1/*-
2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
| 1/*-
2 * Copyright (c) 1999 Poul-Henning Kamp.
3 * Copyright (c) 2008 Bjoern A. Zeeb.
4 * Copyright (c) 2009 James Gritton.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
|
30__FBSDID("$FreeBSD: head/sys/kern/kern_jail.c 194118 2009-06-13 15:39:12Z jamie $");
| 30__FBSDID("$FreeBSD: head/sys/kern/kern_jail.c 194251 2009-06-15 18:59:29Z jamie $");
|
31
32#include "opt_compat.h"
33#include "opt_ddb.h"
34#include "opt_inet.h"
35#include "opt_inet6.h"
36
37#include <sys/param.h>
38#include <sys/types.h>
39#include <sys/kernel.h>
40#include <sys/systm.h>
41#include <sys/errno.h>
42#include <sys/sysproto.h>
43#include <sys/malloc.h>
44#include <sys/osd.h>
45#include <sys/priv.h>
46#include <sys/proc.h>
47#include <sys/taskqueue.h>
48#include <sys/fcntl.h>
49#include <sys/jail.h>
50#include <sys/lock.h>
51#include <sys/mutex.h>
52#include <sys/sx.h>
53#include <sys/sysent.h>
54#include <sys/namei.h>
55#include <sys/mount.h>
56#include <sys/queue.h>
57#include <sys/socket.h>
58#include <sys/syscallsubr.h>
59#include <sys/sysctl.h>
60#include <sys/vnode.h>
61#include <sys/vimage.h>
62#include <net/if.h>
63#include <netinet/in.h>
64#ifdef DDB
65#include <ddb/ddb.h>
66#ifdef INET6
67#include <netinet6/in6_var.h>
68#endif /* INET6 */
69#endif /* DDB */
70
71#include <security/mac/mac_framework.h>
72
73MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
74
75/* prison0 describes what is "real" about the system. */
76struct prison prison0 = {
77 .pr_id = 0,
78 .pr_name = "0",
79 .pr_ref = 1,
80 .pr_uref = 1,
81 .pr_path = "/",
82 .pr_securelevel = -1,
83 .pr_hostuuid = "00000000-0000-0000-0000-000000000000",
84 .pr_children = LIST_HEAD_INITIALIZER(&prison0.pr_children),
85 .pr_flags = PR_HOST,
86 .pr_allow = PR_ALLOW_ALL,
87};
88MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
89
90/* allprison and lastprid are protected by allprison_lock. */
91struct sx allprison_lock;
92SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
93struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
94int lastprid = 0;
95
96static int do_jail_attach(struct thread *td, struct prison *pr);
97static void prison_complete(void *context, int pending);
98static void prison_deref(struct prison *pr, int flags);
99static char *prison_path(struct prison *pr1, struct prison *pr2);
100static void prison_remove_one(struct prison *pr);
101#ifdef INET
102static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
103static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
104#endif
105#ifdef INET6
106static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
107static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
108#endif
109
110/* Flags for prison_deref */
111#define PD_DEREF 0x01
112#define PD_DEUREF 0x02
113#define PD_LOCKED 0x04
114#define PD_LIST_SLOCKED 0x08
115#define PD_LIST_XLOCKED 0x10
116
117/*
118 * Parameter names corresponding to PR_* flag values
119 */
120static char *pr_flag_names[] = {
121 [0] = "persist",
122 "host",
123#ifdef INET
124 "ip4",
125#endif
126#ifdef INET6
127 [3] = "ip6",
128#endif
| 31
32#include "opt_compat.h"
33#include "opt_ddb.h"
34#include "opt_inet.h"
35#include "opt_inet6.h"
36
37#include <sys/param.h>
38#include <sys/types.h>
39#include <sys/kernel.h>
40#include <sys/systm.h>
41#include <sys/errno.h>
42#include <sys/sysproto.h>
43#include <sys/malloc.h>
44#include <sys/osd.h>
45#include <sys/priv.h>
46#include <sys/proc.h>
47#include <sys/taskqueue.h>
48#include <sys/fcntl.h>
49#include <sys/jail.h>
50#include <sys/lock.h>
51#include <sys/mutex.h>
52#include <sys/sx.h>
53#include <sys/sysent.h>
54#include <sys/namei.h>
55#include <sys/mount.h>
56#include <sys/queue.h>
57#include <sys/socket.h>
58#include <sys/syscallsubr.h>
59#include <sys/sysctl.h>
60#include <sys/vnode.h>
61#include <sys/vimage.h>
62#include <net/if.h>
63#include <netinet/in.h>
64#ifdef DDB
65#include <ddb/ddb.h>
66#ifdef INET6
67#include <netinet6/in6_var.h>
68#endif /* INET6 */
69#endif /* DDB */
70
71#include <security/mac/mac_framework.h>
72
73MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
74
75/* prison0 describes what is "real" about the system. */
76struct prison prison0 = {
77 .pr_id = 0,
78 .pr_name = "0",
79 .pr_ref = 1,
80 .pr_uref = 1,
81 .pr_path = "/",
82 .pr_securelevel = -1,
83 .pr_hostuuid = "00000000-0000-0000-0000-000000000000",
84 .pr_children = LIST_HEAD_INITIALIZER(&prison0.pr_children),
85 .pr_flags = PR_HOST,
86 .pr_allow = PR_ALLOW_ALL,
87};
88MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
89
90/* allprison and lastprid are protected by allprison_lock. */
91struct sx allprison_lock;
92SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
93struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
94int lastprid = 0;
95
96static int do_jail_attach(struct thread *td, struct prison *pr);
97static void prison_complete(void *context, int pending);
98static void prison_deref(struct prison *pr, int flags);
99static char *prison_path(struct prison *pr1, struct prison *pr2);
100static void prison_remove_one(struct prison *pr);
101#ifdef INET
102static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
103static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
104#endif
105#ifdef INET6
106static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
107static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
108#endif
109
110/* Flags for prison_deref */
111#define PD_DEREF 0x01
112#define PD_DEUREF 0x02
113#define PD_LOCKED 0x04
114#define PD_LIST_SLOCKED 0x08
115#define PD_LIST_XLOCKED 0x10
116
117/*
118 * Parameter names corresponding to PR_* flag values
119 */
120static char *pr_flag_names[] = {
121 [0] = "persist",
122 "host",
123#ifdef INET
124 "ip4",
125#endif
126#ifdef INET6
127 [3] = "ip6",
128#endif
|
| 129#ifdef VIMAGE
130 [4] = "vnet",
131#endif
|
129};
130
131static char *pr_flag_nonames[] = {
132 [0] = "nopersist",
133 "nohost",
134#ifdef INET
135 "noip4",
136#endif
137#ifdef INET6
138 [3] = "noip6",
139#endif
| 132};
133
134static char *pr_flag_nonames[] = {
135 [0] = "nopersist",
136 "nohost",
137#ifdef INET
138 "noip4",
139#endif
140#ifdef INET6
141 [3] = "noip6",
142#endif
|
| 143#ifdef VIMAGE
144 [4] = "novnet",
145#endif
|
140};
141
142static char *pr_allow_names[] = {
143 "allow.set_hostname",
144 "allow.sysvipc",
145 "allow.raw_sockets",
146 "allow.chflags",
147 "allow.mount",
148 "allow.quotas",
149 "allow.jails",
150 "allow.socket_af",
151};
152
153static char *pr_allow_nonames[] = {
154 "allow.noset_hostname",
155 "allow.nosysvipc",
156 "allow.noraw_sockets",
157 "allow.nochflags",
158 "allow.nomount",
159 "allow.noquotas",
160 "allow.nojails",
161 "allow.nosocket_af",
162};
163
164#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
165static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
166static int jail_default_enforce_statfs = 2;
167#if defined(INET) || defined(INET6)
168static unsigned jail_max_af_ips = 255;
169#endif
170
171#ifdef INET
172static int
173qcmp_v4(const void *ip1, const void *ip2)
174{
175 in_addr_t iaa, iab;
176
177 /*
178 * We need to compare in HBO here to get the list sorted as expected
179 * by the result of the code. Sorting NBO addresses gives you
180 * interesting results. If you do not understand, do not try.
181 */
182 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
183 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
184
185 /*
186 * Do not simply return the difference of the two numbers, the int is
187 * not wide enough.
188 */
189 if (iaa > iab)
190 return (1);
191 else if (iaa < iab)
192 return (-1);
193 else
194 return (0);
195}
196#endif
197
198#ifdef INET6
199static int
200qcmp_v6(const void *ip1, const void *ip2)
201{
202 const struct in6_addr *ia6a, *ia6b;
203 int i, rc;
204
205 ia6a = (const struct in6_addr *)ip1;
206 ia6b = (const struct in6_addr *)ip2;
207
208 rc = 0;
209 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
210 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
211 rc = 1;
212 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
213 rc = -1;
214 }
215 return (rc);
216}
217#endif
218
219/*
220 * struct jail_args {
221 * struct jail *jail;
222 * };
223 */
224int
225jail(struct thread *td, struct jail_args *uap)
226{
227 uint32_t version;
228 int error;
229 struct jail j;
230
231 error = copyin(uap->jail, &version, sizeof(uint32_t));
232 if (error)
233 return (error);
234
235 switch (version) {
236 case 0:
237 {
238 struct jail_v0 j0;
239
240 /* FreeBSD single IPv4 jails. */
241 bzero(&j, sizeof(struct jail));
242 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
243 if (error)
244 return (error);
245 j.version = j0.version;
246 j.path = j0.path;
247 j.hostname = j0.hostname;
248 j.ip4s = j0.ip_number;
249 break;
250 }
251
252 case 1:
253 /*
254 * Version 1 was used by multi-IPv4 jail implementations
255 * that never made it into the official kernel.
256 */
257 return (EINVAL);
258
259 case 2: /* JAIL_API_VERSION */
260 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
261 error = copyin(uap->jail, &j, sizeof(struct jail));
262 if (error)
263 return (error);
264 break;
265
266 default:
267 /* Sci-Fi jails are not supported, sorry. */
268 return (EINVAL);
269 }
270 return (kern_jail(td, &j));
271}
272
273int
274kern_jail(struct thread *td, struct jail *j)
275{
276 struct iovec optiov[2 * (4
277 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
278#ifdef INET
279 + 1
280#endif
281#ifdef INET6
282 + 1
283#endif
284 )];
285 struct uio opt;
286 char *u_path, *u_hostname, *u_name;
287#ifdef INET
288 uint32_t ip4s;
289 struct in_addr *u_ip4;
290#endif
291#ifdef INET6
292 struct in6_addr *u_ip6;
293#endif
294 size_t tmplen;
295 int error, enforce_statfs, fi;
296
297 bzero(&optiov, sizeof(optiov));
298 opt.uio_iov = optiov;
299 opt.uio_iovcnt = 0;
300 opt.uio_offset = -1;
301 opt.uio_resid = -1;
302 opt.uio_segflg = UIO_SYSSPACE;
303 opt.uio_rw = UIO_READ;
304 opt.uio_td = td;
305
306 /* Set permissions for top-level jails from sysctls. */
307 if (!jailed(td->td_ucred)) {
308 for (fi = 0; fi < sizeof(pr_allow_names) /
309 sizeof(pr_allow_names[0]); fi++) {
310 optiov[opt.uio_iovcnt].iov_base =
311 (jail_default_allow & (1 << fi))
312 ? pr_allow_names[fi] : pr_allow_nonames[fi];
313 optiov[opt.uio_iovcnt].iov_len =
314 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
315 opt.uio_iovcnt += 2;
316 }
317 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
318 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
319 opt.uio_iovcnt++;
320 enforce_statfs = jail_default_enforce_statfs;
321 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
322 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
323 opt.uio_iovcnt++;
324 }
325
326 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
327#ifdef INET
328 ip4s = (j->version == 0) ? 1 : j->ip4s;
329 if (ip4s > jail_max_af_ips)
330 return (EINVAL);
331 tmplen += ip4s * sizeof(struct in_addr);
332#else
333 if (j->ip4s > 0)
334 return (EINVAL);
335#endif
336#ifdef INET6
337 if (j->ip6s > jail_max_af_ips)
338 return (EINVAL);
339 tmplen += j->ip6s * sizeof(struct in6_addr);
340#else
341 if (j->ip6s > 0)
342 return (EINVAL);
343#endif
344 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
345 u_hostname = u_path + MAXPATHLEN;
346 u_name = u_hostname + MAXHOSTNAMELEN;
347#ifdef INET
348 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
349#endif
350#ifdef INET6
351#ifdef INET
352 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
353#else
354 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
355#endif
356#endif
357 optiov[opt.uio_iovcnt].iov_base = "path";
358 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
359 opt.uio_iovcnt++;
360 optiov[opt.uio_iovcnt].iov_base = u_path;
361 error = copyinstr(j->path, u_path, MAXPATHLEN,
362 &optiov[opt.uio_iovcnt].iov_len);
363 if (error) {
364 free(u_path, M_TEMP);
365 return (error);
366 }
367 opt.uio_iovcnt++;
368 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
369 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
370 opt.uio_iovcnt++;
371 optiov[opt.uio_iovcnt].iov_base = u_hostname;
372 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
373 &optiov[opt.uio_iovcnt].iov_len);
374 if (error) {
375 free(u_path, M_TEMP);
376 return (error);
377 }
378 opt.uio_iovcnt++;
379 if (j->jailname != NULL) {
380 optiov[opt.uio_iovcnt].iov_base = "name";
381 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
382 opt.uio_iovcnt++;
383 optiov[opt.uio_iovcnt].iov_base = u_name;
384 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
385 &optiov[opt.uio_iovcnt].iov_len);
386 if (error) {
387 free(u_path, M_TEMP);
388 return (error);
389 }
390 opt.uio_iovcnt++;
391 }
392#ifdef INET
393 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
394 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
395 opt.uio_iovcnt++;
396 optiov[opt.uio_iovcnt].iov_base = u_ip4;
397 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
398 if (j->version == 0)
399 u_ip4->s_addr = j->ip4s;
400 else {
401 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
402 if (error) {
403 free(u_path, M_TEMP);
404 return (error);
405 }
406 }
407 opt.uio_iovcnt++;
408#endif
409#ifdef INET6
410 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
411 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
412 opt.uio_iovcnt++;
413 optiov[opt.uio_iovcnt].iov_base = u_ip6;
414 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
415 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
416 if (error) {
417 free(u_path, M_TEMP);
418 return (error);
419 }
420 opt.uio_iovcnt++;
421#endif
422 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
423 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
424 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
425 free(u_path, M_TEMP);
426 return (error);
427}
428
429
430/*
431 * struct jail_set_args {
432 * struct iovec *iovp;
433 * unsigned int iovcnt;
434 * int flags;
435 * };
436 */
437int
438jail_set(struct thread *td, struct jail_set_args *uap)
439{
440 struct uio *auio;
441 int error;
442
443 /* Check that we have an even number of iovecs. */
444 if (uap->iovcnt & 1)
445 return (EINVAL);
446
447 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
448 if (error)
449 return (error);
450 error = kern_jail_set(td, auio, uap->flags);
451 free(auio, M_IOV);
452 return (error);
453}
454
455int
456kern_jail_set(struct thread *td, struct uio *optuio, int flags)
457{
458 struct nameidata nd;
459#ifdef INET
460 struct in_addr *ip4;
461#endif
462#ifdef INET6
463 struct in6_addr *ip6;
464#endif
465 struct vfsopt *opt;
466 struct vfsoptlist *opts;
467 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
468 struct vnode *root;
469 char *domain, *errmsg, *host, *name, *p, *path, *uuid;
470#if defined(INET) || defined(INET6)
471 void *op;
472#endif
473 unsigned long hid;
474 size_t namelen, onamelen;
475 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
476 int gotenforce, gothid, gotslevel, fi, jid, len;
477 int slevel, vfslocked;
478#if defined(INET) || defined(INET6)
479 int ii, ij;
480#endif
481#ifdef INET
482 int ip4s, ip4a, redo_ip4;
483#endif
484#ifdef INET6
485 int ip6s, ip6a, redo_ip6;
486#endif
487 unsigned pr_flags, ch_flags;
488 unsigned pr_allow, ch_allow, tallow;
489 char numbuf[12];
490
491 error = priv_check(td, PRIV_JAIL_SET);
492 if (!error && (flags & JAIL_ATTACH))
493 error = priv_check(td, PRIV_JAIL_ATTACH);
494 if (error)
495 return (error);
496 mypr = ppr = td->td_ucred->cr_prison;
497 if ((flags & JAIL_CREATE) && !(mypr->pr_allow & PR_ALLOW_JAILS))
498 return (EPERM);
499 if (flags & ~JAIL_SET_MASK)
500 return (EINVAL);
501
502 /*
503 * Check all the parameters before committing to anything. Not all
504 * errors can be caught early, but we may as well try. Also, this
505 * takes care of some expensive stuff (path lookup) before getting
506 * the allprison lock.
507 *
508 * XXX Jails are not filesystems, and jail parameters are not mount
509 * options. But it makes more sense to re-use the vfsopt code
510 * than duplicate it under a different name.
511 */
512 error = vfs_buildopts(optuio, &opts);
513 if (error)
514 return (error);
515#ifdef INET
516 ip4a = 0;
517 ip4 = NULL;
518#endif
519#ifdef INET6
520 ip6a = 0;
521 ip6 = NULL;
522#endif
523
524#if defined(INET) || defined(INET6)
525 again:
526#endif
527 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
528 if (error == ENOENT)
529 jid = 0;
530 else if (error != 0)
531 goto done_free;
532
533 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
534 if (error == ENOENT)
535 gotslevel = 0;
536 else if (error != 0)
537 goto done_free;
538 else
539 gotslevel = 1;
540
541 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
542 gotenforce = (error == 0);
543 if (gotenforce) {
544 if (enforce < 0 || enforce > 2)
545 return (EINVAL);
546 } else if (error != ENOENT)
547 goto done_free;
548
549 pr_flags = ch_flags = 0;
550 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
551 fi++) {
552 if (pr_flag_names[fi] == NULL)
553 continue;
554 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
555 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
556 }
557 ch_flags |= pr_flags;
558 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
559 && !(pr_flags & PR_PERSIST)) {
560 error = EINVAL;
561 vfs_opterror(opts, "new jail must persist or attach");
562 goto done_errmsg;
563 }
| 146};
147
148static char *pr_allow_names[] = {
149 "allow.set_hostname",
150 "allow.sysvipc",
151 "allow.raw_sockets",
152 "allow.chflags",
153 "allow.mount",
154 "allow.quotas",
155 "allow.jails",
156 "allow.socket_af",
157};
158
159static char *pr_allow_nonames[] = {
160 "allow.noset_hostname",
161 "allow.nosysvipc",
162 "allow.noraw_sockets",
163 "allow.nochflags",
164 "allow.nomount",
165 "allow.noquotas",
166 "allow.nojails",
167 "allow.nosocket_af",
168};
169
170#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
171static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
172static int jail_default_enforce_statfs = 2;
173#if defined(INET) || defined(INET6)
174static unsigned jail_max_af_ips = 255;
175#endif
176
177#ifdef INET
178static int
179qcmp_v4(const void *ip1, const void *ip2)
180{
181 in_addr_t iaa, iab;
182
183 /*
184 * We need to compare in HBO here to get the list sorted as expected
185 * by the result of the code. Sorting NBO addresses gives you
186 * interesting results. If you do not understand, do not try.
187 */
188 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
189 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
190
191 /*
192 * Do not simply return the difference of the two numbers, the int is
193 * not wide enough.
194 */
195 if (iaa > iab)
196 return (1);
197 else if (iaa < iab)
198 return (-1);
199 else
200 return (0);
201}
202#endif
203
204#ifdef INET6
205static int
206qcmp_v6(const void *ip1, const void *ip2)
207{
208 const struct in6_addr *ia6a, *ia6b;
209 int i, rc;
210
211 ia6a = (const struct in6_addr *)ip1;
212 ia6b = (const struct in6_addr *)ip2;
213
214 rc = 0;
215 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
216 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
217 rc = 1;
218 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
219 rc = -1;
220 }
221 return (rc);
222}
223#endif
224
225/*
226 * struct jail_args {
227 * struct jail *jail;
228 * };
229 */
230int
231jail(struct thread *td, struct jail_args *uap)
232{
233 uint32_t version;
234 int error;
235 struct jail j;
236
237 error = copyin(uap->jail, &version, sizeof(uint32_t));
238 if (error)
239 return (error);
240
241 switch (version) {
242 case 0:
243 {
244 struct jail_v0 j0;
245
246 /* FreeBSD single IPv4 jails. */
247 bzero(&j, sizeof(struct jail));
248 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
249 if (error)
250 return (error);
251 j.version = j0.version;
252 j.path = j0.path;
253 j.hostname = j0.hostname;
254 j.ip4s = j0.ip_number;
255 break;
256 }
257
258 case 1:
259 /*
260 * Version 1 was used by multi-IPv4 jail implementations
261 * that never made it into the official kernel.
