
599 * XXX Jails are not filesystems, and jail parameters are not mount
600 * options. But it makes more sense to re-use the vfsopt code
601 * than duplicate it under a different name.
603 error = vfs_buildopts(optuio, &opts);
614 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
620 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
629 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
637 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
642 else if (enforce < 0 || enforce > 2) {
648 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
656 pr_flags = ch_flags = 0;
657 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
659 if (pr_flag_names[fi] == NULL)
661 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
662 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
664 ch_flags |= pr_flags;
665 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
667 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
674 case JAIL_SYS_DISABLE:
675 if (!pr_flag_jailsys[fi].disable) {
679 pr_flags |= pr_flag_jailsys[fi].disable;
682 pr_flags |= pr_flag_jailsys[fi].new;
684 case JAIL_SYS_INHERIT:
691 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
693 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
694 && !(pr_flags & PR_PERSIST)) {
696 vfs_opterror(opts, "new jail must persist or attach");
700 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
702 vfs_opterror(opts, "vnet cannot be changed after creation");
707 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
709 vfs_opterror(opts, "ip4 cannot be changed after creation");
714 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
716 vfs_opterror(opts, "ip6 cannot be changed after creation");
721 pr_allow = ch_allow = 0;
722 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
724 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
725 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
727 ch_allow |= pr_allow;
729 error = vfs_getopt(opts, "name", (void **)&name, &len);
735 if (len == 0 || name[len - 1] != '\0') {
739 if (len > MAXHOSTNAMELEN) {
740 error = ENAMETOOLONG;
745 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
753 if (len == 0 || host[len - 1] != '\0') {
757 if (len > MAXHOSTNAMELEN) {
758 error = ENAMETOOLONG;
763 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
771 if (len == 0 || domain[len - 1] != '\0') {
775 if (len > MAXHOSTNAMELEN) {
776 error = ENAMETOOLONG;
781 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
789 if (len == 0 || uuid[len - 1] != '\0') {
793 if (len > HOSTUUIDLEN) {
794 error = ENAMETOOLONG;
799 #ifdef COMPAT_FREEBSD32
800 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
803 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
807 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
819 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
824 else if (ip4s & (sizeof(*ip4) - 1)) {
828 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
830 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
832 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
833 ip4s /= sizeof(*ip4);
834 if (ip4s > jail_max_af_ips) {
836 vfs_opterror(opts, "too many IPv4 addresses");
839 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
840 bcopy(op, ip4, ip4s * sizeof(*ip4));
842 * IP addresses are all sorted but ip[0] to preserve
843 * the primary IP address as given from userland.
844 * This special IP is used for unbound outgoing
845 * connections as well for "loopback" traffic in case
846 * source address selection cannot find any more fitting
847 * address to connect from.
850 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
852 * Check for duplicate addresses and do some simple
853 * zero and broadcast checks. If users give other bogus
854 * addresses it is their problem.
856 * We do not have to care about byte order for these
857 * checks so we will do them in NBO.
859 for (ii = 0; ii < ip4s; ii++) {
860 if (ip4[ii].s_addr == INADDR_ANY ||
861 ip4[ii].s_addr == INADDR_BROADCAST) {
866 (ip4[0].s_addr == ip4[ii+1].s_addr ||
867 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
877 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
882 else if (ip6s & (sizeof(*ip6) - 1)) {
886 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
888 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
890 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
891 ip6s /= sizeof(*ip6);
892 if (ip6s > jail_max_af_ips) {
894 vfs_opterror(opts, "too many IPv6 addresses");
897 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
898 bcopy(op, ip6, ip6s * sizeof(*ip6));
900 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
901 for (ii = 0; ii < ip6s; ii++) {
902 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
907 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
908 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
918 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
919 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
922 "vnet jails cannot have IP address restrictions");
927 fullpath_disabled = 0;
929 error = vfs_getopt(opts, "path", (void **)&path, &len);
935 if (flags & JAIL_UPDATE) {
938 "path cannot be changed after creation");
941 if (len == 0 || path[len - 1] != '\0') {
945 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
951 NDFREE(&nd, NDF_ONLY_PNBUF);
952 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
953 strlcpy(g_path, path, MAXPATHLEN);
954 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
957 else if (error == ENODEV) {
958 /* proceed if sysctl debug.disablefullpath == 1 */
959 fullpath_disabled = 1;
960 if (len < 2 || (len == 2 && path[0] == '/'))
963 /* exit on other errors */
966 if (root->v_type != VDIR) {
972 if (fullpath_disabled) {
973 /* Leave room for a real-root full pathname. */
974 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
975 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
976 error = ENAMETOOLONG;
982 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
988 if (flags & JAIL_UPDATE) {
991 "osrelease cannot be changed after creation");
994 if (len == 0 || len >= OSRELEASELEN) {
997 "osrelease string must be 1-%d bytes long",
1003 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
1004 if (error == ENOENT)
1006 else if (error != 0)
1009 if (flags & JAIL_UPDATE) {
1012 "osreldate cannot be changed after creation");
1017 vfs_opterror(opts, "osreldate cannot be 0");
1023 * Grab the allprison lock before letting modules check their
1024 * parameters. Once we have it, do not let go so we'll have a
1025 * consistent view of the OSD list.
1027 sx_xlock(&allprison_lock);
1028 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
1030 goto done_unlock_list;
1032 /* By now, all parameters should have been noted. */
1033 TAILQ_FOREACH(opt, opts, link) {
1034 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1036 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1037 goto done_unlock_list;
1042 * See if we are creating a new record or updating an existing one.
1043 * This abuses the file error codes ENOENT and EEXIST.
1045 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
1048 vfs_opterror(opts, "no valid operation (create or update)");
1049 goto done_unlock_list;
1053 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1054 namelc = strrchr(name, '.');
1055 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1061 * See if a requested jid already exists. There is an
1062 * information leak here if the jid exists but is not within
1063 * the caller's jail hierarchy. Jail creators will get EEXIST
1064 * even though they cannot see the jail, and CREATE | UPDATE
1065 * will return ENOENT which is not normally a valid error.
1069 vfs_opterror(opts, "negative jid");
1070 goto done_unlock_list;
1072 pr = prison_find(jid);
1074 ppr = pr->pr_parent;
1075 /* Create: jid must not exist. */
1076 if (cuflags == JAIL_CREATE) {
1077 mtx_unlock(&pr->pr_mtx);
1079 vfs_opterror(opts, "jail %d already exists",
1081 goto done_unlock_list;
1083 if (!prison_ischild(mypr, pr)) {
1084 mtx_unlock(&pr->pr_mtx);
1086 } else if (pr->pr_uref == 0) {
1087 if (!(flags & JAIL_DYING)) {
1088 mtx_unlock(&pr->pr_mtx);
1090 vfs_opterror(opts, "jail %d is dying",
1092 goto done_unlock_list;
1093 } else if ((flags & JAIL_ATTACH) ||
1094 (pr_flags & PR_PERSIST)) {
1096 * A dying jail might be resurrected
1097 * (via attach or persist), but first
1098 * it must determine if another jail
1099 * has claimed its name. Accomplish
1100 * this by implicitly re-setting the
1104 name = prison_name(mypr, pr);
1109 /* Update: jid must exist. */
1110 if (cuflags == JAIL_UPDATE) {
1112 vfs_opterror(opts, "jail %d not found", jid);
1113 goto done_unlock_list;
1118 * If the caller provided a name, look for a jail by that name.
1119 * This has different semantics for creates and updates keyed by jid
1120 * (where the name must not already exist in a different jail),
1121 * and updates keyed by the name itself (where the name must exist
1122 * because that is the jail being updated).
1125 namelc = strrchr(name, '.');
1130 * This is a hierarchical name. Split it into the
1131 * parent and child names, and make sure the parent
1132 * exists or matches an already found jail.
1136 if (strncmp(name, ppr->pr_name, namelc - name)
1137 || ppr->pr_name[namelc - name] != '\0') {
1138 mtx_unlock(&pr->pr_mtx);
1141 "cannot change jail's parent");
1142 goto done_unlock_list;
1145 ppr = prison_find_name(mypr, name);
1149 "jail \"%s\" not found", name);
1150 goto done_unlock_list;
1152 mtx_unlock(&ppr->pr_mtx);
1156 if (name[0] != '\0') {
1158 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1161 FOREACH_PRISON_CHILD(ppr, tpr) {
1162 if (tpr != pr && tpr->pr_ref > 0 &&
1163 !strcmp(tpr->pr_name + namelen, name)) {
1165 cuflags != JAIL_CREATE) {
1166 mtx_lock(&tpr->pr_mtx);
1167 if (tpr->pr_ref > 0) {
1172 if (tpr->pr_uref > 0) {
1178 mtx_unlock(&tpr->pr_mtx);
1179 } else if (tpr->pr_uref > 0) {
1181 * Create, or update(jid):
1182 * name must not exist in an
1183 * active sibling jail.
1187 mtx_unlock(&pr->pr_mtx);
1189 "jail \"%s\" already exists",
1191 goto done_unlock_list;
1195 /* If no active jail is found, use a dying one. */
1196 if (deadpr != NULL && pr == NULL) {
1197 if (flags & JAIL_DYING) {
1198 mtx_lock(&deadpr->pr_mtx);
1199 if (deadpr->pr_ref == 0) {
1200 mtx_unlock(&deadpr->pr_mtx);
1204 } else if (cuflags == JAIL_UPDATE) {
1207 "jail \"%s\" is dying", name);
1208 goto done_unlock_list;
1211 /* Update: name must exist if no jid. */
1212 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1214 vfs_opterror(opts, "jail \"%s\" not found",
1216 goto done_unlock_list;
1220 /* Update: must provide a jid or name. */
1221 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1223 vfs_opterror(opts, "update specified no jail");
1224 goto done_unlock_list;
1227 /* If there's no prison to update, create a new one and link it in. */
1229 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1230 if (tpr->pr_childcount >= tpr->pr_childmax) {
1232 vfs_opterror(opts, "prison limit exceeded");
1233 goto done_unlock_list;
1236 mtx_lock(&ppr->pr_mtx);
1237 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1238 mtx_unlock(&ppr->pr_mtx);
1240 vfs_opterror(opts, "parent jail went away!");
1241 goto done_unlock_list;
1245 mtx_unlock(&ppr->pr_mtx);
1246 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1248 /* Find the next free jid. */
1251 if (jid == JAIL_MAX)
1253 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1254 if (tpr->pr_id < jid)
1256 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1257 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1260 if (jid == lastprid) {
1263 "no available jail IDs");
1265 prison_deref(ppr, PD_DEREF |
1266 PD_DEUREF | PD_LIST_XLOCKED);
1275 * The jail already has a jid (that did not yet exist),
1276 * so just find where to insert it.
