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",
211 const size_t pr_allow_names_size = sizeof(pr_allow_names);
213 static char *pr_allow_nonames[] = {
214 "allow.noset_hostname",
216 "allow.noraw_sockets",
221 "allow.mount.nodevfs",
222 "allow.mount.nonullfs",
224 "allow.mount.noprocfs",
225 "allow.mount.notmpfs",
227 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
229 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
230 #define JAIL_DEFAULT_ENFORCE_STATFS 2
231 #define JAIL_DEFAULT_DEVFS_RSNUM 0
232 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
233 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
234 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
235 #if defined(INET) || defined(INET6)
236 static unsigned jail_max_af_ips = 255;
241 qcmp_v4(const void *ip1, const void *ip2)
246 * We need to compare in HBO here to get the list sorted as expected
247 * by the result of the code. Sorting NBO addresses gives you
248 * interesting results. If you do not understand, do not try.
250 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
251 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
254 * Do not simply return the difference of the two numbers, the int is
268 qcmp_v6(const void *ip1, const void *ip2)
270 const struct in6_addr *ia6a, *ia6b;
273 ia6a = (const struct in6_addr *)ip1;
274 ia6b = (const struct in6_addr *)ip2;
277 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
278 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
280 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
293 sys_jail(struct thread *td, struct jail_args *uap)
299 error = copyin(uap->jail, &version, sizeof(uint32_t));
308 /* FreeBSD single IPv4 jails. */
309 bzero(&j, sizeof(struct jail));
310 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
313 j.version = j0.version;
315 j.hostname = j0.hostname;
316 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
322 * Version 1 was used by multi-IPv4 jail implementations
323 * that never made it into the official kernel.
327 case 2: /* JAIL_API_VERSION */
328 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
329 error = copyin(uap->jail, &j, sizeof(struct jail));
335 /* Sci-Fi jails are not supported, sorry. */
338 return (kern_jail(td, &j));
342 kern_jail(struct thread *td, struct jail *j)
344 struct iovec optiov[2 * (4
345 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
354 char *u_path, *u_hostname, *u_name;
357 struct in_addr *u_ip4;
360 struct in6_addr *u_ip6;
363 int error, enforce_statfs, fi;
365 bzero(&optiov, sizeof(optiov));
366 opt.uio_iov = optiov;
370 opt.uio_segflg = UIO_SYSSPACE;
371 opt.uio_rw = UIO_READ;
374 /* Set permissions for top-level jails from sysctls. */
375 if (!jailed(td->td_ucred)) {
376 for (fi = 0; fi < sizeof(pr_allow_names) /
377 sizeof(pr_allow_names[0]); fi++) {
378 optiov[opt.uio_iovcnt].iov_base =
379 (jail_default_allow & (1 << fi))
380 ? pr_allow_names[fi] : pr_allow_nonames[fi];
381 optiov[opt.uio_iovcnt].iov_len =
382 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
385 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
386 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
388 enforce_statfs = jail_default_enforce_statfs;
389 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
390 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
394 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
396 ip4s = (j->version == 0) ? 1 : j->ip4s;
397 if (ip4s > jail_max_af_ips)
399 tmplen += ip4s * sizeof(struct in_addr);
405 if (j->ip6s > jail_max_af_ips)
407 tmplen += j->ip6s * sizeof(struct in6_addr);
412 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
413 u_hostname = u_path + MAXPATHLEN;
414 u_name = u_hostname + MAXHOSTNAMELEN;
416 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
420 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
422 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
425 optiov[opt.uio_iovcnt].iov_base = "path";
426 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
428 optiov[opt.uio_iovcnt].iov_base = u_path;
429 error = copyinstr(j->path, u_path, MAXPATHLEN,
430 &optiov[opt.uio_iovcnt].iov_len);
432 free(u_path, M_TEMP);
436 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
437 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
439 optiov[opt.uio_iovcnt].iov_base = u_hostname;
440 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
441 &optiov[opt.uio_iovcnt].iov_len);
443 free(u_path, M_TEMP);
447 if (j->jailname != NULL) {
448 optiov[opt.uio_iovcnt].iov_base = "name";
449 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
451 optiov[opt.uio_iovcnt].iov_base = u_name;
452 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
453 &optiov[opt.uio_iovcnt].iov_len);
455 free(u_path, M_TEMP);
461 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
462 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
464 optiov[opt.uio_iovcnt].iov_base = u_ip4;
465 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
467 u_ip4->s_addr = j->ip4s;
469 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
471 free(u_path, M_TEMP);
478 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
479 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
481 optiov[opt.uio_iovcnt].iov_base = u_ip6;
482 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
483 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
485 free(u_path, M_TEMP);
490 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
491 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
492 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
493 free(u_path, M_TEMP);
499 * struct jail_set_args {
500 * struct iovec *iovp;
501 * unsigned int iovcnt;
506 sys_jail_set(struct thread *td, struct jail_set_args *uap)
511 /* Check that we have an even number of iovecs. */
515 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
518 error = kern_jail_set(td, auio, uap->flags);
524 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
531 struct in6_addr *ip6;
534 struct vfsoptlist *opts;
535 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
537 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
539 #if defined(INET) || defined(INET6)
544 size_t namelen, onamelen;
545 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
546 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
547 int fi, jid, jsys, len, level;
548 int childmax, rsnum, slevel;
549 int fullpath_disabled;
550 #if defined(INET) || defined(INET6)
559 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
563 error = priv_check(td, PRIV_JAIL_SET);
564 if (!error && (flags & JAIL_ATTACH))
565 error = priv_check(td, PRIV_JAIL_ATTACH);
568 mypr = ppr = td->td_ucred->cr_prison;
569 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
571 if (flags & ~JAIL_SET_MASK)
575 * Check all the parameters before committing to anything. Not all
576 * errors can be caught early, but we may as well try. Also, this
577 * takes care of some expensive stuff (path lookup) before getting
578 * the allprison lock.
580 * XXX Jails are not filesystems, and jail parameters are not mount
581 * options. But it makes more sense to re-use the vfsopt code
582 * than duplicate it under a different name.
584 error = vfs_buildopts(optuio, &opts);
595 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
601 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
610 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
618 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
623 else if (enforce < 0 || enforce > 2) {
629 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
637 pr_flags = ch_flags = 0;
638 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
640 if (pr_flag_names[fi] == NULL)
642 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
643 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
645 ch_flags |= pr_flags;
646 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
648 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
655 case JAIL_SYS_DISABLE:
656 if (!pr_flag_jailsys[fi].disable) {
660 pr_flags |= pr_flag_jailsys[fi].disable;
663 pr_flags |= pr_flag_jailsys[fi].new;
665 case JAIL_SYS_INHERIT:
672 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
674 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
675 && !(pr_flags & PR_PERSIST)) {
677 vfs_opterror(opts, "new jail must persist or attach");
681 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
683 vfs_opterror(opts, "vnet cannot be changed after creation");
688 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
690 vfs_opterror(opts, "ip4 cannot be changed after creation");
695 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
697 vfs_opterror(opts, "ip6 cannot be changed after creation");
702 pr_allow = ch_allow = 0;
703 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
705 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
706 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
708 ch_allow |= pr_allow;
710 error = vfs_getopt(opts, "name", (void **)&name, &len);
716 if (len == 0 || name[len - 1] != '\0') {
720 if (len > MAXHOSTNAMELEN) {
721 error = ENAMETOOLONG;
726 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
734 if (len == 0 || host[len - 1] != '\0') {
738 if (len > MAXHOSTNAMELEN) {
739 error = ENAMETOOLONG;
744 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
752 if (len == 0 || domain[len - 1] != '\0') {
756 if (len > MAXHOSTNAMELEN) {
757 error = ENAMETOOLONG;
762 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
770 if (len == 0 || uuid[len - 1] != '\0') {
774 if (len > HOSTUUIDLEN) {
775 error = ENAMETOOLONG;
780 #ifdef COMPAT_FREEBSD32
781 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
784 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
788 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
800 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
802 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
805 else if (ip4s & (sizeof(*ip4) - 1)) {
809 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
811 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
813 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
814 ip4s /= sizeof(*ip4);
815 if (ip4s > jail_max_af_ips) {
817 vfs_opterror(opts, "too many IPv4 addresses");
820 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
821 bcopy(op, ip4, ip4s * sizeof(*ip4));
823 * IP addresses are all sorted but ip[0] to preserve
824 * the primary IP address as given from userland.
825 * This special IP is used for unbound outgoing
826 * connections as well for "loopback" traffic in case
827 * source address selection cannot find any more fitting
828 * address to connect from.
831 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
833 * Check for duplicate addresses and do some simple
834 * zero and broadcast checks. If users give other bogus
835 * addresses it is their problem.
837 * We do not have to care about byte order for these
838 * checks so we will do them in NBO.
840 for (ii = 0; ii < ip4s; ii++) {
841 if (ip4[ii].s_addr == INADDR_ANY ||
842 ip4[ii].s_addr == INADDR_BROADCAST) {
847 (ip4[0].s_addr == ip4[ii+1].s_addr ||
848 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
858 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
860 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
863 else if (ip6s & (sizeof(*ip6) - 1)) {
867 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
869 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
871 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
872 ip6s /= sizeof(*ip6);
873 if (ip6s > jail_max_af_ips) {
875 vfs_opterror(opts, "too many IPv6 addresses");
878 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
879 bcopy(op, ip6, ip6s * sizeof(*ip6));
881 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
882 for (ii = 0; ii < ip6s; ii++) {
883 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
888 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
889 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
899 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
900 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
903 "vnet jails cannot have IP address restrictions");
908 fullpath_disabled = 0;
910 error = vfs_getopt(opts, "path", (void **)&path, &len);
916 if (flags & JAIL_UPDATE) {
919 "path cannot be changed after creation");
922 if (len == 0 || path[len - 1] != '\0') {
926 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
932 NDFREE(&nd, NDF_ONLY_PNBUF);
933 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
934 strlcpy(g_path, path, MAXPATHLEN);
935 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
938 else if (error == ENODEV) {
939 /* proceed if sysctl debug.disablefullpath == 1 */
940 fullpath_disabled = 1;
941 if (len < 2 || (len == 2 && path[0] == '/'))
944 /* exit on other errors */
947 if (root->v_type != VDIR) {
953 if (fullpath_disabled) {
954 /* Leave room for a real-root full pathname. */
955 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
956 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
957 error = ENAMETOOLONG;
964 * Grab the allprison lock before letting modules check their
965 * parameters. Once we have it, do not let go so we'll have a
966 * consistent view of the OSD list.
968 sx_xlock(&allprison_lock);
969 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
971 goto done_unlock_list;
973 /* By now, all parameters should have been noted. */
974 TAILQ_FOREACH(opt, opts, link) {
975 if (!opt->seen && strcmp(opt->name, "errmsg")) {
977 vfs_opterror(opts, "unknown parameter: %s", opt->name);
978 goto done_unlock_list;
983 * See if we are creating a new record or updating an existing one.
