2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 1999 Poul-Henning Kamp.
5 * Copyright (c) 2008 Bjoern A. Zeeb.
6 * Copyright (c) 2009 James Gritton.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
38 #include <sys/param.h>
39 #include <sys/types.h>
40 #include <sys/kernel.h>
41 #include <sys/systm.h>
42 #include <sys/errno.h>
43 #include <sys/sysproto.h>
44 #include <sys/malloc.h>
48 #include <sys/taskqueue.h>
49 #include <sys/fcntl.h>
52 #include <sys/mutex.h>
53 #include <sys/racct.h>
55 #include <sys/refcount.h>
57 #include <sys/sysent.h>
58 #include <sys/namei.h>
59 #include <sys/mount.h>
60 #include <sys/queue.h>
61 #include <sys/socket.h>
62 #include <sys/syscallsubr.h>
63 #include <sys/sysctl.h>
64 #include <sys/vnode.h>
69 #include <netinet/in.h>
75 #include <security/mac/mac_framework.h>
77 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
79 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
80 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
82 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
85 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
87 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
91 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
93 #define _PR_IP_SADDRSEL 0
97 /* prison0 describes what is "real" about the system. */
98 struct prison prison0 = {
104 .pr_securelevel = -1,
106 .pr_childmax = JAIL_MAX,
107 .pr_hostuuid = DEFAULT_HOSTUUID,
108 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
110 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
112 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
114 .pr_allow = PR_ALLOW_ALL,
116 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
123 struct jailsys_flags {
129 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
130 struct sx allprison_lock;
131 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
132 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
133 LIST_HEAD(, prison_racct) allprison_racct;
136 static int do_jail_attach(struct thread *td, struct prison *pr);
137 static void prison_complete(void *context, int pending);
138 static void prison_deref(struct prison *pr, int flags);
139 static char *prison_path(struct prison *pr1, struct prison *pr2);
140 static void prison_remove_one(struct prison *pr);
142 static void prison_racct_attach(struct prison *pr);
143 static void prison_racct_modify(struct prison *pr);
144 static void prison_racct_detach(struct prison *pr);
147 /* Flags for prison_deref */
148 #define PD_DEREF 0x01
149 #define PD_DEUREF 0x02
150 #define PD_LOCKED 0x04
151 #define PD_LIST_SLOCKED 0x08
152 #define PD_LIST_XLOCKED 0x10
155 * Parameter names corresponding to PR_* flag values. Size values are for kvm
156 * as we cannot figure out the size of a sparse array, or an array without a
159 static struct bool_flags pr_flag_bool[] = {
160 {"persist", "nopersist", PR_PERSIST},
162 {"ip4.saddrsel", "ip4.nosaddrsel", PR_IP4_SADDRSEL},
165 {"ip6.saddrsel", "ip6.nosaddrsel", PR_IP6_SADDRSEL},
168 const size_t pr_flag_bool_size = sizeof(pr_flag_bool);
170 static struct jailsys_flags pr_flag_jailsys[] = {
171 {"host", 0, PR_HOST},
173 {"vnet", 0, PR_VNET},
176 {"ip4", PR_IP4_USER, PR_IP4_USER},
179 {"ip6", PR_IP6_USER, PR_IP6_USER},
182 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
184 static struct bool_flags pr_flag_allow[] = {
185 {"allow.set_hostname", "allow.noset_hostname", PR_ALLOW_SET_HOSTNAME},
186 {"allow.sysvipc", "allow.nosysvipc", PR_ALLOW_SYSVIPC},
187 {"allow.raw_sockets", "allow.noraw_sockets", PR_ALLOW_RAW_SOCKETS},
188 {"allow.chflags", "allow.nochflags", PR_ALLOW_CHFLAGS},
189 {"allow.mount", "allow.nomount", PR_ALLOW_MOUNT},
190 {"allow.quotas", "allow.noquotas", PR_ALLOW_QUOTAS},
191 {"allow.socket_af", "allow.nosocket_af", PR_ALLOW_SOCKET_AF},
192 {"allow.mount.devfs", "allow.mount.nodevfs", PR_ALLOW_MOUNT_DEVFS},
193 {"allow.mount.nullfs", "allow.mount.nonullfs", PR_ALLOW_MOUNT_NULLFS},
194 {"allow.mount.zfs", "allow.mount.nozfs", PR_ALLOW_MOUNT_ZFS},
195 {"allow.mount.procfs", "allow.mount.noprocfs", PR_ALLOW_MOUNT_PROCFS},
196 {"allow.mount.tmpfs", "allow.mount.notmpfs", PR_ALLOW_MOUNT_TMPFS},
197 {"allow.mount.fdescfs", "allow.mount.nofdescfs",
198 PR_ALLOW_MOUNT_FDESCFS},
199 {"allow.mount.linprocfs", "allow.mount.nolinprocfs",
200 PR_ALLOW_MOUNT_LINPROCFS},
201 {"allow.mount.linsysfs", "allow.mount.nolinsysfs",
202 PR_ALLOW_MOUNT_LINSYSFS},
203 {"allow.reserved_ports", "allow.noreserved_ports",
204 PR_ALLOW_RESERVED_PORTS},
206 const size_t pr_flag_allow_size = sizeof(pr_flag_allow);
208 #define JAIL_DEFAULT_ALLOW (PR_ALLOW_SET_HOSTNAME | PR_ALLOW_RESERVED_PORTS)
209 #define JAIL_DEFAULT_ENFORCE_STATFS 2
210 #define JAIL_DEFAULT_DEVFS_RSNUM 0
211 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
212 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
213 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
214 #if defined(INET) || defined(INET6)
215 static unsigned jail_max_af_ips = 255;
219 * Initialize the parts of prison0 that can't be static-initialized with
220 * constants. This is called from proc0_init() after creating thread0 cpuset.
226 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
227 prison0.pr_osreldate = osreldate;
228 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
237 sys_jail(struct thread *td, struct jail_args *uap)
243 error = copyin(uap->jail, &version, sizeof(uint32_t));
252 /* FreeBSD single IPv4 jails. */
253 bzero(&j, sizeof(struct jail));
254 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
257 j.version = j0.version;
259 j.hostname = j0.hostname;
260 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
266 * Version 1 was used by multi-IPv4 jail implementations
267 * that never made it into the official kernel.
271 case 2: /* JAIL_API_VERSION */
272 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
273 error = copyin(uap->jail, &j, sizeof(struct jail));
279 /* Sci-Fi jails are not supported, sorry. */
282 return (kern_jail(td, &j));
286 kern_jail(struct thread *td, struct jail *j)
288 struct iovec optiov[2 * (4 + nitems(pr_flag_allow)
297 char *u_path, *u_hostname, *u_name;
298 struct bool_flags *bf;
301 struct in_addr *u_ip4;
304 struct in6_addr *u_ip6;
307 int error, enforce_statfs;
309 bzero(&optiov, sizeof(optiov));
310 opt.uio_iov = optiov;
314 opt.uio_segflg = UIO_SYSSPACE;
315 opt.uio_rw = UIO_READ;
318 /* Set permissions for top-level jails from sysctls. */
319 if (!jailed(td->td_ucred)) {
320 for (bf = pr_flag_allow;
321 bf < pr_flag_allow + nitems(pr_flag_allow);
323 optiov[opt.uio_iovcnt].iov_base = __DECONST(char *,
324 (jail_default_allow & bf->flag)
325 ? bf->name : bf->noname);
326 optiov[opt.uio_iovcnt].iov_len =
327 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
330 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
331 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
333 enforce_statfs = jail_default_enforce_statfs;
334 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
335 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
339 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
341 ip4s = (j->version == 0) ? 1 : j->ip4s;
342 if (ip4s > jail_max_af_ips)
344 tmplen += ip4s * sizeof(struct in_addr);
350 if (j->ip6s > jail_max_af_ips)
352 tmplen += j->ip6s * sizeof(struct in6_addr);
357 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
358 u_hostname = u_path + MAXPATHLEN;
359 u_name = u_hostname + MAXHOSTNAMELEN;
361 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
365 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
367 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
370 optiov[opt.uio_iovcnt].iov_base = "path";
371 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
373 optiov[opt.uio_iovcnt].iov_base = u_path;
374 error = copyinstr(j->path, u_path, MAXPATHLEN,
375 &optiov[opt.uio_iovcnt].iov_len);
377 free(u_path, M_TEMP);
381 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
382 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
384 optiov[opt.uio_iovcnt].iov_base = u_hostname;
385 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
386 &optiov[opt.uio_iovcnt].iov_len);
388 free(u_path, M_TEMP);
392 if (j->jailname != NULL) {
393 optiov[opt.uio_iovcnt].iov_base = "name";
394 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
396 optiov[opt.uio_iovcnt].iov_base = u_name;
397 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
398 &optiov[opt.uio_iovcnt].iov_len);
400 free(u_path, M_TEMP);
406 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
407 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
409 optiov[opt.uio_iovcnt].iov_base = u_ip4;
410 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
412 u_ip4->s_addr = j->ip4s;
414 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
416 free(u_path, M_TEMP);
423 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
424 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
426 optiov[opt.uio_iovcnt].iov_base = u_ip6;
427 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
428 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
430 free(u_path, M_TEMP);
435 KASSERT(opt.uio_iovcnt <= nitems(optiov),
436 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
437 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
438 free(u_path, M_TEMP);
444 * struct jail_set_args {
445 * struct iovec *iovp;
446 * unsigned int iovcnt;
451 sys_jail_set(struct thread *td, struct jail_set_args *uap)
456 /* Check that we have an even number of iovecs. */
460 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
463 error = kern_jail_set(td, auio, uap->flags);
469 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
476 struct in6_addr *ip6;
479 struct vfsoptlist *opts;
480 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
482 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
483 char *g_path, *osrelstr;
484 struct bool_flags *bf;
485 struct jailsys_flags *jsf;
486 #if defined(INET) || defined(INET6)
491 size_t namelen, onamelen, pnamelen;
492 int born, created, cuflags, descend, enforce;
493 int error, errmsg_len, errmsg_pos;
494 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
495 int jid, jsys, len, level;
496 int childmax, osreldt, rsnum, slevel;
497 int fullpath_disabled;
498 #if defined(INET) || defined(INET6)
507 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
511 error = priv_check(td, PRIV_JAIL_SET);
512 if (!error && (flags & JAIL_ATTACH))
513 error = priv_check(td, PRIV_JAIL_ATTACH);
516 mypr = td->td_ucred->cr_prison;
517 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
519 if (flags & ~JAIL_SET_MASK)
523 * Check all the parameters before committing to anything. Not all
524 * errors can be caught early, but we may as well try. Also, this
525 * takes care of some expensive stuff (path lookup) before getting
526 * the allprison lock.
