xref: /freebsd/sys/kern/kern_jail.c (revision 6af83ee0d2941d18880b6aaa2b4facd1d30c6106)
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 mp_fixme("these variables need a lock")
44 
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");
49 
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");
54 
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");
59 
60 int	jail_getfsstatroot_only = 1;
61 SYSCTL_INT(_security_jail, OID_AUTO, getfsstatroot_only, CTLFLAG_RW,
62     &jail_getfsstatroot_only, 0,
63     "Processes see only their root file system in getfsstat()");
64 
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");
69 
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");
74 
75 /* allprison, lastprid, and prisoncount are protected by allprison_mtx. */
76 struct	prisonlist allprison;
77 struct	mtx allprison_mtx;
78 int	lastprid = 0;
79 int	prisoncount = 0;
80 
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);
85 
86 static void
87 init_prison(void *data __unused)
88 {
89 
90 	mtx_init(&allprison_mtx, "allprison", NULL, MTX_DEF);
91 	LIST_INIT(&allprison);
92 }
93 
94 SYSINIT(prison, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_prison, NULL);
95 
96 /*
97  * MPSAFE
98  *
99  * struct jail_args {
100  *	struct jail *jail;
101  * };
102  */
103 int
104 jail(struct thread *td, struct jail_args *uap)
105 {
106 	struct nameidata nd;
107 	struct prison *pr, *tpr;
108 	struct jail j;
109 	struct jail_attach_args jaa;
110 	int error, tryprid;
111 
112 	error = copyin(uap->jail, &j, sizeof(j));
113 	if (error)
114 		return (error);
115 	if (j.version != 0)
116 		return (EINVAL);
117 
118 	MALLOC(pr, struct prison *, sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
119 	mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF);
120 	pr->pr_ref = 1;
121 	error = copyinstr(j.path, &pr->pr_path, sizeof(pr->pr_path), 0);
122 	if (error)
123 		goto e_killmtx;
124 	mtx_lock(&Giant);
125 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, pr->pr_path, td);
126 	error = namei(&nd);
127 	if (error) {
128 		mtx_unlock(&Giant);
129 		goto e_killmtx;
130 	}
131 	pr->pr_root = nd.ni_vp;
132 	VOP_UNLOCK(nd.ni_vp, 0, td);
133 	NDFREE(&nd, NDF_ONLY_PNBUF);
134 	mtx_unlock(&Giant);
135 	error = copyinstr(j.hostname, &pr->pr_host, sizeof(pr->pr_host), 0);
136 	if (error)
137 		goto e_dropvnref;
138 	pr->pr_ip = j.ip_number;
139 	pr->pr_linux = NULL;
140 	pr->pr_securelevel = securelevel;
141 
142 	/* Determine next pr_id and add prison to allprison list. */
143 	mtx_lock(&allprison_mtx);
144 	tryprid = lastprid + 1;
145 	if (tryprid == JAIL_MAX)
146 		tryprid = 1;
147 next:
148 	LIST_FOREACH(tpr, &allprison, pr_list) {
149 		if (tpr->pr_id == tryprid) {
150 			tryprid++;
151 			if (tryprid == JAIL_MAX) {
152 				mtx_unlock(&allprison_mtx);
153 				error = EAGAIN;
154 				goto e_dropvnref;
155 			}
156 			goto next;
157 		}
158 	}
159 	pr->pr_id = jaa.jid = lastprid = tryprid;
160 	LIST_INSERT_HEAD(&allprison, pr, pr_list);
161 	prisoncount++;
162 	mtx_unlock(&allprison_mtx);
163 
164 	error = jail_attach(td, &jaa);
165 	if (error)
166 		goto e_dropprref;
167 	mtx_lock(&pr->pr_mtx);
168 	pr->pr_ref--;
169 	mtx_unlock(&pr->pr_mtx);
170 	td->td_retval[0] = jaa.jid;
171 	return (0);
172 e_dropprref:
173 	mtx_lock(&allprison_mtx);
174 	LIST_REMOVE(pr, pr_list);
175 	prisoncount--;
176 	mtx_unlock(&allprison_mtx);
177 e_dropvnref:
178 	mtx_lock(&Giant);
179 	vrele(pr->pr_root);
180 	mtx_unlock(&Giant);
181 e_killmtx:
182 	mtx_destroy(&pr->pr_mtx);
183 	FREE(pr, M_PRISON);
184 	return (error);
185 }
186 
187 /*
188  * MPSAFE
189  *
190  * struct jail_attach_args {
191  *	int jid;
192  * };
193  */
194 int
195 jail_attach(struct thread *td, struct jail_attach_args *uap)
196 {
197 	struct proc *p;
198 	struct ucred *newcred, *oldcred;
199 	struct prison *pr;
200 	int error;
201 
202 	/*
203 	 * XXX: Note that there is a slight race here if two threads
204 	 * in the same privileged process attempt to attach to two
205 	 * different jails at the same time.  It is important for
206 	 * user processes not to do this, or they might end up with
207 	 * a process root from one prison, but attached to the jail
208 	 * of another.