984 * This abuses the file error codes ENOENT and EEXIST.
986 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
989 vfs_opterror(opts, "no valid operation (create or update)");
990 goto done_unlock_list;
994 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
995 namelc = strrchr(name, '.');
996 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1002 * See if a requested jid already exists. There is an
1003 * information leak here if the jid exists but is not within
1004 * the caller's jail hierarchy. Jail creators will get EEXIST
1005 * even though they cannot see the jail, and CREATE | UPDATE
1006 * will return ENOENT which is not normally a valid error.
1010 vfs_opterror(opts, "negative jid");
1011 goto done_unlock_list;
1013 pr = prison_find(jid);
1015 ppr = pr->pr_parent;
1016 /* Create: jid must not exist. */
1017 if (cuflags == JAIL_CREATE) {
1018 mtx_unlock(&pr->pr_mtx);
1020 vfs_opterror(opts, "jail %d already exists",
1022 goto done_unlock_list;
1024 if (!prison_ischild(mypr, pr)) {
1025 mtx_unlock(&pr->pr_mtx);
1027 } else if (pr->pr_uref == 0) {
1028 if (!(flags & JAIL_DYING)) {
1029 mtx_unlock(&pr->pr_mtx);
1031 vfs_opterror(opts, "jail %d is dying",
1033 goto done_unlock_list;
1034 } else if ((flags & JAIL_ATTACH) ||
1035 (pr_flags & PR_PERSIST)) {
1037 * A dying jail might be resurrected
1038 * (via attach or persist), but first
1039 * it must determine if another jail
1040 * has claimed its name. Accomplish
1041 * this by implicitly re-setting the
1045 name = prison_name(mypr, pr);
1050 /* Update: jid must exist. */
1051 if (cuflags == JAIL_UPDATE) {
1053 vfs_opterror(opts, "jail %d not found", jid);
1054 goto done_unlock_list;
1059 * If the caller provided a name, look for a jail by that name.
1060 * This has different semantics for creates and updates keyed by jid
1061 * (where the name must not already exist in a different jail),
1062 * and updates keyed by the name itself (where the name must exist
1063 * because that is the jail being updated).
1066 namelc = strrchr(name, '.');
1071 * This is a hierarchical name. Split it into the
1072 * parent and child names, and make sure the parent
1073 * exists or matches an already found jail.
1077 if (strncmp(name, ppr->pr_name, namelc - name)
1078 || ppr->pr_name[namelc - name] != '\0') {
1079 mtx_unlock(&pr->pr_mtx);
1082 "cannot change jail's parent");
1083 goto done_unlock_list;
1086 ppr = prison_find_name(mypr, name);
1090 "jail \"%s\" not found", name);
1091 goto done_unlock_list;
1093 mtx_unlock(&ppr->pr_mtx);
1097 if (name[0] != '\0') {
1099 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1102 FOREACH_PRISON_CHILD(ppr, tpr) {
1103 if (tpr != pr && tpr->pr_ref > 0 &&
1104 !strcmp(tpr->pr_name + namelen, name)) {
1106 cuflags != JAIL_CREATE) {
1107 mtx_lock(&tpr->pr_mtx);
1108 if (tpr->pr_ref > 0) {
1113 if (tpr->pr_uref > 0) {
1119 mtx_unlock(&tpr->pr_mtx);
1120 } else if (tpr->pr_uref > 0) {
1122 * Create, or update(jid):
1123 * name must not exist in an
1124 * active sibling jail.
1128 mtx_unlock(&pr->pr_mtx);
1130 "jail \"%s\" already exists",
1132 goto done_unlock_list;
1136 /* If no active jail is found, use a dying one. */
1137 if (deadpr != NULL && pr == NULL) {
1138 if (flags & JAIL_DYING) {
1139 mtx_lock(&deadpr->pr_mtx);
1140 if (deadpr->pr_ref == 0) {
1141 mtx_unlock(&deadpr->pr_mtx);
1145 } else if (cuflags == JAIL_UPDATE) {
1148 "jail \"%s\" is dying", name);
1149 goto done_unlock_list;
1152 /* Update: name must exist if no jid. */
1153 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1155 vfs_opterror(opts, "jail \"%s\" not found",
1157 goto done_unlock_list;
1161 /* Update: must provide a jid or name. */
1162 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1164 vfs_opterror(opts, "update specified no jail");
1165 goto done_unlock_list;
1168 /* If there's no prison to update, create a new one and link it in. */
1170 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1171 if (tpr->pr_childcount >= tpr->pr_childmax) {
1173 vfs_opterror(opts, "prison limit exceeded");
1174 goto done_unlock_list;
1177 mtx_lock(&ppr->pr_mtx);
1178 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1179 mtx_unlock(&ppr->pr_mtx);
1181 vfs_opterror(opts, "parent jail went away!");
1182 goto done_unlock_list;
1186 mtx_unlock(&ppr->pr_mtx);
1187 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1189 /* Find the next free jid. */
1192 if (jid == JAIL_MAX)
1194 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1195 if (tpr->pr_id < jid)
1197 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1198 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1201 if (jid == lastprid) {
1204 "no available jail IDs");
1206 prison_deref(ppr, PD_DEREF |
1207 PD_DEUREF | PD_LIST_XLOCKED);
1216 * The jail already has a jid (that did not yet exist),
1217 * so just find where to insert it.
1219 TAILQ_FOREACH(tpr, &allprison, pr_list)
1220 if (tpr->pr_id >= jid) {
1221 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1226 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1227 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1228 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1229 tpr->pr_childcount++;
1231 pr->pr_parent = ppr;
1234 /* Set some default values, and inherit some from the parent. */
1239 root = mypr->pr_root;
1242 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1243 pr->pr_flags |= PR_HOST;
1244 #if defined(INET) || defined(INET6)
1246 if (!(pr_flags & PR_VNET))
1250 if (!(ch_flags & PR_IP4_USER))
1252 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1253 else if (!(pr_flags & PR_IP4_USER)) {
1254 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1255 if (ppr->pr_ip4 != NULL) {
1256 pr->pr_ip4s = ppr->pr_ip4s;
1257 pr->pr_ip4 = malloc(pr->pr_ip4s *
1258 sizeof(struct in_addr), M_PRISON,
1260 bcopy(ppr->pr_ip4, pr->pr_ip4,
1261 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1266 if (!(ch_flags & PR_IP6_USER))
1268 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1269 else if (!(pr_flags & PR_IP6_USER)) {
1270 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1271 if (ppr->pr_ip6 != NULL) {
1272 pr->pr_ip6s = ppr->pr_ip6s;
1273 pr->pr_ip6 = malloc(pr->pr_ip6s *
1274 sizeof(struct in6_addr), M_PRISON,
1276 bcopy(ppr->pr_ip6, pr->pr_ip6,
1277 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1283 /* Source address selection is always on by default. */
1284 pr->pr_flags |= _PR_IP_SADDRSEL;
1286 pr->pr_securelevel = ppr->pr_securelevel;
1287 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1288 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1289 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1291 LIST_INIT(&pr->pr_children);
1292 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1295 /* Allocate a new vnet if specified. */
1296 pr->pr_vnet = (pr_flags & PR_VNET)
1297 ? vnet_alloc() : ppr->pr_vnet;
1300 * Allocate a dedicated cpuset for each jail.
1301 * Unlike other initial settings, this may return an erorr.
1303 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1305 prison_deref(pr, PD_LIST_XLOCKED);
1309 mtx_lock(&pr->pr_mtx);
1311 * New prisons do not yet have a reference, because we do not
1312 * want other to see the incomplete prison once the
1313 * allprison_lock is downgraded.
1318 * Grab a reference for existing prisons, to ensure they
1319 * continue to exist for the duration of the call.
1322 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1323 if ((pr->pr_flags & PR_VNET) &&
1324 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1327 "vnet jails cannot have IP address restrictions");
1328 goto done_deref_locked;
1332 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1335 "ip4 cannot be changed after creation");
1336 goto done_deref_locked;
1340 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1343 "ip6 cannot be changed after creation");
1344 goto done_deref_locked;
1349 /* Do final error checking before setting anything. */
1351 if (slevel < ppr->pr_securelevel) {
1353 goto done_deref_locked;
1357 if (childmax >= ppr->pr_childmax) {
1359 goto done_deref_locked;
1363 if (enforce < ppr->pr_enforce_statfs) {
1365 goto done_deref_locked;
1370 * devfs_rsnum is a uint16_t
1372 if (rsnum < 0 || rsnum > 65535) {
1374 goto done_deref_locked;
1377 * Nested jails always inherit parent's devfs ruleset
1379 if (jailed(td->td_ucred)) {
1380 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1382 goto done_deref_locked;
1384 rsnum = ppr->pr_devfs_rsnum;
1389 if (ppr->pr_flags & PR_IP4) {
1391 * Make sure the new set of IP addresses is a
1392 * subset of the parent's list. Don't worry
1393 * about the parent being unlocked, as any
1394 * setting is done with allprison_lock held.
1396 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1397 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1399 if (ij == ppr->pr_ip4s) {
1401 goto done_deref_locked;
1404 for (ii = ij = 1; ii < ip4s; ii++) {
1405 if (ip4[ii].s_addr ==
1406 ppr->pr_ip4[0].s_addr)
1408 for (; ij < ppr->pr_ip4s; ij++)
1409 if (ip4[ii].s_addr ==
1410 ppr->pr_ip4[ij].s_addr)
1412 if (ij == ppr->pr_ip4s)
1415 if (ij == ppr->pr_ip4s) {
1417 goto done_deref_locked;
1422 * Check for conflicting IP addresses. We permit them
1423 * if there is no more than one IP on each jail. If
1424 * there is a duplicate on a jail with more than one
1425 * IP stop checking and return error.
1429 for (; tppr != &prison0; tppr = tppr->pr_parent)
1430 if (tppr->pr_flags & PR_VNET)
1433 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1436 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1438 tpr->pr_uref == 0) {
1442 if (!(tpr->pr_flags & PR_IP4_USER))
1445 if (tpr->pr_ip4 == NULL ||
1446 (ip4s == 1 && tpr->pr_ip4s == 1))
1448 for (ii = 0; ii < ip4s; ii++) {
1449 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1452 "IPv4 addresses clash");
1453 goto done_deref_locked;
1461 if (ppr->pr_flags & PR_IP6) {
1463 * Make sure the new set of IP addresses is a
1464 * subset of the parent's list.