1278 TAILQ_FOREACH(tpr, &allprison, pr_list)
1279 if (tpr->pr_id >= jid) {
1280 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1285 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1286 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1287 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1288 tpr->pr_childcount++;
1290 pr->pr_parent = ppr;
1293 /* Set some default values, and inherit some from the parent. */
1298 root = mypr->pr_root;
1301 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1302 pr->pr_flags |= PR_HOST;
1303 #if defined(INET) || defined(INET6)
1305 if (!(pr_flags & PR_VNET))
1309 if (!(ch_flags & PR_IP4_USER))
1311 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1312 else if (!(pr_flags & PR_IP4_USER)) {
1313 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1314 if (ppr->pr_ip4 != NULL) {
1315 pr->pr_ip4s = ppr->pr_ip4s;
1316 pr->pr_ip4 = malloc(pr->pr_ip4s *
1317 sizeof(struct in_addr), M_PRISON,
1319 bcopy(ppr->pr_ip4, pr->pr_ip4,
1320 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1325 if (!(ch_flags & PR_IP6_USER))
1327 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1328 else if (!(pr_flags & PR_IP6_USER)) {
1329 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1330 if (ppr->pr_ip6 != NULL) {
1331 pr->pr_ip6s = ppr->pr_ip6s;
1332 pr->pr_ip6 = malloc(pr->pr_ip6s *
1333 sizeof(struct in6_addr), M_PRISON,
1335 bcopy(ppr->pr_ip6, pr->pr_ip6,
1336 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1342 /* Source address selection is always on by default. */
1343 pr->pr_flags |= _PR_IP_SADDRSEL;
1345 pr->pr_securelevel = ppr->pr_securelevel;
1346 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1347 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1348 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1350 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1351 if (osrelstr == NULL)
1352 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1354 strcpy(pr->pr_osrelease, osrelstr);
1356 LIST_INIT(&pr->pr_children);
1357 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1360 /* Allocate a new vnet if specified. */
1361 pr->pr_vnet = (pr_flags & PR_VNET)
1362 ? vnet_alloc() : ppr->pr_vnet;
1365 * Allocate a dedicated cpuset for each jail.
1366 * Unlike other initial settings, this may return an erorr.
1368 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1370 prison_deref(pr, PD_LIST_XLOCKED);
1374 mtx_lock(&pr->pr_mtx);
1376 * New prisons do not yet have a reference, because we do not
1377 * want other to see the incomplete prison once the
1378 * allprison_lock is downgraded.
1383 * Grab a reference for existing prisons, to ensure they
1384 * continue to exist for the duration of the call.
1387 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1388 if ((pr->pr_flags & PR_VNET) &&
1389 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1392 "vnet jails cannot have IP address restrictions");
1393 goto done_deref_locked;
1397 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1400 "ip4 cannot be changed after creation");
1401 goto done_deref_locked;
1405 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1408 "ip6 cannot be changed after creation");
1409 goto done_deref_locked;
1414 /* Do final error checking before setting anything. */
1416 if (slevel < ppr->pr_securelevel) {
1418 goto done_deref_locked;
1422 if (childmax >= ppr->pr_childmax) {
1424 goto done_deref_locked;
1428 if (enforce < ppr->pr_enforce_statfs) {
1430 goto done_deref_locked;
1435 * devfs_rsnum is a uint16_t
1437 if (rsnum < 0 || rsnum > 65535) {
1439 goto done_deref_locked;
1442 * Nested jails always inherit parent's devfs ruleset
1444 if (jailed(td->td_ucred)) {
1445 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1447 goto done_deref_locked;
1449 rsnum = ppr->pr_devfs_rsnum;
1454 if (ppr->pr_flags & PR_IP4) {
1456 * Make sure the new set of IP addresses is a
1457 * subset of the parent's list. Don't worry
1458 * about the parent being unlocked, as any
1459 * setting is done with allprison_lock held.
1461 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1462 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1464 if (ij == ppr->pr_ip4s) {
1466 goto done_deref_locked;
1469 for (ii = ij = 1; ii < ip4s; ii++) {
1470 if (ip4[ii].s_addr ==
1471 ppr->pr_ip4[0].s_addr)
1473 for (; ij < ppr->pr_ip4s; ij++)
1474 if (ip4[ii].s_addr ==
1475 ppr->pr_ip4[ij].s_addr)
1477 if (ij == ppr->pr_ip4s)
1480 if (ij == ppr->pr_ip4s) {
1482 goto done_deref_locked;
1487 * Check for conflicting IP addresses. We permit them
1488 * if there is no more than one IP on each jail. If
1489 * there is a duplicate on a jail with more than one
1490 * IP stop checking and return error.
1494 for (; tppr != &prison0; tppr = tppr->pr_parent)
1495 if (tppr->pr_flags & PR_VNET)
1498 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1501 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1503 tpr->pr_uref == 0) {
1507 if (!(tpr->pr_flags & PR_IP4_USER))
1510 if (tpr->pr_ip4 == NULL ||
1511 (ip4s == 1 && tpr->pr_ip4s == 1))
1513 for (ii = 0; ii < ip4s; ii++) {
1514 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1517 "IPv4 addresses clash");
1518 goto done_deref_locked;
1526 if (ppr->pr_flags & PR_IP6) {
1528 * Make sure the new set of IP addresses is a
1529 * subset of the parent's list.
1531 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1532 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1535 if (ij == ppr->pr_ip6s) {
1537 goto done_deref_locked;
1540 for (ii = ij = 1; ii < ip6s; ii++) {
1541 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1544 for (; ij < ppr->pr_ip6s; ij++)
1545 if (IN6_ARE_ADDR_EQUAL(
1546 &ip6[ii], &ppr->pr_ip6[ij]))
1548 if (ij == ppr->pr_ip6s)
1551 if (ij == ppr->pr_ip6s) {
1553 goto done_deref_locked;
1557 /* Check for conflicting IP addresses. */
1560 for (; tppr != &prison0; tppr = tppr->pr_parent)
1561 if (tppr->pr_flags & PR_VNET)
1564 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1567 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1569 tpr->pr_uref == 0) {
1573 if (!(tpr->pr_flags & PR_IP6_USER))
1576 if (tpr->pr_ip6 == NULL ||
1577 (ip6s == 1 && tpr->pr_ip6s == 1))
1579 for (ii = 0; ii < ip6s; ii++) {
1580 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1583 "IPv6 addresses clash");
1584 goto done_deref_locked;
1590 onamelen = namelen = 0;
1592 /* Give a default name of the jid. Also allow the name to be
1593 * explicitly the jid - but not any other number, and only in
1594 * normal form (no leading zero/etc).
1596 if (name[0] == '\0')
1597 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1598 else if ((strtoul(namelc, &p, 10) != jid ||
1599 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1602 "name cannot be numeric (unless it is the jid)");
1603 goto done_deref_locked;
1606 * Make sure the name isn't too long for the prison or its
1609 onamelen = strlen(pr->pr_name);
1610 namelen = strlen(name);
1611 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1612 error = ENAMETOOLONG;
1613 goto done_deref_locked;
1615 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1616 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1617 sizeof(pr->pr_name)) {
1618 error = ENAMETOOLONG;
1619 goto done_deref_locked;
1623 if (pr_allow & ~ppr->pr_allow) {
1625 goto done_deref_locked;
1628 /* Set the parameters of the prison. */
1631 if (pr_flags & PR_IP4_USER) {
1632 pr->pr_flags |= PR_IP4;
1633 free(pr->pr_ip4, M_PRISON);
1637 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1639 if (tpr->pr_flags & PR_VNET) {
1644 if (prison_restrict_ip4(tpr, NULL)) {
1653 if (pr_flags & PR_IP6_USER) {
1654 pr->pr_flags |= PR_IP6;
1655 free(pr->pr_ip6, M_PRISON);
1659 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1661 if (tpr->pr_flags & PR_VNET) {
1666 if (prison_restrict_ip6(tpr, NULL)) {
1674 pr->pr_securelevel = slevel;
1675 /* Set all child jails to be at least this level. */
1676 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1677 if (tpr->pr_securelevel < slevel)
1678 tpr->pr_securelevel = slevel;
1681 pr->pr_childmax = childmax;
1682 /* Set all child jails to under this limit. */
1683 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1684 if (tpr->pr_childmax > childmax - level)
1685 tpr->pr_childmax = childmax > level
1686 ? childmax - level : 0;
1689 pr->pr_enforce_statfs = enforce;
1690 /* Pass this restriction on to the children. */
1691 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1692 if (tpr->pr_enforce_statfs < enforce)
1693 tpr->pr_enforce_statfs = enforce;
1696 pr->pr_devfs_rsnum = rsnum;
1697 /* Pass this restriction on to the children. */
1698 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1699 tpr->pr_devfs_rsnum = rsnum;
1702 if (ppr == &prison0)
1703 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1705 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1706 ppr->pr_name, name);
1707 /* Change this component of child names. */
1708 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1709 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1710 strlen(tpr->pr_name + onamelen) + 1);
1711 bcopy(pr->pr_name, tpr->pr_name, namelen);
1715 /* Try to keep a real-rooted full pathname. */
1716 if (fullpath_disabled && path[0] == '/' &&
1717 strcmp(mypr->pr_path, "/"))
1718 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1719 mypr->pr_path, path);
1721 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1724 if (PR_HOST & ch_flags & ~pr_flags) {
1725 if (pr->pr_flags & PR_HOST) {
1727 * Copy the parent's host info. As with pr_ip4 above,
1728 * the lack of a lock on the parent is not a problem;
1729 * it is always set with allprison_lock at least
1730 * shared, and is held exclusively here.
