2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include "opt_compat.h"
35 #include "opt_inet6.h"
37 #include <sys/param.h>
38 #include <sys/types.h>
39 #include <sys/kernel.h>
40 #include <sys/systm.h>
41 #include <sys/errno.h>
42 #include <sys/sysproto.h>
43 #include <sys/malloc.h>
47 #include <sys/taskqueue.h>
48 #include <sys/fcntl.h>
51 #include <sys/mutex.h>
52 #include <sys/racct.h>
53 #include <sys/refcount.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
67 #include <netinet/in.h>
72 #include <netinet6/in6_var.h>
76 #include <security/mac/mac_framework.h>
78 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
81 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
92 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
94 #define _PR_IP_SADDRSEL 0
98 /* prison0 describes what is "real" about the system. */
99 struct prison prison0 = {
105 .pr_securelevel = -1,
107 .pr_childmax = JAIL_MAX,
108 .pr_hostuuid = DEFAULT_HOSTUUID,
109 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
111 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
113 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
115 .pr_allow = PR_ALLOW_ALL,
117 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
119 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
120 struct sx allprison_lock;
121 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
122 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
123 LIST_HEAD(, prison_racct) allprison_racct;
126 static int do_jail_attach(struct thread *td, struct prison *pr);
127 static void prison_complete(void *context, int pending);
128 static void prison_deref(struct prison *pr, int flags);
129 static char *prison_path(struct prison *pr1, struct prison *pr2);
130 static void prison_remove_one(struct prison *pr);
132 static void prison_racct_attach(struct prison *pr);
133 static void prison_racct_modify(struct prison *pr);
134 static void prison_racct_detach(struct prison *pr);
137 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
138 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
141 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
142 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
145 /* Flags for prison_deref */
146 #define PD_DEREF 0x01
147 #define PD_DEUREF 0x02
148 #define PD_LOCKED 0x04
149 #define PD_LIST_SLOCKED 0x08
150 #define PD_LIST_XLOCKED 0x10
153 * Parameter names corresponding to PR_* flag values. Size values are for kvm
154 * as we cannot figure out the size of a sparse array, or an array without a
157 static char *pr_flag_names[] = {
160 [7] = "ip4.saddrsel",
163 [8] = "ip6.saddrsel",
166 const size_t pr_flag_names_size = sizeof(pr_flag_names);
168 static char *pr_flag_nonames[] = {
171 [7] = "ip4.nosaddrsel",
174 [8] = "ip6.nosaddrsel",
177 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
179 struct jailsys_flags {
183 } pr_flag_jailsys[] = {
184 { "host", 0, PR_HOST },
186 { "vnet", 0, PR_VNET },
189 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
192 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
195 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
197 static char *pr_allow_names[] = {
198 "allow.set_hostname",
206 "allow.mount.nullfs",
208 "allow.mount.procfs",
210 "allow.mount.fdescfs",
212 const size_t pr_allow_names_size = sizeof(pr_allow_names);
214 static char *pr_allow_nonames[] = {
215 "allow.noset_hostname",
217 "allow.noraw_sockets",
222 "allow.mount.nodevfs",
223 "allow.mount.nonullfs",
225 "allow.mount.noprocfs",
226 "allow.mount.notmpfs",
227 "allow.mount.nofdescfs",
229 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
231 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
232 #define JAIL_DEFAULT_ENFORCE_STATFS 2
233 #define JAIL_DEFAULT_DEVFS_RSNUM 0
234 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
235 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
236 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
237 #if defined(INET) || defined(INET6)
238 static unsigned jail_max_af_ips = 255;
243 qcmp_v4(const void *ip1, const void *ip2)
248 * We need to compare in HBO here to get the list sorted as expected
249 * by the result of the code. Sorting NBO addresses gives you
250 * interesting results. If you do not understand, do not try.
252 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
253 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
256 * Do not simply return the difference of the two numbers, the int is
270 qcmp_v6(const void *ip1, const void *ip2)
272 const struct in6_addr *ia6a, *ia6b;
275 ia6a = (const struct in6_addr *)ip1;
276 ia6b = (const struct in6_addr *)ip2;
279 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
280 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
282 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
295 sys_jail(struct thread *td, struct jail_args *uap)
301 error = copyin(uap->jail, &version, sizeof(uint32_t));
310 /* FreeBSD single IPv4 jails. */
311 bzero(&j, sizeof(struct jail));
312 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
315 j.version = j0.version;
317 j.hostname = j0.hostname;
318 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
324 * Version 1 was used by multi-IPv4 jail implementations
325 * that never made it into the official kernel.
329 case 2: /* JAIL_API_VERSION */
330 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
331 error = copyin(uap->jail, &j, sizeof(struct jail));
337 /* Sci-Fi jails are not supported, sorry. */
340 return (kern_jail(td, &j));
344 kern_jail(struct thread *td, struct jail *j)
346 struct iovec optiov[2 * (4
347 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
356 char *u_path, *u_hostname, *u_name;
359 struct in_addr *u_ip4;
362 struct in6_addr *u_ip6;
365 int error, enforce_statfs, fi;
367 bzero(&optiov, sizeof(optiov));
368 opt.uio_iov = optiov;
372 opt.uio_segflg = UIO_SYSSPACE;
373 opt.uio_rw = UIO_READ;
376 /* Set permissions for top-level jails from sysctls. */
377 if (!jailed(td->td_ucred)) {
378 for (fi = 0; fi < sizeof(pr_allow_names) /
379 sizeof(pr_allow_names[0]); fi++) {
380 optiov[opt.uio_iovcnt].iov_base =
381 (jail_default_allow & (1 << fi))
382 ? pr_allow_names[fi] : pr_allow_nonames[fi];
383 optiov[opt.uio_iovcnt].iov_len =
384 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
387 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
388 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
390 enforce_statfs = jail_default_enforce_statfs;
391 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
392 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
396 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
398 ip4s = (j->version == 0) ? 1 : j->ip4s;
399 if (ip4s > jail_max_af_ips)
401 tmplen += ip4s * sizeof(struct in_addr);
407 if (j->ip6s > jail_max_af_ips)
409 tmplen += j->ip6s * sizeof(struct in6_addr);
414 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
415 u_hostname = u_path + MAXPATHLEN;
416 u_name = u_hostname + MAXHOSTNAMELEN;
418 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
422 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
424 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
427 optiov[opt.uio_iovcnt].iov_base = "path";
428 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
430 optiov[opt.uio_iovcnt].iov_base = u_path;
431 error = copyinstr(j->path, u_path, MAXPATHLEN,
432 &optiov[opt.uio_iovcnt].iov_len);
434 free(u_path, M_TEMP);
438 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
439 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
441 optiov[opt.uio_iovcnt].iov_base = u_hostname;
442 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
443 &optiov[opt.uio_iovcnt].iov_len);
445 free(u_path, M_TEMP);
449 if (j->jailname != NULL) {
450 optiov[opt.uio_iovcnt].iov_base = "name";
451 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
453 optiov[opt.uio_iovcnt].iov_base = u_name;
454 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
455 &optiov[opt.uio_iovcnt].iov_len);
457 free(u_path, M_TEMP);
463 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
464 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
466 optiov[opt.uio_iovcnt].iov_base = u_ip4;
467 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
469 u_ip4->s_addr = j->ip4s;
471 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
473 free(u_path, M_TEMP);
480 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
481 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
483 optiov[opt.uio_iovcnt].iov_base = u_ip6;
484 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
485 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
487 free(u_path, M_TEMP);
492 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
493 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
494 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
495 free(u_path, M_TEMP);
501 * struct jail_set_args {
502 * struct iovec *iovp;
503 * unsigned int iovcnt;
508 sys_jail_set(struct thread *td, struct jail_set_args *uap)
513 /* Check that we have an even number of iovecs. */
517 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
520 error = kern_jail_set(td, auio, uap->flags);
526 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
533 struct in6_addr *ip6;
536 struct vfsoptlist *opts;
537 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
539 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
541 #if defined(INET) || defined(INET6)
546 size_t namelen, onamelen;
547 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
548 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
549 int fi, jid, jsys, len, level;
550 int childmax, rsnum, slevel;
551 int fullpath_disabled;
552 #if defined(INET) || defined(INET6)
561 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
565 error = priv_check(td, PRIV_JAIL_SET);
566 if (!error && (flags & JAIL_ATTACH))
567 error = priv_check(td, PRIV_JAIL_ATTACH);
570 mypr = ppr = td->td_ucred->cr_prison;
571 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
573 if (flags & ~JAIL_SET_MASK)
577 * Check all the parameters before committing to anything. Not all
578 * errors can be caught early, but we may as well try. Also, this
579 * takes care of some expensive stuff (path lookup) before getting
580 * the allprison lock.
582 * XXX Jails are not filesystems, and jail parameters are not mount
583 * options. But it makes more sense to re-use the vfsopt code
584 * than duplicate it under a different name.
