2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_compat.h"
35 #include "opt_inet6.h"
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
67 #include <netinet/in.h>
72 #include <netinet6/in6_var.h>
76 #include <security/mac/mac_framework.h>
78 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
81 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
92 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
94 #define _PR_IP_SADDRSEL 0
98 /* prison0 describes what is "real" about the system. */
99 struct prison prison0 = {
105 .pr_securelevel = -1,
107 .pr_childmax = JAIL_MAX,
108 .pr_hostuuid = DEFAULT_HOSTUUID,
109 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
111 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
113 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
115 .pr_allow = PR_ALLOW_ALL,
117 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
119 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
120 struct sx allprison_lock;
121 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
122 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
123 LIST_HEAD(, prison_racct) allprison_racct;
126 static int do_jail_attach(struct thread *td, struct prison *pr);
127 static void prison_complete(void *context, int pending);
128 static void prison_deref(struct prison *pr, int flags);
129 static char *prison_path(struct prison *pr1, struct prison *pr2);
130 static void prison_remove_one(struct prison *pr);
132 static void prison_racct_attach(struct prison *pr);
133 static void prison_racct_modify(struct prison *pr);
134 static void prison_racct_detach(struct prison *pr);
137 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
138 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
141 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
142 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
145 /* Flags for prison_deref */
146 #define PD_DEREF 0x01
147 #define PD_DEUREF 0x02
148 #define PD_LOCKED 0x04
149 #define PD_LIST_SLOCKED 0x08
150 #define PD_LIST_XLOCKED 0x10
153 * Parameter names corresponding to PR_* flag values. Size values are for kvm
154 * as we cannot figure out the size of a sparse array, or an array without a
157 static char *pr_flag_names[] = {
160 [7] = "ip4.saddrsel",
163 [8] = "ip6.saddrsel",
166 const size_t pr_flag_names_size = sizeof(pr_flag_names);
168 static char *pr_flag_nonames[] = {
171 [7] = "ip4.nosaddrsel",
174 [8] = "ip6.nosaddrsel",
177 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
179 struct jailsys_flags {
183 } pr_flag_jailsys[] = {
184 { "host", 0, PR_HOST },
186 { "vnet", 0, PR_VNET },
189 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
192 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
195 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
197 static char *pr_allow_names[] = {
198 "allow.set_hostname",
206 "allow.mount.nullfs",
208 "allow.mount.procfs",
210 const size_t pr_allow_names_size = sizeof(pr_allow_names);
212 static char *pr_allow_nonames[] = {
213 "allow.noset_hostname",
215 "allow.noraw_sockets",
220 "allow.mount.nodevfs",
221 "allow.mount.nonullfs",
223 "allow.mount.noprocfs",
225 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
227 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
228 #define JAIL_DEFAULT_ENFORCE_STATFS 2
229 #define JAIL_DEFAULT_DEVFS_RSNUM 0
230 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
231 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
232 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
233 #if defined(INET) || defined(INET6)
234 static unsigned jail_max_af_ips = 255;
239 qcmp_v4(const void *ip1, const void *ip2)
244 * We need to compare in HBO here to get the list sorted as expected
245 * by the result of the code. Sorting NBO addresses gives you
246 * interesting results. If you do not understand, do not try.
248 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
249 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
252 * Do not simply return the difference of the two numbers, the int is
266 qcmp_v6(const void *ip1, const void *ip2)
268 const struct in6_addr *ia6a, *ia6b;
271 ia6a = (const struct in6_addr *)ip1;
272 ia6b = (const struct in6_addr *)ip2;
275 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
276 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
278 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
291 sys_jail(struct thread *td, struct jail_args *uap)
297 error = copyin(uap->jail, &version, sizeof(uint32_t));
306 /* FreeBSD single IPv4 jails. */
307 bzero(&j, sizeof(struct jail));
308 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
311 j.version = j0.version;
313 j.hostname = j0.hostname;
314 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
320 * Version 1 was used by multi-IPv4 jail implementations
321 * that never made it into the official kernel.
325 case 2: /* JAIL_API_VERSION */
326 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
327 error = copyin(uap->jail, &j, sizeof(struct jail));
333 /* Sci-Fi jails are not supported, sorry. */
336 return (kern_jail(td, &j));
340 kern_jail(struct thread *td, struct jail *j)
342 struct iovec optiov[2 * (4
343 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
352 char *u_path, *u_hostname, *u_name;
355 struct in_addr *u_ip4;
358 struct in6_addr *u_ip6;
361 int error, enforce_statfs, fi;
363 bzero(&optiov, sizeof(optiov));
364 opt.uio_iov = optiov;
368 opt.uio_segflg = UIO_SYSSPACE;
369 opt.uio_rw = UIO_READ;
372 /* Set permissions for top-level jails from sysctls. */
373 if (!jailed(td->td_ucred)) {
374 for (fi = 0; fi < sizeof(pr_allow_names) /
375 sizeof(pr_allow_names[0]); fi++) {
376 optiov[opt.uio_iovcnt].iov_base =
377 (jail_default_allow & (1 << fi))
378 ? pr_allow_names[fi] : pr_allow_nonames[fi];
379 optiov[opt.uio_iovcnt].iov_len =
380 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
383 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
384 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
386 enforce_statfs = jail_default_enforce_statfs;
387 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
388 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
392 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
394 ip4s = (j->version == 0) ? 1 : j->ip4s;
395 if (ip4s > jail_max_af_ips)
397 tmplen += ip4s * sizeof(struct in_addr);
403 if (j->ip6s > jail_max_af_ips)
405 tmplen += j->ip6s * sizeof(struct in6_addr);
410 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
411 u_hostname = u_path + MAXPATHLEN;
412 u_name = u_hostname + MAXHOSTNAMELEN;
414 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
418 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
420 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
423 optiov[opt.uio_iovcnt].iov_base = "path";
424 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
426 optiov[opt.uio_iovcnt].iov_base = u_path;
427 error = copyinstr(j->path, u_path, MAXPATHLEN,
428 &optiov[opt.uio_iovcnt].iov_len);
430 free(u_path, M_TEMP);
434 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
435 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
437 optiov[opt.uio_iovcnt].iov_base = u_hostname;
438 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
439 &optiov[opt.uio_iovcnt].iov_len);
441 free(u_path, M_TEMP);
445 if (j->jailname != NULL) {
446 optiov[opt.uio_iovcnt].iov_base = "name";
447 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
449 optiov[opt.uio_iovcnt].iov_base = u_name;
450 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
451 &optiov[opt.uio_iovcnt].iov_len);
453 free(u_path, M_TEMP);
459 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
460 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
462 optiov[opt.uio_iovcnt].iov_base = u_ip4;
463 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
465 u_ip4->s_addr = j->ip4s;
467 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
469 free(u_path, M_TEMP);
476 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
477 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
479 optiov[opt.uio_iovcnt].iov_base = u_ip6;
480 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
481 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
483 free(u_path, M_TEMP);
488 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
489 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
490 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
491 free(u_path, M_TEMP);
497 * struct jail_set_args {
498 * struct iovec *iovp;
499 * unsigned int iovcnt;
504 sys_jail_set(struct thread *td, struct jail_set_args *uap)
509 /* Check that we have an even number of iovecs. */
513 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
516 error = kern_jail_set(td, auio, uap->flags);
522 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
529 struct in6_addr *ip6;
532 struct vfsoptlist *opts;
533 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
535 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
537 #if defined(INET) || defined(INET6)
542 size_t namelen, onamelen;
543 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
544 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
545 int fi, jid, jsys, len, level;
546 int childmax, rsnum, slevel, vfslocked;
547 int fullpath_disabled;
548 #if defined(INET) || defined(INET6)
557 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
561 error = priv_check(td, PRIV_JAIL_SET);
562 if (!error && (flags & JAIL_ATTACH))
563 error = priv_check(td, PRIV_JAIL_ATTACH);
566 mypr = ppr = td->td_ucred->cr_prison;
567 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
569 if (flags & ~JAIL_SET_MASK)
573 * Check all the parameters before committing to anything. Not all
574 * errors can be caught early, but we may as well try. Also, this
575 * takes care of some expensive stuff (path lookup) before getting
576 * the allprison lock.
578 * XXX Jails are not filesystems, and jail parameters are not mount
579 * options. But it makes more sense to re-use the vfsopt code
580 * than duplicate it under a different name.
582 error = vfs_buildopts(optuio, &opts);
593 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
599 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
608 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
616 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
621 else if (enforce < 0 || enforce > 2) {
627 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
635 pr_flags = ch_flags = 0;
636 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
638 if (pr_flag_names[fi] == NULL)
640 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
641 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
643 ch_flags |= pr_flags;
644 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
646 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
653 case JAIL_SYS_DISABLE:
654 if (!pr_flag_jailsys[fi].disable) {
658 pr_flags |= pr_flag_jailsys[fi].disable;
661 pr_flags |= pr_flag_jailsys[fi].new;
663 case JAIL_SYS_INHERIT:
670 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
672 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
673 && !(pr_flags & PR_PERSIST)) {
675 vfs_opterror(opts, "new jail must persist or attach");
679 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
681 vfs_opterror(opts, "vnet cannot be changed after creation");
686 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
688 vfs_opterror(opts, "ip4 cannot be changed after creation");
693 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
695 vfs_opterror(opts, "ip6 cannot be changed after creation");
700 pr_allow = ch_allow = 0;
701 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
703 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
704 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
706 ch_allow |= pr_allow;
708 error = vfs_getopt(opts, "name", (void **)&name, &len);
714 if (len == 0 || name[len - 1] != '\0') {
718 if (len > MAXHOSTNAMELEN) {
719 error = ENAMETOOLONG;
724 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
732 if (len == 0 || host[len - 1] != '\0') {
736 if (len > MAXHOSTNAMELEN) {
737 error = ENAMETOOLONG;
742 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
750 if (len == 0 || domain[len - 1] != '\0') {
754 if (len > MAXHOSTNAMELEN) {
755 error = ENAMETOOLONG;
760 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
768 if (len == 0 || uuid[len - 1] != '\0') {
772 if (len > HOSTUUIDLEN) {
773 error = ENAMETOOLONG;
778 #ifdef COMPAT_FREEBSD32
779 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
782 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
786 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
798 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
800 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
803 else if (ip4s & (sizeof(*ip4) - 1)) {
807 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
809 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
811 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
812 ip4s /= sizeof(*ip4);
813 if (ip4s > jail_max_af_ips) {
815 vfs_opterror(opts, "too many IPv4 addresses");
818 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
819 bcopy(op, ip4, ip4s * sizeof(*ip4));
821 * IP addresses are all sorted but ip[0] to preserve
822 * the primary IP address as given from userland.
