2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_compat.h"
35 #include "opt_inet6.h"
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
67 #include <netinet/in.h>
72 #include <netinet6/in6_var.h>
76 #include <security/mac/mac_framework.h>
78 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
81 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
92 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
94 #define _PR_IP_SADDRSEL 0
98 /* prison0 describes what is "real" about the system. */
99 struct prison prison0 = {
105 .pr_securelevel = -1,
107 .pr_childmax = JAIL_MAX,
108 .pr_hostuuid = DEFAULT_HOSTUUID,
109 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
111 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
113 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
115 .pr_allow = PR_ALLOW_ALL,
117 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
119 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
120 struct sx allprison_lock;
121 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
122 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
123 LIST_HEAD(, prison_racct) allprison_racct;
126 static int do_jail_attach(struct thread *td, struct prison *pr);
127 static void prison_complete(void *context, int pending);
128 static void prison_deref(struct prison *pr, int flags);
129 static char *prison_path(struct prison *pr1, struct prison *pr2);
130 static void prison_remove_one(struct prison *pr);
132 static void prison_racct_attach(struct prison *pr);
133 static void prison_racct_modify(struct prison *pr);
134 static void prison_racct_detach(struct prison *pr);
137 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
138 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
141 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
142 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
145 /* Flags for prison_deref */
146 #define PD_DEREF 0x01
147 #define PD_DEUREF 0x02
148 #define PD_LOCKED 0x04
149 #define PD_LIST_SLOCKED 0x08
150 #define PD_LIST_XLOCKED 0x10
153 * Parameter names corresponding to PR_* flag values. Size values are for kvm
154 * as we cannot figure out the size of a sparse array, or an array without a
157 static char *pr_flag_names[] = {
160 [7] = "ip4.saddrsel",
163 [8] = "ip6.saddrsel",
166 const size_t pr_flag_names_size = sizeof(pr_flag_names);
168 static char *pr_flag_nonames[] = {
171 [7] = "ip4.nosaddrsel",
174 [8] = "ip6.nosaddrsel",
177 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
179 struct jailsys_flags {
183 } pr_flag_jailsys[] = {
184 { "host", 0, PR_HOST },
186 { "vnet", 0, PR_VNET },
189 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
192 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
195 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
197 static char *pr_allow_names[] = {
198 "allow.set_hostname",
206 "allow.mount.nullfs",
208 "allow.mount.procfs",
210 "allow.mount.fdescfs",
212 const size_t pr_allow_names_size = sizeof(pr_allow_names);
214 static char *pr_allow_nonames[] = {
215 "allow.noset_hostname",
217 "allow.noraw_sockets",
222 "allow.mount.nodevfs",
223 "allow.mount.nonullfs",
225 "allow.mount.noprocfs",
226 "allow.mount.notmpfs",
227 "allow.mount.nofdescfs",
229 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
231 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
232 #define JAIL_DEFAULT_ENFORCE_STATFS 2
233 #define JAIL_DEFAULT_DEVFS_RSNUM 0
234 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
235 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
236 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
237 #if defined(INET) || defined(INET6)
238 static unsigned jail_max_af_ips = 255;
242 * Initialize the parts of prison0 that can't be static-initialized with
243 * constants. This is called from proc0_init() after creating thread0 cpuset.
249 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
250 prison0.pr_osreldate = osreldate;
251 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
256 qcmp_v4(const void *ip1, const void *ip2)
261 * We need to compare in HBO here to get the list sorted as expected
262 * by the result of the code. Sorting NBO addresses gives you
263 * interesting results. If you do not understand, do not try.
265 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
266 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
269 * Do not simply return the difference of the two numbers, the int is
283 qcmp_v6(const void *ip1, const void *ip2)
285 const struct in6_addr *ia6a, *ia6b;
288 ia6a = (const struct in6_addr *)ip1;
289 ia6b = (const struct in6_addr *)ip2;
292 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
293 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
295 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
308 sys_jail(struct thread *td, struct jail_args *uap)
314 error = copyin(uap->jail, &version, sizeof(uint32_t));
323 /* FreeBSD single IPv4 jails. */
324 bzero(&j, sizeof(struct jail));
325 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
328 j.version = j0.version;
330 j.hostname = j0.hostname;
331 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
337 * Version 1 was used by multi-IPv4 jail implementations
338 * that never made it into the official kernel.
342 case 2: /* JAIL_API_VERSION */
343 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
344 error = copyin(uap->jail, &j, sizeof(struct jail));
350 /* Sci-Fi jails are not supported, sorry. */
353 return (kern_jail(td, &j));
357 kern_jail(struct thread *td, struct jail *j)
359 struct iovec optiov[2 * (4
360 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
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 < sizeof(pr_allow_names) /
392 sizeof(pr_allow_names[0]); fi++) {
393 optiov[opt.uio_iovcnt].iov_base =
394 (jail_default_allow & (1 << fi))
395 ? pr_allow_names[fi] : pr_allow_nonames[fi];
396 optiov[opt.uio_iovcnt].iov_len =
397 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
400 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
401 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
403 enforce_statfs = jail_default_enforce_statfs;
404 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
405 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
409 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
411 ip4s = (j->version == 0) ? 1 : j->ip4s;
412 if (ip4s > jail_max_af_ips)
414 tmplen += ip4s * sizeof(struct in_addr);
420 if (j->ip6s > jail_max_af_ips)
422 tmplen += j->ip6s * sizeof(struct in6_addr);
427 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
428 u_hostname = u_path + MAXPATHLEN;
429 u_name = u_hostname + MAXHOSTNAMELEN;
431 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
435 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
437 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
440 optiov[opt.uio_iovcnt].iov_base = "path";
441 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
443 optiov[opt.uio_iovcnt].iov_base = u_path;
444 error = copyinstr(j->path, u_path, MAXPATHLEN,
445 &optiov[opt.uio_iovcnt].iov_len);
447 free(u_path, M_TEMP);
451 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
452 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
454 optiov[opt.uio_iovcnt].iov_base = u_hostname;
455 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
456 &optiov[opt.uio_iovcnt].iov_len);
458 free(u_path, M_TEMP);
462 if (j->jailname != NULL) {
463 optiov[opt.uio_iovcnt].iov_base = "name";
464 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
466 optiov[opt.uio_iovcnt].iov_base = u_name;
467 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
468 &optiov[opt.uio_iovcnt].iov_len);
470 free(u_path, M_TEMP);
476 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
477 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
479 optiov[opt.uio_iovcnt].iov_base = u_ip4;
480 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
482 u_ip4->s_addr = j->ip4s;
484 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
486 free(u_path, M_TEMP);
493 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
494 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
496 optiov[opt.uio_iovcnt].iov_base = u_ip6;
497 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
498 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
500 free(u_path, M_TEMP);
505 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
506 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
507 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
508 free(u_path, M_TEMP);
514 * struct jail_set_args {
515 * struct iovec *iovp;
516 * unsigned int iovcnt;
521 sys_jail_set(struct thread *td, struct jail_set_args *uap)
526 /* Check that we have an even number of iovecs. */
530 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
533 error = kern_jail_set(td, auio, uap->flags);
539 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
546 struct in6_addr *ip6;
549 struct vfsoptlist *opts;
550 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
552 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
553 char *g_path, *osrelstr;
554 #if defined(INET) || defined(INET6)
559 size_t namelen, onamelen;
560 int created, cuflags, descend, enforce, 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 = ppr = 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 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
616 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
625 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
633 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
638 else if (enforce < 0 || enforce > 2) {
644 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
652 pr_flags = ch_flags = 0;
653 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
655 if (pr_flag_names[fi] == NULL)
657 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
658 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
660 ch_flags |= pr_flags;
661 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
663 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
670 case JAIL_SYS_DISABLE:
671 if (!pr_flag_jailsys[fi].disable) {
675 pr_flags |= pr_flag_jailsys[fi].disable;
678 pr_flags |= pr_flag_jailsys[fi].new;
680 case JAIL_SYS_INHERIT:
687 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
689 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
690 && !(pr_flags & PR_PERSIST)) {
692 vfs_opterror(opts, "new jail must persist or attach");
696 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
698 vfs_opterror(opts, "vnet cannot be changed after creation");
703 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
705 vfs_opterror(opts, "ip4 cannot be changed after creation");
710 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
712 vfs_opterror(opts, "ip6 cannot be changed after creation");
717 pr_allow = ch_allow = 0;
718 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
720 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
721 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
723 ch_allow |= pr_allow;
725 error = vfs_getopt(opts, "name", (void **)&name, &len);
731 if (len == 0 || name[len - 1] != '\0') {
735 if (len > MAXHOSTNAMELEN) {
736 error = ENAMETOOLONG;
741 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
749 if (len == 0 || host[len - 1] != '\0') {
753 if (len > MAXHOSTNAMELEN) {
754 error = ENAMETOOLONG;
759 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
767 if (len == 0 || domain[len - 1] != '\0') {
771 if (len > MAXHOSTNAMELEN) {
772 error = ENAMETOOLONG;
777 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
785 if (len == 0 || uuid[len - 1] != '\0') {
789 if (len > HOSTUUIDLEN) {
790 error = ENAMETOOLONG;
795 #ifdef COMPAT_FREEBSD32
796 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
799 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
803 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
815 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
820 else if (ip4s & (sizeof(*ip4) - 1)) {
824 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
826 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
828 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
829 ip4s /= sizeof(*ip4);
830 if (ip4s > jail_max_af_ips) {
832 vfs_opterror(opts, "too many IPv4 addresses");
835 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
836 bcopy(op, ip4, ip4s * sizeof(*ip4));
838 * IP addresses are all sorted but ip[0] to preserve
839 * the primary IP address as given from userland.
840 * This special IP is used for unbound outgoing
841 * connections as well for "loopback" traffic in case
842 * source address selection cannot find any more fitting
843 * address to connect from.
846 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
848 * Check for duplicate addresses and do some simple
849 * zero and broadcast checks. If users give other bogus
850 * addresses it is their problem.
852 * We do not have to care about byte order for these
853 * checks so we will do them in NBO.
855 for (ii = 0; ii < ip4s; ii++) {
856 if (ip4[ii].s_addr == INADDR_ANY ||
857 ip4[ii].s_addr == INADDR_BROADCAST) {
862 (ip4[0].s_addr == ip4[ii+1].s_addr ||
863 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
873 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
878 else if (ip6s & (sizeof(*ip6) - 1)) {
882 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
884 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
886 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | 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 fullpath_disabled = 0;
925 error = vfs_getopt(opts, "path", (void **)&path, &len);
931 if (flags & JAIL_UPDATE) {
934 "path cannot be changed after creation");
937 if (len == 0 || path[len - 1] != '\0') {
941 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
947 NDFREE(&nd, NDF_ONLY_PNBUF);
948 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
949 strlcpy(g_path, path, MAXPATHLEN);
950 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
953 else if (error == ENODEV) {
954 /* proceed if sysctl debug.disablefullpath == 1 */
955 fullpath_disabled = 1;
956 if (len < 2 || (len == 2 && path[0] == '/'))
959 /* exit on other errors */
962 if (root->v_type != VDIR) {
968 if (fullpath_disabled) {
969 /* Leave room for a real-root full pathname. */
970 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
971 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
972 error = ENAMETOOLONG;
978 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
984 if (flags & JAIL_UPDATE) {
987 "osrelease cannot be changed after creation");
990 if (len == 0 || len >= OSRELEASELEN) {
993 "osrelease string must be 1-%d bytes long",
999 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
1000 if (error == ENOENT)
1002 else if (error != 0)
1005 if (flags & JAIL_UPDATE) {
1008 "osreldate cannot be changed after creation");
1013 vfs_opterror(opts, "osreldate cannot be 0");
1019 * Grab the allprison lock before letting modules check their
1020 * parameters. Once we have it, do not let go so we'll have a
1021 * consistent view of the OSD list.