586 error = vfs_buildopts(optuio, &opts);
597 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
603 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
612 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
620 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
625 else if (enforce < 0 || enforce > 2) {
631 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
639 pr_flags = ch_flags = 0;
640 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
642 if (pr_flag_names[fi] == NULL)
644 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
645 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
647 ch_flags |= pr_flags;
648 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
650 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
657 case JAIL_SYS_DISABLE:
658 if (!pr_flag_jailsys[fi].disable) {
662 pr_flags |= pr_flag_jailsys[fi].disable;
665 pr_flags |= pr_flag_jailsys[fi].new;
667 case JAIL_SYS_INHERIT:
674 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
676 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
677 && !(pr_flags & PR_PERSIST)) {
679 vfs_opterror(opts, "new jail must persist or attach");
683 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
685 vfs_opterror(opts, "vnet cannot be changed after creation");
690 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
692 vfs_opterror(opts, "ip4 cannot be changed after creation");
697 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
699 vfs_opterror(opts, "ip6 cannot be changed after creation");
704 pr_allow = ch_allow = 0;
705 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
707 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
708 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
710 ch_allow |= pr_allow;
712 error = vfs_getopt(opts, "name", (void **)&name, &len);
718 if (len == 0 || name[len - 1] != '\0') {
722 if (len > MAXHOSTNAMELEN) {
723 error = ENAMETOOLONG;
728 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
736 if (len == 0 || host[len - 1] != '\0') {
740 if (len > MAXHOSTNAMELEN) {
741 error = ENAMETOOLONG;
746 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
754 if (len == 0 || domain[len - 1] != '\0') {
758 if (len > MAXHOSTNAMELEN) {
759 error = ENAMETOOLONG;
764 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
772 if (len == 0 || uuid[len - 1] != '\0') {
776 if (len > HOSTUUIDLEN) {
777 error = ENAMETOOLONG;
782 #ifdef COMPAT_FREEBSD32
783 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
786 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
790 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
802 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
807 else if (ip4s & (sizeof(*ip4) - 1)) {
811 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
813 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
815 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
816 ip4s /= sizeof(*ip4);
817 if (ip4s > jail_max_af_ips) {
819 vfs_opterror(opts, "too many IPv4 addresses");
822 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
823 bcopy(op, ip4, ip4s * sizeof(*ip4));
825 * IP addresses are all sorted but ip[0] to preserve
826 * the primary IP address as given from userland.
827 * This special IP is used for unbound outgoing
828 * connections as well for "loopback" traffic in case
829 * source address selection cannot find any more fitting
830 * address to connect from.
833 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
835 * Check for duplicate addresses and do some simple
836 * zero and broadcast checks. If users give other bogus
837 * addresses it is their problem.
839 * We do not have to care about byte order for these
840 * checks so we will do them in NBO.
842 for (ii = 0; ii < ip4s; ii++) {
843 if (ip4[ii].s_addr == INADDR_ANY ||
844 ip4[ii].s_addr == INADDR_BROADCAST) {
849 (ip4[0].s_addr == ip4[ii+1].s_addr ||
850 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
860 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
865 else if (ip6s & (sizeof(*ip6) - 1)) {
869 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
871 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
873 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
874 ip6s /= sizeof(*ip6);
875 if (ip6s > jail_max_af_ips) {
877 vfs_opterror(opts, "too many IPv6 addresses");
880 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
881 bcopy(op, ip6, ip6s * sizeof(*ip6));
883 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
884 for (ii = 0; ii < ip6s; ii++) {
885 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
890 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
891 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
901 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
902 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
905 "vnet jails cannot have IP address restrictions");
910 fullpath_disabled = 0;
912 error = vfs_getopt(opts, "path", (void **)&path, &len);
918 if (flags & JAIL_UPDATE) {
921 "path cannot be changed after creation");
924 if (len == 0 || path[len - 1] != '\0') {
928 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
934 NDFREE(&nd, NDF_ONLY_PNBUF);
935 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
936 strlcpy(g_path, path, MAXPATHLEN);
937 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
940 else if (error == ENODEV) {
941 /* proceed if sysctl debug.disablefullpath == 1 */
942 fullpath_disabled = 1;
943 if (len < 2 || (len == 2 && path[0] == '/'))
946 /* exit on other errors */
949 if (root->v_type != VDIR) {
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 if (!(flags & JAIL_ATTACH))
1817 sx_sunlock(&allprison_lock);
1818 prison_racct_modify(pr);
1819 if (!(flags & JAIL_ATTACH))
1820 sx_slock(&allprison_lock);
1824 td->td_retval[0] = pr->pr_id;
1827 * Now that it is all there, drop the temporary reference from existing
1828 * prisons. Or add a reference to newly created persistent prisons
1829 * (which was not done earlier so that the prison would not be publicly
1833 prison_deref(pr, (flags & JAIL_ATTACH)
1835 : PD_DEREF | PD_LIST_SLOCKED);
1837 if (pr_flags & PR_PERSIST) {
1838 mtx_lock(&pr->pr_mtx);
1841 mtx_unlock(&pr->pr_mtx);
1843 if (!(flags & JAIL_ATTACH))
1844 sx_sunlock(&allprison_lock);
1850 prison_deref(pr, created
1851 ? PD_LOCKED | PD_LIST_XLOCKED
1852 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1855 sx_xunlock(&allprison_lock);
1861 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1862 if (errmsg_len > 0) {
1863 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1864 if (errmsg_pos > 0) {
1865 if (optuio->uio_segflg == UIO_SYSSPACE)
1867 optuio->uio_iov[errmsg_pos].iov_base,
1871 optuio->uio_iov[errmsg_pos].iov_base,
1878 free(ip4, M_PRISON);
1881 free(ip6, M_PRISON);
1884 free(g_path, M_TEMP);
1891 * struct jail_get_args {
1892 * struct iovec *iovp;
1893 * unsigned int iovcnt;
1898 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1903 /* Check that we have an even number of iovecs. */
1904 if (uap->iovcnt & 1)
1907 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1910 error = kern_jail_get(td, auio, uap->flags);
1912 error = copyout(auio->uio_iov, uap->iovp,
1913 uap->iovcnt * sizeof (struct iovec));
1919 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1921 struct prison *pr, *mypr;
1923 struct vfsoptlist *opts;
1924 char *errmsg, *name;
1925 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1927 if (flags & ~JAIL_GET_MASK)
1930 /* Get the parameter list. */
1931 error = vfs_buildopts(optuio, &opts);
1934 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1935 mypr = td->td_ucred->cr_prison;
1938 * Find the prison specified by one of: lastjid, jid, name.
