2 * ----------------------------------------------------------------------------
3 * "THE BEER-WARE LICENSE" (Revision 42):
4 * <phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you
5 * can do whatever you want with this stuff. If we meet some day, and you think
6 * this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
7 * ----------------------------------------------------------------------------
10 #include <sys/cdefs.h>
11 __FBSDID("$FreeBSD$");
15 #include <sys/param.h>
16 #include <sys/types.h>
17 #include <sys/kernel.h>
18 #include <sys/systm.h>
19 #include <sys/errno.h>
20 #include <sys/sysproto.h>
21 #include <sys/malloc.h>
24 #include <sys/taskqueue.h>
27 #include <sys/mutex.h>
28 #include <sys/namei.h>
29 #include <sys/mount.h>
30 #include <sys/queue.h>
31 #include <sys/socket.h>
32 #include <sys/syscallsubr.h>
33 #include <sys/sysctl.h>
34 #include <sys/vnode.h>
36 #include <netinet/in.h>
38 #include <security/mac/mac_framework.h>
40 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
42 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
45 int jail_set_hostname_allowed = 1;
46 SYSCTL_INT(_security_jail, OID_AUTO, set_hostname_allowed, CTLFLAG_RW,
47 &jail_set_hostname_allowed, 0,
48 "Processes in jail can set their hostnames");
50 int jail_socket_unixiproute_only = 1;
51 SYSCTL_INT(_security_jail, OID_AUTO, socket_unixiproute_only, CTLFLAG_RW,
52 &jail_socket_unixiproute_only, 0,
53 "Processes in jail are limited to creating UNIX/IPv4/route sockets only");
55 int jail_sysvipc_allowed = 0;
56 SYSCTL_INT(_security_jail, OID_AUTO, sysvipc_allowed, CTLFLAG_RW,
57 &jail_sysvipc_allowed, 0,
58 "Processes in jail can use System V IPC primitives");
60 static int jail_enforce_statfs = 2;
61 SYSCTL_INT(_security_jail, OID_AUTO, enforce_statfs, CTLFLAG_RW,
62 &jail_enforce_statfs, 0,
63 "Processes in jail cannot see all mounted file systems");
65 int jail_allow_raw_sockets = 0;
66 SYSCTL_INT(_security_jail, OID_AUTO, allow_raw_sockets, CTLFLAG_RW,
67 &jail_allow_raw_sockets, 0,
68 "Prison root can create raw sockets");
70 int jail_chflags_allowed = 0;
71 SYSCTL_INT(_security_jail, OID_AUTO, chflags_allowed, CTLFLAG_RW,
72 &jail_chflags_allowed, 0,
73 "Processes in jail can alter system file flags");
75 /* allprison, lastprid, and prisoncount are protected by allprison_mtx. */
76 struct prisonlist allprison;
77 struct mtx allprison_mtx;
81 static void init_prison(void *);
82 static void prison_complete(void *context, int pending);
83 static struct prison *prison_find(int);
84 static int sysctl_jail_list(SYSCTL_HANDLER_ARGS);
87 init_prison(void *data __unused)
90 mtx_init(&allprison_mtx, "allprison", NULL, MTX_DEF);
91 LIST_INIT(&allprison);
94 SYSINIT(prison, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_prison, NULL);
102 jail(struct thread *td, struct jail_args *uap)
105 struct prison *pr, *tpr;
107 struct jail_attach_args jaa;
108 int vfslocked, error, tryprid;
110 error = copyin(uap->jail, &j, sizeof(j));
116 MALLOC(pr, struct prison *, sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
117 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF);
119 error = copyinstr(j.path, &pr->pr_path, sizeof(pr->pr_path), 0);
122 NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW | LOCKLEAF, UIO_SYSSPACE,
127 vfslocked = NDHASGIANT(&nd);
128 pr->pr_root = nd.ni_vp;
129 VOP_UNLOCK(nd.ni_vp, 0, td);
130 NDFREE(&nd, NDF_ONLY_PNBUF);
131 VFS_UNLOCK_GIANT(vfslocked);
132 error = copyinstr(j.hostname, &pr->pr_host, sizeof(pr->pr_host), 0);
135 pr->pr_ip = j.ip_number;
137 pr->pr_securelevel = securelevel;
139 /* Determine next pr_id and add prison to allprison list. */
140 mtx_lock(&allprison_mtx);
141 tryprid = lastprid + 1;
142 if (tryprid == JAIL_MAX)
145 LIST_FOREACH(tpr, &allprison, pr_list) {
146 if (tpr->pr_id == tryprid) {
148 if (tryprid == JAIL_MAX) {