1023 sx_xlock(&allprison_lock);
1024 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
1026 goto done_unlock_list;
1028 /* By now, all parameters should have been noted. */
1029 TAILQ_FOREACH(opt, opts, link) {
1030 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1032 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1033 goto done_unlock_list;
1038 * See if we are creating a new record or updating an existing one.
1039 * This abuses the file error codes ENOENT and EEXIST.
1041 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
1044 vfs_opterror(opts, "no valid operation (create or update)");
1045 goto done_unlock_list;
1049 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1050 namelc = strrchr(name, '.');
1051 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1057 * See if a requested jid already exists. There is an
1058 * information leak here if the jid exists but is not within
1059 * the caller's jail hierarchy. Jail creators will get EEXIST
1060 * even though they cannot see the jail, and CREATE | UPDATE
1061 * will return ENOENT which is not normally a valid error.
1065 vfs_opterror(opts, "negative jid");
1066 goto done_unlock_list;
1068 pr = prison_find(jid);
1070 ppr = pr->pr_parent;
1071 /* Create: jid must not exist. */
1072 if (cuflags == JAIL_CREATE) {
1073 mtx_unlock(&pr->pr_mtx);
1075 vfs_opterror(opts, "jail %d already exists",
1077 goto done_unlock_list;
1079 if (!prison_ischild(mypr, pr)) {
1080 mtx_unlock(&pr->pr_mtx);
1082 } else if (pr->pr_uref == 0) {
1083 if (!(flags & JAIL_DYING)) {
1084 mtx_unlock(&pr->pr_mtx);
1086 vfs_opterror(opts, "jail %d is dying",
1088 goto done_unlock_list;
1089 } else if ((flags & JAIL_ATTACH) ||
1090 (pr_flags & PR_PERSIST)) {
1092 * A dying jail might be resurrected
1093 * (via attach or persist), but first
1094 * it must determine if another jail
1095 * has claimed its name. Accomplish
1096 * this by implicitly re-setting the
1100 name = prison_name(mypr, pr);
1105 /* Update: jid must exist. */
1106 if (cuflags == JAIL_UPDATE) {
1108 vfs_opterror(opts, "jail %d not found", jid);
1109 goto done_unlock_list;
1114 * If the caller provided a name, look for a jail by that name.
1115 * This has different semantics for creates and updates keyed by jid
1116 * (where the name must not already exist in a different jail),
1117 * and updates keyed by the name itself (where the name must exist
1118 * because that is the jail being updated).
1121 namelc = strrchr(name, '.');
1126 * This is a hierarchical name. Split it into the
1127 * parent and child names, and make sure the parent
1128 * exists or matches an already found jail.
1132 if (strncmp(name, ppr->pr_name, namelc - name)
1133 || ppr->pr_name[namelc - name] != '\0') {
1134 mtx_unlock(&pr->pr_mtx);
1137 "cannot change jail's parent");
1138 goto done_unlock_list;
1141 ppr = prison_find_name(mypr, name);
1145 "jail \"%s\" not found", name);
1146 goto done_unlock_list;
1148 mtx_unlock(&ppr->pr_mtx);
1152 if (name[0] != '\0') {
1154 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1157 FOREACH_PRISON_CHILD(ppr, tpr) {
1158 if (tpr != pr && tpr->pr_ref > 0 &&
1159 !strcmp(tpr->pr_name + namelen, name)) {
1161 cuflags != JAIL_CREATE) {
1162 mtx_lock(&tpr->pr_mtx);
1163 if (tpr->pr_ref > 0) {
1168 if (tpr->pr_uref > 0) {
1174 mtx_unlock(&tpr->pr_mtx);
1175 } else if (tpr->pr_uref > 0) {
1177 * Create, or update(jid):
1178 * name must not exist in an
1179 * active sibling jail.
1183 mtx_unlock(&pr->pr_mtx);
1185 "jail \"%s\" already exists",
1187 goto done_unlock_list;
1191 /* If no active jail is found, use a dying one. */
1192 if (deadpr != NULL && pr == NULL) {
1193 if (flags & JAIL_DYING) {
1194 mtx_lock(&deadpr->pr_mtx);
1195 if (deadpr->pr_ref == 0) {
1196 mtx_unlock(&deadpr->pr_mtx);
1200 } else if (cuflags == JAIL_UPDATE) {
1203 "jail \"%s\" is dying", name);
1204 goto done_unlock_list;
1207 /* Update: name must exist if no jid. */
1208 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1210 vfs_opterror(opts, "jail \"%s\" not found",
1212 goto done_unlock_list;
1216 /* Update: must provide a jid or name. */
1217 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1219 vfs_opterror(opts, "update specified no jail");
1220 goto done_unlock_list;
1223 /* If there's no prison to update, create a new one and link it in. */
1225 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1226 if (tpr->pr_childcount >= tpr->pr_childmax) {
1228 vfs_opterror(opts, "prison limit exceeded");
1229 goto done_unlock_list;
1232 mtx_lock(&ppr->pr_mtx);
1233 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1234 mtx_unlock(&ppr->pr_mtx);
1236 vfs_opterror(opts, "parent jail went away!");
1237 goto done_unlock_list;
1241 mtx_unlock(&ppr->pr_mtx);
1242 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1244 /* Find the next free jid. */
1247 if (jid == JAIL_MAX)
1249 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1250 if (tpr->pr_id < jid)
1252 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1253 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1256 if (jid == lastprid) {
1259 "no available jail IDs");
1261 prison_deref(ppr, PD_DEREF |
1262 PD_DEUREF | PD_LIST_XLOCKED);
1271 * The jail already has a jid (that did not yet exist),
1272 * so just find where to insert it.
1274 TAILQ_FOREACH(tpr, &allprison, pr_list)
1275 if (tpr->pr_id >= jid) {
1276 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1281 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1282 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1283 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1284 tpr->pr_childcount++;
1286 pr->pr_parent = ppr;
1289 /* Set some default values, and inherit some from the parent. */
1294 root = mypr->pr_root;
1297 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1298 pr->pr_flags |= PR_HOST;
1299 #if defined(INET) || defined(INET6)
1301 if (!(pr_flags & PR_VNET))
1305 if (!(ch_flags & PR_IP4_USER))
1307 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1308 else if (!(pr_flags & PR_IP4_USER)) {
1309 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1310 if (ppr->pr_ip4 != NULL) {
1311 pr->pr_ip4s = ppr->pr_ip4s;
1312 pr->pr_ip4 = malloc(pr->pr_ip4s *
1313 sizeof(struct in_addr), M_PRISON,
1315 bcopy(ppr->pr_ip4, pr->pr_ip4,
1316 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1321 if (!(ch_flags & PR_IP6_USER))
1323 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1324 else if (!(pr_flags & PR_IP6_USER)) {
1325 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1326 if (ppr->pr_ip6 != NULL) {
1327 pr->pr_ip6s = ppr->pr_ip6s;
1328 pr->pr_ip6 = malloc(pr->pr_ip6s *
1329 sizeof(struct in6_addr), M_PRISON,
1331 bcopy(ppr->pr_ip6, pr->pr_ip6,
1332 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1338 /* Source address selection is always on by default. */
1339 pr->pr_flags |= _PR_IP_SADDRSEL;
1341 pr->pr_securelevel = ppr->pr_securelevel;
1342 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1343 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1344 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1346 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1347 if (osrelstr == NULL)
1348 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1350 strcpy(pr->pr_osrelease, osrelstr);
1352 LIST_INIT(&pr->pr_children);
1353 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1356 /* Allocate a new vnet if specified. */
1357 pr->pr_vnet = (pr_flags & PR_VNET)
1358 ? vnet_alloc() : ppr->pr_vnet;
1361 * Allocate a dedicated cpuset for each jail.
1362 * Unlike other initial settings, this may return an erorr.
1364 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1366 prison_deref(pr, PD_LIST_XLOCKED);
1370 mtx_lock(&pr->pr_mtx);
1372 * New prisons do not yet have a reference, because we do not
1373 * want other to see the incomplete prison once the
1374 * allprison_lock is downgraded.
1379 * Grab a reference for existing prisons, to ensure they
1380 * continue to exist for the duration of the call.
1383 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1384 if ((pr->pr_flags & PR_VNET) &&
1385 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1388 "vnet jails cannot have IP address restrictions");
1389 goto done_deref_locked;
1393 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1396 "ip4 cannot be changed after creation");
1397 goto done_deref_locked;
1401 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1404 "ip6 cannot be changed after creation");
1405 goto done_deref_locked;
1410 /* Do final error checking before setting anything. */
1412 if (slevel < ppr->pr_securelevel) {
1414 goto done_deref_locked;
1418 if (childmax >= ppr->pr_childmax) {
1420 goto done_deref_locked;
1424 if (enforce < ppr->pr_enforce_statfs) {
1426 goto done_deref_locked;
1431 * devfs_rsnum is a uint16_t
1433 if (rsnum < 0 || rsnum > 65535) {
1435 goto done_deref_locked;
1438 * Nested jails always inherit parent's devfs ruleset
1440 if (jailed(td->td_ucred)) {
1441 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1443 goto done_deref_locked;
1445 rsnum = ppr->pr_devfs_rsnum;
1450 if (ppr->pr_flags & PR_IP4) {
1452 * Make sure the new set of IP addresses is a
1453 * subset of the parent's list. Don't worry
1454 * about the parent being unlocked, as any
1455 * setting is done with allprison_lock held.
1457 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1458 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1460 if (ij == ppr->pr_ip4s) {
1462 goto done_deref_locked;
1465 for (ii = ij = 1; ii < ip4s; ii++) {
1466 if (ip4[ii].s_addr ==
1467 ppr->pr_ip4[0].s_addr)
1469 for (; ij < ppr->pr_ip4s; ij++)
1470 if (ip4[ii].s_addr ==
1471 ppr->pr_ip4[ij].s_addr)
1473 if (ij == ppr->pr_ip4s)
1476 if (ij == ppr->pr_ip4s) {
1478 goto done_deref_locked;
1483 * Check for conflicting IP addresses. We permit them
1484 * if there is no more than one IP on each jail. If
1485 * there is a duplicate on a jail with more than one
1486 * IP stop checking and return error.