1466 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1467 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1470 if (ij == ppr->pr_ip6s) {
1472 goto done_deref_locked;
1475 for (ii = ij = 1; ii < ip6s; ii++) {
1476 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1479 for (; ij < ppr->pr_ip6s; ij++)
1480 if (IN6_ARE_ADDR_EQUAL(
1481 &ip6[ii], &ppr->pr_ip6[ij]))
1483 if (ij == ppr->pr_ip6s)
1486 if (ij == ppr->pr_ip6s) {
1488 goto done_deref_locked;
1492 /* Check for conflicting IP addresses. */
1495 for (; tppr != &prison0; tppr = tppr->pr_parent)
1496 if (tppr->pr_flags & PR_VNET)
1499 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1502 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1504 tpr->pr_uref == 0) {
1508 if (!(tpr->pr_flags & PR_IP6_USER))
1511 if (tpr->pr_ip6 == NULL ||
1512 (ip6s == 1 && tpr->pr_ip6s == 1))
1514 for (ii = 0; ii < ip6s; ii++) {
1515 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1518 "IPv6 addresses clash");
1519 goto done_deref_locked;
1525 onamelen = namelen = 0;
1527 /* Give a default name of the jid. */
1528 if (name[0] == '\0')
1529 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1530 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1534 "name cannot be numeric (unless it is the jid)");
1535 goto done_deref_locked;
1538 * Make sure the name isn't too long for the prison or its
1541 onamelen = strlen(pr->pr_name);
1542 namelen = strlen(name);
1543 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1544 error = ENAMETOOLONG;
1545 goto done_deref_locked;
1547 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1548 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1549 sizeof(pr->pr_name)) {
1550 error = ENAMETOOLONG;
1551 goto done_deref_locked;
1555 if (pr_allow & ~ppr->pr_allow) {
1557 goto done_deref_locked;
1560 /* Set the parameters of the prison. */
1563 if (pr_flags & PR_IP4_USER) {
1564 pr->pr_flags |= PR_IP4;
1565 free(pr->pr_ip4, M_PRISON);
1569 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1571 if (tpr->pr_flags & PR_VNET) {
1576 if (prison_restrict_ip4(tpr, NULL)) {
1585 if (pr_flags & PR_IP6_USER) {
1586 pr->pr_flags |= PR_IP6;
1587 free(pr->pr_ip6, M_PRISON);
1591 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1593 if (tpr->pr_flags & PR_VNET) {
1598 if (prison_restrict_ip6(tpr, NULL)) {
1606 pr->pr_securelevel = slevel;
1607 /* Set all child jails to be at least this level. */
1608 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1609 if (tpr->pr_securelevel < slevel)
1610 tpr->pr_securelevel = slevel;
1613 pr->pr_childmax = childmax;
1614 /* Set all child jails to under this limit. */
1615 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1616 if (tpr->pr_childmax > childmax - level)
1617 tpr->pr_childmax = childmax > level
1618 ? childmax - level : 0;
1621 pr->pr_enforce_statfs = enforce;
1622 /* Pass this restriction on to the children. */
1623 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1624 if (tpr->pr_enforce_statfs < enforce)
1625 tpr->pr_enforce_statfs = enforce;
1628 pr->pr_devfs_rsnum = rsnum;
1629 /* Pass this restriction on to the children. */
1630 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1631 tpr->pr_devfs_rsnum = rsnum;
1634 if (ppr == &prison0)
1635 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1637 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1638 ppr->pr_name, name);
1639 /* Change this component of child names. */
1640 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1641 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1642 strlen(tpr->pr_name + onamelen) + 1);
1643 bcopy(pr->pr_name, tpr->pr_name, namelen);
1647 /* Try to keep a real-rooted full pathname. */
1648 if (fullpath_disabled && path[0] == '/' &&
1649 strcmp(mypr->pr_path, "/"))
1650 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1651 mypr->pr_path, path);
1653 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1656 if (PR_HOST & ch_flags & ~pr_flags) {
1657 if (pr->pr_flags & PR_HOST) {
1659 * Copy the parent's host info. As with pr_ip4 above,
1660 * the lack of a lock on the parent is not a problem;
1661 * it is always set with allprison_lock at least
1662 * shared, and is held exclusively here.
1664 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1665 sizeof(pr->pr_hostname));
1666 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1667 sizeof(pr->pr_domainname));
1668 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1669 sizeof(pr->pr_hostuuid));
1670 pr->pr_hostid = pr->pr_parent->pr_hostid;
1672 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1673 /* Set this prison, and any descendants without PR_HOST. */
1675 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1677 strlcpy(pr->pr_domainname, domain,
1678 sizeof(pr->pr_domainname));
1680 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1682 pr->pr_hostid = hid;
1683 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1684 if (tpr->pr_flags & PR_HOST)
1688 strlcpy(tpr->pr_hostname,
1690 sizeof(tpr->pr_hostname));
1692 strlcpy(tpr->pr_domainname,
1694 sizeof(tpr->pr_domainname));
1696 strlcpy(tpr->pr_hostuuid,
1698 sizeof(tpr->pr_hostuuid));
1700 tpr->pr_hostid = hid;
1704 if ((tallow = ch_allow & ~pr_allow)) {
1705 /* Clear allow bits in all children. */
1706 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1707 tpr->pr_allow &= ~tallow;
1709 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1711 * Persistent prisons get an extra reference, and prisons losing their
1712 * persist flag lose that reference. Only do this for existing prisons
1713 * for now, so new ones will remain unseen until after the module
1714 * handlers have completed.
1716 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1717 if (pr_flags & PR_PERSIST) {
1725 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1726 mtx_unlock(&pr->pr_mtx);
1730 prison_racct_attach(pr);
1733 /* Locks may have prevented a complete restriction of child IP
1734 * addresses. If so, allocate some more memory and try again.
1739 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1740 mtx_lock(&pr->pr_mtx);
1742 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1744 if (tpr->pr_flags & PR_VNET) {
1749 if (prison_restrict_ip4(tpr, ip4)) {
1756 mtx_unlock(&pr->pr_mtx);
1762 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1763 mtx_lock(&pr->pr_mtx);
1765 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1767 if (tpr->pr_flags & PR_VNET) {
1772 if (prison_restrict_ip6(tpr, ip6)) {
1779 mtx_unlock(&pr->pr_mtx);
1783 /* Let the modules do their work. */
1784 sx_downgrade(&allprison_lock);
1786 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1788 prison_deref(pr, PD_LIST_SLOCKED);
1792 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1794 prison_deref(pr, created
1796 : PD_DEREF | PD_LIST_SLOCKED);
1800 /* Attach this process to the prison if requested. */
1801 if (flags & JAIL_ATTACH) {
1802 mtx_lock(&pr->pr_mtx);
1803 error = do_jail_attach(td, pr);
1805 vfs_opterror(opts, "attach failed");
1807 prison_deref(pr, PD_DEREF);
1814 sx_sunlock(&allprison_lock);
1815 prison_racct_modify(pr);
1816 sx_slock(&allprison_lock);
1820 td->td_retval[0] = pr->pr_id;
1823 * Now that it is all there, drop the temporary reference from existing
1824 * prisons. Or add a reference to newly created persistent prisons
1825 * (which was not done earlier so that the prison would not be publicly
1829 prison_deref(pr, (flags & JAIL_ATTACH)
1831 : PD_DEREF | PD_LIST_SLOCKED);
1833 if (pr_flags & PR_PERSIST) {
1834 mtx_lock(&pr->pr_mtx);
1837 mtx_unlock(&pr->pr_mtx);
1839 if (!(flags & JAIL_ATTACH))
1840 sx_sunlock(&allprison_lock);
1846 prison_deref(pr, created
1847 ? PD_LOCKED | PD_LIST_XLOCKED
1848 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1851 sx_xunlock(&allprison_lock);
1857 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1858 if (errmsg_len > 0) {
1859 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1860 if (errmsg_pos > 0) {
1861 if (optuio->uio_segflg == UIO_SYSSPACE)
1863 optuio->uio_iov[errmsg_pos].iov_base,
1867 optuio->uio_iov[errmsg_pos].iov_base,
1874 free(ip4, M_PRISON);
1877 free(ip6, M_PRISON);
1880 free(g_path, M_TEMP);
1887 * struct jail_get_args {
1888 * struct iovec *iovp;
1889 * unsigned int iovcnt;
1894 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1899 /* Check that we have an even number of iovecs. */
1900 if (uap->iovcnt & 1)
1903 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1906 error = kern_jail_get(td, auio, uap->flags);
1908 error = copyout(auio->uio_iov, uap->iovp,
1909 uap->iovcnt * sizeof (struct iovec));
1915 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1917 struct prison *pr, *mypr;
1919 struct vfsoptlist *opts;
1920 char *errmsg, *name;
1921 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1923 if (flags & ~JAIL_GET_MASK)
1926 /* Get the parameter list. */
1927 error = vfs_buildopts(optuio, &opts);
1930 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1931 mypr = td->td_ucred->cr_prison;
1934 * Find the prison specified by one of: lastjid, jid, name.
