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. Also allow the name to be
1589 * explicitly the jid - but not any other number, and only in
1590 * normal form (no leading zero/etc).
1592 if (name[0] == '\0')
1593 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1594 else if ((strtoul(namelc, &p, 10) != jid ||
1595 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1598 "name cannot be numeric (unless it is the jid)");
1599 goto done_deref_locked;
1602 * Make sure the name isn't too long for the prison or its
1605 onamelen = strlen(pr->pr_name);
1606 namelen = strlen(name);
1607 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1608 error = ENAMETOOLONG;
1609 goto done_deref_locked;
1611 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1612 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1613 sizeof(pr->pr_name)) {
1614 error = ENAMETOOLONG;
1615 goto done_deref_locked;
1619 if (pr_allow & ~ppr->pr_allow) {
1621 goto done_deref_locked;
1624 /* Set the parameters of the prison. */
1627 if (pr_flags & PR_IP4_USER) {
1628 pr->pr_flags |= PR_IP4;
1629 free(pr->pr_ip4, M_PRISON);
1633 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1635 if (tpr->pr_flags & PR_VNET) {
1640 if (prison_restrict_ip4(tpr, NULL)) {
1649 if (pr_flags & PR_IP6_USER) {
1650 pr->pr_flags |= PR_IP6;
1651 free(pr->pr_ip6, M_PRISON);
1655 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1657 if (tpr->pr_flags & PR_VNET) {
1662 if (prison_restrict_ip6(tpr, NULL)) {
1670 pr->pr_securelevel = slevel;
1671 /* Set all child jails to be at least this level. */
1672 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1673 if (tpr->pr_securelevel < slevel)
1674 tpr->pr_securelevel = slevel;
1677 pr->pr_childmax = childmax;
1678 /* Set all child jails to under this limit. */
1679 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1680 if (tpr->pr_childmax > childmax - level)
1681 tpr->pr_childmax = childmax > level
1682 ? childmax - level : 0;
1685 pr->pr_enforce_statfs = enforce;
1686 /* Pass this restriction on to the children. */
1687 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1688 if (tpr->pr_enforce_statfs < enforce)
1689 tpr->pr_enforce_statfs = enforce;
1692 pr->pr_devfs_rsnum = rsnum;
1693 /* Pass this restriction on to the children. */
1694 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1695 tpr->pr_devfs_rsnum = rsnum;
1698 if (ppr == &prison0)
1699 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1701 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1702 ppr->pr_name, name);
1703 /* Change this component of child names. */
1704 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1705 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1706 strlen(tpr->pr_name + onamelen) + 1);
1707 bcopy(pr->pr_name, tpr->pr_name, namelen);
1711 /* Try to keep a real-rooted full pathname. */
1712 if (fullpath_disabled && path[0] == '/' &&
1713 strcmp(mypr->pr_path, "/"))
1714 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1715 mypr->pr_path, path);
1717 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1720 if (PR_HOST & ch_flags & ~pr_flags) {
1721 if (pr->pr_flags & PR_HOST) {
1723 * Copy the parent's host info. As with pr_ip4 above,
1724 * the lack of a lock on the parent is not a problem;
1725 * it is always set with allprison_lock at least
1726 * shared, and is held exclusively here.
1728 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1729 sizeof(pr->pr_hostname));
1730 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1731 sizeof(pr->pr_domainname));
1732 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1733 sizeof(pr->pr_hostuuid));
1734 pr->pr_hostid = pr->pr_parent->pr_hostid;
1736 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1737 /* Set this prison, and any descendants without PR_HOST. */
1739 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1741 strlcpy(pr->pr_domainname, domain,
1742 sizeof(pr->pr_domainname));
1744 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1746 pr->pr_hostid = hid;
1747 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1748 if (tpr->pr_flags & PR_HOST)
1752 strlcpy(tpr->pr_hostname,
1754 sizeof(tpr->pr_hostname));
1756 strlcpy(tpr->pr_domainname,
1758 sizeof(tpr->pr_domainname));
1760 strlcpy(tpr->pr_hostuuid,
1762 sizeof(tpr->pr_hostuuid));
1764 tpr->pr_hostid = hid;
1768 if ((tallow = ch_allow & ~pr_allow)) {
1769 /* Clear allow bits in all children. */
1770 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1771 tpr->pr_allow &= ~tallow;
1773 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1775 * Persistent prisons get an extra reference, and prisons losing their
1776 * persist flag lose that reference. Only do this for existing prisons
1777 * for now, so new ones will remain unseen until after the module
1778 * handlers have completed.
1780 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1781 if (pr_flags & PR_PERSIST) {
1789 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1790 mtx_unlock(&pr->pr_mtx);
1793 if (racct_enable && created)
1794 prison_racct_attach(pr);
1797 /* Locks may have prevented a complete restriction of child IP
1798 * addresses. If so, allocate some more memory and try again.
1803 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1804 mtx_lock(&pr->pr_mtx);
1806 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1808 if (tpr->pr_flags & PR_VNET) {
1813 if (prison_restrict_ip4(tpr, ip4)) {
1820 mtx_unlock(&pr->pr_mtx);
1826 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1827 mtx_lock(&pr->pr_mtx);
1829 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1831 if (tpr->pr_flags & PR_VNET) {
1836 if (prison_restrict_ip6(tpr, ip6)) {
1843 mtx_unlock(&pr->pr_mtx);
1847 /* Let the modules do their work. */
1848 sx_downgrade(&allprison_lock);
1850 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1852 prison_deref(pr, PD_LIST_SLOCKED);
1856 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1858 prison_deref(pr, created
1860 : PD_DEREF | PD_LIST_SLOCKED);
1864 /* Attach this process to the prison if requested. */
1865 if (flags & JAIL_ATTACH) {
1866 mtx_lock(&pr->pr_mtx);
1867 error = do_jail_attach(td, pr);
1869 vfs_opterror(opts, "attach failed");
1871 prison_deref(pr, PD_DEREF);
1877 if (racct_enable && !created) {
1878 if (!(flags & JAIL_ATTACH))
1879 sx_sunlock(&allprison_lock);
1880 prison_racct_modify(pr);
1881 if (!(flags & JAIL_ATTACH))
1882 sx_slock(&allprison_lock);
1886 td->td_retval[0] = pr->pr_id;
1889 * Now that it is all there, drop the temporary reference from existing
1890 * prisons. Or add a reference to newly created persistent prisons
1891 * (which was not done earlier so that the prison would not be publicly
1895 prison_deref(pr, (flags & JAIL_ATTACH)
1897 : PD_DEREF | PD_LIST_SLOCKED);
1899 if (pr_flags & PR_PERSIST) {
1900 mtx_lock(&pr->pr_mtx);
1903 mtx_unlock(&pr->pr_mtx);
1905 if (!(flags & JAIL_ATTACH))
1906 sx_sunlock(&allprison_lock);
1912 prison_deref(pr, created
1913 ? PD_LOCKED | PD_LIST_XLOCKED
1914 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1917 sx_xunlock(&allprison_lock);
1923 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1924 if (errmsg_len > 0) {
1925 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1926 if (errmsg_pos > 0) {
1927 if (optuio->uio_segflg == UIO_SYSSPACE)
1929 optuio->uio_iov[errmsg_pos].iov_base,
1933 optuio->uio_iov[errmsg_pos].iov_base,
1940 free(ip4, M_PRISON);
1943 free(ip6, M_PRISON);
1946 free(g_path, M_TEMP);
1953 * struct jail_get_args {
1954 * struct iovec *iovp;
1955 * unsigned int iovcnt;
1960 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1965 /* Check that we have an even number of iovecs. */
1966 if (uap->iovcnt & 1)
1969 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1972 error = kern_jail_get(td, auio, uap->flags);
1974 error = copyout(auio->uio_iov, uap->iovp,
1975 uap->iovcnt * sizeof (struct iovec));
1981 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1983 struct prison *pr, *mypr;
1985 struct vfsoptlist *opts;
1986 char *errmsg, *name;
1987 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1989 if (flags & ~JAIL_GET_MASK)
1992 /* Get the parameter list. */
1993 error = vfs_buildopts(optuio, &opts);
1996 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1997 mypr = td->td_ucred->cr_prison;
2000 * Find the prison specified by one of: lastjid, jid, name.
