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>
73 #include <security/mac/mac_framework.h>
75 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
77 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
78 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
80 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
83 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
85 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
89 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
91 #define _PR_IP_SADDRSEL 0
95 /* prison0 describes what is "real" about the system. */
96 struct prison prison0 = {
102 .pr_securelevel = -1,
104 .pr_childmax = JAIL_MAX,
105 .pr_hostuuid = DEFAULT_HOSTUUID,
106 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
108 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
110 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
112 .pr_allow = PR_ALLOW_ALL,
114 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
116 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
117 struct sx allprison_lock;
118 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
119 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
120 LIST_HEAD(, prison_racct) allprison_racct;
123 static int do_jail_attach(struct thread *td, struct prison *pr);
124 static void prison_complete(void *context, int pending);
125 static void prison_deref(struct prison *pr, int flags);
126 static char *prison_path(struct prison *pr1, struct prison *pr2);
127 static void prison_remove_one(struct prison *pr);
129 static void prison_racct_attach(struct prison *pr);
130 static void prison_racct_modify(struct prison *pr);
131 static void prison_racct_detach(struct prison *pr);
134 static int _prison_check_ip4(const struct prison *, const struct in_addr *);
135 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
138 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
139 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
142 /* Flags for prison_deref */
143 #define PD_DEREF 0x01
144 #define PD_DEUREF 0x02
145 #define PD_LOCKED 0x04
146 #define PD_LIST_SLOCKED 0x08
147 #define PD_LIST_XLOCKED 0x10
150 * Parameter names corresponding to PR_* flag values. Size values are for kvm
151 * as we cannot figure out the size of a sparse array, or an array without a
154 static char *pr_flag_names[] = {
157 [7] = "ip4.saddrsel",
160 [8] = "ip6.saddrsel",
163 const size_t pr_flag_names_size = sizeof(pr_flag_names);
165 static char *pr_flag_nonames[] = {
168 [7] = "ip4.nosaddrsel",
171 [8] = "ip6.nosaddrsel",
174 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
176 struct jailsys_flags {
180 } pr_flag_jailsys[] = {
181 { "host", 0, PR_HOST },
183 { "vnet", 0, PR_VNET },
186 { "ip4", PR_IP4_USER, PR_IP4_USER },
189 { "ip6", PR_IP6_USER, PR_IP6_USER },
192 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
194 static char *pr_allow_names[] = {
195 "allow.set_hostname",
203 "allow.mount.nullfs",
205 "allow.mount.procfs",
207 "allow.mount.fdescfs",
208 "allow.mount.linprocfs",
209 "allow.mount.linsysfs",
211 const size_t pr_allow_names_size = sizeof(pr_allow_names);
213 static char *pr_allow_nonames[] = {
214 "allow.noset_hostname",
216 "allow.noraw_sockets",
221 "allow.mount.nodevfs",
222 "allow.mount.nonullfs",
224 "allow.mount.noprocfs",
225 "allow.mount.notmpfs",
226 "allow.mount.nofdescfs",
227 "allow.mount.nolinprocfs",
228 "allow.mount.nolinsysfs",
230 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
232 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
233 #define JAIL_DEFAULT_ENFORCE_STATFS 2
234 #define JAIL_DEFAULT_DEVFS_RSNUM 0
235 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
236 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
237 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
238 #if defined(INET) || defined(INET6)
239 static unsigned jail_max_af_ips = 255;
243 * Initialize the parts of prison0 that can't be static-initialized with
244 * constants. This is called from proc0_init() after creating thread0 cpuset.
250 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
251 prison0.pr_osreldate = osreldate;
252 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
257 qcmp_v4(const void *ip1, const void *ip2)
262 * We need to compare in HBO here to get the list sorted as expected
263 * by the result of the code. Sorting NBO addresses gives you
264 * interesting results. If you do not understand, do not try.
266 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
267 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
270 * Do not simply return the difference of the two numbers, the int is
284 qcmp_v6(const void *ip1, const void *ip2)
286 const struct in6_addr *ia6a, *ia6b;
289 ia6a = (const struct in6_addr *)ip1;
290 ia6b = (const struct in6_addr *)ip2;
293 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
294 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
296 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
309 sys_jail(struct thread *td, struct jail_args *uap)
315 error = copyin(uap->jail, &version, sizeof(uint32_t));
324 /* FreeBSD single IPv4 jails. */
325 bzero(&j, sizeof(struct jail));
326 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
329 j.version = j0.version;
331 j.hostname = j0.hostname;
332 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
338 * Version 1 was used by multi-IPv4 jail implementations
339 * that never made it into the official kernel.
343 case 2: /* JAIL_API_VERSION */
344 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
345 error = copyin(uap->jail, &j, sizeof(struct jail));
351 /* Sci-Fi jails are not supported, sorry. */
354 return (kern_jail(td, &j));
358 kern_jail(struct thread *td, struct jail *j)
360 struct iovec optiov[2 * (4 + nitems(pr_allow_names)
369 char *u_path, *u_hostname, *u_name;
372 struct in_addr *u_ip4;
375 struct in6_addr *u_ip6;
378 int error, enforce_statfs, fi;
380 bzero(&optiov, sizeof(optiov));
381 opt.uio_iov = optiov;
385 opt.uio_segflg = UIO_SYSSPACE;
386 opt.uio_rw = UIO_READ;
389 /* Set permissions for top-level jails from sysctls. */
390 if (!jailed(td->td_ucred)) {
391 for (fi = 0; fi < nitems(pr_allow_names); fi++) {
392 optiov[opt.uio_iovcnt].iov_base =
393 (jail_default_allow & (1 << fi))
394 ? pr_allow_names[fi] : pr_allow_nonames[fi];
395 optiov[opt.uio_iovcnt].iov_len =
396 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
399 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
400 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
402 enforce_statfs = jail_default_enforce_statfs;
403 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
404 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
408 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
410 ip4s = (j->version == 0) ? 1 : j->ip4s;
411 if (ip4s > jail_max_af_ips)
413 tmplen += ip4s * sizeof(struct in_addr);
419 if (j->ip6s > jail_max_af_ips)
421 tmplen += j->ip6s * sizeof(struct in6_addr);
426 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
427 u_hostname = u_path + MAXPATHLEN;
428 u_name = u_hostname + MAXHOSTNAMELEN;
430 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
434 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
436 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
439 optiov[opt.uio_iovcnt].iov_base = "path";
440 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
442 optiov[opt.uio_iovcnt].iov_base = u_path;
443 error = copyinstr(j->path, u_path, MAXPATHLEN,
444 &optiov[opt.uio_iovcnt].iov_len);
446 free(u_path, M_TEMP);
450 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
451 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
453 optiov[opt.uio_iovcnt].iov_base = u_hostname;
454 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
455 &optiov[opt.uio_iovcnt].iov_len);
457 free(u_path, M_TEMP);
461 if (j->jailname != NULL) {
462 optiov[opt.uio_iovcnt].iov_base = "name";
463 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
465 optiov[opt.uio_iovcnt].iov_base = u_name;
466 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
467 &optiov[opt.uio_iovcnt].iov_len);
469 free(u_path, M_TEMP);
475 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
476 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
478 optiov[opt.uio_iovcnt].iov_base = u_ip4;
479 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
481 u_ip4->s_addr = j->ip4s;
483 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
485 free(u_path, M_TEMP);
492 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
493 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
495 optiov[opt.uio_iovcnt].iov_base = u_ip6;
496 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
497 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
499 free(u_path, M_TEMP);
504 KASSERT(opt.uio_iovcnt <= nitems(optiov),
505 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
506 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
507 free(u_path, M_TEMP);
513 * struct jail_set_args {
514 * struct iovec *iovp;
515 * unsigned int iovcnt;
520 sys_jail_set(struct thread *td, struct jail_set_args *uap)
525 /* Check that we have an even number of iovecs. */
529 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
532 error = kern_jail_set(td, auio, uap->flags);
538 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
545 struct in6_addr *ip6;
548 struct vfsoptlist *opts;
549 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
551 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
552 char *g_path, *osrelstr;
553 #if defined(INET) || defined(INET6)
558 size_t namelen, onamelen, pnamelen;
559 int born, created, cuflags, descend, enforce;
560 int error, errmsg_len, errmsg_pos;
561 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
562 int fi, jid, jsys, len, level;
563 int childmax, osreldt, rsnum, slevel;
564 int fullpath_disabled;
565 #if defined(INET) || defined(INET6)
574 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
578 error = priv_check(td, PRIV_JAIL_SET);
579 if (!error && (flags & JAIL_ATTACH))
580 error = priv_check(td, PRIV_JAIL_ATTACH);
583 mypr = td->td_ucred->cr_prison;
584 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
586 if (flags & ~JAIL_SET_MASK)
590 * Check all the parameters before committing to anything. Not all
591 * errors can be caught early, but we may as well try. Also, this
592 * takes care of some expensive stuff (path lookup) before getting
593 * the allprison lock.
595 * XXX Jails are not filesystems, and jail parameters are not mount
596 * options. But it makes more sense to re-use the vfsopt code
597 * than duplicate it under a different name.
599 error = vfs_buildopts(optuio, &opts);
610 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
613 vfs_opterror(opts, "no valid operation (create or update)");
617 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
623 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
632 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
640 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
645 else if (enforce < 0 || enforce > 2) {
651 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
659 pr_flags = ch_flags = 0;
660 for (fi = 0; fi < nitems(pr_flag_names); fi++) {
661 if (pr_flag_names[fi] == NULL)
663 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
664 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
666 ch_flags |= pr_flags;
667 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
668 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
675 case JAIL_SYS_DISABLE:
676 if (!pr_flag_jailsys[fi].disable) {
680 pr_flags |= pr_flag_jailsys[fi].disable;
683 pr_flags |= pr_flag_jailsys[fi].new;
685 case JAIL_SYS_INHERIT:
692 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
694 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
695 && !(pr_flags & PR_PERSIST)) {
697 vfs_opterror(opts, "new jail must persist or attach");
701 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
703 vfs_opterror(opts, "vnet cannot be changed after creation");
708 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
710 vfs_opterror(opts, "ip4 cannot be changed after creation");
715 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
717 vfs_opterror(opts, "ip6 cannot be changed after creation");
722 pr_allow = ch_allow = 0;
723 for (fi = 0; fi < nitems(pr_allow_names); fi++) {
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;
829 pr_flags |= PR_IP4_USER;
831 ip4s /= sizeof(*ip4);
832 if (ip4s > jail_max_af_ips) {
834 vfs_opterror(opts, "too many IPv4 addresses");
837 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
838 bcopy(op, ip4, ip4s * sizeof(*ip4));
840 * IP addresses are all sorted but ip[0] to preserve
841 * the primary IP address as given from userland.
842 * This special IP is used for unbound outgoing
843 * connections as well for "loopback" traffic in case
844 * source address selection cannot find any more fitting
845 * address to connect from.
848 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
850 * Check for duplicate addresses and do some simple
851 * zero and broadcast checks. If users give other bogus
852 * addresses it is their problem.
854 * We do not have to care about byte order for these
855 * checks so we will do them in NBO.