1936 sx_slock(&allprison_lock);
1937 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1939 TAILQ_FOREACH(pr, &allprison, pr_list) {
1940 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1941 mtx_lock(&pr->pr_mtx);
1942 if (pr->pr_ref > 0 &&
1943 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1945 mtx_unlock(&pr->pr_mtx);
1951 vfs_opterror(opts, "no jail after %d", jid);
1952 goto done_unlock_list;
1953 } else if (error != ENOENT)
1954 goto done_unlock_list;
1956 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1959 pr = prison_find_child(mypr, jid);
1961 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1962 mtx_unlock(&pr->pr_mtx);
1964 vfs_opterror(opts, "jail %d is dying",
1966 goto done_unlock_list;
1971 vfs_opterror(opts, "jail %d not found", jid);
1972 goto done_unlock_list;
1974 } else if (error != ENOENT)
1975 goto done_unlock_list;
1977 error = vfs_getopt(opts, "name", (void **)&name, &len);
1979 if (len == 0 || name[len - 1] != '\0') {
1981 goto done_unlock_list;
1983 pr = prison_find_name(mypr, name);
1985 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1986 mtx_unlock(&pr->pr_mtx);
1988 vfs_opterror(opts, "jail \"%s\" is dying",
1990 goto done_unlock_list;
1995 vfs_opterror(opts, "jail \"%s\" not found", name);
1996 goto done_unlock_list;
1997 } else if (error != ENOENT)
1998 goto done_unlock_list;
2000 vfs_opterror(opts, "no jail specified");
2002 goto done_unlock_list;
2005 /* Get the parameters of the prison. */
2008 td->td_retval[0] = pr->pr_id;
2009 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2010 if (error != 0 && error != ENOENT)
2012 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2013 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2014 if (error != 0 && error != ENOENT)
2016 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2017 if (error != 0 && error != ENOENT)
2019 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2020 sizeof(pr->pr_cpuset->cs_id));
2021 if (error != 0 && error != ENOENT)
2023 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2024 if (error != 0 && error != ENOENT)
2027 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2028 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2029 if (error != 0 && error != ENOENT)
2033 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2034 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2035 if (error != 0 && error != ENOENT)
2038 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2039 sizeof(pr->pr_securelevel));
2040 if (error != 0 && error != ENOENT)
2042 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2043 sizeof(pr->pr_childcount));
2044 if (error != 0 && error != ENOENT)
2046 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2047 sizeof(pr->pr_childmax));
2048 if (error != 0 && error != ENOENT)
2050 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2051 if (error != 0 && error != ENOENT)
2053 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2054 if (error != 0 && error != ENOENT)
2056 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2057 if (error != 0 && error != ENOENT)
2059 #ifdef COMPAT_FREEBSD32
2060 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2061 uint32_t hid32 = pr->pr_hostid;
2063 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2066 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2067 sizeof(pr->pr_hostid));
2068 if (error != 0 && error != ENOENT)
2070 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2071 sizeof(pr->pr_enforce_statfs));
2072 if (error != 0 && error != ENOENT)
2074 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2075 sizeof(pr->pr_devfs_rsnum));
2076 if (error != 0 && error != ENOENT)
2078 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2080 if (pr_flag_names[fi] == NULL)
2082 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2083 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2084 if (error != 0 && error != ENOENT)
2087 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2088 if (error != 0 && error != ENOENT)
2091 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2094 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2095 i = pr_flag_jailsys[fi].disable &&
2096 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2097 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2100 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2101 if (error != 0 && error != ENOENT)
2104 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2106 if (pr_allow_names[fi] == NULL)
2108 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2109 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2110 if (error != 0 && error != ENOENT)
2113 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2114 if (error != 0 && error != ENOENT)
2117 i = (pr->pr_uref == 0);
2118 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2119 if (error != 0 && error != ENOENT)
2122 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2123 if (error != 0 && error != ENOENT)
2126 /* Get the module parameters. */
2127 mtx_unlock(&pr->pr_mtx);
2129 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2132 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2134 /* By now, all parameters should have been noted. */
2135 TAILQ_FOREACH(opt, opts, link) {
2136 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2138 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2143 /* Write the fetched parameters back to userspace. */
2145 TAILQ_FOREACH(opt, opts, link) {
2146 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2147 pos = 2 * opt->pos + 1;
2148 optuio->uio_iov[pos].iov_len = opt->len;
2149 if (opt->value != NULL) {
2150 if (optuio->uio_segflg == UIO_SYSSPACE) {
2152 optuio->uio_iov[pos].iov_base,
2155 error = copyout(opt->value,
2156 optuio->uio_iov[pos].iov_base,
2167 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2171 sx_sunlock(&allprison_lock);
2173 if (error && errmsg_pos >= 0) {
2174 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2175 errmsg_pos = 2 * errmsg_pos + 1;
2176 if (errmsg_len > 0) {
2177 if (optuio->uio_segflg == UIO_SYSSPACE)
2179 optuio->uio_iov[errmsg_pos].iov_base,
2183 optuio->uio_iov[errmsg_pos].iov_base,
2193 * struct jail_remove_args {
2198 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2200 struct prison *pr, *cpr, *lpr, *tpr;
2203 error = priv_check(td, PRIV_JAIL_REMOVE);
2207 sx_xlock(&allprison_lock);
2208 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2210 sx_xunlock(&allprison_lock);
2214 /* Remove all descendants of this prison, then remove this prison. */
2216 pr->pr_flags |= PR_REMOVE;
2217 if (!LIST_EMPTY(&pr->pr_children)) {
2218 mtx_unlock(&pr->pr_mtx);
2220 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2221 mtx_lock(&cpr->pr_mtx);
2222 if (cpr->pr_ref > 0) {
2225 cpr->pr_flags |= PR_REMOVE;
2227 /* Already removed - do not do it again. */
2230 mtx_unlock(&cpr->pr_mtx);
2232 mtx_lock(&lpr->pr_mtx);
2233 prison_remove_one(lpr);
2234 sx_xlock(&allprison_lock);
2239 mtx_lock(&lpr->pr_mtx);
2240 prison_remove_one(lpr);
2241 sx_xlock(&allprison_lock);
2243 mtx_lock(&pr->pr_mtx);
2245 prison_remove_one(pr);
2250 prison_remove_one(struct prison *pr)
2255 /* If the prison was persistent, it is not anymore. */
2257 if (pr->pr_flags & PR_PERSIST) {
2260 pr->pr_flags &= ~PR_PERSIST;
2264 * jail_remove added a reference. If that's the only one, remove
2267 KASSERT(pr->pr_ref > 0,
2268 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2269 if (pr->pr_ref == 1) {
2271 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2275 mtx_unlock(&pr->pr_mtx);
2276 sx_xunlock(&allprison_lock);
2278 * Kill all processes unfortunate enough to be attached to this prison.
2280 sx_slock(&allproc_lock);
2281 LIST_FOREACH(p, &allproc, p_list) {
2283 if (p->p_state != PRS_NEW && p->p_ucred &&
2284 p->p_ucred->cr_prison == pr)
2285 kern_psignal(p, SIGKILL);
2288 sx_sunlock(&allproc_lock);
2289 /* Remove the temporary reference added by jail_remove. */
2290 prison_deref(pr, deuref | PD_DEREF);
2295 * struct jail_attach_args {
2300 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2305 error = priv_check(td, PRIV_JAIL_ATTACH);
2309 sx_slock(&allprison_lock);
2310 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2312 sx_sunlock(&allprison_lock);
2317 * Do not allow a process to attach to a prison that is not
2318 * considered to be "alive".
2320 if (pr->pr_uref == 0) {
2321 mtx_unlock(&pr->pr_mtx);
2322 sx_sunlock(&allprison_lock);
2326 return (do_jail_attach(td, pr));
2330 do_jail_attach(struct thread *td, struct prison *pr)
2334 struct ucred *newcred, *oldcred;
2338 * XXX: Note that there is a slight race here if two threads
2339 * in the same privileged process attempt to attach to two
2340 * different jails at the same time. It is important for
2341 * user processes not to do this, or they might end up with
2342 * a process root from one prison, but attached to the jail
2347 mtx_unlock(&pr->pr_mtx);
2349 /* Let modules do whatever they need to prepare for attaching. */
2350 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2352 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2355 sx_sunlock(&allprison_lock);
2358 * Reparent the newly attached process to this jail.
2360 ppr = td->td_ucred->cr_prison;
2362 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2366 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2367 if ((error = change_dir(pr->pr_root, td)) != 0)
2370 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2373 VOP_UNLOCK(pr->pr_root, 0);
2374 if ((error = change_root(pr->pr_root, td)))
2379 oldcred = p->p_ucred;
2381 crcopy(newcred, oldcred);
2382 newcred->cr_prison = pr;
2383 p->p_ucred = newcred;
2386 racct_proc_ucred_changed(p, oldcred, newcred);
2389 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2392 VOP_UNLOCK(pr->pr_root, 0);
2394 /* Tell modules this thread is still in its old jail after all. */
2395 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2396 prison_deref(pr, PD_DEREF | PD_DEUREF);
2402 * Returns a locked prison instance, or NULL on failure.
2405 prison_find(int prid)
2409 sx_assert(&allprison_lock, SX_LOCKED);
2410 TAILQ_FOREACH(pr, &allprison, pr_list) {
2411 if (pr->pr_id == prid) {
2412 mtx_lock(&pr->pr_mtx);
2415 mtx_unlock(&pr->pr_mtx);
2422 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2425 prison_find_child(struct prison *mypr, int prid)
2430 sx_assert(&allprison_lock, SX_LOCKED);
2431 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2432 if (pr->pr_id == prid) {
2433 mtx_lock(&pr->pr_mtx);
2436 mtx_unlock(&pr->pr_mtx);
2443 * Look for the name relative to mypr. Returns a locked prison or NULL.
2446 prison_find_name(struct prison *mypr, const char *name)
2448 struct prison *pr, *deadpr;
2452 sx_assert(&allprison_lock, SX_LOCKED);
2453 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2456 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2457 if (!strcmp(pr->pr_name + mylen, name)) {
2458 mtx_lock(&pr->pr_mtx);
2459 if (pr->pr_ref > 0) {
2460 if (pr->pr_uref > 0)
2464 mtx_unlock(&pr->pr_mtx);
2467 /* There was no valid prison - perhaps there was a dying one. */
2468 if (deadpr != NULL) {
2469 mtx_lock(&deadpr->pr_mtx);
2470 if (deadpr->pr_ref == 0) {
2471 mtx_unlock(&deadpr->pr_mtx);
2479 * See if a prison has the specific flag set.
2482 prison_flag(struct ucred *cred, unsigned flag)
2485 /* This is an atomic read, so no locking is necessary. */
2486 return (cred->cr_prison->pr_flags & flag);
2490 prison_allow(struct ucred *cred, unsigned flag)
2493 /* This is an atomic read, so no locking is necessary. */
2494 return (cred->cr_prison->pr_allow & flag);
2498 * Remove a prison reference. If that was the last reference, remove the
2499 * prison itself - but not in this context in case there are locks held.
2502 prison_free_locked(struct prison *pr)
2505 mtx_assert(&pr->pr_mtx, MA_OWNED);
2507 if (pr->pr_ref == 0) {
2508 mtx_unlock(&pr->pr_mtx);
2509 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2510 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2513 mtx_unlock(&pr->pr_mtx);
2517 prison_free(struct prison *pr)
2520 mtx_lock(&pr->pr_mtx);
2521 prison_free_locked(pr);
2525 prison_complete(void *context, int pending)
2528 prison_deref((struct prison *)context, 0);
2532 * Remove a prison reference (usually). This internal version assumes no
2533 * mutexes are held, except perhaps the prison itself. If there are no more
2534 * references, release and delist the prison. On completion, the prison lock
2535 * and the allprison lock are both unlocked.