823 * This special IP is used for unbound outgoing
824 * connections as well for "loopback" traffic in case
825 * source address selection cannot find any more fitting
826 * address to connect from.
829 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
831 * Check for duplicate addresses and do some simple
832 * zero and broadcast checks. If users give other bogus
833 * addresses it is their problem.
835 * We do not have to care about byte order for these
836 * checks so we will do them in NBO.
838 for (ii = 0; ii < ip4s; ii++) {
839 if (ip4[ii].s_addr == INADDR_ANY ||
840 ip4[ii].s_addr == INADDR_BROADCAST) {
845 (ip4[0].s_addr == ip4[ii+1].s_addr ||
846 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
856 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
858 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
861 else if (ip6s & (sizeof(*ip6) - 1)) {
865 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
867 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
869 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
870 ip6s /= sizeof(*ip6);
871 if (ip6s > jail_max_af_ips) {
873 vfs_opterror(opts, "too many IPv6 addresses");
876 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
877 bcopy(op, ip6, ip6s * sizeof(*ip6));
879 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
880 for (ii = 0; ii < ip6s; ii++) {
881 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
886 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
887 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
897 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
898 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
901 "vnet jails cannot have IP address restrictions");
906 fullpath_disabled = 0;
908 error = vfs_getopt(opts, "path", (void **)&path, &len);
914 if (flags & JAIL_UPDATE) {
917 "path cannot be changed after creation");
920 if (len == 0 || path[len - 1] != '\0') {
924 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_SYSSPACE,
929 vfslocked = NDHASGIANT(&nd);
931 NDFREE(&nd, NDF_ONLY_PNBUF);
932 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
933 strlcpy(g_path, path, MAXPATHLEN);
934 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
937 else if (error == ENODEV) {
938 /* proceed if sysctl debug.disablefullpath == 1 */
939 fullpath_disabled = 1;
940 if (len < 2 || (len == 2 && path[0] == '/'))
943 /* exit on other errors */
944 VFS_UNLOCK_GIANT(vfslocked);
947 if (root->v_type != VDIR) {
950 VFS_UNLOCK_GIANT(vfslocked);
954 VFS_UNLOCK_GIANT(vfslocked);
955 if (fullpath_disabled) {
956 /* Leave room for a real-root full pathname. */
957 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
958 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
959 error = ENAMETOOLONG;
966 * Grab the allprison lock before letting modules check their
967 * parameters. Once we have it, do not let go so we'll have a
968 * consistent view of the OSD list.
970 sx_xlock(&allprison_lock);
971 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
973 goto done_unlock_list;
975 /* By now, all parameters should have been noted. */
976 TAILQ_FOREACH(opt, opts, link) {
977 if (!opt->seen && strcmp(opt->name, "errmsg")) {
979 vfs_opterror(opts, "unknown parameter: %s", opt->name);
980 goto done_unlock_list;
985 * See if we are creating a new record or updating an existing one.
986 * This abuses the file error codes ENOENT and EEXIST.
988 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
991 vfs_opterror(opts, "no valid operation (create or update)");
992 goto done_unlock_list;
996 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
997 namelc = strrchr(name, '.');
998 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1004 * See if a requested jid already exists. There is an
1005 * information leak here if the jid exists but is not within
1006 * the caller's jail hierarchy. Jail creators will get EEXIST
1007 * even though they cannot see the jail, and CREATE | UPDATE
1008 * will return ENOENT which is not normally a valid error.
1012 vfs_opterror(opts, "negative jid");
1013 goto done_unlock_list;
1015 pr = prison_find(jid);
1017 ppr = pr->pr_parent;
1018 /* Create: jid must not exist. */
1019 if (cuflags == JAIL_CREATE) {
1020 mtx_unlock(&pr->pr_mtx);
1022 vfs_opterror(opts, "jail %d already exists",
1024 goto done_unlock_list;
1026 if (!prison_ischild(mypr, pr)) {
1027 mtx_unlock(&pr->pr_mtx);
1029 } else if (pr->pr_uref == 0) {
1030 if (!(flags & JAIL_DYING)) {
1031 mtx_unlock(&pr->pr_mtx);
1033 vfs_opterror(opts, "jail %d is dying",
1035 goto done_unlock_list;
1036 } else if ((flags & JAIL_ATTACH) ||
1037 (pr_flags & PR_PERSIST)) {
1039 * A dying jail might be resurrected
1040 * (via attach or persist), but first
1041 * it must determine if another jail
1042 * has claimed its name. Accomplish
1043 * this by implicitly re-setting the
1047 name = prison_name(mypr, pr);
1052 /* Update: jid must exist. */
1053 if (cuflags == JAIL_UPDATE) {
1055 vfs_opterror(opts, "jail %d not found", jid);
1056 goto done_unlock_list;
1061 * If the caller provided a name, look for a jail by that name.
1062 * This has different semantics for creates and updates keyed by jid
1063 * (where the name must not already exist in a different jail),
1064 * and updates keyed by the name itself (where the name must exist
1065 * because that is the jail being updated).
1068 namelc = strrchr(name, '.');
1073 * This is a hierarchical name. Split it into the
1074 * parent and child names, and make sure the parent
1075 * exists or matches an already found jail.
1079 if (strncmp(name, ppr->pr_name, namelc - name)
1080 || ppr->pr_name[namelc - name] != '\0') {
1081 mtx_unlock(&pr->pr_mtx);
1084 "cannot change jail's parent");
1085 goto done_unlock_list;
1088 ppr = prison_find_name(mypr, name);
1092 "jail \"%s\" not found", name);
1093 goto done_unlock_list;
1095 mtx_unlock(&ppr->pr_mtx);
1099 if (name[0] != '\0') {
1101 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1104 FOREACH_PRISON_CHILD(ppr, tpr) {
1105 if (tpr != pr && tpr->pr_ref > 0 &&
1106 !strcmp(tpr->pr_name + namelen, name)) {
1108 cuflags != JAIL_CREATE) {
1109 mtx_lock(&tpr->pr_mtx);
1110 if (tpr->pr_ref > 0) {
1115 if (tpr->pr_uref > 0) {
1121 mtx_unlock(&tpr->pr_mtx);
1122 } else if (tpr->pr_uref > 0) {
1124 * Create, or update(jid):
1125 * name must not exist in an
1126 * active sibling jail.
1130 mtx_unlock(&pr->pr_mtx);
1132 "jail \"%s\" already exists",
1134 goto done_unlock_list;
1138 /* If no active jail is found, use a dying one. */
1139 if (deadpr != NULL && pr == NULL) {
1140 if (flags & JAIL_DYING) {
1141 mtx_lock(&deadpr->pr_mtx);
1142 if (deadpr->pr_ref == 0) {
1143 mtx_unlock(&deadpr->pr_mtx);
1147 } else if (cuflags == JAIL_UPDATE) {
1150 "jail \"%s\" is dying", name);
1151 goto done_unlock_list;
1154 /* Update: name must exist if no jid. */
1155 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1157 vfs_opterror(opts, "jail \"%s\" not found",
1159 goto done_unlock_list;
1163 /* Update: must provide a jid or name. */
1164 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1166 vfs_opterror(opts, "update specified no jail");
1167 goto done_unlock_list;
1170 /* If there's no prison to update, create a new one and link it in. */
1172 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1173 if (tpr->pr_childcount >= tpr->pr_childmax) {
1175 vfs_opterror(opts, "prison limit exceeded");
1176 goto done_unlock_list;
1179 mtx_lock(&ppr->pr_mtx);
1180 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1181 mtx_unlock(&ppr->pr_mtx);
1183 vfs_opterror(opts, "parent jail went away!");
1184 goto done_unlock_list;
1188 mtx_unlock(&ppr->pr_mtx);
1189 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1191 /* Find the next free jid. */
1194 if (jid == JAIL_MAX)
1196 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1197 if (tpr->pr_id < jid)
1199 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1200 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1203 if (jid == lastprid) {
1206 "no available jail IDs");
1208 prison_deref(ppr, PD_DEREF |
1209 PD_DEUREF | PD_LIST_XLOCKED);
1218 * The jail already has a jid (that did not yet exist),
1219 * so just find where to insert it.
1221 TAILQ_FOREACH(tpr, &allprison, pr_list)
1222 if (tpr->pr_id >= jid) {
1223 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1228 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1229 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1230 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1231 tpr->pr_childcount++;
1233 pr->pr_parent = ppr;
1236 /* Set some default values, and inherit some from the parent. */
1241 root = mypr->pr_root;
1244 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1245 pr->pr_flags |= PR_HOST;
1246 #if defined(INET) || defined(INET6)
1248 if (!(pr_flags & PR_VNET))
1252 if (!(ch_flags & PR_IP4_USER))
1254 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1255 else if (!(pr_flags & PR_IP4_USER)) {
1256 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1257 if (ppr->pr_ip4 != NULL) {
1258 pr->pr_ip4s = ppr->pr_ip4s;
1259 pr->pr_ip4 = malloc(pr->pr_ip4s *
1260 sizeof(struct in_addr), M_PRISON,
1262 bcopy(ppr->pr_ip4, pr->pr_ip4,
1263 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1268 if (!(ch_flags & PR_IP6_USER))
1270 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1271 else if (!(pr_flags & PR_IP6_USER)) {
1272 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1273 if (ppr->pr_ip6 != NULL) {
1274 pr->pr_ip6s = ppr->pr_ip6s;
1275 pr->pr_ip6 = malloc(pr->pr_ip6s *
1276 sizeof(struct in6_addr), M_PRISON,
1278 bcopy(ppr->pr_ip6, pr->pr_ip6,
1279 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1285 /* Source address selection is always on by default. */
1286 pr->pr_flags |= _PR_IP_SADDRSEL;
1288 pr->pr_securelevel = ppr->pr_securelevel;
1289 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1290 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1291 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1293 LIST_INIT(&pr->pr_children);
1294 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1297 /* Allocate a new vnet if specified. */
1298 pr->pr_vnet = (pr_flags & PR_VNET)
1299 ? vnet_alloc() : ppr->pr_vnet;
1302 * Allocate a dedicated cpuset for each jail.