857 for (ii = 0; ii < ip4s; ii++) {
858 if (ip4[ii].s_addr == INADDR_ANY ||
859 ip4[ii].s_addr == INADDR_BROADCAST) {
864 (ip4[0].s_addr == ip4[ii+1].s_addr ||
865 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
875 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
880 else if (ip6s & (sizeof(*ip6) - 1)) {
884 ch_flags |= PR_IP6_USER;
885 pr_flags |= PR_IP6_USER;
887 ip6s /= sizeof(*ip6);
888 if (ip6s > jail_max_af_ips) {
890 vfs_opterror(opts, "too many IPv6 addresses");
893 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
894 bcopy(op, ip6, ip6s * sizeof(*ip6));
896 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
897 for (ii = 0; ii < ip6s; ii++) {
898 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
903 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
904 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
914 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
915 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
918 "vnet jails cannot have IP address restrictions");
923 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
929 if (flags & JAIL_UPDATE) {
932 "osrelease cannot be changed after creation");
935 if (len == 0 || len >= OSRELEASELEN) {
938 "osrelease string must be 1-%d bytes long",
944 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
950 if (flags & JAIL_UPDATE) {
953 "osreldate cannot be changed after creation");
958 vfs_opterror(opts, "osreldate cannot be 0");
963 fullpath_disabled = 0;
965 error = vfs_getopt(opts, "path", (void **)&path, &len);
971 if (flags & JAIL_UPDATE) {
974 "path cannot be changed after creation");
977 if (len == 0 || path[len - 1] != '\0') {
981 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
987 NDFREE(&nd, NDF_ONLY_PNBUF);
988 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
989 strlcpy(g_path, path, MAXPATHLEN);
990 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
993 else if (error == ENODEV) {
994 /* proceed if sysctl debug.disablefullpath == 1 */
995 fullpath_disabled = 1;
996 if (len < 2 || (len == 2 && path[0] == '/'))
999 /* exit on other errors */
1002 if (root->v_type != VDIR) {
1007 VOP_UNLOCK(root, 0);
1008 if (fullpath_disabled) {
1009 /* Leave room for a real-root full pathname. */
1010 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
1011 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
1012 error = ENAMETOOLONG;
1019 * Find the specified jail, or at least its parent.
1020 * This abuses the file error codes ENOENT and EEXIST.
1024 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1025 namelc = strrchr(name, '.');
1026 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1030 sx_xlock(&allprison_lock);
1033 * See if a requested jid already exists. There is an
1034 * information leak here if the jid exists but is not within
1035 * the caller's jail hierarchy. Jail creators will get EEXIST
1036 * even though they cannot see the jail, and CREATE | UPDATE
1037 * will return ENOENT which is not normally a valid error.
1041 vfs_opterror(opts, "negative jid");
1042 goto done_unlock_list;
1044 pr = prison_find(jid);
1046 ppr = pr->pr_parent;
1047 /* Create: jid must not exist. */
1048 if (cuflags == JAIL_CREATE) {
1049 mtx_unlock(&pr->pr_mtx);
1051 vfs_opterror(opts, "jail %d already exists",
1053 goto done_unlock_list;
1055 if (!prison_ischild(mypr, pr)) {
1056 mtx_unlock(&pr->pr_mtx);
1058 } else if (pr->pr_uref == 0) {
1059 if (!(flags & JAIL_DYING)) {
1060 mtx_unlock(&pr->pr_mtx);
1062 vfs_opterror(opts, "jail %d is dying",
1064 goto done_unlock_list;
1065 } else if ((flags & JAIL_ATTACH) ||
1066 (pr_flags & PR_PERSIST)) {
1068 * A dying jail might be resurrected
1069 * (via attach or persist), but first
1070 * it must determine if another jail
1071 * has claimed its name. Accomplish
1072 * this by implicitly re-setting the
1076 name = prison_name(mypr, pr);
1081 /* Update: jid must exist. */
1082 if (cuflags == JAIL_UPDATE) {
1084 vfs_opterror(opts, "jail %d not found", jid);
1085 goto done_unlock_list;
1090 * If the caller provided a name, look for a jail by that name.
1091 * This has different semantics for creates and updates keyed by jid
1092 * (where the name must not already exist in a different jail),
1093 * and updates keyed by the name itself (where the name must exist
1094 * because that is the jail being updated).
1098 namelc = strrchr(name, '.');
1103 * This is a hierarchical name. Split it into the
1104 * parent and child names, and make sure the parent
1105 * exists or matches an already found jail.
1108 if (strncmp(name, ppr->pr_name, namelc - name)
1109 || ppr->pr_name[namelc - name] != '\0') {
1110 mtx_unlock(&pr->pr_mtx);
1113 "cannot change jail's parent");
1114 goto done_unlock_list;
1118 ppr = prison_find_name(mypr, name);
1122 "jail \"%s\" not found", name);
1123 goto done_unlock_list;
1125 mtx_unlock(&ppr->pr_mtx);
1130 if (namelc[0] != '\0') {
1132 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1135 FOREACH_PRISON_CHILD(ppr, tpr) {
1136 if (tpr != pr && tpr->pr_ref > 0 &&
1137 !strcmp(tpr->pr_name + pnamelen, namelc)) {
1139 cuflags != JAIL_CREATE) {
1140 mtx_lock(&tpr->pr_mtx);
1141 if (tpr->pr_ref > 0) {
1146 if (tpr->pr_uref > 0) {
1152 mtx_unlock(&tpr->pr_mtx);
1153 } else if (tpr->pr_uref > 0) {
1155 * Create, or update(jid):
1156 * name must not exist in an
1157 * active sibling jail.
1161 mtx_unlock(&pr->pr_mtx);
1163 "jail \"%s\" already exists",
1165 goto done_unlock_list;
1169 /* If no active jail is found, use a dying one. */
1170 if (deadpr != NULL && pr == NULL) {
1171 if (flags & JAIL_DYING) {
1172 mtx_lock(&deadpr->pr_mtx);
1173 if (deadpr->pr_ref == 0) {
1174 mtx_unlock(&deadpr->pr_mtx);
1178 } else if (cuflags == JAIL_UPDATE) {
1181 "jail \"%s\" is dying", name);
1182 goto done_unlock_list;
1185 /* Update: name must exist if no jid. */
1186 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1188 vfs_opterror(opts, "jail \"%s\" not found",
1190 goto done_unlock_list;
1194 /* Update: must provide a jid or name. */
1195 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1197 vfs_opterror(opts, "update specified no jail");
1198 goto done_unlock_list;
1201 /* If there's no prison to update, create a new one and link it in. */
1203 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1204 if (tpr->pr_childcount >= tpr->pr_childmax) {
1206 vfs_opterror(opts, "prison limit exceeded");
1207 goto done_unlock_list;
1210 mtx_lock(&ppr->pr_mtx);
1211 if (ppr->pr_ref == 0) {
1212 mtx_unlock(&ppr->pr_mtx);
1214 vfs_opterror(opts, "jail \"%s\" not found",
1215 prison_name(mypr, ppr));
1216 goto done_unlock_list;
1220 mtx_unlock(&ppr->pr_mtx);
1221 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1223 /* Find the next free jid. */
1226 if (jid == JAIL_MAX)
1228 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1229 if (tpr->pr_id < jid)
1231 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1232 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1235 if (jid == lastprid) {
1238 "no available jail IDs");
1240 prison_deref(ppr, PD_DEREF |
1241 PD_DEUREF | PD_LIST_XLOCKED);
1250 * The jail already has a jid (that did not yet exist),
1251 * so just find where to insert it.
1253 TAILQ_FOREACH(tpr, &allprison, pr_list)
1254 if (tpr->pr_id >= jid) {
1255 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1260 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1261 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1262 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1263 tpr->pr_childcount++;
1265 pr->pr_parent = ppr;
1268 /* Set some default values, and inherit some from the parent. */
1273 root = mypr->pr_root;
1276 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1277 pr->pr_flags |= PR_HOST;
1278 #if defined(INET) || defined(INET6)
1280 if (!(pr_flags & PR_VNET))
1284 if (!(ch_flags & PR_IP4_USER))
1285 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1286 else if (!(pr_flags & PR_IP4_USER)) {
1287 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1288 if (ppr->pr_ip4 != NULL) {
1289 pr->pr_ip4s = ppr->pr_ip4s;
1290 pr->pr_ip4 = malloc(pr->pr_ip4s *
1291 sizeof(struct in_addr), M_PRISON,
1293 bcopy(ppr->pr_ip4, pr->pr_ip4,
1294 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1299 if (!(ch_flags & PR_IP6_USER))
1300 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1301 else if (!(pr_flags & PR_IP6_USER)) {
1302 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1303 if (ppr->pr_ip6 != NULL) {
1304 pr->pr_ip6s = ppr->pr_ip6s;
1305 pr->pr_ip6 = malloc(pr->pr_ip6s *
1306 sizeof(struct in6_addr), M_PRISON,
1308 bcopy(ppr->pr_ip6, pr->pr_ip6,
1309 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1315 /* Source address selection is always on by default. */
1316 pr->pr_flags |= _PR_IP_SADDRSEL;
1318 pr->pr_securelevel = ppr->pr_securelevel;
1319 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1320 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1321 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1323 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1324 if (osrelstr == NULL)
1325 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1327 strcpy(pr->pr_osrelease, osrelstr);
1329 LIST_INIT(&pr->pr_children);
1330 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1331 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1334 /* Allocate a new vnet if specified. */
1335 pr->pr_vnet = (pr_flags & PR_VNET)
1336 ? vnet_alloc() : ppr->pr_vnet;
1339 * Allocate a dedicated cpuset for each jail.
1340 * Unlike other initial settings, this may return an erorr.
1342 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1344 prison_deref(pr, PD_LIST_XLOCKED);
1348 mtx_lock(&pr->pr_mtx);
1350 * New prisons do not yet have a reference, because we do not
1351 * want others to see the incomplete prison once the
1352 * allprison_lock is downgraded.
1357 * Grab a reference for existing prisons, to ensure they
1358 * continue to exist for the duration of the call.
1361 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1362 if ((pr->pr_flags & PR_VNET) &&
1363 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1366 "vnet jails cannot have IP address restrictions");
1367 goto done_deref_locked;
1371 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1374 "ip4 cannot be changed after creation");
1375 goto done_deref_locked;
1379 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1382 "ip6 cannot be changed after creation");
1383 goto done_deref_locked;
1388 /* Do final error checking before setting anything. */
1390 if (slevel < ppr->pr_securelevel) {
1392 goto done_deref_locked;
1396 if (childmax >= ppr->pr_childmax) {
1398 goto done_deref_locked;
1402 if (enforce < ppr->pr_enforce_statfs) {
1404 goto done_deref_locked;
1409 * devfs_rsnum is a uint16_t
1411 if (rsnum < 0 || rsnum > 65535) {
1413 goto done_deref_locked;
1416 * Nested jails always inherit parent's devfs ruleset
1418 if (jailed(td->td_ucred)) {
1419 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1421 goto done_deref_locked;
1423 rsnum = ppr->pr_devfs_rsnum;
1428 if (ppr->pr_flags & PR_IP4) {
1430 * Make sure the new set of IP addresses is a
1431 * subset of the parent's list. Don't worry
1432 * about the parent being unlocked, as any
1433 * setting is done with allprison_lock held.