1490 for (; tppr != &prison0; tppr = tppr->pr_parent)
1491 if (tppr->pr_flags & PR_VNET)
1494 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1497 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1499 tpr->pr_uref == 0) {
1503 if (!(tpr->pr_flags & PR_IP4_USER))
1506 if (tpr->pr_ip4 == NULL ||
1507 (ip4s == 1 && tpr->pr_ip4s == 1))
1509 for (ii = 0; ii < ip4s; ii++) {
1510 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1513 "IPv4 addresses clash");
1514 goto done_deref_locked;
1522 if (ppr->pr_flags & PR_IP6) {
1524 * Make sure the new set of IP addresses is a
1525 * subset of the parent's list.
1527 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1528 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1531 if (ij == ppr->pr_ip6s) {
1533 goto done_deref_locked;
1536 for (ii = ij = 1; ii < ip6s; ii++) {
1537 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1540 for (; ij < ppr->pr_ip6s; ij++)
1541 if (IN6_ARE_ADDR_EQUAL(
1542 &ip6[ii], &ppr->pr_ip6[ij]))
1544 if (ij == ppr->pr_ip6s)
1547 if (ij == ppr->pr_ip6s) {
1549 goto done_deref_locked;
1553 /* Check for conflicting IP addresses. */
1556 for (; tppr != &prison0; tppr = tppr->pr_parent)
1557 if (tppr->pr_flags & PR_VNET)
1560 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1563 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1565 tpr->pr_uref == 0) {
1569 if (!(tpr->pr_flags & PR_IP6_USER))
1572 if (tpr->pr_ip6 == NULL ||
1573 (ip6s == 1 && tpr->pr_ip6s == 1))
1575 for (ii = 0; ii < ip6s; ii++) {
1576 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1579 "IPv6 addresses clash");
1580 goto done_deref_locked;
1586 onamelen = namelen = 0;
1588 /* Give a default name of the jid. */
1589 if (name[0] == '\0')
1590 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1591 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1595 "name cannot be numeric (unless it is the jid)");
1596 goto done_deref_locked;
1599 * Make sure the name isn't too long for the prison or its
1602 onamelen = strlen(pr->pr_name);
1603 namelen = strlen(name);
1604 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1605 error = ENAMETOOLONG;
1606 goto done_deref_locked;
1608 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1609 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1610 sizeof(pr->pr_name)) {
1611 error = ENAMETOOLONG;
1612 goto done_deref_locked;
1616 if (pr_allow & ~ppr->pr_allow) {
1618 goto done_deref_locked;
1621 /* Set the parameters of the prison. */
1624 if (pr_flags & PR_IP4_USER) {
1625 pr->pr_flags |= PR_IP4;
1626 free(pr->pr_ip4, M_PRISON);
1630 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1632 if (tpr->pr_flags & PR_VNET) {
1637 if (prison_restrict_ip4(tpr, NULL)) {
1646 if (pr_flags & PR_IP6_USER) {
1647 pr->pr_flags |= PR_IP6;
1648 free(pr->pr_ip6, M_PRISON);
1652 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1654 if (tpr->pr_flags & PR_VNET) {
1659 if (prison_restrict_ip6(tpr, NULL)) {
1667 pr->pr_securelevel = slevel;
1668 /* Set all child jails to be at least this level. */
1669 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1670 if (tpr->pr_securelevel < slevel)
1671 tpr->pr_securelevel = slevel;
1674 pr->pr_childmax = childmax;
1675 /* Set all child jails to under this limit. */
1676 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1677 if (tpr->pr_childmax > childmax - level)
1678 tpr->pr_childmax = childmax > level
1679 ? childmax - level : 0;
1682 pr->pr_enforce_statfs = enforce;
1683 /* Pass this restriction on to the children. */
1684 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1685 if (tpr->pr_enforce_statfs < enforce)
1686 tpr->pr_enforce_statfs = enforce;
1689 pr->pr_devfs_rsnum = rsnum;
1690 /* Pass this restriction on to the children. */
1691 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1692 tpr->pr_devfs_rsnum = rsnum;
1695 if (ppr == &prison0)
1696 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1698 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1699 ppr->pr_name, name);
1700 /* Change this component of child names. */
1701 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1702 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1703 strlen(tpr->pr_name + onamelen) + 1);
1704 bcopy(pr->pr_name, tpr->pr_name, namelen);
1708 /* Try to keep a real-rooted full pathname. */
1709 if (fullpath_disabled && path[0] == '/' &&
1710 strcmp(mypr->pr_path, "/"))
1711 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1712 mypr->pr_path, path);
1714 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1717 if (PR_HOST & ch_flags & ~pr_flags) {
1718 if (pr->pr_flags & PR_HOST) {
1720 * Copy the parent's host info. As with pr_ip4 above,
1721 * the lack of a lock on the parent is not a problem;
1722 * it is always set with allprison_lock at least
1723 * shared, and is held exclusively here.
1725 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1726 sizeof(pr->pr_hostname));
1727 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1728 sizeof(pr->pr_domainname));
1729 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1730 sizeof(pr->pr_hostuuid));
1731 pr->pr_hostid = pr->pr_parent->pr_hostid;
1733 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1734 /* Set this prison, and any descendants without PR_HOST. */
1736 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1738 strlcpy(pr->pr_domainname, domain,
1739 sizeof(pr->pr_domainname));
1741 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1743 pr->pr_hostid = hid;
1744 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1745 if (tpr->pr_flags & PR_HOST)
1749 strlcpy(tpr->pr_hostname,
1751 sizeof(tpr->pr_hostname));
1753 strlcpy(tpr->pr_domainname,
1755 sizeof(tpr->pr_domainname));
1757 strlcpy(tpr->pr_hostuuid,
1759 sizeof(tpr->pr_hostuuid));
1761 tpr->pr_hostid = hid;
1765 if ((tallow = ch_allow & ~pr_allow)) {
1766 /* Clear allow bits in all children. */
1767 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1768 tpr->pr_allow &= ~tallow;
1770 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1772 * Persistent prisons get an extra reference, and prisons losing their
1773 * persist flag lose that reference. Only do this for existing prisons
1774 * for now, so new ones will remain unseen until after the module
1775 * handlers have completed.
1777 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1778 if (pr_flags & PR_PERSIST) {
1786 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1787 mtx_unlock(&pr->pr_mtx);
1790 if (racct_enable && created)
1791 prison_racct_attach(pr);
1794 /* Locks may have prevented a complete restriction of child IP
1795 * addresses. If so, allocate some more memory and try again.
1800 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1801 mtx_lock(&pr->pr_mtx);
1803 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1805 if (tpr->pr_flags & PR_VNET) {
1810 if (prison_restrict_ip4(tpr, ip4)) {
1817 mtx_unlock(&pr->pr_mtx);
1823 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1824 mtx_lock(&pr->pr_mtx);
1826 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1828 if (tpr->pr_flags & PR_VNET) {
1833 if (prison_restrict_ip6(tpr, ip6)) {
1840 mtx_unlock(&pr->pr_mtx);
1844 /* Let the modules do their work. */
1845 sx_downgrade(&allprison_lock);
1847 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1849 prison_deref(pr, PD_LIST_SLOCKED);
1853 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1855 prison_deref(pr, created
1857 : PD_DEREF | PD_LIST_SLOCKED);
1861 /* Attach this process to the prison if requested. */
1862 if (flags & JAIL_ATTACH) {
1863 mtx_lock(&pr->pr_mtx);
1864 error = do_jail_attach(td, pr);
1866 vfs_opterror(opts, "attach failed");
1868 prison_deref(pr, PD_DEREF);
1874 if (racct_enable && !created) {
1875 if (!(flags & JAIL_ATTACH))
1876 sx_sunlock(&allprison_lock);
1877 prison_racct_modify(pr);
1878 if (!(flags & JAIL_ATTACH))
1879 sx_slock(&allprison_lock);
1883 td->td_retval[0] = pr->pr_id;
1886 * Now that it is all there, drop the temporary reference from existing
1887 * prisons. Or add a reference to newly created persistent prisons
1888 * (which was not done earlier so that the prison would not be publicly
1892 prison_deref(pr, (flags & JAIL_ATTACH)
1894 : PD_DEREF | PD_LIST_SLOCKED);
1896 if (pr_flags & PR_PERSIST) {
1897 mtx_lock(&pr->pr_mtx);
1900 mtx_unlock(&pr->pr_mtx);
1902 if (!(flags & JAIL_ATTACH))
1903 sx_sunlock(&allprison_lock);
1909 prison_deref(pr, created
1910 ? PD_LOCKED | PD_LIST_XLOCKED
1911 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1914 sx_xunlock(&allprison_lock);
1920 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1921 if (errmsg_len > 0) {
1922 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1923 if (errmsg_pos > 0) {
1924 if (optuio->uio_segflg == UIO_SYSSPACE)
1926 optuio->uio_iov[errmsg_pos].iov_base,
1930 optuio->uio_iov[errmsg_pos].iov_base,
1937 free(ip4, M_PRISON);
1940 free(ip6, M_PRISON);
1943 free(g_path, M_TEMP);
1950 * struct jail_get_args {
1951 * struct iovec *iovp;
1952 * unsigned int iovcnt;
1957 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1962 /* Check that we have an even number of iovecs. */
1963 if (uap->iovcnt & 1)
1966 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1969 error = kern_jail_get(td, auio, uap->flags);
1971 error = copyout(auio->uio_iov, uap->iovp,
1972 uap->iovcnt * sizeof (struct iovec));
1978 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1980 struct prison *pr, *mypr;
1982 struct vfsoptlist *opts;
1983 char *errmsg, *name;
1984 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1986 if (flags & ~JAIL_GET_MASK)
1989 /* Get the parameter list. */
1990 error = vfs_buildopts(optuio, &opts);
1993 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1994 mypr = td->td_ucred->cr_prison;
1997 * Find the prison specified by one of: lastjid, jid, name.