1303 * Unlike other initial settings, this may return an erorr.
1305 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1307 prison_deref(pr, PD_LIST_XLOCKED);
1311 mtx_lock(&pr->pr_mtx);
1313 * New prisons do not yet have a reference, because we do not
1314 * want other to see the incomplete prison once the
1315 * allprison_lock is downgraded.
1320 * Grab a reference for existing prisons, to ensure they
1321 * continue to exist for the duration of the call.
1324 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1325 if ((pr->pr_flags & PR_VNET) &&
1326 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1329 "vnet jails cannot have IP address restrictions");
1330 goto done_deref_locked;
1334 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1337 "ip4 cannot be changed after creation");
1338 goto done_deref_locked;
1342 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1345 "ip6 cannot be changed after creation");
1346 goto done_deref_locked;
1351 /* Do final error checking before setting anything. */
1353 if (slevel < ppr->pr_securelevel) {
1355 goto done_deref_locked;
1359 if (childmax >= ppr->pr_childmax) {
1361 goto done_deref_locked;
1365 if (enforce < ppr->pr_enforce_statfs) {
1367 goto done_deref_locked;
1372 * devfs_rsnum is a uint16_t
1374 if (rsnum < 0 || rsnum > 65535) {
1376 goto done_deref_locked;
1379 * Nested jails always inherit parent's devfs ruleset
1381 if (jailed(td->td_ucred)) {
1382 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1384 goto done_deref_locked;
1386 rsnum = ppr->pr_devfs_rsnum;
1391 if (ppr->pr_flags & PR_IP4) {
1393 * Make sure the new set of IP addresses is a
1394 * subset of the parent's list. Don't worry
1395 * about the parent being unlocked, as any
1396 * setting is done with allprison_lock held.
1398 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1399 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1401 if (ij == ppr->pr_ip4s) {
1403 goto done_deref_locked;
1406 for (ii = ij = 1; ii < ip4s; ii++) {
1407 if (ip4[ii].s_addr ==
1408 ppr->pr_ip4[0].s_addr)
1410 for (; ij < ppr->pr_ip4s; ij++)
1411 if (ip4[ii].s_addr ==
1412 ppr->pr_ip4[ij].s_addr)
1414 if (ij == ppr->pr_ip4s)
1417 if (ij == ppr->pr_ip4s) {
1419 goto done_deref_locked;
1424 * Check for conflicting IP addresses. We permit them
1425 * if there is no more than one IP on each jail. If
1426 * there is a duplicate on a jail with more than one
1427 * IP stop checking and return error.
1431 for (; tppr != &prison0; tppr = tppr->pr_parent)
1432 if (tppr->pr_flags & PR_VNET)
1435 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1438 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1440 tpr->pr_uref == 0) {
1444 if (!(tpr->pr_flags & PR_IP4_USER))
1447 if (tpr->pr_ip4 == NULL ||
1448 (ip4s == 1 && tpr->pr_ip4s == 1))
1450 for (ii = 0; ii < ip4s; ii++) {
1451 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1454 "IPv4 addresses clash");
1455 goto done_deref_locked;
1463 if (ppr->pr_flags & PR_IP6) {
1465 * Make sure the new set of IP addresses is a
1466 * subset of the parent's list.
1468 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1469 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1472 if (ij == ppr->pr_ip6s) {
1474 goto done_deref_locked;
1477 for (ii = ij = 1; ii < ip6s; ii++) {
1478 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1481 for (; ij < ppr->pr_ip6s; ij++)
1482 if (IN6_ARE_ADDR_EQUAL(
1483 &ip6[ii], &ppr->pr_ip6[ij]))
1485 if (ij == ppr->pr_ip6s)
1488 if (ij == ppr->pr_ip6s) {
1490 goto done_deref_locked;
1494 /* Check for conflicting IP addresses. */
1497 for (; tppr != &prison0; tppr = tppr->pr_parent)
1498 if (tppr->pr_flags & PR_VNET)
1501 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1504 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1506 tpr->pr_uref == 0) {
1510 if (!(tpr->pr_flags & PR_IP6_USER))
1513 if (tpr->pr_ip6 == NULL ||
1514 (ip6s == 1 && tpr->pr_ip6s == 1))
1516 for (ii = 0; ii < ip6s; ii++) {
1517 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1520 "IPv6 addresses clash");
1521 goto done_deref_locked;
1527 onamelen = namelen = 0;
1529 /* Give a default name of the jid. */
1530 if (name[0] == '\0')
1531 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1532 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid &&
1536 "name cannot be numeric (unless it is the jid)");
1537 goto done_deref_locked;
1540 * Make sure the name isn't too long for the prison or its
1543 onamelen = strlen(pr->pr_name);
1544 namelen = strlen(name);
1545 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1546 error = ENAMETOOLONG;
1547 goto done_deref_locked;
1549 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1550 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1551 sizeof(pr->pr_name)) {
1552 error = ENAMETOOLONG;
1553 goto done_deref_locked;
1557 if (pr_allow & ~ppr->pr_allow) {
1559 goto done_deref_locked;
1562 /* Set the parameters of the prison. */
1565 if (pr_flags & PR_IP4_USER) {
1566 pr->pr_flags |= PR_IP4;
1567 free(pr->pr_ip4, M_PRISON);
1571 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1573 if (tpr->pr_flags & PR_VNET) {
1578 if (prison_restrict_ip4(tpr, NULL)) {
1587 if (pr_flags & PR_IP6_USER) {
1588 pr->pr_flags |= PR_IP6;
1589 free(pr->pr_ip6, M_PRISON);
1593 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1595 if (tpr->pr_flags & PR_VNET) {
1600 if (prison_restrict_ip6(tpr, NULL)) {
1608 pr->pr_securelevel = slevel;
1609 /* Set all child jails to be at least this level. */
1610 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1611 if (tpr->pr_securelevel < slevel)
1612 tpr->pr_securelevel = slevel;
1615 pr->pr_childmax = childmax;
1616 /* Set all child jails to under this limit. */
1617 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1618 if (tpr->pr_childmax > childmax - level)
1619 tpr->pr_childmax = childmax > level
1620 ? childmax - level : 0;
1623 pr->pr_enforce_statfs = enforce;
1624 /* Pass this restriction on to the children. */
1625 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1626 if (tpr->pr_enforce_statfs < enforce)
1627 tpr->pr_enforce_statfs = enforce;
1630 pr->pr_devfs_rsnum = rsnum;
1631 /* Pass this restriction on to the children. */
1632 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1633 tpr->pr_devfs_rsnum = rsnum;
1636 if (ppr == &prison0)
1637 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1639 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1640 ppr->pr_name, name);
1641 /* Change this component of child names. */
1642 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1643 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1644 strlen(tpr->pr_name + onamelen) + 1);
1645 bcopy(pr->pr_name, tpr->pr_name, namelen);
1649 /* Try to keep a real-rooted full pathname. */
1650 if (fullpath_disabled && path[0] == '/' &&
1651 strcmp(mypr->pr_path, "/"))
1652 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1653 mypr->pr_path, path);
1655 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1658 if (PR_HOST & ch_flags & ~pr_flags) {
1659 if (pr->pr_flags & PR_HOST) {
1661 * Copy the parent's host info. As with pr_ip4 above,
1662 * the lack of a lock on the parent is not a problem;
1663 * it is always set with allprison_lock at least
1664 * shared, and is held exclusively here.
1666 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1667 sizeof(pr->pr_hostname));
1668 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1669 sizeof(pr->pr_domainname));
1670 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1671 sizeof(pr->pr_hostuuid));
1672 pr->pr_hostid = pr->pr_parent->pr_hostid;
1674 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1675 /* Set this prison, and any descendants without PR_HOST. */
1677 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1679 strlcpy(pr->pr_domainname, domain,
1680 sizeof(pr->pr_domainname));
1682 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1684 pr->pr_hostid = hid;
1685 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1686 if (tpr->pr_flags & PR_HOST)
1690 strlcpy(tpr->pr_hostname,
1692 sizeof(tpr->pr_hostname));
1694 strlcpy(tpr->pr_domainname,
1696 sizeof(tpr->pr_domainname));
1698 strlcpy(tpr->pr_hostuuid,
1700 sizeof(tpr->pr_hostuuid));
1702 tpr->pr_hostid = hid;
1706 if ((tallow = ch_allow & ~pr_allow)) {
1707 /* Clear allow bits in all children. */
1708 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1709 tpr->pr_allow &= ~tallow;
1711 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1713 * Persistent prisons get an extra reference, and prisons losing their
1714 * persist flag lose that reference. Only do this for existing prisons
1715 * for now, so new ones will remain unseen until after the module
1716 * handlers have completed.
1718 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1719 if (pr_flags & PR_PERSIST) {
1727 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1728 mtx_unlock(&pr->pr_mtx);
1732 prison_racct_attach(pr);
1735 /* Locks may have prevented a complete restriction of child IP
1736 * addresses. If so, allocate some more memory and try again.