1940 sx_slock(&allprison_lock);
1941 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1943 TAILQ_FOREACH(pr, &allprison, pr_list) {
1944 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1945 mtx_lock(&pr->pr_mtx);
1946 if (pr->pr_ref > 0 &&
1947 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1949 mtx_unlock(&pr->pr_mtx);
1955 vfs_opterror(opts, "no jail after %d", jid);
1956 goto done_unlock_list;
1957 } else if (error != ENOENT)
1958 goto done_unlock_list;
1960 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1963 pr = prison_find_child(mypr, jid);
1965 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1966 mtx_unlock(&pr->pr_mtx);
1968 vfs_opterror(opts, "jail %d is dying",
1970 goto done_unlock_list;
1975 vfs_opterror(opts, "jail %d not found", jid);
1976 goto done_unlock_list;
1978 } else if (error != ENOENT)
1979 goto done_unlock_list;
1981 error = vfs_getopt(opts, "name", (void **)&name, &len);
1983 if (len == 0 || name[len - 1] != '\0') {
1985 goto done_unlock_list;
1987 pr = prison_find_name(mypr, name);
1989 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1990 mtx_unlock(&pr->pr_mtx);
1992 vfs_opterror(opts, "jail \"%s\" is dying",
1994 goto done_unlock_list;
1999 vfs_opterror(opts, "jail \"%s\" not found", name);
2000 goto done_unlock_list;
2001 } else if (error != ENOENT)
2002 goto done_unlock_list;
2004 vfs_opterror(opts, "no jail specified");
2006 goto done_unlock_list;
2009 /* Get the parameters of the prison. */
2012 td->td_retval[0] = pr->pr_id;
2013 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2014 if (error != 0 && error != ENOENT)
2016 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2017 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2018 if (error != 0 && error != ENOENT)
2020 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2021 if (error != 0 && error != ENOENT)
2023 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2024 sizeof(pr->pr_cpuset->cs_id));
2025 if (error != 0 && error != ENOENT)
2027 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2028 if (error != 0 && error != ENOENT)
2031 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2032 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2033 if (error != 0 && error != ENOENT)
2037 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2038 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2039 if (error != 0 && error != ENOENT)
2042 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2043 sizeof(pr->pr_securelevel));
2044 if (error != 0 && error != ENOENT)
2046 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2047 sizeof(pr->pr_childcount));
2048 if (error != 0 && error != ENOENT)
2050 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2051 sizeof(pr->pr_childmax));
2052 if (error != 0 && error != ENOENT)
2054 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2055 if (error != 0 && error != ENOENT)
2057 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2058 if (error != 0 && error != ENOENT)
2060 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2061 if (error != 0 && error != ENOENT)
2063 #ifdef COMPAT_FREEBSD32
2064 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2065 uint32_t hid32 = pr->pr_hostid;
2067 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2070 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2071 sizeof(pr->pr_hostid));
2072 if (error != 0 && error != ENOENT)
2074 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2075 sizeof(pr->pr_enforce_statfs));
2076 if (error != 0 && error != ENOENT)
2078 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2079 sizeof(pr->pr_devfs_rsnum));
2080 if (error != 0 && error != ENOENT)
2082 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2084 if (pr_flag_names[fi] == NULL)
2086 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2087 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2088 if (error != 0 && error != ENOENT)
2091 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2092 if (error != 0 && error != ENOENT)
2095 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2098 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2099 i = pr_flag_jailsys[fi].disable &&
2100 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2101 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2104 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2105 if (error != 0 && error != ENOENT)
2108 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2110 if (pr_allow_names[fi] == NULL)
2112 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2113 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2114 if (error != 0 && error != ENOENT)
2117 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2118 if (error != 0 && error != ENOENT)
2121 i = (pr->pr_uref == 0);
2122 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2123 if (error != 0 && error != ENOENT)
2126 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2127 if (error != 0 && error != ENOENT)
2130 /* Get the module parameters. */
2131 mtx_unlock(&pr->pr_mtx);
2133 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2136 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2138 /* By now, all parameters should have been noted. */
2139 TAILQ_FOREACH(opt, opts, link) {
2140 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2142 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2147 /* Write the fetched parameters back to userspace. */
2149 TAILQ_FOREACH(opt, opts, link) {
2150 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2151 pos = 2 * opt->pos + 1;
2152 optuio->uio_iov[pos].iov_len = opt->len;
2153 if (opt->value != NULL) {
2154 if (optuio->uio_segflg == UIO_SYSSPACE) {
2156 optuio->uio_iov[pos].iov_base,
2159 error = copyout(opt->value,
2160 optuio->uio_iov[pos].iov_base,
2171 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2175 sx_sunlock(&allprison_lock);
2177 if (error && errmsg_pos >= 0) {
2178 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2179 errmsg_pos = 2 * errmsg_pos + 1;
2180 if (errmsg_len > 0) {
2181 if (optuio->uio_segflg == UIO_SYSSPACE)
2183 optuio->uio_iov[errmsg_pos].iov_base,
2187 optuio->uio_iov[errmsg_pos].iov_base,
2197 * struct jail_remove_args {
2202 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2204 struct prison *pr, *cpr, *lpr, *tpr;
2207 error = priv_check(td, PRIV_JAIL_REMOVE);
2211 sx_xlock(&allprison_lock);
2212 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2214 sx_xunlock(&allprison_lock);
2218 /* Remove all descendants of this prison, then remove this prison. */
2220 pr->pr_flags |= PR_REMOVE;
2221 if (!LIST_EMPTY(&pr->pr_children)) {
2222 mtx_unlock(&pr->pr_mtx);
2224 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2225 mtx_lock(&cpr->pr_mtx);
2226 if (cpr->pr_ref > 0) {
2229 cpr->pr_flags |= PR_REMOVE;
2231 /* Already removed - do not do it again. */
2234 mtx_unlock(&cpr->pr_mtx);
2236 mtx_lock(&lpr->pr_mtx);
2237 prison_remove_one(lpr);
2238 sx_xlock(&allprison_lock);
2243 mtx_lock(&lpr->pr_mtx);
2244 prison_remove_one(lpr);
2245 sx_xlock(&allprison_lock);
2247 mtx_lock(&pr->pr_mtx);
2249 prison_remove_one(pr);
2254 prison_remove_one(struct prison *pr)
2259 /* If the prison was persistent, it is not anymore. */
2261 if (pr->pr_flags & PR_PERSIST) {
2264 pr->pr_flags &= ~PR_PERSIST;
2268 * jail_remove added a reference. If that's the only one, remove
2271 KASSERT(pr->pr_ref > 0,
2272 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2273 if (pr->pr_ref == 1) {
2275 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2279 mtx_unlock(&pr->pr_mtx);
2280 sx_xunlock(&allprison_lock);
2282 * Kill all processes unfortunate enough to be attached to this prison.
2284 sx_slock(&allproc_lock);
2285 LIST_FOREACH(p, &allproc, p_list) {
2287 if (p->p_state != PRS_NEW && p->p_ucred &&
2288 p->p_ucred->cr_prison == pr)
2289 kern_psignal(p, SIGKILL);
2292 sx_sunlock(&allproc_lock);
2293 /* Remove the temporary reference added by jail_remove. */
2294 prison_deref(pr, deuref | PD_DEREF);
2299 * struct jail_attach_args {
2304 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2309 error = priv_check(td, PRIV_JAIL_ATTACH);
2313 sx_slock(&allprison_lock);
2314 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2316 sx_sunlock(&allprison_lock);
2321 * Do not allow a process to attach to a prison that is not
2322 * considered to be "alive".
2324 if (pr->pr_uref == 0) {
2325 mtx_unlock(&pr->pr_mtx);
2326 sx_sunlock(&allprison_lock);
2330 return (do_jail_attach(td, pr));
2334 do_jail_attach(struct thread *td, struct prison *pr)
2338 struct ucred *newcred, *oldcred;
2342 * XXX: Note that there is a slight race here if two threads
2343 * in the same privileged process attempt to attach to two
2344 * different jails at the same time. It is important for
2345 * user processes not to do this, or they might end up with
2346 * a process root from one prison, but attached to the jail
2351 mtx_unlock(&pr->pr_mtx);
2353 /* Let modules do whatever they need to prepare for attaching. */
2354 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2356 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2359 sx_sunlock(&allprison_lock);
2362 * Reparent the newly attached process to this jail.
2364 ppr = td->td_ucred->cr_prison;
2366 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2370 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2371 if ((error = change_dir(pr->pr_root, td)) != 0)
2374 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2377 VOP_UNLOCK(pr->pr_root, 0);
2378 if ((error = change_root(pr->pr_root, td)))
2383 oldcred = p->p_ucred;
2385 crcopy(newcred, oldcred);
2386 newcred->cr_prison = pr;
2387 p->p_ucred = newcred;
2390 racct_proc_ucred_changed(p, oldcred, newcred);
2393 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2396 VOP_UNLOCK(pr->pr_root, 0);
2398 /* Tell modules this thread is still in its old jail after all. */
2399 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2400 prison_deref(pr, PD_DEREF | PD_DEUREF);
2406 * Returns a locked prison instance, or NULL on failure.
2409 prison_find(int prid)
2413 sx_assert(&allprison_lock, SX_LOCKED);
2414 TAILQ_FOREACH(pr, &allprison, pr_list) {
2415 if (pr->pr_id == prid) {
2416 mtx_lock(&pr->pr_mtx);
2419 mtx_unlock(&pr->pr_mtx);
2426 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2429 prison_find_child(struct prison *mypr, int prid)
2434 sx_assert(&allprison_lock, SX_LOCKED);
2435 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2436 if (pr->pr_id == prid) {
2437 mtx_lock(&pr->pr_mtx);
2440 mtx_unlock(&pr->pr_mtx);
2447 * Look for the name relative to mypr. Returns a locked prison or NULL.
2450 prison_find_name(struct prison *mypr, const char *name)
2452 struct prison *pr, *deadpr;
2456 sx_assert(&allprison_lock, SX_LOCKED);
2457 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2460 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2461 if (!strcmp(pr->pr_name + mylen, name)) {
2462 mtx_lock(&pr->pr_mtx);
2463 if (pr->pr_ref > 0) {
2464 if (pr->pr_uref > 0)
2468 mtx_unlock(&pr->pr_mtx);
2471 /* There was no valid prison - perhaps there was a dying one. */
2472 if (deadpr != NULL) {
2473 mtx_lock(&deadpr->pr_mtx);
2474 if (deadpr->pr_ref == 0) {
2475 mtx_unlock(&deadpr->pr_mtx);
2483 * See if a prison has the specific flag set.
2486 prison_flag(struct ucred *cred, unsigned flag)
2489 /* This is an atomic read, so no locking is necessary. */
2490 return (cred->cr_prison->pr_flags & flag);
2494 prison_allow(struct ucred *cred, unsigned flag)
2497 /* This is an atomic read, so no locking is necessary. */
2498 return (cred->cr_prison->pr_allow & flag);
2502 * Remove a prison reference. If that was the last reference, remove the
2503 * prison itself - but not in this context in case there are locks held.