528 * XXX Jails are not filesystems, and jail parameters are not mount
529 * options. But it makes more sense to re-use the vfsopt code
530 * than duplicate it under a different name.
532 error = vfs_buildopts(optuio, &opts);
543 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
546 vfs_opterror(opts, "no valid operation (create or update)");
550 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
556 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
565 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
573 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
578 else if (enforce < 0 || enforce > 2) {
584 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
592 pr_flags = ch_flags = 0;
593 for (bf = pr_flag_bool;
594 bf < pr_flag_bool + nitems(pr_flag_bool);
596 vfs_flagopt(opts, bf->name, &pr_flags, bf->flag);
597 vfs_flagopt(opts, bf->noname, &ch_flags, bf->flag);
599 ch_flags |= pr_flags;
600 for (jsf = pr_flag_jailsys;
601 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
603 error = vfs_copyopt(opts, jsf->name, &jsys, sizeof(jsys));
609 case JAIL_SYS_DISABLE:
614 pr_flags |= jsf->disable;
617 pr_flags |= jsf->new;
619 case JAIL_SYS_INHERIT:
625 ch_flags |= jsf->new | jsf->disable;
627 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
628 && !(pr_flags & PR_PERSIST)) {
630 vfs_opterror(opts, "new jail must persist or attach");
634 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
636 vfs_opterror(opts, "vnet cannot be changed after creation");
641 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
643 vfs_opterror(opts, "ip4 cannot be changed after creation");
648 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
650 vfs_opterror(opts, "ip6 cannot be changed after creation");
655 pr_allow = ch_allow = 0;
656 for (bf = pr_flag_allow;
657 bf < pr_flag_allow + nitems(pr_flag_allow);
659 vfs_flagopt(opts, bf->name, &pr_allow, bf->flag);
660 vfs_flagopt(opts, bf->noname, &ch_allow, bf->flag);
662 ch_allow |= pr_allow;
664 error = vfs_getopt(opts, "name", (void **)&name, &len);
670 if (len == 0 || name[len - 1] != '\0') {
674 if (len > MAXHOSTNAMELEN) {
675 error = ENAMETOOLONG;
680 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
688 if (len == 0 || host[len - 1] != '\0') {
692 if (len > MAXHOSTNAMELEN) {
693 error = ENAMETOOLONG;
698 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
706 if (len == 0 || domain[len - 1] != '\0') {
710 if (len > MAXHOSTNAMELEN) {
711 error = ENAMETOOLONG;
716 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
724 if (len == 0 || uuid[len - 1] != '\0') {
728 if (len > HOSTUUIDLEN) {
729 error = ENAMETOOLONG;
734 #ifdef COMPAT_FREEBSD32
735 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
738 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
742 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
754 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
759 else if (ip4s & (sizeof(*ip4) - 1)) {
763 ch_flags |= PR_IP4_USER;
764 pr_flags |= PR_IP4_USER;
766 ip4s /= sizeof(*ip4);
767 if (ip4s > jail_max_af_ips) {
769 vfs_opterror(opts, "too many IPv4 addresses");
772 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
773 bcopy(op, ip4, ip4s * sizeof(*ip4));
775 * IP addresses are all sorted but ip[0] to preserve
776 * the primary IP address as given from userland.
777 * This special IP is used for unbound outgoing
778 * connections as well for "loopback" traffic in case
779 * source address selection cannot find any more fitting
780 * address to connect from.
783 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4),
786 * Check for duplicate addresses and do some simple
787 * zero and broadcast checks. If users give other bogus
788 * addresses it is their problem.
790 * We do not have to care about byte order for these
791 * checks so we will do them in NBO.
793 for (ii = 0; ii < ip4s; ii++) {
794 if (ip4[ii].s_addr == INADDR_ANY ||
795 ip4[ii].s_addr == INADDR_BROADCAST) {
800 (ip4[0].s_addr == ip4[ii+1].s_addr ||
801 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
811 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
816 else if (ip6s & (sizeof(*ip6) - 1)) {
820 ch_flags |= PR_IP6_USER;
821 pr_flags |= PR_IP6_USER;
823 ip6s /= sizeof(*ip6);
824 if (ip6s > jail_max_af_ips) {
826 vfs_opterror(opts, "too many IPv6 addresses");
829 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
830 bcopy(op, ip6, ip6s * sizeof(*ip6));
832 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6),
834 for (ii = 0; ii < ip6s; ii++) {
835 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
840 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
841 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
851 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
852 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
855 "vnet jails cannot have IP address restrictions");
860 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
866 if (flags & JAIL_UPDATE) {
869 "osrelease cannot be changed after creation");
872 if (len == 0 || len >= OSRELEASELEN) {
875 "osrelease string must be 1-%d bytes long",
881 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
887 if (flags & JAIL_UPDATE) {
890 "osreldate cannot be changed after creation");
895 vfs_opterror(opts, "osreldate cannot be 0");
900 fullpath_disabled = 0;
902 error = vfs_getopt(opts, "path", (void **)&path, &len);
908 if (flags & JAIL_UPDATE) {
911 "path cannot be changed after creation");
914 if (len == 0 || path[len - 1] != '\0') {
918 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
924 NDFREE(&nd, NDF_ONLY_PNBUF);
925 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
926 strlcpy(g_path, path, MAXPATHLEN);
927 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
930 else if (error == ENODEV) {
931 /* proceed if sysctl debug.disablefullpath == 1 */
932 fullpath_disabled = 1;
933 if (len < 2 || (len == 2 && path[0] == '/'))
936 /* exit on other errors */
939 if (root->v_type != VDIR) {
945 if (fullpath_disabled) {
946 /* Leave room for a real-root full pathname. */
947 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
948 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
949 error = ENAMETOOLONG;
957 * Find the specified jail, or at least its parent.
958 * This abuses the file error codes ENOENT and EEXIST.
962 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
963 namelc = strrchr(name, '.');
964 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
968 sx_xlock(&allprison_lock);
971 * See if a requested jid already exists. There is an
972 * information leak here if the jid exists but is not within
973 * the caller's jail hierarchy. Jail creators will get EEXIST
974 * even though they cannot see the jail, and CREATE | UPDATE
975 * will return ENOENT which is not normally a valid error.
979 vfs_opterror(opts, "negative jid");
980 goto done_unlock_list;
982 pr = prison_find(jid);
985 /* Create: jid must not exist. */
986 if (cuflags == JAIL_CREATE) {
987 mtx_unlock(&pr->pr_mtx);
989 vfs_opterror(opts, "jail %d already exists",
991 goto done_unlock_list;
993 if (!prison_ischild(mypr, pr)) {
994 mtx_unlock(&pr->pr_mtx);
996 } else if (pr->pr_uref == 0) {
997 if (!(flags & JAIL_DYING)) {
998 mtx_unlock(&pr->pr_mtx);
1000 vfs_opterror(opts, "jail %d is dying",
1002 goto done_unlock_list;
1003 } else if ((flags & JAIL_ATTACH) ||
1004 (pr_flags & PR_PERSIST)) {
1006 * A dying jail might be resurrected
1007 * (via attach or persist), but first
1008 * it must determine if another jail
1009 * has claimed its name. Accomplish
1010 * this by implicitly re-setting the
1014 name = prison_name(mypr, pr);
1019 /* Update: jid must exist. */
1020 if (cuflags == JAIL_UPDATE) {
1022 vfs_opterror(opts, "jail %d not found", jid);
1023 goto done_unlock_list;
1028 * If the caller provided a name, look for a jail by that name.
1029 * This has different semantics for creates and updates keyed by jid
1030 * (where the name must not already exist in a different jail),
1031 * and updates keyed by the name itself (where the name must exist
1032 * because that is the jail being updated).
1036 namelc = strrchr(name, '.');
1041 * This is a hierarchical name. Split it into the
1042 * parent and child names, and make sure the parent
1043 * exists or matches an already found jail.
1046 if (strncmp(name, ppr->pr_name, namelc - name)
1047 || ppr->pr_name[namelc - name] != '\0') {
1048 mtx_unlock(&pr->pr_mtx);
1051 "cannot change jail's parent");
1052 goto done_unlock_list;
1056 ppr = prison_find_name(mypr, name);
1060 "jail \"%s\" not found", name);
1061 goto done_unlock_list;
1063 mtx_unlock(&ppr->pr_mtx);
1068 if (namelc[0] != '\0') {
1070 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1073 FOREACH_PRISON_CHILD(ppr, tpr) {
1074 if (tpr != pr && tpr->pr_ref > 0 &&
1075 !strcmp(tpr->pr_name + pnamelen, namelc)) {
1077 cuflags != JAIL_CREATE) {
1078 mtx_lock(&tpr->pr_mtx);
1079 if (tpr->pr_ref > 0) {
1084 if (tpr->pr_uref > 0) {
1090 mtx_unlock(&tpr->pr_mtx);
1091 } else if (tpr->pr_uref > 0) {
1093 * Create, or update(jid):
1094 * name must not exist in an
1095 * active sibling jail.
1099 mtx_unlock(&pr->pr_mtx);
1101 "jail \"%s\" already exists",
1103 goto done_unlock_list;
1107 /* If no active jail is found, use a dying one. */
1108 if (deadpr != NULL && pr == NULL) {
1109 if (flags & JAIL_DYING) {
1110 mtx_lock(&deadpr->pr_mtx);
1111 if (deadpr->pr_ref == 0) {
1112 mtx_unlock(&deadpr->pr_mtx);
1116 } else if (cuflags == JAIL_UPDATE) {
1119 "jail \"%s\" is dying", name);
1120 goto done_unlock_list;
1123 /* Update: name must exist if no jid. */
1124 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1126 vfs_opterror(opts, "jail \"%s\" not found",
1128 goto done_unlock_list;
1132 /* Update: must provide a jid or name. */
1133 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1135 vfs_opterror(opts, "update specified no jail");
1136 goto done_unlock_list;
1139 /* If there's no prison to update, create a new one and link it in. */
1141 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1142 if (tpr->pr_childcount >= tpr->pr_childmax) {
1144 vfs_opterror(opts, "prison limit exceeded");
1145 goto done_unlock_list;
1148 mtx_lock(&ppr->pr_mtx);
1149 if (ppr->pr_ref == 0) {
1150 mtx_unlock(&ppr->pr_mtx);
1152 vfs_opterror(opts, "jail \"%s\" not found",
1153 prison_name(mypr, ppr));
1154 goto done_unlock_list;
1158 mtx_unlock(&ppr->pr_mtx);
1159 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1161 /* Find the next free jid. */
1164 if (jid == JAIL_MAX)
1166 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1167 if (tpr->pr_id < jid)
1169 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1170 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1173 if (jid == lastprid) {
1176 "no available jail IDs");
1178 prison_deref(ppr, PD_DEREF |
1179 PD_DEUREF | PD_LIST_XLOCKED);
1188 * The jail already has a jid (that did not yet exist),
1189 * so just find where to insert it.