1741 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1742 mtx_lock(&pr->pr_mtx);
1744 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1746 if (tpr->pr_flags & PR_VNET) {
1751 if (prison_restrict_ip4(tpr, ip4)) {
1758 mtx_unlock(&pr->pr_mtx);
1764 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1765 mtx_lock(&pr->pr_mtx);
1767 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1769 if (tpr->pr_flags & PR_VNET) {
1774 if (prison_restrict_ip6(tpr, ip6)) {
1781 mtx_unlock(&pr->pr_mtx);
1785 /* Let the modules do their work. */
1786 sx_downgrade(&allprison_lock);
1788 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1790 prison_deref(pr, PD_LIST_SLOCKED);
1794 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1796 prison_deref(pr, created
1798 : PD_DEREF | PD_LIST_SLOCKED);
1802 /* Attach this process to the prison if requested. */
1803 if (flags & JAIL_ATTACH) {
1804 mtx_lock(&pr->pr_mtx);
1805 error = do_jail_attach(td, pr);
1807 vfs_opterror(opts, "attach failed");
1809 prison_deref(pr, PD_DEREF);
1816 sx_sunlock(&allprison_lock);
1817 prison_racct_modify(pr);
1818 sx_slock(&allprison_lock);
1822 td->td_retval[0] = pr->pr_id;
1825 * Now that it is all there, drop the temporary reference from existing
1826 * prisons. Or add a reference to newly created persistent prisons
1827 * (which was not done earlier so that the prison would not be publicly
1831 prison_deref(pr, (flags & JAIL_ATTACH)
1833 : PD_DEREF | PD_LIST_SLOCKED);
1835 if (pr_flags & PR_PERSIST) {
1836 mtx_lock(&pr->pr_mtx);
1839 mtx_unlock(&pr->pr_mtx);
1841 if (!(flags & JAIL_ATTACH))
1842 sx_sunlock(&allprison_lock);
1848 prison_deref(pr, created
1849 ? PD_LOCKED | PD_LIST_XLOCKED
1850 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1853 sx_xunlock(&allprison_lock);
1856 vfslocked = VFS_LOCK_GIANT(root->v_mount);
1858 VFS_UNLOCK_GIANT(vfslocked);
1862 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1863 if (errmsg_len > 0) {
1864 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1865 if (errmsg_pos > 0) {
1866 if (optuio->uio_segflg == UIO_SYSSPACE)
1868 optuio->uio_iov[errmsg_pos].iov_base,
1872 optuio->uio_iov[errmsg_pos].iov_base,
1879 free(ip4, M_PRISON);
1882 free(ip6, M_PRISON);
1885 free(g_path, M_TEMP);
1892 * struct jail_get_args {
1893 * struct iovec *iovp;
1894 * unsigned int iovcnt;
1899 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1904 /* Check that we have an even number of iovecs. */
1905 if (uap->iovcnt & 1)
1908 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1911 error = kern_jail_get(td, auio, uap->flags);
1913 error = copyout(auio->uio_iov, uap->iovp,
1914 uap->iovcnt * sizeof (struct iovec));
1920 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1922 struct prison *pr, *mypr;
1924 struct vfsoptlist *opts;
1925 char *errmsg, *name;
1926 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1928 if (flags & ~JAIL_GET_MASK)
1931 /* Get the parameter list. */
1932 error = vfs_buildopts(optuio, &opts);
1935 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1936 mypr = td->td_ucred->cr_prison;
1939 * Find the prison specified by one of: lastjid, jid, name.
1941 sx_slock(&allprison_lock);
1942 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1944 TAILQ_FOREACH(pr, &allprison, pr_list) {
1945 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1946 mtx_lock(&pr->pr_mtx);
1947 if (pr->pr_ref > 0 &&
1948 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1950 mtx_unlock(&pr->pr_mtx);
1956 vfs_opterror(opts, "no jail after %d", jid);
1957 goto done_unlock_list;
1958 } else if (error != ENOENT)
1959 goto done_unlock_list;
1961 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1964 pr = prison_find_child(mypr, jid);
1966 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1967 mtx_unlock(&pr->pr_mtx);
1969 vfs_opterror(opts, "jail %d is dying",
1971 goto done_unlock_list;
1976 vfs_opterror(opts, "jail %d not found", jid);
1977 goto done_unlock_list;
1979 } else if (error != ENOENT)
1980 goto done_unlock_list;
1982 error = vfs_getopt(opts, "name", (void **)&name, &len);
1984 if (len == 0 || name[len - 1] != '\0') {
1986 goto done_unlock_list;
1988 pr = prison_find_name(mypr, name);
1990 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1991 mtx_unlock(&pr->pr_mtx);
1993 vfs_opterror(opts, "jail \"%s\" is dying",
1995 goto done_unlock_list;
2000 vfs_opterror(opts, "jail \"%s\" not found", name);
2001 goto done_unlock_list;
2002 } else if (error != ENOENT)
2003 goto done_unlock_list;
2005 vfs_opterror(opts, "no jail specified");
2007 goto done_unlock_list;
2010 /* Get the parameters of the prison. */
2013 td->td_retval[0] = pr->pr_id;
2014 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2015 if (error != 0 && error != ENOENT)
2017 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2018 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2019 if (error != 0 && error != ENOENT)
2021 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2022 if (error != 0 && error != ENOENT)
2024 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2025 sizeof(pr->pr_cpuset->cs_id));
2026 if (error != 0 && error != ENOENT)
2028 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2029 if (error != 0 && error != ENOENT)
2032 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2033 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2034 if (error != 0 && error != ENOENT)
2038 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2039 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2040 if (error != 0 && error != ENOENT)
2043 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2044 sizeof(pr->pr_securelevel));
2045 if (error != 0 && error != ENOENT)
2047 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2048 sizeof(pr->pr_childcount));
2049 if (error != 0 && error != ENOENT)
2051 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2052 sizeof(pr->pr_childmax));
2053 if (error != 0 && error != ENOENT)
2055 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2056 if (error != 0 && error != ENOENT)
2058 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2059 if (error != 0 && error != ENOENT)
2061 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2062 if (error != 0 && error != ENOENT)
2064 #ifdef COMPAT_FREEBSD32
2065 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2066 uint32_t hid32 = pr->pr_hostid;
2068 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2071 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2072 sizeof(pr->pr_hostid));
2073 if (error != 0 && error != ENOENT)
2075 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2076 sizeof(pr->pr_enforce_statfs));
2077 if (error != 0 && error != ENOENT)
2079 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2080 sizeof(pr->pr_devfs_rsnum));
2081 if (error != 0 && error != ENOENT)
2083 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2085 if (pr_flag_names[fi] == NULL)
2087 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2088 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2089 if (error != 0 && error != ENOENT)
2092 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2093 if (error != 0 && error != ENOENT)
2096 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2099 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2100 i = pr_flag_jailsys[fi].disable &&
2101 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2102 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2105 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2106 if (error != 0 && error != ENOENT)
2109 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2111 if (pr_allow_names[fi] == NULL)
2113 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2114 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2115 if (error != 0 && error != ENOENT)
2118 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2119 if (error != 0 && error != ENOENT)
2122 i = (pr->pr_uref == 0);
2123 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2124 if (error != 0 && error != ENOENT)
2127 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2128 if (error != 0 && error != ENOENT)
2131 /* Get the module parameters. */
2132 mtx_unlock(&pr->pr_mtx);
2134 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2137 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2139 /* By now, all parameters should have been noted. */
2140 TAILQ_FOREACH(opt, opts, link) {
2141 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2143 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2148 /* Write the fetched parameters back to userspace. */
2150 TAILQ_FOREACH(opt, opts, link) {
2151 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2152 pos = 2 * opt->pos + 1;
2153 optuio->uio_iov[pos].iov_len = opt->len;
2154 if (opt->value != NULL) {
2155 if (optuio->uio_segflg == UIO_SYSSPACE) {
2157 optuio->uio_iov[pos].iov_base,
2160 error = copyout(opt->value,
2161 optuio->uio_iov[pos].iov_base,
2172 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2176 sx_sunlock(&allprison_lock);
2178 if (error && errmsg_pos >= 0) {
2179 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2180 errmsg_pos = 2 * errmsg_pos + 1;
2181 if (errmsg_len > 0) {
2182 if (optuio->uio_segflg == UIO_SYSSPACE)
2184 optuio->uio_iov[errmsg_pos].iov_base,
2188 optuio->uio_iov[errmsg_pos].iov_base,
2198 * struct jail_remove_args {
2203 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2205 struct prison *pr, *cpr, *lpr, *tpr;
2208 error = priv_check(td, PRIV_JAIL_REMOVE);
2212 sx_xlock(&allprison_lock);
2213 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2215 sx_xunlock(&allprison_lock);
2219 /* Remove all descendants of this prison, then remove this prison. */
2221 pr->pr_flags |= PR_REMOVE;
2222 if (!LIST_EMPTY(&pr->pr_children)) {
2223 mtx_unlock(&pr->pr_mtx);
2225 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2226 mtx_lock(&cpr->pr_mtx);
2227 if (cpr->pr_ref > 0) {
2230 cpr->pr_flags |= PR_REMOVE;
2232 /* Already removed - do not do it again. */
2235 mtx_unlock(&cpr->pr_mtx);
2237 mtx_lock(&lpr->pr_mtx);
2238 prison_remove_one(lpr);
2239 sx_xlock(&allprison_lock);
2244 mtx_lock(&lpr->pr_mtx);
2245 prison_remove_one(lpr);
2246 sx_xlock(&allprison_lock);
2248 mtx_lock(&pr->pr_mtx);
2250 prison_remove_one(pr);
2255 prison_remove_one(struct prison *pr)
2260 /* If the prison was persistent, it is not anymore. */
2262 if (pr->pr_flags & PR_PERSIST) {
2265 pr->pr_flags &= ~PR_PERSIST;
2269 * jail_remove added a reference. If that's the only one, remove
2272 KASSERT(pr->pr_ref > 0,
2273 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2274 if (pr->pr_ref == 1) {
2276 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2280 mtx_unlock(&pr->pr_mtx);
2281 sx_xunlock(&allprison_lock);
2283 * Kill all processes unfortunate enough to be attached to this prison.
2285 sx_slock(&allproc_lock);
2286 LIST_FOREACH(p, &allproc, p_list) {
2288 if (p->p_state != PRS_NEW && p->p_ucred &&
2289 p->p_ucred->cr_prison == pr)
2290 kern_psignal(p, SIGKILL);
2293 sx_sunlock(&allproc_lock);
2294 /* Remove the temporary reference added by jail_remove. */
2295 prison_deref(pr, deuref | PD_DEREF);
2300 * struct jail_attach_args {
2305 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2310 error = priv_check(td, PRIV_JAIL_ATTACH);
2314 sx_slock(&allprison_lock);
2315 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2317 sx_sunlock(&allprison_lock);
2322 * Do not allow a process to attach to a prison that is not
2323 * considered to be "alive".
2325 if (pr->pr_uref == 0) {
2326 mtx_unlock(&pr->pr_mtx);
2327 sx_sunlock(&allprison_lock);
2331 return (do_jail_attach(td, pr));
2335 do_jail_attach(struct thread *td, struct prison *pr)
2339 struct ucred *newcred, *oldcred;
2340 int vfslocked, error;
2343 * XXX: Note that there is a slight race here if two threads
2344 * in the same privileged process attempt to attach to two
2345 * different jails at the same time. It is important for
2346 * user processes not to do this, or they might end up with
2347 * a process root from one prison, but attached to the jail
2352 mtx_unlock(&pr->pr_mtx);
2354 /* Let modules do whatever they need to prepare for attaching. */
2355 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2357 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2360 sx_sunlock(&allprison_lock);
2363 * Reparent the newly attached process to this jail.