209 	 */
210 	error = suser(td);
211 	if (error)
212 		return (error);
213 
214 	p = td->td_proc;
215 	mtx_lock(&allprison_mtx);
216 	pr = prison_find(uap->jid);
217 	if (pr == NULL) {
218 		mtx_unlock(&allprison_mtx);
219 		return (EINVAL);
220 	}
221 	pr->pr_ref++;
222 	mtx_unlock(&pr->pr_mtx);
223 	mtx_unlock(&allprison_mtx);
224 
225 	mtx_lock(&Giant);
226 	vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY, td);
227 	if ((error = change_dir(pr->pr_root, td)) != 0)
228 		goto e_unlock;
229 #ifdef MAC
230 	if ((error = mac_check_vnode_chroot(td->td_ucred, pr->pr_root)))
231 		goto e_unlock;
232 #endif
233 	VOP_UNLOCK(pr->pr_root, 0, td);
234 	change_root(pr->pr_root, td);
235 	mtx_unlock(&Giant);
236 
237 	newcred = crget();
238 	PROC_LOCK(p);
239 	oldcred = p->p_ucred;
240 	setsugid(p);
241 	crcopy(newcred, oldcred);
242 	newcred->cr_prison = pr;
243 	p->p_ucred = newcred;
244 	PROC_UNLOCK(p);
245 	crfree(oldcred);
246 	return (0);
247 e_unlock:
248 	VOP_UNLOCK(pr->pr_root, 0, td);
249 	mtx_unlock(&Giant);
250 	mtx_lock(&pr->pr_mtx);
251 	pr->pr_ref--;
252 	mtx_unlock(&pr->pr_mtx);
253 	return (error);
254 }
255 
256 /*
257  * Returns a locked prison instance, or NULL on failure.
258  */
259 static struct prison *
260 prison_find(int prid)
261 {
262 	struct prison *pr;
263 
264 	mtx_assert(&allprison_mtx, MA_OWNED);
265 	LIST_FOREACH(pr, &allprison, pr_list) {
266 		if (pr->pr_id == prid) {
267 			mtx_lock(&pr->pr_mtx);
268 			return (pr);
269 		}
270 	}
271 	return (NULL);
272 }
273 
274 void
275 prison_free(struct prison *pr)
276 {
277 
278 	mtx_lock(&allprison_mtx);
279 	mtx_lock(&pr->pr_mtx);
280 	pr->pr_ref--;
281 	if (pr->pr_ref == 0) {
282 		LIST_REMOVE(pr, pr_list);
283 		mtx_unlock(&pr->pr_mtx);
284 		prisoncount--;
285 		mtx_unlock(&allprison_mtx);
286 
287 		TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
288 		taskqueue_enqueue(taskqueue_swi, &pr->pr_task);
289 		return;
290 	}
291 	mtx_unlock(&pr->pr_mtx);
292 	mtx_unlock(&allprison_mtx);
293 }
294 
295 static void
296 prison_complete(void *context, int pending)
297 {
298 	struct prison *pr;
299 
300 	pr = (struct prison *)context;
301 
302 	mtx_lock(&Giant);
303 	vrele(pr->pr_root);
304 	mtx_unlock(&Giant);
305 
306 	mtx_destroy(&pr->pr_mtx);
307 	if (pr->pr_linux != NULL)
308 		FREE(pr->pr_linux, M_PRISON);
309 	FREE(pr, M_PRISON);
310 }
311 
312 void
313 prison_hold(struct prison *pr)
314 {
315 
316 	mtx_lock(&pr->pr_mtx);
317 	pr->pr_ref++;
318 	mtx_unlock(&pr->pr_mtx);
319 }
320 
321 u_int32_t
322 prison_getip(struct ucred *cred)
323 {
324 
325 	return (cred->cr_prison->pr_ip);
326 }
327 
328 int
329 prison_ip(struct ucred *cred, int flag, u_int32_t *ip)
330 {
331 	u_int32_t tmp;
332 
333 	if (!jailed(cred))
334 		return (0);
335 	if (flag)
336 		tmp = *ip;
337 	else
338 		tmp = ntohl(*ip);
339 	if (tmp == INADDR_ANY) {
340 		if (flag)
341 			*ip = cred->cr_prison->pr_ip;
342 		else
343 			*ip = htonl(cred->cr_prison->pr_ip);
344 		return (0);
345 	}
346 	if (tmp == INADDR_LOOPBACK) {
347 		if (flag)
348 			*ip = cred->cr_prison->pr_ip;
349 		else
350 			*ip = htonl(cred->cr_prison->pr_ip);
351 		return (0);
352 	}
353 	if (cred->cr_prison->pr_ip != tmp)
354 		return (1);
355 	return (0);
356 }
357 
358 void
359 prison_remote_ip(struct ucred *cred, int flag, u_int32_t *ip)
360 {
361 	u_int32_t tmp;
362 
363 	if (!jailed(cred))
364 		return;
365 	if (flag)
366 		tmp = *ip;
367 	else
368 		tmp = ntohl(*ip);
369 	if (tmp == INADDR_LOOPBACK) {
370 		if (flag)
371 			*ip = cred->cr_prison->pr_ip;