262 */
263 return (EINVAL);
264
265 case 2: /* JAIL_API_VERSION */
266 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
267 error = copyin(uap->jail, &j, sizeof(struct jail));
268 if (error)
269 return (error);
270 break;
271
272 default:
273 /* Sci-Fi jails are not supported, sorry. */
274 return (EINVAL);
275 }
276 return (kern_jail(td, &j));
277}
278
279int
280kern_jail(struct thread *td, struct jail *j)
281{
282 struct iovec optiov[2 * (4
283 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
284#ifdef INET
285 + 1
286#endif
287#ifdef INET6
288 + 1
289#endif
290 )];
291 struct uio opt;
292 char *u_path, *u_hostname, *u_name;
293#ifdef INET
294 uint32_t ip4s;
295 struct in_addr *u_ip4;
296#endif
297#ifdef INET6
298 struct in6_addr *u_ip6;
299#endif
300 size_t tmplen;
301 int error, enforce_statfs, fi;
302
303 bzero(&optiov, sizeof(optiov));
304 opt.uio_iov = optiov;
305 opt.uio_iovcnt = 0;
306 opt.uio_offset = -1;
307 opt.uio_resid = -1;
308 opt.uio_segflg = UIO_SYSSPACE;
309 opt.uio_rw = UIO_READ;
310 opt.uio_td = td;
311
312 /* Set permissions for top-level jails from sysctls. */
313 if (!jailed(td->td_ucred)) {
314 for (fi = 0; fi < sizeof(pr_allow_names) /
315 sizeof(pr_allow_names[0]); fi++) {
316 optiov[opt.uio_iovcnt].iov_base =
317 (jail_default_allow & (1 << fi))
318 ? pr_allow_names[fi] : pr_allow_nonames[fi];
319 optiov[opt.uio_iovcnt].iov_len =
320 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
321 opt.uio_iovcnt += 2;
322 }
323 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
324 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
325 opt.uio_iovcnt++;
326 enforce_statfs = jail_default_enforce_statfs;
327 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
328 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
329 opt.uio_iovcnt++;
330 }
331
332 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
333#ifdef INET
334 ip4s = (j->version == 0) ? 1 : j->ip4s;
335 if (ip4s > jail_max_af_ips)
336 return (EINVAL);
337 tmplen += ip4s * sizeof(struct in_addr);
338#else
339 if (j->ip4s > 0)
340 return (EINVAL);
341#endif
342#ifdef INET6
343 if (j->ip6s > jail_max_af_ips)
344 return (EINVAL);
345 tmplen += j->ip6s * sizeof(struct in6_addr);
346#else
347 if (j->ip6s > 0)
348 return (EINVAL);
349#endif
350 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
351 u_hostname = u_path + MAXPATHLEN;
352 u_name = u_hostname + MAXHOSTNAMELEN;
353#ifdef INET
354 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
355#endif
356#ifdef INET6
357#ifdef INET
358 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
359#else
360 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
361#endif
362#endif
363 optiov[opt.uio_iovcnt].iov_base = "path";
364 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
365 opt.uio_iovcnt++;
366 optiov[opt.uio_iovcnt].iov_base = u_path;
367 error = copyinstr(j->path, u_path, MAXPATHLEN,
368 &optiov[opt.uio_iovcnt].iov_len);
369 if (error) {
370 free(u_path, M_TEMP);
371 return (error);
372 }
373 opt.uio_iovcnt++;
374 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
375 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
376 opt.uio_iovcnt++;
377 optiov[opt.uio_iovcnt].iov_base = u_hostname;
378 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
379 &optiov[opt.uio_iovcnt].iov_len);
380 if (error) {
381 free(u_path, M_TEMP);
382 return (error);
383 }
384 opt.uio_iovcnt++;
385 if (j->jailname != NULL) {
386 optiov[opt.uio_iovcnt].iov_base = "name";
387 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
388 opt.uio_iovcnt++;
389 optiov[opt.uio_iovcnt].iov_base = u_name;
390 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
391 &optiov[opt.uio_iovcnt].iov_len);
392 if (error) {
393 free(u_path, M_TEMP);
394 return (error);
395 }
396 opt.uio_iovcnt++;
397 }
398#ifdef INET
399 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
400 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
401 opt.uio_iovcnt++;
402 optiov[opt.uio_iovcnt].iov_base = u_ip4;
403 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
404 if (j->version == 0)
405 u_ip4->s_addr = j->ip4s;
406 else {
407 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
408 if (error) {
409 free(u_path, M_TEMP);
410 return (error);
411 }
412 }
413 opt.uio_iovcnt++;
414#endif
415#ifdef INET6
416 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
417 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
418 opt.uio_iovcnt++;
419 optiov[opt.uio_iovcnt].iov_base = u_ip6;
420 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
421 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
422 if (error) {
423 free(u_path, M_TEMP);
424 return (error);
425 }
426 opt.uio_iovcnt++;
427#endif
428 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
429 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
430 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
431 free(u_path, M_TEMP);
432 return (error);
433}
434
435
436/*
437 * struct jail_set_args {
438 * struct iovec *iovp;
439 * unsigned int iovcnt;
440 * int flags;
441 * };
442 */
443int
444jail_set(struct thread *td, struct jail_set_args *uap)
445{
446 struct uio *auio;
447 int error;
448
449 /* Check that we have an even number of iovecs. */
450 if (uap->iovcnt & 1)
451 return (EINVAL);
452
453 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
454 if (error)
455 return (error);
456 error = kern_jail_set(td, auio, uap->flags);
457 free(auio, M_IOV);
458 return (error);
459}
460
461int
462kern_jail_set(struct thread *td, struct uio *optuio, int flags)
463{
464 struct nameidata nd;
465#ifdef INET
466 struct in_addr *ip4;
467#endif
468#ifdef INET6
469 struct in6_addr *ip6;
470#endif
471 struct vfsopt *opt;
472 struct vfsoptlist *opts;
473 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
474 struct vnode *root;
475 char *domain, *errmsg, *host, *name, *p, *path, *uuid;
476#if defined(INET) || defined(INET6)
477 void *op;
478#endif
479 unsigned long hid;
480 size_t namelen, onamelen;
481 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
482 int gotenforce, gothid, gotslevel, fi, jid, len;
483 int slevel, vfslocked;
484#if defined(INET) || defined(INET6)
485 int ii, ij;
486#endif
487#ifdef INET
488 int ip4s, ip4a, redo_ip4;
489#endif
490#ifdef INET6
491 int ip6s, ip6a, redo_ip6;
492#endif
493 unsigned pr_flags, ch_flags;
494 unsigned pr_allow, ch_allow, tallow;
495 char numbuf[12];
496
497 error = priv_check(td, PRIV_JAIL_SET);
498 if (!error && (flags & JAIL_ATTACH))
499 error = priv_check(td, PRIV_JAIL_ATTACH);
500 if (error)
501 return (error);
502 mypr = ppr = td->td_ucred->cr_prison;
503 if ((flags & JAIL_CREATE) && !(mypr->pr_allow & PR_ALLOW_JAILS))
504 return (EPERM);
505 if (flags & ~JAIL_SET_MASK)
506 return (EINVAL);
507
508 /*
509 * Check all the parameters before committing to anything. Not all
510 * errors can be caught early, but we may as well try. Also, this
511 * takes care of some expensive stuff (path lookup) before getting
512 * the allprison lock.
513 *
514 * XXX Jails are not filesystems, and jail parameters are not mount
515 * options. But it makes more sense to re-use the vfsopt code
516 * than duplicate it under a different name.
517 */
518 error = vfs_buildopts(optuio, &opts);
519 if (error)
520 return (error);
521#ifdef INET
522 ip4a = 0;
523 ip4 = NULL;
524#endif
525#ifdef INET6
526 ip6a = 0;
527 ip6 = NULL;
528#endif
529
530#if defined(INET) || defined(INET6)
531 again:
532#endif
533 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
534 if (error == ENOENT)
535 jid = 0;
536 else if (error != 0)
537 goto done_free;
538
539 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
540 if (error == ENOENT)
541 gotslevel = 0;
542 else if (error != 0)
543 goto done_free;
544 else
545 gotslevel = 1;
546
547 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
548 gotenforce = (error == 0);
549 if (gotenforce) {
550 if (enforce < 0 || enforce > 2)
551 return (EINVAL);
552 } else if (error != ENOENT)
553 goto done_free;
554
555 pr_flags = ch_flags = 0;
556 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
557 fi++) {
558 if (pr_flag_names[fi] == NULL)
559 continue;
560 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
561 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
562 }
563 ch_flags |= pr_flags;
564 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
565 && !(pr_flags & PR_PERSIST)) {
566 error = EINVAL;
567 vfs_opterror(opts, "new jail must persist or attach");
568 goto done_errmsg;
569 }
|
| 570#ifdef VIMAGE
571 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
572 error = EINVAL;
573 vfs_opterror(opts, "vnet cannot be changed after creation");
574 goto done_errmsg;
575 }
576#endif
|
564
565 pr_allow = ch_allow = 0;
566 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
567 fi++) {
568 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
569 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
570 }
571 ch_allow |= pr_allow;
572
573 error = vfs_getopt(opts, "name", (void **)&name, &len);
574 if (error == ENOENT)
575 name = NULL;
576 else if (error != 0)
577 goto done_free;
578 else {
579 if (len == 0 || name[len - 1] != '\0') {
580 error = EINVAL;
581 goto done_free;
582 }
583 if (len > MAXHOSTNAMELEN) {
584 error = ENAMETOOLONG;
585 goto done_free;
586 }
587 }
588
589 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
590 if (error == ENOENT)
591 host = NULL;
592 else if (error != 0)
593 goto done_free;
594 else {
595 ch_flags |= PR_HOST;
596 pr_flags |= PR_HOST;
597 if (len == 0 || host[len - 1] != '\0') {
598 error = EINVAL;
599 goto done_free;
600 }
601 if (len > MAXHOSTNAMELEN) {
602 error = ENAMETOOLONG;
603 goto done_free;
604 }
605 }
606
607 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
608 if (error == ENOENT)
609 domain = NULL;
610 else if (error != 0)
611 goto done_free;
612 else {
613 ch_flags |= PR_HOST;
614 pr_flags |= PR_HOST;
615 if (len == 0 || domain[len - 1] != '\0') {
616 error = EINVAL;
617 goto done_free;
618 }
619 if (len > MAXHOSTNAMELEN) {
620 error = ENAMETOOLONG;
621 goto done_free;
622 }
623 }
624
625 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
626 if (error == ENOENT)
627 uuid = NULL;
628 else if (error != 0)
629 goto done_free;
630 else {
631 ch_flags |= PR_HOST;
632 pr_flags |= PR_HOST;
633 if (len == 0 || uuid[len - 1] != '\0') {
634 error = EINVAL;
635 goto done_free;
636 }
637 if (len > HOSTUUIDLEN) {
638 error = ENAMETOOLONG;
639 goto done_free;
640 }
641 }
642
643#ifdef COMPAT_IA32
644 if (td->td_proc->p_sysent->sv_flags & SV_IA32) {
645 uint32_t hid32;
646
647 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
648 hid = hid32;
649 } else
650#endif
651 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
652 if (error == ENOENT)
653 gothid = 0;
654 else if (error != 0)
655 goto done_free;
656 else {
657 gothid = 1;
658 ch_flags |= PR_HOST;
659 pr_flags |= PR_HOST;
660 }
661
662 /* This might be the second time around for this option. */
663#ifdef INET
664 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
665 if (error == ENOENT)
666 ip4s = -1;
667 else if (error != 0)
668 goto done_free;
669 else if (ip4s & (sizeof(*ip4) - 1)) {
670 error = EINVAL;
671 goto done_free;
672 } else {
673 ch_flags |= PR_IP4_USER;
674 pr_flags |= PR_IP4_USER;
675 if (ip4s > 0) {
676 ip4s /= sizeof(*ip4);
677 if (ip4s > jail_max_af_ips) {
678 error = EINVAL;
679 vfs_opterror(opts, "too many IPv4 addresses");
680 goto done_errmsg;
681 }
682 if (ip4a < ip4s) {
683 ip4a = ip4s;
684 free(ip4, M_PRISON);
685 ip4 = NULL;
686 }
687 if (ip4 == NULL)
688 ip4 = malloc(ip4a * sizeof(*ip4), M_PRISON,
689 M_WAITOK);
690 bcopy(op, ip4, ip4s * sizeof(*ip4));
691 /*
692 * IP addresses are all sorted but ip[0] to preserve
693 * the primary IP address as given from userland.
694 * This special IP is used for unbound outgoing
695 * connections as well for "loopback" traffic.
696 */
697 if (ip4s > 1)
698 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
699 /*
700 * Check for duplicate addresses and do some simple
701 * zero and broadcast checks. If users give other bogus
702 * addresses it is their problem.
703 *
704 * We do not have to care about byte order for these
705 * checks so we will do them in NBO.
706 */
707 for (ii = 0; ii < ip4s; ii++) {
708 if (ip4[ii].s_addr == INADDR_ANY ||
709 ip4[ii].s_addr == INADDR_BROADCAST) {
710 error = EINVAL;
711 goto done_free;
712 }
713 if ((ii+1) < ip4s &&
714 (ip4[0].s_addr == ip4[ii+1].s_addr ||
715 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
716 error = EINVAL;
717 goto done_free;
718 }
719 }
720 }
721 }
722#endif
723
724#ifdef INET6
725 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
726 if (error == ENOENT)
727 ip6s = -1;
728 else if (error != 0)
729 goto done_free;
730 else if (ip6s & (sizeof(*ip6) - 1)) {
731 error = EINVAL;
732 goto done_free;
733 } else {
734 ch_flags |= PR_IP6_USER;
735 pr_flags |= PR_IP6_USER;
736 if (ip6s > 0) {
737 ip6s /= sizeof(*ip6);
738 if (ip6s > jail_max_af_ips) {
739 error = EINVAL;
740 vfs_opterror(opts, "too many IPv6 addresses");
741 goto done_errmsg;
742 }
743 if (ip6a < ip6s) {
744 ip6a = ip6s;
745 free(ip6, M_PRISON);
746 ip6 = NULL;
747 }
748 if (ip6 == NULL)
749 ip6 = malloc(ip6a * sizeof(*ip6), M_PRISON,
750 M_WAITOK);
751 bcopy(op, ip6, ip6s * sizeof(*ip6));
752 if (ip6s > 1)
753 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
754 for (ii = 0; ii < ip6s; ii++) {
755 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
756 error = EINVAL;
757 goto done_free;
758 }
759 if ((ii+1) < ip6s &&
760 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
761 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
762 {
763 error = EINVAL;
764 goto done_free;
765 }
766 }
767 }
768 }
769#endif
770
771 root = NULL;
772 error = vfs_getopt(opts, "path", (void **)&path, &len);
773 if (error == ENOENT)
774 path = NULL;
775 else if (error != 0)
776 goto done_free;
777 else {
778 if (flags & JAIL_UPDATE) {
779 error = EINVAL;
780 vfs_opterror(opts,
781 "path cannot be changed after creation");
782 goto done_errmsg;
783 }
784 if (len == 0 || path[len - 1] != '\0') {
785 error = EINVAL;
786 goto done_free;
787 }
788 if (len < 2 || (len == 2 && path[0] == '/'))
789 path = NULL;
790 else {
791 /* Leave room for a real-root full pathname. */
792 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
793 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
794 error = ENAMETOOLONG;
795 goto done_free;
796 }
797 NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
798 path, td);
799 error = namei(&nd);
800 if (error)
801 goto done_free;
802 vfslocked = NDHASGIANT(&nd);
803 root = nd.ni_vp;
804 NDFREE(&nd, NDF_ONLY_PNBUF);
805 if (root->v_type != VDIR) {
806 error = ENOTDIR;
807 vrele(root);
808 VFS_UNLOCK_GIANT(vfslocked);
809 goto done_free;
810 }
811 VFS_UNLOCK_GIANT(vfslocked);
812 }
813 }
814
815 /*
816 * Grab the allprison lock before letting modules check their
817 * parameters. Once we have it, do not let go so we'll have a
818 * consistent view of the OSD list.
819 */
820 sx_xlock(&allprison_lock);
821 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
822 if (error)
823 goto done_unlock_list;
824
825 /* By now, all parameters should have been noted. */
826 TAILQ_FOREACH(opt, opts, link) {
827 if (!opt->seen && strcmp(opt->name, "errmsg")) {
828 error = EINVAL;
829 vfs_opterror(opts, "unknown parameter: %s", opt->name);
830 goto done_unlock_list;
831 }
832 }
833
834 /*
835 * See if we are creating a new record or updating an existing one.
836 * This abuses the file error codes ENOENT and EEXIST.
837 */
838 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
839 if (!cuflags) {
840 error = EINVAL;
841 vfs_opterror(opts, "no valid operation (create or update)");
842 goto done_unlock_list;
843 }
844 pr = NULL;
845 if (jid != 0) {
846 /*
847 * See if a requested jid already exists. There is an
848 * information leak here if the jid exists but is not within
849 * the caller's jail hierarchy. Jail creators will get EEXIST
850 * even though they cannot see the jail, and CREATE | UPDATE
851 * will return ENOENT which is not normally a valid error.
852 */
853 if (jid < 0) {
854 error = EINVAL;
855 vfs_opterror(opts, "negative jid");
856 goto done_unlock_list;
857 }
858 pr = prison_find(jid);
859 if (pr != NULL) {
860 ppr = pr->pr_parent;
861 /* Create: jid must not exist. */
862 if (cuflags == JAIL_CREATE) {
863 mtx_unlock(&pr->pr_mtx);
864 error = EEXIST;
865 vfs_opterror(opts, "jail %d already exists",
866 jid);
867 goto done_unlock_list;
868 }
869 if (!prison_ischild(mypr, pr)) {
870 mtx_unlock(&pr->pr_mtx);
871 pr = NULL;
872 } else if (pr->pr_uref == 0) {
873 if (!(flags & JAIL_DYING)) {
874 mtx_unlock(&pr->pr_mtx);
875 error = ENOENT;
876 vfs_opterror(opts, "jail %d is dying",
877 jid);
878 goto done_unlock_list;
879 } else if ((flags & JAIL_ATTACH) ||
880 (pr_flags & PR_PERSIST)) {
881 /*
882 * A dying jail might be resurrected
883 * (via attach or persist), but first
884 * it must determine if another jail
885 * has claimed its name. Accomplish
886 * this by implicitly re-setting the
887 * name.
888 */
889 if (name == NULL)
890 name = prison_name(mypr, pr);
891 }
892 }
893 }
894 if (pr == NULL) {
895 /* Update: jid must exist. */
896 if (cuflags == JAIL_UPDATE) {
897 error = ENOENT;
898 vfs_opterror(opts, "jail %d not found", jid);
899 goto done_unlock_list;
900 }
901 }
902 }
903 /*
904 * If the caller provided a name, look for a jail by that name.
905 * This has different semantics for creates and updates keyed by jid
906 * (where the name must not already exist in a different jail),
907 * and updates keyed by the name itself (where the name must exist
908 * because that is the jail being updated).
909 */
910 if (name != NULL) {
911 p = strrchr(name, '.');
912 if (p != NULL) {
913 /*
914 * This is a hierarchical name. Split it into the
915 * parent and child names, and make sure the parent
916 * exists or matches an already found jail.
917 */
918 *p = '\0';
919 if (pr != NULL) {
920 if (strncmp(name, ppr->pr_name, p - name) ||
921 ppr->pr_name[p - name] != '\0') {
922 mtx_unlock(&pr->pr_mtx);
923 error = EINVAL;
924 vfs_opterror(opts,
925 "cannot change jail's parent");
926 goto done_unlock_list;
927 }
928 } else {
929 ppr = prison_find_name(mypr, name);
930 if (ppr == NULL) {
931 error = ENOENT;
932 vfs_opterror(opts,
933 "jail \"%s\" not found", name);
934 goto done_unlock_list;
935 }
936 mtx_unlock(&ppr->pr_mtx);
937 }
938 name = p + 1;
939 }
940 if (name[0] != '\0') {
941 namelen =
942 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
943 name_again:
944 deadpr = NULL;
945 FOREACH_PRISON_CHILD(ppr, tpr) {
946 if (tpr != pr && tpr->pr_ref > 0 &&
947 !strcmp(tpr->pr_name + namelen, name)) {
948 if (pr == NULL &&
949 cuflags != JAIL_CREATE) {
950 mtx_lock(&tpr->pr_mtx);
951 if (tpr->pr_ref > 0) {
952 /*
953 * Use this jail
954 * for updates.