1999 sx_slock(&allprison_lock);
2000 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2002 TAILQ_FOREACH(pr, &allprison, pr_list) {
2003 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2004 mtx_lock(&pr->pr_mtx);
2005 if (pr->pr_ref > 0 &&
2006 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2008 mtx_unlock(&pr->pr_mtx);
2014 vfs_opterror(opts, "no jail after %d", jid);
2015 goto done_unlock_list;
2016 } else if (error != ENOENT)
2017 goto done_unlock_list;
2019 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2022 pr = prison_find_child(mypr, jid);
2024 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2025 mtx_unlock(&pr->pr_mtx);
2027 vfs_opterror(opts, "jail %d is dying",
2029 goto done_unlock_list;
2034 vfs_opterror(opts, "jail %d not found", jid);
2035 goto done_unlock_list;
2037 } else if (error != ENOENT)
2038 goto done_unlock_list;
2040 error = vfs_getopt(opts, "name", (void **)&name, &len);
2042 if (len == 0 || name[len - 1] != '\0') {
2044 goto done_unlock_list;
2046 pr = prison_find_name(mypr, name);
2048 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2049 mtx_unlock(&pr->pr_mtx);
2051 vfs_opterror(opts, "jail \"%s\" is dying",
2053 goto done_unlock_list;
2058 vfs_opterror(opts, "jail \"%s\" not found", name);
2059 goto done_unlock_list;
2060 } else if (error != ENOENT)
2061 goto done_unlock_list;
2063 vfs_opterror(opts, "no jail specified");
2065 goto done_unlock_list;
2068 /* Get the parameters of the prison. */
2071 td->td_retval[0] = pr->pr_id;
2072 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2073 if (error != 0 && error != ENOENT)
2075 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2076 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2077 if (error != 0 && error != ENOENT)
2079 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2080 if (error != 0 && error != ENOENT)
2082 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2083 sizeof(pr->pr_cpuset->cs_id));
2084 if (error != 0 && error != ENOENT)
2086 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2087 if (error != 0 && error != ENOENT)
2090 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2091 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2092 if (error != 0 && error != ENOENT)
2096 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2097 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2098 if (error != 0 && error != ENOENT)
2101 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2102 sizeof(pr->pr_securelevel));
2103 if (error != 0 && error != ENOENT)
2105 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2106 sizeof(pr->pr_childcount));
2107 if (error != 0 && error != ENOENT)
2109 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2110 sizeof(pr->pr_childmax));
2111 if (error != 0 && error != ENOENT)
2113 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2114 if (error != 0 && error != ENOENT)
2116 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2117 if (error != 0 && error != ENOENT)
2119 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2120 if (error != 0 && error != ENOENT)
2122 #ifdef COMPAT_FREEBSD32
2123 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2124 uint32_t hid32 = pr->pr_hostid;
2126 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2129 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2130 sizeof(pr->pr_hostid));
2131 if (error != 0 && error != ENOENT)
2133 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2134 sizeof(pr->pr_enforce_statfs));
2135 if (error != 0 && error != ENOENT)
2137 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2138 sizeof(pr->pr_devfs_rsnum));
2139 if (error != 0 && error != ENOENT)
2141 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2143 if (pr_flag_names[fi] == NULL)
2145 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2146 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2147 if (error != 0 && error != ENOENT)
2150 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2151 if (error != 0 && error != ENOENT)
2154 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2157 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2158 i = pr_flag_jailsys[fi].disable &&
2159 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2160 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2163 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2164 if (error != 0 && error != ENOENT)
2167 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2169 if (pr_allow_names[fi] == NULL)
2171 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2172 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2173 if (error != 0 && error != ENOENT)
2176 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2177 if (error != 0 && error != ENOENT)
2180 i = (pr->pr_uref == 0);
2181 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2182 if (error != 0 && error != ENOENT)
2185 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2186 if (error != 0 && error != ENOENT)
2188 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2189 sizeof(pr->pr_osreldate));
2190 if (error != 0 && error != ENOENT)
2192 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2193 if (error != 0 && error != ENOENT)
2196 /* Get the module parameters. */
2197 mtx_unlock(&pr->pr_mtx);
2199 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2202 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2204 /* By now, all parameters should have been noted. */
2205 TAILQ_FOREACH(opt, opts, link) {
2206 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2208 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2213 /* Write the fetched parameters back to userspace. */
2215 TAILQ_FOREACH(opt, opts, link) {
2216 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2217 pos = 2 * opt->pos + 1;
2218 optuio->uio_iov[pos].iov_len = opt->len;
2219 if (opt->value != NULL) {
2220 if (optuio->uio_segflg == UIO_SYSSPACE) {
2222 optuio->uio_iov[pos].iov_base,
2225 error = copyout(opt->value,
2226 optuio->uio_iov[pos].iov_base,
2237 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2241 sx_sunlock(&allprison_lock);
2243 if (error && errmsg_pos >= 0) {
2244 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2245 errmsg_pos = 2 * errmsg_pos + 1;
2246 if (errmsg_len > 0) {
2247 if (optuio->uio_segflg == UIO_SYSSPACE)
2249 optuio->uio_iov[errmsg_pos].iov_base,
2253 optuio->uio_iov[errmsg_pos].iov_base,
2263 * struct jail_remove_args {
2268 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2270 struct prison *pr, *cpr, *lpr, *tpr;
2273 error = priv_check(td, PRIV_JAIL_REMOVE);
2277 sx_xlock(&allprison_lock);
2278 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2280 sx_xunlock(&allprison_lock);
2284 /* Remove all descendants of this prison, then remove this prison. */
2286 pr->pr_flags |= PR_REMOVE;
2287 if (!LIST_EMPTY(&pr->pr_children)) {
2288 mtx_unlock(&pr->pr_mtx);
2290 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2291 mtx_lock(&cpr->pr_mtx);
2292 if (cpr->pr_ref > 0) {
2295 cpr->pr_flags |= PR_REMOVE;
2297 /* Already removed - do not do it again. */
2300 mtx_unlock(&cpr->pr_mtx);
2302 mtx_lock(&lpr->pr_mtx);
2303 prison_remove_one(lpr);
2304 sx_xlock(&allprison_lock);
2309 mtx_lock(&lpr->pr_mtx);
2310 prison_remove_one(lpr);
2311 sx_xlock(&allprison_lock);
2313 mtx_lock(&pr->pr_mtx);
2315 prison_remove_one(pr);
2320 prison_remove_one(struct prison *pr)
2325 /* If the prison was persistent, it is not anymore. */
2327 if (pr->pr_flags & PR_PERSIST) {
2330 pr->pr_flags &= ~PR_PERSIST;
2334 * jail_remove added a reference. If that's the only one, remove
2337 KASSERT(pr->pr_ref > 0,
2338 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2339 if (pr->pr_ref == 1) {
2341 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2345 mtx_unlock(&pr->pr_mtx);
2346 sx_xunlock(&allprison_lock);
2348 * Kill all processes unfortunate enough to be attached to this prison.
2350 sx_slock(&allproc_lock);
2351 LIST_FOREACH(p, &allproc, p_list) {
2353 if (p->p_state != PRS_NEW && p->p_ucred &&
2354 p->p_ucred->cr_prison == pr)
2355 kern_psignal(p, SIGKILL);
2358 sx_sunlock(&allproc_lock);
2359 /* Remove the temporary reference added by jail_remove. */
2360 prison_deref(pr, deuref | PD_DEREF);
2365 * struct jail_attach_args {
2370 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2375 error = priv_check(td, PRIV_JAIL_ATTACH);
2379 sx_slock(&allprison_lock);
2380 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2382 sx_sunlock(&allprison_lock);
2387 * Do not allow a process to attach to a prison that is not
2388 * considered to be "alive".
2390 if (pr->pr_uref == 0) {
2391 mtx_unlock(&pr->pr_mtx);
2392 sx_sunlock(&allprison_lock);
2396 return (do_jail_attach(td, pr));
2400 do_jail_attach(struct thread *td, struct prison *pr)
2404 struct ucred *newcred, *oldcred;
2408 * XXX: Note that there is a slight race here if two threads
2409 * in the same privileged process attempt to attach to two
2410 * different jails at the same time. It is important for
2411 * user processes not to do this, or they might end up with
2412 * a process root from one prison, but attached to the jail
2417 mtx_unlock(&pr->pr_mtx);
2419 /* Let modules do whatever they need to prepare for attaching. */
2420 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2422 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2425 sx_sunlock(&allprison_lock);
2428 * Reparent the newly attached process to this jail.
2430 ppr = td->td_ucred->cr_prison;
2432 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2436 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2437 if ((error = change_dir(pr->pr_root, td)) != 0)
2440 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2443 VOP_UNLOCK(pr->pr_root, 0);
2444 if ((error = change_root(pr->pr_root, td)))
2449 oldcred = p->p_ucred;
2451 crcopy(newcred, oldcred);
2452 newcred->cr_prison = pr;
2453 p->p_ucred = newcred;
2456 racct_proc_ucred_changed(p, oldcred, newcred);
2459 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2462 VOP_UNLOCK(pr->pr_root, 0);
2464 /* Tell modules this thread is still in its old jail after all. */
2465 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2466 prison_deref(pr, PD_DEREF | PD_DEUREF);
2472 * Returns a locked prison instance, or NULL on failure.
2475 prison_find(int prid)
2479 sx_assert(&allprison_lock, SX_LOCKED);
2480 TAILQ_FOREACH(pr, &allprison, pr_list) {
2481 if (pr->pr_id == prid) {
2482 mtx_lock(&pr->pr_mtx);
2485 mtx_unlock(&pr->pr_mtx);
2492 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2495 prison_find_child(struct prison *mypr, int prid)
2500 sx_assert(&allprison_lock, SX_LOCKED);
2501 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2502 if (pr->pr_id == prid) {
2503 mtx_lock(&pr->pr_mtx);
2506 mtx_unlock(&pr->pr_mtx);
2513 * Look for the name relative to mypr. Returns a locked prison or NULL.
2516 prison_find_name(struct prison *mypr, const char *name)
2518 struct prison *pr, *deadpr;
2522 sx_assert(&allprison_lock, SX_LOCKED);
2523 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2526 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2527 if (!strcmp(pr->pr_name + mylen, name)) {
2528 mtx_lock(&pr->pr_mtx);
2529 if (pr->pr_ref > 0) {
2530 if (pr->pr_uref > 0)
2534 mtx_unlock(&pr->pr_mtx);
2537 /* There was no valid prison - perhaps there was a dying one. */
2538 if (deadpr != NULL) {
2539 mtx_lock(&deadpr->pr_mtx);
2540 if (deadpr->pr_ref == 0) {
2541 mtx_unlock(&deadpr->pr_mtx);
2549 * See if a prison has the specific flag set.
2552 prison_flag(struct ucred *cred, unsigned flag)
2555 /* This is an atomic read, so no locking is necessary. */
2556 return (cred->cr_prison->pr_flags & flag);
2560 prison_allow(struct ucred *cred, unsigned flag)
2563 /* This is an atomic read, so no locking is necessary. */
2564 return (cred->cr_prison->pr_allow & flag);
2568 * Remove a prison reference. If that was the last reference, remove the
2569 * prison itself - but not in this context in case there are locks held.
2572 prison_free_locked(struct prison *pr)
2575 mtx_assert(&pr->pr_mtx, MA_OWNED);
2577 if (pr->pr_ref == 0) {
2578 mtx_unlock(&pr->pr_mtx);
2579 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2580 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2583 mtx_unlock(&pr->pr_mtx);
2587 prison_free(struct prison *pr)
2590 mtx_lock(&pr->pr_mtx);
2591 prison_free_locked(pr);
2595 prison_complete(void *context, int pending)
2598 prison_deref((struct prison *)context, 0);
2602 * Remove a prison reference (usually). This internal version assumes no
2603 * mutexes are held, except perhaps the prison itself. If there are no more
2604 * references, release and delist the prison. On completion, the prison lock
2605 * and the allprison lock are both unlocked.