1732 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1733 sizeof(pr->pr_hostname));
1734 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1735 sizeof(pr->pr_domainname));
1736 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1737 sizeof(pr->pr_hostuuid));
1738 pr->pr_hostid = pr->pr_parent->pr_hostid;
1740 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1741 /* Set this prison, and any descendants without PR_HOST. */
1743 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1745 strlcpy(pr->pr_domainname, domain,
1746 sizeof(pr->pr_domainname));
1748 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1750 pr->pr_hostid = hid;
1751 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1752 if (tpr->pr_flags & PR_HOST)
1756 strlcpy(tpr->pr_hostname,
1758 sizeof(tpr->pr_hostname));
1760 strlcpy(tpr->pr_domainname,
1762 sizeof(tpr->pr_domainname));
1764 strlcpy(tpr->pr_hostuuid,
1766 sizeof(tpr->pr_hostuuid));
1768 tpr->pr_hostid = hid;
1772 if ((tallow = ch_allow & ~pr_allow)) {
1773 /* Clear allow bits in all children. */
1774 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1775 tpr->pr_allow &= ~tallow;
1777 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1779 * Persistent prisons get an extra reference, and prisons losing their
1780 * persist flag lose that reference. Only do this for existing prisons
1781 * for now, so new ones will remain unseen until after the module
1782 * handlers have completed.
1784 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1785 if (pr_flags & PR_PERSIST) {
1793 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1794 mtx_unlock(&pr->pr_mtx);
1797 if (racct_enable && created)
1798 prison_racct_attach(pr);
1801 /* Locks may have prevented a complete restriction of child IP
1802 * addresses. If so, allocate some more memory and try again.
1807 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1808 mtx_lock(&pr->pr_mtx);
1810 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1812 if (tpr->pr_flags & PR_VNET) {
1817 if (prison_restrict_ip4(tpr, ip4)) {
1824 mtx_unlock(&pr->pr_mtx);
1830 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1831 mtx_lock(&pr->pr_mtx);
1833 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1835 if (tpr->pr_flags & PR_VNET) {
1840 if (prison_restrict_ip6(tpr, ip6)) {
1847 mtx_unlock(&pr->pr_mtx);
1851 /* Let the modules do their work. */
1852 sx_downgrade(&allprison_lock);
1854 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1856 prison_deref(pr, PD_LIST_SLOCKED);
1860 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1862 prison_deref(pr, created
1864 : PD_DEREF | PD_LIST_SLOCKED);
1868 /* Attach this process to the prison if requested. */
1869 if (flags & JAIL_ATTACH) {
1870 mtx_lock(&pr->pr_mtx);
1871 error = do_jail_attach(td, pr);
1873 vfs_opterror(opts, "attach failed");
1875 prison_deref(pr, PD_DEREF);
1881 if (racct_enable && !created) {
1882 if (!(flags & JAIL_ATTACH))
1883 sx_sunlock(&allprison_lock);
1884 prison_racct_modify(pr);
1885 if (!(flags & JAIL_ATTACH))
1886 sx_slock(&allprison_lock);
1890 td->td_retval[0] = pr->pr_id;
1893 * Now that it is all there, drop the temporary reference from existing
1894 * prisons. Or add a reference to newly created persistent prisons
1895 * (which was not done earlier so that the prison would not be publicly
1899 prison_deref(pr, (flags & JAIL_ATTACH)
1901 : PD_DEREF | PD_LIST_SLOCKED);
1903 if (pr_flags & PR_PERSIST) {
1904 mtx_lock(&pr->pr_mtx);
1907 mtx_unlock(&pr->pr_mtx);
1909 if (!(flags & JAIL_ATTACH))
1910 sx_sunlock(&allprison_lock);
1916 prison_deref(pr, created
1917 ? PD_LOCKED | PD_LIST_XLOCKED
1918 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1921 sx_xunlock(&allprison_lock);
1927 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1928 if (errmsg_len > 0) {
1929 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1930 if (errmsg_pos > 0) {
1931 if (optuio->uio_segflg == UIO_SYSSPACE)
1933 optuio->uio_iov[errmsg_pos].iov_base,
1937 optuio->uio_iov[errmsg_pos].iov_base,
1944 free(ip4, M_PRISON);
1947 free(ip6, M_PRISON);
1950 free(g_path, M_TEMP);
1957 * struct jail_get_args {
1958 * struct iovec *iovp;
1959 * unsigned int iovcnt;
1964 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1969 /* Check that we have an even number of iovecs. */
1970 if (uap->iovcnt & 1)
1973 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1976 error = kern_jail_get(td, auio, uap->flags);
1978 error = copyout(auio->uio_iov, uap->iovp,
1979 uap->iovcnt * sizeof (struct iovec));
1985 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1987 struct prison *pr, *mypr;
1989 struct vfsoptlist *opts;
1990 char *errmsg, *name;
1991 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1993 if (flags & ~JAIL_GET_MASK)
1996 /* Get the parameter list. */
1997 error = vfs_buildopts(optuio, &opts);
2000 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
2001 mypr = td->td_ucred->cr_prison;
2004 * Find the prison specified by one of: lastjid, jid, name.
2006 sx_slock(&allprison_lock);
2007 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2009 TAILQ_FOREACH(pr, &allprison, pr_list) {
2010 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2011 mtx_lock(&pr->pr_mtx);
2012 if (pr->pr_ref > 0 &&
2013 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2015 mtx_unlock(&pr->pr_mtx);
2021 vfs_opterror(opts, "no jail after %d", jid);
2022 goto done_unlock_list;
2023 } else if (error != ENOENT)
2024 goto done_unlock_list;
2026 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2029 pr = prison_find_child(mypr, jid);
2031 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2032 mtx_unlock(&pr->pr_mtx);
2034 vfs_opterror(opts, "jail %d is dying",
2036 goto done_unlock_list;
2041 vfs_opterror(opts, "jail %d not found", jid);
2042 goto done_unlock_list;
2044 } else if (error != ENOENT)
2045 goto done_unlock_list;
2047 error = vfs_getopt(opts, "name", (void **)&name, &len);
2049 if (len == 0 || name[len - 1] != '\0') {
2051 goto done_unlock_list;
2053 pr = prison_find_name(mypr, name);
2055 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2056 mtx_unlock(&pr->pr_mtx);
2058 vfs_opterror(opts, "jail \"%s\" is dying",
2060 goto done_unlock_list;
2065 vfs_opterror(opts, "jail \"%s\" not found", name);
2066 goto done_unlock_list;
2067 } else if (error != ENOENT)
2068 goto done_unlock_list;
2070 vfs_opterror(opts, "no jail specified");
2072 goto done_unlock_list;
2075 /* Get the parameters of the prison. */
2078 td->td_retval[0] = pr->pr_id;
2079 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2080 if (error != 0 && error != ENOENT)
2082 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2083 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2084 if (error != 0 && error != ENOENT)
2086 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2087 if (error != 0 && error != ENOENT)
2089 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2090 sizeof(pr->pr_cpuset->cs_id));
2091 if (error != 0 && error != ENOENT)
2093 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2094 if (error != 0 && error != ENOENT)
2097 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2098 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2099 if (error != 0 && error != ENOENT)
2103 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2104 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2105 if (error != 0 && error != ENOENT)
2108 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2109 sizeof(pr->pr_securelevel));
2110 if (error != 0 && error != ENOENT)
2112 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2113 sizeof(pr->pr_childcount));
2114 if (error != 0 && error != ENOENT)
2116 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2117 sizeof(pr->pr_childmax));
2118 if (error != 0 && error != ENOENT)
2120 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2121 if (error != 0 && error != ENOENT)
2123 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2124 if (error != 0 && error != ENOENT)
2126 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2127 if (error != 0 && error != ENOENT)
2129 #ifdef COMPAT_FREEBSD32
2130 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2131 uint32_t hid32 = pr->pr_hostid;
2133 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2136 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2137 sizeof(pr->pr_hostid));
2138 if (error != 0 && error != ENOENT)
2140 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2141 sizeof(pr->pr_enforce_statfs));
2142 if (error != 0 && error != ENOENT)
2144 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2145 sizeof(pr->pr_devfs_rsnum));
2146 if (error != 0 && error != ENOENT)
2148 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2150 if (pr_flag_names[fi] == NULL)
2152 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2153 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2154 if (error != 0 && error != ENOENT)
2157 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2158 if (error != 0 && error != ENOENT)
2161 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2164 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2165 i = pr_flag_jailsys[fi].disable &&
2166 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2167 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2170 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2171 if (error != 0 && error != ENOENT)
2174 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2176 if (pr_allow_names[fi] == NULL)
2178 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2179 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2180 if (error != 0 && error != ENOENT)
2183 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2184 if (error != 0 && error != ENOENT)
2187 i = (pr->pr_uref == 0);
2188 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2189 if (error != 0 && error != ENOENT)
2192 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2193 if (error != 0 && error != ENOENT)
2195 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2196 sizeof(pr->pr_osreldate));
2197 if (error != 0 && error != ENOENT)
2199 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2200 if (error != 0 && error != ENOENT)
2203 /* Get the module parameters. */
2204 mtx_unlock(&pr->pr_mtx);
2206 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2209 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2211 /* By now, all parameters should have been noted. */
2212 TAILQ_FOREACH(opt, opts, link) {
2213 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2215 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2220 /* Write the fetched parameters back to userspace. */
2222 TAILQ_FOREACH(opt, opts, link) {
2223 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2224 pos = 2 * opt->pos + 1;
2225 optuio->uio_iov[pos].iov_len = opt->len;
2226 if (opt->value != NULL) {
2227 if (optuio->uio_segflg == UIO_SYSSPACE) {
2229 optuio->uio_iov[pos].iov_base,
2232 error = copyout(opt->value,
2233 optuio->uio_iov[pos].iov_base,
2244 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2248 sx_sunlock(&allprison_lock);
2250 if (error && errmsg_pos >= 0) {
2251 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2252 errmsg_pos = 2 * errmsg_pos + 1;
2253 if (errmsg_len > 0) {
2254 if (optuio->uio_segflg == UIO_SYSSPACE)
2256 optuio->uio_iov[errmsg_pos].iov_base,
2260 optuio->uio_iov[errmsg_pos].iov_base,
2270 * struct jail_remove_args {
2275 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2277 struct prison *pr, *cpr, *lpr, *tpr;
2280 error = priv_check(td, PRIV_JAIL_REMOVE);
2284 sx_xlock(&allprison_lock);
2285 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2287 sx_xunlock(&allprison_lock);
2291 /* Remove all descendants of this prison, then remove this prison. */
2293 pr->pr_flags |= PR_REMOVE;
2294 if (!LIST_EMPTY(&pr->pr_children)) {
2295 mtx_unlock(&pr->pr_mtx);
2297 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2298 mtx_lock(&cpr->pr_mtx);
2299 if (cpr->pr_ref > 0) {
2302 cpr->pr_flags |= PR_REMOVE;
2304 /* Already removed - do not do it again. */
2307 mtx_unlock(&cpr->pr_mtx);
2309 mtx_lock(&lpr->pr_mtx);
2310 prison_remove_one(lpr);
2311 sx_xlock(&allprison_lock);
2316 mtx_lock(&lpr->pr_mtx);
2317 prison_remove_one(lpr);
2318 sx_xlock(&allprison_lock);
2320 mtx_lock(&pr->pr_mtx);
2322 prison_remove_one(pr);
2327 prison_remove_one(struct prison *pr)
2332 /* If the prison was persistent, it is not anymore. */
2334 if (pr->pr_flags & PR_PERSIST) {
2337 pr->pr_flags &= ~PR_PERSIST;
2341 * jail_remove added a reference. If that's the only one, remove
2344 KASSERT(pr->pr_ref > 0,
2345 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2346 if (pr->pr_ref == 1) {
2348 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2352 mtx_unlock(&pr->pr_mtx);
2353 sx_xunlock(&allprison_lock);
2355 * Kill all processes unfortunate enough to be attached to this prison.