955 */
956 if (tpr->pr_uref > 0) {
957 pr = tpr;
958 break;
959 }
960 deadpr = tpr;
961 }
962 mtx_unlock(&tpr->pr_mtx);
963 } else if (tpr->pr_uref > 0) {
964 /*
965 * Create, or update(jid):
966 * name must not exist in an
967 * active sibling jail.
968 */
969 error = EEXIST;
970 if (pr != NULL)
971 mtx_unlock(&pr->pr_mtx);
972 vfs_opterror(opts,
973 "jail \"%s\" already exists",
974 name);
975 goto done_unlock_list;
976 }
977 }
978 }
979 /* If no active jail is found, use a dying one. */
980 if (deadpr != NULL && pr == NULL) {
981 if (flags & JAIL_DYING) {
982 mtx_lock(&deadpr->pr_mtx);
983 if (deadpr->pr_ref == 0) {
984 mtx_unlock(&deadpr->pr_mtx);
985 goto name_again;
986 }
987 pr = deadpr;
988 } else if (cuflags == JAIL_UPDATE) {
989 error = ENOENT;
990 vfs_opterror(opts,
991 "jail \"%s\" is dying", name);
992 goto done_unlock_list;
993 }
994 }
995 /* Update: name must exist if no jid. */
996 else if (cuflags == JAIL_UPDATE && pr == NULL) {
997 error = ENOENT;
998 vfs_opterror(opts, "jail \"%s\" not found",
999 name);
1000 goto done_unlock_list;
1001 }
1002 }
1003 }
1004 /* Update: must provide a jid or name. */
1005 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1006 error = ENOENT;
1007 vfs_opterror(opts, "update specified no jail");
1008 goto done_unlock_list;
1009 }
1010
1011 /* If there's no prison to update, create a new one and link it in. */
1012 if (pr == NULL) {
1013 created = 1;
1014 mtx_lock(&ppr->pr_mtx);
1015 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1016 mtx_unlock(&ppr->pr_mtx);
1017 error = ENOENT;
1018 vfs_opterror(opts, "parent jail went away!");
1019 goto done_unlock_list;
1020 }
1021 ppr->pr_ref++;
1022 ppr->pr_uref++;
1023 mtx_unlock(&ppr->pr_mtx);
1024 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1025 if (jid == 0) {
1026 /* Find the next free jid. */
1027 jid = lastprid + 1;
1028 findnext:
1029 if (jid == JAIL_MAX)
1030 jid = 1;
1031 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1032 if (tpr->pr_id < jid)
1033 continue;
1034 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1035 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1036 break;
1037 }
1038 if (jid == lastprid) {
1039 error = EAGAIN;
1040 vfs_opterror(opts,
1041 "no available jail IDs");
1042 free(pr, M_PRISON);
1043 prison_deref(ppr, PD_DEREF |
1044 PD_DEUREF | PD_LIST_XLOCKED);
1045 goto done_releroot;
1046 }
1047 jid++;
1048 goto findnext;
1049 }
1050 lastprid = jid;
1051 } else {
1052 /*
1053 * The jail already has a jid (that did not yet exist),
1054 * so just find where to insert it.
1055 */
1056 TAILQ_FOREACH(tpr, &allprison, pr_list)
1057 if (tpr->pr_id >= jid) {
1058 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1059 break;
1060 }
1061 }
1062 if (tpr == NULL)
1063 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1064 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1065 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1066 tpr->pr_prisoncount++;
1067
1068 pr->pr_parent = ppr;
1069 pr->pr_id = jid;
1070
1071 /* Set some default values, and inherit some from the parent. */
1072 if (name == NULL)
1073 name = "";
1074 if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1075 if (host == NULL)
1076 host = ppr->pr_hostname;
1077 if (domain == NULL)
1078 domain = ppr->pr_domainname;
1079 if (uuid == NULL)
1080 uuid = ppr->pr_hostuuid;
1081 if (!gothid)
1082 hid = ppr->pr_hostid;
1083 }
1084 if (path == NULL) {
1085 path = "/";
1086 root = mypr->pr_root;
1087 vref(root);
1088 }
1089#ifdef INET
1090 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1091 pr->pr_ip4s = ppr->pr_ip4s;
1092 if (ppr->pr_ip4 != NULL) {
1093 pr->pr_ip4 = malloc(pr->pr_ip4s *
1094 sizeof(struct in_addr), M_PRISON, M_WAITOK);
1095 bcopy(ppr->pr_ip4, pr->pr_ip4,
1096 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1097 }
1098#endif
1099#ifdef INET6
1100 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1101 pr->pr_ip6s = ppr->pr_ip6s;
1102 if (ppr->pr_ip6 != NULL) {
1103 pr->pr_ip6 = malloc(pr->pr_ip6s *
1104 sizeof(struct in6_addr), M_PRISON, M_WAITOK);
1105 bcopy(ppr->pr_ip6, pr->pr_ip6,
1106 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1107 }
1108#endif
1109 pr->pr_securelevel = ppr->pr_securelevel;
1110 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1111 pr->pr_enforce_statfs = ppr->pr_enforce_statfs;
1112
1113 LIST_INIT(&pr->pr_children);
1114 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1115
| 577
578 pr_allow = ch_allow = 0;
579 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
580 fi++) {
581 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
582 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
583 }
584 ch_allow |= pr_allow;
585
586 error = vfs_getopt(opts, "name", (void **)&name, &len);
587 if (error == ENOENT)
588 name = NULL;
589 else if (error != 0)
590 goto done_free;
591 else {
592 if (len == 0 || name[len - 1] != '\0') {
593 error = EINVAL;
594 goto done_free;
595 }
596 if (len > MAXHOSTNAMELEN) {
597 error = ENAMETOOLONG;
598 goto done_free;
599 }
600 }
601
602 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
603 if (error == ENOENT)
604 host = NULL;
605 else if (error != 0)
606 goto done_free;
607 else {
608 ch_flags |= PR_HOST;
609 pr_flags |= PR_HOST;
610 if (len == 0 || host[len - 1] != '\0') {
611 error = EINVAL;
612 goto done_free;
613 }
614 if (len > MAXHOSTNAMELEN) {
615 error = ENAMETOOLONG;
616 goto done_free;
617 }
618 }
619
620 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
621 if (error == ENOENT)
622 domain = NULL;
623 else if (error != 0)
624 goto done_free;
625 else {
626 ch_flags |= PR_HOST;
627 pr_flags |= PR_HOST;
628 if (len == 0 || domain[len - 1] != '\0') {
629 error = EINVAL;
630 goto done_free;
631 }
632 if (len > MAXHOSTNAMELEN) {
633 error = ENAMETOOLONG;
634 goto done_free;
635 }
636 }
637
638 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
639 if (error == ENOENT)
640 uuid = NULL;
641 else if (error != 0)
642 goto done_free;
643 else {
644 ch_flags |= PR_HOST;
645 pr_flags |= PR_HOST;
646 if (len == 0 || uuid[len - 1] != '\0') {
647 error = EINVAL;
648 goto done_free;
649 }
650 if (len > HOSTUUIDLEN) {
651 error = ENAMETOOLONG;
652 goto done_free;
653 }
654 }
655
656#ifdef COMPAT_IA32
657 if (td->td_proc->p_sysent->sv_flags & SV_IA32) {
658 uint32_t hid32;
659
660 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
661 hid = hid32;
662 } else
663#endif
664 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
665 if (error == ENOENT)
666 gothid = 0;
667 else if (error != 0)
668 goto done_free;
669 else {
670 gothid = 1;
671 ch_flags |= PR_HOST;
672 pr_flags |= PR_HOST;
673 }
674
675 /* This might be the second time around for this option. */
676#ifdef INET
677 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
678 if (error == ENOENT)
679 ip4s = -1;
680 else if (error != 0)
681 goto done_free;
682 else if (ip4s & (sizeof(*ip4) - 1)) {
683 error = EINVAL;
684 goto done_free;
685 } else {
686 ch_flags |= PR_IP4_USER;
687 pr_flags |= PR_IP4_USER;
688 if (ip4s > 0) {
689 ip4s /= sizeof(*ip4);
690 if (ip4s > jail_max_af_ips) {
691 error = EINVAL;
692 vfs_opterror(opts, "too many IPv4 addresses");
693 goto done_errmsg;
694 }
695 if (ip4a < ip4s) {
696 ip4a = ip4s;
697 free(ip4, M_PRISON);
698 ip4 = NULL;
699 }
700 if (ip4 == NULL)
701 ip4 = malloc(ip4a * sizeof(*ip4), M_PRISON,
702 M_WAITOK);
703 bcopy(op, ip4, ip4s * sizeof(*ip4));
704 /*
705 * IP addresses are all sorted but ip[0] to preserve
706 * the primary IP address as given from userland.
707 * This special IP is used for unbound outgoing
708 * connections as well for "loopback" traffic.
709 */
710 if (ip4s > 1)
711 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
712 /*
713 * Check for duplicate addresses and do some simple
714 * zero and broadcast checks. If users give other bogus
715 * addresses it is their problem.
716 *
717 * We do not have to care about byte order for these
718 * checks so we will do them in NBO.
719 */
720 for (ii = 0; ii < ip4s; ii++) {
721 if (ip4[ii].s_addr == INADDR_ANY ||
722 ip4[ii].s_addr == INADDR_BROADCAST) {
723 error = EINVAL;
724 goto done_free;
725 }
726 if ((ii+1) < ip4s &&
727 (ip4[0].s_addr == ip4[ii+1].s_addr ||
728 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
729 error = EINVAL;
730 goto done_free;
731 }
732 }
733 }
734 }
735#endif
736
737#ifdef INET6
738 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
739 if (error == ENOENT)
740 ip6s = -1;
741 else if (error != 0)
742 goto done_free;
743 else if (ip6s & (sizeof(*ip6) - 1)) {
744 error = EINVAL;
745 goto done_free;
746 } else {
747 ch_flags |= PR_IP6_USER;
748 pr_flags |= PR_IP6_USER;
749 if (ip6s > 0) {
750 ip6s /= sizeof(*ip6);
751 if (ip6s > jail_max_af_ips) {
752 error = EINVAL;
753 vfs_opterror(opts, "too many IPv6 addresses");
754 goto done_errmsg;
755 }
756 if (ip6a < ip6s) {
757 ip6a = ip6s;
758 free(ip6, M_PRISON);
759 ip6 = NULL;
760 }
761 if (ip6 == NULL)
762 ip6 = malloc(ip6a * sizeof(*ip6), M_PRISON,
763 M_WAITOK);
764 bcopy(op, ip6, ip6s * sizeof(*ip6));
765 if (ip6s > 1)
766 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
767 for (ii = 0; ii < ip6s; ii++) {
768 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
769 error = EINVAL;
770 goto done_free;
771 }
772 if ((ii+1) < ip6s &&
773 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
774 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
775 {
776 error = EINVAL;
777 goto done_free;
778 }
779 }
780 }
781 }
782#endif
783
784 root = NULL;
785 error = vfs_getopt(opts, "path", (void **)&path, &len);
786 if (error == ENOENT)
787 path = NULL;
788 else if (error != 0)
789 goto done_free;
790 else {
791 if (flags & JAIL_UPDATE) {
792 error = EINVAL;
793 vfs_opterror(opts,
794 "path cannot be changed after creation");
795 goto done_errmsg;
796 }
797 if (len == 0 || path[len - 1] != '\0') {
798 error = EINVAL;
799 goto done_free;
800 }
801 if (len < 2 || (len == 2 && path[0] == '/'))
802 path = NULL;
803 else {
804 /* Leave room for a real-root full pathname. */
805 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
806 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
807 error = ENAMETOOLONG;
808 goto done_free;
809 }
810 NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
811 path, td);
812 error = namei(&nd);
813 if (error)
814 goto done_free;
815 vfslocked = NDHASGIANT(&nd);
816 root = nd.ni_vp;
817 NDFREE(&nd, NDF_ONLY_PNBUF);
818 if (root->v_type != VDIR) {
819 error = ENOTDIR;
820 vrele(root);
821 VFS_UNLOCK_GIANT(vfslocked);
822 goto done_free;
823 }
824 VFS_UNLOCK_GIANT(vfslocked);
825 }
826 }
827
828 /*
829 * Grab the allprison lock before letting modules check their
830 * parameters. Once we have it, do not let go so we'll have a
831 * consistent view of the OSD list.
832 */
833 sx_xlock(&allprison_lock);
834 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
835 if (error)
836 goto done_unlock_list;
837
838 /* By now, all parameters should have been noted. */
839 TAILQ_FOREACH(opt, opts, link) {
840 if (!opt->seen && strcmp(opt->name, "errmsg")) {
841 error = EINVAL;
842 vfs_opterror(opts, "unknown parameter: %s", opt->name);
843 goto done_unlock_list;
844 }
845 }
846
847 /*
848 * See if we are creating a new record or updating an existing one.
849 * This abuses the file error codes ENOENT and EEXIST.
850 */
851 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
852 if (!cuflags) {
853 error = EINVAL;
854 vfs_opterror(opts, "no valid operation (create or update)");
855 goto done_unlock_list;
856 }
857 pr = NULL;
858 if (jid != 0) {
859 /*
860 * See if a requested jid already exists. There is an
861 * information leak here if the jid exists but is not within
862 * the caller's jail hierarchy. Jail creators will get EEXIST
863 * even though they cannot see the jail, and CREATE | UPDATE
864 * will return ENOENT which is not normally a valid error.
865 */
866 if (jid < 0) {
867 error = EINVAL;
868 vfs_opterror(opts, "negative jid");
869 goto done_unlock_list;
870 }
871 pr = prison_find(jid);
872 if (pr != NULL) {
873 ppr = pr->pr_parent;
874 /* Create: jid must not exist. */
875 if (cuflags == JAIL_CREATE) {
876 mtx_unlock(&pr->pr_mtx);
877 error = EEXIST;
878 vfs_opterror(opts, "jail %d already exists",
879 jid);
880 goto done_unlock_list;
881 }
882 if (!prison_ischild(mypr, pr)) {
883 mtx_unlock(&pr->pr_mtx);
884 pr = NULL;
885 } else if (pr->pr_uref == 0) {
886 if (!(flags & JAIL_DYING)) {
887 mtx_unlock(&pr->pr_mtx);
888 error = ENOENT;
889 vfs_opterror(opts, "jail %d is dying",
890 jid);
891 goto done_unlock_list;
892 } else if ((flags & JAIL_ATTACH) ||
893 (pr_flags & PR_PERSIST)) {
894 /*
895 * A dying jail might be resurrected
896 * (via attach or persist), but first
897 * it must determine if another jail
898 * has claimed its name. Accomplish
899 * this by implicitly re-setting the
900 * name.
901 */
902 if (name == NULL)
903 name = prison_name(mypr, pr);
904 }
905 }
906 }
907 if (pr == NULL) {
908 /* Update: jid must exist. */
909 if (cuflags == JAIL_UPDATE) {
910 error = ENOENT;
911 vfs_opterror(opts, "jail %d not found", jid);
912 goto done_unlock_list;
913 }
914 }
915 }
916 /*
917 * If the caller provided a name, look for a jail by that name.
918 * This has different semantics for creates and updates keyed by jid
919 * (where the name must not already exist in a different jail),
920 * and updates keyed by the name itself (where the name must exist
921 * because that is the jail being updated).
922 */
923 if (name != NULL) {
924 p = strrchr(name, '.');
925 if (p != NULL) {
926 /*
927 * This is a hierarchical name. Split it into the
928 * parent and child names, and make sure the parent
929 * exists or matches an already found jail.
930 */
931 *p = '\0';
932 if (pr != NULL) {
933 if (strncmp(name, ppr->pr_name, p - name) ||
934 ppr->pr_name[p - name] != '\0') {
935 mtx_unlock(&pr->pr_mtx);
936 error = EINVAL;
937 vfs_opterror(opts,
938 "cannot change jail's parent");
939 goto done_unlock_list;
940 }
941 } else {
942 ppr = prison_find_name(mypr, name);
943 if (ppr == NULL) {
944 error = ENOENT;
945 vfs_opterror(opts,
946 "jail \"%s\" not found", name);
947 goto done_unlock_list;
948 }
949 mtx_unlock(&ppr->pr_mtx);
950 }
951 name = p + 1;
952 }
953 if (name[0] != '\0') {
954 namelen =
955 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
956 name_again:
957 deadpr = NULL;
958 FOREACH_PRISON_CHILD(ppr, tpr) {
959 if (tpr != pr && tpr->pr_ref > 0 &&
960 !strcmp(tpr->pr_name + namelen, name)) {
961 if (pr == NULL &&
962 cuflags != JAIL_CREATE) {
963 mtx_lock(&tpr->pr_mtx);
964 if (tpr->pr_ref > 0) {
965 /*
966 * Use this jail
967 * for updates.
968 */
969 if (tpr->pr_uref > 0) {
970 pr = tpr;
971 break;
972 }
973 deadpr = tpr;
974 }
975 mtx_unlock(&tpr->pr_mtx);
976 } else if (tpr->pr_uref > 0) {
977 /*
978 * Create, or update(jid):
979 * name must not exist in an
980 * active sibling jail.
981 */
982 error = EEXIST;
983 if (pr != NULL)
984 mtx_unlock(&pr->pr_mtx);
985 vfs_opterror(opts,
986 "jail \"%s\" already exists",
987 name);
988 goto done_unlock_list;
989 }
990 }
991 }
992 /* If no active jail is found, use a dying one. */
993 if (deadpr != NULL && pr == NULL) {
994 if (flags & JAIL_DYING) {
995 mtx_lock(&deadpr->pr_mtx);
996 if (deadpr->pr_ref == 0) {
997 mtx_unlock(&deadpr->pr_mtx);
998 goto name_again;
999 }
1000 pr = deadpr;
1001 } else if (cuflags == JAIL_UPDATE) {
1002 error = ENOENT;
1003 vfs_opterror(opts,
1004 "jail \"%s\" is dying", name);
1005 goto done_unlock_list;
1006 }
1007 }
1008 /* Update: name must exist if no jid. */
1009 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1010 error = ENOENT;
1011 vfs_opterror(opts, "jail \"%s\" not found",
1012 name);
1013 goto done_unlock_list;
1014 }
1015 }
1016 }
1017 /* Update: must provide a jid or name. */
1018 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1019 error = ENOENT;
1020 vfs_opterror(opts, "update specified no jail");
1021 goto done_unlock_list;
1022 }
1023
1024 /* If there's no prison to update, create a new one and link it in. */
1025 if (pr == NULL) {
1026 created = 1;
1027 mtx_lock(&ppr->pr_mtx);
1028 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1029 mtx_unlock(&ppr->pr_mtx);
1030 error = ENOENT;
1031 vfs_opterror(opts, "parent jail went away!");
1032 goto done_unlock_list;
1033 }
1034 ppr->pr_ref++;
1035 ppr->pr_uref++;
1036 mtx_unlock(&ppr->pr_mtx);
1037 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1038 if (jid == 0) {
1039 /* Find the next free jid. */
1040 jid = lastprid + 1;
1041 findnext:
1042 if (jid == JAIL_MAX)
1043 jid = 1;
1044 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1045 if (tpr->pr_id < jid)
1046 continue;
1047 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1048 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1049 break;
1050 }
1051 if (jid == lastprid) {
1052 error = EAGAIN;
1053 vfs_opterror(opts,
1054 "no available jail IDs");
1055 free(pr, M_PRISON);
1056 prison_deref(ppr, PD_DEREF |
1057 PD_DEUREF | PD_LIST_XLOCKED);
1058 goto done_releroot;
1059 }
1060 jid++;
1061 goto findnext;
1062 }
1063 lastprid = jid;
1064 } else {
1065 /*
1066 * The jail already has a jid (that did not yet exist),
1067 * so just find where to insert it.