2506 prison_free_locked(struct prison *pr)
2509 mtx_assert(&pr->pr_mtx, MA_OWNED);
2511 if (pr->pr_ref == 0) {
2512 mtx_unlock(&pr->pr_mtx);
2513 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2514 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2517 mtx_unlock(&pr->pr_mtx);
2521 prison_free(struct prison *pr)
2524 mtx_lock(&pr->pr_mtx);
2525 prison_free_locked(pr);
2529 prison_complete(void *context, int pending)
2532 prison_deref((struct prison *)context, 0);
2536 * Remove a prison reference (usually). This internal version assumes no
2537 * mutexes are held, except perhaps the prison itself. If there are no more
2538 * references, release and delist the prison. On completion, the prison lock
2539 * and the allprison lock are both unlocked.
2542 prison_deref(struct prison *pr, int flags)
2544 struct prison *ppr, *tpr;
2546 if (!(flags & PD_LOCKED))
2547 mtx_lock(&pr->pr_mtx);
2549 if (flags & PD_DEUREF) {
2551 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2553 if (flags & PD_DEREF)
2555 /* If the prison still has references, nothing else to do. */
2556 if (pr->pr_ref > 0) {
2557 mtx_unlock(&pr->pr_mtx);
2558 if (flags & PD_LIST_SLOCKED)
2559 sx_sunlock(&allprison_lock);
2560 else if (flags & PD_LIST_XLOCKED)
2561 sx_xunlock(&allprison_lock);
2565 mtx_unlock(&pr->pr_mtx);
2566 if (flags & PD_LIST_SLOCKED) {
2567 if (!sx_try_upgrade(&allprison_lock)) {
2568 sx_sunlock(&allprison_lock);
2569 sx_xlock(&allprison_lock);
2571 } else if (!(flags & PD_LIST_XLOCKED))
2572 sx_xlock(&allprison_lock);
2574 TAILQ_REMOVE(&allprison, pr, pr_list);
2575 LIST_REMOVE(pr, pr_sibling);
2576 ppr = pr->pr_parent;
2577 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2578 tpr->pr_childcount--;
2579 sx_xunlock(&allprison_lock);
2582 if (pr->pr_vnet != ppr->pr_vnet)
2583 vnet_destroy(pr->pr_vnet);
2585 if (pr->pr_root != NULL)
2587 mtx_destroy(&pr->pr_mtx);
2589 free(pr->pr_ip4, M_PRISON);
2592 free(pr->pr_ip6, M_PRISON);
2594 if (pr->pr_cpuset != NULL)
2595 cpuset_rel(pr->pr_cpuset);
2598 prison_racct_detach(pr);
2602 /* Removing a prison frees a reference on its parent. */
2604 mtx_lock(&pr->pr_mtx);
2605 flags = PD_DEREF | PD_DEUREF;
2610 prison_hold_locked(struct prison *pr)
2613 mtx_assert(&pr->pr_mtx, MA_OWNED);
2614 KASSERT(pr->pr_ref > 0,
2615 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2620 prison_hold(struct prison *pr)
2623 mtx_lock(&pr->pr_mtx);
2624 prison_hold_locked(pr);
2625 mtx_unlock(&pr->pr_mtx);
2629 prison_proc_hold(struct prison *pr)
2632 mtx_lock(&pr->pr_mtx);
2633 KASSERT(pr->pr_uref > 0,
2634 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2636 mtx_unlock(&pr->pr_mtx);
2640 prison_proc_free(struct prison *pr)
2643 mtx_lock(&pr->pr_mtx);
2644 KASSERT(pr->pr_uref > 0,
2645 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2646 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2652 * Restrict a prison's IP address list with its parent's, possibly replacing
2653 * it. Return true if the replacement buffer was used (or would have been).
2656 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2661 ppr = pr->pr_parent;
2662 if (!(pr->pr_flags & PR_IP4_USER)) {
2663 /* This has no user settings, so just copy the parent's list. */
2664 if (pr->pr_ip4s < ppr->pr_ip4s) {
2666 * There's no room for the parent's list. Use the
2667 * new list buffer, which is assumed to be big enough
2668 * (if it was passed). If there's no buffer, try to
2672 if (newip4 == NULL) {
2673 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2674 M_PRISON, M_NOWAIT);
2678 if (newip4 != NULL) {
2679 bcopy(ppr->pr_ip4, newip4,
2680 ppr->pr_ip4s * sizeof(*newip4));
2681 free(pr->pr_ip4, M_PRISON);
2682 pr->pr_ip4 = newip4;
2683 pr->pr_ip4s = ppr->pr_ip4s;
2687 pr->pr_ip4s = ppr->pr_ip4s;
2688 if (pr->pr_ip4s > 0)
2689 bcopy(ppr->pr_ip4, pr->pr_ip4,
2690 pr->pr_ip4s * sizeof(*newip4));
2691 else if (pr->pr_ip4 != NULL) {
2692 free(pr->pr_ip4, M_PRISON);
2695 } else if (pr->pr_ip4s > 0) {
2696 /* Remove addresses that aren't in the parent. */
2697 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2698 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2700 if (ij < ppr->pr_ip4s)
2703 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2704 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2707 for (ij = 1; ii < pr->pr_ip4s; ) {
2708 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2712 switch (ij >= ppr->pr_ip4s ? -1 :
2713 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2715 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2716 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2727 if (pr->pr_ip4s == 0) {
2728 pr->pr_flags |= PR_IP4_DISABLE;
2729 free(pr->pr_ip4, M_PRISON);
2737 * Pass back primary IPv4 address of this jail.
2739 * If not restricted return success but do not alter the address. Caller has
2740 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2742 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2743 * Address returned in NBO.
2746 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2750 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2751 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2753 pr = cred->cr_prison;
2754 if (!(pr->pr_flags & PR_IP4))
2756 mtx_lock(&pr->pr_mtx);
2757 if (!(pr->pr_flags & PR_IP4)) {
2758 mtx_unlock(&pr->pr_mtx);
2761 if (pr->pr_ip4 == NULL) {
2762 mtx_unlock(&pr->pr_mtx);
2763 return (EAFNOSUPPORT);
2766 ia->s_addr = pr->pr_ip4[0].s_addr;
2767 mtx_unlock(&pr->pr_mtx);
2772 * Return 1 if we should do proper source address selection or are not jailed.
2773 * We will return 0 if we should bypass source address selection in favour
2774 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2776 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2779 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2785 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2786 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2791 pr = cred->cr_prison;
2792 if (pr->pr_flags & PR_IP4_SADDRSEL)
2795 lia.s_addr = INADDR_ANY;
2796 error = prison_get_ip4(cred, &lia);
2799 if (lia.s_addr == INADDR_ANY)
2802 ia->s_addr = lia.s_addr;
2807 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2810 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2817 * No need to lock since the PR_IP4_USER flag can't be altered for
2820 while (pr1 != &prison0 &&
2822 !(pr1->pr_flags & PR_VNET) &&
2824 !(pr1->pr_flags & PR_IP4_USER))
2825 pr1 = pr1->pr_parent;
2826 while (pr2 != &prison0 &&
2828 !(pr2->pr_flags & PR_VNET) &&
2830 !(pr2->pr_flags & PR_IP4_USER))
2831 pr2 = pr2->pr_parent;
2832 return (pr1 == pr2);
2836 * Make sure our (source) address is set to something meaningful to this
2839 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2840 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2841 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2844 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2850 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2851 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2853 pr = cred->cr_prison;
2854 if (!(pr->pr_flags & PR_IP4))
2856 mtx_lock(&pr->pr_mtx);
2857 if (!(pr->pr_flags & PR_IP4)) {
2858 mtx_unlock(&pr->pr_mtx);
2861 if (pr->pr_ip4 == NULL) {
2862 mtx_unlock(&pr->pr_mtx);
2863 return (EAFNOSUPPORT);
2866 ia0.s_addr = ntohl(ia->s_addr);
2867 if (ia0.s_addr == INADDR_LOOPBACK) {
2868 ia->s_addr = pr->pr_ip4[0].s_addr;
2869 mtx_unlock(&pr->pr_mtx);
2873 if (ia0.s_addr == INADDR_ANY) {
2875 * In case there is only 1 IPv4 address, bind directly.
2877 if (pr->pr_ip4s == 1)
2878 ia->s_addr = pr->pr_ip4[0].s_addr;
2879 mtx_unlock(&pr->pr_mtx);
2883 error = _prison_check_ip4(pr, ia);
2884 mtx_unlock(&pr->pr_mtx);
2889 * Rewrite destination address in case we will connect to loopback address.
2891 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2892 * Address passed in in NBO and returned in NBO.
2895 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2899 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2900 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2902 pr = cred->cr_prison;
2903 if (!(pr->pr_flags & PR_IP4))
2905 mtx_lock(&pr->pr_mtx);
2906 if (!(pr->pr_flags & PR_IP4)) {
2907 mtx_unlock(&pr->pr_mtx);
2910 if (pr->pr_ip4 == NULL) {
2911 mtx_unlock(&pr->pr_mtx);
2912 return (EAFNOSUPPORT);
2915 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2916 ia->s_addr = pr->pr_ip4[0].s_addr;
2917 mtx_unlock(&pr->pr_mtx);
2922 * Return success because nothing had to be changed.