2538 prison_deref(struct prison *pr, int flags)
2540 struct prison *ppr, *tpr;
2542 if (!(flags & PD_LOCKED))
2543 mtx_lock(&pr->pr_mtx);
2545 if (flags & PD_DEUREF) {
2547 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2549 if (flags & PD_DEREF)
2551 /* If the prison still has references, nothing else to do. */
2552 if (pr->pr_ref > 0) {
2553 mtx_unlock(&pr->pr_mtx);
2554 if (flags & PD_LIST_SLOCKED)
2555 sx_sunlock(&allprison_lock);
2556 else if (flags & PD_LIST_XLOCKED)
2557 sx_xunlock(&allprison_lock);
2561 mtx_unlock(&pr->pr_mtx);
2562 if (flags & PD_LIST_SLOCKED) {
2563 if (!sx_try_upgrade(&allprison_lock)) {
2564 sx_sunlock(&allprison_lock);
2565 sx_xlock(&allprison_lock);
2567 } else if (!(flags & PD_LIST_XLOCKED))
2568 sx_xlock(&allprison_lock);
2570 TAILQ_REMOVE(&allprison, pr, pr_list);
2571 LIST_REMOVE(pr, pr_sibling);
2572 ppr = pr->pr_parent;
2573 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2574 tpr->pr_childcount--;
2575 sx_xunlock(&allprison_lock);
2578 if (pr->pr_vnet != ppr->pr_vnet)
2579 vnet_destroy(pr->pr_vnet);
2581 if (pr->pr_root != NULL)
2583 mtx_destroy(&pr->pr_mtx);
2585 free(pr->pr_ip4, M_PRISON);
2588 free(pr->pr_ip6, M_PRISON);
2590 if (pr->pr_cpuset != NULL)
2591 cpuset_rel(pr->pr_cpuset);
2594 prison_racct_detach(pr);
2598 /* Removing a prison frees a reference on its parent. */
2600 mtx_lock(&pr->pr_mtx);
2601 flags = PD_DEREF | PD_DEUREF;
2606 prison_hold_locked(struct prison *pr)
2609 mtx_assert(&pr->pr_mtx, MA_OWNED);
2610 KASSERT(pr->pr_ref > 0,
2611 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2616 prison_hold(struct prison *pr)
2619 mtx_lock(&pr->pr_mtx);
2620 prison_hold_locked(pr);
2621 mtx_unlock(&pr->pr_mtx);
2625 prison_proc_hold(struct prison *pr)
2628 mtx_lock(&pr->pr_mtx);
2629 KASSERT(pr->pr_uref > 0,
2630 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2632 mtx_unlock(&pr->pr_mtx);
2636 prison_proc_free(struct prison *pr)
2639 mtx_lock(&pr->pr_mtx);
2640 KASSERT(pr->pr_uref > 0,
2641 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2642 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2648 * Restrict a prison's IP address list with its parent's, possibly replacing
2649 * it. Return true if the replacement buffer was used (or would have been).
2652 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2657 ppr = pr->pr_parent;
2658 if (!(pr->pr_flags & PR_IP4_USER)) {
2659 /* This has no user settings, so just copy the parent's list. */
2660 if (pr->pr_ip4s < ppr->pr_ip4s) {
2662 * There's no room for the parent's list. Use the
2663 * new list buffer, which is assumed to be big enough
2664 * (if it was passed). If there's no buffer, try to
2668 if (newip4 == NULL) {
2669 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2670 M_PRISON, M_NOWAIT);
2674 if (newip4 != NULL) {
2675 bcopy(ppr->pr_ip4, newip4,
2676 ppr->pr_ip4s * sizeof(*newip4));
2677 free(pr->pr_ip4, M_PRISON);
2678 pr->pr_ip4 = newip4;
2679 pr->pr_ip4s = ppr->pr_ip4s;
2683 pr->pr_ip4s = ppr->pr_ip4s;
2684 if (pr->pr_ip4s > 0)
2685 bcopy(ppr->pr_ip4, pr->pr_ip4,
2686 pr->pr_ip4s * sizeof(*newip4));
2687 else if (pr->pr_ip4 != NULL) {
2688 free(pr->pr_ip4, M_PRISON);
2691 } else if (pr->pr_ip4s > 0) {
2692 /* Remove addresses that aren't in the parent. */
2693 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2694 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2696 if (ij < ppr->pr_ip4s)
2699 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2700 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2703 for (ij = 1; ii < pr->pr_ip4s; ) {
2704 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2708 switch (ij >= ppr->pr_ip4s ? -1 :
2709 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2711 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2712 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2723 if (pr->pr_ip4s == 0) {
2724 pr->pr_flags |= PR_IP4_DISABLE;
2725 free(pr->pr_ip4, M_PRISON);
2733 * Pass back primary IPv4 address of this jail.
2735 * If not restricted return success but do not alter the address. Caller has
2736 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2738 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2739 * Address returned in NBO.
2742 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2746 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2747 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2749 pr = cred->cr_prison;
2750 if (!(pr->pr_flags & PR_IP4))
2752 mtx_lock(&pr->pr_mtx);
2753 if (!(pr->pr_flags & PR_IP4)) {
2754 mtx_unlock(&pr->pr_mtx);
2757 if (pr->pr_ip4 == NULL) {
2758 mtx_unlock(&pr->pr_mtx);
2759 return (EAFNOSUPPORT);
2762 ia->s_addr = pr->pr_ip4[0].s_addr;
2763 mtx_unlock(&pr->pr_mtx);
2768 * Return 1 if we should do proper source address selection or are not jailed.
2769 * We will return 0 if we should bypass source address selection in favour
2770 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2772 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2775 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2781 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2782 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2787 pr = cred->cr_prison;
2788 if (pr->pr_flags & PR_IP4_SADDRSEL)
2791 lia.s_addr = INADDR_ANY;
2792 error = prison_get_ip4(cred, &lia);
2795 if (lia.s_addr == INADDR_ANY)
2798 ia->s_addr = lia.s_addr;
2803 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2806 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2813 * No need to lock since the PR_IP4_USER flag can't be altered for
2816 while (pr1 != &prison0 &&
2818 !(pr1->pr_flags & PR_VNET) &&
2820 !(pr1->pr_flags & PR_IP4_USER))
2821 pr1 = pr1->pr_parent;
2822 while (pr2 != &prison0 &&
2824 !(pr2->pr_flags & PR_VNET) &&
2826 !(pr2->pr_flags & PR_IP4_USER))
2827 pr2 = pr2->pr_parent;
2828 return (pr1 == pr2);
2832 * Make sure our (source) address is set to something meaningful to this
2835 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2836 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2837 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2840 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2846 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2847 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2849 pr = cred->cr_prison;
2850 if (!(pr->pr_flags & PR_IP4))
2852 mtx_lock(&pr->pr_mtx);
2853 if (!(pr->pr_flags & PR_IP4)) {
2854 mtx_unlock(&pr->pr_mtx);
2857 if (pr->pr_ip4 == NULL) {
2858 mtx_unlock(&pr->pr_mtx);
2859 return (EAFNOSUPPORT);
2862 ia0.s_addr = ntohl(ia->s_addr);
2863 if (ia0.s_addr == INADDR_LOOPBACK) {
2864 ia->s_addr = pr->pr_ip4[0].s_addr;
2865 mtx_unlock(&pr->pr_mtx);
2869 if (ia0.s_addr == INADDR_ANY) {
2871 * In case there is only 1 IPv4 address, bind directly.
2873 if (pr->pr_ip4s == 1)
2874 ia->s_addr = pr->pr_ip4[0].s_addr;
2875 mtx_unlock(&pr->pr_mtx);
2879 error = _prison_check_ip4(pr, ia);
2880 mtx_unlock(&pr->pr_mtx);
2885 * Rewrite destination address in case we will connect to loopback address.
2887 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2888 * Address passed in in NBO and returned in NBO.
2891 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2895 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2896 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2898 pr = cred->cr_prison;
2899 if (!(pr->pr_flags & PR_IP4))
2901 mtx_lock(&pr->pr_mtx);
2902 if (!(pr->pr_flags & PR_IP4)) {
2903 mtx_unlock(&pr->pr_mtx);
2906 if (pr->pr_ip4 == NULL) {
2907 mtx_unlock(&pr->pr_mtx);
2908 return (EAFNOSUPPORT);
2911 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2912 ia->s_addr = pr->pr_ip4[0].s_addr;
2913 mtx_unlock(&pr->pr_mtx);
2918 * Return success because nothing had to be changed.
2920 mtx_unlock(&pr->pr_mtx);
2925 * Check if given address belongs to the jail referenced by cred/prison.
2927 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2928 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2929 * doesn't allow IPv4. Address passed in in NBO.
2932 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2937 * Check the primary IP.
2939 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2943 * All the other IPs are sorted so we can do a binary search.
2946 z = pr->pr_ip4s - 2;
2949 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2958 return (EADDRNOTAVAIL);
2962 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2967 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2968 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2970 pr = cred->cr_prison;
2971 if (!(pr->pr_flags & PR_IP4))
2973 mtx_lock(&pr->pr_mtx);
2974 if (!(pr->pr_flags & PR_IP4)) {
2975 mtx_unlock(&pr->pr_mtx);
2978 if (pr->pr_ip4 == NULL) {
2979 mtx_unlock(&pr->pr_mtx);
2980 return (EAFNOSUPPORT);
2983 error = _prison_check_ip4(pr, ia);
2984 mtx_unlock(&pr->pr_mtx);
2991 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2996 ppr = pr->pr_parent;
2997 if (!(pr->pr_flags & PR_IP6_USER)) {
2998 /* This has no user settings, so just copy the parent's list. */
2999 if (pr->pr_ip6s < ppr->pr_ip6s) {
3001 * There's no room for the parent's list. Use the
3002 * new list buffer, which is assumed to be big enough
3003 * (if it was passed). If there's no buffer, try to
3007 if (newip6 == NULL) {
3008 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3009 M_PRISON, M_NOWAIT);
3013 if (newip6 != NULL) {
3014 bcopy(ppr->pr_ip6, newip6,
3015 ppr->pr_ip6s * sizeof(*newip6));
3016 free(pr->pr_ip6, M_PRISON);
3017 pr->pr_ip6 = newip6;
3018 pr->pr_ip6s = ppr->pr_ip6s;
3022 pr->pr_ip6s = ppr->pr_ip6s;
3023 if (pr->pr_ip6s > 0)
3024 bcopy(ppr->pr_ip6, pr->pr_ip6,
3025 pr->pr_ip6s * sizeof(*newip6));
3026 else if (pr->pr_ip6 != NULL) {
3027 free(pr->pr_ip6, M_PRISON);
3030 } else if (pr->pr_ip6s > 0) {
3031 /* Remove addresses that aren't in the parent. */
3032 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3033 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3036 if (ij < ppr->pr_ip6s)
3039 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3040 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3043 for (ij = 1; ii < pr->pr_ip6s; ) {
3044 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3049 switch (ij >= ppr->pr_ip4s ? -1 :
3050 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3052 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3053 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3064 if (pr->pr_ip6s == 0) {
3065 pr->pr_flags |= PR_IP6_DISABLE;
3066 free(pr->pr_ip6, M_PRISON);
3074 * Pass back primary IPv6 address for this jail.
3076 * If not restricted return success but do not alter the address. Caller has
3077 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3079 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3082 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3086 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3087 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3089 pr = cred->cr_prison;
3090 if (!(pr->pr_flags & PR_IP6))
3092 mtx_lock(&pr->pr_mtx);
3093 if (!(pr->pr_flags & PR_IP6)) {
3094 mtx_unlock(&pr->pr_mtx);
3097 if (pr->pr_ip6 == NULL) {
3098 mtx_unlock(&pr->pr_mtx);
3099 return (EAFNOSUPPORT);
3102 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3103 mtx_unlock(&pr->pr_mtx);
3108 * Return 1 if we should do proper source address selection or are not jailed.