1191 TAILQ_FOREACH(tpr, &allprison, pr_list)
1192 if (tpr->pr_id >= jid) {
1193 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1198 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1199 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1200 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1201 tpr->pr_childcount++;
1203 pr->pr_parent = ppr;
1206 /* Set some default values, and inherit some from the parent. */
1211 root = mypr->pr_root;
1214 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1215 pr->pr_flags |= PR_HOST;
1216 #if defined(INET) || defined(INET6)
1218 if (!(pr_flags & PR_VNET))
1222 if (!(ch_flags & PR_IP4_USER))
1223 pr->pr_flags |= PR_IP4 | PR_IP4_USER;
1224 else if (!(pr_flags & PR_IP4_USER)) {
1225 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1226 if (ppr->pr_ip4 != NULL) {
1227 pr->pr_ip4s = ppr->pr_ip4s;
1228 pr->pr_ip4 = malloc(pr->pr_ip4s *
1229 sizeof(struct in_addr), M_PRISON,
1231 bcopy(ppr->pr_ip4, pr->pr_ip4,
1232 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1237 if (!(ch_flags & PR_IP6_USER))
1238 pr->pr_flags |= PR_IP6 | PR_IP6_USER;
1239 else if (!(pr_flags & PR_IP6_USER)) {
1240 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1241 if (ppr->pr_ip6 != NULL) {
1242 pr->pr_ip6s = ppr->pr_ip6s;
1243 pr->pr_ip6 = malloc(pr->pr_ip6s *
1244 sizeof(struct in6_addr), M_PRISON,
1246 bcopy(ppr->pr_ip6, pr->pr_ip6,
1247 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1253 /* Source address selection is always on by default. */
1254 pr->pr_flags |= _PR_IP_SADDRSEL;
1256 pr->pr_securelevel = ppr->pr_securelevel;
1257 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1258 pr->pr_enforce_statfs = jail_default_enforce_statfs;
1259 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1261 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1262 if (osrelstr == NULL)
1263 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1265 strcpy(pr->pr_osrelease, osrelstr);
1267 LIST_INIT(&pr->pr_children);
1268 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1269 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
1272 /* Allocate a new vnet if specified. */
1273 pr->pr_vnet = (pr_flags & PR_VNET)
1274 ? vnet_alloc() : ppr->pr_vnet;
1277 * Allocate a dedicated cpuset for each jail.
1278 * Unlike other initial settings, this may return an erorr.
1280 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1282 prison_deref(pr, PD_LIST_XLOCKED);
1286 mtx_lock(&pr->pr_mtx);
1288 * New prisons do not yet have a reference, because we do not
1289 * want others to see the incomplete prison once the
1290 * allprison_lock is downgraded.
1295 * Grab a reference for existing prisons, to ensure they
1296 * continue to exist for the duration of the call.
1299 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1300 if ((pr->pr_flags & PR_VNET) &&
1301 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1304 "vnet jails cannot have IP address restrictions");
1305 goto done_deref_locked;
1309 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1312 "ip4 cannot be changed after creation");
1313 goto done_deref_locked;
1317 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1320 "ip6 cannot be changed after creation");
1321 goto done_deref_locked;
1326 /* Do final error checking before setting anything. */
1328 if (slevel < ppr->pr_securelevel) {
1330 goto done_deref_locked;
1334 if (childmax >= ppr->pr_childmax) {
1336 goto done_deref_locked;
1340 if (enforce < ppr->pr_enforce_statfs) {
1342 goto done_deref_locked;
1347 * devfs_rsnum is a uint16_t
1349 if (rsnum < 0 || rsnum > 65535) {
1351 goto done_deref_locked;
1354 * Nested jails always inherit parent's devfs ruleset
1356 if (jailed(td->td_ucred)) {
1357 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1359 goto done_deref_locked;
1361 rsnum = ppr->pr_devfs_rsnum;
1366 if (ppr->pr_flags & PR_IP4) {
1368 * Make sure the new set of IP addresses is a
1369 * subset of the parent's list. Don't worry
1370 * about the parent being unlocked, as any
1371 * setting is done with allprison_lock held.
1373 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1374 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1376 if (ij == ppr->pr_ip4s) {
1378 goto done_deref_locked;
1381 for (ii = ij = 1; ii < ip4s; ii++) {
1382 if (ip4[ii].s_addr ==
1383 ppr->pr_ip4[0].s_addr)
1385 for (; ij < ppr->pr_ip4s; ij++)
1386 if (ip4[ii].s_addr ==
1387 ppr->pr_ip4[ij].s_addr)
1389 if (ij == ppr->pr_ip4s)
1392 if (ij == ppr->pr_ip4s) {
1394 goto done_deref_locked;
1399 * Check for conflicting IP addresses. We permit them
1400 * if there is no more than one IP on each jail. If
1401 * there is a duplicate on a jail with more than one
1402 * IP stop checking and return error.
1406 for (; tppr != &prison0; tppr = tppr->pr_parent)
1407 if (tppr->pr_flags & PR_VNET)
1410 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1413 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1415 tpr->pr_uref == 0) {
1419 if (!(tpr->pr_flags & PR_IP4_USER))
1422 if (tpr->pr_ip4 == NULL ||
1423 (ip4s == 1 && tpr->pr_ip4s == 1))
1425 for (ii = 0; ii < ip4s; ii++) {
1426 if (prison_check_ip4_locked(tpr, &ip4[ii]) ==
1430 "IPv4 addresses clash");
1431 goto done_deref_locked;
1439 if (ppr->pr_flags & PR_IP6) {
1441 * Make sure the new set of IP addresses is a
1442 * subset of the parent's list.
1444 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1445 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1448 if (ij == ppr->pr_ip6s) {
1450 goto done_deref_locked;
1453 for (ii = ij = 1; ii < ip6s; ii++) {
1454 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1457 for (; ij < ppr->pr_ip6s; ij++)
1458 if (IN6_ARE_ADDR_EQUAL(
1459 &ip6[ii], &ppr->pr_ip6[ij]))
1461 if (ij == ppr->pr_ip6s)
1464 if (ij == ppr->pr_ip6s) {
1466 goto done_deref_locked;
1470 /* Check for conflicting IP addresses. */
1473 for (; tppr != &prison0; tppr = tppr->pr_parent)
1474 if (tppr->pr_flags & PR_VNET)
1477 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1480 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1482 tpr->pr_uref == 0) {
1486 if (!(tpr->pr_flags & PR_IP6_USER))
1489 if (tpr->pr_ip6 == NULL ||
1490 (ip6s == 1 && tpr->pr_ip6s == 1))
1492 for (ii = 0; ii < ip6s; ii++) {
1493 if (prison_check_ip6_locked(tpr, &ip6[ii]) ==
1497 "IPv6 addresses clash");
1498 goto done_deref_locked;
1504 onamelen = namelen = 0;
1505 if (namelc != NULL) {
1506 /* Give a default name of the jid. Also allow the name to be
1507 * explicitly the jid - but not any other number, and only in
1508 * normal form (no leading zero/etc).
1510 if (namelc[0] == '\0')
1511 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid);
1512 else if ((strtoul(namelc, &p, 10) != jid ||
1513 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1516 "name cannot be numeric (unless it is the jid)");
1517 goto done_deref_locked;
1520 * Make sure the name isn't too long for the prison or its
1523 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1524 onamelen = strlen(pr->pr_name + pnamelen);
1525 namelen = strlen(namelc);
1526 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) {
1527 error = ENAMETOOLONG;
1528 goto done_deref_locked;
1530 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1531 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1532 sizeof(pr->pr_name)) {
1533 error = ENAMETOOLONG;
1534 goto done_deref_locked;
1538 if (pr_allow & ~ppr->pr_allow) {
1540 goto done_deref_locked;
1544 * Let modules check their parameters. This requires unlocking and
1545 * then re-locking the prison, but this is still a valid state as long
1546 * as allprison_lock remains xlocked.
1548 mtx_unlock(&pr->pr_mtx);
1549 error = osd_jail_call(pr, PR_METHOD_CHECK, opts);
1551 prison_deref(pr, created
1553 : PD_DEREF | PD_LIST_XLOCKED);
1556 mtx_lock(&pr->pr_mtx);
1558 /* At this point, all valid parameters should have been noted. */
1559 TAILQ_FOREACH(opt, opts, link) {
1560 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1562 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1563 goto done_deref_locked;
1567 /* Set the parameters of the prison. */
1570 if (pr_flags & PR_IP4_USER) {
1571 pr->pr_flags |= PR_IP4;
1572 free(pr->pr_ip4, M_PRISON);
1576 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1578 if (tpr->pr_flags & PR_VNET) {
1583 if (prison_restrict_ip4(tpr, NULL)) {
1592 if (pr_flags & PR_IP6_USER) {
1593 pr->pr_flags |= PR_IP6;
1594 free(pr->pr_ip6, M_PRISON);
1598 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1600 if (tpr->pr_flags & PR_VNET) {
1605 if (prison_restrict_ip6(tpr, NULL)) {
1613 pr->pr_securelevel = slevel;
1614 /* Set all child jails to be at least this level. */
1615 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1616 if (tpr->pr_securelevel < slevel)
1617 tpr->pr_securelevel = slevel;
1620 pr->pr_childmax = childmax;
1621 /* Set all child jails to under this limit. */
1622 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1623 if (tpr->pr_childmax > childmax - level)
1624 tpr->pr_childmax = childmax > level
1625 ? childmax - level : 0;
1628 pr->pr_enforce_statfs = enforce;
1629 /* Pass this restriction on to the children. */
1630 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1631 if (tpr->pr_enforce_statfs < enforce)
1632 tpr->pr_enforce_statfs = enforce;
1635 pr->pr_devfs_rsnum = rsnum;
1636 /* Pass this restriction on to the children. */
1637 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1638 tpr->pr_devfs_rsnum = rsnum;
1640 if (namelc != NULL) {
1641 if (ppr == &prison0)
1642 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name));
1644 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1645 ppr->pr_name, namelc);
1646 /* Change this component of child names. */
1647 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1648 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1649 strlen(tpr->pr_name + onamelen) + 1);
1650 bcopy(pr->pr_name, tpr->pr_name, namelen);
1654 /* Try to keep a real-rooted full pathname. */
1655 if (fullpath_disabled && path[0] == '/' &&
1656 strcmp(mypr->pr_path, "/"))
1657 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1658 mypr->pr_path, path);
1660 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1663 if (PR_HOST & ch_flags & ~pr_flags) {
1664 if (pr->pr_flags & PR_HOST) {
1666 * Copy the parent's host info. As with pr_ip4 above,
1667 * the lack of a lock on the parent is not a problem;
1668 * it is always set with allprison_lock at least
1669 * shared, and is held exclusively here.