1435 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1436 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1438 if (ij == ppr->pr_ip4s) {
1440 goto done_deref_locked;
1443 for (ii = ij = 1; ii < ip4s; ii++) {
1444 if (ip4[ii].s_addr ==
1445 ppr->pr_ip4[0].s_addr)
1447 for (; ij < ppr->pr_ip4s; ij++)
1448 if (ip4[ii].s_addr ==
1449 ppr->pr_ip4[ij].s_addr)
1451 if (ij == ppr->pr_ip4s)
1454 if (ij == ppr->pr_ip4s) {
1456 goto done_deref_locked;
1461 * Check for conflicting IP addresses. We permit them
1462 * if there is no more than one IP on each jail. If
1463 * there is a duplicate on a jail with more than one
1464 * IP stop checking and return error.
1468 for (; tppr != &prison0; tppr = tppr->pr_parent)
1469 if (tppr->pr_flags & PR_VNET)
1472 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1475 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1477 tpr->pr_uref == 0) {
1481 if (!(tpr->pr_flags & PR_IP4_USER))
1484 if (tpr->pr_ip4 == NULL ||
1485 (ip4s == 1 && tpr->pr_ip4s == 1))
1487 for (ii = 0; ii < ip4s; ii++) {
1488 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1491 "IPv4 addresses clash");
1492 goto done_deref_locked;
1500 if (ppr->pr_flags & PR_IP6) {
1502 * Make sure the new set of IP addresses is a
1503 * subset of the parent's list.
1505 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1506 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1509 if (ij == ppr->pr_ip6s) {
1511 goto done_deref_locked;
1514 for (ii = ij = 1; ii < ip6s; ii++) {
1515 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1518 for (; ij < ppr->pr_ip6s; ij++)
1519 if (IN6_ARE_ADDR_EQUAL(
1520 &ip6[ii], &ppr->pr_ip6[ij]))
1522 if (ij == ppr->pr_ip6s)
1525 if (ij == ppr->pr_ip6s) {
1527 goto done_deref_locked;
1531 /* Check for conflicting IP addresses. */
1534 for (; tppr != &prison0; tppr = tppr->pr_parent)
1535 if (tppr->pr_flags & PR_VNET)
1538 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1541 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1543 tpr->pr_uref == 0) {
1547 if (!(tpr->pr_flags & PR_IP6_USER))
1550 if (tpr->pr_ip6 == NULL ||
1551 (ip6s == 1 && tpr->pr_ip6s == 1))
1553 for (ii = 0; ii < ip6s; ii++) {
1554 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1557 "IPv6 addresses clash");
1558 goto done_deref_locked;
1564 onamelen = namelen = 0;
1565 if (namelc != NULL) {
1566 /* Give a default name of the jid. Also allow the name to be
1567 * explicitly the jid - but not any other number, and only in
1568 * normal form (no leading zero/etc).
1570 if (namelc[0] == '\0')
1571 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1572 else if ((strtoul(namelc, &p, 10) != jid ||
1573 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1576 "name cannot be numeric (unless it is the jid)");
1577 goto done_deref_locked;
1580 * Make sure the name isn't too long for the prison or its
1583 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1584 onamelen = strlen(pr->pr_name + pnamelen);
1585 namelen = strlen(namelc);
1586 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1587 error = ENAMETOOLONG;
1588 goto done_deref_locked;
1590 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1591 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1592 sizeof(pr->pr_name)) {
1593 error = ENAMETOOLONG;
1594 goto done_deref_locked;
1598 if (pr_allow & ~ppr->pr_allow) {
1600 goto done_deref_locked;
1604 * Let modules check their parameters. This requires unlocking and
1605 * then re-locking the prison, but this is still a valid state as long
1606 * as allprison_lock remains xlocked.
1608 mtx_unlock(&pr->pr_mtx);
1609 error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1611 prison_deref(pr, created
1613 : PD_DEREF | PD_LIST_XLOCKED);
1616 mtx_lock(&pr->pr_mtx);
1618 /* At this point, all valid parameters should have been noted. */
1619 TAILQ_FOREACH(opt, opts, link) {
1620 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1622 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1623 goto done_deref_locked;
1627 /* Set the parameters of the prison. */
1630 if (pr_flags & PR_IP4_USER) {
1631 pr->pr_flags |= PR_IP4;
1632 free(pr->pr_ip4, M_PRISON);
1636 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1638 if (tpr->pr_flags & PR_VNET) {
1643 if (prison_restrict_ip4(tpr, NULL)) {
1652 if (pr_flags & PR_IP6_USER) {
1653 pr->pr_flags |= PR_IP6;
1654 free(pr->pr_ip6, M_PRISON);
1658 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1660 if (tpr->pr_flags & PR_VNET) {
1665 if (prison_restrict_ip6(tpr, NULL)) {
1673 pr->pr_securelevel = slevel;
1674 /* Set all child jails to be at least this level. */
1675 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1676 if (tpr->pr_securelevel < slevel)
1677 tpr->pr_securelevel = slevel;
1680 pr->pr_childmax = childmax;
1681 /* Set all child jails to under this limit. */
1682 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1683 if (tpr->pr_childmax > childmax - level)
1684 tpr->pr_childmax = childmax > level
1685 ? childmax - level : 0;
1688 pr->pr_enforce_statfs = enforce;
1689 /* Pass this restriction on to the children. */
1690 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1691 if (tpr->pr_enforce_statfs < enforce)
1692 tpr->pr_enforce_statfs = enforce;
1695 pr->pr_devfs_rsnum = rsnum;
1696 /* Pass this restriction on to the children. */
1697 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1698 tpr->pr_devfs_rsnum = rsnum;
1700 if (namelc != NULL) {
1701 if (ppr == &prison0)
1702 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1704 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1705 ppr->pr_name, namelc);
1706 /* Change this component of child names. */
1707 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1708 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1709 strlen(tpr->pr_name + onamelen) + 1);
1710 bcopy(pr->pr_name, tpr->pr_name, namelen);
1714 /* Try to keep a real-rooted full pathname. */
1715 if (fullpath_disabled && path[0] == '/' &&
1716 strcmp(mypr->pr_path, "/"))
1717 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1718 mypr->pr_path, path);
1720 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1723 if (PR_HOST & ch_flags & ~pr_flags) {
1724 if (pr->pr_flags & PR_HOST) {
1726 * Copy the parent's host info. As with pr_ip4 above,
1727 * the lack of a lock on the parent is not a problem;
1728 * it is always set with allprison_lock at least
1729 * shared, and is held exclusively here.
1731 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1732 sizeof(pr->pr_hostname));
1733 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1734 sizeof(pr->pr_domainname));
1735 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1736 sizeof(pr->pr_hostuuid));
1737 pr->pr_hostid = pr->pr_parent->pr_hostid;
1739 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1740 /* Set this prison, and any descendants without PR_HOST. */
1742 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1744 strlcpy(pr->pr_domainname, domain,
1745 sizeof(pr->pr_domainname));
1747 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1749 pr->pr_hostid = hid;
1750 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1751 if (tpr->pr_flags & PR_HOST)
1755 strlcpy(tpr->pr_hostname,
1757 sizeof(tpr->pr_hostname));
1759 strlcpy(tpr->pr_domainname,
1761 sizeof(tpr->pr_domainname));
1763 strlcpy(tpr->pr_hostuuid,
1765 sizeof(tpr->pr_hostuuid));
1767 tpr->pr_hostid = hid;
1771 if ((tallow = ch_allow & ~pr_allow)) {
1772 /* Clear allow bits in all children. */
1773 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1774 tpr->pr_allow &= ~tallow;
1776 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1778 * Persistent prisons get an extra reference, and prisons losing their
1779 * persist flag lose that reference. Only do this for existing prisons
1780 * for now, so new ones will remain unseen until after the module
1781 * handlers have completed.
1783 born = pr->pr_uref == 0;
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 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1857 prison_deref(pr, created
1859 : PD_DEREF | PD_LIST_SLOCKED);
1863 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1866 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1867 prison_deref(pr, created
1869 : PD_DEREF | PD_LIST_SLOCKED);
1873 /* Attach this process to the prison if requested. */
1874 if (flags & JAIL_ATTACH) {
1875 mtx_lock(&pr->pr_mtx);
1876 error = do_jail_attach(td, pr);
1878 vfs_opterror(opts, "attach failed");
1880 prison_deref(pr, PD_DEREF);
1886 if (racct_enable && !created) {
1887 if (!(flags & JAIL_ATTACH))
1888 sx_sunlock(&allprison_lock);
1889 prison_racct_modify(pr);
1890 if (!(flags & JAIL_ATTACH))
1891 sx_slock(&allprison_lock);
1895 td->td_retval[0] = pr->pr_id;
1898 * Now that it is all there, drop the temporary reference from existing
1899 * prisons. Or add a reference to newly created persistent prisons
1900 * (which was not done earlier so that the prison would not be publicly
1904 prison_deref(pr, (flags & JAIL_ATTACH)
1906 : PD_DEREF | PD_LIST_SLOCKED);
1908 if (pr_flags & PR_PERSIST) {
1909 mtx_lock(&pr->pr_mtx);
1912 mtx_unlock(&pr->pr_mtx);
1914 if (!(flags & JAIL_ATTACH))
1915 sx_sunlock(&allprison_lock);
1921 prison_deref(pr, created
1922 ? PD_LOCKED | PD_LIST_XLOCKED
1923 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1926 sx_xunlock(&allprison_lock);
1932 if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1933 &errmsg_len) == 0 && errmsg_len > 0) {
1934 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1935 if (optuio->uio_segflg == UIO_SYSSPACE)
1937 optuio->uio_iov[errmsg_pos].iov_base,
1941 optuio->uio_iov[errmsg_pos].iov_base,
1947 free(ip4, M_PRISON);
1950 free(ip6, M_PRISON);
1953 free(g_path, M_TEMP);
1960 * struct jail_get_args {
1961 * struct iovec *iovp;
1962 * unsigned int iovcnt;
1967 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1972 /* Check that we have an even number of iovecs. */
1973 if (uap->iovcnt & 1)
1976 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1979 error = kern_jail_get(td, auio, uap->flags);
1981 error = copyout(auio->uio_iov, uap->iovp,
1982 uap->iovcnt * sizeof (struct iovec));
1988 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1990 struct prison *pr, *mypr;
1992 struct vfsoptlist *opts;
1993 char *errmsg, *name;
1994 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1996 if (flags & ~JAIL_GET_MASK)
1999 /* Get the parameter list. */
2000 error = vfs_buildopts(optuio, &opts);
2003 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
2004 mypr = td->td_ucred->cr_prison;
2007 * Find the prison specified by one of: lastjid, jid, name.