372 		else
373 			*ip = htonl(cred->cr_prison->pr_ip);
374 		return;
375 	}
376 	return;
377 }
378 
379 int
380 prison_if(struct ucred *cred, struct sockaddr *sa)
381 {
382 	struct sockaddr_in *sai;
383 	int ok;
384 
385 	sai = (struct sockaddr_in *)sa;
386 	if ((sai->sin_family != AF_INET) && jail_socket_unixiproute_only)
387 		ok = 1;
388 	else if (sai->sin_family != AF_INET)
389 		ok = 0;
390 	else if (cred->cr_prison->pr_ip != ntohl(sai->sin_addr.s_addr))
391 		ok = 1;
392 	else
393 		ok = 0;
394 	return (ok);
395 }
396 
397 /*
398  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
399  */
400 int
401 prison_check(struct ucred *cred1, struct ucred *cred2)
402 {
403 
404 	if (jailed(cred1)) {
405 		if (!jailed(cred2))
406 			return (ESRCH);
407 		if (cred2->cr_prison != cred1->cr_prison)
408 			return (ESRCH);
409 	}
410 
411 	return (0);
412 }
413 
414 /*
415  * Return 1 if the passed credential is in a jail, otherwise 0.
416  */
417 int
418 jailed(struct ucred *cred)
419 {
420 
421 	return (cred->cr_prison != NULL);
422 }
423 
424 /*
425  * Return the correct hostname for the passed credential.
426  */
427 void
428 getcredhostname(struct ucred *cred, char *buf, size_t size)
429 {
430 
431 	if (jailed(cred)) {
432 		mtx_lock(&cred->cr_prison->pr_mtx);
433 		strlcpy(buf, cred->cr_prison->pr_host, size);
434 		mtx_unlock(&cred->cr_prison->pr_mtx);
435 	} else
436 		strlcpy(buf, hostname, size);
437 }
438 
439 /*
440  * Return 1 if the passed credential can "see" the passed mountpoint
441  * when performing a getfsstat(); otherwise, 0.
442  */
443 int
444 prison_check_mount(struct ucred *cred, struct mount *mp)
445 {
446 
447 	if (jail_getfsstatroot_only && cred->cr_prison != NULL) {
448 		if (cred->cr_prison->pr_root->v_mount != mp)
449 			return (0);
450 	}
451 	return (1);
452 }
453 
454 static int
455 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
456 {
457 	struct xprison *xp, *sxp;
458 	struct prison *pr;
459 	int count, error;
460 
461 	mtx_assert(&Giant, MA_OWNED);
462 	if (jailed(req->td->td_ucred))
463 		return (0);
464 retry:
465 	mtx_lock(&allprison_mtx);
466 	count = prisoncount;
467 	mtx_unlock(&allprison_mtx);
468 
469 	if (count == 0)
470 		return (0);
471 
472 	sxp = xp = malloc(sizeof(*xp) * count, M_TEMP, M_WAITOK | M_ZERO);
473 	mtx_lock(&allprison_mtx);
474 	if (count != prisoncount) {
475 		mtx_unlock(&allprison_mtx);
476 		free(sxp, M_TEMP);
477 		goto retry;
478 	}
479 
480 	LIST_FOREACH(pr, &allprison, pr_list) {
481 		mtx_lock(&pr->pr_mtx);
482 		xp->pr_version = XPRISON_VERSION;
483 		xp->pr_id = pr->pr_id;
484 		strlcpy(xp->pr_path, pr->pr_path, sizeof(xp->pr_path));
485 		strlcpy(xp->pr_host, pr->pr_host, sizeof(xp->pr_host));
486 		xp->pr_ip = pr->pr_ip;
487 		mtx_unlock(&pr->pr_mtx);
488 		xp++;
489 	}
490 	mtx_unlock(&allprison_mtx);
491 
492 	error = SYSCTL_OUT(req, sxp, sizeof(*sxp) * count);
493 	free(sxp, M_TEMP);
494 	if (error)
495 		return (error);
496 	return (0);
497 }
498 
499 SYSCTL_OID(_security_jail, OID_AUTO, list, CTLTYPE_STRUCT | CTLFLAG_RD,
500     NULL, 0, sysctl_jail_list, "S", "List of active jails");
501 
502 static int
503 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
504 {
505 	int error, injail;
506 
507 	injail = jailed(req->td->td_ucred);
508 	error = SYSCTL_OUT(req, &injail, sizeof(injail));
509 
510 	return (error);
511 }
512 SYSCTL_PROC(_security_jail, OID_AUTO, jailed, CTLTYPE_INT | CTLFLAG_RD,
513     NULL, 0, sysctl_jail_jailed, "I", "Process in jail?");
514