2357 sx_slock(&allproc_lock);
2358 LIST_FOREACH(p, &allproc, p_list) {
2360 if (p->p_state != PRS_NEW && p->p_ucred &&
2361 p->p_ucred->cr_prison == pr)
2362 kern_psignal(p, SIGKILL);
2365 sx_sunlock(&allproc_lock);
2366 /* Remove the temporary reference added by jail_remove. */
2367 prison_deref(pr, deuref | PD_DEREF);
2372 * struct jail_attach_args {
2377 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2382 error = priv_check(td, PRIV_JAIL_ATTACH);
2386 sx_slock(&allprison_lock);
2387 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2389 sx_sunlock(&allprison_lock);
2394 * Do not allow a process to attach to a prison that is not
2395 * considered to be "alive".
2397 if (pr->pr_uref == 0) {
2398 mtx_unlock(&pr->pr_mtx);
2399 sx_sunlock(&allprison_lock);
2403 return (do_jail_attach(td, pr));
2407 do_jail_attach(struct thread *td, struct prison *pr)
2411 struct ucred *newcred, *oldcred;
2415 * XXX: Note that there is a slight race here if two threads
2416 * in the same privileged process attempt to attach to two
2417 * different jails at the same time. It is important for
2418 * user processes not to do this, or they might end up with
2419 * a process root from one prison, but attached to the jail
2424 mtx_unlock(&pr->pr_mtx);
2426 /* Let modules do whatever they need to prepare for attaching. */
2427 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2429 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2432 sx_sunlock(&allprison_lock);
2435 * Reparent the newly attached process to this jail.
2437 ppr = td->td_ucred->cr_prison;
2439 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2443 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2444 if ((error = change_dir(pr->pr_root, td)) != 0)
2447 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2450 VOP_UNLOCK(pr->pr_root, 0);
2451 if ((error = change_root(pr->pr_root, td)))
2456 oldcred = p->p_ucred;
2458 crcopy(newcred, oldcred);
2459 newcred->cr_prison = pr;
2460 p->p_ucred = newcred;
2463 racct_proc_ucred_changed(p, oldcred, newcred);
2466 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2469 VOP_UNLOCK(pr->pr_root, 0);
2471 /* Tell modules this thread is still in its old jail after all. */
2472 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2473 prison_deref(pr, PD_DEREF | PD_DEUREF);
2479 * Returns a locked prison instance, or NULL on failure.
2482 prison_find(int prid)
2486 sx_assert(&allprison_lock, SX_LOCKED);
2487 TAILQ_FOREACH(pr, &allprison, pr_list) {
2488 if (pr->pr_id == prid) {
2489 mtx_lock(&pr->pr_mtx);
2492 mtx_unlock(&pr->pr_mtx);
2499 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2502 prison_find_child(struct prison *mypr, int prid)
2507 sx_assert(&allprison_lock, SX_LOCKED);
2508 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2509 if (pr->pr_id == prid) {
2510 mtx_lock(&pr->pr_mtx);
2513 mtx_unlock(&pr->pr_mtx);
2520 * Look for the name relative to mypr. Returns a locked prison or NULL.
2523 prison_find_name(struct prison *mypr, const char *name)
2525 struct prison *pr, *deadpr;
2529 sx_assert(&allprison_lock, SX_LOCKED);
2530 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2533 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2534 if (!strcmp(pr->pr_name + mylen, name)) {
2535 mtx_lock(&pr->pr_mtx);
2536 if (pr->pr_ref > 0) {
2537 if (pr->pr_uref > 0)
2541 mtx_unlock(&pr->pr_mtx);
2544 /* There was no valid prison - perhaps there was a dying one. */
2545 if (deadpr != NULL) {
2546 mtx_lock(&deadpr->pr_mtx);
2547 if (deadpr->pr_ref == 0) {
2548 mtx_unlock(&deadpr->pr_mtx);
2556 * See if a prison has the specific flag set.
2559 prison_flag(struct ucred *cred, unsigned flag)
2562 /* This is an atomic read, so no locking is necessary. */
2563 return (cred->cr_prison->pr_flags & flag);
2567 prison_allow(struct ucred *cred, unsigned flag)
2570 /* This is an atomic read, so no locking is necessary. */
2571 return (cred->cr_prison->pr_allow & flag);
2575 * Remove a prison reference. If that was the last reference, remove the
2576 * prison itself - but not in this context in case there are locks held.
2579 prison_free_locked(struct prison *pr)
2582 mtx_assert(&pr->pr_mtx, MA_OWNED);
2584 if (pr->pr_ref == 0) {
2585 mtx_unlock(&pr->pr_mtx);
2586 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2587 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2590 mtx_unlock(&pr->pr_mtx);
2594 prison_free(struct prison *pr)
2597 mtx_lock(&pr->pr_mtx);
2598 prison_free_locked(pr);
2602 prison_complete(void *context, int pending)
2605 prison_deref((struct prison *)context, 0);
2609 * Remove a prison reference (usually). This internal version assumes no
2610 * mutexes are held, except perhaps the prison itself. If there are no more
2611 * references, release and delist the prison. On completion, the prison lock
2612 * and the allprison lock are both unlocked.
2615 prison_deref(struct prison *pr, int flags)
2617 struct prison *ppr, *tpr;
2619 if (!(flags & PD_LOCKED))
2620 mtx_lock(&pr->pr_mtx);
2622 if (flags & PD_DEUREF) {
2624 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2626 if (flags & PD_DEREF)
2628 /* If the prison still has references, nothing else to do. */
2629 if (pr->pr_ref > 0) {
2630 mtx_unlock(&pr->pr_mtx);
2631 if (flags & PD_LIST_SLOCKED)
2632 sx_sunlock(&allprison_lock);
2633 else if (flags & PD_LIST_XLOCKED)
2634 sx_xunlock(&allprison_lock);
2638 mtx_unlock(&pr->pr_mtx);
2639 if (flags & PD_LIST_SLOCKED) {
2640 if (!sx_try_upgrade(&allprison_lock)) {
2641 sx_sunlock(&allprison_lock);
2642 sx_xlock(&allprison_lock);
2644 } else if (!(flags & PD_LIST_XLOCKED))
2645 sx_xlock(&allprison_lock);
2647 TAILQ_REMOVE(&allprison, pr, pr_list);
2648 LIST_REMOVE(pr, pr_sibling);
2649 ppr = pr->pr_parent;
2650 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2651 tpr->pr_childcount--;
2652 sx_xunlock(&allprison_lock);
2655 if (pr->pr_vnet != ppr->pr_vnet)
2656 vnet_destroy(pr->pr_vnet);
2658 if (pr->pr_root != NULL)
2660 mtx_destroy(&pr->pr_mtx);
2662 free(pr->pr_ip4, M_PRISON);
2665 free(pr->pr_ip6, M_PRISON);
2667 if (pr->pr_cpuset != NULL)
2668 cpuset_rel(pr->pr_cpuset);
2672 prison_racct_detach(pr);
2676 /* Removing a prison frees a reference on its parent. */
2678 mtx_lock(&pr->pr_mtx);
2679 flags = PD_DEREF | PD_DEUREF;
2684 prison_hold_locked(struct prison *pr)
2687 mtx_assert(&pr->pr_mtx, MA_OWNED);
2688 KASSERT(pr->pr_ref > 0,
2689 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2694 prison_hold(struct prison *pr)
2697 mtx_lock(&pr->pr_mtx);
2698 prison_hold_locked(pr);
2699 mtx_unlock(&pr->pr_mtx);
2703 prison_proc_hold(struct prison *pr)
2706 mtx_lock(&pr->pr_mtx);
2707 KASSERT(pr->pr_uref > 0,
2708 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2710 mtx_unlock(&pr->pr_mtx);
2714 prison_proc_free(struct prison *pr)
2717 mtx_lock(&pr->pr_mtx);
2718 KASSERT(pr->pr_uref > 0,
2719 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2720 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2726 * Restrict a prison's IP address list with its parent's, possibly replacing
2727 * it. Return true if the replacement buffer was used (or would have been).
2730 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2735 ppr = pr->pr_parent;
2736 if (!(pr->pr_flags & PR_IP4_USER)) {
2737 /* This has no user settings, so just copy the parent's list. */
2738 if (pr->pr_ip4s < ppr->pr_ip4s) {
2740 * There's no room for the parent's list. Use the
2741 * new list buffer, which is assumed to be big enough
2742 * (if it was passed). If there's no buffer, try to
2746 if (newip4 == NULL) {
2747 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2748 M_PRISON, M_NOWAIT);
2752 if (newip4 != NULL) {
2753 bcopy(ppr->pr_ip4, newip4,
2754 ppr->pr_ip4s * sizeof(*newip4));
2755 free(pr->pr_ip4, M_PRISON);
2756 pr->pr_ip4 = newip4;
2757 pr->pr_ip4s = ppr->pr_ip4s;
2761 pr->pr_ip4s = ppr->pr_ip4s;
2762 if (pr->pr_ip4s > 0)
2763 bcopy(ppr->pr_ip4, pr->pr_ip4,
2764 pr->pr_ip4s * sizeof(*newip4));
2765 else if (pr->pr_ip4 != NULL) {
2766 free(pr->pr_ip4, M_PRISON);
2769 } else if (pr->pr_ip4s > 0) {
2770 /* Remove addresses that aren't in the parent. */
2771 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2772 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2774 if (ij < ppr->pr_ip4s)
2777 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2778 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2781 for (ij = 1; ii < pr->pr_ip4s; ) {
2782 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2786 switch (ij >= ppr->pr_ip4s ? -1 :
2787 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2789 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2790 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2801 if (pr->pr_ip4s == 0) {
2802 pr->pr_flags |= PR_IP4_DISABLE;
2803 free(pr->pr_ip4, M_PRISON);
2811 * Pass back primary IPv4 address of this jail.
