xref: /freebsd/sys/kern/kern_jail.c (revision ceaec73d406831b1251babb61675df0a1aa54a31)
1 /*-
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  * ----------------------------------------------------------------------------
8  */
9 
10 #include <sys/cdefs.h>
11 __FBSDID("$FreeBSD$");
12 
13 #include "opt_mac.h"
14 
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/mac.h>
22 #include <sys/malloc.h>
23 #include <sys/proc.h>
24 #include <sys/taskqueue.h>
25 #include <sys/jail.h>
26 #include <sys/lock.h>
27 #include <sys/mutex.h>
28 #include <sys/namei.h>
29 #include <sys/queue.h>
30 #include <sys/socket.h>
31 #include <sys/syscallsubr.h>
32 #include <sys/sysctl.h>
33 #include <sys/vnode.h>
34 #include <net/if.h>
35 #include <netinet/in.h>
36 
37 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
38 
39 SYSCTL_DECL(_security);
40 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
41     "Jail rules");
42 
43 int	jail_set_hostname_allowed = 1;
44 SYSCTL_INT(_security_jail, OID_AUTO, set_hostname_allowed, CTLFLAG_RW,
45     &jail_set_hostname_allowed, 0,
46     "Processes in jail can set their hostnames");
47 
48 int	jail_socket_unixiproute_only = 1;
49 SYSCTL_INT(_security_jail, OID_AUTO, socket_unixiproute_only, CTLFLAG_RW,
50     &jail_socket_unixiproute_only, 0,
51     "Processes in jail are limited to creating UNIX/IPv4/route sockets only");
52 
53 int	jail_sysvipc_allowed = 0;
54 SYSCTL_INT(_security_jail, OID_AUTO, sysvipc_allowed, CTLFLAG_RW,
55     &jail_sysvipc_allowed, 0,
56     "Processes in jail can use System V IPC primitives");
57 
58 int	jail_getfsstatroot_only = 1;
59 SYSCTL_INT(_security_jail, OID_AUTO, getfsstatroot_only, CTLFLAG_RW,
60     &jail_getfsstatroot_only, 0,
61     "Processes see only their root file system in getfsstat()");
62 
63 int	jail_allow_raw_sockets = 0;
64 SYSCTL_INT(_security_jail, OID_AUTO, allow_raw_sockets, CTLFLAG_RW,
65     &jail_allow_raw_sockets, 0,
66     "Prison root can create raw sockets");
67 
68 int	jail_chflags_allowed = 0;
69 SYSCTL_INT(_security_jail, OID_AUTO, chflags_allowed, CTLFLAG_RW,
70     &jail_chflags_allowed, 0,
71     "Processes in jail can alter system file flags");
72 
73 /* allprison, lastprid, and prisoncount are protected by allprison_mtx. */
74 struct	prisonlist allprison;
75 struct	mtx allprison_mtx;
76 int	lastprid = 0;
77 int	prisoncount = 0;
78 
79 static void		 init_prison(void *);
80 static void		 prison_complete(void *context, int pending);
81 static struct prison	*prison_find(int);
82 static int		 sysctl_jail_list(SYSCTL_HANDLER_ARGS);
83 
84 static void
85 init_prison(void *data __unused)
86 {
87 
88 	mtx_init(&allprison_mtx, "allprison", NULL, MTX_DEF);
89 	LIST_INIT(&allprison);
90 }
91 
92 SYSINIT(prison, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_prison, NULL);
93 
94 /*
95  * MPSAFE
96  *
97  * struct jail_args {
98  *	struct jail *jail;
99  * };
100  */
101 int
102 jail(struct thread *td, struct jail_args *uap)
103 {
104 	struct nameidata nd;
105 	struct prison *pr, *tpr;
106 	struct jail j;
107 	struct jail_attach_args jaa;
108 	int error, tryprid;
109 
110 	error = copyin(uap->jail, &j, sizeof(j));
111 	if (error)
112 		return (error);
113 	if (j.version != 0)
114 		return (EINVAL);
115 
116 	MALLOC(pr, struct prison *, sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
117 	mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF);
118 	pr->pr_ref = 1;
119 	error = copyinstr(j.path, &pr->pr_path, sizeof(pr->pr_path), 0);
120 	if (error)