2002 sx_slock(&allprison_lock);
2003 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2005 TAILQ_FOREACH(pr, &allprison, pr_list) {
2006 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2007 mtx_lock(&pr->pr_mtx);
2008 if (pr->pr_ref > 0 &&
2009 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2011 mtx_unlock(&pr->pr_mtx);
2017 vfs_opterror(opts, "no jail after %d", jid);
2018 goto done_unlock_list;
2019 } else if (error != ENOENT)
2020 goto done_unlock_list;
2022 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2025 pr = prison_find_child(mypr, jid);
2027 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2028 mtx_unlock(&pr->pr_mtx);
2030 vfs_opterror(opts, "jail %d is dying",
2032 goto done_unlock_list;
2037 vfs_opterror(opts, "jail %d not found", jid);
2038 goto done_unlock_list;
2040 } else if (error != ENOENT)
2041 goto done_unlock_list;
2043 error = vfs_getopt(opts, "name", (void **)&name, &len);
2045 if (len == 0 || name[len - 1] != '\0') {
2047 goto done_unlock_list;
2049 pr = prison_find_name(mypr, name);
2051 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2052 mtx_unlock(&pr->pr_mtx);
2054 vfs_opterror(opts, "jail \"%s\" is dying",
2056 goto done_unlock_list;
2061 vfs_opterror(opts, "jail \"%s\" not found", name);
2062 goto done_unlock_list;
2063 } else if (error != ENOENT)
2064 goto done_unlock_list;
2066 vfs_opterror(opts, "no jail specified");
2068 goto done_unlock_list;
2071 /* Get the parameters of the prison. */
2074 td->td_retval[0] = pr->pr_id;
2075 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2076 if (error != 0 && error != ENOENT)
2078 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2079 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2080 if (error != 0 && error != ENOENT)
2082 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2083 if (error != 0 && error != ENOENT)
2085 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2086 sizeof(pr->pr_cpuset->cs_id));
2087 if (error != 0 && error != ENOENT)
2089 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2090 if (error != 0 && error != ENOENT)
2093 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2094 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2095 if (error != 0 && error != ENOENT)
2099 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2100 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2101 if (error != 0 && error != ENOENT)
2104 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2105 sizeof(pr->pr_securelevel));
2106 if (error != 0 && error != ENOENT)
2108 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2109 sizeof(pr->pr_childcount));
2110 if (error != 0 && error != ENOENT)
2112 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2113 sizeof(pr->pr_childmax));
2114 if (error != 0 && error != ENOENT)
2116 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2117 if (error != 0 && error != ENOENT)
2119 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2120 if (error != 0 && error != ENOENT)
2122 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2123 if (error != 0 && error != ENOENT)
2125 #ifdef COMPAT_FREEBSD32
2126 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2127 uint32_t hid32 = pr->pr_hostid;
2129 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2132 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2133 sizeof(pr->pr_hostid));
2134 if (error != 0 && error != ENOENT)
2136 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2137 sizeof(pr->pr_enforce_statfs));
2138 if (error != 0 && error != ENOENT)
2140 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2141 sizeof(pr->pr_devfs_rsnum));
2142 if (error != 0 && error != ENOENT)
2144 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2146 if (pr_flag_names[fi] == NULL)
2148 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2149 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2150 if (error != 0 && error != ENOENT)
2153 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2154 if (error != 0 && error != ENOENT)
2157 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2160 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2161 i = pr_flag_jailsys[fi].disable &&
2162 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2163 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2166 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2167 if (error != 0 && error != ENOENT)
2170 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2172 if (pr_allow_names[fi] == NULL)
2174 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2175 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2176 if (error != 0 && error != ENOENT)
2179 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2180 if (error != 0 && error != ENOENT)
2183 i = (pr->pr_uref == 0);
2184 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2185 if (error != 0 && error != ENOENT)
2188 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2189 if (error != 0 && error != ENOENT)
2191 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2192 sizeof(pr->pr_osreldate));
2193 if (error != 0 && error != ENOENT)
2195 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2196 if (error != 0 && error != ENOENT)
2199 /* Get the module parameters. */
2200 mtx_unlock(&pr->pr_mtx);
2202 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2205 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2207 /* By now, all parameters should have been noted. */
2208 TAILQ_FOREACH(opt, opts, link) {
2209 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2211 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2216 /* Write the fetched parameters back to userspace. */
2218 TAILQ_FOREACH(opt, opts, link) {
2219 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2220 pos = 2 * opt->pos + 1;
2221 optuio->uio_iov[pos].iov_len = opt->len;
2222 if (opt->value != NULL) {
2223 if (optuio->uio_segflg == UIO_SYSSPACE) {
2225 optuio->uio_iov[pos].iov_base,
2228 error = copyout(opt->value,
2229 optuio->uio_iov[pos].iov_base,
2240 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2244 sx_sunlock(&allprison_lock);
2246 if (error && errmsg_pos >= 0) {
2247 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2248 errmsg_pos = 2 * errmsg_pos + 1;
2249 if (errmsg_len > 0) {
2250 if (optuio->uio_segflg == UIO_SYSSPACE)
2252 optuio->uio_iov[errmsg_pos].iov_base,
2256 optuio->uio_iov[errmsg_pos].iov_base,
2266 * struct jail_remove_args {
2271 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2273 struct prison *pr, *cpr, *lpr, *tpr;
2276 error = priv_check(td, PRIV_JAIL_REMOVE);
2280 sx_xlock(&allprison_lock);
2281 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2283 sx_xunlock(&allprison_lock);
2287 /* Remove all descendants of this prison, then remove this prison. */
2289 pr->pr_flags |= PR_REMOVE;
2290 if (!LIST_EMPTY(&pr->pr_children)) {
2291 mtx_unlock(&pr->pr_mtx);
2293 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2294 mtx_lock(&cpr->pr_mtx);
2295 if (cpr->pr_ref > 0) {
2298 cpr->pr_flags |= PR_REMOVE;
2300 /* Already removed - do not do it again. */
2303 mtx_unlock(&cpr->pr_mtx);
2305 mtx_lock(&lpr->pr_mtx);
2306 prison_remove_one(lpr);
2307 sx_xlock(&allprison_lock);
2312 mtx_lock(&lpr->pr_mtx);
2313 prison_remove_one(lpr);
2314 sx_xlock(&allprison_lock);
2316 mtx_lock(&pr->pr_mtx);
2318 prison_remove_one(pr);
2323 prison_remove_one(struct prison *pr)
2328 /* If the prison was persistent, it is not anymore. */
2330 if (pr->pr_flags & PR_PERSIST) {
2333 pr->pr_flags &= ~PR_PERSIST;
2337 * jail_remove added a reference. If that's the only one, remove
2340 KASSERT(pr->pr_ref > 0,
2341 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2342 if (pr->pr_ref == 1) {
2344 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2348 mtx_unlock(&pr->pr_mtx);
2349 sx_xunlock(&allprison_lock);
2351 * Kill all processes unfortunate enough to be attached to this prison.
2353 sx_slock(&allproc_lock);
2354 LIST_FOREACH(p, &allproc, p_list) {
2356 if (p->p_state != PRS_NEW && p->p_ucred &&
2357 p->p_ucred->cr_prison == pr)
2358 kern_psignal(p, SIGKILL);
2361 sx_sunlock(&allproc_lock);
2362 /* Remove the temporary reference added by jail_remove. */
2363 prison_deref(pr, deuref | PD_DEREF);
2368 * struct jail_attach_args {
2373 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2378 error = priv_check(td, PRIV_JAIL_ATTACH);
2382 sx_slock(&allprison_lock);
2383 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2385 sx_sunlock(&allprison_lock);
2390 * Do not allow a process to attach to a prison that is not
2391 * considered to be "alive".