2813 * If not restricted return success but do not alter the address. Caller has
2814 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2816 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2817 * Address returned in NBO.
2820 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2824 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2825 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2827 pr = cred->cr_prison;
2828 if (!(pr->pr_flags & PR_IP4))
2830 mtx_lock(&pr->pr_mtx);
2831 if (!(pr->pr_flags & PR_IP4)) {
2832 mtx_unlock(&pr->pr_mtx);
2835 if (pr->pr_ip4 == NULL) {
2836 mtx_unlock(&pr->pr_mtx);
2837 return (EAFNOSUPPORT);
2840 ia->s_addr = pr->pr_ip4[0].s_addr;
2841 mtx_unlock(&pr->pr_mtx);
2846 * Return 1 if we should do proper source address selection or are not jailed.
2847 * We will return 0 if we should bypass source address selection in favour
2848 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2850 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2853 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2859 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2860 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2865 pr = cred->cr_prison;
2866 if (pr->pr_flags & PR_IP4_SADDRSEL)
2869 lia.s_addr = INADDR_ANY;
2870 error = prison_get_ip4(cred, &lia);
2873 if (lia.s_addr == INADDR_ANY)
2876 ia->s_addr = lia.s_addr;
2881 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2884 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2891 * No need to lock since the PR_IP4_USER flag can't be altered for
2894 while (pr1 != &prison0 &&
2896 !(pr1->pr_flags & PR_VNET) &&
2898 !(pr1->pr_flags & PR_IP4_USER))
2899 pr1 = pr1->pr_parent;
2900 while (pr2 != &prison0 &&
2902 !(pr2->pr_flags & PR_VNET) &&
2904 !(pr2->pr_flags & PR_IP4_USER))
2905 pr2 = pr2->pr_parent;
2906 return (pr1 == pr2);
2910 * Make sure our (source) address is set to something meaningful to this
2913 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2914 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2915 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2918 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2924 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2925 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2927 pr = cred->cr_prison;
2928 if (!(pr->pr_flags & PR_IP4))
2930 mtx_lock(&pr->pr_mtx);
2931 if (!(pr->pr_flags & PR_IP4)) {
2932 mtx_unlock(&pr->pr_mtx);
2935 if (pr->pr_ip4 == NULL) {
2936 mtx_unlock(&pr->pr_mtx);
2937 return (EAFNOSUPPORT);
2940 ia0.s_addr = ntohl(ia->s_addr);
2941 if (ia0.s_addr == INADDR_LOOPBACK) {
2942 ia->s_addr = pr->pr_ip4[0].s_addr;
2943 mtx_unlock(&pr->pr_mtx);
2947 if (ia0.s_addr == INADDR_ANY) {
2949 * In case there is only 1 IPv4 address, bind directly.
2951 if (pr->pr_ip4s == 1)
2952 ia->s_addr = pr->pr_ip4[0].s_addr;
2953 mtx_unlock(&pr->pr_mtx);
2957 error = _prison_check_ip4(pr, ia);
2958 mtx_unlock(&pr->pr_mtx);
2963 * Rewrite destination address in case we will connect to loopback address.
2965 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2966 * Address passed in in NBO and returned in NBO.
2969 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2973 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2974 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2976 pr = cred->cr_prison;
2977 if (!(pr->pr_flags & PR_IP4))
2979 mtx_lock(&pr->pr_mtx);
2980 if (!(pr->pr_flags & PR_IP4)) {
2981 mtx_unlock(&pr->pr_mtx);
2984 if (pr->pr_ip4 == NULL) {
2985 mtx_unlock(&pr->pr_mtx);
2986 return (EAFNOSUPPORT);
2989 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2990 ia->s_addr = pr->pr_ip4[0].s_addr;
2991 mtx_unlock(&pr->pr_mtx);
2996 * Return success because nothing had to be changed.
2998 mtx_unlock(&pr->pr_mtx);
3003 * Check if given address belongs to the jail referenced by cred/prison.
3005 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
3006 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3007 * doesn't allow IPv4. Address passed in in NBO.
3010 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
3015 * Check the primary IP.
3017 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3021 * All the other IPs are sorted so we can do a binary search.
3024 z = pr->pr_ip4s - 2;
3027 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3036 return (EADDRNOTAVAIL);
3040 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
3045 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3046 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3048 pr = cred->cr_prison;
3049 if (!(pr->pr_flags & PR_IP4))
3051 mtx_lock(&pr->pr_mtx);
3052 if (!(pr->pr_flags & PR_IP4)) {
3053 mtx_unlock(&pr->pr_mtx);
3056 if (pr->pr_ip4 == NULL) {
3057 mtx_unlock(&pr->pr_mtx);
3058 return (EAFNOSUPPORT);
3061 error = _prison_check_ip4(pr, ia);
3062 mtx_unlock(&pr->pr_mtx);
3069 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3074 ppr = pr->pr_parent;
3075 if (!(pr->pr_flags & PR_IP6_USER)) {
3076 /* This has no user settings, so just copy the parent's list. */
3077 if (pr->pr_ip6s < ppr->pr_ip6s) {
3079 * There's no room for the parent's list. Use the
3080 * new list buffer, which is assumed to be big enough
3081 * (if it was passed). If there's no buffer, try to
3085 if (newip6 == NULL) {
3086 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3087 M_PRISON, M_NOWAIT);
3091 if (newip6 != NULL) {
3092 bcopy(ppr->pr_ip6, newip6,
3093 ppr->pr_ip6s * sizeof(*newip6));
3094 free(pr->pr_ip6, M_PRISON);
3095 pr->pr_ip6 = newip6;
3096 pr->pr_ip6s = ppr->pr_ip6s;
3100 pr->pr_ip6s = ppr->pr_ip6s;
3101 if (pr->pr_ip6s > 0)
3102 bcopy(ppr->pr_ip6, pr->pr_ip6,
3103 pr->pr_ip6s * sizeof(*newip6));
3104 else if (pr->pr_ip6 != NULL) {
3105 free(pr->pr_ip6, M_PRISON);
3108 } else if (pr->pr_ip6s > 0) {
3109 /* Remove addresses that aren't in the parent. */
3110 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3111 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3114 if (ij < ppr->pr_ip6s)
3117 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3118 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3121 for (ij = 1; ii < pr->pr_ip6s; ) {
3122 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3127 switch (ij >= ppr->pr_ip6s ? -1 :
3128 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3130 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3131 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3142 if (pr->pr_ip6s == 0) {
3143 pr->pr_flags |= PR_IP6_DISABLE;
3144 free(pr->pr_ip6, M_PRISON);
3152 * Pass back primary IPv6 address for this jail.
3154 * If not restricted return success but do not alter the address. Caller has
3155 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3157 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3160 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3164 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3165 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3167 pr = cred->cr_prison;
3168 if (!(pr->pr_flags & PR_IP6))
3170 mtx_lock(&pr->pr_mtx);
3171 if (!(pr->pr_flags & PR_IP6)) {
3172 mtx_unlock(&pr->pr_mtx);
3175 if (pr->pr_ip6 == NULL) {
3176 mtx_unlock(&pr->pr_mtx);
3177 return (EAFNOSUPPORT);
3180 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3181 mtx_unlock(&pr->pr_mtx);
3186 * Return 1 if we should do proper source address selection or are not jailed.
3187 * We will return 0 if we should bypass source address selection in favour
3188 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3190 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3193 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3196 struct in6_addr lia6;
3199 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3200 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3205 pr = cred->cr_prison;
3206 if (pr->pr_flags & PR_IP6_SADDRSEL)
3210 error = prison_get_ip6(cred, &lia6);
3213 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3216 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3221 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3224 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3230 while (pr1 != &prison0 &&
3232 !(pr1->pr_flags & PR_VNET) &&
3234 !(pr1->pr_flags & PR_IP6_USER))
3235 pr1 = pr1->pr_parent;
3236 while (pr2 != &prison0 &&
3238 !(pr2->pr_flags & PR_VNET) &&
3240 !(pr2->pr_flags & PR_IP6_USER))
3241 pr2 = pr2->pr_parent;
3242 return (pr1 == pr2);
3246 * Make sure our (source) address is set to something meaningful to this jail.
3248 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3249 * when needed while binding.
3251 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3252 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3253 * doesn't allow IPv6.
3256 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3261 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3262 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3264 pr = cred->cr_prison;
3265 if (!(pr->pr_flags & PR_IP6))
3267 mtx_lock(&pr->pr_mtx);
3268 if (!(pr->pr_flags & PR_IP6)) {
3269 mtx_unlock(&pr->pr_mtx);
3272 if (pr->pr_ip6 == NULL) {
3273 mtx_unlock(&pr->pr_mtx);
3274 return (EAFNOSUPPORT);
3277 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3278 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3279 mtx_unlock(&pr->pr_mtx);
3283 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3285 * In case there is only 1 IPv6 address, and v6only is true,
3286 * then bind directly.
3288 if (v6only != 0 && pr->pr_ip6s == 1)
3289 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3290 mtx_unlock(&pr->pr_mtx);
3294 error = _prison_check_ip6(pr, ia6);
3295 mtx_unlock(&pr->pr_mtx);
3300 * Rewrite destination address in case we will connect to loopback address.
3302 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3305 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3309 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3310 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3312 pr = cred->cr_prison;
3313 if (!(pr->pr_flags & PR_IP6))
3315 mtx_lock(&pr->pr_mtx);
3316 if (!(pr->pr_flags & PR_IP6)) {
3317 mtx_unlock(&pr->pr_mtx);
3320 if (pr->pr_ip6 == NULL) {
3321 mtx_unlock(&pr->pr_mtx);
3322 return (EAFNOSUPPORT);
3325 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3326 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3327 mtx_unlock(&pr->pr_mtx);
3332 * Return success because nothing had to be changed.