121 		goto e_killmtx;
122 	mtx_lock(&Giant);
123 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, pr->pr_path, td);
124 	error = namei(&nd);
125 	if (error) {
126 		mtx_unlock(&Giant);
127 		goto e_killmtx;
128 	}
129 	pr->pr_root = nd.ni_vp;
130 	VOP_UNLOCK(nd.ni_vp, 0, td);
131 	NDFREE(&nd, NDF_ONLY_PNBUF);
132 	mtx_unlock(&Giant);
133 	error = copyinstr(j.hostname, &pr->pr_host, sizeof(pr->pr_host), 0);
134 	if (error)
135 		goto e_dropvnref;
136 	pr->pr_ip = j.ip_number;
137 	pr->pr_linux = NULL;
138 	pr->pr_securelevel = securelevel;
139 
140 	/* Determine next pr_id and add prison to allprison list. */
141 	mtx_lock(&allprison_mtx);
142 	tryprid = lastprid + 1;
143 	if (tryprid == JAIL_MAX)
144 		tryprid = 1;
145 next:
146 	LIST_FOREACH(tpr, &allprison, pr_list) {
147 		if (tpr->pr_id == tryprid) {
148 			tryprid++;
149 			if (tryprid == JAIL_MAX) {
150 				mtx_unlock(&allprison_mtx);
151 				error = EAGAIN;
152 				goto e_dropvnref;
153 			}
154 			goto next;
155 		}
156 	}
157 	pr->pr_id = jaa.jid = lastprid = tryprid;
158 	LIST_INSERT_HEAD(&allprison, pr, pr_list);
159 	prisoncount++;
160 	mtx_unlock(&allprison_mtx);
161 
162 	error = jail_attach(td, &jaa);
163 	if (error)
164 		goto e_dropprref;
165 	mtx_lock(&pr->pr_mtx);
166 	pr->pr_ref--;
167 	mtx_unlock(&pr->pr_mtx);
168 	td->td_retval[0] = jaa.jid;
169 	return (0);
170 e_dropprref:
171 	mtx_lock(&allprison_mtx);
172 	LIST_REMOVE(pr, pr_list);
173 	prisoncount--;
174 	mtx_unlock(&allprison_mtx);
175 e_dropvnref:
176 	mtx_lock(&Giant);
177 	vrele(pr->pr_root);
178 	mtx_unlock(&Giant);
179 e_killmtx:
180 	mtx_destroy(&pr->pr_mtx);
181 	FREE(pr, M_PRISON);
182 	return (error);
183 }
184 
185 /*
186  * MPSAFE
187  *
188  * struct jail_attach_args {
189  *	int jid;
190  * };
191  */
192 int
193 jail_attach(struct thread *td, struct jail_attach_args *uap)
194 {
195 	struct proc *p;
196 	struct ucred *newcred, *oldcred;
197 	struct prison *pr;
198 	int error;
199 
200 	/*
201 	 * XXX: Note that there is a slight race here if two threads
202 	 * in the same privileged process attempt to attach to two
203 	 * different jails at the same time.  It is important for
204 	 * user processes not to do this, or they might end up with
205 	 * a process root from one prison, but attached to the jail
206 	 * of another.
207 	 */
208 	error = suser(td);
209 	if (error)
210 		return (error);
211 
212 	p = td->td_proc;
213 	mtx_lock(&allprison_mtx);
214 	pr = prison_find(uap->jid);
215 	if (pr == NULL) {
216 		mtx_unlock(&allprison_mtx);
217 		return (EINVAL);
218 	}
219 	pr->pr_ref++;
220 	mtx_unlock(&pr->pr_mtx);
221 	mtx_unlock(&allprison_mtx);
222 
223 	mtx_lock(&Giant);
224 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY, td);
225 	if ((error = change_dir(pr->pr_root, td)) != 0)
226 		goto e_unlock;
227 #ifdef MAC
228 	if ((error = mac_check_vnode_chroot(td->td_ucred, pr->pr_root)))
229 		goto e_unlock;
230 #endif
231 	VOP_UNLOCK(pr->pr_root, 0, td);
232 	change_root(pr->pr_root, td);
233 	mtx_unlock(&Giant);
234 
235 	newcred = crget();
236 	PROC_LOCK(p);
237 	oldcred = p->p_ucred;
238 	setsugid(p);
239 	crcopy(newcred, oldcred);
240 	newcred->cr_prison = pr;
241 	p->p_ucred = newcred;
242 	PROC_UNLOCK(p);
243 	crfree(oldcred);
244 	return (0);
245 e_unlock:
246 	VOP_UNLOCK(pr->pr_root, 0, td);
247 	mtx_unlock(&Giant);
248 	mtx_lock(&pr->pr_mtx);
249 	pr->pr_ref--;
250 	mtx_unlock(&pr->pr_mtx);
251 	return (error);
252 }
253 
254 /*
255  * Returns a locked prison instance, or NULL on failure.