1068 */
1069 TAILQ_FOREACH(tpr, &allprison, pr_list)
1070 if (tpr->pr_id >= jid) {
1071 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1072 break;
1073 }
1074 }
1075 if (tpr == NULL)
1076 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1077 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1078 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1079 tpr->pr_prisoncount++;
1080
1081 pr->pr_parent = ppr;
1082 pr->pr_id = jid;
1083
1084 /* Set some default values, and inherit some from the parent. */
1085 if (name == NULL)
1086 name = "";
1087 if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1088 if (host == NULL)
1089 host = ppr->pr_hostname;
1090 if (domain == NULL)
1091 domain = ppr->pr_domainname;
1092 if (uuid == NULL)
1093 uuid = ppr->pr_hostuuid;
1094 if (!gothid)
1095 hid = ppr->pr_hostid;
1096 }
1097 if (path == NULL) {
1098 path = "/";
1099 root = mypr->pr_root;
1100 vref(root);
1101 }
1102#ifdef INET
1103 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1104 pr->pr_ip4s = ppr->pr_ip4s;
1105 if (ppr->pr_ip4 != NULL) {
1106 pr->pr_ip4 = malloc(pr->pr_ip4s *
1107 sizeof(struct in_addr), M_PRISON, M_WAITOK);
1108 bcopy(ppr->pr_ip4, pr->pr_ip4,
1109 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1110 }
1111#endif
1112#ifdef INET6
1113 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1114 pr->pr_ip6s = ppr->pr_ip6s;
1115 if (ppr->pr_ip6 != NULL) {
1116 pr->pr_ip6 = malloc(pr->pr_ip6s *
1117 sizeof(struct in6_addr), M_PRISON, M_WAITOK);
1118 bcopy(ppr->pr_ip6, pr->pr_ip6,
1119 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1120 }
1121#endif
1122 pr->pr_securelevel = ppr->pr_securelevel;
1123 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1124 pr->pr_enforce_statfs = ppr->pr_enforce_statfs;
1125
1126 LIST_INIT(&pr->pr_children);
1127 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1128
|
| 1129#ifdef VIMAGE
1130 /* Allocate a new vnet if specified. */
1131 pr->pr_vnet = (pr_flags & PR_VNET)
1132 ? vnet_alloc() : ppr->pr_vnet;
1133#endif
|
1116 /*
1117 * Allocate a dedicated cpuset for each jail.
1118 * Unlike other initial settings, this may return an erorr.
1119 */
1120 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1121 if (error) {
1122 prison_deref(pr, PD_LIST_XLOCKED);
1123 goto done_releroot;
1124 }
1125
1126 mtx_lock(&pr->pr_mtx);
1127 /*
1128 * New prisons do not yet have a reference, because we do not
1129 * want other to see the incomplete prison once the
1130 * allprison_lock is downgraded.
1131 */
1132 } else {
1133 created = 0;
1134 /*
1135 * Grab a reference for existing prisons, to ensure they
1136 * continue to exist for the duration of the call.
1137 */
1138 pr->pr_ref++;
1139 }
1140
1141 /* Do final error checking before setting anything. */
1142 if (gotslevel) {
1143 if (slevel < ppr->pr_securelevel) {
1144 error = EPERM;
1145 goto done_deref_locked;
1146 }
1147 }
1148 if (gotenforce) {
1149 if (enforce < ppr->pr_enforce_statfs) {
1150 error = EPERM;
1151 goto done_deref_locked;
1152 }
1153 }
1154#ifdef INET
1155 if (ch_flags & PR_IP4_USER) {
1156 if (ppr->pr_flags & PR_IP4) {
1157 if (!(pr_flags & PR_IP4_USER)) {
1158 /*
1159 * Silently ignore attempts to make the IP
1160 * addresses unrestricted when the parent is
1161 * restricted; in other words, interpret
1162 * "unrestricted" as "as unrestricted as
1163 * possible".
1164 */
1165 ip4s = ppr->pr_ip4s;
1166 if (ip4s == 0) {
1167 free(ip4, M_PRISON);
1168 ip4 = NULL;
1169 } else if (ip4s <= ip4a) {
1170 /* Inherit the parent's address(es). */
1171 bcopy(ppr->pr_ip4, ip4,
1172 ip4s * sizeof(*ip4));
1173 } else {
1174 /*
1175 * There's no room for the parent's
1176 * address list. Allocate some more.
1177 */
1178 ip4a = ip4s;
1179 free(ip4, M_PRISON);
1180 ip4 = malloc(ip4a * sizeof(*ip4),
1181 M_PRISON, M_NOWAIT);
1182 if (ip4 != NULL)
1183 bcopy(ppr->pr_ip4, ip4,
1184 ip4s * sizeof(*ip4));
1185 else {
1186 /* Allocation failed without
1187 * sleeping. Unlocking the
1188 * prison now will invalidate
1189 * some checks and prematurely
1190 * show an unfinished new jail.
1191 * So let go of everything and
1192 * start over.
1193 */
1194 prison_deref(pr, created
1195 ? PD_LOCKED |
1196 PD_LIST_XLOCKED
1197 : PD_DEREF | PD_LOCKED |
1198 PD_LIST_XLOCKED);
1199 if (root != NULL) {
1200 vfslocked =
1201 VFS_LOCK_GIANT(
1202 root->v_mount);
1203 vrele(root);
1204 VFS_UNLOCK_GIANT(
1205 vfslocked);
1206 }
1207 ip4 = malloc(ip4a *
1208 sizeof(*ip4), M_PRISON,
1209 M_WAITOK);
1210 goto again;
1211 }
1212 }
1213 } else if (ip4s > 0) {
1214 /*
1215 * Make sure the new set of IP addresses is a
1216 * subset of the parent's list. Don't worry
1217 * about the parent being unlocked, as any
1218 * setting is done with allprison_lock held.
1219 */
1220 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1221 if (ip4[0].s_addr ==
1222 ppr->pr_ip4[ij].s_addr)
1223 break;
1224 if (ij == ppr->pr_ip4s) {
1225 error = EPERM;
1226 goto done_deref_locked;
1227 }
1228 if (ip4s > 1) {
1229 for (ii = ij = 1; ii < ip4s; ii++) {
1230 if (ip4[ii].s_addr ==
1231 ppr->pr_ip4[0].s_addr)
1232 continue;
1233 for (; ij < ppr->pr_ip4s; ij++)
1234 if (ip4[ii].s_addr ==
1235 ppr->pr_ip4[ij].s_addr)
1236 break;
1237 if (ij == ppr->pr_ip4s)
1238 break;
1239 }
1240 if (ij == ppr->pr_ip4s) {
1241 error = EPERM;
1242 goto done_deref_locked;
1243 }
1244 }
1245 }
1246 }
1247 if (ip4s > 0) {
1248 /*
1249 * Check for conflicting IP addresses. We permit them
1250 * if there is no more than one IP on each jail. If
1251 * there is a duplicate on a jail with more than one
1252 * IP stop checking and return error.
1253 */
1254 FOREACH_PRISON_DESCENDANT(&prison0, tpr, descend) {
1255 if (tpr == pr || tpr->pr_uref == 0) {
1256 descend = 0;
1257 continue;
1258 }
1259 if (!(tpr->pr_flags & PR_IP4_USER))
1260 continue;
1261 descend = 0;
1262 if (tpr->pr_ip4 == NULL ||
1263 (ip4s == 1 && tpr->pr_ip4s == 1))
1264 continue;
1265 for (ii = 0; ii < ip4s; ii++) {
1266 if (_prison_check_ip4(tpr,
1267 &ip4[ii]) == 0) {
1268 error = EADDRINUSE;
1269 vfs_opterror(opts,
1270 "IPv4 addresses clash");
1271 goto done_deref_locked;
1272 }
1273 }
1274 }
1275 }
1276 }
1277#endif
1278#ifdef INET6
1279 if (ch_flags & PR_IP6_USER) {
1280 if (ppr->pr_flags & PR_IP6) {
1281 if (!(pr_flags & PR_IP6_USER)) {
1282 /*
1283 * Silently ignore attempts to make the IP
1284 * addresses unrestricted when the parent is
1285 * restricted.
1286 */
1287 ip6s = ppr->pr_ip6s;
1288 if (ip6s == 0) {
1289 free(ip6, M_PRISON);
1290 ip6 = NULL;
1291 } else if (ip6s <= ip6a) {
1292 /* Inherit the parent's address(es). */
1293 bcopy(ppr->pr_ip6, ip6,
1294 ip6s * sizeof(*ip6));
1295 } else {
1296 /*
1297 * There's no room for the parent's
1298 * address list.
1299 */
1300 ip6a = ip6s;
1301 free(ip6, M_PRISON);
1302 ip6 = malloc(ip6a * sizeof(*ip6),
1303 M_PRISON, M_NOWAIT);
1304 if (ip6 != NULL)
1305 bcopy(ppr->pr_ip6, ip6,
1306 ip6s * sizeof(*ip6));
1307 else {
1308 prison_deref(pr, created
1309 ? PD_LOCKED |
1310 PD_LIST_XLOCKED
1311 : PD_DEREF | PD_LOCKED |
1312 PD_LIST_XLOCKED);
1313 if (root != NULL) {
1314 vfslocked =
1315 VFS_LOCK_GIANT(
1316 root->v_mount);
1317 vrele(root);
1318 VFS_UNLOCK_GIANT(
1319 vfslocked);
1320 }
1321 ip6 = malloc(ip6a *
1322 sizeof(*ip6), M_PRISON,
1323 M_WAITOK);
1324 goto again;
1325 }
1326 }
1327 } else if (ip6s > 0) {
1328 /*
1329 * Make sure the new set of IP addresses is a
1330 * subset of the parent's list.
1331 */
1332 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1333 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1334 &ppr->pr_ip6[ij]))
1335 break;
1336 if (ij == ppr->pr_ip6s) {
1337 error = EPERM;
1338 goto done_deref_locked;
1339 }
1340 if (ip6s > 1) {
1341 for (ii = ij = 1; ii < ip6s; ii++) {
1342 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1343 &ppr->pr_ip6[0]))
1344 continue;
1345 for (; ij < ppr->pr_ip6s; ij++)
1346 if (IN6_ARE_ADDR_EQUAL(
1347 &ip6[ii],
1348 &ppr->pr_ip6[ij]))
1349 break;
1350 if (ij == ppr->pr_ip6s)
1351 break;
1352 }
1353 if (ij == ppr->pr_ip6s) {
1354 error = EPERM;
1355 goto done_deref_locked;
1356 }
1357 }
1358 }
1359 }
1360 if (ip6s > 0) {
1361 /* Check for conflicting IP addresses. */
1362 FOREACH_PRISON_DESCENDANT(&prison0, tpr, descend) {
1363 if (tpr == pr || tpr->pr_uref == 0) {
1364 descend = 0;
1365 continue;
1366 }
1367 if (!(tpr->pr_flags & PR_IP6_USER))
1368 continue;
1369 descend = 0;
1370 if (tpr->pr_ip6 == NULL ||
1371 (ip6s == 1 && tpr->pr_ip6s == 1))
1372 continue;
1373 for (ii = 0; ii < ip6s; ii++) {
1374 if (_prison_check_ip6(tpr,
1375 &ip6[ii]) == 0) {
1376 error = EADDRINUSE;
1377 vfs_opterror(opts,
1378 "IPv6 addresses clash");
1379 goto done_deref_locked;
1380 }
1381 }
1382 }
1383 }
1384 }
1385#endif
1386 onamelen = namelen = 0;
1387 if (name != NULL) {
1388 /* Give a default name of the jid. */
1389 if (name[0] == '\0')
1390 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1391 else if (strtoul(name, &p, 10) != jid && *p == '\0') {
1392 error = EINVAL;
1393 vfs_opterror(opts, "name cannot be numeric");
1394 goto done_deref_locked;
1395 }
1396 /*
1397 * Make sure the name isn't too long for the prison or its
1398 * children.
1399 */
1400 onamelen = strlen(pr->pr_name);
1401 namelen = strlen(name);
1402 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1403 error = ENAMETOOLONG;
1404 goto done_deref_locked;
1405 }
1406 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1407 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1408 sizeof(pr->pr_name)) {
1409 error = ENAMETOOLONG;
1410 goto done_deref_locked;
1411 }
1412 }
1413 }
1414 if (pr_allow & ~ppr->pr_allow) {
1415 error = EPERM;
1416 goto done_deref_locked;
1417 }
1418
1419 /* Set the parameters of the prison. */
1420#ifdef INET
1421 redo_ip4 = 0;
1422 if (ch_flags & PR_IP4_USER) {
1423 if (pr_flags & PR_IP4_USER) {
1424 /* Some restriction set. */
1425 pr->pr_flags |= PR_IP4;
1426 if (ip4s >= 0) {
1427 free(pr->pr_ip4, M_PRISON);
1428 pr->pr_ip4s = ip4s;
1429 pr->pr_ip4 = ip4;
1430 ip4 = NULL;
1431 }
1432 } else if (ppr->pr_flags & PR_IP4) {
1433 /* This restriction cleared, but keep inherited. */
1434 free(pr->pr_ip4, M_PRISON);
1435 pr->pr_ip4s = ip4s;
1436 pr->pr_ip4 = ip4;
1437 ip4 = NULL;
1438 } else {
1439 /* Restriction cleared, now unrestricted. */
1440 pr->pr_flags &= ~PR_IP4;
1441 free(pr->pr_ip4, M_PRISON);
1442 pr->pr_ip4s = 0;
1443 }
1444 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1445 if (prison_restrict_ip4(tpr, NULL)) {
1446 redo_ip4 = 1;
1447 descend = 0;
1448 }
1449 }
1450 }
1451#endif
1452#ifdef INET6
1453 redo_ip6 = 0;
1454 if (ch_flags & PR_IP6_USER) {
1455 if (pr_flags & PR_IP6_USER) {
1456 /* Some restriction set. */
1457 pr->pr_flags |= PR_IP6;
1458 if (ip6s >= 0) {
1459 free(pr->pr_ip6, M_PRISON);
1460 pr->pr_ip6s = ip6s;
1461 pr->pr_ip6 = ip6;
1462 ip6 = NULL;
1463 }
1464 } else if (ppr->pr_flags & PR_IP6) {
1465 /* This restriction cleared, but keep inherited. */
1466 free(pr->pr_ip6, M_PRISON);
1467 pr->pr_ip6s = ip6s;
1468 pr->pr_ip6 = ip6;
1469 ip6 = NULL;
1470 } else {
1471 /* Restriction cleared, now unrestricted. */
1472 pr->pr_flags &= ~PR_IP6;
1473 free(pr->pr_ip6, M_PRISON);
1474 pr->pr_ip6s = 0;
1475 }
1476 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1477 if (prison_restrict_ip6(tpr, NULL)) {
1478 redo_ip6 = 1;
1479 descend = 0;
1480 }
1481 }
1482 }
1483#endif
1484 if (gotslevel) {
1485 pr->pr_securelevel = slevel;
1486 /* Set all child jails to be at least this level. */
1487 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1488 if (tpr->pr_securelevel < slevel)
1489 tpr->pr_securelevel = slevel;
1490 }
1491 if (gotenforce) {
1492 pr->pr_enforce_statfs = enforce;
1493 /* Pass this restriction on to the children. */
1494 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1495 if (tpr->pr_enforce_statfs < enforce)
1496 tpr->pr_enforce_statfs = enforce;
1497 }
1498 if (name != NULL) {
1499 if (ppr == &prison0)
1500 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1501 else
1502 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1503 ppr->pr_name, name);
1504 /* Change this component of child names. */
1505 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1506 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1507 strlen(tpr->pr_name + onamelen) + 1);
1508 bcopy(pr->pr_name, tpr->pr_name, namelen);
1509 }
1510 }
1511 if (path != NULL) {
1512 /* Try to keep a real-rooted full pathname. */
1513 if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
1514 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1515 mypr->pr_path, path);
1516 else
1517 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1518 pr->pr_root = root;
1519 }
1520 if (PR_HOST & ch_flags & ~pr_flags) {
1521 if (pr->pr_flags & PR_HOST) {
1522 /*
1523 * Copy the parent's host info. As with pr_ip4 above,
1524 * the lack of a lock on the parent is not a problem;
1525 * it is always set with allprison_lock at least
1526 * shared, and is held exclusively here.
1527 */
1528 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1529 sizeof(pr->pr_hostname));
1530 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1531 sizeof(pr->pr_domainname));
1532 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1533 sizeof(pr->pr_hostuuid));
1534 pr->pr_hostid = pr->pr_parent->pr_hostid;
1535 }
1536 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1537 /* Set this prison, and any descendants without PR_HOST. */
1538 if (host != NULL)
1539 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1540 if (domain != NULL)
1541 strlcpy(pr->pr_domainname, domain,
1542 sizeof(pr->pr_domainname));
1543 if (uuid != NULL)
1544 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1545 if (gothid)
1546 pr->pr_hostid = hid;
1547 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1548 if (tpr->pr_flags & PR_HOST)
1549 descend = 0;
1550 else {
1551 if (host != NULL)
1552 strlcpy(tpr->pr_hostname,
1553 pr->pr_hostname,
1554 sizeof(tpr->pr_hostname));
1555 if (domain != NULL)
1556 strlcpy(tpr->pr_domainname,
1557 pr->pr_domainname,
1558 sizeof(tpr->pr_domainname));
1559 if (uuid != NULL)
1560 strlcpy(tpr->pr_hostuuid,
1561 pr->pr_hostuuid,
1562 sizeof(tpr->pr_hostuuid));
1563 if (gothid)
1564 tpr->pr_hostid = hid;
1565 }
1566 }
1567 }
1568 if ((tallow = ch_allow & ~pr_allow)) {
1569 /* Clear allow bits in all children. */
1570 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1571 tpr->pr_allow &= ~tallow;
1572 }
1573 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1574 /*
1575 * Persistent prisons get an extra reference, and prisons losing their
1576 * persist flag lose that reference. Only do this for existing prisons
1577 * for now, so new ones will remain unseen until after the module
1578 * handlers have completed.
1579 */
1580 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1581 if (pr_flags & PR_PERSIST) {
1582 pr->pr_ref++;
1583 pr->pr_uref++;
1584 } else {
1585 pr->pr_ref--;
1586 pr->pr_uref--;
1587 }
1588 }
1589 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1590 mtx_unlock(&pr->pr_mtx);
1591
1592 /* Locks may have prevented a complete restriction of child IP
1593 * addresses. If so, allocate some more memory and try again.
1594 */
1595#ifdef INET
1596 while (redo_ip4) {
1597 ip4s = pr->pr_ip4s;
1598 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1599 mtx_lock(&pr->pr_mtx);
1600 redo_ip4 = 0;
1601 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1602 if (prison_restrict_ip4(tpr, ip4)) {
1603 if (ip4 != NULL)
1604 ip4 = NULL;
1605 else
1606 redo_ip4 = 1;
1607 }
1608 }
1609 mtx_unlock(&pr->pr_mtx);
1610 }
1611#endif
1612#ifdef INET6
1613 while (redo_ip6) {
1614 ip6s = pr->pr_ip6s;
1615 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1616 mtx_lock(&pr->pr_mtx);
1617 redo_ip6 = 0;
1618 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1619 if (prison_restrict_ip6(tpr, ip6)) {
1620 if (ip6 != NULL)
1621 ip6 = NULL;
1622 else
1623 redo_ip6 = 1;
1624 }
1625 }
1626 mtx_unlock(&pr->pr_mtx);
1627 }
1628#endif
1629
1630 /* Let the modules do their work. */
1631 sx_downgrade(&allprison_lock);
1632 if (created) {
1633 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1634 if (error) {
1635 prison_deref(pr, PD_LIST_SLOCKED);
1636 goto done_errmsg;
1637 }
1638 }
1639 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1640 if (error) {
1641 prison_deref(pr, created
1642 ? PD_LIST_SLOCKED
1643 : PD_DEREF | PD_LIST_SLOCKED);
1644 goto done_errmsg;
1645 }
1646
1647 /* Attach this process to the prison if requested. */
1648 if (flags & JAIL_ATTACH) {
1649 mtx_lock(&pr->pr_mtx);
1650 error = do_jail_attach(td, pr);
1651 if (error) {
1652 vfs_opterror(opts, "attach failed");
1653 if (!created)
1654 prison_deref(pr, PD_DEREF);
1655 goto done_errmsg;
1656 }
1657 }
1658
1659 /*
1660 * Now that it is all there, drop the temporary reference from existing
1661 * prisons. Or add a reference to newly created persistent prisons
1662 * (which was not done earlier so that the prison would not be publicly
1663 * visible).