2924 mtx_unlock(&pr->pr_mtx);
2929 * Check if given address belongs to the jail referenced by cred/prison.
2931 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2932 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2933 * doesn't allow IPv4. Address passed in in NBO.
2936 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2941 * Check the primary IP.
2943 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2947 * All the other IPs are sorted so we can do a binary search.
2950 z = pr->pr_ip4s - 2;
2953 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2962 return (EADDRNOTAVAIL);
2966 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2971 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2972 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2974 pr = cred->cr_prison;
2975 if (!(pr->pr_flags & PR_IP4))
2977 mtx_lock(&pr->pr_mtx);
2978 if (!(pr->pr_flags & PR_IP4)) {
2979 mtx_unlock(&pr->pr_mtx);
2982 if (pr->pr_ip4 == NULL) {
2983 mtx_unlock(&pr->pr_mtx);
2984 return (EAFNOSUPPORT);
2987 error = _prison_check_ip4(pr, ia);
2988 mtx_unlock(&pr->pr_mtx);
2995 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3000 ppr = pr->pr_parent;
3001 if (!(pr->pr_flags & PR_IP6_USER)) {
3002 /* This has no user settings, so just copy the parent's list. */
3003 if (pr->pr_ip6s < ppr->pr_ip6s) {
3005 * There's no room for the parent's list. Use the
3006 * new list buffer, which is assumed to be big enough
3007 * (if it was passed). If there's no buffer, try to
3011 if (newip6 == NULL) {
3012 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3013 M_PRISON, M_NOWAIT);
3017 if (newip6 != NULL) {
3018 bcopy(ppr->pr_ip6, newip6,
3019 ppr->pr_ip6s * sizeof(*newip6));
3020 free(pr->pr_ip6, M_PRISON);
3021 pr->pr_ip6 = newip6;
3022 pr->pr_ip6s = ppr->pr_ip6s;
3026 pr->pr_ip6s = ppr->pr_ip6s;
3027 if (pr->pr_ip6s > 0)
3028 bcopy(ppr->pr_ip6, pr->pr_ip6,
3029 pr->pr_ip6s * sizeof(*newip6));
3030 else if (pr->pr_ip6 != NULL) {
3031 free(pr->pr_ip6, M_PRISON);
3034 } else if (pr->pr_ip6s > 0) {
3035 /* Remove addresses that aren't in the parent. */
3036 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3037 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3040 if (ij < ppr->pr_ip6s)
3043 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3044 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3047 for (ij = 1; ii < pr->pr_ip6s; ) {
3048 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3053 switch (ij >= ppr->pr_ip6s ? -1 :
3054 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3056 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3057 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3068 if (pr->pr_ip6s == 0) {
3069 pr->pr_flags |= PR_IP6_DISABLE;
3070 free(pr->pr_ip6, M_PRISON);
3078 * Pass back primary IPv6 address for this jail.
3080 * If not restricted return success but do not alter the address. Caller has
3081 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3083 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3086 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3090 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3091 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3093 pr = cred->cr_prison;
3094 if (!(pr->pr_flags & PR_IP6))
3096 mtx_lock(&pr->pr_mtx);
3097 if (!(pr->pr_flags & PR_IP6)) {
3098 mtx_unlock(&pr->pr_mtx);
3101 if (pr->pr_ip6 == NULL) {
3102 mtx_unlock(&pr->pr_mtx);
3103 return (EAFNOSUPPORT);
3106 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3107 mtx_unlock(&pr->pr_mtx);
3112 * Return 1 if we should do proper source address selection or are not jailed.
3113 * We will return 0 if we should bypass source address selection in favour
3114 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3116 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3119 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3122 struct in6_addr lia6;
3125 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3126 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3131 pr = cred->cr_prison;
3132 if (pr->pr_flags & PR_IP6_SADDRSEL)
3136 error = prison_get_ip6(cred, &lia6);
3139 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3142 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3147 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3150 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3156 while (pr1 != &prison0 &&
3158 !(pr1->pr_flags & PR_VNET) &&
3160 !(pr1->pr_flags & PR_IP6_USER))
3161 pr1 = pr1->pr_parent;
3162 while (pr2 != &prison0 &&
3164 !(pr2->pr_flags & PR_VNET) &&
3166 !(pr2->pr_flags & PR_IP6_USER))
3167 pr2 = pr2->pr_parent;
3168 return (pr1 == pr2);
3172 * Make sure our (source) address is set to something meaningful to this jail.
3174 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3175 * when needed while binding.
3177 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3178 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3179 * doesn't allow IPv6.
3182 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3187 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3188 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3190 pr = cred->cr_prison;
3191 if (!(pr->pr_flags & PR_IP6))
3193 mtx_lock(&pr->pr_mtx);
3194 if (!(pr->pr_flags & PR_IP6)) {
3195 mtx_unlock(&pr->pr_mtx);
3198 if (pr->pr_ip6 == NULL) {
3199 mtx_unlock(&pr->pr_mtx);
3200 return (EAFNOSUPPORT);
3203 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3204 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3205 mtx_unlock(&pr->pr_mtx);
3209 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3211 * In case there is only 1 IPv6 address, and v6only is true,
3212 * then bind directly.
3214 if (v6only != 0 && pr->pr_ip6s == 1)
3215 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3216 mtx_unlock(&pr->pr_mtx);
3220 error = _prison_check_ip6(pr, ia6);
3221 mtx_unlock(&pr->pr_mtx);
3226 * Rewrite destination address in case we will connect to loopback address.
3228 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3231 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3235 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3236 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3238 pr = cred->cr_prison;
3239 if (!(pr->pr_flags & PR_IP6))
3241 mtx_lock(&pr->pr_mtx);
3242 if (!(pr->pr_flags & PR_IP6)) {
3243 mtx_unlock(&pr->pr_mtx);
3246 if (pr->pr_ip6 == NULL) {
3247 mtx_unlock(&pr->pr_mtx);
3248 return (EAFNOSUPPORT);
3251 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3252 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3253 mtx_unlock(&pr->pr_mtx);
3258 * Return success because nothing had to be changed.
3260 mtx_unlock(&pr->pr_mtx);
3265 * Check if given address belongs to the jail referenced by cred/prison.
3267 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3268 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3269 * doesn't allow IPv6.
3272 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3277 * Check the primary IP.
3279 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3283 * All the other IPs are sorted so we can do a binary search.
3286 z = pr->pr_ip6s - 2;
3289 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3298 return (EADDRNOTAVAIL);
3302 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3307 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3308 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3310 pr = cred->cr_prison;
3311 if (!(pr->pr_flags & PR_IP6))
3313 mtx_lock(&pr->pr_mtx);
3314 if (!(pr->pr_flags & PR_IP6)) {
3315 mtx_unlock(&pr->pr_mtx);
3318 if (pr->pr_ip6 == NULL) {
3319 mtx_unlock(&pr->pr_mtx);
3320 return (EAFNOSUPPORT);
3323 error = _prison_check_ip6(pr, ia6);
3324 mtx_unlock(&pr->pr_mtx);
3330 * Check if a jail supports the given address family.
3332 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3336 prison_check_af(struct ucred *cred, int af)
3341 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3343 pr = cred->cr_prison;
3345 /* Prisons with their own network stack are not limited. */
3346 if (prison_owns_vnet(cred))
3355 if (pr->pr_flags & PR_IP4)
3357 mtx_lock(&pr->pr_mtx);
3358 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3359 error = EAFNOSUPPORT;
3360 mtx_unlock(&pr->pr_mtx);
3366 if (pr->pr_flags & PR_IP6)
3368 mtx_lock(&pr->pr_mtx);
3369 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3370 error = EAFNOSUPPORT;
3371 mtx_unlock(&pr->pr_mtx);
3379 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3380 error = EAFNOSUPPORT;
3386 * Check if given address belongs to the jail referenced by cred (wrapper to
3387 * prison_check_ip[46]).
3389 * Returns 0 if jail doesn't restrict the address family or if address belongs
3390 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3391 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3394 prison_if(struct ucred *cred, struct sockaddr *sa)
3397 struct sockaddr_in *sai;
3400 struct sockaddr_in6 *sai6;
3404 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3405 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3408 if (prison_owns_vnet(cred))
3413 switch (sa->sa_family)
3417 sai = (struct sockaddr_in *)sa;
3418 error = prison_check_ip4(cred, &sai->sin_addr);
3423 sai6 = (struct sockaddr_in6 *)sa;
3424 error = prison_check_ip6(cred, &sai6->sin6_addr);
3428 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3429 error = EAFNOSUPPORT;
3435 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3438 prison_check(struct ucred *cred1, struct ucred *cred2)
3441 return ((cred1->cr_prison == cred2->cr_prison ||
3442 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3446 * Return 1 if p2 is a child of p1, otherwise 0.
3449 prison_ischild(struct prison *pr1, struct prison *pr2)
3452 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3459 * Return 1 if the passed credential is in a jail, otherwise 0.
3462 jailed(struct ucred *cred)
3465 return (cred->cr_prison != &prison0);
3469 * Return 1 if the passed credential is in a jail and that jail does not
3470 * have its own virtual network stack, otherwise 0.