3109 * We will return 0 if we should bypass source address selection in favour
3110 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3112 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3115 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3118 struct in6_addr lia6;
3121 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3122 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3127 pr = cred->cr_prison;
3128 if (pr->pr_flags & PR_IP6_SADDRSEL)
3132 error = prison_get_ip6(cred, &lia6);
3135 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3138 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3143 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3146 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3152 while (pr1 != &prison0 &&
3154 !(pr1->pr_flags & PR_VNET) &&
3156 !(pr1->pr_flags & PR_IP6_USER))
3157 pr1 = pr1->pr_parent;
3158 while (pr2 != &prison0 &&
3160 !(pr2->pr_flags & PR_VNET) &&
3162 !(pr2->pr_flags & PR_IP6_USER))
3163 pr2 = pr2->pr_parent;
3164 return (pr1 == pr2);
3168 * Make sure our (source) address is set to something meaningful to this jail.
3170 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3171 * when needed while binding.
3173 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3174 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3175 * doesn't allow IPv6.
3178 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3183 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3184 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3186 pr = cred->cr_prison;
3187 if (!(pr->pr_flags & PR_IP6))
3189 mtx_lock(&pr->pr_mtx);
3190 if (!(pr->pr_flags & PR_IP6)) {
3191 mtx_unlock(&pr->pr_mtx);
3194 if (pr->pr_ip6 == NULL) {
3195 mtx_unlock(&pr->pr_mtx);
3196 return (EAFNOSUPPORT);
3199 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3200 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3201 mtx_unlock(&pr->pr_mtx);
3205 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3207 * In case there is only 1 IPv6 address, and v6only is true,
3208 * then bind directly.
3210 if (v6only != 0 && pr->pr_ip6s == 1)
3211 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3212 mtx_unlock(&pr->pr_mtx);
3216 error = _prison_check_ip6(pr, ia6);
3217 mtx_unlock(&pr->pr_mtx);
3222 * Rewrite destination address in case we will connect to loopback address.
3224 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3227 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3231 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3232 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3234 pr = cred->cr_prison;
3235 if (!(pr->pr_flags & PR_IP6))
3237 mtx_lock(&pr->pr_mtx);
3238 if (!(pr->pr_flags & PR_IP6)) {
3239 mtx_unlock(&pr->pr_mtx);
3242 if (pr->pr_ip6 == NULL) {
3243 mtx_unlock(&pr->pr_mtx);
3244 return (EAFNOSUPPORT);
3247 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3248 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3249 mtx_unlock(&pr->pr_mtx);
3254 * Return success because nothing had to be changed.
3256 mtx_unlock(&pr->pr_mtx);
3261 * Check if given address belongs to the jail referenced by cred/prison.
3263 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3264 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3265 * doesn't allow IPv6.
3268 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3273 * Check the primary IP.
3275 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3279 * All the other IPs are sorted so we can do a binary search.
3282 z = pr->pr_ip6s - 2;
3285 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3294 return (EADDRNOTAVAIL);
3298 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3303 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3304 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3306 pr = cred->cr_prison;
3307 if (!(pr->pr_flags & PR_IP6))
3309 mtx_lock(&pr->pr_mtx);
3310 if (!(pr->pr_flags & PR_IP6)) {
3311 mtx_unlock(&pr->pr_mtx);
3314 if (pr->pr_ip6 == NULL) {
3315 mtx_unlock(&pr->pr_mtx);
3316 return (EAFNOSUPPORT);
3319 error = _prison_check_ip6(pr, ia6);
3320 mtx_unlock(&pr->pr_mtx);
3326 * Check if a jail supports the given address family.
3328 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3332 prison_check_af(struct ucred *cred, int af)
3337 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3339 pr = cred->cr_prison;
3341 /* Prisons with their own network stack are not limited. */
3342 if (prison_owns_vnet(cred))
3351 if (pr->pr_flags & PR_IP4)
3353 mtx_lock(&pr->pr_mtx);
3354 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3355 error = EAFNOSUPPORT;
3356 mtx_unlock(&pr->pr_mtx);
3362 if (pr->pr_flags & PR_IP6)
3364 mtx_lock(&pr->pr_mtx);
3365 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3366 error = EAFNOSUPPORT;
3367 mtx_unlock(&pr->pr_mtx);
3375 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3376 error = EAFNOSUPPORT;
3382 * Check if given address belongs to the jail referenced by cred (wrapper to
3383 * prison_check_ip[46]).
3385 * Returns 0 if jail doesn't restrict the address family or if address belongs
3386 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3387 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3390 prison_if(struct ucred *cred, struct sockaddr *sa)
3393 struct sockaddr_in *sai;
3396 struct sockaddr_in6 *sai6;
3400 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3401 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3404 if (prison_owns_vnet(cred))
3409 switch (sa->sa_family)
3413 sai = (struct sockaddr_in *)sa;
3414 error = prison_check_ip4(cred, &sai->sin_addr);
3419 sai6 = (struct sockaddr_in6 *)sa;
3420 error = prison_check_ip6(cred, &sai6->sin6_addr);
3424 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3425 error = EAFNOSUPPORT;
3431 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3434 prison_check(struct ucred *cred1, struct ucred *cred2)
3437 return ((cred1->cr_prison == cred2->cr_prison ||
3438 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3442 * Return 1 if p2 is a child of p1, otherwise 0.
3445 prison_ischild(struct prison *pr1, struct prison *pr2)
3448 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3455 * Return 1 if the passed credential is in a jail, otherwise 0.
3458 jailed(struct ucred *cred)
3461 return (cred->cr_prison != &prison0);
3465 * Return 1 if the passed credential is in a jail and that jail does not
3466 * have its own virtual network stack, otherwise 0.
3469 jailed_without_vnet(struct ucred *cred)
3475 if (prison_owns_vnet(cred))
3483 * Return the correct hostname (domainname, et al) for the passed credential.
3486 getcredhostname(struct ucred *cred, char *buf, size_t size)
3491 * A NULL credential can be used to shortcut to the physical
3492 * system's hostname.
3494 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3495 mtx_lock(&pr->pr_mtx);
3496 strlcpy(buf, pr->pr_hostname, size);
3497 mtx_unlock(&pr->pr_mtx);
3501 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3504 mtx_lock(&cred->cr_prison->pr_mtx);
3505 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3506 mtx_unlock(&cred->cr_prison->pr_mtx);
3510 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3513 mtx_lock(&cred->cr_prison->pr_mtx);
3514 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3515 mtx_unlock(&cred->cr_prison->pr_mtx);
3519 getcredhostid(struct ucred *cred, unsigned long *hostid)
3522 mtx_lock(&cred->cr_prison->pr_mtx);
3523 *hostid = cred->cr_prison->pr_hostid;
3524 mtx_unlock(&cred->cr_prison->pr_mtx);
3529 * Determine whether the prison represented by cred owns
3530 * its vnet rather than having it inherited.
3532 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3535 prison_owns_vnet(struct ucred *cred)
3539 * vnets cannot be added/removed after jail creation,
3540 * so no need to lock here.
3542 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3547 * Determine whether the subject represented by cred can "see"
3548 * status of a mount point.
3549 * Returns: 0 for permitted, ENOENT otherwise.
3550 * XXX: This function should be called cr_canseemount() and should be
3551 * placed in kern_prot.c.
3554 prison_canseemount(struct ucred *cred, struct mount *mp)
3560 pr = cred->cr_prison;
3561 if (pr->pr_enforce_statfs == 0)
3563 if (pr->pr_root->v_mount == mp)
3565 if (pr->pr_enforce_statfs == 2)
3568 * If jail's chroot directory is set to "/" we should be able to see
3569 * all mount-points from inside a jail.
3570 * This is ugly check, but this is the only situation when jail's
3571 * directory ends with '/'.
3573 if (strcmp(pr->pr_path, "/") == 0)
3575 len = strlen(pr->pr_path);
3577 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3580 * Be sure that we don't have situation where jail's root directory
3581 * is "/some/path" and mount point is "/some/pathpath".
3583 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3589 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3591 char jpath[MAXPATHLEN];
3595 pr = cred->cr_prison;
3596 if (pr->pr_enforce_statfs == 0)
3598 if (prison_canseemount(cred, mp) != 0) {
3599 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3600 strlcpy(sp->f_mntonname, "[restricted]",
3601 sizeof(sp->f_mntonname));
3604 if (pr->pr_root->v_mount == mp) {
3606 * Clear current buffer data, so we are sure nothing from
3607 * the valid path left there.
3609 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3610 *sp->f_mntonname = '/';
3614 * If jail's chroot directory is set to "/" we should be able to see
3615 * all mount-points from inside a jail.
3617 if (strcmp(pr->pr_path, "/") == 0)
3619 len = strlen(pr->pr_path);
3620 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3622 * Clear current buffer data, so we are sure nothing from
3623 * the valid path left there.
3625 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3626 if (*jpath == '\0') {
3627 /* Should never happen. */
3628 *sp->f_mntonname = '/';
3630 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3635 * Check with permission for a specific privilege is granted within jail. We
3636 * have a specific list of accepted privileges; the rest are denied.
3639 prison_priv_check(struct ucred *cred, int priv)
3647 * Privileges specific to prisons with a virtual network stack.
3648 * There might be a duplicate entry here in case the privilege
3649 * is only granted conditionally in the legacy jail case.
3654 * NFS-specific privileges.
3656 case PRIV_NFS_DAEMON:
3657 case PRIV_NFS_LOCKD:
3660 * Network stack privileges.
3662 case PRIV_NET_BRIDGE:
3665 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3666 case PRIV_NET_ROUTE:
3668 case PRIV_NET_SETIFMTU:
3669 case PRIV_NET_SETIFFLAGS:
3670 case PRIV_NET_SETIFCAP:
3671 case PRIV_NET_SETIFDESCR:
3672 case PRIV_NET_SETIFNAME :
3673 case PRIV_NET_SETIFMETRIC:
3674 case PRIV_NET_SETIFPHYS:
3675 case PRIV_NET_SETIFMAC:
3676 case PRIV_NET_ADDMULTI:
3677 case PRIV_NET_DELMULTI:
3678 case PRIV_NET_HWIOCTL:
3679 case PRIV_NET_SETLLADDR:
3680 case PRIV_NET_ADDIFGROUP:
3681 case PRIV_NET_DELIFGROUP:
3682 case PRIV_NET_IFCREATE:
3683 case PRIV_NET_IFDESTROY:
3684 case PRIV_NET_ADDIFADDR:
3685 case PRIV_NET_DELIFADDR:
3688 case PRIV_NET_SETIFVNET:
3689 case PRIV_NET_SETIFFIB:
3692 * 802.11-related privileges.