1664 */
1665 if (!created) {
1666 prison_deref(pr, (flags & JAIL_ATTACH)
1667 ? PD_DEREF
1668 : PD_DEREF | PD_LIST_SLOCKED);
1669 } else {
1670 if (pr_flags & PR_PERSIST) {
1671 mtx_lock(&pr->pr_mtx);
1672 pr->pr_ref++;
1673 pr->pr_uref++;
1674 mtx_unlock(&pr->pr_mtx);
1675 }
1676 if (!(flags & JAIL_ATTACH))
1677 sx_sunlock(&allprison_lock);
1678 }
1679 td->td_retval[0] = pr->pr_id;
1680 goto done_errmsg;
1681
1682 done_deref_locked:
1683 prison_deref(pr, created
1684 ? PD_LOCKED | PD_LIST_XLOCKED
1685 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1686 goto done_releroot;
1687 done_unlock_list:
1688 sx_xunlock(&allprison_lock);
1689 done_releroot:
1690 if (root != NULL) {
1691 vfslocked = VFS_LOCK_GIANT(root->v_mount);
1692 vrele(root);
1693 VFS_UNLOCK_GIANT(vfslocked);
1694 }
1695 done_errmsg:
1696 if (error) {
1697 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1698 if (errmsg_len > 0) {
1699 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1700 if (errmsg_pos > 0) {
1701 if (optuio->uio_segflg == UIO_SYSSPACE)
1702 bcopy(errmsg,
1703 optuio->uio_iov[errmsg_pos].iov_base,
1704 errmsg_len);
1705 else
1706 copyout(errmsg,
1707 optuio->uio_iov[errmsg_pos].iov_base,
1708 errmsg_len);
1709 }
1710 }
1711 }
1712 done_free:
1713#ifdef INET
1714 free(ip4, M_PRISON);
1715#endif
1716#ifdef INET6
1717 free(ip6, M_PRISON);
1718#endif
1719 vfs_freeopts(opts);
1720 return (error);
1721}
1722
1723
1724/*
1725 * struct jail_get_args {
1726 * struct iovec *iovp;
1727 * unsigned int iovcnt;
1728 * int flags;
1729 * };
1730 */
1731int
1732jail_get(struct thread *td, struct jail_get_args *uap)
1733{
1734 struct uio *auio;
1735 int error;
1736
1737 /* Check that we have an even number of iovecs. */
1738 if (uap->iovcnt & 1)
1739 return (EINVAL);
1740
1741 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1742 if (error)
1743 return (error);
1744 error = kern_jail_get(td, auio, uap->flags);
1745 if (error == 0)
1746 error = copyout(auio->uio_iov, uap->iovp,
1747 uap->iovcnt * sizeof (struct iovec));
1748 free(auio, M_IOV);
1749 return (error);
1750}
1751
1752int
1753kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1754{
1755 struct prison *pr, *mypr;
1756 struct vfsopt *opt;
1757 struct vfsoptlist *opts;
1758 char *errmsg, *name;
1759 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1760
1761 if (flags & ~JAIL_GET_MASK)
1762 return (EINVAL);
1763
1764 /* Get the parameter list. */
1765 error = vfs_buildopts(optuio, &opts);
1766 if (error)
1767 return (error);
1768 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1769 mypr = td->td_ucred->cr_prison;
1770
1771 /*
1772 * Find the prison specified by one of: lastjid, jid, name.
1773 */
1774 sx_slock(&allprison_lock);
1775 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1776 if (error == 0) {
1777 TAILQ_FOREACH(pr, &allprison, pr_list) {
1778 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1779 mtx_lock(&pr->pr_mtx);
1780 if (pr->pr_ref > 0 &&
1781 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1782 break;
1783 mtx_unlock(&pr->pr_mtx);
1784 }
1785 }
1786 if (pr != NULL)
1787 goto found_prison;
1788 error = ENOENT;
1789 vfs_opterror(opts, "no jail after %d", jid);
1790 goto done_unlock_list;
1791 } else if (error != ENOENT)
1792 goto done_unlock_list;
1793
1794 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1795 if (error == 0) {
1796 if (jid != 0) {
1797 pr = prison_find_child(mypr, jid);
1798 if (pr != NULL) {
1799 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1800 mtx_unlock(&pr->pr_mtx);
1801 error = ENOENT;
1802 vfs_opterror(opts, "jail %d is dying",
1803 jid);
1804 goto done_unlock_list;
1805 }
1806 goto found_prison;
1807 }
1808 error = ENOENT;
1809 vfs_opterror(opts, "jail %d not found", jid);
1810 goto done_unlock_list;
1811 }
1812 } else if (error != ENOENT)
1813 goto done_unlock_list;
1814
1815 error = vfs_getopt(opts, "name", (void **)&name, &len);
1816 if (error == 0) {
1817 if (len == 0 || name[len - 1] != '\0') {
1818 error = EINVAL;
1819 goto done_unlock_list;
1820 }
1821 pr = prison_find_name(mypr, name);
1822 if (pr != NULL) {
1823 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1824 mtx_unlock(&pr->pr_mtx);
1825 error = ENOENT;
1826 vfs_opterror(opts, "jail \"%s\" is dying",
1827 name);
1828 goto done_unlock_list;
1829 }
1830 goto found_prison;
1831 }
1832 error = ENOENT;
1833 vfs_opterror(opts, "jail \"%s\" not found", name);
1834 goto done_unlock_list;
1835 } else if (error != ENOENT)
1836 goto done_unlock_list;
1837
1838 vfs_opterror(opts, "no jail specified");
1839 error = ENOENT;
1840 goto done_unlock_list;
1841
1842 found_prison:
1843 /* Get the parameters of the prison. */
1844 pr->pr_ref++;
1845 locked = PD_LOCKED;
1846 td->td_retval[0] = pr->pr_id;
1847 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
1848 if (error != 0 && error != ENOENT)
1849 goto done_deref;
1850 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
1851 error = vfs_setopt(opts, "parent", &i, sizeof(i));
1852 if (error != 0 && error != ENOENT)
1853 goto done_deref;
1854 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
1855 if (error != 0 && error != ENOENT)
1856 goto done_deref;
1857 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
1858 sizeof(pr->pr_cpuset->cs_id));
1859 if (error != 0 && error != ENOENT)
1860 goto done_deref;
1861 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
1862 if (error != 0 && error != ENOENT)
1863 goto done_deref;
1864#ifdef INET
1865 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
1866 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1867 if (error != 0 && error != ENOENT)
1868 goto done_deref;
1869#endif
1870#ifdef INET6
1871 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
1872 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1873 if (error != 0 && error != ENOENT)
1874 goto done_deref;
1875#endif
1876 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
1877 sizeof(pr->pr_securelevel));
1878 if (error != 0 && error != ENOENT)
1879 goto done_deref;
1880 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
1881 if (error != 0 && error != ENOENT)
1882 goto done_deref;
1883 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
1884 if (error != 0 && error != ENOENT)
1885 goto done_deref;
1886 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
1887 if (error != 0 && error != ENOENT)
1888 goto done_deref;
1889#ifdef COMPAT_IA32
1890 if (td->td_proc->p_sysent->sv_flags & SV_IA32) {
1891 uint32_t hid32 = pr->pr_hostid;
1892
1893 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
1894 } else
1895#endif
1896 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
1897 sizeof(pr->pr_hostid));
1898 if (error != 0 && error != ENOENT)
1899 goto done_deref;
1900 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
1901 sizeof(pr->pr_enforce_statfs));
1902 if (error != 0 && error != ENOENT)
1903 goto done_deref;
1904 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
1905 fi++) {
1906 if (pr_flag_names[fi] == NULL)
1907 continue;
1908 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
1909 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
1910 if (error != 0 && error != ENOENT)
1911 goto done_deref;
1912 i = !i;
1913 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
1914 if (error != 0 && error != ENOENT)
1915 goto done_deref;
1916 }
1917 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
1918 fi++) {
1919 if (pr_allow_names[fi] == NULL)
1920 continue;
1921 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
1922 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
1923 if (error != 0 && error != ENOENT)
1924 goto done_deref;
1925 i = !i;
1926 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
1927 if (error != 0 && error != ENOENT)
1928 goto done_deref;
1929 }
1930 i = (pr->pr_uref == 0);
1931 error = vfs_setopt(opts, "dying", &i, sizeof(i));
1932 if (error != 0 && error != ENOENT)
1933 goto done_deref;
1934 i = !i;
1935 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
1936 if (error != 0 && error != ENOENT)
1937 goto done_deref;
1938
1939 /* Get the module parameters. */
1940 mtx_unlock(&pr->pr_mtx);
1941 locked = 0;
1942 error = osd_jail_call(pr, PR_METHOD_GET, opts);
1943 if (error)
1944 goto done_deref;
1945 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
1946
1947 /* By now, all parameters should have been noted. */
1948 TAILQ_FOREACH(opt, opts, link) {
1949 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1950 error = EINVAL;
1951 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1952 goto done_errmsg;
1953 }
1954 }
1955
1956 /* Write the fetched parameters back to userspace. */
1957 error = 0;
1958 TAILQ_FOREACH(opt, opts, link) {
1959 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
1960 pos = 2 * opt->pos + 1;
1961 optuio->uio_iov[pos].iov_len = opt->len;
1962 if (opt->value != NULL) {
1963 if (optuio->uio_segflg == UIO_SYSSPACE) {
1964 bcopy(opt->value,
1965 optuio->uio_iov[pos].iov_base,
1966 opt->len);
1967 } else {
1968 error = copyout(opt->value,
1969 optuio->uio_iov[pos].iov_base,
1970 opt->len);
1971 if (error)
1972 break;
1973 }
1974 }
1975 }
1976 }
1977 goto done_errmsg;
1978
1979 done_deref:
1980 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
1981 goto done_errmsg;
1982
1983 done_unlock_list:
1984 sx_sunlock(&allprison_lock);
1985 done_errmsg:
1986 if (error && errmsg_pos >= 0) {
1987 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1988 errmsg_pos = 2 * errmsg_pos + 1;
1989 if (errmsg_len > 0) {
1990 if (optuio->uio_segflg == UIO_SYSSPACE)
1991 bcopy(errmsg,
1992 optuio->uio_iov[errmsg_pos].iov_base,
1993 errmsg_len);
1994 else
1995 copyout(errmsg,
1996 optuio->uio_iov[errmsg_pos].iov_base,
1997 errmsg_len);
1998 }
1999 }
2000 vfs_freeopts(opts);
2001 return (error);
2002}
2003
2004
2005/*
2006 * struct jail_remove_args {
2007 * int jid;
2008 * };
2009 */
2010int
2011jail_remove(struct thread *td, struct jail_remove_args *uap)
2012{
2013 struct prison *pr, *cpr, *lpr, *tpr;
2014 int descend, error;
2015
2016 error = priv_check(td, PRIV_JAIL_REMOVE);
2017 if (error)
2018 return (error);
2019
2020 sx_xlock(&allprison_lock);
2021 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2022 if (pr == NULL) {
2023 sx_xunlock(&allprison_lock);
2024 return (EINVAL);
2025 }
2026
2027 /* Remove all descendants of this prison, then remove this prison. */
2028 pr->pr_ref++;
2029 pr->pr_flags |= PR_REMOVE;
2030 if (!LIST_EMPTY(&pr->pr_children)) {
2031 mtx_unlock(&pr->pr_mtx);
2032 lpr = NULL;
2033 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2034 mtx_lock(&cpr->pr_mtx);
2035 if (cpr->pr_ref > 0) {
2036 tpr = cpr;
2037 cpr->pr_ref++;
2038 cpr->pr_flags |= PR_REMOVE;
2039 } else {
2040 /* Already removed - do not do it again. */
2041 tpr = NULL;
2042 }
2043 mtx_unlock(&cpr->pr_mtx);
2044 if (lpr != NULL) {
2045 mtx_lock(&lpr->pr_mtx);
2046 prison_remove_one(lpr);
2047 sx_xlock(&allprison_lock);
2048 }
2049 lpr = tpr;
2050 }
2051 if (lpr != NULL) {
2052 mtx_lock(&lpr->pr_mtx);
2053 prison_remove_one(lpr);
2054 sx_xlock(&allprison_lock);
2055 }
2056 mtx_lock(&pr->pr_mtx);
2057 }
2058 prison_remove_one(pr);
2059 return (0);
2060}
2061
2062static void
2063prison_remove_one(struct prison *pr)
2064{
2065 struct proc *p;
2066 int deuref;
2067
2068 /* If the prison was persistent, it is not anymore. */
2069 deuref = 0;
2070 if (pr->pr_flags & PR_PERSIST) {
2071 pr->pr_ref--;
2072 deuref = PD_DEUREF;
2073 pr->pr_flags &= ~PR_PERSIST;
2074 }
2075
2076 /*
2077 * jail_remove added a reference. If that's the only one, remove
2078 * the prison now.
2079 */
2080 KASSERT(pr->pr_ref > 0,
2081 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2082 if (pr->pr_ref == 1) {
2083 prison_deref(pr,
2084 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2085 return;
2086 }
2087
2088 mtx_unlock(&pr->pr_mtx);
2089 sx_xunlock(&allprison_lock);
2090 /*
2091 * Kill all processes unfortunate enough to be attached to this prison.
2092 */
2093 sx_slock(&allproc_lock);
2094 LIST_FOREACH(p, &allproc, p_list) {
2095 PROC_LOCK(p);
2096 if (p->p_state != PRS_NEW && p->p_ucred &&
2097 p->p_ucred->cr_prison == pr)
2098 psignal(p, SIGKILL);
2099 PROC_UNLOCK(p);
2100 }
2101 sx_sunlock(&allproc_lock);
2102 /* Remove the temporary reference added by jail_remove. */
2103 prison_deref(pr, deuref | PD_DEREF);
2104}
2105
2106
2107/*
2108 * struct jail_attach_args {
2109 * int jid;
2110 * };
2111 */
2112int
2113jail_attach(struct thread *td, struct jail_attach_args *uap)
2114{
2115 struct prison *pr;
2116 int error;
2117
2118 error = priv_check(td, PRIV_JAIL_ATTACH);
2119 if (error)
2120 return (error);
2121
2122 sx_slock(&allprison_lock);
2123 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2124 if (pr == NULL) {
2125 sx_sunlock(&allprison_lock);
2126 return (EINVAL);
2127 }
2128
2129 /*
2130 * Do not allow a process to attach to a prison that is not
2131 * considered to be "alive".
2132 */
2133 if (pr->pr_uref == 0) {
2134 mtx_unlock(&pr->pr_mtx);
2135 sx_sunlock(&allprison_lock);
2136 return (EINVAL);
2137 }
2138
2139 return (do_jail_attach(td, pr));
2140}
2141
2142static int
2143do_jail_attach(struct thread *td, struct prison *pr)
2144{
2145 struct prison *ppr;
2146 struct proc *p;
2147 struct ucred *newcred, *oldcred;
2148 int vfslocked, error;
2149
2150 /*
2151 * XXX: Note that there is a slight race here if two threads
2152 * in the same privileged process attempt to attach to two
2153 * different jails at the same time. It is important for
2154 * user processes not to do this, or they might end up with
2155 * a process root from one prison, but attached to the jail
2156 * of another.
2157 */
2158 pr->pr_ref++;
2159 pr->pr_uref++;
2160 mtx_unlock(&pr->pr_mtx);
2161
2162 /* Let modules do whatever they need to prepare for attaching. */
2163 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2164 if (error) {
2165 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2166 return (error);
2167 }
2168 sx_sunlock(&allprison_lock);
2169
2170 /*
2171 * Reparent the newly attached process to this jail.
2172 */
2173 ppr = td->td_ucred->cr_prison;
2174 p = td->td_proc;
2175 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2176 if (error)
2177 goto e_revert_osd;
2178
2179 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2180 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2181 if ((error = change_dir(pr->pr_root, td)) != 0)
2182 goto e_unlock;
2183#ifdef MAC
2184 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2185 goto e_unlock;
2186#endif
2187 VOP_UNLOCK(pr->pr_root, 0);
2188 if ((error = change_root(pr->pr_root, td)))
2189 goto e_unlock_giant;
2190 VFS_UNLOCK_GIANT(vfslocked);
2191
2192 newcred = crget();
2193 PROC_LOCK(p);
2194 oldcred = p->p_ucred;
2195 setsugid(p);
2196 crcopy(newcred, oldcred);
2197 newcred->cr_prison = pr;
2198 p->p_ucred = newcred;
2199 PROC_UNLOCK(p);
2200 crfree(oldcred);
2201 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2202 return (0);
2203 e_unlock:
2204 VOP_UNLOCK(pr->pr_root, 0);
2205 e_unlock_giant:
2206 VFS_UNLOCK_GIANT(vfslocked);
2207 e_revert_osd:
2208 /* Tell modules this thread is still in its old jail after all. */
2209 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2210 prison_deref(pr, PD_DEREF | PD_DEUREF);
2211 return (error);
2212}
2213
2214
2215/*
2216 * Returns a locked prison instance, or NULL on failure.
2217 */
2218struct prison *
2219prison_find(int prid)
2220{
2221 struct prison *pr;
2222
2223 sx_assert(&allprison_lock, SX_LOCKED);
2224 TAILQ_FOREACH(pr, &allprison, pr_list) {
2225 if (pr->pr_id == prid) {
2226 mtx_lock(&pr->pr_mtx);
2227 if (pr->pr_ref > 0)
2228 return (pr);
2229 mtx_unlock(&pr->pr_mtx);
2230 }
2231 }
2232 return (NULL);
2233}
2234
2235/*
2236 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2237 */
2238struct prison *
2239prison_find_child(struct prison *mypr, int prid)
2240{
2241 struct prison *pr;
2242 int descend;
2243
2244 sx_assert(&allprison_lock, SX_LOCKED);
2245 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2246 if (pr->pr_id == prid) {
2247 mtx_lock(&pr->pr_mtx);
2248 if (pr->pr_ref > 0)
2249 return (pr);
2250 mtx_unlock(&pr->pr_mtx);
2251 }
2252 }
2253 return (NULL);
2254}
2255
2256/*
2257 * Look for the name relative to mypr. Returns a locked prison or NULL.
2258 */
2259struct prison *
2260prison_find_name(struct prison *mypr, const char *name)
2261{
2262 struct prison *pr, *deadpr;
2263 size_t mylen;
2264 int descend;
2265
2266 sx_assert(&allprison_lock, SX_LOCKED);
2267 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2268 again:
2269 deadpr = NULL;
2270 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2271 if (!strcmp(pr->pr_name + mylen, name)) {
2272 mtx_lock(&pr->pr_mtx);
2273 if (pr->pr_ref > 0) {
2274 if (pr->pr_uref > 0)
2275 return (pr);
2276 deadpr = pr;
2277 }
2278 mtx_unlock(&pr->pr_mtx);
2279 }
2280 }
2281 /* There was no valid prison - perhaps there was a dying one. */
2282 if (deadpr != NULL) {
2283 mtx_lock(&deadpr->pr_mtx);
2284 if (deadpr->pr_ref == 0) {
2285 mtx_unlock(&deadpr->pr_mtx);
2286 goto again;
2287 }
2288 }
2289 return (deadpr);
2290}
2291
2292/*
2293 * See if a prison has the specific flag set.
2294 */
2295int
2296prison_flag(struct ucred *cred, unsigned flag)
2297{
2298
2299 /* This is an atomic read, so no locking is necessary. */
2300 return (cred->cr_prison->pr_flags & flag);
2301}
2302
2303int
2304prison_allow(struct ucred *cred, unsigned flag)
2305{
2306
2307 /* This is an atomic read, so no locking is necessary. */
2308 return (cred->cr_prison->pr_allow & flag);
2309}
2310
2311/*
2312 * Remove a prison reference. If that was the last reference, remove the
2313 * prison itself - but not in this context in case there are locks held.
2314 */
2315void
2316prison_free_locked(struct prison *pr)
2317{
2318
2319 mtx_assert(&pr->pr_mtx, MA_OWNED);
2320 pr->pr_ref--;
2321 if (pr->pr_ref == 0) {
2322 mtx_unlock(&pr->pr_mtx);
2323 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2324 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2325 return;
2326 }
2327 mtx_unlock(&pr->pr_mtx);
2328}
2329
2330void
2331prison_free(struct prison *pr)
2332{
2333
2334 mtx_lock(&pr->pr_mtx);
2335 prison_free_locked(pr);
2336}
2337
2338static void
2339prison_complete(void *context, int pending)
2340{
2341
2342 prison_deref((struct prison *)context, 0);
2343}
2344
2345/*
2346 * Remove a prison reference (usually). This internal version assumes no
2347 * mutexes are held, except perhaps the prison itself. If there are no more
2348 * references, release and delist the prison. On completion, the prison lock
2349 * and the allprison lock are both unlocked.