3334 mtx_unlock(&pr->pr_mtx);
3339 * Check if given address belongs to the jail referenced by cred/prison.
3341 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3342 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3343 * doesn't allow IPv6.
3346 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3351 * Check the primary IP.
3353 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3357 * All the other IPs are sorted so we can do a binary search.
3360 z = pr->pr_ip6s - 2;
3363 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3372 return (EADDRNOTAVAIL);
3376 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3381 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3382 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3384 pr = cred->cr_prison;
3385 if (!(pr->pr_flags & PR_IP6))
3387 mtx_lock(&pr->pr_mtx);
3388 if (!(pr->pr_flags & PR_IP6)) {
3389 mtx_unlock(&pr->pr_mtx);
3392 if (pr->pr_ip6 == NULL) {
3393 mtx_unlock(&pr->pr_mtx);
3394 return (EAFNOSUPPORT);
3397 error = _prison_check_ip6(pr, ia6);
3398 mtx_unlock(&pr->pr_mtx);
3404 * Check if a jail supports the given address family.
3406 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3410 prison_check_af(struct ucred *cred, int af)
3415 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3417 pr = cred->cr_prison;
3419 /* Prisons with their own network stack are not limited. */
3420 if (prison_owns_vnet(cred))
3429 if (pr->pr_flags & PR_IP4)
3431 mtx_lock(&pr->pr_mtx);
3432 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3433 error = EAFNOSUPPORT;
3434 mtx_unlock(&pr->pr_mtx);
3440 if (pr->pr_flags & PR_IP6)
3442 mtx_lock(&pr->pr_mtx);
3443 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3444 error = EAFNOSUPPORT;
3445 mtx_unlock(&pr->pr_mtx);
3453 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3454 error = EAFNOSUPPORT;
3460 * Check if given address belongs to the jail referenced by cred (wrapper to
3461 * prison_check_ip[46]).
3463 * Returns 0 if jail doesn't restrict the address family or if address belongs
3464 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3465 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3468 prison_if(struct ucred *cred, struct sockaddr *sa)
3471 struct sockaddr_in *sai;
3474 struct sockaddr_in6 *sai6;
3478 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3479 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3482 if (prison_owns_vnet(cred))
3487 switch (sa->sa_family)
3491 sai = (struct sockaddr_in *)sa;
3492 error = prison_check_ip4(cred, &sai->sin_addr);
3497 sai6 = (struct sockaddr_in6 *)sa;
3498 error = prison_check_ip6(cred, &sai6->sin6_addr);
3502 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3503 error = EAFNOSUPPORT;
3509 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3512 prison_check(struct ucred *cred1, struct ucred *cred2)
3515 return ((cred1->cr_prison == cred2->cr_prison ||
3516 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3520 * Return 1 if p2 is a child of p1, otherwise 0.
3523 prison_ischild(struct prison *pr1, struct prison *pr2)
3526 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3533 * Return 1 if the passed credential is in a jail, otherwise 0.
3536 jailed(struct ucred *cred)
3539 return (cred->cr_prison != &prison0);
3543 * Return 1 if the passed credential is in a jail and that jail does not
3544 * have its own virtual network stack, otherwise 0.
3547 jailed_without_vnet(struct ucred *cred)
3553 if (prison_owns_vnet(cred))
3561 * Return the correct hostname (domainname, et al) for the passed credential.
3564 getcredhostname(struct ucred *cred, char *buf, size_t size)
3569 * A NULL credential can be used to shortcut to the physical
3570 * system's hostname.
3572 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3573 mtx_lock(&pr->pr_mtx);
3574 strlcpy(buf, pr->pr_hostname, size);
3575 mtx_unlock(&pr->pr_mtx);
3579 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3582 mtx_lock(&cred->cr_prison->pr_mtx);
3583 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3584 mtx_unlock(&cred->cr_prison->pr_mtx);
3588 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3591 mtx_lock(&cred->cr_prison->pr_mtx);
3592 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3593 mtx_unlock(&cred->cr_prison->pr_mtx);
3597 getcredhostid(struct ucred *cred, unsigned long *hostid)
3600 mtx_lock(&cred->cr_prison->pr_mtx);
3601 *hostid = cred->cr_prison->pr_hostid;
3602 mtx_unlock(&cred->cr_prison->pr_mtx);
3607 * Determine whether the prison represented by cred owns
3608 * its vnet rather than having it inherited.
3610 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3613 prison_owns_vnet(struct ucred *cred)
3617 * vnets cannot be added/removed after jail creation,
3618 * so no need to lock here.
3620 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3625 * Determine whether the subject represented by cred can "see"
3626 * status of a mount point.
3627 * Returns: 0 for permitted, ENOENT otherwise.
3628 * XXX: This function should be called cr_canseemount() and should be
3629 * placed in kern_prot.c.
3632 prison_canseemount(struct ucred *cred, struct mount *mp)
3638 pr = cred->cr_prison;
3639 if (pr->pr_enforce_statfs == 0)
3641 if (pr->pr_root->v_mount == mp)
3643 if (pr->pr_enforce_statfs == 2)
3646 * If jail's chroot directory is set to "/" we should be able to see
3647 * all mount-points from inside a jail.
3648 * This is ugly check, but this is the only situation when jail's
3649 * directory ends with '/'.
3651 if (strcmp(pr->pr_path, "/") == 0)
3653 len = strlen(pr->pr_path);
3655 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3658 * Be sure that we don't have situation where jail's root directory
3659 * is "/some/path" and mount point is "/some/pathpath".
3661 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3667 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3669 char jpath[MAXPATHLEN];
3673 pr = cred->cr_prison;
3674 if (pr->pr_enforce_statfs == 0)
3676 if (prison_canseemount(cred, mp) != 0) {
3677 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3678 strlcpy(sp->f_mntonname, "[restricted]",
3679 sizeof(sp->f_mntonname));
3682 if (pr->pr_root->v_mount == mp) {
3684 * Clear current buffer data, so we are sure nothing from
3685 * the valid path left there.
3687 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3688 *sp->f_mntonname = '/';
3692 * If jail's chroot directory is set to "/" we should be able to see
3693 * all mount-points from inside a jail.
3695 if (strcmp(pr->pr_path, "/") == 0)
3697 len = strlen(pr->pr_path);
3698 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3700 * Clear current buffer data, so we are sure nothing from
3701 * the valid path left there.
3703 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3704 if (*jpath == '\0') {
3705 /* Should never happen. */
3706 *sp->f_mntonname = '/';
3708 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3713 * Check with permission for a specific privilege is granted within jail. We
3714 * have a specific list of accepted privileges; the rest are denied.
3717 prison_priv_check(struct ucred *cred, int priv)
3725 * Privileges specific to prisons with a virtual network stack.
3726 * There might be a duplicate entry here in case the privilege
3727 * is only granted conditionally in the legacy jail case.
3732 * NFS-specific privileges.
3734 case PRIV_NFS_DAEMON:
3735 case PRIV_NFS_LOCKD:
3738 * Network stack privileges.
3740 case PRIV_NET_BRIDGE:
3743 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3744 case PRIV_NET_ROUTE:
3746 case PRIV_NET_SETIFMTU:
3747 case PRIV_NET_SETIFFLAGS:
3748 case PRIV_NET_SETIFCAP:
3749 case PRIV_NET_SETIFDESCR:
3750 case PRIV_NET_SETIFNAME :
3751 case PRIV_NET_SETIFMETRIC:
3752 case PRIV_NET_SETIFPHYS:
3753 case PRIV_NET_SETIFMAC:
3754 case PRIV_NET_ADDMULTI:
3755 case PRIV_NET_DELMULTI:
3756 case PRIV_NET_HWIOCTL:
3757 case PRIV_NET_SETLLADDR:
3758 case PRIV_NET_ADDIFGROUP:
3759 case PRIV_NET_DELIFGROUP:
3760 case PRIV_NET_IFCREATE:
3761 case PRIV_NET_IFDESTROY:
3762 case PRIV_NET_ADDIFADDR:
3763 case PRIV_NET_DELIFADDR:
3766 case PRIV_NET_SETIFVNET:
3767 case PRIV_NET_SETIFFIB:
3770 * 802.11-related privileges.
3772 case PRIV_NET80211_GETKEY:
3774 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3779 * AppleTalk privileges.
3781 case PRIV_NETATALK_RESERVEDPORT:
3786 case PRIV_NETATM_CFG:
3787 case PRIV_NETATM_ADD:
3788 case PRIV_NETATM_DEL:
3789 case PRIV_NETATM_SET:
3792 * Bluetooth privileges.
3794 case PRIV_NETBLUETOOTH_RAW:
3798 * Netgraph and netgraph module privileges.
3800 case PRIV_NETGRAPH_CONTROL:
3802 case PRIV_NETGRAPH_TTY:
3806 * IPv4 and IPv6 privileges.
3808 case PRIV_NETINET_IPFW:
3809 case PRIV_NETINET_DIVERT:
3810 case PRIV_NETINET_PF:
3811 case PRIV_NETINET_DUMMYNET:
3812 case PRIV_NETINET_CARP:
3813 case PRIV_NETINET_MROUTE:
3814 case PRIV_NETINET_RAW:
3815 case PRIV_NETINET_ADDRCTRL6:
3816 case PRIV_NETINET_ND6:
3817 case PRIV_NETINET_SCOPE6:
3818 case PRIV_NETINET_ALIFETIME6:
3819 case PRIV_NETINET_IPSEC:
3820 case PRIV_NETINET_BINDANY:
3824 * IPX/SPX privileges.
3826 case PRIV_NETIPX_RESERVEDPORT:
3827 case PRIV_NETIPX_RAW:
3841 * No default: or deny here.
3842 * In case of no permit fall through to next switch().
3844 if (cred->cr_prison->pr_flags & PR_VNET)
3852 * Allow ktrace privileges for root in jail.
3858 * Allow jailed processes to configure audit identity and
3859 * submit audit records (login, etc). In the future we may
3860 * want to further refine the relationship between audit and
3863 case PRIV_AUDIT_GETAUDIT:
3864 case PRIV_AUDIT_SETAUDIT:
3865 case PRIV_AUDIT_SUBMIT:
3869 * Allow jailed processes to manipulate process UNIX
3870 * credentials in any way they see fit.