2365 ppr = td->td_ucred->cr_prison;
2367 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2371 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2372 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2373 if ((error = change_dir(pr->pr_root, td)) != 0)
2376 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2379 VOP_UNLOCK(pr->pr_root, 0);
2380 if ((error = change_root(pr->pr_root, td)))
2381 goto e_unlock_giant;
2382 VFS_UNLOCK_GIANT(vfslocked);
2386 oldcred = p->p_ucred;
2388 crcopy(newcred, oldcred);
2389 newcred->cr_prison = pr;
2390 p->p_ucred = newcred;
2393 racct_proc_ucred_changed(p, oldcred, newcred);
2396 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2399 VOP_UNLOCK(pr->pr_root, 0);
2401 VFS_UNLOCK_GIANT(vfslocked);
2403 /* Tell modules this thread is still in its old jail after all. */
2404 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2405 prison_deref(pr, PD_DEREF | PD_DEUREF);
2411 * Returns a locked prison instance, or NULL on failure.
2414 prison_find(int prid)
2418 sx_assert(&allprison_lock, SX_LOCKED);
2419 TAILQ_FOREACH(pr, &allprison, pr_list) {
2420 if (pr->pr_id == prid) {
2421 mtx_lock(&pr->pr_mtx);
2424 mtx_unlock(&pr->pr_mtx);
2431 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2434 prison_find_child(struct prison *mypr, int prid)
2439 sx_assert(&allprison_lock, SX_LOCKED);
2440 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2441 if (pr->pr_id == prid) {
2442 mtx_lock(&pr->pr_mtx);
2445 mtx_unlock(&pr->pr_mtx);
2452 * Look for the name relative to mypr. Returns a locked prison or NULL.
2455 prison_find_name(struct prison *mypr, const char *name)
2457 struct prison *pr, *deadpr;
2461 sx_assert(&allprison_lock, SX_LOCKED);
2462 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2465 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2466 if (!strcmp(pr->pr_name + mylen, name)) {
2467 mtx_lock(&pr->pr_mtx);
2468 if (pr->pr_ref > 0) {
2469 if (pr->pr_uref > 0)
2473 mtx_unlock(&pr->pr_mtx);
2476 /* There was no valid prison - perhaps there was a dying one. */
2477 if (deadpr != NULL) {
2478 mtx_lock(&deadpr->pr_mtx);
2479 if (deadpr->pr_ref == 0) {
2480 mtx_unlock(&deadpr->pr_mtx);
2488 * See if a prison has the specific flag set.
2491 prison_flag(struct ucred *cred, unsigned flag)
2494 /* This is an atomic read, so no locking is necessary. */
2495 return (cred->cr_prison->pr_flags & flag);
2499 prison_allow(struct ucred *cred, unsigned flag)
2502 /* This is an atomic read, so no locking is necessary. */
2503 return (cred->cr_prison->pr_allow & flag);
2507 * Remove a prison reference. If that was the last reference, remove the
2508 * prison itself - but not in this context in case there are locks held.
2511 prison_free_locked(struct prison *pr)
2514 mtx_assert(&pr->pr_mtx, MA_OWNED);
2516 if (pr->pr_ref == 0) {
2517 mtx_unlock(&pr->pr_mtx);
2518 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2519 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2522 mtx_unlock(&pr->pr_mtx);
2526 prison_free(struct prison *pr)
2529 mtx_lock(&pr->pr_mtx);
2530 prison_free_locked(pr);
2534 prison_complete(void *context, int pending)
2537 prison_deref((struct prison *)context, 0);
2541 * Remove a prison reference (usually). This internal version assumes no
2542 * mutexes are held, except perhaps the prison itself. If there are no more
2543 * references, release and delist the prison. On completion, the prison lock
2544 * and the allprison lock are both unlocked.
2547 prison_deref(struct prison *pr, int flags)
2549 struct prison *ppr, *tpr;
2552 if (!(flags & PD_LOCKED))
2553 mtx_lock(&pr->pr_mtx);
2555 if (flags & PD_DEUREF) {
2557 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2559 if (flags & PD_DEREF)
2561 /* If the prison still has references, nothing else to do. */
2562 if (pr->pr_ref > 0) {
2563 mtx_unlock(&pr->pr_mtx);
2564 if (flags & PD_LIST_SLOCKED)
2565 sx_sunlock(&allprison_lock);
2566 else if (flags & PD_LIST_XLOCKED)
2567 sx_xunlock(&allprison_lock);
2571 mtx_unlock(&pr->pr_mtx);
2572 if (flags & PD_LIST_SLOCKED) {
2573 if (!sx_try_upgrade(&allprison_lock)) {
2574 sx_sunlock(&allprison_lock);
2575 sx_xlock(&allprison_lock);
2577 } else if (!(flags & PD_LIST_XLOCKED))
2578 sx_xlock(&allprison_lock);
2580 TAILQ_REMOVE(&allprison, pr, pr_list);
2581 LIST_REMOVE(pr, pr_sibling);
2582 ppr = pr->pr_parent;
2583 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2584 tpr->pr_childcount--;
2585 sx_xunlock(&allprison_lock);
2588 if (pr->pr_vnet != ppr->pr_vnet)
2589 vnet_destroy(pr->pr_vnet);
2591 if (pr->pr_root != NULL) {
2592 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2594 VFS_UNLOCK_GIANT(vfslocked);
2596 mtx_destroy(&pr->pr_mtx);
2598 free(pr->pr_ip4, M_PRISON);
2601 free(pr->pr_ip6, M_PRISON);
2603 if (pr->pr_cpuset != NULL)
2604 cpuset_rel(pr->pr_cpuset);
2607 prison_racct_detach(pr);
2611 /* Removing a prison frees a reference on its parent. */
2613 mtx_lock(&pr->pr_mtx);
2614 flags = PD_DEREF | PD_DEUREF;
2619 prison_hold_locked(struct prison *pr)
2622 mtx_assert(&pr->pr_mtx, MA_OWNED);
2623 KASSERT(pr->pr_ref > 0,
2624 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2629 prison_hold(struct prison *pr)
2632 mtx_lock(&pr->pr_mtx);
2633 prison_hold_locked(pr);
2634 mtx_unlock(&pr->pr_mtx);
2638 prison_proc_hold(struct prison *pr)
2641 mtx_lock(&pr->pr_mtx);
2642 KASSERT(pr->pr_uref > 0,
2643 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2645 mtx_unlock(&pr->pr_mtx);
2649 prison_proc_free(struct prison *pr)
2652 mtx_lock(&pr->pr_mtx);
2653 KASSERT(pr->pr_uref > 0,
2654 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2655 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2661 * Restrict a prison's IP address list with its parent's, possibly replacing
2662 * it. Return true if the replacement buffer was used (or would have been).
2665 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2670 ppr = pr->pr_parent;
2671 if (!(pr->pr_flags & PR_IP4_USER)) {
2672 /* This has no user settings, so just copy the parent's list. */
2673 if (pr->pr_ip4s < ppr->pr_ip4s) {
2675 * There's no room for the parent's list. Use the
2676 * new list buffer, which is assumed to be big enough
2677 * (if it was passed). If there's no buffer, try to
2681 if (newip4 == NULL) {
2682 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2683 M_PRISON, M_NOWAIT);
2687 if (newip4 != NULL) {
2688 bcopy(ppr->pr_ip4, newip4,
2689 ppr->pr_ip4s * sizeof(*newip4));
2690 free(pr->pr_ip4, M_PRISON);
2691 pr->pr_ip4 = newip4;
2692 pr->pr_ip4s = ppr->pr_ip4s;
2696 pr->pr_ip4s = ppr->pr_ip4s;
2697 if (pr->pr_ip4s > 0)
2698 bcopy(ppr->pr_ip4, pr->pr_ip4,
2699 pr->pr_ip4s * sizeof(*newip4));
2700 else if (pr->pr_ip4 != NULL) {
2701 free(pr->pr_ip4, M_PRISON);
2704 } else if (pr->pr_ip4s > 0) {
2705 /* Remove addresses that aren't in the parent. */
2706 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2707 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2709 if (ij < ppr->pr_ip4s)
2712 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2713 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2716 for (ij = 1; ii < pr->pr_ip4s; ) {
2717 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2721 switch (ij >= ppr->pr_ip4s ? -1 :
2722 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2724 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2725 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2736 if (pr->pr_ip4s == 0) {
2737 pr->pr_flags |= PR_IP4_DISABLE;
2738 free(pr->pr_ip4, M_PRISON);
2746 * Pass back primary IPv4 address of this jail.
2748 * If not restricted return success but do not alter the address. Caller has
2749 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2751 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2752 * Address returned in NBO.
2755 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2759 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2760 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2762 pr = cred->cr_prison;
2763 if (!(pr->pr_flags & PR_IP4))
2765 mtx_lock(&pr->pr_mtx);
2766 if (!(pr->pr_flags & PR_IP4)) {
2767 mtx_unlock(&pr->pr_mtx);
2770 if (pr->pr_ip4 == NULL) {
2771 mtx_unlock(&pr->pr_mtx);
2772 return (EAFNOSUPPORT);
2775 ia->s_addr = pr->pr_ip4[0].s_addr;
2776 mtx_unlock(&pr->pr_mtx);
2781 * Return 1 if we should do proper source address selection or are not jailed.