2350 */
2351static void
2352prison_deref(struct prison *pr, int flags)
2353{
2354 struct prison *ppr, *tpr;
2355 int vfslocked;
2356
2357 if (!(flags & PD_LOCKED))
2358 mtx_lock(&pr->pr_mtx);
2359 /* Decrement the user references in a separate loop. */
2360 if (flags & PD_DEUREF) {
2361 for (tpr = pr;; tpr = tpr->pr_parent) {
2362 if (tpr != pr)
2363 mtx_lock(&tpr->pr_mtx);
2364 if (--tpr->pr_uref > 0)
2365 break;
2366 KASSERT(tpr != &prison0, ("prison0 pr_uref=0"));
2367 mtx_unlock(&tpr->pr_mtx);
2368 }
2369 /* Done if there were only user references to remove. */
2370 if (!(flags & PD_DEREF)) {
2371 mtx_unlock(&tpr->pr_mtx);
2372 if (flags & PD_LIST_SLOCKED)
2373 sx_sunlock(&allprison_lock);
2374 else if (flags & PD_LIST_XLOCKED)
2375 sx_xunlock(&allprison_lock);
2376 return;
2377 }
2378 if (tpr != pr) {
2379 mtx_unlock(&tpr->pr_mtx);
2380 mtx_lock(&pr->pr_mtx);
2381 }
2382 }
2383
2384 for (;;) {
2385 if (flags & PD_DEREF)
2386 pr->pr_ref--;
2387 /* If the prison still has references, nothing else to do. */
2388 if (pr->pr_ref > 0) {
2389 mtx_unlock(&pr->pr_mtx);
2390 if (flags & PD_LIST_SLOCKED)
2391 sx_sunlock(&allprison_lock);
2392 else if (flags & PD_LIST_XLOCKED)
2393 sx_xunlock(&allprison_lock);
2394 return;
2395 }
2396
2397 mtx_unlock(&pr->pr_mtx);
2398 if (flags & PD_LIST_SLOCKED) {
2399 if (!sx_try_upgrade(&allprison_lock)) {
2400 sx_sunlock(&allprison_lock);
2401 sx_xlock(&allprison_lock);
2402 }
2403 } else if (!(flags & PD_LIST_XLOCKED))
2404 sx_xlock(&allprison_lock);
2405
2406 TAILQ_REMOVE(&allprison, pr, pr_list);
2407 LIST_REMOVE(pr, pr_sibling);
2408 ppr = pr->pr_parent;
2409 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2410 tpr->pr_prisoncount--;
2411 sx_downgrade(&allprison_lock);
2412
| 1134 /*
1135 * Allocate a dedicated cpuset for each jail.
1136 * Unlike other initial settings, this may return an erorr.
1137 */
1138 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1139 if (error) {
1140 prison_deref(pr, PD_LIST_XLOCKED);
1141 goto done_releroot;
1142 }
1143
1144 mtx_lock(&pr->pr_mtx);
1145 /*
1146 * New prisons do not yet have a reference, because we do not
1147 * want other to see the incomplete prison once the
1148 * allprison_lock is downgraded.
1149 */
1150 } else {
1151 created = 0;
1152 /*
1153 * Grab a reference for existing prisons, to ensure they
1154 * continue to exist for the duration of the call.
1155 */
1156 pr->pr_ref++;
1157 }
1158
1159 /* Do final error checking before setting anything. */
1160 if (gotslevel) {
1161 if (slevel < ppr->pr_securelevel) {
1162 error = EPERM;
1163 goto done_deref_locked;
1164 }
1165 }
1166 if (gotenforce) {
1167 if (enforce < ppr->pr_enforce_statfs) {
1168 error = EPERM;
1169 goto done_deref_locked;
1170 }
1171 }
1172#ifdef INET
1173 if (ch_flags & PR_IP4_USER) {
1174 if (ppr->pr_flags & PR_IP4) {
1175 if (!(pr_flags & PR_IP4_USER)) {
1176 /*
1177 * Silently ignore attempts to make the IP
1178 * addresses unrestricted when the parent is
1179 * restricted; in other words, interpret
1180 * "unrestricted" as "as unrestricted as
1181 * possible".
1182 */
1183 ip4s = ppr->pr_ip4s;
1184 if (ip4s == 0) {
1185 free(ip4, M_PRISON);
1186 ip4 = NULL;
1187 } else if (ip4s <= ip4a) {
1188 /* Inherit the parent's address(es). */
1189 bcopy(ppr->pr_ip4, ip4,
1190 ip4s * sizeof(*ip4));
1191 } else {
1192 /*
1193 * There's no room for the parent's
1194 * address list. Allocate some more.
1195 */
1196 ip4a = ip4s;
1197 free(ip4, M_PRISON);
1198 ip4 = malloc(ip4a * sizeof(*ip4),
1199 M_PRISON, M_NOWAIT);
1200 if (ip4 != NULL)
1201 bcopy(ppr->pr_ip4, ip4,
1202 ip4s * sizeof(*ip4));
1203 else {
1204 /* Allocation failed without
1205 * sleeping. Unlocking the
1206 * prison now will invalidate
1207 * some checks and prematurely
1208 * show an unfinished new jail.
1209 * So let go of everything and
1210 * start over.
1211 */
1212 prison_deref(pr, created
1213 ? PD_LOCKED |
1214 PD_LIST_XLOCKED
1215 : PD_DEREF | PD_LOCKED |
1216 PD_LIST_XLOCKED);
1217 if (root != NULL) {
1218 vfslocked =
1219 VFS_LOCK_GIANT(
1220 root->v_mount);
1221 vrele(root);
1222 VFS_UNLOCK_GIANT(
1223 vfslocked);
1224 }
1225 ip4 = malloc(ip4a *
1226 sizeof(*ip4), M_PRISON,
1227 M_WAITOK);
1228 goto again;
1229 }
1230 }
1231 } else if (ip4s > 0) {
1232 /*
1233 * Make sure the new set of IP addresses is a
1234 * subset of the parent's list. Don't worry
1235 * about the parent being unlocked, as any
1236 * setting is done with allprison_lock held.
1237 */
1238 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1239 if (ip4[0].s_addr ==
1240 ppr->pr_ip4[ij].s_addr)
1241 break;
1242 if (ij == ppr->pr_ip4s) {
1243 error = EPERM;
1244 goto done_deref_locked;
1245 }
1246 if (ip4s > 1) {
1247 for (ii = ij = 1; ii < ip4s; ii++) {
1248 if (ip4[ii].s_addr ==
1249 ppr->pr_ip4[0].s_addr)
1250 continue;
1251 for (; ij < ppr->pr_ip4s; ij++)
1252 if (ip4[ii].s_addr ==
1253 ppr->pr_ip4[ij].s_addr)
1254 break;
1255 if (ij == ppr->pr_ip4s)
1256 break;
1257 }
1258 if (ij == ppr->pr_ip4s) {
1259 error = EPERM;
1260 goto done_deref_locked;
1261 }
1262 }
1263 }
1264 }
1265 if (ip4s > 0) {
1266 /*
1267 * Check for conflicting IP addresses. We permit them
1268 * if there is no more than one IP on each jail. If
1269 * there is a duplicate on a jail with more than one
1270 * IP stop checking and return error.
1271 */
1272 FOREACH_PRISON_DESCENDANT(&prison0, tpr, descend) {
1273 if (tpr == pr || tpr->pr_uref == 0) {
1274 descend = 0;
1275 continue;
1276 }
1277 if (!(tpr->pr_flags & PR_IP4_USER))
1278 continue;
1279 descend = 0;
1280 if (tpr->pr_ip4 == NULL ||
1281 (ip4s == 1 && tpr->pr_ip4s == 1))
1282 continue;
1283 for (ii = 0; ii < ip4s; ii++) {
1284 if (_prison_check_ip4(tpr,
1285 &ip4[ii]) == 0) {
1286 error = EADDRINUSE;
1287 vfs_opterror(opts,
1288 "IPv4 addresses clash");
1289 goto done_deref_locked;
1290 }
1291 }
1292 }
1293 }
1294 }
1295#endif
1296#ifdef INET6
1297 if (ch_flags & PR_IP6_USER) {
1298 if (ppr->pr_flags & PR_IP6) {
1299 if (!(pr_flags & PR_IP6_USER)) {
1300 /*
1301 * Silently ignore attempts to make the IP
1302 * addresses unrestricted when the parent is
1303 * restricted.
1304 */
1305 ip6s = ppr->pr_ip6s;
1306 if (ip6s == 0) {
1307 free(ip6, M_PRISON);
1308 ip6 = NULL;
1309 } else if (ip6s <= ip6a) {
1310 /* Inherit the parent's address(es). */
1311 bcopy(ppr->pr_ip6, ip6,
1312 ip6s * sizeof(*ip6));
1313 } else {
1314 /*
1315 * There's no room for the parent's
1316 * address list.
1317 */
1318 ip6a = ip6s;
1319 free(ip6, M_PRISON);
1320 ip6 = malloc(ip6a * sizeof(*ip6),
1321 M_PRISON, M_NOWAIT);
1322 if (ip6 != NULL)
1323 bcopy(ppr->pr_ip6, ip6,
1324 ip6s * sizeof(*ip6));
1325 else {
1326 prison_deref(pr, created
1327 ? PD_LOCKED |
1328 PD_LIST_XLOCKED
1329 : PD_DEREF | PD_LOCKED |
1330 PD_LIST_XLOCKED);
1331 if (root != NULL) {
1332 vfslocked =
1333 VFS_LOCK_GIANT(
1334 root->v_mount);
1335 vrele(root);
1336 VFS_UNLOCK_GIANT(
1337 vfslocked);
1338 }
1339 ip6 = malloc(ip6a *
1340 sizeof(*ip6), M_PRISON,
1341 M_WAITOK);
1342 goto again;
1343 }
1344 }
1345 } else if (ip6s > 0) {
1346 /*
1347 * Make sure the new set of IP addresses is a
1348 * subset of the parent's list.
1349 */
1350 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1351 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1352 &ppr->pr_ip6[ij]))
1353 break;
1354 if (ij == ppr->pr_ip6s) {
1355 error = EPERM;
1356 goto done_deref_locked;
1357 }
1358 if (ip6s > 1) {
1359 for (ii = ij = 1; ii < ip6s; ii++) {
1360 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1361 &ppr->pr_ip6[0]))
1362 continue;
1363 for (; ij < ppr->pr_ip6s; ij++)
1364 if (IN6_ARE_ADDR_EQUAL(
1365 &ip6[ii],
1366 &ppr->pr_ip6[ij]))
1367 break;
1368 if (ij == ppr->pr_ip6s)
1369 break;
1370 }
1371 if (ij == ppr->pr_ip6s) {
1372 error = EPERM;
1373 goto done_deref_locked;
1374 }
1375 }
1376 }
1377 }
1378 if (ip6s > 0) {
1379 /* Check for conflicting IP addresses. */
1380 FOREACH_PRISON_DESCENDANT(&prison0, tpr, descend) {
1381 if (tpr == pr || tpr->pr_uref == 0) {
1382 descend = 0;
1383 continue;
1384 }
1385 if (!(tpr->pr_flags & PR_IP6_USER))
1386 continue;
1387 descend = 0;
1388 if (tpr->pr_ip6 == NULL ||
1389 (ip6s == 1 && tpr->pr_ip6s == 1))
1390 continue;
1391 for (ii = 0; ii < ip6s; ii++) {
1392 if (_prison_check_ip6(tpr,
1393 &ip6[ii]) == 0) {
1394 error = EADDRINUSE;
1395 vfs_opterror(opts,
1396 "IPv6 addresses clash");
1397 goto done_deref_locked;
1398 }
1399 }
1400 }
1401 }
1402 }
1403#endif
1404 onamelen = namelen = 0;
1405 if (name != NULL) {
1406 /* Give a default name of the jid. */
1407 if (name[0] == '\0')
1408 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1409 else if (strtoul(name, &p, 10) != jid && *p == '\0') {
1410 error = EINVAL;
1411 vfs_opterror(opts, "name cannot be numeric");
1412 goto done_deref_locked;
1413 }
1414 /*
1415 * Make sure the name isn't too long for the prison or its
1416 * children.
1417 */
1418 onamelen = strlen(pr->pr_name);
1419 namelen = strlen(name);
1420 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1421 error = ENAMETOOLONG;
1422 goto done_deref_locked;
1423 }
1424 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1425 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1426 sizeof(pr->pr_name)) {
1427 error = ENAMETOOLONG;
1428 goto done_deref_locked;
1429 }
1430 }
1431 }
1432 if (pr_allow & ~ppr->pr_allow) {
1433 error = EPERM;
1434 goto done_deref_locked;
1435 }
1436
1437 /* Set the parameters of the prison. */
1438#ifdef INET
1439 redo_ip4 = 0;
1440 if (ch_flags & PR_IP4_USER) {
1441 if (pr_flags & PR_IP4_USER) {
1442 /* Some restriction set. */
1443 pr->pr_flags |= PR_IP4;
1444 if (ip4s >= 0) {
1445 free(pr->pr_ip4, M_PRISON);
1446 pr->pr_ip4s = ip4s;
1447 pr->pr_ip4 = ip4;
1448 ip4 = NULL;
1449 }
1450 } else if (ppr->pr_flags & PR_IP4) {
1451 /* This restriction cleared, but keep inherited. */
1452 free(pr->pr_ip4, M_PRISON);
1453 pr->pr_ip4s = ip4s;
1454 pr->pr_ip4 = ip4;
1455 ip4 = NULL;
1456 } else {
1457 /* Restriction cleared, now unrestricted. */
1458 pr->pr_flags &= ~PR_IP4;
1459 free(pr->pr_ip4, M_PRISON);
1460 pr->pr_ip4s = 0;
1461 }
1462 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1463 if (prison_restrict_ip4(tpr, NULL)) {
1464 redo_ip4 = 1;
1465 descend = 0;
1466 }
1467 }
1468 }
1469#endif
1470#ifdef INET6
1471 redo_ip6 = 0;
1472 if (ch_flags & PR_IP6_USER) {
1473 if (pr_flags & PR_IP6_USER) {
1474 /* Some restriction set. */
1475 pr->pr_flags |= PR_IP6;
1476 if (ip6s >= 0) {
1477 free(pr->pr_ip6, M_PRISON);
1478 pr->pr_ip6s = ip6s;
1479 pr->pr_ip6 = ip6;
1480 ip6 = NULL;
1481 }
1482 } else if (ppr->pr_flags & PR_IP6) {
1483 /* This restriction cleared, but keep inherited. */
1484 free(pr->pr_ip6, M_PRISON);
1485 pr->pr_ip6s = ip6s;
1486 pr->pr_ip6 = ip6;
1487 ip6 = NULL;
1488 } else {
1489 /* Restriction cleared, now unrestricted. */
1490 pr->pr_flags &= ~PR_IP6;
1491 free(pr->pr_ip6, M_PRISON);
1492 pr->pr_ip6s = 0;
1493 }
1494 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1495 if (prison_restrict_ip6(tpr, NULL)) {
1496 redo_ip6 = 1;
1497 descend = 0;
1498 }
1499 }
1500 }
1501#endif
1502 if (gotslevel) {
1503 pr->pr_securelevel = slevel;
1504 /* Set all child jails to be at least this level. */
1505 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1506 if (tpr->pr_securelevel < slevel)
1507 tpr->pr_securelevel = slevel;
1508 }
1509 if (gotenforce) {
1510 pr->pr_enforce_statfs = enforce;
1511 /* Pass this restriction on to the children. */
1512 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1513 if (tpr->pr_enforce_statfs < enforce)
1514 tpr->pr_enforce_statfs = enforce;
1515 }
1516 if (name != NULL) {
1517 if (ppr == &prison0)
1518 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1519 else
1520 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1521 ppr->pr_name, name);
1522 /* Change this component of child names. */
1523 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1524 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1525 strlen(tpr->pr_name + onamelen) + 1);
1526 bcopy(pr->pr_name, tpr->pr_name, namelen);
1527 }
1528 }
1529 if (path != NULL) {
1530 /* Try to keep a real-rooted full pathname. */
1531 if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
1532 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1533 mypr->pr_path, path);
1534 else
1535 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1536 pr->pr_root = root;
1537 }
1538 if (PR_HOST & ch_flags & ~pr_flags) {
1539 if (pr->pr_flags & PR_HOST) {
1540 /*
1541 * Copy the parent's host info. As with pr_ip4 above,
1542 * the lack of a lock on the parent is not a problem;
1543 * it is always set with allprison_lock at least
1544 * shared, and is held exclusively here.
1545 */
1546 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1547 sizeof(pr->pr_hostname));
1548 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1549 sizeof(pr->pr_domainname));
1550 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1551 sizeof(pr->pr_hostuuid));
1552 pr->pr_hostid = pr->pr_parent->pr_hostid;
1553 }
1554 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1555 /* Set this prison, and any descendants without PR_HOST. */
1556 if (host != NULL)
1557 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1558 if (domain != NULL)
1559 strlcpy(pr->pr_domainname, domain,
1560 sizeof(pr->pr_domainname));
1561 if (uuid != NULL)
1562 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1563 if (gothid)
1564 pr->pr_hostid = hid;
1565 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1566 if (tpr->pr_flags & PR_HOST)
1567 descend = 0;
1568 else {
1569 if (host != NULL)
1570 strlcpy(tpr->pr_hostname,
1571 pr->pr_hostname,
1572 sizeof(tpr->pr_hostname));
1573 if (domain != NULL)
1574 strlcpy(tpr->pr_domainname,
1575 pr->pr_domainname,
1576 sizeof(tpr->pr_domainname));
1577 if (uuid != NULL)
1578 strlcpy(tpr->pr_hostuuid,
1579 pr->pr_hostuuid,
1580 sizeof(tpr->pr_hostuuid));
1581 if (gothid)
1582 tpr->pr_hostid = hid;
1583 }
1584 }
1585 }
1586 if ((tallow = ch_allow & ~pr_allow)) {
1587 /* Clear allow bits in all children. */
1588 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1589 tpr->pr_allow &= ~tallow;
1590 }
1591 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1592 /*
1593 * Persistent prisons get an extra reference, and prisons losing their
1594 * persist flag lose that reference. Only do this for existing prisons
1595 * for now, so new ones will remain unseen until after the module
1596 * handlers have completed.
1597 */
1598 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1599 if (pr_flags & PR_PERSIST) {
1600 pr->pr_ref++;
1601 pr->pr_uref++;
1602 } else {
1603 pr->pr_ref--;
1604 pr->pr_uref--;
1605 }
1606 }
1607 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1608 mtx_unlock(&pr->pr_mtx);
1609
1610 /* Locks may have prevented a complete restriction of child IP
1611 * addresses. If so, allocate some more memory and try again.
1612 */
1613#ifdef INET
1614 while (redo_ip4) {
1615 ip4s = pr->pr_ip4s;
1616 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1617 mtx_lock(&pr->pr_mtx);
1618 redo_ip4 = 0;
1619 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1620 if (prison_restrict_ip4(tpr, ip4)) {
1621 if (ip4 != NULL)
1622 ip4 = NULL;
1623 else
1624 redo_ip4 = 1;
1625 }
1626 }
1627 mtx_unlock(&pr->pr_mtx);
1628 }
1629#endif
1630#ifdef INET6
1631 while (redo_ip6) {
1632 ip6s = pr->pr_ip6s;
1633 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1634 mtx_lock(&pr->pr_mtx);
1635 redo_ip6 = 0;
1636 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1637 if (prison_restrict_ip6(tpr, ip6)) {
1638 if (ip6 != NULL)
1639 ip6 = NULL;
1640 else
1641 redo_ip6 = 1;
1642 }
1643 }
1644 mtx_unlock(&pr->pr_mtx);
1645 }
1646#endif
1647
1648 /* Let the modules do their work. */
1649 sx_downgrade(&allprison_lock);
1650 if (created) {
1651 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1652 if (error) {
1653 prison_deref(pr, PD_LIST_SLOCKED);
1654 goto done_errmsg;
1655 }
1656 }
1657 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1658 if (error) {
1659 prison_deref(pr, created
1660 ? PD_LIST_SLOCKED
1661 : PD_DEREF | PD_LIST_SLOCKED);
1662 goto done_errmsg;
1663 }
1664
1665 /* Attach this process to the prison if requested. */
1666 if (flags & JAIL_ATTACH) {
1667 mtx_lock(&pr->pr_mtx);
1668 error = do_jail_attach(td, pr);
1669 if (error) {
1670 vfs_opterror(opts, "attach failed");
1671 if (!created)
1672 prison_deref(pr, PD_DEREF);
1673 goto done_errmsg;
1674 }
1675 }
1676
1677 /*
1678 * Now that it is all there, drop the temporary reference from existing
1679 * prisons. Or add a reference to newly created persistent prisons
1680 * (which was not done earlier so that the prison would not be publicly
1681 * visible).