149 mtx_unlock(&allprison_mtx);
156 pr->pr_id = jaa.jid = lastprid = tryprid;
157 LIST_INSERT_HEAD(&allprison, pr, pr_list);
159 mtx_unlock(&allprison_mtx);
161 error = jail_attach(td, &jaa);
164 mtx_lock(&pr->pr_mtx);
166 mtx_unlock(&pr->pr_mtx);
167 td->td_retval[0] = jaa.jid;
170 mtx_lock(&allprison_mtx);
171 LIST_REMOVE(pr, pr_list);
173 mtx_unlock(&allprison_mtx);
175 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
177 VFS_UNLOCK_GIANT(vfslocked);
179 mtx_destroy(&pr->pr_mtx);
185 * struct jail_attach_args {
190 jail_attach(struct thread *td, struct jail_attach_args *uap)
193 struct ucred *newcred, *oldcred;
195 int vfslocked, error;
198 * XXX: Note that there is a slight race here if two threads
199 * in the same privileged process attempt to attach to two
200 * different jails at the same time. It is important for
201 * user processes not to do this, or they might end up with
202 * a process root from one prison, but attached to the jail
205 error = priv_check(td, PRIV_JAIL_ATTACH);
210 mtx_lock(&allprison_mtx);
211 pr = prison_find(uap->jid);
213 mtx_unlock(&allprison_mtx);
217 mtx_unlock(&pr->pr_mtx);
218 mtx_unlock(&allprison_mtx);
220 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
221 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY, td);
222 if ((error = change_dir(pr->pr_root, td)) != 0)
225 if ((error = mac_check_vnode_chroot(td->td_ucred, pr->pr_root)))
228 VOP_UNLOCK(pr->pr_root, 0, td);
229 change_root(pr->pr_root, td);
230 VFS_UNLOCK_GIANT(vfslocked);
234 oldcred = p->p_ucred;
236 crcopy(newcred, oldcred);
237 newcred->cr_prison = pr;
238 p->p_ucred = newcred;
243 VOP_UNLOCK(pr->pr_root, 0, td);
244 VFS_UNLOCK_GIANT(vfslocked);
245 mtx_lock(&pr->pr_mtx);
247 mtx_unlock(&pr->pr_mtx);
252 * Returns a locked prison instance, or NULL on failure.
254 static struct prison *
255 prison_find(int prid)
259 mtx_assert(&allprison_mtx, MA_OWNED);
260 LIST_FOREACH(pr, &allprison, pr_list) {
261 if (pr->pr_id == prid) {
262 mtx_lock(&pr->pr_mtx);
270 prison_free(struct prison *pr)
273 mtx_lock(&allprison_mtx);
274 mtx_lock(&pr->pr_mtx);
276 if (pr->pr_ref == 0) {
277 LIST_REMOVE(pr, pr_list);
278 mtx_unlock(&pr->pr_mtx);
280 mtx_unlock(&allprison_mtx);
282 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
283 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
286 mtx_unlock(&pr->pr_mtx);
287 mtx_unlock(&allprison_mtx);
291 prison_complete(void *context, int pending)
296 pr = (struct prison *)context;
298 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
300 VFS_UNLOCK_GIANT(vfslocked);
302 mtx_destroy(&pr->pr_mtx);
303 if (pr->pr_linux != NULL)
304 FREE(pr->pr_linux, M_PRISON);
309 prison_hold(struct prison *pr)
312 mtx_lock(&pr->pr_mtx);
314 mtx_unlock(&pr->pr_mtx);
318 prison_getip(struct ucred *cred)
321 return (cred->cr_prison->pr_ip);
325 prison_ip(struct ucred *cred, int flag, u_int32_t *ip)
335 if (tmp == INADDR_ANY) {
337 *ip = cred->cr_prison->pr_ip;
339 *ip = htonl(cred->cr_prison->pr_ip);
342 if (tmp == INADDR_LOOPBACK) {
344 *ip = cred->cr_prison->pr_ip;
346 *ip = htonl(cred->cr_prison->pr_ip);
349 if (cred->cr_prison->pr_ip != tmp)
355 prison_remote_ip(struct ucred *cred, int flag, u_int32_t *ip)
365 if (tmp == INADDR_LOOPBACK) {
367 *ip = cred->cr_prison->pr_ip;
369 *ip = htonl(cred->cr_prison->pr_ip);
376 prison_if(struct ucred *cred, struct sockaddr *sa)
378 struct sockaddr_in *sai;
381 sai = (struct sockaddr_in *)sa;
382 if ((sai->sin_family != AF_INET) && jail_socket_unixiproute_only)
384 else if (sai->sin_family != AF_INET)
386 else if (cred->cr_prison->pr_ip != ntohl(sai->sin_addr.s_addr))
394 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
397 prison_check(struct ucred *cred1, struct ucred *cred2)
403 if (cred2->cr_prison != cred1->cr_prison)
411 * Return 1 if the passed credential is in a jail, otherwise 0.