2393 if (pr->pr_uref == 0) {
2394 mtx_unlock(&pr->pr_mtx);
2395 sx_sunlock(&allprison_lock);
2399 return (do_jail_attach(td, pr));
2403 do_jail_attach(struct thread *td, struct prison *pr)
2407 struct ucred *newcred, *oldcred;
2411 * XXX: Note that there is a slight race here if two threads
2412 * in the same privileged process attempt to attach to two
2413 * different jails at the same time. It is important for
2414 * user processes not to do this, or they might end up with
2415 * a process root from one prison, but attached to the jail
2420 mtx_unlock(&pr->pr_mtx);
2422 /* Let modules do whatever they need to prepare for attaching. */
2423 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2425 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2428 sx_sunlock(&allprison_lock);
2431 * Reparent the newly attached process to this jail.
2433 ppr = td->td_ucred->cr_prison;
2435 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2439 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2440 if ((error = change_dir(pr->pr_root, td)) != 0)
2443 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2446 VOP_UNLOCK(pr->pr_root, 0);
2447 if ((error = change_root(pr->pr_root, td)))
2452 oldcred = p->p_ucred;
2454 crcopy(newcred, oldcred);
2455 newcred->cr_prison = pr;
2456 p->p_ucred = newcred;
2459 racct_proc_ucred_changed(p, oldcred, newcred);
2462 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2465 VOP_UNLOCK(pr->pr_root, 0);
2467 /* Tell modules this thread is still in its old jail after all. */
2468 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2469 prison_deref(pr, PD_DEREF | PD_DEUREF);
2475 * Returns a locked prison instance, or NULL on failure.
2478 prison_find(int prid)
2482 sx_assert(&allprison_lock, SX_LOCKED);
2483 TAILQ_FOREACH(pr, &allprison, pr_list) {
2484 if (pr->pr_id == prid) {
2485 mtx_lock(&pr->pr_mtx);
2488 mtx_unlock(&pr->pr_mtx);
2495 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2498 prison_find_child(struct prison *mypr, int prid)
2503 sx_assert(&allprison_lock, SX_LOCKED);
2504 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2505 if (pr->pr_id == prid) {
2506 mtx_lock(&pr->pr_mtx);
2509 mtx_unlock(&pr->pr_mtx);
2516 * Look for the name relative to mypr. Returns a locked prison or NULL.
2519 prison_find_name(struct prison *mypr, const char *name)
2521 struct prison *pr, *deadpr;
2525 sx_assert(&allprison_lock, SX_LOCKED);
2526 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2529 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2530 if (!strcmp(pr->pr_name + mylen, name)) {
2531 mtx_lock(&pr->pr_mtx);
2532 if (pr->pr_ref > 0) {
2533 if (pr->pr_uref > 0)
2537 mtx_unlock(&pr->pr_mtx);
2540 /* There was no valid prison - perhaps there was a dying one. */
2541 if (deadpr != NULL) {
2542 mtx_lock(&deadpr->pr_mtx);
2543 if (deadpr->pr_ref == 0) {
2544 mtx_unlock(&deadpr->pr_mtx);
2552 * See if a prison has the specific flag set.
2555 prison_flag(struct ucred *cred, unsigned flag)
2558 /* This is an atomic read, so no locking is necessary. */
2559 return (cred->cr_prison->pr_flags & flag);
2563 prison_allow(struct ucred *cred, unsigned flag)
2566 /* This is an atomic read, so no locking is necessary. */
2567 return (cred->cr_prison->pr_allow & flag);
2571 * Remove a prison reference. If that was the last reference, remove the
2572 * prison itself - but not in this context in case there are locks held.
2575 prison_free_locked(struct prison *pr)
2578 mtx_assert(&pr->pr_mtx, MA_OWNED);
2580 if (pr->pr_ref == 0) {
2581 mtx_unlock(&pr->pr_mtx);
2582 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2583 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2586 mtx_unlock(&pr->pr_mtx);
2590 prison_free(struct prison *pr)
2593 mtx_lock(&pr->pr_mtx);
2594 prison_free_locked(pr);
2598 prison_complete(void *context, int pending)
2601 prison_deref((struct prison *)context, 0);
2605 * Remove a prison reference (usually). This internal version assumes no
2606 * mutexes are held, except perhaps the prison itself. If there are no more
2607 * references, release and delist the prison. On completion, the prison lock
2608 * and the allprison lock are both unlocked.
2611 prison_deref(struct prison *pr, int flags)
2613 struct prison *ppr, *tpr;
2615 if (!(flags & PD_LOCKED))
2616 mtx_lock(&pr->pr_mtx);
2618 if (flags & PD_DEUREF) {
2620 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2622 if (flags & PD_DEREF)
2624 /* If the prison still has references, nothing else to do. */
2625 if (pr->pr_ref > 0) {
2626 mtx_unlock(&pr->pr_mtx);
2627 if (flags & PD_LIST_SLOCKED)
2628 sx_sunlock(&allprison_lock);
2629 else if (flags & PD_LIST_XLOCKED)
2630 sx_xunlock(&allprison_lock);
2634 mtx_unlock(&pr->pr_mtx);
2635 if (flags & PD_LIST_SLOCKED) {
2636 if (!sx_try_upgrade(&allprison_lock)) {
2637 sx_sunlock(&allprison_lock);
2638 sx_xlock(&allprison_lock);
2640 } else if (!(flags & PD_LIST_XLOCKED))
2641 sx_xlock(&allprison_lock);
2643 TAILQ_REMOVE(&allprison, pr, pr_list);
2644 LIST_REMOVE(pr, pr_sibling);
2645 ppr = pr->pr_parent;
2646 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2647 tpr->pr_childcount--;
2648 sx_xunlock(&allprison_lock);
2651 if (pr->pr_vnet != ppr->pr_vnet)
2652 vnet_destroy(pr->pr_vnet);
2654 if (pr->pr_root != NULL)
2656 mtx_destroy(&pr->pr_mtx);
2658 free(pr->pr_ip4, M_PRISON);
2661 free(pr->pr_ip6, M_PRISON);
2663 if (pr->pr_cpuset != NULL)
2664 cpuset_rel(pr->pr_cpuset);
2668 prison_racct_detach(pr);
2672 /* Removing a prison frees a reference on its parent. */
2674 mtx_lock(&pr->pr_mtx);
2675 flags = PD_DEREF | PD_DEUREF;
2680 prison_hold_locked(struct prison *pr)
2683 mtx_assert(&pr->pr_mtx, MA_OWNED);
2684 KASSERT(pr->pr_ref > 0,
2685 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2690 prison_hold(struct prison *pr)
2693 mtx_lock(&pr->pr_mtx);
2694 prison_hold_locked(pr);
2695 mtx_unlock(&pr->pr_mtx);
2699 prison_proc_hold(struct prison *pr)
2702 mtx_lock(&pr->pr_mtx);
2703 KASSERT(pr->pr_uref > 0,
2704 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2706 mtx_unlock(&pr->pr_mtx);
2710 prison_proc_free(struct prison *pr)
2713 mtx_lock(&pr->pr_mtx);
2714 KASSERT(pr->pr_uref > 0,
2715 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2716 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2722 * Restrict a prison's IP address list with its parent's, possibly replacing
2723 * it. Return true if the replacement buffer was used (or would have been).
2726 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2731 ppr = pr->pr_parent;
2732 if (!(pr->pr_flags & PR_IP4_USER)) {
2733 /* This has no user settings, so just copy the parent's list. */
2734 if (pr->pr_ip4s < ppr->pr_ip4s) {
2736 * There's no room for the parent's list. Use the
2737 * new list buffer, which is assumed to be big enough
2738 * (if it was passed). If there's no buffer, try to
2742 if (newip4 == NULL) {
2743 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2744 M_PRISON, M_NOWAIT);
2748 if (newip4 != NULL) {
2749 bcopy(ppr->pr_ip4, newip4,
2750 ppr->pr_ip4s * sizeof(*newip4));
2751 free(pr->pr_ip4, M_PRISON);
2752 pr->pr_ip4 = newip4;
2753 pr->pr_ip4s = ppr->pr_ip4s;
2757 pr->pr_ip4s = ppr->pr_ip4s;
2758 if (pr->pr_ip4s > 0)
2759 bcopy(ppr->pr_ip4, pr->pr_ip4,
2760 pr->pr_ip4s * sizeof(*newip4));
2761 else if (pr->pr_ip4 != NULL) {
2762 free(pr->pr_ip4, M_PRISON);
2765 } else if (pr->pr_ip4s > 0) {
2766 /* Remove addresses that aren't in the parent. */
2767 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2768 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2770 if (ij < ppr->pr_ip4s)
2773 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2774 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2777 for (ij = 1; ii < pr->pr_ip4s; ) {
2778 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2782 switch (ij >= ppr->pr_ip4s ? -1 :
2783 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2785 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2786 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2797 if (pr->pr_ip4s == 0) {
2798 pr->pr_flags |= PR_IP4_DISABLE;
2799 free(pr->pr_ip4, M_PRISON);
2807 * Pass back primary IPv4 address of this jail.