2009 sx_slock(&allprison_lock);
2010 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2012 TAILQ_FOREACH(pr, &allprison, pr_list) {
2013 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2014 mtx_lock(&pr->pr_mtx);
2015 if (pr->pr_ref > 0 &&
2016 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2018 mtx_unlock(&pr->pr_mtx);
2024 vfs_opterror(opts, "no jail after %d", jid);
2025 goto done_unlock_list;
2026 } else if (error != ENOENT)
2027 goto done_unlock_list;
2029 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2032 pr = prison_find_child(mypr, jid);
2034 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2035 mtx_unlock(&pr->pr_mtx);
2037 vfs_opterror(opts, "jail %d is dying",
2039 goto done_unlock_list;
2044 vfs_opterror(opts, "jail %d not found", jid);
2045 goto done_unlock_list;
2047 } else if (error != ENOENT)
2048 goto done_unlock_list;
2050 error = vfs_getopt(opts, "name", (void **)&name, &len);
2052 if (len == 0 || name[len - 1] != '\0') {
2054 goto done_unlock_list;
2056 pr = prison_find_name(mypr, name);
2058 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2059 mtx_unlock(&pr->pr_mtx);
2061 vfs_opterror(opts, "jail \"%s\" is dying",
2063 goto done_unlock_list;
2068 vfs_opterror(opts, "jail \"%s\" not found", name);
2069 goto done_unlock_list;
2070 } else if (error != ENOENT)
2071 goto done_unlock_list;
2073 vfs_opterror(opts, "no jail specified");
2075 goto done_unlock_list;
2078 /* Get the parameters of the prison. */
2081 td->td_retval[0] = pr->pr_id;
2082 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2083 if (error != 0 && error != ENOENT)
2085 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2086 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2087 if (error != 0 && error != ENOENT)
2089 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2090 if (error != 0 && error != ENOENT)
2092 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2093 sizeof(pr->pr_cpuset->cs_id));
2094 if (error != 0 && error != ENOENT)
2096 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2097 if (error != 0 && error != ENOENT)
2100 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2101 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2102 if (error != 0 && error != ENOENT)
2106 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2107 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2108 if (error != 0 && error != ENOENT)
2111 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2112 sizeof(pr->pr_securelevel));
2113 if (error != 0 && error != ENOENT)
2115 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2116 sizeof(pr->pr_childcount));
2117 if (error != 0 && error != ENOENT)
2119 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2120 sizeof(pr->pr_childmax));
2121 if (error != 0 && error != ENOENT)
2123 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2124 if (error != 0 && error != ENOENT)
2126 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2127 if (error != 0 && error != ENOENT)
2129 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2130 if (error != 0 && error != ENOENT)
2132 #ifdef COMPAT_FREEBSD32
2133 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2134 uint32_t hid32 = pr->pr_hostid;
2136 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2139 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2140 sizeof(pr->pr_hostid));
2141 if (error != 0 && error != ENOENT)
2143 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2144 sizeof(pr->pr_enforce_statfs));
2145 if (error != 0 && error != ENOENT)
2147 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2148 sizeof(pr->pr_devfs_rsnum));
2149 if (error != 0 && error != ENOENT)
2151 for (fi = 0; fi < nitems(pr_flag_names); fi++) {
2152 if (pr_flag_names[fi] == NULL)
2154 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2155 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2156 if (error != 0 && error != ENOENT)
2159 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2160 if (error != 0 && error != ENOENT)
2163 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
2165 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2166 i = pr_flag_jailsys[fi].disable &&
2167 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2168 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2171 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2172 if (error != 0 && error != ENOENT)
2175 for (fi = 0; fi < nitems(pr_allow_names); fi++) {
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 if (!LIST_EMPTY(&pr->pr_children)) {
2294 mtx_unlock(&pr->pr_mtx);
2296 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2297 mtx_lock(&cpr->pr_mtx);
2298 if (cpr->pr_ref > 0) {
2302 /* Already removed - do not do it again. */
2305 mtx_unlock(&cpr->pr_mtx);
2307 mtx_lock(&lpr->pr_mtx);
2308 prison_remove_one(lpr);
2309 sx_xlock(&allprison_lock);
2314 mtx_lock(&lpr->pr_mtx);
2315 prison_remove_one(lpr);
2316 sx_xlock(&allprison_lock);
2318 mtx_lock(&pr->pr_mtx);
2320 prison_remove_one(pr);
2325 prison_remove_one(struct prison *pr)
2330 /* If the prison was persistent, it is not anymore. */
2332 if (pr->pr_flags & PR_PERSIST) {
2335 pr->pr_flags &= ~PR_PERSIST;
2339 * jail_remove added a reference. If that's the only one, remove
2342 KASSERT(pr->pr_ref > 0,
2343 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2344 if (pr->pr_ref == 1) {
2346 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2350 mtx_unlock(&pr->pr_mtx);
2351 sx_xunlock(&allprison_lock);
2353 * Kill all processes unfortunate enough to be attached to this prison.
2355 sx_slock(&allproc_lock);
2356 LIST_FOREACH(p, &allproc, p_list) {
2358 if (p->p_state != PRS_NEW && p->p_ucred &&
2359 p->p_ucred->cr_prison == pr)
2360 kern_psignal(p, SIGKILL);
2363 sx_sunlock(&allproc_lock);
2364 /* Remove the temporary reference added by jail_remove. */
2365 prison_deref(pr, deuref | PD_DEREF);
2370 * struct jail_attach_args {
2375 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2380 error = priv_check(td, PRIV_JAIL_ATTACH);
2385 * Start with exclusive hold on allprison_lock to ensure that a possible
2386 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2387 * But then immediately downgrade it since we don't need to stop
2390 sx_xlock(&allprison_lock);
2391 sx_downgrade(&allprison_lock);
2392 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2394 sx_sunlock(&allprison_lock);
2399 * Do not allow a process to attach to a prison that is not
2400 * considered to be "alive".
2402 if (pr->pr_uref == 0) {
2403 mtx_unlock(&pr->pr_mtx);
2404 sx_sunlock(&allprison_lock);
2408 return (do_jail_attach(td, pr));
2412 do_jail_attach(struct thread *td, struct prison *pr)
2415 struct ucred *newcred, *oldcred;
2419 * XXX: Note that there is a slight race here if two threads
2420 * in the same privileged process attempt to attach to two
2421 * different jails at the same time. It is important for
2422 * user processes not to do this, or they might end up with
2423 * a process root from one prison, but attached to the jail
2428 mtx_unlock(&pr->pr_mtx);
2430 /* Let modules do whatever they need to prepare for attaching. */
2431 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2433 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2436 sx_sunlock(&allprison_lock);
2439 * Reparent the newly attached process to this jail.
2442 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2446 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2447 if ((error = change_dir(pr->pr_root, td)) != 0)
2450 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2453 VOP_UNLOCK(pr->pr_root, 0);
2454 if ((error = pwd_chroot(td, pr->pr_root)))
2459 oldcred = crcopysafe(p, newcred);
2460 newcred->cr_prison = pr;
2461 proc_set_cred(p, newcred);
2465 racct_proc_ucred_changed(p, oldcred, newcred);
2467 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2472 VOP_UNLOCK(pr->pr_root, 0);
2474 /* Tell modules this thread is still in its old jail after all. */
2475 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2476 prison_deref(pr, PD_DEREF | PD_DEUREF);
2482 * Returns a locked prison instance, or NULL on failure.
2485 prison_find(int prid)
2489 sx_assert(&allprison_lock, SX_LOCKED);
2490 TAILQ_FOREACH(pr, &allprison, pr_list) {
2491 if (pr->pr_id == prid) {
2492 mtx_lock(&pr->pr_mtx);
2495 mtx_unlock(&pr->pr_mtx);
2502 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2505 prison_find_child(struct prison *mypr, int prid)
2510 sx_assert(&allprison_lock, SX_LOCKED);
2511 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2512 if (pr->pr_id == prid) {
2513 mtx_lock(&pr->pr_mtx);
2516 mtx_unlock(&pr->pr_mtx);
2523 * Look for the name relative to mypr. Returns a locked prison or NULL.
2526 prison_find_name(struct prison *mypr, const char *name)
2528 struct prison *pr, *deadpr;
2532 sx_assert(&allprison_lock, SX_LOCKED);
2533 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2536 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2537 if (!strcmp(pr->pr_name + mylen, name)) {
2538 mtx_lock(&pr->pr_mtx);
2539 if (pr->pr_ref > 0) {
2540 if (pr->pr_uref > 0)
2544 mtx_unlock(&pr->pr_mtx);
2547 /* There was no valid prison - perhaps there was a dying one. */
2548 if (deadpr != NULL) {
2549 mtx_lock(&deadpr->pr_mtx);
2550 if (deadpr->pr_ref == 0) {
2551 mtx_unlock(&deadpr->pr_mtx);
2559 * See if a prison has the specific flag set.
2562 prison_flag(struct ucred *cred, unsigned flag)
2565 /* This is an atomic read, so no locking is necessary. */
2566 return (cred->cr_prison->pr_flags & flag);
2570 prison_allow(struct ucred *cred, unsigned flag)
2573 /* This is an atomic read, so no locking is necessary. */
2574 return (cred->cr_prison->pr_allow & flag);
2578 * Remove a prison reference. If that was the last reference, remove the
2579 * prison itself - but not in this context in case there are locks held.
2582 prison_free_locked(struct prison *pr)
2586 mtx_assert(&pr->pr_mtx, MA_OWNED);
2588 mtx_unlock(&pr->pr_mtx);
2590 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2594 prison_free(struct prison *pr)
2597 mtx_lock(&pr->pr_mtx);
2598 prison_free_locked(pr);
2602 * Complete a call to either prison_free or prison_proc_free.
2605 prison_complete(void *context, int pending)
2607 struct prison *pr = context;
2609 sx_xlock(&allprison_lock);
2610 mtx_lock(&pr->pr_mtx);
2611 prison_deref(pr, pr->pr_uref
2612 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2613 : PD_LOCKED | PD_LIST_XLOCKED);
2617 * Remove a prison reference (usually). This internal version assumes no
2618 * mutexes are held, except perhaps the prison itself. If there are no more
2619 * references, release and delist the prison. On completion, the prison lock
2620 * and the allprison lock are both unlocked.
2623 prison_deref(struct prison *pr, int flags)
2625 struct prison *ppr, *tpr;
2628 if (!(flags & PD_LOCKED))
2629 mtx_lock(&pr->pr_mtx);
2631 if (flags & PD_DEUREF) {
2632 KASSERT(pr->pr_uref > 0,
2633 ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2636 lasturef = pr->pr_uref == 0;
2639 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2642 if (flags & PD_DEREF) {
2643 KASSERT(pr->pr_ref > 0,
2644 ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2649 mtx_unlock(&pr->pr_mtx);
2652 * Tell the modules if the last user reference was removed
2653 * (even it sticks around in dying state).