1671 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1672 sizeof(pr->pr_hostname));
1673 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1674 sizeof(pr->pr_domainname));
1675 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1676 sizeof(pr->pr_hostuuid));
1677 pr->pr_hostid = pr->pr_parent->pr_hostid;
1679 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1680 /* Set this prison, and any descendants without PR_HOST. */
1682 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1684 strlcpy(pr->pr_domainname, domain,
1685 sizeof(pr->pr_domainname));
1687 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1689 pr->pr_hostid = hid;
1690 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1691 if (tpr->pr_flags & PR_HOST)
1695 strlcpy(tpr->pr_hostname,
1697 sizeof(tpr->pr_hostname));
1699 strlcpy(tpr->pr_domainname,
1701 sizeof(tpr->pr_domainname));
1703 strlcpy(tpr->pr_hostuuid,
1705 sizeof(tpr->pr_hostuuid));
1707 tpr->pr_hostid = hid;
1711 if ((tallow = ch_allow & ~pr_allow)) {
1712 /* Clear allow bits in all children. */
1713 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1714 tpr->pr_allow &= ~tallow;
1716 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1718 * Persistent prisons get an extra reference, and prisons losing their
1719 * persist flag lose that reference. Only do this for existing prisons
1720 * for now, so new ones will remain unseen until after the module
1721 * handlers have completed.
1723 born = pr->pr_uref == 0;
1724 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1725 if (pr_flags & PR_PERSIST) {
1733 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1734 mtx_unlock(&pr->pr_mtx);
1737 if (racct_enable && created)
1738 prison_racct_attach(pr);
1741 /* Locks may have prevented a complete restriction of child IP
1742 * addresses. If so, allocate some more memory and try again.
1747 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1748 mtx_lock(&pr->pr_mtx);
1750 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1752 if (tpr->pr_flags & PR_VNET) {
1757 if (prison_restrict_ip4(tpr, ip4)) {
1764 mtx_unlock(&pr->pr_mtx);
1770 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1771 mtx_lock(&pr->pr_mtx);
1773 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1775 if (tpr->pr_flags & PR_VNET) {
1780 if (prison_restrict_ip6(tpr, ip6)) {
1787 mtx_unlock(&pr->pr_mtx);
1791 /* Let the modules do their work. */
1792 sx_downgrade(&allprison_lock);
1794 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1796 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1797 prison_deref(pr, created
1799 : PD_DEREF | PD_LIST_SLOCKED);
1803 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1806 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
1807 prison_deref(pr, created
1809 : PD_DEREF | PD_LIST_SLOCKED);
1813 /* Attach this process to the prison if requested. */
1814 if (flags & JAIL_ATTACH) {
1815 mtx_lock(&pr->pr_mtx);
1816 error = do_jail_attach(td, pr);
1818 vfs_opterror(opts, "attach failed");
1820 prison_deref(pr, PD_DEREF);
1826 if (racct_enable && !created) {
1827 if (!(flags & JAIL_ATTACH))
1828 sx_sunlock(&allprison_lock);
1829 prison_racct_modify(pr);
1830 if (!(flags & JAIL_ATTACH))
1831 sx_slock(&allprison_lock);
1835 td->td_retval[0] = pr->pr_id;
1838 * Now that it is all there, drop the temporary reference from existing
1839 * prisons. Or add a reference to newly created persistent prisons
1840 * (which was not done earlier so that the prison would not be publicly
1844 prison_deref(pr, (flags & JAIL_ATTACH)
1846 : PD_DEREF | PD_LIST_SLOCKED);
1848 if (pr_flags & PR_PERSIST) {
1849 mtx_lock(&pr->pr_mtx);
1852 mtx_unlock(&pr->pr_mtx);
1854 if (!(flags & JAIL_ATTACH))
1855 sx_sunlock(&allprison_lock);
1861 prison_deref(pr, created
1862 ? PD_LOCKED | PD_LIST_XLOCKED
1863 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1866 sx_xunlock(&allprison_lock);
1872 if (vfs_getopt(opts, "errmsg", (void **)&errmsg,
1873 &errmsg_len) == 0 && errmsg_len > 0) {
1874 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1875 if (optuio->uio_segflg == UIO_SYSSPACE)
1877 optuio->uio_iov[errmsg_pos].iov_base,
1881 optuio->uio_iov[errmsg_pos].iov_base,
1887 free(ip4, M_PRISON);
1890 free(ip6, M_PRISON);
1893 free(g_path, M_TEMP);
1900 * struct jail_get_args {
1901 * struct iovec *iovp;
1902 * unsigned int iovcnt;
1907 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1912 /* Check that we have an even number of iovecs. */
1913 if (uap->iovcnt & 1)
1916 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1919 error = kern_jail_get(td, auio, uap->flags);
1921 error = copyout(auio->uio_iov, uap->iovp,
1922 uap->iovcnt * sizeof (struct iovec));
1928 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1930 struct bool_flags *bf;
1931 struct jailsys_flags *jsf;
1932 struct prison *pr, *mypr;
1934 struct vfsoptlist *opts;
1935 char *errmsg, *name;
1936 int error, errmsg_len, errmsg_pos, i, jid, len, locked, pos;
1939 if (flags & ~JAIL_GET_MASK)
1942 /* Get the parameter list. */
1943 error = vfs_buildopts(optuio, &opts);
1946 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1947 mypr = td->td_ucred->cr_prison;
1950 * Find the prison specified by one of: lastjid, jid, name.
1952 sx_slock(&allprison_lock);
1953 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1955 TAILQ_FOREACH(pr, &allprison, pr_list) {
1956 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1957 mtx_lock(&pr->pr_mtx);
1958 if (pr->pr_ref > 0 &&
1959 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1961 mtx_unlock(&pr->pr_mtx);
1967 vfs_opterror(opts, "no jail after %d", jid);
1968 goto done_unlock_list;
1969 } else if (error != ENOENT)
1970 goto done_unlock_list;
1972 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1975 pr = prison_find_child(mypr, jid);
1977 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1978 mtx_unlock(&pr->pr_mtx);
1980 vfs_opterror(opts, "jail %d is dying",
1982 goto done_unlock_list;
1987 vfs_opterror(opts, "jail %d not found", jid);
1988 goto done_unlock_list;
1990 } else if (error != ENOENT)
1991 goto done_unlock_list;
1993 error = vfs_getopt(opts, "name", (void **)&name, &len);
1995 if (len == 0 || name[len - 1] != '\0') {
1997 goto done_unlock_list;
1999 pr = prison_find_name(mypr, name);
2001 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2002 mtx_unlock(&pr->pr_mtx);
2004 vfs_opterror(opts, "jail \"%s\" is dying",
2006 goto done_unlock_list;
2011 vfs_opterror(opts, "jail \"%s\" not found", name);
2012 goto done_unlock_list;
2013 } else if (error != ENOENT)
2014 goto done_unlock_list;
2016 vfs_opterror(opts, "no jail specified");
2018 goto done_unlock_list;
2021 /* Get the parameters of the prison. */
2024 td->td_retval[0] = pr->pr_id;
2025 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2026 if (error != 0 && error != ENOENT)
2028 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2029 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2030 if (error != 0 && error != ENOENT)
2032 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2033 if (error != 0 && error != ENOENT)
2035 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2036 sizeof(pr->pr_cpuset->cs_id));
2037 if (error != 0 && error != ENOENT)
2039 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2040 if (error != 0 && error != ENOENT)
2043 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2044 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2045 if (error != 0 && error != ENOENT)
2049 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2050 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2051 if (error != 0 && error != ENOENT)
2054 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2055 sizeof(pr->pr_securelevel));
2056 if (error != 0 && error != ENOENT)
2058 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2059 sizeof(pr->pr_childcount));
2060 if (error != 0 && error != ENOENT)
2062 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2063 sizeof(pr->pr_childmax));
2064 if (error != 0 && error != ENOENT)
2066 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2067 if (error != 0 && error != ENOENT)
2069 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2070 if (error != 0 && error != ENOENT)
2072 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2073 if (error != 0 && error != ENOENT)
2075 #ifdef COMPAT_FREEBSD32
2076 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2077 uint32_t hid32 = pr->pr_hostid;
2079 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2082 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2083 sizeof(pr->pr_hostid));
2084 if (error != 0 && error != ENOENT)
2086 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2087 sizeof(pr->pr_enforce_statfs));
2088 if (error != 0 && error != ENOENT)
2090 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2091 sizeof(pr->pr_devfs_rsnum));
2092 if (error != 0 && error != ENOENT)
2094 for (bf = pr_flag_bool;
2095 bf < pr_flag_bool + nitems(pr_flag_bool);
2097 i = (pr->pr_flags & bf->flag) ? 1 : 0;
2098 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2099 if (error != 0 && error != ENOENT)
2102 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2103 if (error != 0 && error != ENOENT)
2106 for (jsf = pr_flag_jailsys;
2107 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
2109 f = pr->pr_flags & (jsf->disable | jsf->new);
2110 i = (f != 0 && f == jsf->disable) ? JAIL_SYS_DISABLE
2111 : (f == jsf->new) ? JAIL_SYS_NEW
2113 error = vfs_setopt(opts, jsf->name, &i, sizeof(i));
2114 if (error != 0 && error != ENOENT)
2117 for (bf = pr_flag_allow;
2118 bf < pr_flag_allow + nitems(pr_flag_allow);
2120 i = (pr->pr_allow & bf->flag) ? 1 : 0;
2121 error = vfs_setopt(opts, bf->name, &i, sizeof(i));
2122 if (error != 0 && error != ENOENT)
2125 error = vfs_setopt(opts, bf->noname, &i, sizeof(i));
2126 if (error != 0 && error != ENOENT)
2129 i = (pr->pr_uref == 0);
2130 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2131 if (error != 0 && error != ENOENT)
2134 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2135 if (error != 0 && error != ENOENT)
2137 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2138 sizeof(pr->pr_osreldate));
2139 if (error != 0 && error != ENOENT)
2141 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2142 if (error != 0 && error != ENOENT)
2145 /* Get the module parameters. */
2146 mtx_unlock(&pr->pr_mtx);
2148 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2151 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2153 /* By now, all parameters should have been noted. */
2154 TAILQ_FOREACH(opt, opts, link) {
2155 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2157 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2162 /* Write the fetched parameters back to userspace. */
2164 TAILQ_FOREACH(opt, opts, link) {
2165 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2166 pos = 2 * opt->pos + 1;
2167 optuio->uio_iov[pos].iov_len = opt->len;
2168 if (opt->value != NULL) {
2169 if (optuio->uio_segflg == UIO_SYSSPACE) {
2171 optuio->uio_iov[pos].iov_base,
2174 error = copyout(opt->value,
2175 optuio->uio_iov[pos].iov_base,
2186 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2190 sx_sunlock(&allprison_lock);
2192 if (error && errmsg_pos >= 0) {
2193 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2194 errmsg_pos = 2 * errmsg_pos + 1;
2195 if (errmsg_len > 0) {
2196 if (optuio->uio_segflg == UIO_SYSSPACE)
2198 optuio->uio_iov[errmsg_pos].iov_base,
2202 optuio->uio_iov[errmsg_pos].iov_base,
2212 * struct jail_remove_args {
2217 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2219 struct prison *pr, *cpr, *lpr, *tpr;
2222 error = priv_check(td, PRIV_JAIL_REMOVE);
2226 sx_xlock(&allprison_lock);
2227 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2229 sx_xunlock(&allprison_lock);
2233 /* Remove all descendants of this prison, then remove this prison. */
2235 if (!LIST_EMPTY(&pr->pr_children)) {
2236 mtx_unlock(&pr->pr_mtx);
2238 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2239 mtx_lock(&cpr->pr_mtx);
2240 if (cpr->pr_ref > 0) {
2244 /* Already removed - do not do it again. */
2247 mtx_unlock(&cpr->pr_mtx);
2249 mtx_lock(&lpr->pr_mtx);
2250 prison_remove_one(lpr);
2251 sx_xlock(&allprison_lock);
2256 mtx_lock(&lpr->pr_mtx);
2257 prison_remove_one(lpr);
2258 sx_xlock(&allprison_lock);
2260 mtx_lock(&pr->pr_mtx);
2262 prison_remove_one(pr);
2267 prison_remove_one(struct prison *pr)
2272 /* If the prison was persistent, it is not anymore. */
2274 if (pr->pr_flags & PR_PERSIST) {
2277 pr->pr_flags &= ~PR_PERSIST;
2281 * jail_remove added a reference. If that's the only one, remove
2284 KASSERT(pr->pr_ref > 0,
2285 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2286 if (pr->pr_ref == 1) {
2288 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2292 mtx_unlock(&pr->pr_mtx);
2293 sx_xunlock(&allprison_lock);
2295 * Kill all processes unfortunate enough to be attached to this prison.