2809 * If not restricted return success but do not alter the address. Caller has
2810 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2812 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2813 * Address returned in NBO.
2816 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2820 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2821 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2823 pr = cred->cr_prison;
2824 if (!(pr->pr_flags & PR_IP4))
2826 mtx_lock(&pr->pr_mtx);
2827 if (!(pr->pr_flags & PR_IP4)) {
2828 mtx_unlock(&pr->pr_mtx);
2831 if (pr->pr_ip4 == NULL) {
2832 mtx_unlock(&pr->pr_mtx);
2833 return (EAFNOSUPPORT);
2836 ia->s_addr = pr->pr_ip4[0].s_addr;
2837 mtx_unlock(&pr->pr_mtx);
2842 * Return 1 if we should do proper source address selection or are not jailed.
2843 * We will return 0 if we should bypass source address selection in favour
2844 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2846 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2849 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2855 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2856 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2861 pr = cred->cr_prison;
2862 if (pr->pr_flags & PR_IP4_SADDRSEL)
2865 lia.s_addr = INADDR_ANY;
2866 error = prison_get_ip4(cred, &lia);
2869 if (lia.s_addr == INADDR_ANY)
2872 ia->s_addr = lia.s_addr;
2877 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2880 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2887 * No need to lock since the PR_IP4_USER flag can't be altered for
2890 while (pr1 != &prison0 &&
2892 !(pr1->pr_flags & PR_VNET) &&
2894 !(pr1->pr_flags & PR_IP4_USER))
2895 pr1 = pr1->pr_parent;
2896 while (pr2 != &prison0 &&
2898 !(pr2->pr_flags & PR_VNET) &&
2900 !(pr2->pr_flags & PR_IP4_USER))
2901 pr2 = pr2->pr_parent;
2902 return (pr1 == pr2);
2906 * Make sure our (source) address is set to something meaningful to this
2909 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2910 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2911 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2914 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2920 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2921 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2923 pr = cred->cr_prison;
2924 if (!(pr->pr_flags & PR_IP4))
2926 mtx_lock(&pr->pr_mtx);
2927 if (!(pr->pr_flags & PR_IP4)) {
2928 mtx_unlock(&pr->pr_mtx);
2931 if (pr->pr_ip4 == NULL) {
2932 mtx_unlock(&pr->pr_mtx);
2933 return (EAFNOSUPPORT);
2936 ia0.s_addr = ntohl(ia->s_addr);
2937 if (ia0.s_addr == INADDR_LOOPBACK) {
2938 ia->s_addr = pr->pr_ip4[0].s_addr;
2939 mtx_unlock(&pr->pr_mtx);
2943 if (ia0.s_addr == INADDR_ANY) {
2945 * In case there is only 1 IPv4 address, bind directly.
2947 if (pr->pr_ip4s == 1)
2948 ia->s_addr = pr->pr_ip4[0].s_addr;
2949 mtx_unlock(&pr->pr_mtx);
2953 error = _prison_check_ip4(pr, ia);
2954 mtx_unlock(&pr->pr_mtx);
2959 * Rewrite destination address in case we will connect to loopback address.
2961 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2962 * Address passed in in NBO and returned in NBO.
2965 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2969 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2970 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2972 pr = cred->cr_prison;
2973 if (!(pr->pr_flags & PR_IP4))
2975 mtx_lock(&pr->pr_mtx);
2976 if (!(pr->pr_flags & PR_IP4)) {
2977 mtx_unlock(&pr->pr_mtx);
2980 if (pr->pr_ip4 == NULL) {
2981 mtx_unlock(&pr->pr_mtx);
2982 return (EAFNOSUPPORT);
2985 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2986 ia->s_addr = pr->pr_ip4[0].s_addr;
2987 mtx_unlock(&pr->pr_mtx);
2992 * Return success because nothing had to be changed.
2994 mtx_unlock(&pr->pr_mtx);
2999 * Check if given address belongs to the jail referenced by cred/prison.
3001 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
3002 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3003 * doesn't allow IPv4. Address passed in in NBO.
3006 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
3011 * Check the primary IP.
3013 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3017 * All the other IPs are sorted so we can do a binary search.
3020 z = pr->pr_ip4s - 2;
3023 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3032 return (EADDRNOTAVAIL);
3036 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
3041 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3042 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3044 pr = cred->cr_prison;
3045 if (!(pr->pr_flags & PR_IP4))
3047 mtx_lock(&pr->pr_mtx);
3048 if (!(pr->pr_flags & PR_IP4)) {
3049 mtx_unlock(&pr->pr_mtx);
3052 if (pr->pr_ip4 == NULL) {
3053 mtx_unlock(&pr->pr_mtx);
3054 return (EAFNOSUPPORT);
3057 error = _prison_check_ip4(pr, ia);
3058 mtx_unlock(&pr->pr_mtx);
3065 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3070 ppr = pr->pr_parent;
3071 if (!(pr->pr_flags & PR_IP6_USER)) {
3072 /* This has no user settings, so just copy the parent's list. */
3073 if (pr->pr_ip6s < ppr->pr_ip6s) {
3075 * There's no room for the parent's list. Use the
3076 * new list buffer, which is assumed to be big enough
3077 * (if it was passed). If there's no buffer, try to
3081 if (newip6 == NULL) {
3082 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3083 M_PRISON, M_NOWAIT);
3087 if (newip6 != NULL) {
3088 bcopy(ppr->pr_ip6, newip6,
3089 ppr->pr_ip6s * sizeof(*newip6));
3090 free(pr->pr_ip6, M_PRISON);
3091 pr->pr_ip6 = newip6;
3092 pr->pr_ip6s = ppr->pr_ip6s;
3096 pr->pr_ip6s = ppr->pr_ip6s;
3097 if (pr->pr_ip6s > 0)
3098 bcopy(ppr->pr_ip6, pr->pr_ip6,
3099 pr->pr_ip6s * sizeof(*newip6));
3100 else if (pr->pr_ip6 != NULL) {
3101 free(pr->pr_ip6, M_PRISON);
3104 } else if (pr->pr_ip6s > 0) {
3105 /* Remove addresses that aren't in the parent. */
3106 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3107 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3110 if (ij < ppr->pr_ip6s)
3113 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3114 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3117 for (ij = 1; ii < pr->pr_ip6s; ) {
3118 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3123 switch (ij >= ppr->pr_ip6s ? -1 :
3124 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3126 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3127 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3138 if (pr->pr_ip6s == 0) {
3139 pr->pr_flags |= PR_IP6_DISABLE;
3140 free(pr->pr_ip6, M_PRISON);
3148 * Pass back primary IPv6 address for this jail.
3150 * If not restricted return success but do not alter the address. Caller has
3151 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3153 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3156 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3160 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3161 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3163 pr = cred->cr_prison;
3164 if (!(pr->pr_flags & PR_IP6))
3166 mtx_lock(&pr->pr_mtx);
3167 if (!(pr->pr_flags & PR_IP6)) {
3168 mtx_unlock(&pr->pr_mtx);
3171 if (pr->pr_ip6 == NULL) {
3172 mtx_unlock(&pr->pr_mtx);
3173 return (EAFNOSUPPORT);
3176 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3177 mtx_unlock(&pr->pr_mtx);
3182 * Return 1 if we should do proper source address selection or are not jailed.