3694 case PRIV_NET80211_GETKEY:
3696 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3701 * AppleTalk privileges.
3703 case PRIV_NETATALK_RESERVEDPORT:
3708 case PRIV_NETATM_CFG:
3709 case PRIV_NETATM_ADD:
3710 case PRIV_NETATM_DEL:
3711 case PRIV_NETATM_SET:
3714 * Bluetooth privileges.
3716 case PRIV_NETBLUETOOTH_RAW:
3720 * Netgraph and netgraph module privileges.
3722 case PRIV_NETGRAPH_CONTROL:
3724 case PRIV_NETGRAPH_TTY:
3728 * IPv4 and IPv6 privileges.
3730 case PRIV_NETINET_IPFW:
3731 case PRIV_NETINET_DIVERT:
3732 case PRIV_NETINET_PF:
3733 case PRIV_NETINET_DUMMYNET:
3734 case PRIV_NETINET_CARP:
3735 case PRIV_NETINET_MROUTE:
3736 case PRIV_NETINET_RAW:
3737 case PRIV_NETINET_ADDRCTRL6:
3738 case PRIV_NETINET_ND6:
3739 case PRIV_NETINET_SCOPE6:
3740 case PRIV_NETINET_ALIFETIME6:
3741 case PRIV_NETINET_IPSEC:
3742 case PRIV_NETINET_BINDANY:
3746 * IPX/SPX privileges.
3748 case PRIV_NETIPX_RESERVEDPORT:
3749 case PRIV_NETIPX_RAW:
3763 * No default: or deny here.
3764 * In case of no permit fall through to next switch().
3766 if (cred->cr_prison->pr_flags & PR_VNET)
3774 * Allow ktrace privileges for root in jail.
3780 * Allow jailed processes to configure audit identity and
3781 * submit audit records (login, etc). In the future we may
3782 * want to further refine the relationship between audit and
3785 case PRIV_AUDIT_GETAUDIT:
3786 case PRIV_AUDIT_SETAUDIT:
3787 case PRIV_AUDIT_SUBMIT:
3791 * Allow jailed processes to manipulate process UNIX
3792 * credentials in any way they see fit.
3794 case PRIV_CRED_SETUID:
3795 case PRIV_CRED_SETEUID:
3796 case PRIV_CRED_SETGID:
3797 case PRIV_CRED_SETEGID:
3798 case PRIV_CRED_SETGROUPS:
3799 case PRIV_CRED_SETREUID:
3800 case PRIV_CRED_SETREGID:
3801 case PRIV_CRED_SETRESUID:
3802 case PRIV_CRED_SETRESGID:
3805 * Jail implements visibility constraints already, so allow
3806 * jailed root to override uid/gid-based constraints.
3808 case PRIV_SEEOTHERGIDS:
3809 case PRIV_SEEOTHERUIDS:
3812 * Jail implements inter-process debugging limits already, so
3813 * allow jailed root various debugging privileges.
3815 case PRIV_DEBUG_DIFFCRED:
3816 case PRIV_DEBUG_SUGID:
3817 case PRIV_DEBUG_UNPRIV:
3820 * Allow jail to set various resource limits and login
3821 * properties, and for now, exceed process resource limits.
3823 case PRIV_PROC_LIMIT:
3824 case PRIV_PROC_SETLOGIN:
3825 case PRIV_PROC_SETRLIMIT:
3828 * System V and POSIX IPC privileges are granted in jail.
3831 case PRIV_IPC_WRITE:
3832 case PRIV_IPC_ADMIN:
3833 case PRIV_IPC_MSGSIZE:
3837 * Jail operations within a jail work on child jails.
3839 case PRIV_JAIL_ATTACH:
3841 case PRIV_JAIL_REMOVE:
3844 * Jail implements its own inter-process limits, so allow
3845 * root processes in jail to change scheduling on other
3846 * processes in the same jail. Likewise for signalling.
3848 case PRIV_SCHED_DIFFCRED:
3849 case PRIV_SCHED_CPUSET:
3850 case PRIV_SIGNAL_DIFFCRED:
3851 case PRIV_SIGNAL_SUGID:
3854 * Allow jailed processes to write to sysctls marked as jail
3857 case PRIV_SYSCTL_WRITEJAIL:
3860 * Allow root in jail to manage a variety of quota
3861 * properties. These should likely be conditional on a
3862 * configuration option.
3864 case PRIV_VFS_GETQUOTA:
3865 case PRIV_VFS_SETQUOTA:
3868 * Since Jail relies on chroot() to implement file system
3869 * protections, grant many VFS privileges to root in jail.
3870 * Be careful to exclude mount-related and NFS-related
3874 case PRIV_VFS_WRITE:
3875 case PRIV_VFS_ADMIN:
3877 case PRIV_VFS_LOOKUP:
3878 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3879 case PRIV_VFS_CHFLAGS_DEV:
3880 case PRIV_VFS_CHOWN:
3881 case PRIV_VFS_CHROOT:
3882 case PRIV_VFS_RETAINSUGID:
3883 case PRIV_VFS_FCHROOT:
3885 case PRIV_VFS_SETGID:
3887 case PRIV_VFS_STICKYFILE:
3890 * As in the non-jail case, non-root users are expected to be
3891 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3892 * exists in the jail and they have permission to access it).
3894 case PRIV_KMEM_READ:
3898 * Depending on the global setting, allow privilege of
3899 * setting system flags.
3901 case PRIV_VFS_SYSFLAGS:
3902 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3908 * Depending on the global setting, allow privilege of
3909 * mounting/unmounting file systems.
3911 case PRIV_VFS_MOUNT:
3912 case PRIV_VFS_UNMOUNT:
3913 case PRIV_VFS_MOUNT_NONUSER:
3914 case PRIV_VFS_MOUNT_OWNER:
3915 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3916 cred->cr_prison->pr_enforce_statfs < 2)
3922 * Allow jailed root to bind reserved ports and reuse in-use
3925 case PRIV_NETINET_RESERVEDPORT:
3926 case PRIV_NETINET_REUSEPORT:
3930 * Allow jailed root to set certian IPv4/6 (option) headers.
3932 case PRIV_NETINET_SETHDROPTS:
3936 * Conditionally allow creating raw sockets in jail.
3938 case PRIV_NETINET_RAW:
3939 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3945 * Since jail implements its own visibility limits on netstat
3946 * sysctls, allow getcred. This allows identd to work in
3949 case PRIV_NETINET_GETCRED:
3953 * Allow jailed root to set loginclass.
3955 case PRIV_PROC_SETLOGINCLASS:
3960 * In all remaining cases, deny the privilege request. This
3961 * includes almost all network privileges, many system
3962 * configuration privileges.
3969 * Return the part of pr2's name that is relative to pr1, or the whole name
3970 * if it does not directly follow.
3974 prison_name(struct prison *pr1, struct prison *pr2)
3978 /* Jails see themselves as "0" (if they see themselves at all). */
3981 name = pr2->pr_name;
3982 if (prison_ischild(pr1, pr2)) {
3984 * pr1 isn't locked (and allprison_lock may not be either)
3985 * so its length can't be counted on. But the number of dots
3986 * can be counted on - and counted.
3988 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3989 name = strchr(name, '.') + 1;
3995 * Return the part of pr2's path that is relative to pr1, or the whole path
3996 * if it does not directly follow.
3999 prison_path(struct prison *pr1, struct prison *pr2)
4001 char *path1, *path2;
4004 path1 = pr1->pr_path;
4005 path2 = pr2->pr_path;
4006 if (!strcmp(path1, "/"))
4008 len1 = strlen(path1);
4009 if (strncmp(path1, path2, len1))
4011 if (path2[len1] == '\0')
4013 if (path2[len1] == '/')
4014 return (path2 + len1);
4020 * Jail-related sysctls.
4022 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4026 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4029 struct prison *pr, *cpr;
4031 struct in_addr *ip4 = NULL;
4035 struct in6_addr *ip6 = NULL;
4040 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4041 pr = req->td->td_ucred->cr_prison;
4043 sx_slock(&allprison_lock);
4044 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4045 #if defined(INET) || defined(INET6)
4048 mtx_lock(&cpr->pr_mtx);
4050 if (cpr->pr_ip4s > 0) {
4051 if (ip4s < cpr->pr_ip4s) {
4052 ip4s = cpr->pr_ip4s;
4053 mtx_unlock(&cpr->pr_mtx);
4054 ip4 = realloc(ip4, ip4s *
4055 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4058 bcopy(cpr->pr_ip4, ip4,
4059 cpr->pr_ip4s * sizeof(struct in_addr));
4063 if (cpr->pr_ip6s > 0) {
4064 if (ip6s < cpr->pr_ip6s) {
4065 ip6s = cpr->pr_ip6s;
4066 mtx_unlock(&cpr->pr_mtx);
4067 ip6 = realloc(ip6, ip6s *
4068 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4071 bcopy(cpr->pr_ip6, ip6,
4072 cpr->pr_ip6s * sizeof(struct in6_addr));
4075 if (cpr->pr_ref == 0) {
4076 mtx_unlock(&cpr->pr_mtx);
4079 bzero(xp, sizeof(*xp));
4080 xp->pr_version = XPRISON_VERSION;
4081 xp->pr_id = cpr->pr_id;
4082 xp->pr_state = cpr->pr_uref > 0
4083 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4084 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4085 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4086 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4088 xp->pr_ip4s = cpr->pr_ip4s;
4091 xp->pr_ip6s = cpr->pr_ip6s;
4093 mtx_unlock(&cpr->pr_mtx);
4094 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4098 if (xp->pr_ip4s > 0) {
4099 error = SYSCTL_OUT(req, ip4,
4100 xp->pr_ip4s * sizeof(struct in_addr));
4106 if (xp->pr_ip6s > 0) {
4107 error = SYSCTL_OUT(req, ip6,
4108 xp->pr_ip6s * sizeof(struct in6_addr));
4114 sx_sunlock(&allprison_lock);
4125 SYSCTL_OID(_security_jail, OID_AUTO, list,
4126 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4127 sysctl_jail_list, "S", "List of active jails");
4130 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4134 injail = jailed(req->td->td_ucred);
4135 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4140 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4141 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4142 sysctl_jail_jailed, "I", "Process in jail?");
4145 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4147 int error, havevnet;
4149 struct ucred *cred = req->td->td_ucred;
4151 havevnet = jailed(cred) && prison_owns_vnet(cred);
4155 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4160 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4161 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4162 sysctl_jail_vnet, "I", "Jail owns VNET?");
4164 #if defined(INET) || defined(INET6)
4165 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4166 &jail_max_af_ips, 0,
4167 "Number of IP addresses a jail may have at most per address family");
4171 * Default parameters for jail(2) compatability. For historical reasons,
4172 * the sysctl names have varying similarity to the parameter names. Prisons
4173 * just see their own parameters, and can't change them.