2608 prison_deref(struct prison *pr, int flags)
2610 struct prison *ppr, *tpr;
2612 if (!(flags & PD_LOCKED))
2613 mtx_lock(&pr->pr_mtx);
2615 if (flags & PD_DEUREF) {
2617 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2619 if (flags & PD_DEREF)
2621 /* If the prison still has references, nothing else to do. */
2622 if (pr->pr_ref > 0) {
2623 mtx_unlock(&pr->pr_mtx);
2624 if (flags & PD_LIST_SLOCKED)
2625 sx_sunlock(&allprison_lock);
2626 else if (flags & PD_LIST_XLOCKED)
2627 sx_xunlock(&allprison_lock);
2631 mtx_unlock(&pr->pr_mtx);
2632 if (flags & PD_LIST_SLOCKED) {
2633 if (!sx_try_upgrade(&allprison_lock)) {
2634 sx_sunlock(&allprison_lock);
2635 sx_xlock(&allprison_lock);
2637 } else if (!(flags & PD_LIST_XLOCKED))
2638 sx_xlock(&allprison_lock);
2640 TAILQ_REMOVE(&allprison, pr, pr_list);
2641 LIST_REMOVE(pr, pr_sibling);
2642 ppr = pr->pr_parent;
2643 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2644 tpr->pr_childcount--;
2645 sx_xunlock(&allprison_lock);
2648 if (pr->pr_vnet != ppr->pr_vnet)
2649 vnet_destroy(pr->pr_vnet);
2651 if (pr->pr_root != NULL)
2653 mtx_destroy(&pr->pr_mtx);
2655 free(pr->pr_ip4, M_PRISON);
2658 free(pr->pr_ip6, M_PRISON);
2660 if (pr->pr_cpuset != NULL)
2661 cpuset_rel(pr->pr_cpuset);
2665 prison_racct_detach(pr);
2669 /* Removing a prison frees a reference on its parent. */
2671 mtx_lock(&pr->pr_mtx);
2672 flags = PD_DEREF | PD_DEUREF;
2677 prison_hold_locked(struct prison *pr)
2680 mtx_assert(&pr->pr_mtx, MA_OWNED);
2681 KASSERT(pr->pr_ref > 0,
2682 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2687 prison_hold(struct prison *pr)
2690 mtx_lock(&pr->pr_mtx);
2691 prison_hold_locked(pr);
2692 mtx_unlock(&pr->pr_mtx);
2696 prison_proc_hold(struct prison *pr)
2699 mtx_lock(&pr->pr_mtx);
2700 KASSERT(pr->pr_uref > 0,
2701 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2703 mtx_unlock(&pr->pr_mtx);
2707 prison_proc_free(struct prison *pr)
2710 mtx_lock(&pr->pr_mtx);
2711 KASSERT(pr->pr_uref > 0,
2712 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2713 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2719 * Restrict a prison's IP address list with its parent's, possibly replacing
2720 * it. Return true if the replacement buffer was used (or would have been).
2723 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2728 ppr = pr->pr_parent;
2729 if (!(pr->pr_flags & PR_IP4_USER)) {
2730 /* This has no user settings, so just copy the parent's list. */
2731 if (pr->pr_ip4s < ppr->pr_ip4s) {
2733 * There's no room for the parent's list. Use the
2734 * new list buffer, which is assumed to be big enough
2735 * (if it was passed). If there's no buffer, try to
2739 if (newip4 == NULL) {
2740 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2741 M_PRISON, M_NOWAIT);
2745 if (newip4 != NULL) {
2746 bcopy(ppr->pr_ip4, newip4,
2747 ppr->pr_ip4s * sizeof(*newip4));
2748 free(pr->pr_ip4, M_PRISON);
2749 pr->pr_ip4 = newip4;
2750 pr->pr_ip4s = ppr->pr_ip4s;
2754 pr->pr_ip4s = ppr->pr_ip4s;
2755 if (pr->pr_ip4s > 0)
2756 bcopy(ppr->pr_ip4, pr->pr_ip4,
2757 pr->pr_ip4s * sizeof(*newip4));
2758 else if (pr->pr_ip4 != NULL) {
2759 free(pr->pr_ip4, M_PRISON);
2762 } else if (pr->pr_ip4s > 0) {
2763 /* Remove addresses that aren't in the parent. */
2764 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2765 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2767 if (ij < ppr->pr_ip4s)
2770 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2771 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2774 for (ij = 1; ii < pr->pr_ip4s; ) {
2775 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2779 switch (ij >= ppr->pr_ip4s ? -1 :
2780 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2782 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2783 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2794 if (pr->pr_ip4s == 0) {
2795 pr->pr_flags |= PR_IP4_DISABLE;
2796 free(pr->pr_ip4, M_PRISON);
2804 * Pass back primary IPv4 address of this jail.
2806 * If not restricted return success but do not alter the address. Caller has
2807 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2809 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2810 * Address returned in NBO.
2813 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2817 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2818 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2820 pr = cred->cr_prison;
2821 if (!(pr->pr_flags & PR_IP4))
2823 mtx_lock(&pr->pr_mtx);
2824 if (!(pr->pr_flags & PR_IP4)) {
2825 mtx_unlock(&pr->pr_mtx);
2828 if (pr->pr_ip4 == NULL) {
2829 mtx_unlock(&pr->pr_mtx);
2830 return (EAFNOSUPPORT);
2833 ia->s_addr = pr->pr_ip4[0].s_addr;
2834 mtx_unlock(&pr->pr_mtx);
2839 * Return 1 if we should do proper source address selection or are not jailed.
2840 * We will return 0 if we should bypass source address selection in favour
2841 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2843 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2846 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2852 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2853 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2858 pr = cred->cr_prison;
2859 if (pr->pr_flags & PR_IP4_SADDRSEL)
2862 lia.s_addr = INADDR_ANY;
2863 error = prison_get_ip4(cred, &lia);
2866 if (lia.s_addr == INADDR_ANY)
2869 ia->s_addr = lia.s_addr;
2874 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2877 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2884 * No need to lock since the PR_IP4_USER flag can't be altered for
2887 while (pr1 != &prison0 &&
2889 !(pr1->pr_flags & PR_VNET) &&
2891 !(pr1->pr_flags & PR_IP4_USER))
2892 pr1 = pr1->pr_parent;
2893 while (pr2 != &prison0 &&
2895 !(pr2->pr_flags & PR_VNET) &&
2897 !(pr2->pr_flags & PR_IP4_USER))
2898 pr2 = pr2->pr_parent;
2899 return (pr1 == pr2);
2903 * Make sure our (source) address is set to something meaningful to this
2906 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2907 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2908 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2911 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2917 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2918 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2920 pr = cred->cr_prison;
2921 if (!(pr->pr_flags & PR_IP4))
2923 mtx_lock(&pr->pr_mtx);
2924 if (!(pr->pr_flags & PR_IP4)) {
2925 mtx_unlock(&pr->pr_mtx);
2928 if (pr->pr_ip4 == NULL) {
2929 mtx_unlock(&pr->pr_mtx);
2930 return (EAFNOSUPPORT);
2933 ia0.s_addr = ntohl(ia->s_addr);
2934 if (ia0.s_addr == INADDR_LOOPBACK) {
2935 ia->s_addr = pr->pr_ip4[0].s_addr;
2936 mtx_unlock(&pr->pr_mtx);
2940 if (ia0.s_addr == INADDR_ANY) {
2942 * In case there is only 1 IPv4 address, bind directly.
2944 if (pr->pr_ip4s == 1)
2945 ia->s_addr = pr->pr_ip4[0].s_addr;
2946 mtx_unlock(&pr->pr_mtx);
2950 error = _prison_check_ip4(pr, ia);
2951 mtx_unlock(&pr->pr_mtx);
2956 * Rewrite destination address in case we will connect to loopback address.
2958 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2959 * Address passed in in NBO and returned in NBO.
2962 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2966 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2967 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2969 pr = cred->cr_prison;
2970 if (!(pr->pr_flags & PR_IP4))
2972 mtx_lock(&pr->pr_mtx);
2973 if (!(pr->pr_flags & PR_IP4)) {
2974 mtx_unlock(&pr->pr_mtx);
2977 if (pr->pr_ip4 == NULL) {
2978 mtx_unlock(&pr->pr_mtx);
2979 return (EAFNOSUPPORT);
2982 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2983 ia->s_addr = pr->pr_ip4[0].s_addr;
2984 mtx_unlock(&pr->pr_mtx);
2989 * Return success because nothing had to be changed.
2991 mtx_unlock(&pr->pr_mtx);
2996 * Check if given address belongs to the jail referenced by cred/prison.
2998 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2999 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3000 * doesn't allow IPv4. Address passed in in NBO.
3003 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
3008 * Check the primary IP.
3010 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3014 * All the other IPs are sorted so we can do a binary search.
3017 z = pr->pr_ip4s - 2;
3020 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3029 return (EADDRNOTAVAIL);
3033 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
3038 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3039 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3041 pr = cred->cr_prison;
3042 if (!(pr->pr_flags & PR_IP4))
3044 mtx_lock(&pr->pr_mtx);
3045 if (!(pr->pr_flags & PR_IP4)) {
3046 mtx_unlock(&pr->pr_mtx);
3049 if (pr->pr_ip4 == NULL) {
3050 mtx_unlock(&pr->pr_mtx);
3051 return (EAFNOSUPPORT);
3054 error = _prison_check_ip4(pr, ia);
3055 mtx_unlock(&pr->pr_mtx);
3062 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3067 ppr = pr->pr_parent;
3068 if (!(pr->pr_flags & PR_IP6_USER)) {
3069 /* This has no user settings, so just copy the parent's list. */
3070 if (pr->pr_ip6s < ppr->pr_ip6s) {
3072 * There's no room for the parent's list. Use the
3073 * new list buffer, which is assumed to be big enough
3074 * (if it was passed). If there's no buffer, try to
3078 if (newip6 == NULL) {
3079 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3080 M_PRISON, M_NOWAIT);
3084 if (newip6 != NULL) {
3085 bcopy(ppr->pr_ip6, newip6,
3086 ppr->pr_ip6s * sizeof(*newip6));
3087 free(pr->pr_ip6, M_PRISON);
3088 pr->pr_ip6 = newip6;
3089 pr->pr_ip6s = ppr->pr_ip6s;
3093 pr->pr_ip6s = ppr->pr_ip6s;
3094 if (pr->pr_ip6s > 0)
3095 bcopy(ppr->pr_ip6, pr->pr_ip6,
3096 pr->pr_ip6s * sizeof(*newip6));
3097 else if (pr->pr_ip6 != NULL) {
3098 free(pr->pr_ip6, M_PRISON);
3101 } else if (pr->pr_ip6s > 0) {
3102 /* Remove addresses that aren't in the parent. */
3103 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3104 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3107 if (ij < ppr->pr_ip6s)
3110 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3111 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3114 for (ij = 1; ii < pr->pr_ip6s; ) {
3115 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3120 switch (ij >= ppr->pr_ip6s ? -1 :
3121 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3123 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3124 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3135 if (pr->pr_ip6s == 0) {
3136 pr->pr_flags |= PR_IP6_DISABLE;
3137 free(pr->pr_ip6, M_PRISON);
3145 * Pass back primary IPv6 address for this jail.
3147 * If not restricted return success but do not alter the address. Caller has
3148 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3150 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3153 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3157 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3158 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3160 pr = cred->cr_prison;
3161 if (!(pr->pr_flags & PR_IP6))
3163 mtx_lock(&pr->pr_mtx);
3164 if (!(pr->pr_flags & PR_IP6)) {
3165 mtx_unlock(&pr->pr_mtx);
3168 if (pr->pr_ip6 == NULL) {
3169 mtx_unlock(&pr->pr_mtx);
3170 return (EAFNOSUPPORT);
3173 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3174 mtx_unlock(&pr->pr_mtx);
3179 * Return 1 if we should do proper source address selection or are not jailed.