256  */
257 static struct prison *
258 prison_find(int prid)
259 {
260 	struct prison *pr;
261 
262 	mtx_assert(&allprison_mtx, MA_OWNED);
263 	LIST_FOREACH(pr, &allprison, pr_list) {
264 		if (pr->pr_id == prid) {
265 			mtx_lock(&pr->pr_mtx);
266 			return (pr);
267 		}
268 	}
269 	return (NULL);
270 }
271 
272 void
273 prison_free(struct prison *pr)
274 {
275 
276 	mtx_lock(&allprison_mtx);
277 	mtx_lock(&pr->pr_mtx);
278 	pr->pr_ref--;
279 	if (pr->pr_ref == 0) {
280 		LIST_REMOVE(pr, pr_list);
281 		mtx_unlock(&pr->pr_mtx);
282 		prisoncount--;
283 		mtx_unlock(&allprison_mtx);
284 
285 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
286 		taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
287 		return;
288 	}
289 	mtx_unlock(&pr->pr_mtx);
290 	mtx_unlock(&allprison_mtx);
291 }
292 
293 static void
294 prison_complete(void *context, int pending)
295 {
296 	struct prison *pr;
297 
298 	pr = (struct prison *)context;
299 
300 	mtx_lock(&Giant);
301 	vrele(pr->pr_root);
302 	mtx_unlock(&Giant);
303 
304 	mtx_destroy(&pr->pr_mtx);
305 	if (pr->pr_linux != NULL)
306 		FREE(pr->pr_linux, M_PRISON);
307 	FREE(pr, M_PRISON);
308 }
309 
310 void
311 prison_hold(struct prison *pr)
312 {
313 
314 	mtx_lock(&pr->pr_mtx);
315 	pr->pr_ref++;
316 	mtx_unlock(&pr->pr_mtx);
317 }
318 
319 u_int32_t
320 prison_getip(struct ucred *cred)
321 {
322 
323 	return (cred->cr_prison->pr_ip);
324 }
325 
326 int
327 prison_ip(struct ucred *cred, int flag, u_int32_t *ip)
328 {
329 	u_int32_t tmp;
330 
331 	if (!jailed(cred))
332 		return (0);
333 	if (flag)
334 		tmp = *ip;
335 	else
336 		tmp = ntohl(*ip);
337 	if (tmp == INADDR_ANY) {
338 		if (flag)
339 			*ip = cred->cr_prison->pr_ip;
340 		else
341 			*ip = htonl(cred->cr_prison->pr_ip);
342 		return (0);
343 	}
344 	if (tmp == INADDR_LOOPBACK) {
345 		if (flag)
346 			*ip = cred->cr_prison->pr_ip;
347 		else
348 			*ip = htonl(cred->cr_prison->pr_ip);
349 		return (0);
350 	}
351 	if (cred->cr_prison->pr_ip != tmp)
352 		return (1);
353 	return (0);
354 }
355 
356 void
357 prison_remote_ip(struct ucred *cred, int flag, u_int32_t *ip)
358 {
359 	u_int32_t tmp;
360 
361 	if (!jailed(cred))
362 		return;
363 	if (flag)
364 		tmp = *ip;
365 	else
366 		tmp = ntohl(*ip);
367 	if (tmp == INADDR_LOOPBACK) {
368 		if (flag)
369 			*ip = cred->cr_prison->pr_ip;
370 		else
371 			*ip = htonl(cred->cr_prison->pr_ip);
372 		return;
373 	}
374 	return;
375 }
376 
377 int
378 prison_if(struct ucred *cred, struct sockaddr *sa)
379 {
380 	struct sockaddr_in *sai;
381 	int ok;
382 
383 	sai = (struct sockaddr_in *)sa;
384 	if ((sai->sin_family != AF_INET) && jail_socket_unixiproute_only)
385 		ok = 1;
386 	else if (sai->sin_family != AF_INET)
387 		ok = 0;
388 	else if (cred->cr_prison->pr_ip != ntohl(sai->sin_addr.s_addr))
389 		ok = 1;
390 	else
391 		ok = 0;
392 	return (ok);
393 }
394 
395 /*
396  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
397  */
398 int
399 prison_check(struct ucred *cred1, struct ucred *cred2)
400 {
401 
402 	if (jailed(cred1)) {
403 		if (!jailed(cred2))
404 			return (ESRCH);
405 		if (cred2->cr_prison != cred1->cr_prison)
406 			return (ESRCH);
407 	}
408 
409 	return (0);
410 }
411 
412 /*
413  * Return 1 if the passed credential is in a jail, otherwise 0.