2297 sx_slock(&allproc_lock);
2298 LIST_FOREACH(p, &allproc, p_list) {
2300 if (p->p_state != PRS_NEW && p->p_ucred &&
2301 p->p_ucred->cr_prison == pr)
2302 kern_psignal(p, SIGKILL);
2305 sx_sunlock(&allproc_lock);
2306 /* Remove the temporary reference added by jail_remove. */
2307 prison_deref(pr, deuref | PD_DEREF);
2312 * struct jail_attach_args {
2317 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2322 error = priv_check(td, PRIV_JAIL_ATTACH);
2327 * Start with exclusive hold on allprison_lock to ensure that a possible
2328 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove.
2329 * But then immediately downgrade it since we don't need to stop
2332 sx_xlock(&allprison_lock);
2333 sx_downgrade(&allprison_lock);
2334 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2336 sx_sunlock(&allprison_lock);
2341 * Do not allow a process to attach to a prison that is not
2342 * considered to be "alive".
2344 if (pr->pr_uref == 0) {
2345 mtx_unlock(&pr->pr_mtx);
2346 sx_sunlock(&allprison_lock);
2350 return (do_jail_attach(td, pr));
2354 do_jail_attach(struct thread *td, struct prison *pr)
2357 struct ucred *newcred, *oldcred;
2361 * XXX: Note that there is a slight race here if two threads
2362 * in the same privileged process attempt to attach to two
2363 * different jails at the same time. It is important for
2364 * user processes not to do this, or they might end up with
2365 * a process root from one prison, but attached to the jail
2370 mtx_unlock(&pr->pr_mtx);
2372 /* Let modules do whatever they need to prepare for attaching. */
2373 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2375 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2378 sx_sunlock(&allprison_lock);
2381 * Reparent the newly attached process to this jail.
2384 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2388 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2389 if ((error = change_dir(pr->pr_root, td)) != 0)
2392 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2395 VOP_UNLOCK(pr->pr_root, 0);
2396 if ((error = pwd_chroot(td, pr->pr_root)))
2401 oldcred = crcopysafe(p, newcred);
2402 newcred->cr_prison = pr;
2403 proc_set_cred(p, newcred);
2406 racct_proc_ucred_changed(p, oldcred, newcred);
2411 rctl_proc_ucred_changed(p, newcred);
2414 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF);
2419 VOP_UNLOCK(pr->pr_root, 0);
2421 /* Tell modules this thread is still in its old jail after all. */
2422 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td);
2423 prison_deref(pr, PD_DEREF | PD_DEUREF);
2429 * Returns a locked prison instance, or NULL on failure.
2432 prison_find(int prid)
2436 sx_assert(&allprison_lock, SX_LOCKED);
2437 TAILQ_FOREACH(pr, &allprison, pr_list) {
2438 if (pr->pr_id == prid) {
2439 mtx_lock(&pr->pr_mtx);
2442 mtx_unlock(&pr->pr_mtx);
2449 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2452 prison_find_child(struct prison *mypr, int prid)
2457 sx_assert(&allprison_lock, SX_LOCKED);
2458 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2459 if (pr->pr_id == prid) {
2460 mtx_lock(&pr->pr_mtx);
2463 mtx_unlock(&pr->pr_mtx);
2470 * Look for the name relative to mypr. Returns a locked prison or NULL.
2473 prison_find_name(struct prison *mypr, const char *name)
2475 struct prison *pr, *deadpr;
2479 sx_assert(&allprison_lock, SX_LOCKED);
2480 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2483 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2484 if (!strcmp(pr->pr_name + mylen, name)) {
2485 mtx_lock(&pr->pr_mtx);
2486 if (pr->pr_ref > 0) {
2487 if (pr->pr_uref > 0)
2491 mtx_unlock(&pr->pr_mtx);
2494 /* There was no valid prison - perhaps there was a dying one. */
2495 if (deadpr != NULL) {
2496 mtx_lock(&deadpr->pr_mtx);
2497 if (deadpr->pr_ref == 0) {
2498 mtx_unlock(&deadpr->pr_mtx);
2506 * See if a prison has the specific flag set.
2509 prison_flag(struct ucred *cred, unsigned flag)
2512 /* This is an atomic read, so no locking is necessary. */
2513 return (cred->cr_prison->pr_flags & flag);
2517 prison_allow(struct ucred *cred, unsigned flag)
2520 /* This is an atomic read, so no locking is necessary. */
2521 return (cred->cr_prison->pr_allow & flag);
2525 * Remove a prison reference. If that was the last reference, remove the
2526 * prison itself - but not in this context in case there are locks held.
2529 prison_free_locked(struct prison *pr)
2533 mtx_assert(&pr->pr_mtx, MA_OWNED);
2535 mtx_unlock(&pr->pr_mtx);
2537 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2541 prison_free(struct prison *pr)
2544 mtx_lock(&pr->pr_mtx);
2545 prison_free_locked(pr);
2549 * Complete a call to either prison_free or prison_proc_free.
2552 prison_complete(void *context, int pending)
2554 struct prison *pr = context;
2556 sx_xlock(&allprison_lock);
2557 mtx_lock(&pr->pr_mtx);
2558 prison_deref(pr, pr->pr_uref
2559 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED
2560 : PD_LOCKED | PD_LIST_XLOCKED);
2564 * Remove a prison reference (usually). This internal version assumes no
2565 * mutexes are held, except perhaps the prison itself. If there are no more
2566 * references, release and delist the prison. On completion, the prison lock
2567 * and the allprison lock are both unlocked.
2570 prison_deref(struct prison *pr, int flags)
2572 struct prison *ppr, *tpr;
2575 if (!(flags & PD_LOCKED))
2576 mtx_lock(&pr->pr_mtx);
2578 if (flags & PD_DEUREF) {
2579 KASSERT(pr->pr_uref > 0,
2580 ("prison_deref PD_DEUREF on a dead prison (jid=%d)",
2583 lasturef = pr->pr_uref == 0;
2586 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2589 if (flags & PD_DEREF) {
2590 KASSERT(pr->pr_ref > 0,
2591 ("prison_deref PD_DEREF on a dead prison (jid=%d)",
2596 mtx_unlock(&pr->pr_mtx);
2599 * Tell the modules if the last user reference was removed
2600 * (even it sticks around in dying state).
2603 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) {
2604 sx_xlock(&allprison_lock);
2605 flags |= PD_LIST_XLOCKED;
2607 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL);
2608 mtx_lock(&pr->pr_mtx);
2610 mtx_unlock(&pr->pr_mtx);
2613 /* If the prison still has references, nothing else to do. */
2615 if (flags & PD_LIST_SLOCKED)
2616 sx_sunlock(&allprison_lock);
2617 else if (flags & PD_LIST_XLOCKED)
2618 sx_xunlock(&allprison_lock);
2622 if (flags & PD_LIST_SLOCKED) {
2623 if (!sx_try_upgrade(&allprison_lock)) {
2624 sx_sunlock(&allprison_lock);
2625 sx_xlock(&allprison_lock);
2627 } else if (!(flags & PD_LIST_XLOCKED))
2628 sx_xlock(&allprison_lock);
2630 TAILQ_REMOVE(&allprison, pr, pr_list);
2631 LIST_REMOVE(pr, pr_sibling);
2632 ppr = pr->pr_parent;
2633 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2634 tpr->pr_childcount--;
2635 sx_xunlock(&allprison_lock);
2638 if (pr->pr_vnet != ppr->pr_vnet)
2639 vnet_destroy(pr->pr_vnet);
2641 if (pr->pr_root != NULL)
2643 mtx_destroy(&pr->pr_mtx);
2645 free(pr->pr_ip4, M_PRISON);
2648 free(pr->pr_ip6, M_PRISON);
2650 if (pr->pr_cpuset != NULL)
2651 cpuset_rel(pr->pr_cpuset);
2655 prison_racct_detach(pr);
2659 /* Removing a prison frees a reference on its parent. */
2661 mtx_lock(&pr->pr_mtx);
2662 flags = PD_DEREF | PD_DEUREF;
2667 prison_hold_locked(struct prison *pr)
2670 mtx_assert(&pr->pr_mtx, MA_OWNED);
2671 KASSERT(pr->pr_ref > 0,
2672 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2677 prison_hold(struct prison *pr)
2680 mtx_lock(&pr->pr_mtx);
2681 prison_hold_locked(pr);
2682 mtx_unlock(&pr->pr_mtx);
2686 prison_proc_hold(struct prison *pr)
2689 mtx_lock(&pr->pr_mtx);
2690 KASSERT(pr->pr_uref > 0,
2691 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2693 mtx_unlock(&pr->pr_mtx);
2697 prison_proc_free(struct prison *pr)
2700 mtx_lock(&pr->pr_mtx);
2701 KASSERT(pr->pr_uref > 0,
2702 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2703 if (pr->pr_uref > 1)
2707 * Don't remove the last user reference in this context, which
2708 * is expected to be a process that is not only locked, but
2712 mtx_unlock(&pr->pr_mtx);
2713 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2716 mtx_unlock(&pr->pr_mtx);
2720 * Check if a jail supports the given address family.