3872 case PRIV_CRED_SETUID:
3873 case PRIV_CRED_SETEUID:
3874 case PRIV_CRED_SETGID:
3875 case PRIV_CRED_SETEGID:
3876 case PRIV_CRED_SETGROUPS:
3877 case PRIV_CRED_SETREUID:
3878 case PRIV_CRED_SETREGID:
3879 case PRIV_CRED_SETRESUID:
3880 case PRIV_CRED_SETRESGID:
3883 * Jail implements visibility constraints already, so allow
3884 * jailed root to override uid/gid-based constraints.
3886 case PRIV_SEEOTHERGIDS:
3887 case PRIV_SEEOTHERUIDS:
3890 * Jail implements inter-process debugging limits already, so
3891 * allow jailed root various debugging privileges.
3893 case PRIV_DEBUG_DIFFCRED:
3894 case PRIV_DEBUG_SUGID:
3895 case PRIV_DEBUG_UNPRIV:
3898 * Allow jail to set various resource limits and login
3899 * properties, and for now, exceed process resource limits.
3901 case PRIV_PROC_LIMIT:
3902 case PRIV_PROC_SETLOGIN:
3903 case PRIV_PROC_SETRLIMIT:
3906 * System V and POSIX IPC privileges are granted in jail.
3909 case PRIV_IPC_WRITE:
3910 case PRIV_IPC_ADMIN:
3911 case PRIV_IPC_MSGSIZE:
3915 * Jail operations within a jail work on child jails.
3917 case PRIV_JAIL_ATTACH:
3919 case PRIV_JAIL_REMOVE:
3922 * Jail implements its own inter-process limits, so allow
3923 * root processes in jail to change scheduling on other
3924 * processes in the same jail. Likewise for signalling.
3926 case PRIV_SCHED_DIFFCRED:
3927 case PRIV_SCHED_CPUSET:
3928 case PRIV_SIGNAL_DIFFCRED:
3929 case PRIV_SIGNAL_SUGID:
3932 * Allow jailed processes to write to sysctls marked as jail
3935 case PRIV_SYSCTL_WRITEJAIL:
3938 * Allow root in jail to manage a variety of quota
3939 * properties. These should likely be conditional on a
3940 * configuration option.
3942 case PRIV_VFS_GETQUOTA:
3943 case PRIV_VFS_SETQUOTA:
3946 * Since Jail relies on chroot() to implement file system
3947 * protections, grant many VFS privileges to root in jail.
3948 * Be careful to exclude mount-related and NFS-related
3952 case PRIV_VFS_WRITE:
3953 case PRIV_VFS_ADMIN:
3955 case PRIV_VFS_LOOKUP:
3956 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3957 case PRIV_VFS_CHFLAGS_DEV:
3958 case PRIV_VFS_CHOWN:
3959 case PRIV_VFS_CHROOT:
3960 case PRIV_VFS_RETAINSUGID:
3961 case PRIV_VFS_FCHROOT:
3963 case PRIV_VFS_SETGID:
3965 case PRIV_VFS_STICKYFILE:
3968 * As in the non-jail case, non-root users are expected to be
3969 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3970 * exists in the jail and they have permission to access it).
3972 case PRIV_KMEM_READ:
3976 * Depending on the global setting, allow privilege of
3977 * setting system flags.
3979 case PRIV_VFS_SYSFLAGS:
3980 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3986 * Depending on the global setting, allow privilege of
3987 * mounting/unmounting file systems.
3989 case PRIV_VFS_MOUNT:
3990 case PRIV_VFS_UNMOUNT:
3991 case PRIV_VFS_MOUNT_NONUSER:
3992 case PRIV_VFS_MOUNT_OWNER:
3993 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3994 cred->cr_prison->pr_enforce_statfs < 2)
4000 * Allow jailed root to bind reserved ports and reuse in-use
4003 case PRIV_NETINET_RESERVEDPORT:
4004 case PRIV_NETINET_REUSEPORT:
4008 * Allow jailed root to set certian IPv4/6 (option) headers.
4010 case PRIV_NETINET_SETHDROPTS:
4014 * Conditionally allow creating raw sockets in jail.
4016 case PRIV_NETINET_RAW:
4017 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4023 * Since jail implements its own visibility limits on netstat
4024 * sysctls, allow getcred. This allows identd to work in
4027 case PRIV_NETINET_GETCRED:
4031 * Allow jailed root to set loginclass.
4033 case PRIV_PROC_SETLOGINCLASS:
4038 * In all remaining cases, deny the privilege request. This
4039 * includes almost all network privileges, many system
4040 * configuration privileges.
4047 * Return the part of pr2's name that is relative to pr1, or the whole name
4048 * if it does not directly follow.
4052 prison_name(struct prison *pr1, struct prison *pr2)
4056 /* Jails see themselves as "0" (if they see themselves at all). */
4059 name = pr2->pr_name;
4060 if (prison_ischild(pr1, pr2)) {
4062 * pr1 isn't locked (and allprison_lock may not be either)
4063 * so its length can't be counted on. But the number of dots
4064 * can be counted on - and counted.
4066 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4067 name = strchr(name, '.') + 1;
4073 * Return the part of pr2's path that is relative to pr1, or the whole path
4074 * if it does not directly follow.
4077 prison_path(struct prison *pr1, struct prison *pr2)
4079 char *path1, *path2;
4082 path1 = pr1->pr_path;
4083 path2 = pr2->pr_path;
4084 if (!strcmp(path1, "/"))
4086 len1 = strlen(path1);
4087 if (strncmp(path1, path2, len1))
4089 if (path2[len1] == '\0')
4091 if (path2[len1] == '/')
4092 return (path2 + len1);
4098 * Jail-related sysctls.
4100 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4104 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4107 struct prison *pr, *cpr;
4109 struct in_addr *ip4 = NULL;
4113 struct in6_addr *ip6 = NULL;
4118 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4119 pr = req->td->td_ucred->cr_prison;
4121 sx_slock(&allprison_lock);
4122 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4123 #if defined(INET) || defined(INET6)
4126 mtx_lock(&cpr->pr_mtx);
4128 if (cpr->pr_ip4s > 0) {
4129 if (ip4s < cpr->pr_ip4s) {
4130 ip4s = cpr->pr_ip4s;
4131 mtx_unlock(&cpr->pr_mtx);
4132 ip4 = realloc(ip4, ip4s *
4133 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4136 bcopy(cpr->pr_ip4, ip4,
4137 cpr->pr_ip4s * sizeof(struct in_addr));
4141 if (cpr->pr_ip6s > 0) {
4142 if (ip6s < cpr->pr_ip6s) {
4143 ip6s = cpr->pr_ip6s;
4144 mtx_unlock(&cpr->pr_mtx);
4145 ip6 = realloc(ip6, ip6s *
4146 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4149 bcopy(cpr->pr_ip6, ip6,
4150 cpr->pr_ip6s * sizeof(struct in6_addr));
4153 if (cpr->pr_ref == 0) {
4154 mtx_unlock(&cpr->pr_mtx);
4157 bzero(xp, sizeof(*xp));
4158 xp->pr_version = XPRISON_VERSION;
4159 xp->pr_id = cpr->pr_id;
4160 xp->pr_state = cpr->pr_uref > 0
4161 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4162 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4163 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4164 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4166 xp->pr_ip4s = cpr->pr_ip4s;
4169 xp->pr_ip6s = cpr->pr_ip6s;
4171 mtx_unlock(&cpr->pr_mtx);
4172 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4176 if (xp->pr_ip4s > 0) {
4177 error = SYSCTL_OUT(req, ip4,
4178 xp->pr_ip4s * sizeof(struct in_addr));
4184 if (xp->pr_ip6s > 0) {
4185 error = SYSCTL_OUT(req, ip6,
4186 xp->pr_ip6s * sizeof(struct in6_addr));
4192 sx_sunlock(&allprison_lock);
4203 SYSCTL_OID(_security_jail, OID_AUTO, list,
4204 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4205 sysctl_jail_list, "S", "List of active jails");
4208 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4212 injail = jailed(req->td->td_ucred);
4213 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4218 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4219 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4220 sysctl_jail_jailed, "I", "Process in jail?");
4223 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4225 int error, havevnet;
4227 struct ucred *cred = req->td->td_ucred;
4229 havevnet = jailed(cred) && prison_owns_vnet(cred);
4233 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4238 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4239 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4240 sysctl_jail_vnet, "I", "Jail owns VNET?");
4242 #if defined(INET) || defined(INET6)
4243 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4244 &jail_max_af_ips, 0,
4245 "Number of IP addresses a jail may have at most per address family");
4249 * Default parameters for jail(2) compatability. For historical reasons,
4250 * the sysctl names have varying similarity to the parameter names. Prisons
4251 * just see their own parameters, and can't change them.
4254 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4257 int allow, error, i;
4259 pr = req->td->td_ucred->cr_prison;
4260 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4262 /* Get the current flag value, and convert it to a boolean. */
4263 i = (allow & arg2) ? 1 : 0;
4266 error = sysctl_handle_int(oidp, &i, 0, req);
4267 if (error || !req->newptr)
4273 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4276 mtx_lock(&prison0.pr_mtx);
4277 jail_default_allow = (jail_default_allow & ~arg2) | i;
4278 mtx_unlock(&prison0.pr_mtx);
4282 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4283 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4284 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4285 "Processes in jail can set their hostnames");
4286 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4287 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4288 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4289 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4290 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4291 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4292 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4293 "Processes in jail can use System V IPC primitives");
4294 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4295 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4296 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4297 "Prison root can create raw sockets");
4298 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4299 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4300 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4301 "Processes in jail can alter system file flags");
4302 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4303 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4304 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4305 "Processes in jail can mount/unmount jail-friendly file systems");
4306 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4307 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4308 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4309 "Processes in jail can mount the devfs file system");
4310 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4311 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4312 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4313 "Processes in jail can mount the fdescfs file system");
4314 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4315 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4316 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4317 "Processes in jail can mount the nullfs file system");
4318 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4319 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4320 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4321 "Processes in jail can mount the procfs file system");
4322 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
4323 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4324 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
4325 "Processes in jail can mount the linprocfs file system");
4326 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
4327 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4328 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
4329 "Processes in jail can mount the linsysfs file system");
4330 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4331 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4332 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4333 "Processes in jail can mount the tmpfs file system");
4334 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4335 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4336 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4337 "Processes in jail can mount the zfs file system");
4340 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4345 pr = req->td->td_ucred->cr_prison;
4346 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4347 error = sysctl_handle_int(oidp, &level, 0, req);
4348 if (error || !req->newptr)
4350 *(int *)arg1 = level;
4354 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4355 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4356 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4357 sysctl_jail_default_level, "I",
4358 "Processes in jail cannot see all mounted file systems");
4360 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4361 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4362 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4363 sysctl_jail_default_level, "I",
4364 "Ruleset for the devfs filesystem in jail");
4367 * Nodes to describe jail parameters. Maximum length of string parameters
4368 * is returned in the string itself, and the other parameters exist merely
4369 * to make themselves and their types known.