3473 jailed_without_vnet(struct ucred *cred)
3479 if (prison_owns_vnet(cred))
3487 * Return the correct hostname (domainname, et al) for the passed credential.
3490 getcredhostname(struct ucred *cred, char *buf, size_t size)
3495 * A NULL credential can be used to shortcut to the physical
3496 * system's hostname.
3498 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3499 mtx_lock(&pr->pr_mtx);
3500 strlcpy(buf, pr->pr_hostname, size);
3501 mtx_unlock(&pr->pr_mtx);
3505 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3508 mtx_lock(&cred->cr_prison->pr_mtx);
3509 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3510 mtx_unlock(&cred->cr_prison->pr_mtx);
3514 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3517 mtx_lock(&cred->cr_prison->pr_mtx);
3518 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3519 mtx_unlock(&cred->cr_prison->pr_mtx);
3523 getcredhostid(struct ucred *cred, unsigned long *hostid)
3526 mtx_lock(&cred->cr_prison->pr_mtx);
3527 *hostid = cred->cr_prison->pr_hostid;
3528 mtx_unlock(&cred->cr_prison->pr_mtx);
3533 * Determine whether the prison represented by cred owns
3534 * its vnet rather than having it inherited.
3536 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3539 prison_owns_vnet(struct ucred *cred)
3543 * vnets cannot be added/removed after jail creation,
3544 * so no need to lock here.
3546 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3551 * Determine whether the subject represented by cred can "see"
3552 * status of a mount point.
3553 * Returns: 0 for permitted, ENOENT otherwise.
3554 * XXX: This function should be called cr_canseemount() and should be
3555 * placed in kern_prot.c.
3558 prison_canseemount(struct ucred *cred, struct mount *mp)
3564 pr = cred->cr_prison;
3565 if (pr->pr_enforce_statfs == 0)
3567 if (pr->pr_root->v_mount == mp)
3569 if (pr->pr_enforce_statfs == 2)
3572 * If jail's chroot directory is set to "/" we should be able to see
3573 * all mount-points from inside a jail.
3574 * This is ugly check, but this is the only situation when jail's
3575 * directory ends with '/'.
3577 if (strcmp(pr->pr_path, "/") == 0)
3579 len = strlen(pr->pr_path);
3581 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3584 * Be sure that we don't have situation where jail's root directory
3585 * is "/some/path" and mount point is "/some/pathpath".
3587 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3593 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3595 char jpath[MAXPATHLEN];
3599 pr = cred->cr_prison;
3600 if (pr->pr_enforce_statfs == 0)
3602 if (prison_canseemount(cred, mp) != 0) {
3603 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3604 strlcpy(sp->f_mntonname, "[restricted]",
3605 sizeof(sp->f_mntonname));
3608 if (pr->pr_root->v_mount == mp) {
3610 * Clear current buffer data, so we are sure nothing from
3611 * the valid path left there.
3613 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3614 *sp->f_mntonname = '/';
3618 * If jail's chroot directory is set to "/" we should be able to see
3619 * all mount-points from inside a jail.
3621 if (strcmp(pr->pr_path, "/") == 0)
3623 len = strlen(pr->pr_path);
3624 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3626 * Clear current buffer data, so we are sure nothing from
3627 * the valid path left there.
3629 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3630 if (*jpath == '\0') {
3631 /* Should never happen. */
3632 *sp->f_mntonname = '/';
3634 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3639 * Check with permission for a specific privilege is granted within jail. We
3640 * have a specific list of accepted privileges; the rest are denied.
3643 prison_priv_check(struct ucred *cred, int priv)
3651 * Privileges specific to prisons with a virtual network stack.
3652 * There might be a duplicate entry here in case the privilege
3653 * is only granted conditionally in the legacy jail case.
3658 * NFS-specific privileges.
3660 case PRIV_NFS_DAEMON:
3661 case PRIV_NFS_LOCKD:
3664 * Network stack privileges.
3666 case PRIV_NET_BRIDGE:
3669 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3670 case PRIV_NET_ROUTE:
3672 case PRIV_NET_SETIFMTU:
3673 case PRIV_NET_SETIFFLAGS:
3674 case PRIV_NET_SETIFCAP:
3675 case PRIV_NET_SETIFDESCR:
3676 case PRIV_NET_SETIFNAME :
3677 case PRIV_NET_SETIFMETRIC:
3678 case PRIV_NET_SETIFPHYS:
3679 case PRIV_NET_SETIFMAC:
3680 case PRIV_NET_ADDMULTI:
3681 case PRIV_NET_DELMULTI:
3682 case PRIV_NET_HWIOCTL:
3683 case PRIV_NET_SETLLADDR:
3684 case PRIV_NET_ADDIFGROUP:
3685 case PRIV_NET_DELIFGROUP:
3686 case PRIV_NET_IFCREATE:
3687 case PRIV_NET_IFDESTROY:
3688 case PRIV_NET_ADDIFADDR:
3689 case PRIV_NET_DELIFADDR:
3692 case PRIV_NET_SETIFVNET:
3693 case PRIV_NET_SETIFFIB:
3696 * 802.11-related privileges.
3698 case PRIV_NET80211_GETKEY:
3700 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3705 * AppleTalk privileges.
3707 case PRIV_NETATALK_RESERVEDPORT:
3712 case PRIV_NETATM_CFG:
3713 case PRIV_NETATM_ADD:
3714 case PRIV_NETATM_DEL:
3715 case PRIV_NETATM_SET:
3718 * Bluetooth privileges.
3720 case PRIV_NETBLUETOOTH_RAW:
3724 * Netgraph and netgraph module privileges.
3726 case PRIV_NETGRAPH_CONTROL:
3728 case PRIV_NETGRAPH_TTY:
3732 * IPv4 and IPv6 privileges.
3734 case PRIV_NETINET_IPFW:
3735 case PRIV_NETINET_DIVERT:
3736 case PRIV_NETINET_PF:
3737 case PRIV_NETINET_DUMMYNET:
3738 case PRIV_NETINET_CARP:
3739 case PRIV_NETINET_MROUTE:
3740 case PRIV_NETINET_RAW:
3741 case PRIV_NETINET_ADDRCTRL6:
3742 case PRIV_NETINET_ND6:
3743 case PRIV_NETINET_SCOPE6:
3744 case PRIV_NETINET_ALIFETIME6:
3745 case PRIV_NETINET_IPSEC:
3746 case PRIV_NETINET_BINDANY:
3750 * IPX/SPX privileges.
3752 case PRIV_NETIPX_RESERVEDPORT:
3753 case PRIV_NETIPX_RAW:
3767 * No default: or deny here.
3768 * In case of no permit fall through to next switch().
3770 if (cred->cr_prison->pr_flags & PR_VNET)
3778 * Allow ktrace privileges for root in jail.
3784 * Allow jailed processes to configure audit identity and
3785 * submit audit records (login, etc). In the future we may
3786 * want to further refine the relationship between audit and
3789 case PRIV_AUDIT_GETAUDIT:
3790 case PRIV_AUDIT_SETAUDIT:
3791 case PRIV_AUDIT_SUBMIT:
3795 * Allow jailed processes to manipulate process UNIX
3796 * credentials in any way they see fit.
3798 case PRIV_CRED_SETUID:
3799 case PRIV_CRED_SETEUID:
3800 case PRIV_CRED_SETGID:
3801 case PRIV_CRED_SETEGID:
3802 case PRIV_CRED_SETGROUPS:
3803 case PRIV_CRED_SETREUID:
3804 case PRIV_CRED_SETREGID:
3805 case PRIV_CRED_SETRESUID:
3806 case PRIV_CRED_SETRESGID:
3809 * Jail implements visibility constraints already, so allow
3810 * jailed root to override uid/gid-based constraints.
3812 case PRIV_SEEOTHERGIDS:
3813 case PRIV_SEEOTHERUIDS:
3816 * Jail implements inter-process debugging limits already, so
3817 * allow jailed root various debugging privileges.
3819 case PRIV_DEBUG_DIFFCRED:
3820 case PRIV_DEBUG_SUGID:
3821 case PRIV_DEBUG_UNPRIV:
3824 * Allow jail to set various resource limits and login
3825 * properties, and for now, exceed process resource limits.
3827 case PRIV_PROC_LIMIT:
3828 case PRIV_PROC_SETLOGIN:
3829 case PRIV_PROC_SETRLIMIT:
3832 * System V and POSIX IPC privileges are granted in jail.
3835 case PRIV_IPC_WRITE:
3836 case PRIV_IPC_ADMIN:
3837 case PRIV_IPC_MSGSIZE:
3841 * Jail operations within a jail work on child jails.
3843 case PRIV_JAIL_ATTACH:
3845 case PRIV_JAIL_REMOVE:
3848 * Jail implements its own inter-process limits, so allow
3849 * root processes in jail to change scheduling on other
3850 * processes in the same jail. Likewise for signalling.