2656 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2657 sx_xlock(&allprison_lock);
2658 flags |= PD_LIST_XLOCKED;
2660 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2661 mtx_lock(&pr->pr_mtx);
2663 mtx_unlock(&pr->pr_mtx);
2666 /* If the prison still has references, nothing else to do. */
2668 if (flags & PD_LIST_SLOCKED)
2669 sx_sunlock(&allprison_lock);
2670 else if (flags & PD_LIST_XLOCKED)
2671 sx_xunlock(&allprison_lock);
2675 if (flags & PD_LIST_SLOCKED) {
2676 if (!sx_try_upgrade(&allprison_lock)) {
2677 sx_sunlock(&allprison_lock);
2678 sx_xlock(&allprison_lock);
2680 } else if (!(flags & PD_LIST_XLOCKED))
2681 sx_xlock(&allprison_lock);
2683 TAILQ_REMOVE(&allprison, pr, pr_list);
2684 LIST_REMOVE(pr, pr_sibling);
2685 ppr = pr->pr_parent;
2686 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2687 tpr->pr_childcount--;
2688 sx_xunlock(&allprison_lock);
2691 if (pr->pr_vnet != ppr->pr_vnet)
2692 vnet_destroy(pr->pr_vnet);
2694 if (pr->pr_root != NULL)
2696 mtx_destroy(&pr->pr_mtx);
2698 free(pr->pr_ip4, M_PRISON);
2701 free(pr->pr_ip6, M_PRISON);
2703 if (pr->pr_cpuset != NULL)
2704 cpuset_rel(pr->pr_cpuset);
2708 prison_racct_detach(pr);
2712 /* Removing a prison frees a reference on its parent. */
2714 mtx_lock(&pr->pr_mtx);
2715 flags = PD_DEREF | PD_DEUREF;
2720 prison_hold_locked(struct prison *pr)
2723 mtx_assert(&pr->pr_mtx, MA_OWNED);
2724 KASSERT(pr->pr_ref > 0,
2725 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2730 prison_hold(struct prison *pr)
2733 mtx_lock(&pr->pr_mtx);
2734 prison_hold_locked(pr);
2735 mtx_unlock(&pr->pr_mtx);
2739 prison_proc_hold(struct prison *pr)
2742 mtx_lock(&pr->pr_mtx);
2743 KASSERT(pr->pr_uref > 0,
2744 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2746 mtx_unlock(&pr->pr_mtx);
2750 prison_proc_free(struct prison *pr)
2753 mtx_lock(&pr->pr_mtx);
2754 KASSERT(pr->pr_uref > 0,
2755 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2756 if (pr->pr_uref > 1)
2760 * Don't remove the last user reference in this context, which
2761 * is expected to be a process that is not only locked, but
2765 mtx_unlock(&pr->pr_mtx);
2766 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2769 mtx_unlock(&pr->pr_mtx);
2775 * Restrict a prison's IP address list with its parent's, possibly replacing
2776 * it. Return true if the replacement buffer was used (or would have been).
2779 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2784 ppr = pr->pr_parent;
2785 if (!(pr->pr_flags & PR_IP4_USER)) {
2786 /* This has no user settings, so just copy the parent's list. */
2787 if (pr->pr_ip4s < ppr->pr_ip4s) {
2789 * There's no room for the parent's list. Use the
2790 * new list buffer, which is assumed to be big enough
2791 * (if it was passed). If there's no buffer, try to
2795 if (newip4 == NULL) {
2796 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2797 M_PRISON, M_NOWAIT);
2801 if (newip4 != NULL) {
2802 bcopy(ppr->pr_ip4, newip4,
2803 ppr->pr_ip4s * sizeof(*newip4));
2804 free(pr->pr_ip4, M_PRISON);
2805 pr->pr_ip4 = newip4;
2806 pr->pr_ip4s = ppr->pr_ip4s;
2810 pr->pr_ip4s = ppr->pr_ip4s;
2811 if (pr->pr_ip4s > 0)
2812 bcopy(ppr->pr_ip4, pr->pr_ip4,
2813 pr->pr_ip4s * sizeof(*newip4));
2814 else if (pr->pr_ip4 != NULL) {
2815 free(pr->pr_ip4, M_PRISON);
2818 } else if (pr->pr_ip4s > 0) {
2819 /* Remove addresses that aren't in the parent. */
2820 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2821 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2823 if (ij < ppr->pr_ip4s)
2826 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2827 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2830 for (ij = 1; ii < pr->pr_ip4s; ) {
2831 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2835 switch (ij >= ppr->pr_ip4s ? -1 :
2836 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2838 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2839 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2850 if (pr->pr_ip4s == 0) {
2851 free(pr->pr_ip4, M_PRISON);
2859 * Pass back primary IPv4 address of this jail.
2861 * If not restricted return success but do not alter the address. Caller has
2862 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2864 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2865 * Address returned in NBO.
2868 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2872 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2873 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2875 pr = cred->cr_prison;
2876 if (!(pr->pr_flags & PR_IP4))
2878 mtx_lock(&pr->pr_mtx);
2879 if (!(pr->pr_flags & PR_IP4)) {
2880 mtx_unlock(&pr->pr_mtx);
2883 if (pr->pr_ip4 == NULL) {
2884 mtx_unlock(&pr->pr_mtx);
2885 return (EAFNOSUPPORT);
2888 ia->s_addr = pr->pr_ip4[0].s_addr;
2889 mtx_unlock(&pr->pr_mtx);
2894 * Return 1 if we should do proper source address selection or are not jailed.
2895 * We will return 0 if we should bypass source address selection in favour
2896 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2898 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2901 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2907 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2908 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2913 pr = cred->cr_prison;
2914 if (pr->pr_flags & PR_IP4_SADDRSEL)
2917 lia.s_addr = INADDR_ANY;
2918 error = prison_get_ip4(cred, &lia);
2921 if (lia.s_addr == INADDR_ANY)
2924 ia->s_addr = lia.s_addr;
2929 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2932 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2939 * No need to lock since the PR_IP4_USER flag can't be altered for
2942 while (pr1 != &prison0 &&
2944 !(pr1->pr_flags & PR_VNET) &&
2946 !(pr1->pr_flags & PR_IP4_USER))
2947 pr1 = pr1->pr_parent;
2948 while (pr2 != &prison0 &&
2950 !(pr2->pr_flags & PR_VNET) &&
2952 !(pr2->pr_flags & PR_IP4_USER))
2953 pr2 = pr2->pr_parent;
2954 return (pr1 == pr2);
2958 * Make sure our (source) address is set to something meaningful to this
2961 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2962 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2963 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2966 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2972 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2973 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2975 pr = cred->cr_prison;
2976 if (!(pr->pr_flags & PR_IP4))
2978 mtx_lock(&pr->pr_mtx);
2979 if (!(pr->pr_flags & PR_IP4)) {
2980 mtx_unlock(&pr->pr_mtx);
2983 if (pr->pr_ip4 == NULL) {
2984 mtx_unlock(&pr->pr_mtx);
2985 return (EAFNOSUPPORT);
2988 ia0.s_addr = ntohl(ia->s_addr);
2989 if (ia0.s_addr == INADDR_LOOPBACK) {
2990 ia->s_addr = pr->pr_ip4[0].s_addr;
2991 mtx_unlock(&pr->pr_mtx);
2995 if (ia0.s_addr == INADDR_ANY) {
2997 * In case there is only 1 IPv4 address, bind directly.
2999 if (pr->pr_ip4s == 1)
3000 ia->s_addr = pr->pr_ip4[0].s_addr;
3001 mtx_unlock(&pr->pr_mtx);
3005 error = _prison_check_ip4(pr, ia);
3006 mtx_unlock(&pr->pr_mtx);
3011 * Rewrite destination address in case we will connect to loopback address.
3013 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
3014 * Address passed in in NBO and returned in NBO.
3017 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
3021 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3022 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3024 pr = cred->cr_prison;
3025 if (!(pr->pr_flags & PR_IP4))
3027 mtx_lock(&pr->pr_mtx);
3028 if (!(pr->pr_flags & PR_IP4)) {
3029 mtx_unlock(&pr->pr_mtx);
3032 if (pr->pr_ip4 == NULL) {
3033 mtx_unlock(&pr->pr_mtx);
3034 return (EAFNOSUPPORT);
3037 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
3038 ia->s_addr = pr->pr_ip4[0].s_addr;
3039 mtx_unlock(&pr->pr_mtx);
3044 * Return success because nothing had to be changed.
3046 mtx_unlock(&pr->pr_mtx);
3051 * Check if given address belongs to the jail referenced by cred/prison.
3053 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
3054 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3055 * doesn't allow IPv4. Address passed in in NBO.
3058 _prison_check_ip4(const struct prison *pr, const struct in_addr *ia)
3063 * Check the primary IP.
3065 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3069 * All the other IPs are sorted so we can do a binary search.
3072 z = pr->pr_ip4s - 2;
3075 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3084 return (EADDRNOTAVAIL);
3088 prison_check_ip4(const struct ucred *cred, const struct in_addr *ia)
3093 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3094 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3096 pr = cred->cr_prison;
3097 if (!(pr->pr_flags & PR_IP4))
3099 mtx_lock(&pr->pr_mtx);
3100 if (!(pr->pr_flags & PR_IP4)) {
3101 mtx_unlock(&pr->pr_mtx);
3104 if (pr->pr_ip4 == NULL) {
3105 mtx_unlock(&pr->pr_mtx);
3106 return (EAFNOSUPPORT);
3109 error = _prison_check_ip4(pr, ia);
3110 mtx_unlock(&pr->pr_mtx);
3117 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3122 ppr = pr->pr_parent;
3123 if (!(pr->pr_flags & PR_IP6_USER)) {
3124 /* This has no user settings, so just copy the parent's list. */
3125 if (pr->pr_ip6s < ppr->pr_ip6s) {
3127 * There's no room for the parent's list. Use the
3128 * new list buffer, which is assumed to be big enough
3129 * (if it was passed). If there's no buffer, try to
3133 if (newip6 == NULL) {
3134 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3135 M_PRISON, M_NOWAIT);
3139 if (newip6 != NULL) {
3140 bcopy(ppr->pr_ip6, newip6,
3141 ppr->pr_ip6s * sizeof(*newip6));
3142 free(pr->pr_ip6, M_PRISON);
3143 pr->pr_ip6 = newip6;
3144 pr->pr_ip6s = ppr->pr_ip6s;
3148 pr->pr_ip6s = ppr->pr_ip6s;
3149 if (pr->pr_ip6s > 0)
3150 bcopy(ppr->pr_ip6, pr->pr_ip6,
3151 pr->pr_ip6s * sizeof(*newip6));
3152 else if (pr->pr_ip6 != NULL) {
3153 free(pr->pr_ip6, M_PRISON);
3156 } else if (pr->pr_ip6s > 0) {
3157 /* Remove addresses that aren't in the parent. */
3158 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3159 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3162 if (ij < ppr->pr_ip6s)
3165 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3166 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3169 for (ij = 1; ii < pr->pr_ip6s; ) {
3170 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3175 switch (ij >= ppr->pr_ip6s ? -1 :
3176 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3178 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3179 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3190 if (pr->pr_ip6s == 0) {
3191 free(pr->pr_ip6, M_PRISON);
3199 * Pass back primary IPv6 address for this jail.
3201 * If not restricted return success but do not alter the address. Caller has
3202 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3204 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3207 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3211 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3212 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3214 pr = cred->cr_prison;
3215 if (!(pr->pr_flags & PR_IP6))
3217 mtx_lock(&pr->pr_mtx);
3218 if (!(pr->pr_flags & PR_IP6)) {
3219 mtx_unlock(&pr->pr_mtx);
3222 if (pr->pr_ip6 == NULL) {
3223 mtx_unlock(&pr->pr_mtx);
3224 return (EAFNOSUPPORT);
3227 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3228 mtx_unlock(&pr->pr_mtx);
3233 * Return 1 if we should do proper source address selection or are not jailed.