2782 * We will return 0 if we should bypass source address selection in favour
2783 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2785 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2788 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2794 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2795 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2800 pr = cred->cr_prison;
2801 if (pr->pr_flags & PR_IP4_SADDRSEL)
2804 lia.s_addr = INADDR_ANY;
2805 error = prison_get_ip4(cred, &lia);
2808 if (lia.s_addr == INADDR_ANY)
2811 ia->s_addr = lia.s_addr;
2816 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2819 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2826 * No need to lock since the PR_IP4_USER flag can't be altered for
2829 while (pr1 != &prison0 &&
2831 !(pr1->pr_flags & PR_VNET) &&
2833 !(pr1->pr_flags & PR_IP4_USER))
2834 pr1 = pr1->pr_parent;
2835 while (pr2 != &prison0 &&
2837 !(pr2->pr_flags & PR_VNET) &&
2839 !(pr2->pr_flags & PR_IP4_USER))
2840 pr2 = pr2->pr_parent;
2841 return (pr1 == pr2);
2845 * Make sure our (source) address is set to something meaningful to this
2848 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2849 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2850 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2853 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2859 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2860 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2862 pr = cred->cr_prison;
2863 if (!(pr->pr_flags & PR_IP4))
2865 mtx_lock(&pr->pr_mtx);
2866 if (!(pr->pr_flags & PR_IP4)) {
2867 mtx_unlock(&pr->pr_mtx);
2870 if (pr->pr_ip4 == NULL) {
2871 mtx_unlock(&pr->pr_mtx);
2872 return (EAFNOSUPPORT);
2875 ia0.s_addr = ntohl(ia->s_addr);
2876 if (ia0.s_addr == INADDR_LOOPBACK) {
2877 ia->s_addr = pr->pr_ip4[0].s_addr;
2878 mtx_unlock(&pr->pr_mtx);
2882 if (ia0.s_addr == INADDR_ANY) {
2884 * In case there is only 1 IPv4 address, bind directly.
2886 if (pr->pr_ip4s == 1)
2887 ia->s_addr = pr->pr_ip4[0].s_addr;
2888 mtx_unlock(&pr->pr_mtx);
2892 error = _prison_check_ip4(pr, ia);
2893 mtx_unlock(&pr->pr_mtx);
2898 * Rewrite destination address in case we will connect to loopback address.
2900 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2901 * Address passed in in NBO and returned in NBO.
2904 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2908 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2909 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2911 pr = cred->cr_prison;
2912 if (!(pr->pr_flags & PR_IP4))
2914 mtx_lock(&pr->pr_mtx);
2915 if (!(pr->pr_flags & PR_IP4)) {
2916 mtx_unlock(&pr->pr_mtx);
2919 if (pr->pr_ip4 == NULL) {
2920 mtx_unlock(&pr->pr_mtx);
2921 return (EAFNOSUPPORT);
2924 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2925 ia->s_addr = pr->pr_ip4[0].s_addr;
2926 mtx_unlock(&pr->pr_mtx);
2931 * Return success because nothing had to be changed.
2933 mtx_unlock(&pr->pr_mtx);
2938 * Check if given address belongs to the jail referenced by cred/prison.
2940 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2941 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2942 * doesn't allow IPv4. Address passed in in NBO.
2945 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2950 * Check the primary IP.
2952 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2956 * All the other IPs are sorted so we can do a binary search.
2959 z = pr->pr_ip4s - 2;
2962 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2971 return (EADDRNOTAVAIL);
2975 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2980 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2981 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2983 pr = cred->cr_prison;
2984 if (!(pr->pr_flags & PR_IP4))
2986 mtx_lock(&pr->pr_mtx);
2987 if (!(pr->pr_flags & PR_IP4)) {
2988 mtx_unlock(&pr->pr_mtx);
2991 if (pr->pr_ip4 == NULL) {
2992 mtx_unlock(&pr->pr_mtx);
2993 return (EAFNOSUPPORT);
2996 error = _prison_check_ip4(pr, ia);
2997 mtx_unlock(&pr->pr_mtx);
3004 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3009 ppr = pr->pr_parent;
3010 if (!(pr->pr_flags & PR_IP6_USER)) {
3011 /* This has no user settings, so just copy the parent's list. */
3012 if (pr->pr_ip6s < ppr->pr_ip6s) {
3014 * There's no room for the parent's list. Use the
3015 * new list buffer, which is assumed to be big enough
3016 * (if it was passed). If there's no buffer, try to
3020 if (newip6 == NULL) {
3021 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3022 M_PRISON, M_NOWAIT);
3026 if (newip6 != NULL) {
3027 bcopy(ppr->pr_ip6, newip6,
3028 ppr->pr_ip6s * sizeof(*newip6));
3029 free(pr->pr_ip6, M_PRISON);
3030 pr->pr_ip6 = newip6;
3031 pr->pr_ip6s = ppr->pr_ip6s;
3035 pr->pr_ip6s = ppr->pr_ip6s;
3036 if (pr->pr_ip6s > 0)
3037 bcopy(ppr->pr_ip6, pr->pr_ip6,
3038 pr->pr_ip6s * sizeof(*newip6));
3039 else if (pr->pr_ip6 != NULL) {
3040 free(pr->pr_ip6, M_PRISON);
3043 } else if (pr->pr_ip6s > 0) {
3044 /* Remove addresses that aren't in the parent. */
3045 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3046 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3049 if (ij < ppr->pr_ip6s)
3052 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3053 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3056 for (ij = 1; ii < pr->pr_ip6s; ) {
3057 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3062 switch (ij >= ppr->pr_ip6s ? -1 :
3063 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3065 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3066 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3077 if (pr->pr_ip6s == 0) {
3078 pr->pr_flags |= PR_IP6_DISABLE;
3079 free(pr->pr_ip6, M_PRISON);
3087 * Pass back primary IPv6 address for this jail.
3089 * If not restricted return success but do not alter the address. Caller has
3090 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3092 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3095 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3099 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3100 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3102 pr = cred->cr_prison;
3103 if (!(pr->pr_flags & PR_IP6))
3105 mtx_lock(&pr->pr_mtx);
3106 if (!(pr->pr_flags & PR_IP6)) {
3107 mtx_unlock(&pr->pr_mtx);
3110 if (pr->pr_ip6 == NULL) {
3111 mtx_unlock(&pr->pr_mtx);
3112 return (EAFNOSUPPORT);
3115 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3116 mtx_unlock(&pr->pr_mtx);
3121 * Return 1 if we should do proper source address selection or are not jailed.
3122 * We will return 0 if we should bypass source address selection in favour
3123 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3125 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3128 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3131 struct in6_addr lia6;
3134 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3135 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3140 pr = cred->cr_prison;
3141 if (pr->pr_flags & PR_IP6_SADDRSEL)
3145 error = prison_get_ip6(cred, &lia6);
3148 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3151 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3156 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3159 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3165 while (pr1 != &prison0 &&
3167 !(pr1->pr_flags & PR_VNET) &&
3169 !(pr1->pr_flags & PR_IP6_USER))
3170 pr1 = pr1->pr_parent;
3171 while (pr2 != &prison0 &&
3173 !(pr2->pr_flags & PR_VNET) &&
3175 !(pr2->pr_flags & PR_IP6_USER))
3176 pr2 = pr2->pr_parent;
3177 return (pr1 == pr2);
3181 * Make sure our (source) address is set to something meaningful to this jail.
3183 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3184 * when needed while binding.
3186 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3187 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3188 * doesn't allow IPv6.
3191 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3196 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3197 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3199 pr = cred->cr_prison;
3200 if (!(pr->pr_flags & PR_IP6))
3202 mtx_lock(&pr->pr_mtx);
3203 if (!(pr->pr_flags & PR_IP6)) {
3204 mtx_unlock(&pr->pr_mtx);
3207 if (pr->pr_ip6 == NULL) {
3208 mtx_unlock(&pr->pr_mtx);
3209 return (EAFNOSUPPORT);
3212 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3213 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3214 mtx_unlock(&pr->pr_mtx);
3218 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3220 * In case there is only 1 IPv6 address, and v6only is true,
3221 * then bind directly.
3223 if (v6only != 0 && pr->pr_ip6s == 1)
3224 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3225 mtx_unlock(&pr->pr_mtx);
3229 error = _prison_check_ip6(pr, ia6);
3230 mtx_unlock(&pr->pr_mtx);
3235 * Rewrite destination address in case we will connect to loopback address.
3237 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3240 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3244 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3245 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3247 pr = cred->cr_prison;
3248 if (!(pr->pr_flags & PR_IP6))
3250 mtx_lock(&pr->pr_mtx);
3251 if (!(pr->pr_flags & PR_IP6)) {
3252 mtx_unlock(&pr->pr_mtx);
3255 if (pr->pr_ip6 == NULL) {
3256 mtx_unlock(&pr->pr_mtx);
3257 return (EAFNOSUPPORT);
3260 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3261 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3262 mtx_unlock(&pr->pr_mtx);
3267 * Return success because nothing had to be changed.
3269 mtx_unlock(&pr->pr_mtx);
3274 * Check if given address belongs to the jail referenced by cred/prison.
3276 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3277 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3278 * doesn't allow IPv6.
3281 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3286 * Check the primary IP.
3288 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3292 * All the other IPs are sorted so we can do a binary search.
3295 z = pr->pr_ip6s - 2;
3298 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3307 return (EADDRNOTAVAIL);
3311 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3316 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3317 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3319 pr = cred->cr_prison;
3320 if (!(pr->pr_flags & PR_IP6))
3322 mtx_lock(&pr->pr_mtx);
3323 if (!(pr->pr_flags & PR_IP6)) {
3324 mtx_unlock(&pr->pr_mtx);
3327 if (pr->pr_ip6 == NULL) {
3328 mtx_unlock(&pr->pr_mtx);
3329 return (EAFNOSUPPORT);
3332 error = _prison_check_ip6(pr, ia6);
3333 mtx_unlock(&pr->pr_mtx);
3339 * Check if a jail supports the given address family.
3341 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3345 prison_check_af(struct ucred *cred, int af)
3350 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3352 pr = cred->cr_prison;
3354 /* Prisons with their own network stack are not limited. */
3355 if (prison_owns_vnet(cred))
3364 if (pr->pr_flags & PR_IP4)
3366 mtx_lock(&pr->pr_mtx);
3367 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3368 error = EAFNOSUPPORT;
3369 mtx_unlock(&pr->pr_mtx);
3375 if (pr->pr_flags & PR_IP6)
3377 mtx_lock(&pr->pr_mtx);
3378 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3379 error = EAFNOSUPPORT;
3380 mtx_unlock(&pr->pr_mtx);
3388 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3389 error = EAFNOSUPPORT;
3395 * Check if given address belongs to the jail referenced by cred (wrapper to
3396 * prison_check_ip[46]).
3398 * Returns 0 if jail doesn't restrict the address family or if address belongs
3399 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3400 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3403 prison_if(struct ucred *cred, struct sockaddr *sa)
3406 struct sockaddr_in *sai;
3409 struct sockaddr_in6 *sai6;
3413 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3414 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3417 if (prison_owns_vnet(cred))
3422 switch (sa->sa_family)
3426 sai = (struct sockaddr_in *)sa;
3427 error = prison_check_ip4(cred, &sai->sin_addr);
3432 sai6 = (struct sockaddr_in6 *)sa;
3433 error = prison_check_ip6(cred, &sai6->sin6_addr);
3437 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3438 error = EAFNOSUPPORT;
3444 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3447 prison_check(struct ucred *cred1, struct ucred *cred2)
3450 return ((cred1->cr_prison == cred2->cr_prison ||
3451 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3455 * Return 1 if p2 is a child of p1, otherwise 0.