2722 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
2726 prison_check_af(struct ucred *cred, int af)
2731 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2733 pr = cred->cr_prison;
2735 /* Prisons with their own network stack are not limited. */
2736 if (prison_owns_vnet(cred))
2745 if (pr->pr_flags & PR_IP4)
2747 mtx_lock(&pr->pr_mtx);
2748 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
2749 error = EAFNOSUPPORT;
2750 mtx_unlock(&pr->pr_mtx);
2756 if (pr->pr_flags & PR_IP6)
2758 mtx_lock(&pr->pr_mtx);
2759 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
2760 error = EAFNOSUPPORT;
2761 mtx_unlock(&pr->pr_mtx);
2769 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
2770 error = EAFNOSUPPORT;
2776 * Check if given address belongs to the jail referenced by cred (wrapper to
2777 * prison_check_ip[46]).
2779 * Returns 0 if jail doesn't restrict the address family or if address belongs
2780 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
2781 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
2784 prison_if(struct ucred *cred, struct sockaddr *sa)
2787 struct sockaddr_in *sai;
2790 struct sockaddr_in6 *sai6;
2794 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2795 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
2798 if (prison_owns_vnet(cred))
2803 switch (sa->sa_family)
2807 sai = (struct sockaddr_in *)sa;
2808 error = prison_check_ip4(cred, &sai->sin_addr);
2813 sai6 = (struct sockaddr_in6 *)sa;
2814 error = prison_check_ip6(cred, &sai6->sin6_addr);
2818 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
2819 error = EAFNOSUPPORT;
2825 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
2828 prison_check(struct ucred *cred1, struct ucred *cred2)
2831 return ((cred1->cr_prison == cred2->cr_prison ||
2832 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
2836 * Return 1 if p2 is a child of p1, otherwise 0.
2839 prison_ischild(struct prison *pr1, struct prison *pr2)
2842 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
2849 * Return 1 if the passed credential is in a jail, otherwise 0.
2852 jailed(struct ucred *cred)
2855 return (cred->cr_prison != &prison0);
2859 * Return 1 if the passed credential is in a jail and that jail does not
2860 * have its own virtual network stack, otherwise 0.
2863 jailed_without_vnet(struct ucred *cred)
2869 if (prison_owns_vnet(cred))
2877 * Return the correct hostname (domainname, et al) for the passed credential.
2880 getcredhostname(struct ucred *cred, char *buf, size_t size)
2885 * A NULL credential can be used to shortcut to the physical
2886 * system's hostname.
2888 pr = (cred != NULL) ? cred->cr_prison : &prison0;
2889 mtx_lock(&pr->pr_mtx);
2890 strlcpy(buf, pr->pr_hostname, size);
2891 mtx_unlock(&pr->pr_mtx);
2895 getcreddomainname(struct ucred *cred, char *buf, size_t size)
2898 mtx_lock(&cred->cr_prison->pr_mtx);
2899 strlcpy(buf, cred->cr_prison->pr_domainname, size);
2900 mtx_unlock(&cred->cr_prison->pr_mtx);
2904 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
2907 mtx_lock(&cred->cr_prison->pr_mtx);
2908 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
2909 mtx_unlock(&cred->cr_prison->pr_mtx);
2913 getcredhostid(struct ucred *cred, unsigned long *hostid)
2916 mtx_lock(&cred->cr_prison->pr_mtx);
2917 *hostid = cred->cr_prison->pr_hostid;
2918 mtx_unlock(&cred->cr_prison->pr_mtx);
2923 * Determine whether the prison represented by cred owns
2924 * its vnet rather than having it inherited.
2926 * Returns 1 in case the prison owns the vnet, 0 otherwise.
2929 prison_owns_vnet(struct ucred *cred)
2933 * vnets cannot be added/removed after jail creation,
2934 * so no need to lock here.
2936 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
2941 * Determine whether the subject represented by cred can "see"
2942 * status of a mount point.
2943 * Returns: 0 for permitted, ENOENT otherwise.
2944 * XXX: This function should be called cr_canseemount() and should be
2945 * placed in kern_prot.c.
2948 prison_canseemount(struct ucred *cred, struct mount *mp)
2954 pr = cred->cr_prison;
2955 if (pr->pr_enforce_statfs == 0)
2957 if (pr->pr_root->v_mount == mp)
2959 if (pr->pr_enforce_statfs == 2)
2962 * If jail's chroot directory is set to "/" we should be able to see
2963 * all mount-points from inside a jail.
2964 * This is ugly check, but this is the only situation when jail's
2965 * directory ends with '/'.
2967 if (strcmp(pr->pr_path, "/") == 0)
2969 len = strlen(pr->pr_path);
2971 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
2974 * Be sure that we don't have situation where jail's root directory
2975 * is "/some/path" and mount point is "/some/pathpath".
2977 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
2983 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
2985 char jpath[MAXPATHLEN];
2989 pr = cred->cr_prison;
2990 if (pr->pr_enforce_statfs == 0)
2992 if (prison_canseemount(cred, mp) != 0) {
2993 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
2994 strlcpy(sp->f_mntonname, "[restricted]",
2995 sizeof(sp->f_mntonname));
2998 if (pr->pr_root->v_mount == mp) {
3000 * Clear current buffer data, so we are sure nothing from
3001 * the valid path left there.
3003 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3004 *sp->f_mntonname = '/';
3008 * If jail's chroot directory is set to "/" we should be able to see
3009 * all mount-points from inside a jail.
3011 if (strcmp(pr->pr_path, "/") == 0)
3013 len = strlen(pr->pr_path);
3014 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3016 * Clear current buffer data, so we are sure nothing from
3017 * the valid path left there.
3019 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3020 if (*jpath == '\0') {
3021 /* Should never happen. */
3022 *sp->f_mntonname = '/';
3024 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3029 * Check with permission for a specific privilege is granted within jail. We
3030 * have a specific list of accepted privileges; the rest are denied.
3033 prison_priv_check(struct ucred *cred, int priv)
3041 * Privileges specific to prisons with a virtual network stack.
3042 * There might be a duplicate entry here in case the privilege
3043 * is only granted conditionally in the legacy jail case.
3048 * NFS-specific privileges.
3050 case PRIV_NFS_DAEMON:
3051 case PRIV_NFS_LOCKD:
3054 * Network stack privileges.
3056 case PRIV_NET_BRIDGE:
3059 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3060 case PRIV_NET_ROUTE:
3062 case PRIV_NET_SETIFMTU:
3063 case PRIV_NET_SETIFFLAGS:
3064 case PRIV_NET_SETIFCAP:
3065 case PRIV_NET_SETIFDESCR:
3066 case PRIV_NET_SETIFNAME :
3067 case PRIV_NET_SETIFMETRIC:
3068 case PRIV_NET_SETIFPHYS:
3069 case PRIV_NET_SETIFMAC:
3070 case PRIV_NET_ADDMULTI:
3071 case PRIV_NET_DELMULTI:
3072 case PRIV_NET_HWIOCTL:
3073 case PRIV_NET_SETLLADDR:
3074 case PRIV_NET_ADDIFGROUP:
3075 case PRIV_NET_DELIFGROUP:
3076 case PRIV_NET_IFCREATE:
3077 case PRIV_NET_IFDESTROY:
3078 case PRIV_NET_ADDIFADDR:
3079 case PRIV_NET_DELIFADDR:
3082 case PRIV_NET_SETIFVNET:
3083 case PRIV_NET_SETIFFIB:
3086 * 802.11-related privileges.
3088 case PRIV_NET80211_GETKEY:
3090 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3097 case PRIV_NETATM_CFG:
3098 case PRIV_NETATM_ADD:
3099 case PRIV_NETATM_DEL:
3100 case PRIV_NETATM_SET:
3103 * Bluetooth privileges.
3105 case PRIV_NETBLUETOOTH_RAW:
3109 * Netgraph and netgraph module privileges.
3111 case PRIV_NETGRAPH_CONTROL:
3113 case PRIV_NETGRAPH_TTY:
3117 * IPv4 and IPv6 privileges.
3119 case PRIV_NETINET_IPFW:
3120 case PRIV_NETINET_DIVERT:
3121 case PRIV_NETINET_PF:
3122 case PRIV_NETINET_DUMMYNET:
3123 case PRIV_NETINET_CARP:
3124 case PRIV_NETINET_MROUTE:
3125 case PRIV_NETINET_RAW:
3126 case PRIV_NETINET_ADDRCTRL6:
3127 case PRIV_NETINET_ND6:
3128 case PRIV_NETINET_SCOPE6:
3129 case PRIV_NETINET_ALIFETIME6:
3130 case PRIV_NETINET_IPSEC:
3131 case PRIV_NETINET_BINDANY:
3146 * No default: or deny here.
3147 * In case of no permit fall through to next switch().
3149 if (cred->cr_prison->pr_flags & PR_VNET)
3157 * Allow ktrace privileges for root in jail.
3163 * Allow jailed processes to configure audit identity and
3164 * submit audit records (login, etc). In the future we may
3165 * want to further refine the relationship between audit and
3168 case PRIV_AUDIT_GETAUDIT:
3169 case PRIV_AUDIT_SETAUDIT:
3170 case PRIV_AUDIT_SUBMIT:
3174 * Allow jailed processes to manipulate process UNIX
3175 * credentials in any way they see fit.
3177 case PRIV_CRED_SETUID:
3178 case PRIV_CRED_SETEUID:
3179 case PRIV_CRED_SETGID:
3180 case PRIV_CRED_SETEGID:
3181 case PRIV_CRED_SETGROUPS:
3182 case PRIV_CRED_SETREUID:
3183 case PRIV_CRED_SETREGID:
3184 case PRIV_CRED_SETRESUID:
3185 case PRIV_CRED_SETRESGID:
3188 * Jail implements visibility constraints already, so allow
3189 * jailed root to override uid/gid-based constraints.