3234 * We will return 0 if we should bypass source address selection in favour
3235 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3237 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3240 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3243 struct in6_addr lia6;
3246 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3247 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3252 pr = cred->cr_prison;
3253 if (pr->pr_flags & PR_IP6_SADDRSEL)
3257 error = prison_get_ip6(cred, &lia6);
3260 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3263 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3268 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3271 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3277 while (pr1 != &prison0 &&
3279 !(pr1->pr_flags & PR_VNET) &&
3281 !(pr1->pr_flags & PR_IP6_USER))
3282 pr1 = pr1->pr_parent;
3283 while (pr2 != &prison0 &&
3285 !(pr2->pr_flags & PR_VNET) &&
3287 !(pr2->pr_flags & PR_IP6_USER))
3288 pr2 = pr2->pr_parent;
3289 return (pr1 == pr2);
3293 * Make sure our (source) address is set to something meaningful to this jail.
3295 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3296 * when needed while binding.
3298 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3299 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3300 * doesn't allow IPv6.
3303 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3308 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3309 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3311 pr = cred->cr_prison;
3312 if (!(pr->pr_flags & PR_IP6))
3314 mtx_lock(&pr->pr_mtx);
3315 if (!(pr->pr_flags & PR_IP6)) {
3316 mtx_unlock(&pr->pr_mtx);
3319 if (pr->pr_ip6 == NULL) {
3320 mtx_unlock(&pr->pr_mtx);
3321 return (EAFNOSUPPORT);
3324 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3325 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3326 mtx_unlock(&pr->pr_mtx);
3330 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3332 * In case there is only 1 IPv6 address, and v6only is true,
3333 * then bind directly.
3335 if (v6only != 0 && pr->pr_ip6s == 1)
3336 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3337 mtx_unlock(&pr->pr_mtx);
3341 error = _prison_check_ip6(pr, ia6);
3342 mtx_unlock(&pr->pr_mtx);
3347 * Rewrite destination address in case we will connect to loopback address.
3349 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3352 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3356 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3357 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3359 pr = cred->cr_prison;
3360 if (!(pr->pr_flags & PR_IP6))
3362 mtx_lock(&pr->pr_mtx);
3363 if (!(pr->pr_flags & PR_IP6)) {
3364 mtx_unlock(&pr->pr_mtx);
3367 if (pr->pr_ip6 == NULL) {
3368 mtx_unlock(&pr->pr_mtx);
3369 return (EAFNOSUPPORT);
3372 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3373 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3374 mtx_unlock(&pr->pr_mtx);
3379 * Return success because nothing had to be changed.
3381 mtx_unlock(&pr->pr_mtx);
3386 * Check if given address belongs to the jail referenced by cred/prison.
3388 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3389 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3390 * doesn't allow IPv6.
3393 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3398 * Check the primary IP.
3400 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3404 * All the other IPs are sorted so we can do a binary search.
3407 z = pr->pr_ip6s - 2;
3410 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3419 return (EADDRNOTAVAIL);
3423 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3428 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3429 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3431 pr = cred->cr_prison;
3432 if (!(pr->pr_flags & PR_IP6))
3434 mtx_lock(&pr->pr_mtx);
3435 if (!(pr->pr_flags & PR_IP6)) {
3436 mtx_unlock(&pr->pr_mtx);
3439 if (pr->pr_ip6 == NULL) {
3440 mtx_unlock(&pr->pr_mtx);
3441 return (EAFNOSUPPORT);
3444 error = _prison_check_ip6(pr, ia6);
3445 mtx_unlock(&pr->pr_mtx);
3451 * Check if a jail supports the given address family.
3453 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3457 prison_check_af(struct ucred *cred, int af)
3462 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3464 pr = cred->cr_prison;
3466 /* Prisons with their own network stack are not limited. */
3467 if (prison_owns_vnet(cred))
3476 if (pr->pr_flags & PR_IP4)
3478 mtx_lock(&pr->pr_mtx);
3479 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3480 error = EAFNOSUPPORT;
3481 mtx_unlock(&pr->pr_mtx);
3487 if (pr->pr_flags & PR_IP6)
3489 mtx_lock(&pr->pr_mtx);
3490 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3491 error = EAFNOSUPPORT;
3492 mtx_unlock(&pr->pr_mtx);
3500 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3501 error = EAFNOSUPPORT;
3507 * Check if given address belongs to the jail referenced by cred (wrapper to
3508 * prison_check_ip[46]).
3510 * Returns 0 if jail doesn't restrict the address family or if address belongs
3511 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3512 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3515 prison_if(struct ucred *cred, struct sockaddr *sa)
3518 struct sockaddr_in *sai;
3521 struct sockaddr_in6 *sai6;
3525 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3526 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3529 if (prison_owns_vnet(cred))
3534 switch (sa->sa_family)
3538 sai = (struct sockaddr_in *)sa;
3539 error = prison_check_ip4(cred, &sai->sin_addr);
3544 sai6 = (struct sockaddr_in6 *)sa;
3545 error = prison_check_ip6(cred, &sai6->sin6_addr);
3549 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3550 error = EAFNOSUPPORT;
3556 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3559 prison_check(struct ucred *cred1, struct ucred *cred2)
3562 return ((cred1->cr_prison == cred2->cr_prison ||
3563 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3567 * Return 1 if p2 is a child of p1, otherwise 0.
3570 prison_ischild(struct prison *pr1, struct prison *pr2)
3573 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3580 * Return 1 if the passed credential is in a jail, otherwise 0.
3583 jailed(struct ucred *cred)
3586 return (cred->cr_prison != &prison0);
3590 * Return 1 if the passed credential is in a jail and that jail does not
3591 * have its own virtual network stack, otherwise 0.
3594 jailed_without_vnet(struct ucred *cred)
3600 if (prison_owns_vnet(cred))
3608 * Return the correct hostname (domainname, et al) for the passed credential.
3611 getcredhostname(struct ucred *cred, char *buf, size_t size)
3616 * A NULL credential can be used to shortcut to the physical
3617 * system's hostname.
3619 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3620 mtx_lock(&pr->pr_mtx);
3621 strlcpy(buf, pr->pr_hostname, size);
3622 mtx_unlock(&pr->pr_mtx);
3626 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3629 mtx_lock(&cred->cr_prison->pr_mtx);
3630 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3631 mtx_unlock(&cred->cr_prison->pr_mtx);
3635 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3638 mtx_lock(&cred->cr_prison->pr_mtx);
3639 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3640 mtx_unlock(&cred->cr_prison->pr_mtx);
3644 getcredhostid(struct ucred *cred, unsigned long *hostid)
3647 mtx_lock(&cred->cr_prison->pr_mtx);
3648 *hostid = cred->cr_prison->pr_hostid;
3649 mtx_unlock(&cred->cr_prison->pr_mtx);
3654 * Determine whether the prison represented by cred owns
3655 * its vnet rather than having it inherited.
3657 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3660 prison_owns_vnet(struct ucred *cred)
3664 * vnets cannot be added/removed after jail creation,
3665 * so no need to lock here.
3667 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3672 * Determine whether the subject represented by cred can "see"
3673 * status of a mount point.
3674 * Returns: 0 for permitted, ENOENT otherwise.
3675 * XXX: This function should be called cr_canseemount() and should be
3676 * placed in kern_prot.c.
3679 prison_canseemount(struct ucred *cred, struct mount *mp)
3685 pr = cred->cr_prison;
3686 if (pr->pr_enforce_statfs == 0)
3688 if (pr->pr_root->v_mount == mp)
3690 if (pr->pr_enforce_statfs == 2)
3693 * If jail's chroot directory is set to "/" we should be able to see
3694 * all mount-points from inside a jail.
3695 * This is ugly check, but this is the only situation when jail's
3696 * directory ends with '/'.
3698 if (strcmp(pr->pr_path, "/") == 0)
3700 len = strlen(pr->pr_path);
3702 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3705 * Be sure that we don't have situation where jail's root directory
3706 * is "/some/path" and mount point is "/some/pathpath".
3708 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3714 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3716 char jpath[MAXPATHLEN];
3720 pr = cred->cr_prison;
3721 if (pr->pr_enforce_statfs == 0)
3723 if (prison_canseemount(cred, mp) != 0) {
3724 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3725 strlcpy(sp->f_mntonname, "[restricted]",
3726 sizeof(sp->f_mntonname));
3729 if (pr->pr_root->v_mount == mp) {
3731 * Clear current buffer data, so we are sure nothing from
3732 * the valid path left there.
3734 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3735 *sp->f_mntonname = '/';
3739 * If jail's chroot directory is set to "/" we should be able to see
3740 * all mount-points from inside a jail.
3742 if (strcmp(pr->pr_path, "/") == 0)
3744 len = strlen(pr->pr_path);
3745 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3747 * Clear current buffer data, so we are sure nothing from
3748 * the valid path left there.
3750 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3751 if (*jpath == '\0') {
3752 /* Should never happen. */
3753 *sp->f_mntonname = '/';
3755 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3760 * Check with permission for a specific privilege is granted within jail. We
3761 * have a specific list of accepted privileges; the rest are denied.
3764 prison_priv_check(struct ucred *cred, int priv)
3772 * Privileges specific to prisons with a virtual network stack.
3773 * There might be a duplicate entry here in case the privilege
3774 * is only granted conditionally in the legacy jail case.
3779 * NFS-specific privileges.
3781 case PRIV_NFS_DAEMON:
3782 case PRIV_NFS_LOCKD:
3785 * Network stack privileges.
3787 case PRIV_NET_BRIDGE:
3790 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3791 case PRIV_NET_ROUTE:
3793 case PRIV_NET_SETIFMTU:
3794 case PRIV_NET_SETIFFLAGS:
3795 case PRIV_NET_SETIFCAP:
3796 case PRIV_NET_SETIFDESCR:
3797 case PRIV_NET_SETIFNAME :
3798 case PRIV_NET_SETIFMETRIC:
3799 case PRIV_NET_SETIFPHYS:
3800 case PRIV_NET_SETIFMAC:
3801 case PRIV_NET_ADDMULTI:
3802 case PRIV_NET_DELMULTI:
3803 case PRIV_NET_HWIOCTL:
3804 case PRIV_NET_SETLLADDR:
3805 case PRIV_NET_ADDIFGROUP:
3806 case PRIV_NET_DELIFGROUP:
3807 case PRIV_NET_IFCREATE:
3808 case PRIV_NET_IFDESTROY:
3809 case PRIV_NET_ADDIFADDR:
3810 case PRIV_NET_DELIFADDR:
3813 case PRIV_NET_SETIFVNET:
3814 case PRIV_NET_SETIFFIB:
3817 * 802.11-related privileges.
3819 case PRIV_NET80211_GETKEY:
3821 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3828 case PRIV_NETATM_CFG:
3829 case PRIV_NETATM_ADD:
3830 case PRIV_NETATM_DEL:
3831 case PRIV_NETATM_SET:
3834 * Bluetooth privileges.
3836 case PRIV_NETBLUETOOTH_RAW:
3840 * Netgraph and netgraph module privileges.