3458 prison_ischild(struct prison *pr1, struct prison *pr2)
3461 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3468 * Return 1 if the passed credential is in a jail, otherwise 0.
3471 jailed(struct ucred *cred)
3474 return (cred->cr_prison != &prison0);
3478 * Return 1 if the passed credential is in a jail and that jail does not
3479 * have its own virtual network stack, otherwise 0.
3482 jailed_without_vnet(struct ucred *cred)
3488 if (prison_owns_vnet(cred))
3496 * Return the correct hostname (domainname, et al) for the passed credential.
3499 getcredhostname(struct ucred *cred, char *buf, size_t size)
3504 * A NULL credential can be used to shortcut to the physical
3505 * system's hostname.
3507 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3508 mtx_lock(&pr->pr_mtx);
3509 strlcpy(buf, pr->pr_hostname, size);
3510 mtx_unlock(&pr->pr_mtx);
3514 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3517 mtx_lock(&cred->cr_prison->pr_mtx);
3518 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3519 mtx_unlock(&cred->cr_prison->pr_mtx);
3523 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3526 mtx_lock(&cred->cr_prison->pr_mtx);
3527 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3528 mtx_unlock(&cred->cr_prison->pr_mtx);
3532 getcredhostid(struct ucred *cred, unsigned long *hostid)
3535 mtx_lock(&cred->cr_prison->pr_mtx);
3536 *hostid = cred->cr_prison->pr_hostid;
3537 mtx_unlock(&cred->cr_prison->pr_mtx);
3542 * Determine whether the prison represented by cred owns
3543 * its vnet rather than having it inherited.
3545 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3548 prison_owns_vnet(struct ucred *cred)
3552 * vnets cannot be added/removed after jail creation,
3553 * so no need to lock here.
3555 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3560 * Determine whether the subject represented by cred can "see"
3561 * status of a mount point.
3562 * Returns: 0 for permitted, ENOENT otherwise.
3563 * XXX: This function should be called cr_canseemount() and should be
3564 * placed in kern_prot.c.
3567 prison_canseemount(struct ucred *cred, struct mount *mp)
3573 pr = cred->cr_prison;
3574 if (pr->pr_enforce_statfs == 0)
3576 if (pr->pr_root->v_mount == mp)
3578 if (pr->pr_enforce_statfs == 2)
3581 * If jail's chroot directory is set to "/" we should be able to see
3582 * all mount-points from inside a jail.
3583 * This is ugly check, but this is the only situation when jail's
3584 * directory ends with '/'.
3586 if (strcmp(pr->pr_path, "/") == 0)
3588 len = strlen(pr->pr_path);
3590 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3593 * Be sure that we don't have situation where jail's root directory
3594 * is "/some/path" and mount point is "/some/pathpath".
3596 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3602 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3604 char jpath[MAXPATHLEN];
3608 pr = cred->cr_prison;
3609 if (pr->pr_enforce_statfs == 0)
3611 if (prison_canseemount(cred, mp) != 0) {
3612 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3613 strlcpy(sp->f_mntonname, "[restricted]",
3614 sizeof(sp->f_mntonname));
3617 if (pr->pr_root->v_mount == mp) {
3619 * Clear current buffer data, so we are sure nothing from
3620 * the valid path left there.
3622 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3623 *sp->f_mntonname = '/';
3627 * If jail's chroot directory is set to "/" we should be able to see
3628 * all mount-points from inside a jail.
3630 if (strcmp(pr->pr_path, "/") == 0)
3632 len = strlen(pr->pr_path);
3633 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3635 * Clear current buffer data, so we are sure nothing from
3636 * the valid path left there.
3638 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3639 if (*jpath == '\0') {
3640 /* Should never happen. */
3641 *sp->f_mntonname = '/';
3643 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3648 * Check with permission for a specific privilege is granted within jail. We
3649 * have a specific list of accepted privileges; the rest are denied.
3652 prison_priv_check(struct ucred *cred, int priv)
3660 * Privileges specific to prisons with a virtual network stack.
3661 * There might be a duplicate entry here in case the privilege
3662 * is only granted conditionally in the legacy jail case.
3667 * NFS-specific privileges.
3669 case PRIV_NFS_DAEMON:
3670 case PRIV_NFS_LOCKD:
3673 * Network stack privileges.
3675 case PRIV_NET_BRIDGE:
3678 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3679 case PRIV_NET_ROUTE:
3681 case PRIV_NET_SETIFMTU:
3682 case PRIV_NET_SETIFFLAGS:
3683 case PRIV_NET_SETIFCAP:
3684 case PRIV_NET_SETIFDESCR:
3685 case PRIV_NET_SETIFNAME :
3686 case PRIV_NET_SETIFMETRIC:
3687 case PRIV_NET_SETIFPHYS:
3688 case PRIV_NET_SETIFMAC:
3689 case PRIV_NET_ADDMULTI:
3690 case PRIV_NET_DELMULTI:
3691 case PRIV_NET_HWIOCTL:
3692 case PRIV_NET_SETLLADDR:
3693 case PRIV_NET_ADDIFGROUP:
3694 case PRIV_NET_DELIFGROUP:
3695 case PRIV_NET_IFCREATE:
3696 case PRIV_NET_IFDESTROY:
3697 case PRIV_NET_ADDIFADDR:
3698 case PRIV_NET_DELIFADDR:
3701 case PRIV_NET_SETIFVNET:
3702 case PRIV_NET_SETIFFIB:
3705 * 802.11-related privileges.
3707 case PRIV_NET80211_GETKEY:
3709 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3714 * AppleTalk privileges.
3716 case PRIV_NETATALK_RESERVEDPORT:
3721 case PRIV_NETATM_CFG:
3722 case PRIV_NETATM_ADD:
3723 case PRIV_NETATM_DEL:
3724 case PRIV_NETATM_SET:
3727 * Bluetooth privileges.
3729 case PRIV_NETBLUETOOTH_RAW:
3733 * Netgraph and netgraph module privileges.
3735 case PRIV_NETGRAPH_CONTROL:
3737 case PRIV_NETGRAPH_TTY:
3741 * IPv4 and IPv6 privileges.
3743 case PRIV_NETINET_IPFW:
3744 case PRIV_NETINET_DIVERT:
3745 case PRIV_NETINET_PF:
3746 case PRIV_NETINET_DUMMYNET:
3747 case PRIV_NETINET_CARP:
3748 case PRIV_NETINET_MROUTE:
3749 case PRIV_NETINET_RAW:
3750 case PRIV_NETINET_ADDRCTRL6:
3751 case PRIV_NETINET_ND6:
3752 case PRIV_NETINET_SCOPE6:
3753 case PRIV_NETINET_ALIFETIME6:
3754 case PRIV_NETINET_IPSEC:
3755 case PRIV_NETINET_BINDANY:
3759 * IPX/SPX privileges.
3761 case PRIV_NETIPX_RESERVEDPORT:
3762 case PRIV_NETIPX_RAW:
3776 * No default: or deny here.
3777 * In case of no permit fall through to next switch().
3779 if (cred->cr_prison->pr_flags & PR_VNET)
3787 * Allow ktrace privileges for root in jail.
3793 * Allow jailed processes to configure audit identity and
3794 * submit audit records (login, etc). In the future we may
3795 * want to further refine the relationship between audit and
3798 case PRIV_AUDIT_GETAUDIT:
3799 case PRIV_AUDIT_SETAUDIT:
3800 case PRIV_AUDIT_SUBMIT:
3804 * Allow jailed processes to manipulate process UNIX
3805 * credentials in any way they see fit.
3807 case PRIV_CRED_SETUID:
3808 case PRIV_CRED_SETEUID:
3809 case PRIV_CRED_SETGID:
3810 case PRIV_CRED_SETEGID:
3811 case PRIV_CRED_SETGROUPS:
3812 case PRIV_CRED_SETREUID:
3813 case PRIV_CRED_SETREGID:
3814 case PRIV_CRED_SETRESUID:
3815 case PRIV_CRED_SETRESGID:
3818 * Jail implements visibility constraints already, so allow
3819 * jailed root to override uid/gid-based constraints.
3821 case PRIV_SEEOTHERGIDS:
3822 case PRIV_SEEOTHERUIDS:
3825 * Jail implements inter-process debugging limits already, so
3826 * allow jailed root various debugging privileges.
3828 case PRIV_DEBUG_DIFFCRED:
3829 case PRIV_DEBUG_SUGID:
3830 case PRIV_DEBUG_UNPRIV:
3833 * Allow jail to set various resource limits and login
3834 * properties, and for now, exceed process resource limits.
3836 case PRIV_PROC_LIMIT:
3837 case PRIV_PROC_SETLOGIN:
3838 case PRIV_PROC_SETRLIMIT:
3841 * System V and POSIX IPC privileges are granted in jail.
3844 case PRIV_IPC_WRITE:
3845 case PRIV_IPC_ADMIN:
3846 case PRIV_IPC_MSGSIZE:
3850 * Jail operations within a jail work on child jails.
3852 case PRIV_JAIL_ATTACH:
3854 case PRIV_JAIL_REMOVE:
3857 * Jail implements its own inter-process limits, so allow
3858 * root processes in jail to change scheduling on other
3859 * processes in the same jail. Likewise for signalling.
3861 case PRIV_SCHED_DIFFCRED:
3862 case PRIV_SCHED_CPUSET:
3863 case PRIV_SIGNAL_DIFFCRED:
3864 case PRIV_SIGNAL_SUGID:
3867 * Allow jailed processes to write to sysctls marked as jail
3870 case PRIV_SYSCTL_WRITEJAIL:
3873 * Allow root in jail to manage a variety of quota
3874 * properties. These should likely be conditional on a
3875 * configuration option.
3877 case PRIV_VFS_GETQUOTA:
3878 case PRIV_VFS_SETQUOTA:
3881 * Since Jail relies on chroot() to implement file system
3882 * protections, grant many VFS privileges to root in jail.