414  */
415 int
416 jailed(struct ucred *cred)
417 {
418 
419 	return (cred->cr_prison != NULL);
420 }
421 
422 /*
423  * Return the correct hostname for the passed credential.
424  */
425 void
426 getcredhostname(struct ucred *cred, char *buf, size_t size)
427 {
428 
429 	if (jailed(cred)) {
430 		mtx_lock(&cred->cr_prison->pr_mtx);
431 		strlcpy(buf, cred->cr_prison->pr_host, size);
432 		mtx_unlock(&cred->cr_prison->pr_mtx);
433 	} else
434 		strlcpy(buf, hostname, size);
435 }
436 
437 /*
438  * Return 1 if the passed credential can "see" the passed mountpoint
439  * when performing a getfsstat(); otherwise, 0.
440  */
441 int
442 prison_check_mount(struct ucred *cred, struct mount *mp)
443 {
444 
445 	if (jail_getfsstatroot_only && cred->cr_prison != NULL) {
446 		if (cred->cr_prison->pr_root->v_mount != mp)
447 			return (0);
448 	}
449 	return (1);
450 }
451 
452 static int
453 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
454 {
455 	struct xprison *xp, *sxp;
456 	struct prison *pr;
457 	int count, error;
458 
459 	mtx_assert(&Giant, MA_OWNED);
460 	if (jailed(req->td->td_ucred))
461 		return (0);
462 retry:
463 	mtx_lock(&allprison_mtx);
464 	count = prisoncount;
465 	mtx_unlock(&allprison_mtx);
466 
467 	if (count == 0)
468 		return (0);
469 
470 	sxp = xp = malloc(sizeof(*xp) * count, M_TEMP, M_WAITOK | M_ZERO);
471 	mtx_lock(&allprison_mtx);
472 	if (count != prisoncount) {
473 		mtx_unlock(&allprison_mtx);
474 		free(sxp, M_TEMP);
475 		goto retry;
476 	}
477 
478 	LIST_FOREACH(pr, &allprison, pr_list) {
479 		mtx_lock(&pr->pr_mtx);
480 		xp->pr_version = XPRISON_VERSION;
481 		xp->pr_id = pr->pr_id;
482 		strlcpy(xp->pr_path, pr->pr_path, sizeof(xp->pr_path));
483 		strlcpy(xp->pr_host, pr->pr_host, sizeof(xp->pr_host));
484 		xp->pr_ip = pr->pr_ip;
485 		mtx_unlock(&pr->pr_mtx);
486 		xp++;
487 	}
488 	mtx_unlock(&allprison_mtx);
489 
490 	error = SYSCTL_OUT(req, sxp, sizeof(*sxp) * count);
491 	free(sxp, M_TEMP);
492 	if (error)
493 		return (error);
494 	return (0);
495 }
496 
497 SYSCTL_OID(_security_jail, OID_AUTO, list, CTLTYPE_STRUCT | CTLFLAG_RD,
498     NULL, 0, sysctl_jail_list, "S", "List of active jails");
499 
500 static int
501 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
502 {
503 	int error, injail;
504 
505 	injail = jailed(req->td->td_ucred);
506 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
507 
508 	return (error);
509 }
510 SYSCTL_PROC(_security_jail, OID_AUTO, jailed, CTLTYPE_INT | CTLFLAG_RD,
511     NULL, 0, sysctl_jail_jailed, "I", "Process in jail?");
512