4371 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4375 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4382 switch (oidp->oid_kind & CTLTYPE)
4388 if (!(req->flags & SCTL_MASK32))
4390 return (SYSCTL_OUT(req, &l, sizeof(l)));
4394 return (SYSCTL_OUT(req, &i, sizeof(i)));
4395 case CTLTYPE_STRING:
4396 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4398 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4399 case CTLTYPE_STRUCT:
4401 return (SYSCTL_OUT(req, &s, sizeof(s)));
4407 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4408 * jail creation time but cannot be changed in an existing jail.
4410 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4411 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4412 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4413 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4414 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4415 "I", "Jail secure level");
4416 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4417 "Jail value for kern.osreldate and uname -K");
4418 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4419 "Jail value for kern.osrelease and uname -r");
4420 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4421 "I", "Jail cannot see all mounted file systems");
4422 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4423 "I", "Ruleset for in-jail devfs mounts");
4424 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4425 "B", "Jail persistence");
4427 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4428 "E,jailsys", "Virtual network stack");
4430 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4431 "B", "Jail is in the process of shutting down");
4433 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4434 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4435 "I", "Current number of child jails");
4436 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4437 "I", "Maximum number of child jails");
4439 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4440 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4442 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4443 "Jail NIS domainname");
4444 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4446 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4447 "LU", "Jail host ID");
4449 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4450 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4453 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4454 "Jail IPv4 address virtualization");
4455 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4456 "S,in_addr,a", "Jail IPv4 addresses");
4457 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4458 "B", "Do (not) use IPv4 source address selection rather than the "
4459 "primary jail IPv4 address.");
4462 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4463 "Jail IPv6 address virtualization");
4464 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4465 "S,in6_addr,a", "Jail IPv6 addresses");
4466 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4467 "B", "Do (not) use IPv6 source address selection rather than the "
4468 "primary jail IPv6 address.");
4471 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4472 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4473 "B", "Jail may set hostname");
4474 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4475 "B", "Jail may use SYSV IPC");
4476 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4477 "B", "Jail may create raw sockets");
4478 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4479 "B", "Jail may alter system file flags");
4480 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4481 "B", "Jail may set file quotas");
4482 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4483 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4485 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4486 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4487 "B", "Jail may mount/unmount jail-friendly file systems in general");
4488 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4489 "B", "Jail may mount the devfs file system");
4490 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4491 "B", "Jail may mount the fdescfs file system");
4492 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4493 "B", "Jail may mount the nullfs file system");
4494 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4495 "B", "Jail may mount the procfs file system");
4496 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
4497 "B", "Jail may mount the linprocfs file system");
4498 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
4499 "B", "Jail may mount the linsysfs file system");
4500 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4501 "B", "Jail may mount the tmpfs file system");
4502 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4503 "B", "Jail may mount the zfs file system");
4507 prison_racct_foreach(void (*callback)(struct racct *racct,
4508 void *arg2, void *arg3), void *arg2, void *arg3)
4510 struct prison_racct *prr;
4512 ASSERT_RACCT_ENABLED();
4514 sx_slock(&allprison_lock);
4515 LIST_FOREACH(prr, &allprison_racct, prr_next)
4516 (callback)(prr->prr_racct, arg2, arg3);
4517 sx_sunlock(&allprison_lock);
4520 static struct prison_racct *
4521 prison_racct_find_locked(const char *name)
4523 struct prison_racct *prr;
4525 ASSERT_RACCT_ENABLED();
4526 sx_assert(&allprison_lock, SA_XLOCKED);
4528 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4531 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4532 if (strcmp(name, prr->prr_name) != 0)
4535 /* Found prison_racct with a matching name? */
4536 prison_racct_hold(prr);
4540 /* Add new prison_racct. */
4541 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4542 racct_create(&prr->prr_racct);
4544 strcpy(prr->prr_name, name);
4545 refcount_init(&prr->prr_refcount, 1);
4546 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4551 struct prison_racct *
4552 prison_racct_find(const char *name)
4554 struct prison_racct *prr;
4556 ASSERT_RACCT_ENABLED();
4558 sx_xlock(&allprison_lock);
4559 prr = prison_racct_find_locked(name);
4560 sx_xunlock(&allprison_lock);
4565 prison_racct_hold(struct prison_racct *prr)
4568 ASSERT_RACCT_ENABLED();
4570 refcount_acquire(&prr->prr_refcount);
4574 prison_racct_free_locked(struct prison_racct *prr)
4577 ASSERT_RACCT_ENABLED();
4578 sx_assert(&allprison_lock, SA_XLOCKED);
4580 if (refcount_release(&prr->prr_refcount)) {
4581 racct_destroy(&prr->prr_racct);
4582 LIST_REMOVE(prr, prr_next);
4583 free(prr, M_PRISON_RACCT);
4588 prison_racct_free(struct prison_racct *prr)
4592 ASSERT_RACCT_ENABLED();
4593 sx_assert(&allprison_lock, SA_UNLOCKED);
4595 old = prr->prr_refcount;
4596 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4599 sx_xlock(&allprison_lock);
4600 prison_racct_free_locked(prr);
4601 sx_xunlock(&allprison_lock);
4605 prison_racct_attach(struct prison *pr)
4607 struct prison_racct *prr;
4609 ASSERT_RACCT_ENABLED();
4610 sx_assert(&allprison_lock, SA_XLOCKED);
4612 prr = prison_racct_find_locked(pr->pr_name);
4613 KASSERT(prr != NULL, ("cannot find prison_racct"));
4615 pr->pr_prison_racct = prr;
4619 * Handle jail renaming. From the racct point of view, renaming means
4620 * moving from one prison_racct to another.
4623 prison_racct_modify(struct prison *pr)
4627 struct prison_racct *oldprr;
4629 ASSERT_RACCT_ENABLED();
4631 sx_slock(&allproc_lock);
4632 sx_xlock(&allprison_lock);
4634 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4635 sx_xunlock(&allprison_lock);
4636 sx_sunlock(&allproc_lock);
4640 oldprr = pr->pr_prison_racct;
4641 pr->pr_prison_racct = NULL;
4643 prison_racct_attach(pr);
4646 * Move resource utilisation records.
4648 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4651 * Force rctl to reattach rules to processes.
4653 FOREACH_PROC_IN_SYSTEM(p) {
4655 cred = crhold(p->p_ucred);
4657 racct_proc_ucred_changed(p, cred, cred);
4661 sx_sunlock(&allproc_lock);
4662 prison_racct_free_locked(oldprr);
4663 sx_xunlock(&allprison_lock);
4667 prison_racct_detach(struct prison *pr)
4670 ASSERT_RACCT_ENABLED();
4671 sx_assert(&allprison_lock, SA_UNLOCKED);
4673 if (pr->pr_prison_racct == NULL)
4675 prison_racct_free(pr->pr_prison_racct);
4676 pr->pr_prison_racct = NULL;
4683 db_show_prison(struct prison *pr)
4686 #if defined(INET) || defined(INET6)
4691 char ip6buf[INET6_ADDRSTRLEN];
4694 db_printf("prison %p:\n", pr);
4695 db_printf(" jid = %d\n", pr->pr_id);
4696 db_printf(" name = %s\n", pr->pr_name);
4697 db_printf(" parent = %p\n", pr->pr_parent);
4698 db_printf(" ref = %d\n", pr->pr_ref);
4699 db_printf(" uref = %d\n", pr->pr_uref);
4700 db_printf(" path = %s\n", pr->pr_path);
4701 db_printf(" cpuset = %d\n", pr->pr_cpuset
4702 ? pr->pr_cpuset->cs_id : -1);
4704 db_printf(" vnet = %p\n", pr->pr_vnet);
4706 db_printf(" root = %p\n", pr->pr_root);
4707 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4708 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4709 db_printf(" children.max = %d\n", pr->pr_childmax);
4710 db_printf(" children.cur = %d\n", pr->pr_childcount);
4711 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4712 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4713 db_printf(" flags = 0x%x", pr->pr_flags);
4714 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4716 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4717 db_printf(" %s", pr_flag_names[fi]);
4718 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4720 jsf = pr->pr_flags &
4721 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4722 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4723 pr_flag_jailsys[fi].disable &&
4724 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4725 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4728 db_printf(" allow = 0x%x", pr->pr_allow);
4729 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4731 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4732 db_printf(" %s", pr_allow_names[fi]);
4734 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4735 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4736 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4737 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4738 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4740 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4741 for (ii = 0; ii < pr->pr_ip4s; ii++)
4742 db_printf(" %s %s\n",
4743 ii == 0 ? "ip4.addr =" : " ",
4744 inet_ntoa(pr->pr_ip4[ii]));
4747 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4748 for (ii = 0; ii < pr->pr_ip6s; ii++)
4749 db_printf(" %s %s\n",
4750 ii == 0 ? "ip6.addr =" : " ",
4751 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4755 DB_SHOW_COMMAND(prison, db_show_prison_command)
4761 * Show all prisons in the list, and prison0 which is not
4764 db_show_prison(&prison0);
4765 if (!db_pager_quit) {
4766 TAILQ_FOREACH(pr, &allprison, pr_list) {
4778 /* Look for a prison with the ID and with references. */
4779 TAILQ_FOREACH(pr, &allprison, pr_list)
4780 if (pr->pr_id == addr && pr->pr_ref > 0)
4783 /* Look again, without requiring a reference. */
4784 TAILQ_FOREACH(pr, &allprison, pr_list)
4785 if (pr->pr_id == addr)
4788 /* Assume address points to a valid prison. */
4789 pr = (struct prison *)addr;