3191 case PRIV_SEEOTHERGIDS:
3192 case PRIV_SEEOTHERUIDS:
3195 * Jail implements inter-process debugging limits already, so
3196 * allow jailed root various debugging privileges.
3198 case PRIV_DEBUG_DIFFCRED:
3199 case PRIV_DEBUG_SUGID:
3200 case PRIV_DEBUG_UNPRIV:
3203 * Allow jail to set various resource limits and login
3204 * properties, and for now, exceed process resource limits.
3206 case PRIV_PROC_LIMIT:
3207 case PRIV_PROC_SETLOGIN:
3208 case PRIV_PROC_SETRLIMIT:
3211 * System V and POSIX IPC privileges are granted in jail.
3214 case PRIV_IPC_WRITE:
3215 case PRIV_IPC_ADMIN:
3216 case PRIV_IPC_MSGSIZE:
3220 * Jail operations within a jail work on child jails.
3222 case PRIV_JAIL_ATTACH:
3224 case PRIV_JAIL_REMOVE:
3227 * Jail implements its own inter-process limits, so allow
3228 * root processes in jail to change scheduling on other
3229 * processes in the same jail. Likewise for signalling.
3231 case PRIV_SCHED_DIFFCRED:
3232 case PRIV_SCHED_CPUSET:
3233 case PRIV_SIGNAL_DIFFCRED:
3234 case PRIV_SIGNAL_SUGID:
3237 * Allow jailed processes to write to sysctls marked as jail
3240 case PRIV_SYSCTL_WRITEJAIL:
3243 * Allow root in jail to manage a variety of quota
3244 * properties. These should likely be conditional on a
3245 * configuration option.
3247 case PRIV_VFS_GETQUOTA:
3248 case PRIV_VFS_SETQUOTA:
3251 * Since Jail relies on chroot() to implement file system
3252 * protections, grant many VFS privileges to root in jail.
3253 * Be careful to exclude mount-related and NFS-related
3257 case PRIV_VFS_WRITE:
3258 case PRIV_VFS_ADMIN:
3260 case PRIV_VFS_LOOKUP:
3261 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3262 case PRIV_VFS_CHFLAGS_DEV:
3263 case PRIV_VFS_CHOWN:
3264 case PRIV_VFS_CHROOT:
3265 case PRIV_VFS_RETAINSUGID:
3266 case PRIV_VFS_FCHROOT:
3268 case PRIV_VFS_SETGID:
3270 case PRIV_VFS_STICKYFILE:
3273 * As in the non-jail case, non-root users are expected to be
3274 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3275 * exists in the jail and they have permission to access it).
3277 case PRIV_KMEM_READ:
3281 * Depending on the global setting, allow privilege of
3282 * setting system flags.
3284 case PRIV_VFS_SYSFLAGS:
3285 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3291 * Depending on the global setting, allow privilege of
3292 * mounting/unmounting file systems.
3294 case PRIV_VFS_MOUNT:
3295 case PRIV_VFS_UNMOUNT:
3296 case PRIV_VFS_MOUNT_NONUSER:
3297 case PRIV_VFS_MOUNT_OWNER:
3298 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3299 cred->cr_prison->pr_enforce_statfs < 2)
3305 * Conditionally allow jailed root to bind reserved ports.
3307 case PRIV_NETINET_RESERVEDPORT:
3308 if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS)
3314 * Allow jailed root to reuse in-use ports.
3316 case PRIV_NETINET_REUSEPORT:
3320 * Allow jailed root to set certain IPv4/6 (option) headers.
3322 case PRIV_NETINET_SETHDROPTS:
3326 * Conditionally allow creating raw sockets in jail.
3328 case PRIV_NETINET_RAW:
3329 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3335 * Since jail implements its own visibility limits on netstat
3336 * sysctls, allow getcred. This allows identd to work in
3339 case PRIV_NETINET_GETCRED:
3343 * Allow jailed root to set loginclass.
3345 case PRIV_PROC_SETLOGINCLASS:
3350 * In all remaining cases, deny the privilege request. This
3351 * includes almost all network privileges, many system
3352 * configuration privileges.
3359 * Return the part of pr2's name that is relative to pr1, or the whole name
3360 * if it does not directly follow.
3364 prison_name(struct prison *pr1, struct prison *pr2)
3368 /* Jails see themselves as "0" (if they see themselves at all). */
3371 name = pr2->pr_name;
3372 if (prison_ischild(pr1, pr2)) {
3374 * pr1 isn't locked (and allprison_lock may not be either)
3375 * so its length can't be counted on. But the number of dots
3376 * can be counted on - and counted.
3378 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3379 name = strchr(name, '.') + 1;
3385 * Return the part of pr2's path that is relative to pr1, or the whole path
3386 * if it does not directly follow.
3389 prison_path(struct prison *pr1, struct prison *pr2)
3391 char *path1, *path2;
3394 path1 = pr1->pr_path;
3395 path2 = pr2->pr_path;
3396 if (!strcmp(path1, "/"))
3398 len1 = strlen(path1);
3399 if (strncmp(path1, path2, len1))
3401 if (path2[len1] == '\0')
3403 if (path2[len1] == '/')
3404 return (path2 + len1);
3410 * Jail-related sysctls.
3412 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3416 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3419 struct prison *pr, *cpr;
3421 struct in_addr *ip4 = NULL;
3425 struct in6_addr *ip6 = NULL;
3430 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3431 pr = req->td->td_ucred->cr_prison;
3433 sx_slock(&allprison_lock);
3434 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3435 #if defined(INET) || defined(INET6)
3438 mtx_lock(&cpr->pr_mtx);
3440 if (cpr->pr_ip4s > 0) {
3441 if (ip4s < cpr->pr_ip4s) {
3442 ip4s = cpr->pr_ip4s;
3443 mtx_unlock(&cpr->pr_mtx);
3444 ip4 = realloc(ip4, ip4s *
3445 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3448 bcopy(cpr->pr_ip4, ip4,
3449 cpr->pr_ip4s * sizeof(struct in_addr));
3453 if (cpr->pr_ip6s > 0) {
3454 if (ip6s < cpr->pr_ip6s) {
3455 ip6s = cpr->pr_ip6s;
3456 mtx_unlock(&cpr->pr_mtx);
3457 ip6 = realloc(ip6, ip6s *
3458 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3461 bcopy(cpr->pr_ip6, ip6,
3462 cpr->pr_ip6s * sizeof(struct in6_addr));
3465 if (cpr->pr_ref == 0) {
3466 mtx_unlock(&cpr->pr_mtx);
3469 bzero(xp, sizeof(*xp));
3470 xp->pr_version = XPRISON_VERSION;
3471 xp->pr_id = cpr->pr_id;
3472 xp->pr_state = cpr->pr_uref > 0
3473 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3474 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3475 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3476 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3478 xp->pr_ip4s = cpr->pr_ip4s;
3481 xp->pr_ip6s = cpr->pr_ip6s;
3483 mtx_unlock(&cpr->pr_mtx);
3484 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3488 if (xp->pr_ip4s > 0) {
3489 error = SYSCTL_OUT(req, ip4,
3490 xp->pr_ip4s * sizeof(struct in_addr));
3496 if (xp->pr_ip6s > 0) {
3497 error = SYSCTL_OUT(req, ip6,
3498 xp->pr_ip6s * sizeof(struct in6_addr));
3504 sx_sunlock(&allprison_lock);
3515 SYSCTL_OID(_security_jail, OID_AUTO, list,
3516 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3517 sysctl_jail_list, "S", "List of active jails");
3520 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3524 injail = jailed(req->td->td_ucred);
3525 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3530 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3531 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3532 sysctl_jail_jailed, "I", "Process in jail?");
3535 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
3537 int error, havevnet;
3539 struct ucred *cred = req->td->td_ucred;
3541 havevnet = jailed(cred) && prison_owns_vnet(cred);
3545 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
3550 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
3551 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3552 sysctl_jail_vnet, "I", "Jail owns VNET?");
3554 #if defined(INET) || defined(INET6)
3555 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3556 &jail_max_af_ips, 0,
3557 "Number of IP addresses a jail may have at most per address family (deprecated)");
3561 * Default parameters for jail(2) compatibility. For historical reasons,
3562 * the sysctl names have varying similarity to the parameter names. Prisons
3563 * just see their own parameters, and can't change them.
3566 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3569 int allow, error, i;
3571 pr = req->td->td_ucred->cr_prison;
3572 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3574 /* Get the current flag value, and convert it to a boolean. */
3575 i = (allow & arg2) ? 1 : 0;
3578 error = sysctl_handle_int(oidp, &i, 0, req);
3579 if (error || !req->newptr)
3585 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3588 mtx_lock(&prison0.pr_mtx);
3589 jail_default_allow = (jail_default_allow & ~arg2) | i;
3590 mtx_unlock(&prison0.pr_mtx);
3594 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3595 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3596 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3597 "Processes in jail can set their hostnames (deprecated)");
3598 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3599 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3600 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3601 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)");
3602 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3603 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3604 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3605 "Processes in jail can use System V IPC primitives (deprecated)");
3606 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3607 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3608 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3609 "Prison root can create raw sockets (deprecated)");
3610 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3611 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3612 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3613 "Processes in jail can alter system file flags (deprecated)");
3614 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3615 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3616 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3617 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)");
3618 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
3619 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3620 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
3621 "Processes in jail can mount the devfs file system (deprecated)");
3622 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
3623 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3624 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
3625 "Processes in jail can mount the fdescfs file system (deprecated)");
3626 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
3627 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3628 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
3629 "Processes in jail can mount the nullfs file system (deprecated)");
3630 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
3631 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3632 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
3633 "Processes in jail can mount the procfs file system (deprecated)");
3634 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
3635 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3636 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
3637 "Processes in jail can mount the linprocfs file system (deprecated)");
3638 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
3639 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3640 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
3641 "Processes in jail can mount the linsysfs file system (deprecated)");
3642 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
3643 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3644 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
3645 "Processes in jail can mount the tmpfs file system (deprecated)");
3646 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
3647 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3648 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
3649 "Processes in jail can mount the zfs file system (deprecated)");
3652 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3657 pr = req->td->td_ucred->cr_prison;
3658 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3659 error = sysctl_handle_int(oidp, &level, 0, req);
3660 if (error || !req->newptr)
3662 *(int *)arg1 = level;
3666 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3667 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3668 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3669 sysctl_jail_default_level, "I",
3670 "Processes in jail cannot see all mounted file systems (deprecated)");
3672 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
3673 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
3674 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
3675 sysctl_jail_default_level, "I",
3676 "Ruleset for the devfs filesystem in jail (deprecated)");
3679 * Nodes to describe jail parameters. Maximum length of string parameters
3680 * is returned in the string itself, and the other parameters exist merely
3681 * to make themselves and their types known.