414 jailed(struct ucred *cred)
417 return (cred->cr_prison != NULL);
421 * Return the correct hostname for the passed credential.
424 getcredhostname(struct ucred *cred, char *buf, size_t size)
428 mtx_lock(&cred->cr_prison->pr_mtx);
429 strlcpy(buf, cred->cr_prison->pr_host, size);
430 mtx_unlock(&cred->cr_prison->pr_mtx);
432 strlcpy(buf, hostname, size);
436 * Determine whether the subject represented by cred can "see"
437 * status of a mount point.
438 * Returns: 0 for permitted, ENOENT otherwise.
439 * XXX: This function should be called cr_canseemount() and should be
440 * placed in kern_prot.c.
443 prison_canseemount(struct ucred *cred, struct mount *mp)
449 if (!jailed(cred) || jail_enforce_statfs == 0)
451 pr = cred->cr_prison;
452 if (pr->pr_root->v_mount == mp)
454 if (jail_enforce_statfs == 2)
457 * If jail's chroot directory is set to "/" we should be able to see
458 * all mount-points from inside a jail.
459 * This is ugly check, but this is the only situation when jail's
460 * directory ends with '/'.
462 if (strcmp(pr->pr_path, "/") == 0)
464 len = strlen(pr->pr_path);
466 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
469 * Be sure that we don't have situation where jail's root directory
470 * is "/some/path" and mount point is "/some/pathpath".
472 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
478 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
480 char jpath[MAXPATHLEN];
484 if (!jailed(cred) || jail_enforce_statfs == 0)
486 pr = cred->cr_prison;
487 if (prison_canseemount(cred, mp) != 0) {
488 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
489 strlcpy(sp->f_mntonname, "[restricted]",
490 sizeof(sp->f_mntonname));
493 if (pr->pr_root->v_mount == mp) {
495 * Clear current buffer data, so we are sure nothing from
496 * the valid path left there.
498 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
499 *sp->f_mntonname = '/';
503 * If jail's chroot directory is set to "/" we should be able to see
504 * all mount-points from inside a jail.
506 if (strcmp(pr->pr_path, "/") == 0)
508 len = strlen(pr->pr_path);
509 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
511 * Clear current buffer data, so we are sure nothing from
512 * the valid path left there.
514 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
515 if (*jpath == '\0') {
516 /* Should never happen. */
517 *sp->f_mntonname = '/';
519 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
524 * Check with permission for a specific privilege is granted within jail. We
525 * have a specific list of accepted privileges; the rest are denied.
528 prison_priv_check(struct ucred *cred, int priv)
537 * Allow ktrace privileges for root in jail.
543 * Allow jailed processes to configure audit identity and
544 * submit audit records (login, etc). In the future we may
545 * want to further refine the relationship between audit and
548 case PRIV_AUDIT_GETAUDIT:
549 case PRIV_AUDIT_SETAUDIT:
550 case PRIV_AUDIT_SUBMIT:
554 * Allow jailed processes to manipulate process UNIX
555 * credentials in any way they see fit.
557 case PRIV_CRED_SETUID:
558 case PRIV_CRED_SETEUID:
559 case PRIV_CRED_SETGID:
560 case PRIV_CRED_SETEGID:
561 case PRIV_CRED_SETGROUPS:
562 case PRIV_CRED_SETREUID:
563 case PRIV_CRED_SETREGID:
564 case PRIV_CRED_SETRESUID:
565 case PRIV_CRED_SETRESGID:
568 * Jail implements visibility constraints already, so allow
569 * jailed root to override uid/gid-based constraints.