3852 case PRIV_SCHED_DIFFCRED:
3853 case PRIV_SCHED_CPUSET:
3854 case PRIV_SIGNAL_DIFFCRED:
3855 case PRIV_SIGNAL_SUGID:
3858 * Allow jailed processes to write to sysctls marked as jail
3861 case PRIV_SYSCTL_WRITEJAIL:
3864 * Allow root in jail to manage a variety of quota
3865 * properties. These should likely be conditional on a
3866 * configuration option.
3868 case PRIV_VFS_GETQUOTA:
3869 case PRIV_VFS_SETQUOTA:
3872 * Since Jail relies on chroot() to implement file system
3873 * protections, grant many VFS privileges to root in jail.
3874 * Be careful to exclude mount-related and NFS-related
3878 case PRIV_VFS_WRITE:
3879 case PRIV_VFS_ADMIN:
3881 case PRIV_VFS_LOOKUP:
3882 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3883 case PRIV_VFS_CHFLAGS_DEV:
3884 case PRIV_VFS_CHOWN:
3885 case PRIV_VFS_CHROOT:
3886 case PRIV_VFS_RETAINSUGID:
3887 case PRIV_VFS_FCHROOT:
3889 case PRIV_VFS_SETGID:
3891 case PRIV_VFS_STICKYFILE:
3894 * As in the non-jail case, non-root users are expected to be
3895 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3896 * exists in the jail and they have permission to access it).
3898 case PRIV_KMEM_READ:
3902 * Depending on the global setting, allow privilege of
3903 * setting system flags.
3905 case PRIV_VFS_SYSFLAGS:
3906 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3912 * Depending on the global setting, allow privilege of
3913 * mounting/unmounting file systems.
3915 case PRIV_VFS_MOUNT:
3916 case PRIV_VFS_UNMOUNT:
3917 case PRIV_VFS_MOUNT_NONUSER:
3918 case PRIV_VFS_MOUNT_OWNER:
3919 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3920 cred->cr_prison->pr_enforce_statfs < 2)
3926 * Allow jailed root to bind reserved ports and reuse in-use
3929 case PRIV_NETINET_RESERVEDPORT:
3930 case PRIV_NETINET_REUSEPORT:
3934 * Allow jailed root to set certian IPv4/6 (option) headers.
3936 case PRIV_NETINET_SETHDROPTS:
3940 * Conditionally allow creating raw sockets in jail.
3942 case PRIV_NETINET_RAW:
3943 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3949 * Since jail implements its own visibility limits on netstat
3950 * sysctls, allow getcred. This allows identd to work in
3953 case PRIV_NETINET_GETCRED:
3957 * Allow jailed root to set loginclass.
3959 case PRIV_PROC_SETLOGINCLASS:
3964 * In all remaining cases, deny the privilege request. This
3965 * includes almost all network privileges, many system
3966 * configuration privileges.
3973 * Return the part of pr2's name that is relative to pr1, or the whole name
3974 * if it does not directly follow.
3978 prison_name(struct prison *pr1, struct prison *pr2)
3982 /* Jails see themselves as "0" (if they see themselves at all). */
3985 name = pr2->pr_name;
3986 if (prison_ischild(pr1, pr2)) {
3988 * pr1 isn't locked (and allprison_lock may not be either)
3989 * so its length can't be counted on. But the number of dots
3990 * can be counted on - and counted.
3992 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3993 name = strchr(name, '.') + 1;
3999 * Return the part of pr2's path that is relative to pr1, or the whole path
4000 * if it does not directly follow.
4003 prison_path(struct prison *pr1, struct prison *pr2)
4005 char *path1, *path2;
4008 path1 = pr1->pr_path;
4009 path2 = pr2->pr_path;
4010 if (!strcmp(path1, "/"))
4012 len1 = strlen(path1);
4013 if (strncmp(path1, path2, len1))
4015 if (path2[len1] == '\0')
4017 if (path2[len1] == '/')
4018 return (path2 + len1);
4024 * Jail-related sysctls.
4026 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4030 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4033 struct prison *pr, *cpr;
4035 struct in_addr *ip4 = NULL;
4039 struct in6_addr *ip6 = NULL;
4044 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4045 pr = req->td->td_ucred->cr_prison;
4047 sx_slock(&allprison_lock);
4048 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4049 #if defined(INET) || defined(INET6)
4052 mtx_lock(&cpr->pr_mtx);
4054 if (cpr->pr_ip4s > 0) {
4055 if (ip4s < cpr->pr_ip4s) {
4056 ip4s = cpr->pr_ip4s;
4057 mtx_unlock(&cpr->pr_mtx);
4058 ip4 = realloc(ip4, ip4s *
4059 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4062 bcopy(cpr->pr_ip4, ip4,
4063 cpr->pr_ip4s * sizeof(struct in_addr));
4067 if (cpr->pr_ip6s > 0) {
4068 if (ip6s < cpr->pr_ip6s) {
4069 ip6s = cpr->pr_ip6s;
4070 mtx_unlock(&cpr->pr_mtx);
4071 ip6 = realloc(ip6, ip6s *
4072 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4075 bcopy(cpr->pr_ip6, ip6,
4076 cpr->pr_ip6s * sizeof(struct in6_addr));
4079 if (cpr->pr_ref == 0) {
4080 mtx_unlock(&cpr->pr_mtx);
4083 bzero(xp, sizeof(*xp));
4084 xp->pr_version = XPRISON_VERSION;
4085 xp->pr_id = cpr->pr_id;
4086 xp->pr_state = cpr->pr_uref > 0
4087 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4088 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4089 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4090 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4092 xp->pr_ip4s = cpr->pr_ip4s;
4095 xp->pr_ip6s = cpr->pr_ip6s;
4097 mtx_unlock(&cpr->pr_mtx);
4098 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4102 if (xp->pr_ip4s > 0) {
4103 error = SYSCTL_OUT(req, ip4,
4104 xp->pr_ip4s * sizeof(struct in_addr));
4110 if (xp->pr_ip6s > 0) {
4111 error = SYSCTL_OUT(req, ip6,
4112 xp->pr_ip6s * sizeof(struct in6_addr));
4118 sx_sunlock(&allprison_lock);
4129 SYSCTL_OID(_security_jail, OID_AUTO, list,
4130 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4131 sysctl_jail_list, "S", "List of active jails");
4134 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4138 injail = jailed(req->td->td_ucred);
4139 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4144 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4145 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4146 sysctl_jail_jailed, "I", "Process in jail?");
4149 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4151 int error, havevnet;
4153 struct ucred *cred = req->td->td_ucred;
4155 havevnet = jailed(cred) && prison_owns_vnet(cred);
4159 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4164 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4165 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4166 sysctl_jail_vnet, "I", "Jail owns VNET?");
4168 #if defined(INET) || defined(INET6)
4169 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4170 &jail_max_af_ips, 0,
4171 "Number of IP addresses a jail may have at most per address family");
4175 * Default parameters for jail(2) compatability. For historical reasons,
4176 * the sysctl names have varying similarity to the parameter names. Prisons
4177 * just see their own parameters, and can't change them.
4180 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4183 int allow, error, i;
4185 pr = req->td->td_ucred->cr_prison;
4186 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4188 /* Get the current flag value, and convert it to a boolean. */
4189 i = (allow & arg2) ? 1 : 0;
4192 error = sysctl_handle_int(oidp, &i, 0, req);
4193 if (error || !req->newptr)
4199 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4202 mtx_lock(&prison0.pr_mtx);
4203 jail_default_allow = (jail_default_allow & ~arg2) | i;
4204 mtx_unlock(&prison0.pr_mtx);
4208 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4209 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4210 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4211 "Processes in jail can set their hostnames");
4212 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4213 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4214 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4215 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4216 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4218 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4219 "Processes in jail can use System V IPC primitives");
4220 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4221 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4222 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4223 "Prison root can create raw sockets");
4224 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4225 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4226 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4227 "Processes in jail can alter system file flags");
4228 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4229 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4230 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4231 "Processes in jail can mount/unmount jail-friendly file systems");
4232 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4233 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4234 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4235 "Processes in jail can mount the devfs file system");
4236 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4237 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4238 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4239 "Processes in jail can mount the fdescfs file system");
4240 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4241 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4242 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4243 "Processes in jail can mount the nullfs file system");
4244 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4245 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4246 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4247 "Processes in jail can mount the procfs file system");
4248 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4249 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4250 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4251 "Processes in jail can mount the tmpfs file system");
4252 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4253 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4254 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4255 "Processes in jail can mount the zfs file system");
4258 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4263 pr = req->td->td_ucred->cr_prison;
4264 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4265 error = sysctl_handle_int(oidp, &level, 0, req);
4266 if (error || !req->newptr)
4268 *(int *)arg1 = level;
4272 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4273 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4274 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4275 sysctl_jail_default_level, "I",
4276 "Processes in jail cannot see all mounted file systems");
4278 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4279 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4280 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4281 sysctl_jail_default_level, "I",
4282 "Ruleset for the devfs filesystem in jail");
4285 * Nodes to describe jail parameters. Maximum length of string parameters
4286 * is returned in the string itself, and the other parameters exist merely
4287 * to make themselves and their types known.