1682 */
1683 if (!created) {
1684 prison_deref(pr, (flags & JAIL_ATTACH)
1685 ? PD_DEREF
1686 : PD_DEREF | PD_LIST_SLOCKED);
1687 } else {
1688 if (pr_flags & PR_PERSIST) {
1689 mtx_lock(&pr->pr_mtx);
1690 pr->pr_ref++;
1691 pr->pr_uref++;
1692 mtx_unlock(&pr->pr_mtx);
1693 }
1694 if (!(flags & JAIL_ATTACH))
1695 sx_sunlock(&allprison_lock);
1696 }
1697 td->td_retval[0] = pr->pr_id;
1698 goto done_errmsg;
1699
1700 done_deref_locked:
1701 prison_deref(pr, created
1702 ? PD_LOCKED | PD_LIST_XLOCKED
1703 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1704 goto done_releroot;
1705 done_unlock_list:
1706 sx_xunlock(&allprison_lock);
1707 done_releroot:
1708 if (root != NULL) {
1709 vfslocked = VFS_LOCK_GIANT(root->v_mount);
1710 vrele(root);
1711 VFS_UNLOCK_GIANT(vfslocked);
1712 }
1713 done_errmsg:
1714 if (error) {
1715 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1716 if (errmsg_len > 0) {
1717 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1718 if (errmsg_pos > 0) {
1719 if (optuio->uio_segflg == UIO_SYSSPACE)
1720 bcopy(errmsg,
1721 optuio->uio_iov[errmsg_pos].iov_base,
1722 errmsg_len);
1723 else
1724 copyout(errmsg,
1725 optuio->uio_iov[errmsg_pos].iov_base,
1726 errmsg_len);
1727 }
1728 }
1729 }
1730 done_free:
1731#ifdef INET
1732 free(ip4, M_PRISON);
1733#endif
1734#ifdef INET6
1735 free(ip6, M_PRISON);
1736#endif
1737 vfs_freeopts(opts);
1738 return (error);
1739}
1740
1741
1742/*
1743 * struct jail_get_args {
1744 * struct iovec *iovp;
1745 * unsigned int iovcnt;
1746 * int flags;
1747 * };
1748 */
1749int
1750jail_get(struct thread *td, struct jail_get_args *uap)
1751{
1752 struct uio *auio;
1753 int error;
1754
1755 /* Check that we have an even number of iovecs. */
1756 if (uap->iovcnt & 1)
1757 return (EINVAL);
1758
1759 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1760 if (error)
1761 return (error);
1762 error = kern_jail_get(td, auio, uap->flags);
1763 if (error == 0)
1764 error = copyout(auio->uio_iov, uap->iovp,
1765 uap->iovcnt * sizeof (struct iovec));
1766 free(auio, M_IOV);
1767 return (error);
1768}
1769
1770int
1771kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1772{
1773 struct prison *pr, *mypr;
1774 struct vfsopt *opt;
1775 struct vfsoptlist *opts;
1776 char *errmsg, *name;
1777 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1778
1779 if (flags & ~JAIL_GET_MASK)
1780 return (EINVAL);
1781
1782 /* Get the parameter list. */
1783 error = vfs_buildopts(optuio, &opts);
1784 if (error)
1785 return (error);
1786 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1787 mypr = td->td_ucred->cr_prison;
1788
1789 /*
1790 * Find the prison specified by one of: lastjid, jid, name.
1791 */
1792 sx_slock(&allprison_lock);
1793 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1794 if (error == 0) {
1795 TAILQ_FOREACH(pr, &allprison, pr_list) {
1796 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1797 mtx_lock(&pr->pr_mtx);
1798 if (pr->pr_ref > 0 &&
1799 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1800 break;
1801 mtx_unlock(&pr->pr_mtx);
1802 }
1803 }
1804 if (pr != NULL)
1805 goto found_prison;
1806 error = ENOENT;
1807 vfs_opterror(opts, "no jail after %d", jid);
1808 goto done_unlock_list;
1809 } else if (error != ENOENT)
1810 goto done_unlock_list;
1811
1812 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1813 if (error == 0) {
1814 if (jid != 0) {
1815 pr = prison_find_child(mypr, jid);
1816 if (pr != NULL) {
1817 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1818 mtx_unlock(&pr->pr_mtx);
1819 error = ENOENT;
1820 vfs_opterror(opts, "jail %d is dying",
1821 jid);
1822 goto done_unlock_list;
1823 }
1824 goto found_prison;
1825 }
1826 error = ENOENT;
1827 vfs_opterror(opts, "jail %d not found", jid);
1828 goto done_unlock_list;
1829 }
1830 } else if (error != ENOENT)
1831 goto done_unlock_list;
1832
1833 error = vfs_getopt(opts, "name", (void **)&name, &len);
1834 if (error == 0) {
1835 if (len == 0 || name[len - 1] != '\0') {
1836 error = EINVAL;
1837 goto done_unlock_list;
1838 }
1839 pr = prison_find_name(mypr, name);
1840 if (pr != NULL) {
1841 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1842 mtx_unlock(&pr->pr_mtx);
1843 error = ENOENT;
1844 vfs_opterror(opts, "jail \"%s\" is dying",
1845 name);
1846 goto done_unlock_list;
1847 }
1848 goto found_prison;
1849 }
1850 error = ENOENT;
1851 vfs_opterror(opts, "jail \"%s\" not found", name);
1852 goto done_unlock_list;
1853 } else if (error != ENOENT)
1854 goto done_unlock_list;
1855
1856 vfs_opterror(opts, "no jail specified");
1857 error = ENOENT;
1858 goto done_unlock_list;
1859
1860 found_prison:
1861 /* Get the parameters of the prison. */
1862 pr->pr_ref++;
1863 locked = PD_LOCKED;
1864 td->td_retval[0] = pr->pr_id;
1865 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
1866 if (error != 0 && error != ENOENT)
1867 goto done_deref;
1868 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
1869 error = vfs_setopt(opts, "parent", &i, sizeof(i));
1870 if (error != 0 && error != ENOENT)
1871 goto done_deref;
1872 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
1873 if (error != 0 && error != ENOENT)
1874 goto done_deref;
1875 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
1876 sizeof(pr->pr_cpuset->cs_id));
1877 if (error != 0 && error != ENOENT)
1878 goto done_deref;
1879 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
1880 if (error != 0 && error != ENOENT)
1881 goto done_deref;
1882#ifdef INET
1883 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
1884 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1885 if (error != 0 && error != ENOENT)
1886 goto done_deref;
1887#endif
1888#ifdef INET6
1889 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
1890 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1891 if (error != 0 && error != ENOENT)
1892 goto done_deref;
1893#endif
1894 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
1895 sizeof(pr->pr_securelevel));
1896 if (error != 0 && error != ENOENT)
1897 goto done_deref;
1898 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
1899 if (error != 0 && error != ENOENT)
1900 goto done_deref;
1901 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
1902 if (error != 0 && error != ENOENT)
1903 goto done_deref;
1904 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
1905 if (error != 0 && error != ENOENT)
1906 goto done_deref;
1907#ifdef COMPAT_IA32
1908 if (td->td_proc->p_sysent->sv_flags & SV_IA32) {
1909 uint32_t hid32 = pr->pr_hostid;
1910
1911 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
1912 } else
1913#endif
1914 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
1915 sizeof(pr->pr_hostid));
1916 if (error != 0 && error != ENOENT)
1917 goto done_deref;
1918 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
1919 sizeof(pr->pr_enforce_statfs));
1920 if (error != 0 && error != ENOENT)
1921 goto done_deref;
1922 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
1923 fi++) {
1924 if (pr_flag_names[fi] == NULL)
1925 continue;
1926 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
1927 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
1928 if (error != 0 && error != ENOENT)
1929 goto done_deref;
1930 i = !i;
1931 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
1932 if (error != 0 && error != ENOENT)
1933 goto done_deref;
1934 }
1935 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
1936 fi++) {
1937 if (pr_allow_names[fi] == NULL)
1938 continue;
1939 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
1940 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
1941 if (error != 0 && error != ENOENT)
1942 goto done_deref;
1943 i = !i;
1944 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
1945 if (error != 0 && error != ENOENT)
1946 goto done_deref;
1947 }
1948 i = (pr->pr_uref == 0);
1949 error = vfs_setopt(opts, "dying", &i, sizeof(i));
1950 if (error != 0 && error != ENOENT)
1951 goto done_deref;
1952 i = !i;
1953 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
1954 if (error != 0 && error != ENOENT)
1955 goto done_deref;
1956
1957 /* Get the module parameters. */
1958 mtx_unlock(&pr->pr_mtx);
1959 locked = 0;
1960 error = osd_jail_call(pr, PR_METHOD_GET, opts);
1961 if (error)
1962 goto done_deref;
1963 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
1964
1965 /* By now, all parameters should have been noted. */
1966 TAILQ_FOREACH(opt, opts, link) {
1967 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1968 error = EINVAL;
1969 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1970 goto done_errmsg;
1971 }
1972 }
1973
1974 /* Write the fetched parameters back to userspace. */
1975 error = 0;
1976 TAILQ_FOREACH(opt, opts, link) {
1977 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
1978 pos = 2 * opt->pos + 1;
1979 optuio->uio_iov[pos].iov_len = opt->len;
1980 if (opt->value != NULL) {
1981 if (optuio->uio_segflg == UIO_SYSSPACE) {
1982 bcopy(opt->value,
1983 optuio->uio_iov[pos].iov_base,
1984 opt->len);
1985 } else {
1986 error = copyout(opt->value,
1987 optuio->uio_iov[pos].iov_base,
1988 opt->len);
1989 if (error)
1990 break;
1991 }
1992 }
1993 }
1994 }
1995 goto done_errmsg;
1996
1997 done_deref:
1998 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
1999 goto done_errmsg;
2000
2001 done_unlock_list:
2002 sx_sunlock(&allprison_lock);
2003 done_errmsg:
2004 if (error && errmsg_pos >= 0) {
2005 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2006 errmsg_pos = 2 * errmsg_pos + 1;
2007 if (errmsg_len > 0) {
2008 if (optuio->uio_segflg == UIO_SYSSPACE)
2009 bcopy(errmsg,
2010 optuio->uio_iov[errmsg_pos].iov_base,
2011 errmsg_len);
2012 else
2013 copyout(errmsg,
2014 optuio->uio_iov[errmsg_pos].iov_base,
2015 errmsg_len);
2016 }
2017 }
2018 vfs_freeopts(opts);
2019 return (error);
2020}
2021
2022
2023/*
2024 * struct jail_remove_args {
2025 * int jid;
2026 * };
2027 */
2028int
2029jail_remove(struct thread *td, struct jail_remove_args *uap)
2030{
2031 struct prison *pr, *cpr, *lpr, *tpr;
2032 int descend, error;
2033
2034 error = priv_check(td, PRIV_JAIL_REMOVE);
2035 if (error)
2036 return (error);
2037
2038 sx_xlock(&allprison_lock);
2039 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2040 if (pr == NULL) {
2041 sx_xunlock(&allprison_lock);
2042 return (EINVAL);
2043 }
2044
2045 /* Remove all descendants of this prison, then remove this prison. */
2046 pr->pr_ref++;
2047 pr->pr_flags |= PR_REMOVE;
2048 if (!LIST_EMPTY(&pr->pr_children)) {
2049 mtx_unlock(&pr->pr_mtx);
2050 lpr = NULL;
2051 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2052 mtx_lock(&cpr->pr_mtx);
2053 if (cpr->pr_ref > 0) {
2054 tpr = cpr;
2055 cpr->pr_ref++;
2056 cpr->pr_flags |= PR_REMOVE;
2057 } else {
2058 /* Already removed - do not do it again. */
2059 tpr = NULL;
2060 }
2061 mtx_unlock(&cpr->pr_mtx);
2062 if (lpr != NULL) {
2063 mtx_lock(&lpr->pr_mtx);
2064 prison_remove_one(lpr);
2065 sx_xlock(&allprison_lock);
2066 }
2067 lpr = tpr;
2068 }
2069 if (lpr != NULL) {
2070 mtx_lock(&lpr->pr_mtx);
2071 prison_remove_one(lpr);
2072 sx_xlock(&allprison_lock);
2073 }
2074 mtx_lock(&pr->pr_mtx);
2075 }
2076 prison_remove_one(pr);
2077 return (0);
2078}
2079
2080static void
2081prison_remove_one(struct prison *pr)
2082{
2083 struct proc *p;
2084 int deuref;
2085
2086 /* If the prison was persistent, it is not anymore. */
2087 deuref = 0;
2088 if (pr->pr_flags & PR_PERSIST) {
2089 pr->pr_ref--;
2090 deuref = PD_DEUREF;
2091 pr->pr_flags &= ~PR_PERSIST;
2092 }
2093
2094 /*
2095 * jail_remove added a reference. If that's the only one, remove
2096 * the prison now.
2097 */
2098 KASSERT(pr->pr_ref > 0,
2099 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2100 if (pr->pr_ref == 1) {
2101 prison_deref(pr,
2102 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2103 return;
2104 }
2105
2106 mtx_unlock(&pr->pr_mtx);
2107 sx_xunlock(&allprison_lock);
2108 /*
2109 * Kill all processes unfortunate enough to be attached to this prison.
2110 */
2111 sx_slock(&allproc_lock);
2112 LIST_FOREACH(p, &allproc, p_list) {
2113 PROC_LOCK(p);
2114 if (p->p_state != PRS_NEW && p->p_ucred &&
2115 p->p_ucred->cr_prison == pr)
2116 psignal(p, SIGKILL);
2117 PROC_UNLOCK(p);
2118 }
2119 sx_sunlock(&allproc_lock);
2120 /* Remove the temporary reference added by jail_remove. */
2121 prison_deref(pr, deuref | PD_DEREF);
2122}
2123
2124
2125/*
2126 * struct jail_attach_args {
2127 * int jid;
2128 * };
2129 */
2130int
2131jail_attach(struct thread *td, struct jail_attach_args *uap)
2132{
2133 struct prison *pr;
2134 int error;
2135
2136 error = priv_check(td, PRIV_JAIL_ATTACH);
2137 if (error)
2138 return (error);
2139
2140 sx_slock(&allprison_lock);
2141 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2142 if (pr == NULL) {
2143 sx_sunlock(&allprison_lock);
2144 return (EINVAL);
2145 }
2146
2147 /*
2148 * Do not allow a process to attach to a prison that is not
2149 * considered to be "alive".
2150 */
2151 if (pr->pr_uref == 0) {
2152 mtx_unlock(&pr->pr_mtx);
2153 sx_sunlock(&allprison_lock);
2154 return (EINVAL);
2155 }
2156
2157 return (do_jail_attach(td, pr));
2158}
2159
2160static int
2161do_jail_attach(struct thread *td, struct prison *pr)
2162{
2163 struct prison *ppr;
2164 struct proc *p;
2165 struct ucred *newcred, *oldcred;
2166 int vfslocked, error;
2167
2168 /*
2169 * XXX: Note that there is a slight race here if two threads
2170 * in the same privileged process attempt to attach to two
2171 * different jails at the same time. It is important for
2172 * user processes not to do this, or they might end up with
2173 * a process root from one prison, but attached to the jail
2174 * of another.
2175 */
2176 pr->pr_ref++;
2177 pr->pr_uref++;
2178 mtx_unlock(&pr->pr_mtx);
2179
2180 /* Let modules do whatever they need to prepare for attaching. */
2181 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2182 if (error) {
2183 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2184 return (error);
2185 }
2186 sx_sunlock(&allprison_lock);
2187
2188 /*
2189 * Reparent the newly attached process to this jail.
2190 */
2191 ppr = td->td_ucred->cr_prison;
2192 p = td->td_proc;
2193 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2194 if (error)
2195 goto e_revert_osd;
2196
2197 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2198 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2199 if ((error = change_dir(pr->pr_root, td)) != 0)
2200 goto e_unlock;
2201#ifdef MAC
2202 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2203 goto e_unlock;
2204#endif
2205 VOP_UNLOCK(pr->pr_root, 0);
2206 if ((error = change_root(pr->pr_root, td)))
2207 goto e_unlock_giant;
2208 VFS_UNLOCK_GIANT(vfslocked);
2209
2210 newcred = crget();
2211 PROC_LOCK(p);
2212 oldcred = p->p_ucred;
2213 setsugid(p);
2214 crcopy(newcred, oldcred);
2215 newcred->cr_prison = pr;
2216 p->p_ucred = newcred;
2217 PROC_UNLOCK(p);
2218 crfree(oldcred);
2219 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2220 return (0);
2221 e_unlock:
2222 VOP_UNLOCK(pr->pr_root, 0);
2223 e_unlock_giant:
2224 VFS_UNLOCK_GIANT(vfslocked);
2225 e_revert_osd:
2226 /* Tell modules this thread is still in its old jail after all. */
2227 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2228 prison_deref(pr, PD_DEREF | PD_DEUREF);
2229 return (error);
2230}
2231
2232
2233/*
2234 * Returns a locked prison instance, or NULL on failure.
2235 */
2236struct prison *
2237prison_find(int prid)
2238{
2239 struct prison *pr;
2240
2241 sx_assert(&allprison_lock, SX_LOCKED);
2242 TAILQ_FOREACH(pr, &allprison, pr_list) {
2243 if (pr->pr_id == prid) {
2244 mtx_lock(&pr->pr_mtx);
2245 if (pr->pr_ref > 0)
2246 return (pr);
2247 mtx_unlock(&pr->pr_mtx);
2248 }
2249 }
2250 return (NULL);
2251}
2252
2253/*
2254 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2255 */
2256struct prison *
2257prison_find_child(struct prison *mypr, int prid)
2258{
2259 struct prison *pr;
2260 int descend;
2261
2262 sx_assert(&allprison_lock, SX_LOCKED);
2263 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2264 if (pr->pr_id == prid) {
2265 mtx_lock(&pr->pr_mtx);
2266 if (pr->pr_ref > 0)
2267 return (pr);
2268 mtx_unlock(&pr->pr_mtx);
2269 }
2270 }
2271 return (NULL);
2272}
2273
2274/*
2275 * Look for the name relative to mypr. Returns a locked prison or NULL.
2276 */
2277struct prison *
2278prison_find_name(struct prison *mypr, const char *name)
2279{
2280 struct prison *pr, *deadpr;
2281 size_t mylen;
2282 int descend;
2283
2284 sx_assert(&allprison_lock, SX_LOCKED);
2285 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2286 again:
2287 deadpr = NULL;
2288 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2289 if (!strcmp(pr->pr_name + mylen, name)) {
2290 mtx_lock(&pr->pr_mtx);
2291 if (pr->pr_ref > 0) {
2292 if (pr->pr_uref > 0)
2293 return (pr);
2294 deadpr = pr;
2295 }
2296 mtx_unlock(&pr->pr_mtx);
2297 }
2298 }
2299 /* There was no valid prison - perhaps there was a dying one. */
2300 if (deadpr != NULL) {
2301 mtx_lock(&deadpr->pr_mtx);
2302 if (deadpr->pr_ref == 0) {
2303 mtx_unlock(&deadpr->pr_mtx);
2304 goto again;
2305 }
2306 }
2307 return (deadpr);
2308}
2309
2310/*
2311 * See if a prison has the specific flag set.
2312 */
2313int
2314prison_flag(struct ucred *cred, unsigned flag)
2315{
2316
2317 /* This is an atomic read, so no locking is necessary. */
2318 return (cred->cr_prison->pr_flags & flag);
2319}
2320
2321int
2322prison_allow(struct ucred *cred, unsigned flag)
2323{
2324
2325 /* This is an atomic read, so no locking is necessary. */
2326 return (cred->cr_prison->pr_allow & flag);
2327}
2328
2329/*
2330 * Remove a prison reference. If that was the last reference, remove the
2331 * prison itself - but not in this context in case there are locks held.