3183 * We will return 0 if we should bypass source address selection in favour
3184 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3186 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3189 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3192 struct in6_addr lia6;
3195 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3196 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3201 pr = cred->cr_prison;
3202 if (pr->pr_flags & PR_IP6_SADDRSEL)
3206 error = prison_get_ip6(cred, &lia6);
3209 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3212 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3217 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3220 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3226 while (pr1 != &prison0 &&
3228 !(pr1->pr_flags & PR_VNET) &&
3230 !(pr1->pr_flags & PR_IP6_USER))
3231 pr1 = pr1->pr_parent;
3232 while (pr2 != &prison0 &&
3234 !(pr2->pr_flags & PR_VNET) &&
3236 !(pr2->pr_flags & PR_IP6_USER))
3237 pr2 = pr2->pr_parent;
3238 return (pr1 == pr2);
3242 * Make sure our (source) address is set to something meaningful to this jail.
3244 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3245 * when needed while binding.
3247 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3248 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3249 * doesn't allow IPv6.
3252 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3257 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3258 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3260 pr = cred->cr_prison;
3261 if (!(pr->pr_flags & PR_IP6))
3263 mtx_lock(&pr->pr_mtx);
3264 if (!(pr->pr_flags & PR_IP6)) {
3265 mtx_unlock(&pr->pr_mtx);
3268 if (pr->pr_ip6 == NULL) {
3269 mtx_unlock(&pr->pr_mtx);
3270 return (EAFNOSUPPORT);
3273 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3274 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3275 mtx_unlock(&pr->pr_mtx);
3279 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3281 * In case there is only 1 IPv6 address, and v6only is true,
3282 * then bind directly.
3284 if (v6only != 0 && pr->pr_ip6s == 1)
3285 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3286 mtx_unlock(&pr->pr_mtx);
3290 error = _prison_check_ip6(pr, ia6);
3291 mtx_unlock(&pr->pr_mtx);
3296 * Rewrite destination address in case we will connect to loopback address.
3298 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3301 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3305 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3306 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3308 pr = cred->cr_prison;
3309 if (!(pr->pr_flags & PR_IP6))
3311 mtx_lock(&pr->pr_mtx);
3312 if (!(pr->pr_flags & PR_IP6)) {
3313 mtx_unlock(&pr->pr_mtx);
3316 if (pr->pr_ip6 == NULL) {
3317 mtx_unlock(&pr->pr_mtx);
3318 return (EAFNOSUPPORT);
3321 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3322 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3323 mtx_unlock(&pr->pr_mtx);
3328 * Return success because nothing had to be changed.
3330 mtx_unlock(&pr->pr_mtx);
3335 * Check if given address belongs to the jail referenced by cred/prison.
3337 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3338 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3339 * doesn't allow IPv6.
3342 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3347 * Check the primary IP.
3349 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3353 * All the other IPs are sorted so we can do a binary search.
3356 z = pr->pr_ip6s - 2;
3359 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3368 return (EADDRNOTAVAIL);
3372 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3377 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3378 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3380 pr = cred->cr_prison;
3381 if (!(pr->pr_flags & PR_IP6))
3383 mtx_lock(&pr->pr_mtx);
3384 if (!(pr->pr_flags & PR_IP6)) {
3385 mtx_unlock(&pr->pr_mtx);
3388 if (pr->pr_ip6 == NULL) {
3389 mtx_unlock(&pr->pr_mtx);
3390 return (EAFNOSUPPORT);
3393 error = _prison_check_ip6(pr, ia6);
3394 mtx_unlock(&pr->pr_mtx);
3400 * Check if a jail supports the given address family.
3402 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3406 prison_check_af(struct ucred *cred, int af)
3411 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3413 pr = cred->cr_prison;
3415 /* Prisons with their own network stack are not limited. */
3416 if (prison_owns_vnet(cred))
3425 if (pr->pr_flags & PR_IP4)
3427 mtx_lock(&pr->pr_mtx);
3428 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3429 error = EAFNOSUPPORT;
3430 mtx_unlock(&pr->pr_mtx);
3436 if (pr->pr_flags & PR_IP6)
3438 mtx_lock(&pr->pr_mtx);
3439 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3440 error = EAFNOSUPPORT;
3441 mtx_unlock(&pr->pr_mtx);
3449 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3450 error = EAFNOSUPPORT;
3456 * Check if given address belongs to the jail referenced by cred (wrapper to
3457 * prison_check_ip[46]).
3459 * Returns 0 if jail doesn't restrict the address family or if address belongs
3460 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3461 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3464 prison_if(struct ucred *cred, struct sockaddr *sa)
3467 struct sockaddr_in *sai;
3470 struct sockaddr_in6 *sai6;
3474 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3475 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3478 if (prison_owns_vnet(cred))
3483 switch (sa->sa_family)
3487 sai = (struct sockaddr_in *)sa;
3488 error = prison_check_ip4(cred, &sai->sin_addr);
3493 sai6 = (struct sockaddr_in6 *)sa;
3494 error = prison_check_ip6(cred, &sai6->sin6_addr);
3498 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3499 error = EAFNOSUPPORT;
3505 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3508 prison_check(struct ucred *cred1, struct ucred *cred2)
3511 return ((cred1->cr_prison == cred2->cr_prison ||
3512 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3516 * Return 1 if p2 is a child of p1, otherwise 0.
3519 prison_ischild(struct prison *pr1, struct prison *pr2)
3522 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3529 * Return 1 if the passed credential is in a jail, otherwise 0.
3532 jailed(struct ucred *cred)
3535 return (cred->cr_prison != &prison0);
3539 * Return 1 if the passed credential is in a jail and that jail does not
3540 * have its own virtual network stack, otherwise 0.
3543 jailed_without_vnet(struct ucred *cred)
3549 if (prison_owns_vnet(cred))
3557 * Return the correct hostname (domainname, et al) for the passed credential.
3560 getcredhostname(struct ucred *cred, char *buf, size_t size)
3565 * A NULL credential can be used to shortcut to the physical
3566 * system's hostname.
3568 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3569 mtx_lock(&pr->pr_mtx);
3570 strlcpy(buf, pr->pr_hostname, size);
3571 mtx_unlock(&pr->pr_mtx);
3575 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3578 mtx_lock(&cred->cr_prison->pr_mtx);
3579 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3580 mtx_unlock(&cred->cr_prison->pr_mtx);
3584 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3587 mtx_lock(&cred->cr_prison->pr_mtx);
3588 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3589 mtx_unlock(&cred->cr_prison->pr_mtx);
3593 getcredhostid(struct ucred *cred, unsigned long *hostid)
3596 mtx_lock(&cred->cr_prison->pr_mtx);
3597 *hostid = cred->cr_prison->pr_hostid;
3598 mtx_unlock(&cred->cr_prison->pr_mtx);
3603 * Determine whether the prison represented by cred owns
3604 * its vnet rather than having it inherited.
3606 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3609 prison_owns_vnet(struct ucred *cred)
3613 * vnets cannot be added/removed after jail creation,
3614 * so no need to lock here.
3616 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3621 * Determine whether the subject represented by cred can "see"
3622 * status of a mount point.
3623 * Returns: 0 for permitted, ENOENT otherwise.
3624 * XXX: This function should be called cr_canseemount() and should be
3625 * placed in kern_prot.c.
3628 prison_canseemount(struct ucred *cred, struct mount *mp)
3634 pr = cred->cr_prison;
3635 if (pr->pr_enforce_statfs == 0)
3637 if (pr->pr_root->v_mount == mp)
3639 if (pr->pr_enforce_statfs == 2)
3642 * If jail's chroot directory is set to "/" we should be able to see
3643 * all mount-points from inside a jail.
3644 * This is ugly check, but this is the only situation when jail's
3645 * directory ends with '/'.
3647 if (strcmp(pr->pr_path, "/") == 0)
3649 len = strlen(pr->pr_path);
3651 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3654 * Be sure that we don't have situation where jail's root directory
3655 * is "/some/path" and mount point is "/some/pathpath".
3657 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3663 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3665 char jpath[MAXPATHLEN];
3669 pr = cred->cr_prison;
3670 if (pr->pr_enforce_statfs == 0)
3672 if (prison_canseemount(cred, mp) != 0) {
3673 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3674 strlcpy(sp->f_mntonname, "[restricted]",
3675 sizeof(sp->f_mntonname));
3678 if (pr->pr_root->v_mount == mp) {
3680 * Clear current buffer data, so we are sure nothing from
3681 * the valid path left there.
3683 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3684 *sp->f_mntonname = '/';
3688 * If jail's chroot directory is set to "/" we should be able to see
3689 * all mount-points from inside a jail.