4176 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4179 int allow, error, i;
4181 pr = req->td->td_ucred->cr_prison;
4182 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4184 /* Get the current flag value, and convert it to a boolean. */
4185 i = (allow & arg2) ? 1 : 0;
4188 error = sysctl_handle_int(oidp, &i, 0, req);
4189 if (error || !req->newptr)
4195 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4198 mtx_lock(&prison0.pr_mtx);
4199 jail_default_allow = (jail_default_allow & ~arg2) | i;
4200 mtx_unlock(&prison0.pr_mtx);
4204 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4205 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4206 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4207 "Processes in jail can set their hostnames");
4208 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4209 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4210 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4211 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4212 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4213 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4214 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4215 "Processes in jail can use System V IPC primitives");
4216 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4218 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4219 "Prison root can create raw sockets");
4220 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4221 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4222 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4223 "Processes in jail can alter system file flags");
4224 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4225 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4226 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4227 "Processes in jail can mount/unmount jail-friendly file systems");
4228 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4229 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4230 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4231 "Processes in jail can mount the devfs file system");
4232 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4233 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4234 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4235 "Processes in jail can mount the nullfs file system");
4236 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4237 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4238 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4239 "Processes in jail can mount the procfs file system");
4240 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4241 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4242 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4243 "Processes in jail can mount the tmpfs file system");
4244 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4245 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4246 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4247 "Processes in jail can mount the zfs file system");
4250 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4255 pr = req->td->td_ucred->cr_prison;
4256 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4257 error = sysctl_handle_int(oidp, &level, 0, req);
4258 if (error || !req->newptr)
4260 *(int *)arg1 = level;
4264 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4265 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4266 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4267 sysctl_jail_default_level, "I",
4268 "Processes in jail cannot see all mounted file systems");
4270 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4271 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4272 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4273 sysctl_jail_default_level, "I",
4274 "Ruleset for the devfs filesystem in jail");
4277 * Nodes to describe jail parameters. Maximum length of string parameters
4278 * is returned in the string itself, and the other parameters exist merely
4279 * to make themselves and their types known.
4281 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4285 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4292 switch (oidp->oid_kind & CTLTYPE)
4298 if (!(req->flags & SCTL_MASK32))
4300 return (SYSCTL_OUT(req, &l, sizeof(l)));
4304 return (SYSCTL_OUT(req, &i, sizeof(i)));
4305 case CTLTYPE_STRING:
4306 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4308 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4309 case CTLTYPE_STRUCT:
4311 return (SYSCTL_OUT(req, &s, sizeof(s)));
4316 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4317 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4318 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4319 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4320 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4321 "I", "Jail secure level");
4322 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4323 "I", "Jail cannot see all mounted file systems");
4324 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4325 "I", "Ruleset for in-jail devfs mounts");
4326 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4327 "B", "Jail persistence");
4329 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4330 "E,jailsys", "Virtual network stack");
4332 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4333 "B", "Jail is in the process of shutting down");
4335 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4336 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4337 "I", "Current number of child jails");
4338 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4339 "I", "Maximum number of child jails");
4341 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4342 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4344 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4345 "Jail NIS domainname");
4346 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4348 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4349 "LU", "Jail host ID");
4351 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4352 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4355 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4356 "Jail IPv4 address virtualization");
4357 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4358 "S,in_addr,a", "Jail IPv4 addresses");
4359 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4360 "B", "Do (not) use IPv4 source address selection rather than the "
4361 "primary jail IPv4 address.");
4364 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4365 "Jail IPv6 address virtualization");
4366 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4367 "S,in6_addr,a", "Jail IPv6 addresses");
4368 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4369 "B", "Do (not) use IPv6 source address selection rather than the "
4370 "primary jail IPv6 address.");
4373 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4374 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4375 "B", "Jail may set hostname");
4376 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4377 "B", "Jail may use SYSV IPC");
4378 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4379 "B", "Jail may create raw sockets");
4380 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4381 "B", "Jail may alter system file flags");
4382 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4383 "B", "Jail may set file quotas");
4384 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4385 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4387 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4388 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4389 "B", "Jail may mount/unmount jail-friendly file systems in general");
4390 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4391 "B", "Jail may mount the devfs file system");
4392 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4393 "B", "Jail may mount the nullfs file system");
4394 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4395 "B", "Jail may mount the procfs file system");
4396 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4397 "B", "Jail may mount the tmpfs file system");
4398 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4399 "B", "Jail may mount the zfs file system");
4402 prison_racct_foreach(void (*callback)(struct racct *racct,
4403 void *arg2, void *arg3), void *arg2, void *arg3)
4405 struct prison_racct *prr;
4407 sx_slock(&allprison_lock);
4408 LIST_FOREACH(prr, &allprison_racct, prr_next)
4409 (callback)(prr->prr_racct, arg2, arg3);
4410 sx_sunlock(&allprison_lock);
4413 static struct prison_racct *
4414 prison_racct_find_locked(const char *name)
4416 struct prison_racct *prr;
4418 sx_assert(&allprison_lock, SA_XLOCKED);
4420 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4423 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4424 if (strcmp(name, prr->prr_name) != 0)
4427 /* Found prison_racct with a matching name? */
4428 prison_racct_hold(prr);
4432 /* Add new prison_racct. */
4433 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4434 racct_create(&prr->prr_racct);
4436 strcpy(prr->prr_name, name);
4437 refcount_init(&prr->prr_refcount, 1);
4438 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4443 struct prison_racct *
4444 prison_racct_find(const char *name)
4446 struct prison_racct *prr;
4448 sx_xlock(&allprison_lock);
4449 prr = prison_racct_find_locked(name);
4450 sx_xunlock(&allprison_lock);
4455 prison_racct_hold(struct prison_racct *prr)
4458 refcount_acquire(&prr->prr_refcount);
4462 prison_racct_free_locked(struct prison_racct *prr)
4465 sx_assert(&allprison_lock, SA_XLOCKED);
4467 if (refcount_release(&prr->prr_refcount)) {
4468 racct_destroy(&prr->prr_racct);
4469 LIST_REMOVE(prr, prr_next);
4470 free(prr, M_PRISON_RACCT);
4475 prison_racct_free(struct prison_racct *prr)
4479 sx_assert(&allprison_lock, SA_UNLOCKED);
4481 old = prr->prr_refcount;
4482 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4485 sx_xlock(&allprison_lock);
4486 prison_racct_free_locked(prr);
4487 sx_xunlock(&allprison_lock);
4492 prison_racct_attach(struct prison *pr)
4494 struct prison_racct *prr;
4496 sx_assert(&allprison_lock, SA_XLOCKED);
4498 prr = prison_racct_find_locked(pr->pr_name);
4499 KASSERT(prr != NULL, ("cannot find prison_racct"));
4501 pr->pr_prison_racct = prr;
4505 * Handle jail renaming. From the racct point of view, renaming means
4506 * moving from one prison_racct to another.
4509 prison_racct_modify(struct prison *pr)
4513 struct prison_racct *oldprr;
4515 sx_slock(&allproc_lock);
4516 sx_xlock(&allprison_lock);
4518 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4519 sx_xunlock(&allprison_lock);
4520 sx_sunlock(&allproc_lock);
4524 oldprr = pr->pr_prison_racct;
4525 pr->pr_prison_racct = NULL;
4527 prison_racct_attach(pr);
4530 * Move resource utilisation records.
4532 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4535 * Force rctl to reattach rules to processes.
4537 FOREACH_PROC_IN_SYSTEM(p) {
4539 cred = crhold(p->p_ucred);
4541 racct_proc_ucred_changed(p, cred, cred);
4545 sx_sunlock(&allproc_lock);
4546 prison_racct_free_locked(oldprr);
4547 sx_xunlock(&allprison_lock);
4551 prison_racct_detach(struct prison *pr)
4554 sx_assert(&allprison_lock, SA_UNLOCKED);
4556 if (pr->pr_prison_racct == NULL)
4558 prison_racct_free(pr->pr_prison_racct);
4559 pr->pr_prison_racct = NULL;
4566 db_show_prison(struct prison *pr)
4569 #if defined(INET) || defined(INET6)
4574 char ip6buf[INET6_ADDRSTRLEN];
4577 db_printf("prison %p:\n", pr);
4578 db_printf(" jid = %d\n", pr->pr_id);
4579 db_printf(" name = %s\n", pr->pr_name);
4580 db_printf(" parent = %p\n", pr->pr_parent);
4581 db_printf(" ref = %d\n", pr->pr_ref);
4582 db_printf(" uref = %d\n", pr->pr_uref);
4583 db_printf(" path = %s\n", pr->pr_path);
4584 db_printf(" cpuset = %d\n", pr->pr_cpuset
4585 ? pr->pr_cpuset->cs_id : -1);
4587 db_printf(" vnet = %p\n", pr->pr_vnet);
4589 db_printf(" root = %p\n", pr->pr_root);
4590 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4591 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4592 db_printf(" children.max = %d\n", pr->pr_childmax);
4593 db_printf(" children.cur = %d\n", pr->pr_childcount);
4594 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4595 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4596 db_printf(" flags = 0x%x", pr->pr_flags);
4597 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4599 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4600 db_printf(" %s", pr_flag_names[fi]);
4601 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4603 jsf = pr->pr_flags &
4604 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4605 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4606 pr_flag_jailsys[fi].disable &&
4607 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4608 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4611 db_printf(" allow = 0x%x", pr->pr_allow);
4612 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4614 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4615 db_printf(" %s", pr_allow_names[fi]);
4617 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4618 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4619 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4620 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4621 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4623 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4624 for (ii = 0; ii < pr->pr_ip4s; ii++)
4625 db_printf(" %s %s\n",
4626 ii == 0 ? "ip4.addr =" : " ",
4627 inet_ntoa(pr->pr_ip4[ii]));
4630 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4631 for (ii = 0; ii < pr->pr_ip6s; ii++)
4632 db_printf(" %s %s\n",
4633 ii == 0 ? "ip6.addr =" : " ",
4634 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4638 DB_SHOW_COMMAND(prison, db_show_prison_command)
4644 * Show all prisons in the list, and prison0 which is not
4647 db_show_prison(&prison0);
4648 if (!db_pager_quit) {
4649 TAILQ_FOREACH(pr, &allprison, pr_list) {
4661 /* Look for a prison with the ID and with references. */
4662 TAILQ_FOREACH(pr, &allprison, pr_list)
4663 if (pr->pr_id == addr && pr->pr_ref > 0)
4666 /* Look again, without requiring a reference. */
4667 TAILQ_FOREACH(pr, &allprison, pr_list)
4668 if (pr->pr_id == addr)
4671 /* Assume address points to a valid prison. */
4672 pr = (struct prison *)addr;