3180 * We will return 0 if we should bypass source address selection in favour
3181 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3183 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3186 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3189 struct in6_addr lia6;
3192 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3193 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3198 pr = cred->cr_prison;
3199 if (pr->pr_flags & PR_IP6_SADDRSEL)
3203 error = prison_get_ip6(cred, &lia6);
3206 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3209 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3214 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3217 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3223 while (pr1 != &prison0 &&
3225 !(pr1->pr_flags & PR_VNET) &&
3227 !(pr1->pr_flags & PR_IP6_USER))
3228 pr1 = pr1->pr_parent;
3229 while (pr2 != &prison0 &&
3231 !(pr2->pr_flags & PR_VNET) &&
3233 !(pr2->pr_flags & PR_IP6_USER))
3234 pr2 = pr2->pr_parent;
3235 return (pr1 == pr2);
3239 * Make sure our (source) address is set to something meaningful to this jail.
3241 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3242 * when needed while binding.
3244 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3245 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3246 * doesn't allow IPv6.
3249 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3254 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3255 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3257 pr = cred->cr_prison;
3258 if (!(pr->pr_flags & PR_IP6))
3260 mtx_lock(&pr->pr_mtx);
3261 if (!(pr->pr_flags & PR_IP6)) {
3262 mtx_unlock(&pr->pr_mtx);
3265 if (pr->pr_ip6 == NULL) {
3266 mtx_unlock(&pr->pr_mtx);
3267 return (EAFNOSUPPORT);
3270 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3271 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3272 mtx_unlock(&pr->pr_mtx);
3276 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3278 * In case there is only 1 IPv6 address, and v6only is true,
3279 * then bind directly.
3281 if (v6only != 0 && pr->pr_ip6s == 1)
3282 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3283 mtx_unlock(&pr->pr_mtx);
3287 error = _prison_check_ip6(pr, ia6);
3288 mtx_unlock(&pr->pr_mtx);
3293 * Rewrite destination address in case we will connect to loopback address.
3295 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3298 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3302 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3303 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3305 pr = cred->cr_prison;
3306 if (!(pr->pr_flags & PR_IP6))
3308 mtx_lock(&pr->pr_mtx);
3309 if (!(pr->pr_flags & PR_IP6)) {
3310 mtx_unlock(&pr->pr_mtx);
3313 if (pr->pr_ip6 == NULL) {
3314 mtx_unlock(&pr->pr_mtx);
3315 return (EAFNOSUPPORT);
3318 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3319 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3320 mtx_unlock(&pr->pr_mtx);
3325 * Return success because nothing had to be changed.
3327 mtx_unlock(&pr->pr_mtx);
3332 * Check if given address belongs to the jail referenced by cred/prison.
3334 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3335 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3336 * doesn't allow IPv6.
3339 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3344 * Check the primary IP.
3346 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3350 * All the other IPs are sorted so we can do a binary search.
3353 z = pr->pr_ip6s - 2;
3356 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3365 return (EADDRNOTAVAIL);
3369 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3374 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3375 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3377 pr = cred->cr_prison;
3378 if (!(pr->pr_flags & PR_IP6))
3380 mtx_lock(&pr->pr_mtx);
3381 if (!(pr->pr_flags & PR_IP6)) {
3382 mtx_unlock(&pr->pr_mtx);
3385 if (pr->pr_ip6 == NULL) {
3386 mtx_unlock(&pr->pr_mtx);
3387 return (EAFNOSUPPORT);
3390 error = _prison_check_ip6(pr, ia6);
3391 mtx_unlock(&pr->pr_mtx);
3397 * Check if a jail supports the given address family.
3399 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3403 prison_check_af(struct ucred *cred, int af)
3408 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3410 pr = cred->cr_prison;
3412 /* Prisons with their own network stack are not limited. */
3413 if (prison_owns_vnet(cred))
3422 if (pr->pr_flags & PR_IP4)
3424 mtx_lock(&pr->pr_mtx);
3425 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3426 error = EAFNOSUPPORT;
3427 mtx_unlock(&pr->pr_mtx);
3433 if (pr->pr_flags & PR_IP6)
3435 mtx_lock(&pr->pr_mtx);
3436 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3437 error = EAFNOSUPPORT;
3438 mtx_unlock(&pr->pr_mtx);
3446 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3447 error = EAFNOSUPPORT;
3453 * Check if given address belongs to the jail referenced by cred (wrapper to
3454 * prison_check_ip[46]).
3456 * Returns 0 if jail doesn't restrict the address family or if address belongs
3457 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3458 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3461 prison_if(struct ucred *cred, struct sockaddr *sa)
3464 struct sockaddr_in *sai;
3467 struct sockaddr_in6 *sai6;
3471 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3472 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3475 if (prison_owns_vnet(cred))
3480 switch (sa->sa_family)
3484 sai = (struct sockaddr_in *)sa;
3485 error = prison_check_ip4(cred, &sai->sin_addr);
3490 sai6 = (struct sockaddr_in6 *)sa;
3491 error = prison_check_ip6(cred, &sai6->sin6_addr);
3495 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3496 error = EAFNOSUPPORT;
3502 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3505 prison_check(struct ucred *cred1, struct ucred *cred2)
3508 return ((cred1->cr_prison == cred2->cr_prison ||
3509 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3513 * Return 1 if p2 is a child of p1, otherwise 0.
3516 prison_ischild(struct prison *pr1, struct prison *pr2)
3519 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3526 * Return 1 if the passed credential is in a jail, otherwise 0.
3529 jailed(struct ucred *cred)
3532 return (cred->cr_prison != &prison0);
3536 * Return 1 if the passed credential is in a jail and that jail does not
3537 * have its own virtual network stack, otherwise 0.
3540 jailed_without_vnet(struct ucred *cred)
3546 if (prison_owns_vnet(cred))
3554 * Return the correct hostname (domainname, et al) for the passed credential.
3557 getcredhostname(struct ucred *cred, char *buf, size_t size)
3562 * A NULL credential can be used to shortcut to the physical
3563 * system's hostname.
3565 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3566 mtx_lock(&pr->pr_mtx);
3567 strlcpy(buf, pr->pr_hostname, size);
3568 mtx_unlock(&pr->pr_mtx);
3572 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3575 mtx_lock(&cred->cr_prison->pr_mtx);
3576 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3577 mtx_unlock(&cred->cr_prison->pr_mtx);
3581 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3584 mtx_lock(&cred->cr_prison->pr_mtx);
3585 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3586 mtx_unlock(&cred->cr_prison->pr_mtx);
3590 getcredhostid(struct ucred *cred, unsigned long *hostid)
3593 mtx_lock(&cred->cr_prison->pr_mtx);
3594 *hostid = cred->cr_prison->pr_hostid;
3595 mtx_unlock(&cred->cr_prison->pr_mtx);
3600 * Determine whether the prison represented by cred owns
3601 * its vnet rather than having it inherited.
3603 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3606 prison_owns_vnet(struct ucred *cred)
3610 * vnets cannot be added/removed after jail creation,
3611 * so no need to lock here.
3613 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3618 * Determine whether the subject represented by cred can "see"
3619 * status of a mount point.
3620 * Returns: 0 for permitted, ENOENT otherwise.
3621 * XXX: This function should be called cr_canseemount() and should be
3622 * placed in kern_prot.c.
3625 prison_canseemount(struct ucred *cred, struct mount *mp)
3631 pr = cred->cr_prison;
3632 if (pr->pr_enforce_statfs == 0)
3634 if (pr->pr_root->v_mount == mp)
3636 if (pr->pr_enforce_statfs == 2)
3639 * If jail's chroot directory is set to "/" we should be able to see
3640 * all mount-points from inside a jail.
3641 * This is ugly check, but this is the only situation when jail's
3642 * directory ends with '/'.
3644 if (strcmp(pr->pr_path, "/") == 0)
3646 len = strlen(pr->pr_path);
3648 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3651 * Be sure that we don't have situation where jail's root directory
3652 * is "/some/path" and mount point is "/some/pathpath".
3654 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3660 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3662 char jpath[MAXPATHLEN];
3666 pr = cred->cr_prison;
3667 if (pr->pr_enforce_statfs == 0)
3669 if (prison_canseemount(cred, mp) != 0) {
3670 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3671 strlcpy(sp->f_mntonname, "[restricted]",
3672 sizeof(sp->f_mntonname));
3675 if (pr->pr_root->v_mount == mp) {
3677 * Clear current buffer data, so we are sure nothing from
3678 * the valid path left there.
3680 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3681 *sp->f_mntonname = '/';
3685 * If jail's chroot directory is set to "/" we should be able to see
3686 * all mount-points from inside a jail.
3688 if (strcmp(pr->pr_path, "/") == 0)
3690 len = strlen(pr->pr_path);
3691 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3693 * Clear current buffer data, so we are sure nothing from
3694 * the valid path left there.
3696 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3697 if (*jpath == '\0') {
3698 /* Should never happen. */
3699 *sp->f_mntonname = '/';
3701 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3706 * Check with permission for a specific privilege is granted within jail. We
3707 * have a specific list of accepted privileges; the rest are denied.
3710 prison_priv_check(struct ucred *cred, int priv)
3718 * Privileges specific to prisons with a virtual network stack.
3719 * There might be a duplicate entry here in case the privilege
3720 * is only granted conditionally in the legacy jail case.
3725 * NFS-specific privileges.
3727 case PRIV_NFS_DAEMON:
3728 case PRIV_NFS_LOCKD:
3731 * Network stack privileges.
3733 case PRIV_NET_BRIDGE:
3736 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3737 case PRIV_NET_ROUTE:
3739 case PRIV_NET_SETIFMTU:
3740 case PRIV_NET_SETIFFLAGS:
3741 case PRIV_NET_SETIFCAP:
3742 case PRIV_NET_SETIFDESCR:
3743 case PRIV_NET_SETIFNAME :
3744 case PRIV_NET_SETIFMETRIC:
3745 case PRIV_NET_SETIFPHYS:
3746 case PRIV_NET_SETIFMAC:
3747 case PRIV_NET_ADDMULTI:
3748 case PRIV_NET_DELMULTI:
3749 case PRIV_NET_HWIOCTL:
3750 case PRIV_NET_SETLLADDR:
3751 case PRIV_NET_ADDIFGROUP:
3752 case PRIV_NET_DELIFGROUP:
3753 case PRIV_NET_IFCREATE:
3754 case PRIV_NET_IFDESTROY:
3755 case PRIV_NET_ADDIFADDR:
3756 case PRIV_NET_DELIFADDR:
3759 case PRIV_NET_SETIFVNET:
3760 case PRIV_NET_SETIFFIB:
3763 * 802.11-related privileges.
3765 case PRIV_NET80211_GETKEY:
3767 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3772 * AppleTalk privileges.