3691 if (strcmp(pr->pr_path, "/") == 0)
3693 len = strlen(pr->pr_path);
3694 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3696 * Clear current buffer data, so we are sure nothing from
3697 * the valid path left there.
3699 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3700 if (*jpath == '\0') {
3701 /* Should never happen. */
3702 *sp->f_mntonname = '/';
3704 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3709 * Check with permission for a specific privilege is granted within jail. We
3710 * have a specific list of accepted privileges; the rest are denied.
3713 prison_priv_check(struct ucred *cred, int priv)
3721 * Privileges specific to prisons with a virtual network stack.
3722 * There might be a duplicate entry here in case the privilege
3723 * is only granted conditionally in the legacy jail case.
3728 * NFS-specific privileges.
3730 case PRIV_NFS_DAEMON:
3731 case PRIV_NFS_LOCKD:
3734 * Network stack privileges.
3736 case PRIV_NET_BRIDGE:
3739 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3740 case PRIV_NET_ROUTE:
3742 case PRIV_NET_SETIFMTU:
3743 case PRIV_NET_SETIFFLAGS:
3744 case PRIV_NET_SETIFCAP:
3745 case PRIV_NET_SETIFDESCR:
3746 case PRIV_NET_SETIFNAME :
3747 case PRIV_NET_SETIFMETRIC:
3748 case PRIV_NET_SETIFPHYS:
3749 case PRIV_NET_SETIFMAC:
3750 case PRIV_NET_ADDMULTI:
3751 case PRIV_NET_DELMULTI:
3752 case PRIV_NET_HWIOCTL:
3753 case PRIV_NET_SETLLADDR:
3754 case PRIV_NET_ADDIFGROUP:
3755 case PRIV_NET_DELIFGROUP:
3756 case PRIV_NET_IFCREATE:
3757 case PRIV_NET_IFDESTROY:
3758 case PRIV_NET_ADDIFADDR:
3759 case PRIV_NET_DELIFADDR:
3762 case PRIV_NET_SETIFVNET:
3763 case PRIV_NET_SETIFFIB:
3766 * 802.11-related privileges.
3768 case PRIV_NET80211_GETKEY:
3770 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3775 * AppleTalk privileges.
3777 case PRIV_NETATALK_RESERVEDPORT:
3782 case PRIV_NETATM_CFG:
3783 case PRIV_NETATM_ADD:
3784 case PRIV_NETATM_DEL:
3785 case PRIV_NETATM_SET:
3788 * Bluetooth privileges.
3790 case PRIV_NETBLUETOOTH_RAW:
3794 * Netgraph and netgraph module privileges.
3796 case PRIV_NETGRAPH_CONTROL:
3798 case PRIV_NETGRAPH_TTY:
3802 * IPv4 and IPv6 privileges.
3804 case PRIV_NETINET_IPFW:
3805 case PRIV_NETINET_DIVERT:
3806 case PRIV_NETINET_PF:
3807 case PRIV_NETINET_DUMMYNET:
3808 case PRIV_NETINET_CARP:
3809 case PRIV_NETINET_MROUTE:
3810 case PRIV_NETINET_RAW:
3811 case PRIV_NETINET_ADDRCTRL6:
3812 case PRIV_NETINET_ND6:
3813 case PRIV_NETINET_SCOPE6:
3814 case PRIV_NETINET_ALIFETIME6:
3815 case PRIV_NETINET_IPSEC:
3816 case PRIV_NETINET_BINDANY:
3820 * IPX/SPX privileges.
3822 case PRIV_NETIPX_RESERVEDPORT:
3823 case PRIV_NETIPX_RAW:
3837 * No default: or deny here.
3838 * In case of no permit fall through to next switch().
3840 if (cred->cr_prison->pr_flags & PR_VNET)
3848 * Allow ktrace privileges for root in jail.
3854 * Allow jailed processes to configure audit identity and
3855 * submit audit records (login, etc). In the future we may
3856 * want to further refine the relationship between audit and
3859 case PRIV_AUDIT_GETAUDIT:
3860 case PRIV_AUDIT_SETAUDIT:
3861 case PRIV_AUDIT_SUBMIT:
3865 * Allow jailed processes to manipulate process UNIX
3866 * credentials in any way they see fit.
3868 case PRIV_CRED_SETUID:
3869 case PRIV_CRED_SETEUID:
3870 case PRIV_CRED_SETGID:
3871 case PRIV_CRED_SETEGID:
3872 case PRIV_CRED_SETGROUPS:
3873 case PRIV_CRED_SETREUID:
3874 case PRIV_CRED_SETREGID:
3875 case PRIV_CRED_SETRESUID:
3876 case PRIV_CRED_SETRESGID:
3879 * Jail implements visibility constraints already, so allow
3880 * jailed root to override uid/gid-based constraints.
3882 case PRIV_SEEOTHERGIDS:
3883 case PRIV_SEEOTHERUIDS:
3886 * Jail implements inter-process debugging limits already, so
3887 * allow jailed root various debugging privileges.
3889 case PRIV_DEBUG_DIFFCRED:
3890 case PRIV_DEBUG_SUGID:
3891 case PRIV_DEBUG_UNPRIV:
3894 * Allow jail to set various resource limits and login
3895 * properties, and for now, exceed process resource limits.
3897 case PRIV_PROC_LIMIT:
3898 case PRIV_PROC_SETLOGIN:
3899 case PRIV_PROC_SETRLIMIT:
3902 * System V and POSIX IPC privileges are granted in jail.
3905 case PRIV_IPC_WRITE:
3906 case PRIV_IPC_ADMIN:
3907 case PRIV_IPC_MSGSIZE:
3911 * Jail operations within a jail work on child jails.
3913 case PRIV_JAIL_ATTACH:
3915 case PRIV_JAIL_REMOVE:
3918 * Jail implements its own inter-process limits, so allow
3919 * root processes in jail to change scheduling on other
3920 * processes in the same jail. Likewise for signalling.
3922 case PRIV_SCHED_DIFFCRED:
3923 case PRIV_SCHED_CPUSET:
3924 case PRIV_SIGNAL_DIFFCRED:
3925 case PRIV_SIGNAL_SUGID:
3928 * Allow jailed processes to write to sysctls marked as jail
3931 case PRIV_SYSCTL_WRITEJAIL:
3934 * Allow root in jail to manage a variety of quota
3935 * properties. These should likely be conditional on a
3936 * configuration option.
3938 case PRIV_VFS_GETQUOTA:
3939 case PRIV_VFS_SETQUOTA:
3942 * Since Jail relies on chroot() to implement file system
3943 * protections, grant many VFS privileges to root in jail.
3944 * Be careful to exclude mount-related and NFS-related
3948 case PRIV_VFS_WRITE:
3949 case PRIV_VFS_ADMIN:
3951 case PRIV_VFS_LOOKUP:
3952 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3953 case PRIV_VFS_CHFLAGS_DEV:
3954 case PRIV_VFS_CHOWN:
3955 case PRIV_VFS_CHROOT:
3956 case PRIV_VFS_RETAINSUGID:
3957 case PRIV_VFS_FCHROOT:
3959 case PRIV_VFS_SETGID:
3961 case PRIV_VFS_STICKYFILE:
3964 * As in the non-jail case, non-root users are expected to be
3965 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3966 * exists in the jail and they have permission to access it).
3968 case PRIV_KMEM_READ:
3972 * Depending on the global setting, allow privilege of
3973 * setting system flags.
3975 case PRIV_VFS_SYSFLAGS:
3976 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3982 * Depending on the global setting, allow privilege of
3983 * mounting/unmounting file systems.
3985 case PRIV_VFS_MOUNT:
3986 case PRIV_VFS_UNMOUNT:
3987 case PRIV_VFS_MOUNT_NONUSER:
3988 case PRIV_VFS_MOUNT_OWNER:
3989 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3990 cred->cr_prison->pr_enforce_statfs < 2)
3996 * Allow jailed root to bind reserved ports and reuse in-use
3999 case PRIV_NETINET_RESERVEDPORT:
4000 case PRIV_NETINET_REUSEPORT:
4004 * Allow jailed root to set certian IPv4/6 (option) headers.