571 case PRIV_SEEOTHERGIDS:
572 case PRIV_SEEOTHERUIDS:
575 * Jail implements inter-process debugging limits already, so
576 * allow jailed root various debugging privileges.
578 case PRIV_DEBUG_DIFFCRED:
579 case PRIV_DEBUG_SUGID:
580 case PRIV_DEBUG_UNPRIV:
583 * Allow jail to set various resource limits and login
584 * properties, and for now, exceed process resource limits.
586 case PRIV_PROC_LIMIT:
587 case PRIV_PROC_SETLOGIN:
588 case PRIV_PROC_SETRLIMIT:
591 * System V and POSIX IPC privileges are granted in jail.
596 case PRIV_IPC_MSGSIZE:
600 * Jail implements its own inter-process limits, so allow
601 * root processes in jail to change scheduling on other
602 * processes in the same jail. Likewise for signalling.
604 case PRIV_SCHED_DIFFCRED:
605 case PRIV_SIGNAL_DIFFCRED:
606 case PRIV_SIGNAL_SUGID:
609 * Allow jailed processes to write to sysctls marked as jail
612 case PRIV_SYSCTL_WRITEJAIL:
615 * Allow root in jail to manage a variety of quota
616 * properties. These should likely be conditional on a
617 * configuration option.
619 case PRIV_VFS_GETQUOTA:
620 case PRIV_VFS_SETQUOTA:
623 * Since Jail relies on chroot() to implement file system
624 * protections, grant many VFS privileges to root in jail.
625 * Be careful to exclude mount-related and NFS-related
632 case PRIV_VFS_LOOKUP:
633 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
634 case PRIV_VFS_CHFLAGS_DEV:
636 case PRIV_VFS_CHROOT:
637 case PRIV_VFS_RETAINSUGID:
638 case PRIV_VFS_FCHROOT:
640 case PRIV_VFS_SETGID:
641 case PRIV_VFS_STICKYFILE:
645 * Depending on the global setting, allow privilege of
646 * setting system flags.
648 case PRIV_VFS_SYSFLAGS:
649 if (jail_chflags_allowed)
655 * Allow jailed root to bind reserved ports.
657 case PRIV_NETINET_RESERVEDPORT:
661 * Conditionally allow creating raw sockets in jail.
663 case PRIV_NETINET_RAW:
664 if (jail_allow_raw_sockets)
670 * Since jail implements its own visibility limits on netstat
671 * sysctls, allow getcred. This allows identd to work in
674 case PRIV_NETINET_GETCRED:
679 * In all remaining cases, deny the privilege request. This
680 * includes almost all network privileges, many system
681 * configuration privileges.
688 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
690 struct xprison *xp, *sxp;
694 if (jailed(req->td->td_ucred))
697 mtx_lock(&allprison_mtx);
699 mtx_unlock(&allprison_mtx);
704 sxp = xp = malloc(sizeof(*xp) * count, M_TEMP, M_WAITOK | M_ZERO);
705 mtx_lock(&allprison_mtx);
706 if (count != prisoncount) {
707 mtx_unlock(&allprison_mtx);
712 LIST_FOREACH(pr, &allprison, pr_list) {
713 mtx_lock(&pr->pr_mtx);
714 xp->pr_version = XPRISON_VERSION;
715 xp->pr_id = pr->pr_id;
716 strlcpy(xp->pr_path, pr->pr_path, sizeof(xp->pr_path));
717 strlcpy(xp->pr_host, pr->pr_host, sizeof(xp->pr_host));
718 xp->pr_ip = pr->pr_ip;
719 mtx_unlock(&pr->pr_mtx);
722 mtx_unlock(&allprison_mtx);
724 error = SYSCTL_OUT(req, sxp, sizeof(*sxp) * count);
729 SYSCTL_OID(_security_jail, OID_AUTO, list, CTLTYPE_STRUCT | CTLFLAG_RD,
730 NULL, 0, sysctl_jail_list, "S", "List of active jails");
733 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
737 injail = jailed(req->td->td_ucred);
738 error = SYSCTL_OUT(req, &injail, sizeof(injail));
742 SYSCTL_PROC(_security_jail, OID_AUTO, jailed, CTLTYPE_INT | CTLFLAG_RD,
743 NULL, 0, sysctl_jail_jailed, "I", "Process in jail?");