3774 case PRIV_NETATALK_RESERVEDPORT:
3779 case PRIV_NETATM_CFG:
3780 case PRIV_NETATM_ADD:
3781 case PRIV_NETATM_DEL:
3782 case PRIV_NETATM_SET:
3785 * Bluetooth privileges.
3787 case PRIV_NETBLUETOOTH_RAW:
3791 * Netgraph and netgraph module privileges.
3793 case PRIV_NETGRAPH_CONTROL:
3795 case PRIV_NETGRAPH_TTY:
3799 * IPv4 and IPv6 privileges.
3801 case PRIV_NETINET_IPFW:
3802 case PRIV_NETINET_DIVERT:
3803 case PRIV_NETINET_PF:
3804 case PRIV_NETINET_DUMMYNET:
3805 case PRIV_NETINET_CARP:
3806 case PRIV_NETINET_MROUTE:
3807 case PRIV_NETINET_RAW:
3808 case PRIV_NETINET_ADDRCTRL6:
3809 case PRIV_NETINET_ND6:
3810 case PRIV_NETINET_SCOPE6:
3811 case PRIV_NETINET_ALIFETIME6:
3812 case PRIV_NETINET_IPSEC:
3813 case PRIV_NETINET_BINDANY:
3817 * IPX/SPX privileges.
3819 case PRIV_NETIPX_RESERVEDPORT:
3820 case PRIV_NETIPX_RAW:
3834 * No default: or deny here.
3835 * In case of no permit fall through to next switch().
3837 if (cred->cr_prison->pr_flags & PR_VNET)
3845 * Allow ktrace privileges for root in jail.
3851 * Allow jailed processes to configure audit identity and
3852 * submit audit records (login, etc). In the future we may
3853 * want to further refine the relationship between audit and
3856 case PRIV_AUDIT_GETAUDIT:
3857 case PRIV_AUDIT_SETAUDIT:
3858 case PRIV_AUDIT_SUBMIT:
3862 * Allow jailed processes to manipulate process UNIX
3863 * credentials in any way they see fit.
3865 case PRIV_CRED_SETUID:
3866 case PRIV_CRED_SETEUID:
3867 case PRIV_CRED_SETGID:
3868 case PRIV_CRED_SETEGID:
3869 case PRIV_CRED_SETGROUPS:
3870 case PRIV_CRED_SETREUID:
3871 case PRIV_CRED_SETREGID:
3872 case PRIV_CRED_SETRESUID:
3873 case PRIV_CRED_SETRESGID:
3876 * Jail implements visibility constraints already, so allow
3877 * jailed root to override uid/gid-based constraints.
3879 case PRIV_SEEOTHERGIDS:
3880 case PRIV_SEEOTHERUIDS:
3883 * Jail implements inter-process debugging limits already, so
3884 * allow jailed root various debugging privileges.
3886 case PRIV_DEBUG_DIFFCRED:
3887 case PRIV_DEBUG_SUGID:
3888 case PRIV_DEBUG_UNPRIV:
3891 * Allow jail to set various resource limits and login
3892 * properties, and for now, exceed process resource limits.
3894 case PRIV_PROC_LIMIT:
3895 case PRIV_PROC_SETLOGIN:
3896 case PRIV_PROC_SETRLIMIT:
3899 * System V and POSIX IPC privileges are granted in jail.
3902 case PRIV_IPC_WRITE:
3903 case PRIV_IPC_ADMIN:
3904 case PRIV_IPC_MSGSIZE:
3908 * Jail operations within a jail work on child jails.
3910 case PRIV_JAIL_ATTACH:
3912 case PRIV_JAIL_REMOVE:
3915 * Jail implements its own inter-process limits, so allow
3916 * root processes in jail to change scheduling on other
3917 * processes in the same jail. Likewise for signalling.
3919 case PRIV_SCHED_DIFFCRED:
3920 case PRIV_SCHED_CPUSET:
3921 case PRIV_SIGNAL_DIFFCRED:
3922 case PRIV_SIGNAL_SUGID:
3925 * Allow jailed processes to write to sysctls marked as jail
3928 case PRIV_SYSCTL_WRITEJAIL:
3931 * Allow root in jail to manage a variety of quota
3932 * properties. These should likely be conditional on a
3933 * configuration option.
3935 case PRIV_VFS_GETQUOTA:
3936 case PRIV_VFS_SETQUOTA:
3939 * Since Jail relies on chroot() to implement file system
3940 * protections, grant many VFS privileges to root in jail.
3941 * Be careful to exclude mount-related and NFS-related
3945 case PRIV_VFS_WRITE:
3946 case PRIV_VFS_ADMIN:
3948 case PRIV_VFS_LOOKUP:
3949 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3950 case PRIV_VFS_CHFLAGS_DEV:
3951 case PRIV_VFS_CHOWN:
3952 case PRIV_VFS_CHROOT:
3953 case PRIV_VFS_RETAINSUGID:
3954 case PRIV_VFS_FCHROOT:
3956 case PRIV_VFS_SETGID:
3958 case PRIV_VFS_STICKYFILE:
3961 * As in the non-jail case, non-root users are expected to be
3962 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3963 * exists in the jail and they have permission to access it).
3965 case PRIV_KMEM_READ:
3969 * Depending on the global setting, allow privilege of
3970 * setting system flags.
3972 case PRIV_VFS_SYSFLAGS:
3973 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3979 * Depending on the global setting, allow privilege of
3980 * mounting/unmounting file systems.
3982 case PRIV_VFS_MOUNT:
3983 case PRIV_VFS_UNMOUNT:
3984 case PRIV_VFS_MOUNT_NONUSER:
3985 case PRIV_VFS_MOUNT_OWNER:
3986 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3987 cred->cr_prison->pr_enforce_statfs < 2)
3993 * Allow jailed root to bind reserved ports and reuse in-use
3996 case PRIV_NETINET_RESERVEDPORT:
3997 case PRIV_NETINET_REUSEPORT:
4001 * Allow jailed root to set certian IPv4/6 (option) headers.
4003 case PRIV_NETINET_SETHDROPTS:
4007 * Conditionally allow creating raw sockets in jail.
4009 case PRIV_NETINET_RAW:
4010 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4016 * Since jail implements its own visibility limits on netstat
4017 * sysctls, allow getcred. This allows identd to work in
4020 case PRIV_NETINET_GETCRED:
4024 * Allow jailed root to set loginclass.
4026 case PRIV_PROC_SETLOGINCLASS:
4031 * In all remaining cases, deny the privilege request. This
4032 * includes almost all network privileges, many system
4033 * configuration privileges.
4040 * Return the part of pr2's name that is relative to pr1, or the whole name
4041 * if it does not directly follow.
4045 prison_name(struct prison *pr1, struct prison *pr2)
4049 /* Jails see themselves as "0" (if they see themselves at all). */
4052 name = pr2->pr_name;
4053 if (prison_ischild(pr1, pr2)) {
4055 * pr1 isn't locked (and allprison_lock may not be either)
4056 * so its length can't be counted on. But the number of dots
4057 * can be counted on - and counted.
4059 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4060 name = strchr(name, '.') + 1;
4066 * Return the part of pr2's path that is relative to pr1, or the whole path
4067 * if it does not directly follow.
4070 prison_path(struct prison *pr1, struct prison *pr2)
4072 char *path1, *path2;
4075 path1 = pr1->pr_path;
4076 path2 = pr2->pr_path;
4077 if (!strcmp(path1, "/"))
4079 len1 = strlen(path1);
4080 if (strncmp(path1, path2, len1))
4082 if (path2[len1] == '\0')
4084 if (path2[len1] == '/')
4085 return (path2 + len1);
4091 * Jail-related sysctls.
4093 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4097 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4100 struct prison *pr, *cpr;
4102 struct in_addr *ip4 = NULL;
4106 struct in6_addr *ip6 = NULL;
4111 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4112 pr = req->td->td_ucred->cr_prison;
4114 sx_slock(&allprison_lock);
4115 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4116 #if defined(INET) || defined(INET6)
4119 mtx_lock(&cpr->pr_mtx);
4121 if (cpr->pr_ip4s > 0) {
4122 if (ip4s < cpr->pr_ip4s) {
4123 ip4s = cpr->pr_ip4s;
4124 mtx_unlock(&cpr->pr_mtx);
4125 ip4 = realloc(ip4, ip4s *
4126 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4129 bcopy(cpr->pr_ip4, ip4,
4130 cpr->pr_ip4s * sizeof(struct in_addr));
4134 if (cpr->pr_ip6s > 0) {
4135 if (ip6s < cpr->pr_ip6s) {
4136 ip6s = cpr->pr_ip6s;
4137 mtx_unlock(&cpr->pr_mtx);
4138 ip6 = realloc(ip6, ip6s *
4139 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4142 bcopy(cpr->pr_ip6, ip6,
4143 cpr->pr_ip6s * sizeof(struct in6_addr));
4146 if (cpr->pr_ref == 0) {
4147 mtx_unlock(&cpr->pr_mtx);
4150 bzero(xp, sizeof(*xp));
4151 xp->pr_version = XPRISON_VERSION;
4152 xp->pr_id = cpr->pr_id;
4153 xp->pr_state = cpr->pr_uref > 0
4154 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4155 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4156 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4157 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4159 xp->pr_ip4s = cpr->pr_ip4s;
4162 xp->pr_ip6s = cpr->pr_ip6s;
4164 mtx_unlock(&cpr->pr_mtx);
4165 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4169 if (xp->pr_ip4s > 0) {
4170 error = SYSCTL_OUT(req, ip4,
4171 xp->pr_ip4s * sizeof(struct in_addr));
4177 if (xp->pr_ip6s > 0) {
4178 error = SYSCTL_OUT(req, ip6,
4179 xp->pr_ip6s * sizeof(struct in6_addr));
4185 sx_sunlock(&allprison_lock);
4196 SYSCTL_OID(_security_jail, OID_AUTO, list,
4197 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4198 sysctl_jail_list, "S", "List of active jails");
4201 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4205 injail = jailed(req->td->td_ucred);
4206 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4211 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4212 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4213 sysctl_jail_jailed, "I", "Process in jail?");
4216 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4218 int error, havevnet;
4220 struct ucred *cred = req->td->td_ucred;
4222 havevnet = jailed(cred) && prison_owns_vnet(cred);
4226 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4231 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4232 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4233 sysctl_jail_vnet, "I", "Jail owns VNET?");
4235 #if defined(INET) || defined(INET6)
4236 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4237 &jail_max_af_ips, 0,
4238 "Number of IP addresses a jail may have at most per address family");
4242 * Default parameters for jail(2) compatability. For historical reasons,
4243 * the sysctl names have varying similarity to the parameter names. Prisons
4244 * just see their own parameters, and can't change them.