4289 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4293 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4300 switch (oidp->oid_kind & CTLTYPE)
4306 if (!(req->flags & SCTL_MASK32))
4308 return (SYSCTL_OUT(req, &l, sizeof(l)));
4312 return (SYSCTL_OUT(req, &i, sizeof(i)));
4313 case CTLTYPE_STRING:
4314 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4316 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4317 case CTLTYPE_STRUCT:
4319 return (SYSCTL_OUT(req, &s, sizeof(s)));
4324 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4325 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4326 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4327 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4328 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4329 "I", "Jail secure level");
4330 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4331 "I", "Jail cannot see all mounted file systems");
4332 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4333 "I", "Ruleset for in-jail devfs mounts");
4334 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4335 "B", "Jail persistence");
4337 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4338 "E,jailsys", "Virtual network stack");
4340 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4341 "B", "Jail is in the process of shutting down");
4343 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4344 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4345 "I", "Current number of child jails");
4346 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4347 "I", "Maximum number of child jails");
4349 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4350 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4352 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4353 "Jail NIS domainname");
4354 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4356 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4357 "LU", "Jail host ID");
4359 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4360 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4363 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4364 "Jail IPv4 address virtualization");
4365 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4366 "S,in_addr,a", "Jail IPv4 addresses");
4367 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4368 "B", "Do (not) use IPv4 source address selection rather than the "
4369 "primary jail IPv4 address.");
4372 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4373 "Jail IPv6 address virtualization");
4374 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4375 "S,in6_addr,a", "Jail IPv6 addresses");
4376 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4377 "B", "Do (not) use IPv6 source address selection rather than the "
4378 "primary jail IPv6 address.");
4381 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4382 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4383 "B", "Jail may set hostname");
4384 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4385 "B", "Jail may use SYSV IPC");
4386 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4387 "B", "Jail may create raw sockets");
4388 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4389 "B", "Jail may alter system file flags");
4390 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4391 "B", "Jail may set file quotas");
4392 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4393 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4395 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4396 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4397 "B", "Jail may mount/unmount jail-friendly file systems in general");
4398 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4399 "B", "Jail may mount the devfs file system");
4400 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4401 "B", "Jail may mount the fdescfs file system");
4402 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4403 "B", "Jail may mount the nullfs file system");
4404 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4405 "B", "Jail may mount the procfs file system");
4406 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4407 "B", "Jail may mount the tmpfs file system");
4408 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4409 "B", "Jail may mount the zfs file system");
4412 prison_racct_foreach(void (*callback)(struct racct *racct,
4413 void *arg2, void *arg3), void *arg2, void *arg3)
4415 struct prison_racct *prr;
4417 sx_slock(&allprison_lock);
4418 LIST_FOREACH(prr, &allprison_racct, prr_next)
4419 (callback)(prr->prr_racct, arg2, arg3);
4420 sx_sunlock(&allprison_lock);
4423 static struct prison_racct *
4424 prison_racct_find_locked(const char *name)
4426 struct prison_racct *prr;
4428 sx_assert(&allprison_lock, SA_XLOCKED);
4430 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4433 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4434 if (strcmp(name, prr->prr_name) != 0)
4437 /* Found prison_racct with a matching name? */
4438 prison_racct_hold(prr);
4442 /* Add new prison_racct. */
4443 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4444 racct_create(&prr->prr_racct);
4446 strcpy(prr->prr_name, name);
4447 refcount_init(&prr->prr_refcount, 1);
4448 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4453 struct prison_racct *
4454 prison_racct_find(const char *name)
4456 struct prison_racct *prr;
4458 sx_xlock(&allprison_lock);
4459 prr = prison_racct_find_locked(name);
4460 sx_xunlock(&allprison_lock);
4465 prison_racct_hold(struct prison_racct *prr)
4468 refcount_acquire(&prr->prr_refcount);
4472 prison_racct_free_locked(struct prison_racct *prr)
4475 sx_assert(&allprison_lock, SA_XLOCKED);
4477 if (refcount_release(&prr->prr_refcount)) {
4478 racct_destroy(&prr->prr_racct);
4479 LIST_REMOVE(prr, prr_next);
4480 free(prr, M_PRISON_RACCT);
4485 prison_racct_free(struct prison_racct *prr)
4489 sx_assert(&allprison_lock, SA_UNLOCKED);
4491 old = prr->prr_refcount;
4492 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4495 sx_xlock(&allprison_lock);
4496 prison_racct_free_locked(prr);
4497 sx_xunlock(&allprison_lock);
4502 prison_racct_attach(struct prison *pr)
4504 struct prison_racct *prr;
4506 sx_assert(&allprison_lock, SA_XLOCKED);
4508 prr = prison_racct_find_locked(pr->pr_name);
4509 KASSERT(prr != NULL, ("cannot find prison_racct"));
4511 pr->pr_prison_racct = prr;
4515 * Handle jail renaming. From the racct point of view, renaming means
4516 * moving from one prison_racct to another.
4519 prison_racct_modify(struct prison *pr)
4523 struct prison_racct *oldprr;
4525 sx_slock(&allproc_lock);
4526 sx_xlock(&allprison_lock);
4528 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4529 sx_xunlock(&allprison_lock);
4530 sx_sunlock(&allproc_lock);
4534 oldprr = pr->pr_prison_racct;
4535 pr->pr_prison_racct = NULL;
4537 prison_racct_attach(pr);
4540 * Move resource utilisation records.
4542 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4545 * Force rctl to reattach rules to processes.
4547 FOREACH_PROC_IN_SYSTEM(p) {
4549 cred = crhold(p->p_ucred);
4551 racct_proc_ucred_changed(p, cred, cred);
4555 sx_sunlock(&allproc_lock);
4556 prison_racct_free_locked(oldprr);
4557 sx_xunlock(&allprison_lock);
4561 prison_racct_detach(struct prison *pr)
4564 sx_assert(&allprison_lock, SA_UNLOCKED);
4566 if (pr->pr_prison_racct == NULL)
4568 prison_racct_free(pr->pr_prison_racct);
4569 pr->pr_prison_racct = NULL;
4576 db_show_prison(struct prison *pr)
4579 #if defined(INET) || defined(INET6)
4584 char ip6buf[INET6_ADDRSTRLEN];
4587 db_printf("prison %p:\n", pr);
4588 db_printf(" jid = %d\n", pr->pr_id);
4589 db_printf(" name = %s\n", pr->pr_name);
4590 db_printf(" parent = %p\n", pr->pr_parent);
4591 db_printf(" ref = %d\n", pr->pr_ref);
4592 db_printf(" uref = %d\n", pr->pr_uref);
4593 db_printf(" path = %s\n", pr->pr_path);
4594 db_printf(" cpuset = %d\n", pr->pr_cpuset
4595 ? pr->pr_cpuset->cs_id : -1);
4597 db_printf(" vnet = %p\n", pr->pr_vnet);
4599 db_printf(" root = %p\n", pr->pr_root);
4600 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4601 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4602 db_printf(" children.max = %d\n", pr->pr_childmax);
4603 db_printf(" children.cur = %d\n", pr->pr_childcount);
4604 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4605 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4606 db_printf(" flags = 0x%x", pr->pr_flags);
4607 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4609 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4610 db_printf(" %s", pr_flag_names[fi]);
4611 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4613 jsf = pr->pr_flags &
4614 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4615 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4616 pr_flag_jailsys[fi].disable &&
4617 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4618 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4621 db_printf(" allow = 0x%x", pr->pr_allow);
4622 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4624 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4625 db_printf(" %s", pr_allow_names[fi]);
4627 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4628 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4629 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4630 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4631 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4633 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4634 for (ii = 0; ii < pr->pr_ip4s; ii++)
4635 db_printf(" %s %s\n",
4636 ii == 0 ? "ip4.addr =" : " ",
4637 inet_ntoa(pr->pr_ip4[ii]));
4640 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4641 for (ii = 0; ii < pr->pr_ip6s; ii++)
4642 db_printf(" %s %s\n",
4643 ii == 0 ? "ip6.addr =" : " ",
4644 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4648 DB_SHOW_COMMAND(prison, db_show_prison_command)
4654 * Show all prisons in the list, and prison0 which is not
4657 db_show_prison(&prison0);
4658 if (!db_pager_quit) {
4659 TAILQ_FOREACH(pr, &allprison, pr_list) {
4671 /* Look for a prison with the ID and with references. */
4672 TAILQ_FOREACH(pr, &allprison, pr_list)
4673 if (pr->pr_id == addr && pr->pr_ref > 0)
4676 /* Look again, without requiring a reference. */
4677 TAILQ_FOREACH(pr, &allprison, pr_list)
4678 if (pr->pr_id == addr)
4681 /* Assume address points to a valid prison. */
4682 pr = (struct prison *)addr;