4006 case PRIV_NETINET_SETHDROPTS:
4010 * Conditionally allow creating raw sockets in jail.
4012 case PRIV_NETINET_RAW:
4013 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4019 * Since jail implements its own visibility limits on netstat
4020 * sysctls, allow getcred. This allows identd to work in
4023 case PRIV_NETINET_GETCRED:
4027 * Allow jailed root to set loginclass.
4029 case PRIV_PROC_SETLOGINCLASS:
4034 * In all remaining cases, deny the privilege request. This
4035 * includes almost all network privileges, many system
4036 * configuration privileges.
4043 * Return the part of pr2's name that is relative to pr1, or the whole name
4044 * if it does not directly follow.
4048 prison_name(struct prison *pr1, struct prison *pr2)
4052 /* Jails see themselves as "0" (if they see themselves at all). */
4055 name = pr2->pr_name;
4056 if (prison_ischild(pr1, pr2)) {
4058 * pr1 isn't locked (and allprison_lock may not be either)
4059 * so its length can't be counted on. But the number of dots
4060 * can be counted on - and counted.
4062 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4063 name = strchr(name, '.') + 1;
4069 * Return the part of pr2's path that is relative to pr1, or the whole path
4070 * if it does not directly follow.
4073 prison_path(struct prison *pr1, struct prison *pr2)
4075 char *path1, *path2;
4078 path1 = pr1->pr_path;
4079 path2 = pr2->pr_path;
4080 if (!strcmp(path1, "/"))
4082 len1 = strlen(path1);
4083 if (strncmp(path1, path2, len1))
4085 if (path2[len1] == '\0')
4087 if (path2[len1] == '/')
4088 return (path2 + len1);
4094 * Jail-related sysctls.
4096 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4100 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4103 struct prison *pr, *cpr;
4105 struct in_addr *ip4 = NULL;
4109 struct in6_addr *ip6 = NULL;
4114 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4115 pr = req->td->td_ucred->cr_prison;
4117 sx_slock(&allprison_lock);
4118 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4119 #if defined(INET) || defined(INET6)
4122 mtx_lock(&cpr->pr_mtx);
4124 if (cpr->pr_ip4s > 0) {
4125 if (ip4s < cpr->pr_ip4s) {
4126 ip4s = cpr->pr_ip4s;
4127 mtx_unlock(&cpr->pr_mtx);
4128 ip4 = realloc(ip4, ip4s *
4129 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4132 bcopy(cpr->pr_ip4, ip4,
4133 cpr->pr_ip4s * sizeof(struct in_addr));
4137 if (cpr->pr_ip6s > 0) {
4138 if (ip6s < cpr->pr_ip6s) {
4139 ip6s = cpr->pr_ip6s;
4140 mtx_unlock(&cpr->pr_mtx);
4141 ip6 = realloc(ip6, ip6s *
4142 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4145 bcopy(cpr->pr_ip6, ip6,
4146 cpr->pr_ip6s * sizeof(struct in6_addr));
4149 if (cpr->pr_ref == 0) {
4150 mtx_unlock(&cpr->pr_mtx);
4153 bzero(xp, sizeof(*xp));
4154 xp->pr_version = XPRISON_VERSION;
4155 xp->pr_id = cpr->pr_id;
4156 xp->pr_state = cpr->pr_uref > 0
4157 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4158 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4159 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4160 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4162 xp->pr_ip4s = cpr->pr_ip4s;
4165 xp->pr_ip6s = cpr->pr_ip6s;
4167 mtx_unlock(&cpr->pr_mtx);
4168 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4172 if (xp->pr_ip4s > 0) {
4173 error = SYSCTL_OUT(req, ip4,
4174 xp->pr_ip4s * sizeof(struct in_addr));
4180 if (xp->pr_ip6s > 0) {
4181 error = SYSCTL_OUT(req, ip6,
4182 xp->pr_ip6s * sizeof(struct in6_addr));
4188 sx_sunlock(&allprison_lock);
4199 SYSCTL_OID(_security_jail, OID_AUTO, list,
4200 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4201 sysctl_jail_list, "S", "List of active jails");
4204 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4208 injail = jailed(req->td->td_ucred);
4209 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4214 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4215 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4216 sysctl_jail_jailed, "I", "Process in jail?");
4219 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4221 int error, havevnet;
4223 struct ucred *cred = req->td->td_ucred;
4225 havevnet = jailed(cred) && prison_owns_vnet(cred);
4229 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4234 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4235 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4236 sysctl_jail_vnet, "I", "Jail owns VNET?");
4238 #if defined(INET) || defined(INET6)
4239 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4240 &jail_max_af_ips, 0,
4241 "Number of IP addresses a jail may have at most per address family");
4245 * Default parameters for jail(2) compatability. For historical reasons,
4246 * the sysctl names have varying similarity to the parameter names. Prisons
4247 * just see their own parameters, and can't change them.
4250 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4253 int allow, error, i;
4255 pr = req->td->td_ucred->cr_prison;
4256 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4258 /* Get the current flag value, and convert it to a boolean. */
4259 i = (allow & arg2) ? 1 : 0;
4262 error = sysctl_handle_int(oidp, &i, 0, req);
4263 if (error || !req->newptr)
4269 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4272 mtx_lock(&prison0.pr_mtx);
4273 jail_default_allow = (jail_default_allow & ~arg2) | i;
4274 mtx_unlock(&prison0.pr_mtx);
4278 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4279 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4280 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4281 "Processes in jail can set their hostnames");
4282 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4283 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4284 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4285 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4286 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4287 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4288 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4289 "Processes in jail can use System V IPC primitives");
4290 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4291 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4292 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4293 "Prison root can create raw sockets");
4294 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4295 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4296 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4297 "Processes in jail can alter system file flags");
4298 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4299 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4300 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4301 "Processes in jail can mount/unmount jail-friendly file systems");
4302 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4303 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4304 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4305 "Processes in jail can mount the devfs file system");
4306 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4307 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4308 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4309 "Processes in jail can mount the fdescfs file system");
4310 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4311 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4312 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4313 "Processes in jail can mount the nullfs file system");
4314 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4315 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4316 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4317 "Processes in jail can mount the procfs file system");
4318 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4319 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4320 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4321 "Processes in jail can mount the tmpfs file system");
4322 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4323 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4324 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4325 "Processes in jail can mount the zfs file system");
4328 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4333 pr = req->td->td_ucred->cr_prison;
4334 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4335 error = sysctl_handle_int(oidp, &level, 0, req);
4336 if (error || !req->newptr)
4338 *(int *)arg1 = level;
4342 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4343 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4344 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4345 sysctl_jail_default_level, "I",
4346 "Processes in jail cannot see all mounted file systems");
4348 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4349 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4350 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4351 sysctl_jail_default_level, "I",
4352 "Ruleset for the devfs filesystem in jail");
4355 * Nodes to describe jail parameters. Maximum length of string parameters
4356 * is returned in the string itself, and the other parameters exist merely
4357 * to make themselves and their types known.
4359 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4363 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4370 switch (oidp->oid_kind & CTLTYPE)
4376 if (!(req->flags & SCTL_MASK32))
4378 return (SYSCTL_OUT(req, &l, sizeof(l)));
4382 return (SYSCTL_OUT(req, &i, sizeof(i)));
4383 case CTLTYPE_STRING:
4384 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4386 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4387 case CTLTYPE_STRUCT:
4389 return (SYSCTL_OUT(req, &s, sizeof(s)));
4395 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4396 * jail creation time but cannot be changed in an existing jail.