3883 * Be careful to exclude mount-related and NFS-related
3887 case PRIV_VFS_WRITE:
3888 case PRIV_VFS_ADMIN:
3890 case PRIV_VFS_LOOKUP:
3891 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3892 case PRIV_VFS_CHFLAGS_DEV:
3893 case PRIV_VFS_CHOWN:
3894 case PRIV_VFS_CHROOT:
3895 case PRIV_VFS_RETAINSUGID:
3896 case PRIV_VFS_FCHROOT:
3898 case PRIV_VFS_SETGID:
3900 case PRIV_VFS_STICKYFILE:
3904 * Depending on the global setting, allow privilege of
3905 * setting system flags.
3907 case PRIV_VFS_SYSFLAGS:
3908 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3914 * Depending on the global setting, allow privilege of
3915 * mounting/unmounting file systems.
3917 case PRIV_VFS_MOUNT:
3918 case PRIV_VFS_UNMOUNT:
3919 case PRIV_VFS_MOUNT_NONUSER:
3920 case PRIV_VFS_MOUNT_OWNER:
3921 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3922 cred->cr_prison->pr_enforce_statfs < 2)
3928 * Allow jailed root to bind reserved ports and reuse in-use
3931 case PRIV_NETINET_RESERVEDPORT:
3932 case PRIV_NETINET_REUSEPORT:
3936 * Allow jailed root to set certian IPv4/6 (option) headers.
3938 case PRIV_NETINET_SETHDROPTS:
3942 * Conditionally allow creating raw sockets in jail.
3944 case PRIV_NETINET_RAW:
3945 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3951 * Since jail implements its own visibility limits on netstat
3952 * sysctls, allow getcred. This allows identd to work in
3955 case PRIV_NETINET_GETCRED:
3959 * Allow jailed root to set loginclass.
3961 case PRIV_PROC_SETLOGINCLASS:
3966 * In all remaining cases, deny the privilege request. This
3967 * includes almost all network privileges, many system
3968 * configuration privileges.
3975 * Return the part of pr2's name that is relative to pr1, or the whole name
3976 * if it does not directly follow.
3980 prison_name(struct prison *pr1, struct prison *pr2)
3984 /* Jails see themselves as "0" (if they see themselves at all). */
3987 name = pr2->pr_name;
3988 if (prison_ischild(pr1, pr2)) {
3990 * pr1 isn't locked (and allprison_lock may not be either)
3991 * so its length can't be counted on. But the number of dots
3992 * can be counted on - and counted.
3994 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3995 name = strchr(name, '.') + 1;
4001 * Return the part of pr2's path that is relative to pr1, or the whole path
4002 * if it does not directly follow.
4005 prison_path(struct prison *pr1, struct prison *pr2)
4007 char *path1, *path2;
4010 path1 = pr1->pr_path;
4011 path2 = pr2->pr_path;
4012 if (!strcmp(path1, "/"))
4014 len1 = strlen(path1);
4015 if (strncmp(path1, path2, len1))
4017 if (path2[len1] == '\0')
4019 if (path2[len1] == '/')
4020 return (path2 + len1);
4026 * Jail-related sysctls.
4028 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4032 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4035 struct prison *pr, *cpr;
4037 struct in_addr *ip4 = NULL;
4041 struct in6_addr *ip6 = NULL;
4046 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4047 pr = req->td->td_ucred->cr_prison;
4049 sx_slock(&allprison_lock);
4050 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4051 #if defined(INET) || defined(INET6)
4054 mtx_lock(&cpr->pr_mtx);
4056 if (cpr->pr_ip4s > 0) {
4057 if (ip4s < cpr->pr_ip4s) {
4058 ip4s = cpr->pr_ip4s;
4059 mtx_unlock(&cpr->pr_mtx);
4060 ip4 = realloc(ip4, ip4s *
4061 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4064 bcopy(cpr->pr_ip4, ip4,
4065 cpr->pr_ip4s * sizeof(struct in_addr));
4069 if (cpr->pr_ip6s > 0) {
4070 if (ip6s < cpr->pr_ip6s) {
4071 ip6s = cpr->pr_ip6s;
4072 mtx_unlock(&cpr->pr_mtx);
4073 ip6 = realloc(ip6, ip6s *
4074 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4077 bcopy(cpr->pr_ip6, ip6,
4078 cpr->pr_ip6s * sizeof(struct in6_addr));
4081 if (cpr->pr_ref == 0) {
4082 mtx_unlock(&cpr->pr_mtx);
4085 bzero(xp, sizeof(*xp));
4086 xp->pr_version = XPRISON_VERSION;
4087 xp->pr_id = cpr->pr_id;
4088 xp->pr_state = cpr->pr_uref > 0
4089 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4090 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4091 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4092 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4094 xp->pr_ip4s = cpr->pr_ip4s;
4097 xp->pr_ip6s = cpr->pr_ip6s;
4099 mtx_unlock(&cpr->pr_mtx);
4100 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4104 if (xp->pr_ip4s > 0) {
4105 error = SYSCTL_OUT(req, ip4,
4106 xp->pr_ip4s * sizeof(struct in_addr));
4112 if (xp->pr_ip6s > 0) {
4113 error = SYSCTL_OUT(req, ip6,
4114 xp->pr_ip6s * sizeof(struct in6_addr));
4120 sx_sunlock(&allprison_lock);
4131 SYSCTL_OID(_security_jail, OID_AUTO, list,
4132 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4133 sysctl_jail_list, "S", "List of active jails");
4136 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4140 injail = jailed(req->td->td_ucred);
4141 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4146 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4147 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4148 sysctl_jail_jailed, "I", "Process in jail?");
4151 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4153 int error, havevnet;
4155 struct ucred *cred = req->td->td_ucred;
4157 havevnet = jailed(cred) && prison_owns_vnet(cred);
4161 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4166 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4167 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4168 sysctl_jail_vnet, "I", "Jail owns VNET?");
4170 #if defined(INET) || defined(INET6)
4171 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4172 &jail_max_af_ips, 0,
4173 "Number of IP addresses a jail may have at most per address family");
4177 * Default parameters for jail(2) compatability. For historical reasons,
4178 * the sysctl names have varying similarity to the parameter names. Prisons
4179 * just see their own parameters, and can't change them.
4182 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4185 int allow, error, i;
4187 pr = req->td->td_ucred->cr_prison;
4188 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4190 /* Get the current flag value, and convert it to a boolean. */
4191 i = (allow & arg2) ? 1 : 0;
4194 error = sysctl_handle_int(oidp, &i, 0, req);
4195 if (error || !req->newptr)
4201 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4204 mtx_lock(&prison0.pr_mtx);
4205 jail_default_allow = (jail_default_allow & ~arg2) | i;
4206 mtx_unlock(&prison0.pr_mtx);
4210 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4211 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4212 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4213 "Processes in jail can set their hostnames");
4214 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4215 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4216 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4217 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4218 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4219 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4220 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4221 "Processes in jail can use System V IPC primitives");
4222 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4223 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4224 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4225 "Prison root can create raw sockets");
4226 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4227 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4228 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4229 "Processes in jail can alter system file flags");
4230 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4231 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4232 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4233 "Processes in jail can mount/unmount jail-friendly file systems");
4234 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4235 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4236 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4237 "Processes in jail can mount the devfs file system");
4238 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4239 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4240 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4241 "Processes in jail can mount the nullfs file system");
4242 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4243 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4244 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4245 "Processes in jail can mount the procfs file system");
4246 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4247 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4248 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4249 "Processes in jail can mount the zfs file system");
4252 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4257 pr = req->td->td_ucred->cr_prison;
4258 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4259 error = sysctl_handle_int(oidp, &level, 0, req);
4260 if (error || !req->newptr)
4262 *(int *)arg1 = level;
4266 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4267 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4268 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4269 sysctl_jail_default_level, "I",
4270 "Processes in jail cannot see all mounted file systems");
4272 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4273 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4274 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4275 sysctl_jail_default_level, "I",
4276 "Ruleset for the devfs filesystem in jail");
4279 * Nodes to describe jail parameters. Maximum length of string parameters
4280 * is returned in the string itself, and the other parameters exist merely
4281 * to make themselves and their types known.
4283 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4287 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4294 switch (oidp->oid_kind & CTLTYPE)
4300 if (!(req->flags & SCTL_MASK32))
4302 return (SYSCTL_OUT(req, &l, sizeof(l)));
4306 return (SYSCTL_OUT(req, &i, sizeof(i)));
4307 case CTLTYPE_STRING:
4308 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4310 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4311 case CTLTYPE_STRUCT:
4313 return (SYSCTL_OUT(req, &s, sizeof(s)));
4318 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4319 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4320 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4321 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4322 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4323 "I", "Jail secure level");
4324 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4325 "I", "Jail cannot see all mounted file systems");
4326 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4327 "I", "Ruleset for in-jail devfs mounts");
4328 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4329 "B", "Jail persistence");
4331 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4332 "E,jailsys", "Virtual network stack");
4334 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4335 "B", "Jail is in the process of shutting down");
4337 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4338 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4339 "I", "Current number of child jails");
4340 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4341 "I", "Maximum number of child jails");
4343 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4344 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4346 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4347 "Jail NIS domainname");
4348 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4350 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4351 "LU", "Jail host ID");
4353 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4354 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4357 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4358 "Jail IPv4 address virtualization");
4359 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4360 "S,in_addr,a", "Jail IPv4 addresses");
4361 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4362 "B", "Do (not) use IPv4 source address selection rather than the "
4363 "primary jail IPv4 address.");
4366 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4367 "Jail IPv6 address virtualization");
4368 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4369 "S,in6_addr,a", "Jail IPv6 addresses");
4370 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4371 "B", "Do (not) use IPv6 source address selection rather than the "
4372 "primary jail IPv6 address.");
4375 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4376 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4377 "B", "Jail may set hostname");
4378 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4379 "B", "Jail may use SYSV IPC");
4380 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4381 "B", "Jail may create raw sockets");
4382 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4383 "B", "Jail may alter system file flags");
4384 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4385 "B", "Jail may set file quotas");
4386 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4387 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4389 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4390 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4391 "B", "Jail may mount/unmount jail-friendly file systems in general");
4392 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4393 "B", "Jail may mount the devfs file system");
4394 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4395 "B", "Jail may mount the nullfs file system");
4396 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4397 "B", "Jail may mount the procfs 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;