3683 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
3687 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3694 switch (oidp->oid_kind & CTLTYPE)
3700 if (!(req->flags & SCTL_MASK32))
3702 return (SYSCTL_OUT(req, &l, sizeof(l)));
3706 return (SYSCTL_OUT(req, &i, sizeof(i)));
3707 case CTLTYPE_STRING:
3708 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
3710 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3711 case CTLTYPE_STRUCT:
3713 return (SYSCTL_OUT(req, &s, sizeof(s)));
3719 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
3720 * jail creation time but cannot be changed in an existing jail.
3722 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3723 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3724 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3725 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3726 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3727 "I", "Jail secure level");
3728 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
3729 "Jail value for kern.osreldate and uname -K");
3730 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
3731 "Jail value for kern.osrelease and uname -r");
3732 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3733 "I", "Jail cannot see all mounted file systems");
3734 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
3735 "I", "Ruleset for in-jail devfs mounts");
3736 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3737 "B", "Jail persistence");
3739 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3740 "E,jailsys", "Virtual network stack");
3742 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3743 "B", "Jail is in the process of shutting down");
3745 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
3746 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
3747 "I", "Current number of child jails");
3748 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
3749 "I", "Maximum number of child jails");
3751 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
3752 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3754 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3755 "Jail NIS domainname");
3756 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3758 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3759 "LU", "Jail host ID");
3761 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3762 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3765 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
3766 "Jail IPv4 address virtualization");
3767 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3768 "S,in_addr,a", "Jail IPv4 addresses");
3769 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3770 "B", "Do (not) use IPv4 source address selection rather than the "
3771 "primary jail IPv4 address.");
3774 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
3775 "Jail IPv6 address virtualization");
3776 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3777 "S,in6_addr,a", "Jail IPv6 addresses");
3778 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
3779 "B", "Do (not) use IPv6 source address selection rather than the "
3780 "primary jail IPv6 address.");
3783 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3784 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3785 "B", "Jail may set hostname");
3786 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3787 "B", "Jail may use SYSV IPC");
3788 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3789 "B", "Jail may create raw sockets");
3790 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3791 "B", "Jail may alter system file flags");
3792 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3793 "B", "Jail may set file quotas");
3794 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3795 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3796 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW,
3797 "B", "Jail may bind sockets to reserved ports");
3799 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
3800 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
3801 "B", "Jail may mount/unmount jail-friendly file systems in general");
3802 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
3803 "B", "Jail may mount the devfs file system");
3804 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
3805 "B", "Jail may mount the fdescfs file system");
3806 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
3807 "B", "Jail may mount the nullfs file system");
3808 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
3809 "B", "Jail may mount the procfs file system");
3810 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
3811 "B", "Jail may mount the linprocfs file system");
3812 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
3813 "B", "Jail may mount the linsysfs file system");
3814 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
3815 "B", "Jail may mount the tmpfs file system");
3816 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
3817 "B", "Jail may mount the zfs file system");
3821 prison_racct_foreach(void (*callback)(struct racct *racct,
3822 void *arg2, void *arg3), void (*pre)(void), void (*post)(void),
3823 void *arg2, void *arg3)
3825 struct prison_racct *prr;
3827 ASSERT_RACCT_ENABLED();
3829 sx_slock(&allprison_lock);
3832 LIST_FOREACH(prr, &allprison_racct, prr_next)
3833 (callback)(prr->prr_racct, arg2, arg3);
3836 sx_sunlock(&allprison_lock);
3839 static struct prison_racct *
3840 prison_racct_find_locked(const char *name)
3842 struct prison_racct *prr;
3844 ASSERT_RACCT_ENABLED();
3845 sx_assert(&allprison_lock, SA_XLOCKED);
3847 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
3850 LIST_FOREACH(prr, &allprison_racct, prr_next) {
3851 if (strcmp(name, prr->prr_name) != 0)
3854 /* Found prison_racct with a matching name? */
3855 prison_racct_hold(prr);
3859 /* Add new prison_racct. */
3860 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
3861 racct_create(&prr->prr_racct);
3863 strcpy(prr->prr_name, name);
3864 refcount_init(&prr->prr_refcount, 1);
3865 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
3870 struct prison_racct *
3871 prison_racct_find(const char *name)
3873 struct prison_racct *prr;
3875 ASSERT_RACCT_ENABLED();
3877 sx_xlock(&allprison_lock);
3878 prr = prison_racct_find_locked(name);
3879 sx_xunlock(&allprison_lock);
3884 prison_racct_hold(struct prison_racct *prr)
3887 ASSERT_RACCT_ENABLED();
3889 refcount_acquire(&prr->prr_refcount);
3893 prison_racct_free_locked(struct prison_racct *prr)
3896 ASSERT_RACCT_ENABLED();
3897 sx_assert(&allprison_lock, SA_XLOCKED);
3899 if (refcount_release(&prr->prr_refcount)) {
3900 racct_destroy(&prr->prr_racct);
3901 LIST_REMOVE(prr, prr_next);
3902 free(prr, M_PRISON_RACCT);
3907 prison_racct_free(struct prison_racct *prr)
3911 ASSERT_RACCT_ENABLED();
3912 sx_assert(&allprison_lock, SA_UNLOCKED);
3914 old = prr->prr_refcount;
3915 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
3918 sx_xlock(&allprison_lock);
3919 prison_racct_free_locked(prr);
3920 sx_xunlock(&allprison_lock);
3924 prison_racct_attach(struct prison *pr)
3926 struct prison_racct *prr;
3928 ASSERT_RACCT_ENABLED();
3929 sx_assert(&allprison_lock, SA_XLOCKED);
3931 prr = prison_racct_find_locked(pr->pr_name);
3932 KASSERT(prr != NULL, ("cannot find prison_racct"));
3934 pr->pr_prison_racct = prr;
3938 * Handle jail renaming. From the racct point of view, renaming means
3939 * moving from one prison_racct to another.
3942 prison_racct_modify(struct prison *pr)
3946 struct prison_racct *oldprr;
3948 ASSERT_RACCT_ENABLED();
3950 sx_slock(&allproc_lock);
3951 sx_xlock(&allprison_lock);
3953 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
3954 sx_xunlock(&allprison_lock);
3955 sx_sunlock(&allproc_lock);
3959 oldprr = pr->pr_prison_racct;
3960 pr->pr_prison_racct = NULL;
3962 prison_racct_attach(pr);
3965 * Move resource utilisation records.
3967 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
3971 * Force rctl to reattach rules to processes.
3973 FOREACH_PROC_IN_SYSTEM(p) {
3975 cred = crhold(p->p_ucred);
3977 rctl_proc_ucred_changed(p, cred);
3982 sx_sunlock(&allproc_lock);
3983 prison_racct_free_locked(oldprr);
3984 sx_xunlock(&allprison_lock);
3988 prison_racct_detach(struct prison *pr)
3991 ASSERT_RACCT_ENABLED();
3992 sx_assert(&allprison_lock, SA_UNLOCKED);
3994 if (pr->pr_prison_racct == NULL)
3996 prison_racct_free(pr->pr_prison_racct);
3997 pr->pr_prison_racct = NULL;
4004 db_show_prison(struct prison *pr)
4006 struct bool_flags *bf;
4007 struct jailsys_flags *jsf;
4008 #if defined(INET) || defined(INET6)
4013 char ip4buf[INET_ADDRSTRLEN];
4016 char ip6buf[INET6_ADDRSTRLEN];
4019 db_printf("prison %p:\n", pr);
4020 db_printf(" jid = %d\n", pr->pr_id);
4021 db_printf(" name = %s\n", pr->pr_name);
4022 db_printf(" parent = %p\n", pr->pr_parent);
4023 db_printf(" ref = %d\n", pr->pr_ref);
4024 db_printf(" uref = %d\n", pr->pr_uref);
4025 db_printf(" path = %s\n", pr->pr_path);
4026 db_printf(" cpuset = %d\n", pr->pr_cpuset
4027 ? pr->pr_cpuset->cs_id : -1);
4029 db_printf(" vnet = %p\n", pr->pr_vnet);
4031 db_printf(" root = %p\n", pr->pr_root);
4032 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4033 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4034 db_printf(" children.max = %d\n", pr->pr_childmax);
4035 db_printf(" children.cur = %d\n", pr->pr_childcount);
4036 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4037 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4038 db_printf(" flags = 0x%x", pr->pr_flags);
4039 for (bf = pr_flag_bool; bf < pr_flag_bool + nitems(pr_flag_bool); bf++)
4040 if (pr->pr_flags & bf->flag)
4041 db_printf(" %s", bf->name);
4042 for (jsf = pr_flag_jailsys;
4043 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys);
4045 f = pr->pr_flags & (jsf->disable | jsf->new);
4046 db_printf(" %-16s= %s\n", jsf->name,
4047 (f != 0 && f == jsf->disable) ? "disable"
4048 : (f == jsf->new) ? "new"
4051 db_printf(" allow = 0x%x", pr->pr_allow);
4052 for (bf = pr_flag_allow;
4053 bf < pr_flag_allow + nitems(pr_flag_allow);
4055 if (pr->pr_allow & bf->flag)
4056 db_printf(" %s", bf->name);
4058 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4059 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4060 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4061 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4062 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4064 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4065 for (ii = 0; ii < pr->pr_ip4s; ii++)
4066 db_printf(" %s %s\n",
4067 ii == 0 ? "ip4.addr =" : " ",
4068 inet_ntoa_r(pr->pr_ip4[ii], ip4buf));
4071 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4072 for (ii = 0; ii < pr->pr_ip6s; ii++)
4073 db_printf(" %s %s\n",
4074 ii == 0 ? "ip6.addr =" : " ",
4075 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4079 DB_SHOW_COMMAND(prison, db_show_prison_command)
4085 * Show all prisons in the list, and prison0 which is not
4088 db_show_prison(&prison0);
4089 if (!db_pager_quit) {
4090 TAILQ_FOREACH(pr, &allprison, pr_list) {
4102 /* Look for a prison with the ID and with references. */
4103 TAILQ_FOREACH(pr, &allprison, pr_list)
4104 if (pr->pr_id == addr && pr->pr_ref > 0)
4107 /* Look again, without requiring a reference. */
4108 TAILQ_FOREACH(pr, &allprison, pr_list)
4109 if (pr->pr_id == addr)
4112 /* Assume address points to a valid prison. */
4113 pr = (struct prison *)addr;