4398 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4399 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4400 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4401 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4402 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4403 "I", "Jail secure level");
4404 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4405 "Jail value for kern.osreldate and uname -K");
4406 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4407 "Jail value for kern.osrelease and uname -r");
4408 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4409 "I", "Jail cannot see all mounted file systems");
4410 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4411 "I", "Ruleset for in-jail devfs mounts");
4412 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4413 "B", "Jail persistence");
4415 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4416 "E,jailsys", "Virtual network stack");
4418 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4419 "B", "Jail is in the process of shutting down");
4421 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4422 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4423 "I", "Current number of child jails");
4424 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4425 "I", "Maximum number of child jails");
4427 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4428 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4430 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4431 "Jail NIS domainname");
4432 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4434 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4435 "LU", "Jail host ID");
4437 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4438 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4441 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4442 "Jail IPv4 address virtualization");
4443 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4444 "S,in_addr,a", "Jail IPv4 addresses");
4445 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4446 "B", "Do (not) use IPv4 source address selection rather than the "
4447 "primary jail IPv4 address.");
4450 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4451 "Jail IPv6 address virtualization");
4452 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4453 "S,in6_addr,a", "Jail IPv6 addresses");
4454 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4455 "B", "Do (not) use IPv6 source address selection rather than the "
4456 "primary jail IPv6 address.");
4459 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4460 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4461 "B", "Jail may set hostname");
4462 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4463 "B", "Jail may use SYSV IPC");
4464 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4465 "B", "Jail may create raw sockets");
4466 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4467 "B", "Jail may alter system file flags");
4468 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4469 "B", "Jail may set file quotas");
4470 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4471 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4473 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4474 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4475 "B", "Jail may mount/unmount jail-friendly file systems in general");
4476 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4477 "B", "Jail may mount the devfs file system");
4478 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4479 "B", "Jail may mount the fdescfs file system");
4480 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4481 "B", "Jail may mount the nullfs file system");
4482 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4483 "B", "Jail may mount the procfs file system");
4484 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4485 "B", "Jail may mount the tmpfs file system");
4486 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4487 "B", "Jail may mount the zfs file system");
4491 prison_racct_foreach(void (*callback)(struct racct *racct,
4492 void *arg2, void *arg3), void *arg2, void *arg3)
4494 struct prison_racct *prr;
4496 ASSERT_RACCT_ENABLED();
4498 sx_slock(&allprison_lock);
4499 LIST_FOREACH(prr, &allprison_racct, prr_next)
4500 (callback)(prr->prr_racct, arg2, arg3);
4501 sx_sunlock(&allprison_lock);
4504 static struct prison_racct *
4505 prison_racct_find_locked(const char *name)
4507 struct prison_racct *prr;
4509 ASSERT_RACCT_ENABLED();
4510 sx_assert(&allprison_lock, SA_XLOCKED);
4512 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4515 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4516 if (strcmp(name, prr->prr_name) != 0)
4519 /* Found prison_racct with a matching name? */
4520 prison_racct_hold(prr);
4524 /* Add new prison_racct. */
4525 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4526 racct_create(&prr->prr_racct);
4528 strcpy(prr->prr_name, name);
4529 refcount_init(&prr->prr_refcount, 1);
4530 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4535 struct prison_racct *
4536 prison_racct_find(const char *name)
4538 struct prison_racct *prr;
4540 ASSERT_RACCT_ENABLED();
4542 sx_xlock(&allprison_lock);
4543 prr = prison_racct_find_locked(name);
4544 sx_xunlock(&allprison_lock);
4549 prison_racct_hold(struct prison_racct *prr)
4552 ASSERT_RACCT_ENABLED();
4554 refcount_acquire(&prr->prr_refcount);
4558 prison_racct_free_locked(struct prison_racct *prr)
4561 ASSERT_RACCT_ENABLED();
4562 sx_assert(&allprison_lock, SA_XLOCKED);
4564 if (refcount_release(&prr->prr_refcount)) {
4565 racct_destroy(&prr->prr_racct);
4566 LIST_REMOVE(prr, prr_next);
4567 free(prr, M_PRISON_RACCT);
4572 prison_racct_free(struct prison_racct *prr)
4576 ASSERT_RACCT_ENABLED();
4577 sx_assert(&allprison_lock, SA_UNLOCKED);
4579 old = prr->prr_refcount;
4580 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4583 sx_xlock(&allprison_lock);
4584 prison_racct_free_locked(prr);
4585 sx_xunlock(&allprison_lock);
4589 prison_racct_attach(struct prison *pr)
4591 struct prison_racct *prr;
4593 ASSERT_RACCT_ENABLED();
4594 sx_assert(&allprison_lock, SA_XLOCKED);
4596 prr = prison_racct_find_locked(pr->pr_name);
4597 KASSERT(prr != NULL, ("cannot find prison_racct"));
4599 pr->pr_prison_racct = prr;
4603 * Handle jail renaming. From the racct point of view, renaming means
4604 * moving from one prison_racct to another.
4607 prison_racct_modify(struct prison *pr)
4611 struct prison_racct *oldprr;
4613 ASSERT_RACCT_ENABLED();
4615 sx_slock(&allproc_lock);
4616 sx_xlock(&allprison_lock);
4618 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4619 sx_xunlock(&allprison_lock);
4620 sx_sunlock(&allproc_lock);
4624 oldprr = pr->pr_prison_racct;
4625 pr->pr_prison_racct = NULL;
4627 prison_racct_attach(pr);
4630 * Move resource utilisation records.
4632 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4635 * Force rctl to reattach rules to processes.
4637 FOREACH_PROC_IN_SYSTEM(p) {
4639 cred = crhold(p->p_ucred);
4641 racct_proc_ucred_changed(p, cred, cred);
4645 sx_sunlock(&allproc_lock);
4646 prison_racct_free_locked(oldprr);
4647 sx_xunlock(&allprison_lock);
4651 prison_racct_detach(struct prison *pr)
4654 ASSERT_RACCT_ENABLED();
4655 sx_assert(&allprison_lock, SA_UNLOCKED);
4657 if (pr->pr_prison_racct == NULL)
4659 prison_racct_free(pr->pr_prison_racct);
4660 pr->pr_prison_racct = NULL;
4667 db_show_prison(struct prison *pr)
4670 #if defined(INET) || defined(INET6)
4675 char ip6buf[INET6_ADDRSTRLEN];
4678 db_printf("prison %p:\n", pr);
4679 db_printf(" jid = %d\n", pr->pr_id);
4680 db_printf(" name = %s\n", pr->pr_name);
4681 db_printf(" parent = %p\n", pr->pr_parent);
4682 db_printf(" ref = %d\n", pr->pr_ref);
4683 db_printf(" uref = %d\n", pr->pr_uref);
4684 db_printf(" path = %s\n", pr->pr_path);
4685 db_printf(" cpuset = %d\n", pr->pr_cpuset
4686 ? pr->pr_cpuset->cs_id : -1);
4688 db_printf(" vnet = %p\n", pr->pr_vnet);
4690 db_printf(" root = %p\n", pr->pr_root);
4691 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4692 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4693 db_printf(" children.max = %d\n", pr->pr_childmax);
4694 db_printf(" children.cur = %d\n", pr->pr_childcount);
4695 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4696 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4697 db_printf(" flags = 0x%x", pr->pr_flags);
4698 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4700 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4701 db_printf(" %s", pr_flag_names[fi]);
4702 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4704 jsf = pr->pr_flags &
4705 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4706 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4707 pr_flag_jailsys[fi].disable &&
4708 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4709 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4712 db_printf(" allow = 0x%x", pr->pr_allow);
4713 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4715 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4716 db_printf(" %s", pr_allow_names[fi]);
4718 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4719 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4720 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4721 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4722 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4724 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4725 for (ii = 0; ii < pr->pr_ip4s; ii++)
4726 db_printf(" %s %s\n",
4727 ii == 0 ? "ip4.addr =" : " ",
4728 inet_ntoa(pr->pr_ip4[ii]));
4731 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4732 for (ii = 0; ii < pr->pr_ip6s; ii++)
4733 db_printf(" %s %s\n",
4734 ii == 0 ? "ip6.addr =" : " ",
4735 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4739 DB_SHOW_COMMAND(prison, db_show_prison_command)
4745 * Show all prisons in the list, and prison0 which is not
4748 db_show_prison(&prison0);
4749 if (!db_pager_quit) {
4750 TAILQ_FOREACH(pr, &allprison, pr_list) {
4762 /* Look for a prison with the ID and with references. */
4763 TAILQ_FOREACH(pr, &allprison, pr_list)
4764 if (pr->pr_id == addr && pr->pr_ref > 0)
4767 /* Look again, without requiring a reference. */
4768 TAILQ_FOREACH(pr, &allprison, pr_list)
4769 if (pr->pr_id == addr)
4772 /* Assume address points to a valid prison. */
4773 pr = (struct prison *)addr;