4247 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4250 int allow, error, i;
4252 pr = req->td->td_ucred->cr_prison;
4253 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4255 /* Get the current flag value, and convert it to a boolean. */
4256 i = (allow & arg2) ? 1 : 0;
4259 error = sysctl_handle_int(oidp, &i, 0, req);
4260 if (error || !req->newptr)
4266 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4269 mtx_lock(&prison0.pr_mtx);
4270 jail_default_allow = (jail_default_allow & ~arg2) | i;
4271 mtx_unlock(&prison0.pr_mtx);
4275 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4276 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4277 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4278 "Processes in jail can set their hostnames");
4279 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4280 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4281 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4282 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4283 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4284 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4285 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4286 "Processes in jail can use System V IPC primitives");
4287 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4288 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4289 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4290 "Prison root can create raw sockets");
4291 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4292 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4293 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4294 "Processes in jail can alter system file flags");
4295 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4296 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4297 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4298 "Processes in jail can mount/unmount jail-friendly file systems");
4299 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4300 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4301 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4302 "Processes in jail can mount the devfs file system");
4303 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4304 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4305 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4306 "Processes in jail can mount the fdescfs file system");
4307 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4308 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4309 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4310 "Processes in jail can mount the nullfs file system");
4311 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4312 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4313 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4314 "Processes in jail can mount the procfs file system");
4315 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4316 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4317 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4318 "Processes in jail can mount the tmpfs file system");
4319 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4320 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4321 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4322 "Processes in jail can mount the zfs file system");
4325 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4330 pr = req->td->td_ucred->cr_prison;
4331 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4332 error = sysctl_handle_int(oidp, &level, 0, req);
4333 if (error || !req->newptr)
4335 *(int *)arg1 = level;
4339 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4340 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4341 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4342 sysctl_jail_default_level, "I",
4343 "Processes in jail cannot see all mounted file systems");
4345 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4346 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4347 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4348 sysctl_jail_default_level, "I",
4349 "Ruleset for the devfs filesystem in jail");
4352 * Nodes to describe jail parameters. Maximum length of string parameters
4353 * is returned in the string itself, and the other parameters exist merely
4354 * to make themselves and their types known.
4356 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4360 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4367 switch (oidp->oid_kind & CTLTYPE)
4373 if (!(req->flags & SCTL_MASK32))
4375 return (SYSCTL_OUT(req, &l, sizeof(l)));
4379 return (SYSCTL_OUT(req, &i, sizeof(i)));
4380 case CTLTYPE_STRING:
4381 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4383 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4384 case CTLTYPE_STRUCT:
4386 return (SYSCTL_OUT(req, &s, sizeof(s)));
4392 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4393 * jail creation time but cannot be changed in an existing jail.
4395 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4396 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4397 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4398 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4399 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4400 "I", "Jail secure level");
4401 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4402 "Jail value for kern.osreldate and uname -K");
4403 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4404 "Jail value for kern.osrelease and uname -r");
4405 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4406 "I", "Jail cannot see all mounted file systems");
4407 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4408 "I", "Ruleset for in-jail devfs mounts");
4409 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4410 "B", "Jail persistence");
4412 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4413 "E,jailsys", "Virtual network stack");
4415 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4416 "B", "Jail is in the process of shutting down");
4418 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4419 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4420 "I", "Current number of child jails");
4421 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4422 "I", "Maximum number of child jails");
4424 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4425 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4427 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4428 "Jail NIS domainname");
4429 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4431 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4432 "LU", "Jail host ID");
4434 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4435 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4438 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4439 "Jail IPv4 address virtualization");
4440 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4441 "S,in_addr,a", "Jail IPv4 addresses");
4442 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4443 "B", "Do (not) use IPv4 source address selection rather than the "
4444 "primary jail IPv4 address.");
4447 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4448 "Jail IPv6 address virtualization");
4449 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4450 "S,in6_addr,a", "Jail IPv6 addresses");
4451 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4452 "B", "Do (not) use IPv6 source address selection rather than the "
4453 "primary jail IPv6 address.");
4456 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4457 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4458 "B", "Jail may set hostname");
4459 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4460 "B", "Jail may use SYSV IPC");
4461 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4462 "B", "Jail may create raw sockets");
4463 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4464 "B", "Jail may alter system file flags");
4465 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4466 "B", "Jail may set file quotas");
4467 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4468 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4470 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4471 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4472 "B", "Jail may mount/unmount jail-friendly file systems in general");
4473 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4474 "B", "Jail may mount the devfs file system");
4475 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4476 "B", "Jail may mount the fdescfs file system");
4477 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4478 "B", "Jail may mount the nullfs file system");
4479 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4480 "B", "Jail may mount the procfs file system");
4481 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4482 "B", "Jail may mount the tmpfs file system");
4483 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4484 "B", "Jail may mount the zfs file system");
4488 prison_racct_foreach(void (*callback)(struct racct *racct,
4489 void *arg2, void *arg3), void *arg2, void *arg3)
4491 struct prison_racct *prr;
4493 ASSERT_RACCT_ENABLED();
4495 sx_slock(&allprison_lock);
4496 LIST_FOREACH(prr, &allprison_racct, prr_next)
4497 (callback)(prr->prr_racct, arg2, arg3);
4498 sx_sunlock(&allprison_lock);
4501 static struct prison_racct *
4502 prison_racct_find_locked(const char *name)
4504 struct prison_racct *prr;
4506 ASSERT_RACCT_ENABLED();
4507 sx_assert(&allprison_lock, SA_XLOCKED);
4509 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4512 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4513 if (strcmp(name, prr->prr_name) != 0)
4516 /* Found prison_racct with a matching name? */
4517 prison_racct_hold(prr);
4521 /* Add new prison_racct. */
4522 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4523 racct_create(&prr->prr_racct);
4525 strcpy(prr->prr_name, name);
4526 refcount_init(&prr->prr_refcount, 1);
4527 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4532 struct prison_racct *
4533 prison_racct_find(const char *name)
4535 struct prison_racct *prr;
4537 ASSERT_RACCT_ENABLED();
4539 sx_xlock(&allprison_lock);
4540 prr = prison_racct_find_locked(name);
4541 sx_xunlock(&allprison_lock);
4546 prison_racct_hold(struct prison_racct *prr)
4549 ASSERT_RACCT_ENABLED();
4551 refcount_acquire(&prr->prr_refcount);
4555 prison_racct_free_locked(struct prison_racct *prr)
4558 ASSERT_RACCT_ENABLED();
4559 sx_assert(&allprison_lock, SA_XLOCKED);
4561 if (refcount_release(&prr->prr_refcount)) {
4562 racct_destroy(&prr->prr_racct);
4563 LIST_REMOVE(prr, prr_next);
4564 free(prr, M_PRISON_RACCT);
4569 prison_racct_free(struct prison_racct *prr)
4573 ASSERT_RACCT_ENABLED();
4574 sx_assert(&allprison_lock, SA_UNLOCKED);
4576 old = prr->prr_refcount;
4577 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4580 sx_xlock(&allprison_lock);
4581 prison_racct_free_locked(prr);
4582 sx_xunlock(&allprison_lock);
4586 prison_racct_attach(struct prison *pr)
4588 struct prison_racct *prr;
4590 ASSERT_RACCT_ENABLED();
4591 sx_assert(&allprison_lock, SA_XLOCKED);
4593 prr = prison_racct_find_locked(pr->pr_name);
4594 KASSERT(prr != NULL, ("cannot find prison_racct"));
4596 pr->pr_prison_racct = prr;
4600 * Handle jail renaming. From the racct point of view, renaming means
4601 * moving from one prison_racct to another.
4604 prison_racct_modify(struct prison *pr)
4608 struct prison_racct *oldprr;
4610 ASSERT_RACCT_ENABLED();
4612 sx_slock(&allproc_lock);
4613 sx_xlock(&allprison_lock);
4615 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4616 sx_xunlock(&allprison_lock);
4617 sx_sunlock(&allproc_lock);
4621 oldprr = pr->pr_prison_racct;
4622 pr->pr_prison_racct = NULL;
4624 prison_racct_attach(pr);
4627 * Move resource utilisation records.
4629 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4632 * Force rctl to reattach rules to processes.
4634 FOREACH_PROC_IN_SYSTEM(p) {
4636 cred = crhold(p->p_ucred);
4638 racct_proc_ucred_changed(p, cred, cred);
4642 sx_sunlock(&allproc_lock);
4643 prison_racct_free_locked(oldprr);
4644 sx_xunlock(&allprison_lock);
4648 prison_racct_detach(struct prison *pr)
4651 ASSERT_RACCT_ENABLED();
4652 sx_assert(&allprison_lock, SA_UNLOCKED);
4654 if (pr->pr_prison_racct == NULL)
4656 prison_racct_free(pr->pr_prison_racct);
4657 pr->pr_prison_racct = NULL;
4664 db_show_prison(struct prison *pr)
4667 #if defined(INET) || defined(INET6)
4672 char ip6buf[INET6_ADDRSTRLEN];
4675 db_printf("prison %p:\n", pr);
4676 db_printf(" jid = %d\n", pr->pr_id);
4677 db_printf(" name = %s\n", pr->pr_name);
4678 db_printf(" parent = %p\n", pr->pr_parent);
4679 db_printf(" ref = %d\n", pr->pr_ref);
4680 db_printf(" uref = %d\n", pr->pr_uref);
4681 db_printf(" path = %s\n", pr->pr_path);
4682 db_printf(" cpuset = %d\n", pr->pr_cpuset
4683 ? pr->pr_cpuset->cs_id : -1);
4685 db_printf(" vnet = %p\n", pr->pr_vnet);
4687 db_printf(" root = %p\n", pr->pr_root);
4688 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4689 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4690 db_printf(" children.max = %d\n", pr->pr_childmax);
4691 db_printf(" children.cur = %d\n", pr->pr_childcount);
4692 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4693 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4694 db_printf(" flags = 0x%x", pr->pr_flags);
4695 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4697 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4698 db_printf(" %s", pr_flag_names[fi]);
4699 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4701 jsf = pr->pr_flags &
4702 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4703 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4704 pr_flag_jailsys[fi].disable &&
4705 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4706 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4709 db_printf(" allow = 0x%x", pr->pr_allow);
4710 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4712 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4713 db_printf(" %s", pr_allow_names[fi]);
4715 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4716 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4717 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4718 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4719 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4721 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4722 for (ii = 0; ii < pr->pr_ip4s; ii++)
4723 db_printf(" %s %s\n",
4724 ii == 0 ? "ip4.addr =" : " ",
4725 inet_ntoa(pr->pr_ip4[ii]));
4728 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4729 for (ii = 0; ii < pr->pr_ip6s; ii++)
4730 db_printf(" %s %s\n",
4731 ii == 0 ? "ip6.addr =" : " ",
4732 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4736 DB_SHOW_COMMAND(prison, db_show_prison_command)
4742 * Show all prisons in the list, and prison0 which is not
4745 db_show_prison(&prison0);
4746 if (!db_pager_quit) {
4747 TAILQ_FOREACH(pr, &allprison, pr_list) {
4759 /* Look for a prison with the ID and with references. */
4760 TAILQ_FOREACH(pr, &allprison, pr_list)
4761 if (pr->pr_id == addr && pr->pr_ref > 0)
4764 /* Look again, without requiring a reference. */
4765 TAILQ_FOREACH(pr, &allprison, pr_list)
4766 if (pr->pr_id == addr)
4769 /* Assume address points to a valid prison. */
4770 pr = (struct prison *)addr;