3842 case PRIV_NETGRAPH_CONTROL:
3844 case PRIV_NETGRAPH_TTY:
3848 * IPv4 and IPv6 privileges.
3850 case PRIV_NETINET_IPFW:
3851 case PRIV_NETINET_DIVERT:
3852 case PRIV_NETINET_PF:
3853 case PRIV_NETINET_DUMMYNET:
3854 case PRIV_NETINET_CARP:
3855 case PRIV_NETINET_MROUTE:
3856 case PRIV_NETINET_RAW:
3857 case PRIV_NETINET_ADDRCTRL6:
3858 case PRIV_NETINET_ND6:
3859 case PRIV_NETINET_SCOPE6:
3860 case PRIV_NETINET_ALIFETIME6:
3861 case PRIV_NETINET_IPSEC:
3862 case PRIV_NETINET_BINDANY:
3877 * No default: or deny here.
3878 * In case of no permit fall through to next switch().
3880 if (cred->cr_prison->pr_flags & PR_VNET)
3888 * Allow ktrace privileges for root in jail.
3894 * Allow jailed processes to configure audit identity and
3895 * submit audit records (login, etc). In the future we may
3896 * want to further refine the relationship between audit and
3899 case PRIV_AUDIT_GETAUDIT:
3900 case PRIV_AUDIT_SETAUDIT:
3901 case PRIV_AUDIT_SUBMIT:
3905 * Allow jailed processes to manipulate process UNIX
3906 * credentials in any way they see fit.
3908 case PRIV_CRED_SETUID:
3909 case PRIV_CRED_SETEUID:
3910 case PRIV_CRED_SETGID:
3911 case PRIV_CRED_SETEGID:
3912 case PRIV_CRED_SETGROUPS:
3913 case PRIV_CRED_SETREUID:
3914 case PRIV_CRED_SETREGID:
3915 case PRIV_CRED_SETRESUID:
3916 case PRIV_CRED_SETRESGID:
3919 * Jail implements visibility constraints already, so allow
3920 * jailed root to override uid/gid-based constraints.
3922 case PRIV_SEEOTHERGIDS:
3923 case PRIV_SEEOTHERUIDS:
3926 * Jail implements inter-process debugging limits already, so
3927 * allow jailed root various debugging privileges.
3929 case PRIV_DEBUG_DIFFCRED:
3930 case PRIV_DEBUG_SUGID:
3931 case PRIV_DEBUG_UNPRIV:
3934 * Allow jail to set various resource limits and login
3935 * properties, and for now, exceed process resource limits.
3937 case PRIV_PROC_LIMIT:
3938 case PRIV_PROC_SETLOGIN:
3939 case PRIV_PROC_SETRLIMIT:
3942 * System V and POSIX IPC privileges are granted in jail.
3945 case PRIV_IPC_WRITE:
3946 case PRIV_IPC_ADMIN:
3947 case PRIV_IPC_MSGSIZE:
3951 * Jail operations within a jail work on child jails.
3953 case PRIV_JAIL_ATTACH:
3955 case PRIV_JAIL_REMOVE:
3958 * Jail implements its own inter-process limits, so allow
3959 * root processes in jail to change scheduling on other
3960 * processes in the same jail. Likewise for signalling.
3962 case PRIV_SCHED_DIFFCRED:
3963 case PRIV_SCHED_CPUSET:
3964 case PRIV_SIGNAL_DIFFCRED:
3965 case PRIV_SIGNAL_SUGID:
3968 * Allow jailed processes to write to sysctls marked as jail
3971 case PRIV_SYSCTL_WRITEJAIL:
3974 * Allow root in jail to manage a variety of quota
3975 * properties. These should likely be conditional on a
3976 * configuration option.
3978 case PRIV_VFS_GETQUOTA:
3979 case PRIV_VFS_SETQUOTA:
3982 * Since Jail relies on chroot() to implement file system
3983 * protections, grant many VFS privileges to root in jail.
3984 * Be careful to exclude mount-related and NFS-related
3988 case PRIV_VFS_WRITE:
3989 case PRIV_VFS_ADMIN:
3991 case PRIV_VFS_LOOKUP:
3992 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3993 case PRIV_VFS_CHFLAGS_DEV:
3994 case PRIV_VFS_CHOWN:
3995 case PRIV_VFS_CHROOT:
3996 case PRIV_VFS_RETAINSUGID:
3997 case PRIV_VFS_FCHROOT:
3999 case PRIV_VFS_SETGID:
4001 case PRIV_VFS_STICKYFILE:
4004 * As in the non-jail case, non-root users are expected to be
4005 * able to read kernel/phyiscal memory (provided /dev/[k]mem
4006 * exists in the jail and they have permission to access it).
4008 case PRIV_KMEM_READ:
4012 * Depending on the global setting, allow privilege of
4013 * setting system flags.
4015 case PRIV_VFS_SYSFLAGS:
4016 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
4022 * Depending on the global setting, allow privilege of
4023 * mounting/unmounting file systems.
4025 case PRIV_VFS_MOUNT:
4026 case PRIV_VFS_UNMOUNT:
4027 case PRIV_VFS_MOUNT_NONUSER:
4028 case PRIV_VFS_MOUNT_OWNER:
4029 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
4030 cred->cr_prison->pr_enforce_statfs < 2)
4036 * Allow jailed root to bind reserved ports and reuse in-use
4039 case PRIV_NETINET_RESERVEDPORT:
4040 case PRIV_NETINET_REUSEPORT:
4044 * Allow jailed root to set certain IPv4/6 (option) headers.
4046 case PRIV_NETINET_SETHDROPTS:
4050 * Conditionally allow creating raw sockets in jail.
4052 case PRIV_NETINET_RAW:
4053 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4059 * Since jail implements its own visibility limits on netstat
4060 * sysctls, allow getcred. This allows identd to work in
4063 case PRIV_NETINET_GETCRED:
4067 * Allow jailed root to set loginclass.
4069 case PRIV_PROC_SETLOGINCLASS:
4074 * In all remaining cases, deny the privilege request. This
4075 * includes almost all network privileges, many system
4076 * configuration privileges.
4083 * Return the part of pr2's name that is relative to pr1, or the whole name
4084 * if it does not directly follow.
4088 prison_name(struct prison *pr1, struct prison *pr2)
4092 /* Jails see themselves as "0" (if they see themselves at all). */
4095 name = pr2->pr_name;
4096 if (prison_ischild(pr1, pr2)) {
4098 * pr1 isn't locked (and allprison_lock may not be either)
4099 * so its length can't be counted on. But the number of dots
4100 * can be counted on - and counted.
4102 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4103 name = strchr(name, '.') + 1;
4109 * Return the part of pr2's path that is relative to pr1, or the whole path
4110 * if it does not directly follow.
4113 prison_path(struct prison *pr1, struct prison *pr2)
4115 char *path1, *path2;
4118 path1 = pr1->pr_path;
4119 path2 = pr2->pr_path;
4120 if (!strcmp(path1, "/"))
4122 len1 = strlen(path1);
4123 if (strncmp(path1, path2, len1))
4125 if (path2[len1] == '\0')
4127 if (path2[len1] == '/')
4128 return (path2 + len1);
4134 * Jail-related sysctls.
4136 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4140 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4143 struct prison *pr, *cpr;
4145 struct in_addr *ip4 = NULL;
4149 struct in6_addr *ip6 = NULL;
4154 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4155 pr = req->td->td_ucred->cr_prison;
4157 sx_slock(&allprison_lock);
4158 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4159 #if defined(INET) || defined(INET6)
4162 mtx_lock(&cpr->pr_mtx);
4164 if (cpr->pr_ip4s > 0) {
4165 if (ip4s < cpr->pr_ip4s) {
4166 ip4s = cpr->pr_ip4s;
4167 mtx_unlock(&cpr->pr_mtx);
4168 ip4 = realloc(ip4, ip4s *
4169 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4172 bcopy(cpr->pr_ip4, ip4,
4173 cpr->pr_ip4s * sizeof(struct in_addr));
4177 if (cpr->pr_ip6s > 0) {
4178 if (ip6s < cpr->pr_ip6s) {
4179 ip6s = cpr->pr_ip6s;
4180 mtx_unlock(&cpr->pr_mtx);
4181 ip6 = realloc(ip6, ip6s *
4182 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4185 bcopy(cpr->pr_ip6, ip6,
4186 cpr->pr_ip6s * sizeof(struct in6_addr));
4189 if (cpr->pr_ref == 0) {
4190 mtx_unlock(&cpr->pr_mtx);
4193 bzero(xp, sizeof(*xp));
4194 xp->pr_version = XPRISON_VERSION;
4195 xp->pr_id = cpr->pr_id;
4196 xp->pr_state = cpr->pr_uref > 0
4197 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4198 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4199 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4200 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4202 xp->pr_ip4s = cpr->pr_ip4s;
4205 xp->pr_ip6s = cpr->pr_ip6s;
4207 mtx_unlock(&cpr->pr_mtx);
4208 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4212 if (xp->pr_ip4s > 0) {
4213 error = SYSCTL_OUT(req, ip4,
4214 xp->pr_ip4s * sizeof(struct in_addr));
4220 if (xp->pr_ip6s > 0) {
4221 error = SYSCTL_OUT(req, ip6,
4222 xp->pr_ip6s * sizeof(struct in6_addr));
4228 sx_sunlock(&allprison_lock);
4239 SYSCTL_OID(_security_jail, OID_AUTO, list,
4240 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4241 sysctl_jail_list, "S", "List of active jails");
4244 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4248 injail = jailed(req->td->td_ucred);
4249 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4254 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4255 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4256 sysctl_jail_jailed, "I", "Process in jail?");
4259 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4261 int error, havevnet;
4263 struct ucred *cred = req->td->td_ucred;
4265 havevnet = jailed(cred) && prison_owns_vnet(cred);
4269 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4274 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4275 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4276 sysctl_jail_vnet, "I", "Jail owns VNET?");
4278 #if defined(INET) || defined(INET6)
4279 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4280 &jail_max_af_ips, 0,
4281 "Number of IP addresses a jail may have at most per address family (deprecated)");
4285 * Default parameters for jail(2) compatibility. For historical reasons,
4286 * the sysctl names have varying similarity to the parameter names. Prisons
4287 * just see their own parameters, and can't change them.
4290 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4293 int allow, error, i;
4295 pr = req->td->td_ucred->cr_prison;
4296 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4298 /* Get the current flag value, and convert it to a boolean. */
4299 i = (allow & arg2) ? 1 : 0;
4302 error = sysctl_handle_int(oidp, &i, 0, req);
4303 if (error || !req->newptr)
4309 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4312 mtx_lock(&prison0.pr_mtx);
4313 jail_default_allow = (jail_default_allow & ~arg2) | i;
4314 mtx_unlock(&prison0.pr_mtx);
4318 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4319 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4320 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4321 "Processes in jail can set their hostnames (deprecated)");
4322 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4323 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4324 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4325 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
4326 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4327 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4328 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4329 "Processes in jail can use System V IPC primitives (deprecated)");
4330 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4331 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4332 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4333 "Prison root can create raw sockets (deprecated)");
4334 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4335 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4336 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4337 "Processes in jail can alter system file flags (deprecated)");
4338 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4339 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4340 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4341 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
4342 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4343 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4344 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4345 "Processes in jail can mount the devfs file system (deprecated)");
4346 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4347 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4348 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4349 "Processes in jail can mount the fdescfs file system (deprecated)");
4350 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4351 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4352 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4353 "Processes in jail can mount the nullfs file system (deprecated)");
4354 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4355 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4356 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4357 "Processes in jail can mount the procfs file system (deprecated)");
4358 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
4359 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4360 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
4361 "Processes in jail can mount the linprocfs file system (deprecated)");
4362 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
4363 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4364 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
4365 "Processes in jail can mount the linsysfs file system (deprecated)");
4366 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4367 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4368 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4369 "Processes in jail can mount the tmpfs file system (deprecated)");
4370 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4371 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4372 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4373 "Processes in jail can mount the zfs file system (deprecated)");
4376 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4381 pr = req->td->td_ucred->cr_prison;
4382 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4383 error = sysctl_handle_int(oidp, &level, 0, req);
4384 if (error || !req->newptr)
4386 *(int *)arg1 = level;
4390 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4391 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4392 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4393 sysctl_jail_default_level, "I",
4394 "Processes in jail cannot see all mounted file systems (deprecated)");
4396 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4397 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4398 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4399 sysctl_jail_default_level, "I",
4400 "Ruleset for the devfs filesystem in jail (deprecated)");
4403 * Nodes to describe jail parameters. Maximum length of string parameters
4404 * is returned in the string itself, and the other parameters exist merely
4405 * to make themselves and their types known.
4407 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4411 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4418 switch (oidp->oid_kind & CTLTYPE)
4424 if (!(req->flags & SCTL_MASK32))
4426 return (SYSCTL_OUT(req, &l, sizeof(l)));
4430 return (SYSCTL_OUT(req, &i, sizeof(i)));
4431 case CTLTYPE_STRING:
4432 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4434 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4435 case CTLTYPE_STRUCT:
4437 return (SYSCTL_OUT(req, &s, sizeof(s)));
4443 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4444 * jail creation time but cannot be changed in an existing jail.
4446 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4447 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4448 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4449 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4450 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4451 "I", "Jail secure level");
4452 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4453 "Jail value for kern.osreldate and uname -K");
4454 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4455 "Jail value for kern.osrelease and uname -r");
4456 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4457 "I", "Jail cannot see all mounted file systems");
4458 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4459 "I", "Ruleset for in-jail devfs mounts");
4460 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4461 "B", "Jail persistence");
4463 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4464 "E,jailsys", "Virtual network stack");
4466 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4467 "B", "Jail is in the process of shutting down");
4469 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4470 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4471 "I", "Current number of child jails");
4472 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4473 "I", "Maximum number of child jails");
4475 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4476 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4478 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4479 "Jail NIS domainname");
4480 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4482 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4483 "LU", "Jail host ID");
4485 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4486 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4489 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4490 "Jail IPv4 address virtualization");
4491 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4492 "S,in_addr,a", "Jail IPv4 addresses");
4493 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4494 "B", "Do (not) use IPv4 source address selection rather than the "
4495 "primary jail IPv4 address.");
4498 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4499 "Jail IPv6 address virtualization");
4500 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4501 "S,in6_addr,a", "Jail IPv6 addresses");
4502 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4503 "B", "Do (not) use IPv6 source address selection rather than the "
4504 "primary jail IPv6 address.");
4507 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4508 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4509 "B", "Jail may set hostname");
4510 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4511 "B", "Jail may use SYSV IPC");
4512 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4513 "B", "Jail may create raw sockets");
4514 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4515 "B", "Jail may alter system file flags");
4516 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4517 "B", "Jail may set file quotas");
4518 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4519 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4521 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4522 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4523 "B", "Jail may mount/unmount jail-friendly file systems in general");
4524 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4525 "B", "Jail may mount the devfs file system");
4526 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4527 "B", "Jail may mount the fdescfs file system");
4528 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4529 "B", "Jail may mount the nullfs file system");
4530 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4531 "B", "Jail may mount the procfs file system");
4532 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
4533 "B", "Jail may mount the linprocfs file system");
4534 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
4535 "B", "Jail may mount the linsysfs file system");
4536 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4537 "B", "Jail may mount the tmpfs file system");
4538 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4539 "B", "Jail may mount the zfs file system");
4543 prison_racct_foreach(void (*callback)(struct racct *racct,
4544 void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
4545 void *arg2, void *arg3)
4547 struct prison_racct *prr;
4549 ASSERT_RACCT_ENABLED();
4551 sx_slock(&allprison_lock);
4554 LIST_FOREACH(prr, &allprison_racct, prr_next)
4555 (callback)(prr->prr_racct, arg2, arg3);
4558 sx_sunlock(&allprison_lock);
4561 static struct prison_racct *
4562 prison_racct_find_locked(const char *name)
4564 struct prison_racct *prr;
4566 ASSERT_RACCT_ENABLED();
4567 sx_assert(&allprison_lock, SA_XLOCKED);
4569 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4572 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4573 if (strcmp(name, prr->prr_name) != 0)
4576 /* Found prison_racct with a matching name? */
4577 prison_racct_hold(prr);
4581 /* Add new prison_racct. */
4582 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4583 racct_create(&prr->prr_racct);
4585 strcpy(prr->prr_name, name);
4586 refcount_init(&prr->prr_refcount, 1);
4587 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4592 struct prison_racct *
4593 prison_racct_find(const char *name)
4595 struct prison_racct *prr;
4597 ASSERT_RACCT_ENABLED();
4599 sx_xlock(&allprison_lock);
4600 prr = prison_racct_find_locked(name);
4601 sx_xunlock(&allprison_lock);
4606 prison_racct_hold(struct prison_racct *prr)
4609 ASSERT_RACCT_ENABLED();
4611 refcount_acquire(&prr->prr_refcount);
4615 prison_racct_free_locked(struct prison_racct *prr)
4618 ASSERT_RACCT_ENABLED();
4619 sx_assert(&allprison_lock, SA_XLOCKED);
4621 if (refcount_release(&prr->prr_refcount)) {
4622 racct_destroy(&prr->prr_racct);
4623 LIST_REMOVE(prr, prr_next);
4624 free(prr, M_PRISON_RACCT);
4629 prison_racct_free(struct prison_racct *prr)
4633 ASSERT_RACCT_ENABLED();
4634 sx_assert(&allprison_lock, SA_UNLOCKED);
4636 old = prr->prr_refcount;
4637 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4640 sx_xlock(&allprison_lock);
4641 prison_racct_free_locked(prr);
4642 sx_xunlock(&allprison_lock);
4646 prison_racct_attach(struct prison *pr)
4648 struct prison_racct *prr;
4650 ASSERT_RACCT_ENABLED();
4651 sx_assert(&allprison_lock, SA_XLOCKED);
4653 prr = prison_racct_find_locked(pr->pr_name);
4654 KASSERT(prr != NULL, ("cannot find prison_racct"));
4656 pr->pr_prison_racct = prr;
4660 * Handle jail renaming. From the racct point of view, renaming means
4661 * moving from one prison_racct to another.
4664 prison_racct_modify(struct prison *pr)
4668 struct prison_racct *oldprr;
4670 ASSERT_RACCT_ENABLED();
4672 sx_slock(&allproc_lock);
4673 sx_xlock(&allprison_lock);
4675 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4676 sx_xunlock(&allprison_lock);
4677 sx_sunlock(&allproc_lock);
4681 oldprr = pr->pr_prison_racct;
4682 pr->pr_prison_racct = NULL;
4684 prison_racct_attach(pr);
4687 * Move resource utilisation records.
4689 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4692 * Force rctl to reattach rules to processes.
4694 FOREACH_PROC_IN_SYSTEM(p) {
4696 cred = crhold(p->p_ucred);
4698 racct_proc_ucred_changed(p, cred, cred);
4702 sx_sunlock(&allproc_lock);
4703 prison_racct_free_locked(oldprr);
4704 sx_xunlock(&allprison_lock);
4708 prison_racct_detach(struct prison *pr)
4711 ASSERT_RACCT_ENABLED();
4712 sx_assert(&allprison_lock, SA_UNLOCKED);
4714 if (pr->pr_prison_racct == NULL)
4716 prison_racct_free(pr->pr_prison_racct);
4717 pr->pr_prison_racct = NULL;
4724 db_show_prison(struct prison *pr)
4727 #if defined(INET) || defined(INET6)
4732 char ip6buf[INET6_ADDRSTRLEN];
4735 db_printf("prison %p:\n", pr);
4736 db_printf(" jid = %d\n", pr->pr_id);
4737 db_printf(" name = %s\n", pr->pr_name);
4738 db_printf(" parent = %p\n", pr->pr_parent);
4739 db_printf(" ref = %d\n", pr->pr_ref);
4740 db_printf(" uref = %d\n", pr->pr_uref);
4741 db_printf(" path = %s\n", pr->pr_path);
4742 db_printf(" cpuset = %d\n", pr->pr_cpuset
4743 ? pr->pr_cpuset->cs_id : -1);
4745 db_printf(" vnet = %p\n", pr->pr_vnet);
4747 db_printf(" root = %p\n", pr->pr_root);
4748 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4749 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4750 db_printf(" children.max = %d\n", pr->pr_childmax);
4751 db_printf(" children.cur = %d\n", pr->pr_childcount);
4752 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4753 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4754 db_printf(" flags = 0x%x", pr->pr_flags);
4755 for (fi = 0; fi < nitems(pr_flag_names); fi++)
4756 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4757 db_printf(" %s", pr_flag_names[fi]);
4758 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) {
4759 jsf = pr->pr_flags &
4760 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4761 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4762 pr_flag_jailsys[fi].disable &&
4763 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4764 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4767 db_printf(" allow = 0x%x", pr->pr_allow);
4768 for (fi = 0; fi < nitems(pr_allow_names); fi++)
4769 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4770 db_printf(" %s", pr_allow_names[fi]);
4772 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4773 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4774 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4775 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4776 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4778 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4779 for (ii = 0; ii < pr->pr_ip4s; ii++)
4780 db_printf(" %s %s\n",
4781 ii == 0 ? "ip4.addr =" : " ",
4782 inet_ntoa(pr->pr_ip4[ii]));
4785 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4786 for (ii = 0; ii < pr->pr_ip6s; ii++)
4787 db_printf(" %s %s\n",
4788 ii == 0 ? "ip6.addr =" : " ",
4789 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4793 DB_SHOW_COMMAND(prison, db_show_prison_command)
4799 * Show all prisons in the list, and prison0 which is not
4802 db_show_prison(&prison0);
4803 if (!db_pager_quit) {
4804 TAILQ_FOREACH(pr, &allprison, pr_list) {
4816 /* Look for a prison with the ID and with references. */
4817 TAILQ_FOREACH(pr, &allprison, pr_list)
4818 if (pr->pr_id == addr && pr->pr_ref > 0)
4821 /* Look again, without requiring a reference. */
4822 TAILQ_FOREACH(pr, &allprison, pr_list)
4823 if (pr->pr_id == addr)
4826 /* Assume address points to a valid prison. */
4827 pr = (struct prison *)addr;