2332 */
2333void
2334prison_free_locked(struct prison *pr)
2335{
2336
2337 mtx_assert(&pr->pr_mtx, MA_OWNED);
2338 pr->pr_ref--;
2339 if (pr->pr_ref == 0) {
2340 mtx_unlock(&pr->pr_mtx);
2341 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2342 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2343 return;
2344 }
2345 mtx_unlock(&pr->pr_mtx);
2346}
2347
2348void
2349prison_free(struct prison *pr)
2350{
2351
2352 mtx_lock(&pr->pr_mtx);
2353 prison_free_locked(pr);
2354}
2355
2356static void
2357prison_complete(void *context, int pending)
2358{
2359
2360 prison_deref((struct prison *)context, 0);
2361}
2362
2363/*
2364 * Remove a prison reference (usually). This internal version assumes no
2365 * mutexes are held, except perhaps the prison itself. If there are no more
2366 * references, release and delist the prison. On completion, the prison lock
2367 * and the allprison lock are both unlocked.
2368 */
2369static void
2370prison_deref(struct prison *pr, int flags)
2371{
2372 struct prison *ppr, *tpr;
2373 int vfslocked;
2374
2375 if (!(flags & PD_LOCKED))
2376 mtx_lock(&pr->pr_mtx);
2377 /* Decrement the user references in a separate loop. */
2378 if (flags & PD_DEUREF) {
2379 for (tpr = pr;; tpr = tpr->pr_parent) {
2380 if (tpr != pr)
2381 mtx_lock(&tpr->pr_mtx);
2382 if (--tpr->pr_uref > 0)
2383 break;
2384 KASSERT(tpr != &prison0, ("prison0 pr_uref=0"));
2385 mtx_unlock(&tpr->pr_mtx);
2386 }
2387 /* Done if there were only user references to remove. */
2388 if (!(flags & PD_DEREF)) {
2389 mtx_unlock(&tpr->pr_mtx);
2390 if (flags & PD_LIST_SLOCKED)
2391 sx_sunlock(&allprison_lock);
2392 else if (flags & PD_LIST_XLOCKED)
2393 sx_xunlock(&allprison_lock);
2394 return;
2395 }
2396 if (tpr != pr) {
2397 mtx_unlock(&tpr->pr_mtx);
2398 mtx_lock(&pr->pr_mtx);
2399 }
2400 }
2401
2402 for (;;) {
2403 if (flags & PD_DEREF)
2404 pr->pr_ref--;
2405 /* If the prison still has references, nothing else to do. */
2406 if (pr->pr_ref > 0) {
2407 mtx_unlock(&pr->pr_mtx);
2408 if (flags & PD_LIST_SLOCKED)
2409 sx_sunlock(&allprison_lock);
2410 else if (flags & PD_LIST_XLOCKED)
2411 sx_xunlock(&allprison_lock);
2412 return;
2413 }
2414
2415 mtx_unlock(&pr->pr_mtx);
2416 if (flags & PD_LIST_SLOCKED) {
2417 if (!sx_try_upgrade(&allprison_lock)) {
2418 sx_sunlock(&allprison_lock);
2419 sx_xlock(&allprison_lock);
2420 }
2421 } else if (!(flags & PD_LIST_XLOCKED))
2422 sx_xlock(&allprison_lock);
2423
2424 TAILQ_REMOVE(&allprison, pr, pr_list);
2425 LIST_REMOVE(pr, pr_sibling);
2426 ppr = pr->pr_parent;
2427 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2428 tpr->pr_prisoncount--;
2429 sx_downgrade(&allprison_lock);
2430
|
| 2431#ifdef VIMAGE
2432 if (pr->pr_flags & PR_VNET)
2433 vnet_destroy(pr->pr_vnet);
2434#endif
|
2413 if (pr->pr_root != NULL) {
2414 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2415 vrele(pr->pr_root);
2416 VFS_UNLOCK_GIANT(vfslocked);
2417 }
2418 mtx_destroy(&pr->pr_mtx);
2419#ifdef INET
2420 free(pr->pr_ip4, M_PRISON);
2421#endif
2422#ifdef INET6
2423 free(pr->pr_ip6, M_PRISON);
2424#endif
2425 if (pr->pr_cpuset != NULL)
2426 cpuset_rel(pr->pr_cpuset);
2427 osd_jail_exit(pr);
2428 free(pr, M_PRISON);
2429
2430 /* Removing a prison frees a reference on its parent. */
2431 pr = ppr;
2432 mtx_lock(&pr->pr_mtx);
2433 flags = PD_DEREF | PD_LIST_SLOCKED;
2434 }
2435}
2436
2437void
2438prison_hold_locked(struct prison *pr)
2439{
2440
2441 mtx_assert(&pr->pr_mtx, MA_OWNED);
2442 KASSERT(pr->pr_ref > 0,
2443 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2444 pr->pr_ref++;
2445}
2446
2447void
2448prison_hold(struct prison *pr)
2449{
2450
2451 mtx_lock(&pr->pr_mtx);
2452 prison_hold_locked(pr);
2453 mtx_unlock(&pr->pr_mtx);
2454}
2455
2456void
2457prison_proc_hold(struct prison *pr)
2458{
2459
2460 mtx_lock(&pr->pr_mtx);
2461 KASSERT(pr->pr_uref > 0,
2462 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2463 pr->pr_uref++;
2464 mtx_unlock(&pr->pr_mtx);
2465}
2466
2467void
2468prison_proc_free(struct prison *pr)
2469{
2470
2471 mtx_lock(&pr->pr_mtx);
2472 KASSERT(pr->pr_uref > 0,
2473 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2474 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2475}
2476
2477
2478#ifdef INET
2479/*
2480 * Restrict a prison's IP address list with its parent's, possibly replacing
2481 * it. Return true if the replacement buffer was used (or would have been).
2482 */
2483static int
2484prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2485{
2486 int ii, ij, used;
2487 struct prison *ppr;
2488
2489 ppr = pr->pr_parent;
2490 if (!(pr->pr_flags & PR_IP4_USER)) {
2491 /* This has no user settings, so just copy the parent's list. */
2492 if (pr->pr_ip4s < ppr->pr_ip4s) {
2493 /*
2494 * There's no room for the parent's list. Use the
2495 * new list buffer, which is assumed to be big enough
2496 * (if it was passed). If there's no buffer, try to
2497 * allocate one.
2498 */
2499 used = 1;
2500 if (newip4 == NULL) {
2501 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2502 M_PRISON, M_NOWAIT);
2503 if (newip4 != NULL)
2504 used = 0;
2505 }
2506 if (newip4 != NULL) {
2507 bcopy(ppr->pr_ip4, newip4,
2508 ppr->pr_ip4s * sizeof(*newip4));
2509 free(pr->pr_ip4, M_PRISON);
2510 pr->pr_ip4 = newip4;
2511 pr->pr_ip4s = ppr->pr_ip4s;
2512 pr->pr_flags |= PR_IP4;
2513 }
2514 return (used);
2515 }
2516 pr->pr_ip4s = ppr->pr_ip4s;
2517 if (pr->pr_ip4s > 0)
2518 bcopy(ppr->pr_ip4, pr->pr_ip4,
2519 pr->pr_ip4s * sizeof(*newip4));
2520 else if (pr->pr_ip4 != NULL) {
2521 free(pr->pr_ip4, M_PRISON);
2522 pr->pr_ip4 = NULL;
2523 }
2524 pr->pr_flags =
2525 (pr->pr_flags & ~PR_IP4) | (ppr->pr_flags & PR_IP4);
2526 } else if (pr->pr_ip4s > 0 && (ppr->pr_flags & PR_IP4)) {
2527 /* Remove addresses that aren't in the parent. */
2528 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2529 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2530 break;
2531 if (ij < ppr->pr_ip4s)
2532 ii = 1;
2533 else {
2534 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2535 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2536 ii = 0;
2537 }
2538 for (ij = 1; ii < pr->pr_ip4s; ) {
2539 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2540 ii++;
2541 continue;
2542 }
2543 switch (ij >= ppr->pr_ip4s ? -1 :
2544 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2545 case -1:
2546 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2547 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2548 break;
2549 case 0:
2550 ii++;
2551 ij++;
2552 break;
2553 case 1:
2554 ij++;
2555 break;
2556 }
2557 }
2558 if (pr->pr_ip4s == 0) {
2559 free(pr->pr_ip4, M_PRISON);
2560 pr->pr_ip4 = NULL;
2561 }
2562 }
2563 return (0);
2564}
2565
2566/*
2567 * Pass back primary IPv4 address of this jail.
2568 *
2569 * If not restricted return success but do not alter the address. Caller has
2570 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2571 *
2572 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2573 * Address returned in NBO.
2574 */
2575int
2576prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2577{
2578 struct prison *pr;
2579
2580 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2581 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2582
2583 pr = cred->cr_prison;
2584 if (!(pr->pr_flags & PR_IP4))
2585 return (0);
2586 mtx_lock(&pr->pr_mtx);
2587 if (!(pr->pr_flags & PR_IP4)) {
2588 mtx_unlock(&pr->pr_mtx);
2589 return (0);
2590 }
2591 if (pr->pr_ip4 == NULL) {
2592 mtx_unlock(&pr->pr_mtx);
2593 return (EAFNOSUPPORT);
2594 }
2595
2596 ia->s_addr = pr->pr_ip4[0].s_addr;
2597 mtx_unlock(&pr->pr_mtx);
2598 return (0);
2599}
2600
2601/*
2602 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2603 */
2604int
2605prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2606{
2607
2608 if (pr1 == pr2)
2609 return (1);
2610
2611 /*
2612 * jail_set maintains an exclusive hold on allprison_lock while it
2613 * changes the IP addresses, so only a shared hold is needed. This is
2614 * easier than locking the two prisons which would require finding the
2615 * proper locking order and end up needing allprison_lock anyway.
2616 */
2617 sx_slock(&allprison_lock);
2618 while (pr1 != &prison0 && !(pr1->pr_flags & PR_IP4_USER))
2619 pr1 = pr1->pr_parent;
2620 while (pr2 != &prison0 && !(pr2->pr_flags & PR_IP4_USER))
2621 pr2 = pr2->pr_parent;
2622 sx_sunlock(&allprison_lock);
2623 return (pr1 == pr2);
2624}
2625
2626/*
2627 * Make sure our (source) address is set to something meaningful to this
2628 * jail.
2629 *
2630 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2631 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2632 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2633 */
2634int
2635prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2636{
2637 struct prison *pr;
2638 struct in_addr ia0;
2639 int error;
2640
2641 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2642 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2643
2644 pr = cred->cr_prison;
2645 if (!(pr->pr_flags & PR_IP4))
2646 return (0);
2647 mtx_lock(&pr->pr_mtx);
2648 if (!(pr->pr_flags & PR_IP4)) {
2649 mtx_unlock(&pr->pr_mtx);
2650 return (0);
2651 }
2652 if (pr->pr_ip4 == NULL) {
2653 mtx_unlock(&pr->pr_mtx);
2654 return (EAFNOSUPPORT);
2655 }
2656
2657 ia0.s_addr = ntohl(ia->s_addr);
2658 if (ia0.s_addr == INADDR_LOOPBACK) {
2659 ia->s_addr = pr->pr_ip4[0].s_addr;
2660 mtx_unlock(&pr->pr_mtx);
2661 return (0);
2662 }
2663
2664 if (ia0.s_addr == INADDR_ANY) {
2665 /*
2666 * In case there is only 1 IPv4 address, bind directly.
2667 */
2668 if (pr->pr_ip4s == 1)
2669 ia->s_addr = pr->pr_ip4[0].s_addr;
2670 mtx_unlock(&pr->pr_mtx);
2671 return (0);
2672 }
2673
2674 error = _prison_check_ip4(pr, ia);
2675 mtx_unlock(&pr->pr_mtx);
2676 return (error);
2677}
2678
2679/*
2680 * Rewrite destination address in case we will connect to loopback address.
2681 *
2682 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2683 * Address passed in in NBO and returned in NBO.
2684 */
2685int
2686prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2687{
2688 struct prison *pr;
2689
2690 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2691 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2692
2693 pr = cred->cr_prison;
2694 if (!(pr->pr_flags & PR_IP4))
2695 return (0);
2696 mtx_lock(&pr->pr_mtx);
2697 if (!(pr->pr_flags & PR_IP4)) {
2698 mtx_unlock(&pr->pr_mtx);
2699 return (0);
2700 }
2701 if (pr->pr_ip4 == NULL) {
2702 mtx_unlock(&pr->pr_mtx);
2703 return (EAFNOSUPPORT);
2704 }
2705
2706 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2707 ia->s_addr = pr->pr_ip4[0].s_addr;
2708 mtx_unlock(&pr->pr_mtx);
2709 return (0);
2710 }
2711
2712 /*
2713 * Return success because nothing had to be changed.
2714 */
2715 mtx_unlock(&pr->pr_mtx);
2716 return (0);
2717}
2718
2719/*
2720 * Check if given address belongs to the jail referenced by cred/prison.
2721 *
2722 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2723 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2724 * doesn't allow IPv4. Address passed in in NBO.
2725 */
2726static int
2727_prison_check_ip4(struct prison *pr, struct in_addr *ia)
2728{
2729 int i, a, z, d;
2730
2731 /*
2732 * Check the primary IP.
2733 */
2734 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2735 return (0);
2736
2737 /*
2738 * All the other IPs are sorted so we can do a binary search.
2739 */
2740 a = 0;
2741 z = pr->pr_ip4s - 2;
2742 while (a <= z) {
2743 i = (a + z) / 2;
2744 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2745 if (d > 0)
2746 z = i - 1;
2747 else if (d < 0)
2748 a = i + 1;
2749 else
2750 return (0);
2751 }
2752
2753 return (EADDRNOTAVAIL);
2754}
2755
2756int
2757prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2758{
2759 struct prison *pr;
2760 int error;
2761
2762 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2763 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2764
2765 pr = cred->cr_prison;
2766 if (!(pr->pr_flags & PR_IP4))
2767 return (0);
2768 mtx_lock(&pr->pr_mtx);
2769 if (!(pr->pr_flags & PR_IP4)) {
2770 mtx_unlock(&pr->pr_mtx);
2771 return (0);
2772 }
2773 if (pr->pr_ip4 == NULL) {
2774 mtx_unlock(&pr->pr_mtx);
2775 return (EAFNOSUPPORT);
2776 }
2777
2778 error = _prison_check_ip4(pr, ia);
2779 mtx_unlock(&pr->pr_mtx);
2780 return (error);
2781}
2782#endif
2783
2784#ifdef INET6
2785static int
2786prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2787{
2788 int ii, ij, used;
2789 struct prison *ppr;
2790
2791 ppr = pr->pr_parent;
2792 if (!(pr->pr_flags & PR_IP6_USER)) {
2793 /* This has no user settings, so just copy the parent's list. */
2794 if (pr->pr_ip6s < ppr->pr_ip6s) {
2795 /*
2796 * There's no room for the parent's list. Use the
2797 * new list buffer, which is assumed to be big enough
2798 * (if it was passed). If there's no buffer, try to
2799 * allocate one.
2800 */
2801 used = 1;
2802 if (newip6 == NULL) {
2803 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
2804 M_PRISON, M_NOWAIT);
2805 if (newip6 != NULL)
2806 used = 0;
2807 }
2808 if (newip6 != NULL) {
2809 bcopy(ppr->pr_ip6, newip6,
2810 ppr->pr_ip6s * sizeof(*newip6));
2811 free(pr->pr_ip6, M_PRISON);
2812 pr->pr_ip6 = newip6;
2813 pr->pr_ip6s = ppr->pr_ip6s;
2814 pr->pr_flags |= PR_IP6;
2815 }
2816 return (used);
2817 }
2818 pr->pr_ip6s = ppr->pr_ip6s;
2819 if (pr->pr_ip6s > 0)
2820 bcopy(ppr->pr_ip6, pr->pr_ip6,
2821 pr->pr_ip6s * sizeof(*newip6));
2822 else if (pr->pr_ip6 != NULL) {
2823 free(pr->pr_ip6, M_PRISON);
2824 pr->pr_ip6 = NULL;
2825 }
2826 pr->pr_flags =
2827 (pr->pr_flags & ~PR_IP6) | (ppr->pr_flags & PR_IP6);
2828 } else if (pr->pr_ip6s > 0 && (ppr->pr_flags & PR_IP6)) {
2829 /* Remove addresses that aren't in the parent. */
2830 for (ij = 0; ij < ppr->pr_ip6s; ij++)
2831 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
2832 &ppr->pr_ip6[ij]))
2833 break;
2834 if (ij < ppr->pr_ip6s)
2835 ii = 1;
2836 else {
2837 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
2838 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
2839 ii = 0;
2840 }
2841 for (ij = 1; ii < pr->pr_ip6s; ) {
2842 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
2843 &ppr->pr_ip6[0])) {
2844 ii++;
2845 continue;
2846 }
2847 switch (ij >= ppr->pr_ip4s ? -1 :
2848 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
2849 case -1:
2850 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
2851 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
2852 break;
2853 case 0:
2854 ii++;
2855 ij++;
2856 break;
2857 case 1:
2858 ij++;
2859 break;
2860 }
2861 }
2862 if (pr->pr_ip6s == 0) {
2863 free(pr->pr_ip6, M_PRISON);
2864 pr->pr_ip6 = NULL;
2865 }
2866 }
2867 return 0;
2868}
2869
2870/*
2871 * Pass back primary IPv6 address for this jail.
2872 *
2873 * If not restricted return success but do not alter the address. Caller has
2874 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
2875 *
2876 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
2877 */
2878int
2879prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
2880{
2881 struct prison *pr;
2882
2883 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2884 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
2885
2886 pr = cred->cr_prison;
2887 if (!(pr->pr_flags & PR_IP6))
2888 return (0);
2889 mtx_lock(&pr->pr_mtx);
2890 if (!(pr->pr_flags & PR_IP6)) {
2891 mtx_unlock(&pr->pr_mtx);
2892 return (0);
2893 }
2894 if (pr->pr_ip6 == NULL) {
2895 mtx_unlock(&pr->pr_mtx);
2896 return (EAFNOSUPPORT);
2897 }
2898
2899 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2900 mtx_unlock(&pr->pr_mtx);
2901 return (0);
2902}
2903
2904/*
2905 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
2906 */
2907int
2908prison_equal_ip6(struct prison *pr1, struct prison *pr2)
2909{
2910
2911 if (pr1 == pr2)
2912 return (1);
2913
2914 sx_slock(&allprison_lock);
2915 while (pr1 != &prison0 && !(pr1->pr_flags & PR_IP6_USER))
2916 pr1 = pr1->pr_parent;
2917 while (pr2 != &prison0 && !(pr2->pr_flags & PR_IP6_USER))
2918 pr2 = pr2->pr_parent;
2919 sx_sunlock(&allprison_lock);
2920 return (pr1 == pr2);
2921}
2922
2923/*
2924 * Make sure our (source) address is set to something meaningful to this jail.
2925 *
2926 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
2927 * when needed while binding.
2928 *
2929 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
2930 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2931 * doesn't allow IPv6.
2932 */
2933int
2934prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
2935{
2936 struct prison *pr;
2937 int error;
2938
2939 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2940 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
2941
2942 pr = cred->cr_prison;
2943 if (!(pr->pr_flags & PR_IP6))
2944 return (0);
2945 mtx_lock(&pr->pr_mtx);
2946 if (!(pr->pr_flags & PR_IP6)) {
2947 mtx_unlock(&pr->pr_mtx);
2948 return (0);
2949 }
2950 if (pr->pr_ip6 == NULL) {
2951 mtx_unlock(&pr->pr_mtx);
2952 return (EAFNOSUPPORT);
2953 }
2954
2955 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
2956 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2957 mtx_unlock(&pr->pr_mtx);
2958 return (0);
2959 }
2960
2961 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
2962 /*
2963 * In case there is only 1 IPv6 address, and v6only is true,
2964 * then bind directly.
2965 */
2966 if (v6only != 0 && pr->pr_ip6s == 1)
2967 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2968 mtx_unlock(&pr->pr_mtx);
2969 return (0);
2970 }
2971
2972 error = _prison_check_ip6(pr, ia6);
2973 mtx_unlock(&pr->pr_mtx);
2974 return (error);
2975}
2976
2977/*
2978 * Rewrite destination address in case we will connect to loopback address.
2979 *
2980 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
2981 */
2982int
2983prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
2984{
2985 struct prison *pr;
2986
2987 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2988 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
2989
2990 pr = cred->cr_prison;
2991 if (!(pr->pr_flags & PR_IP6))
2992 return (0);
2993 mtx_lock(&pr->pr_mtx);
2994 if (!(pr->pr_flags & PR_IP6)) {
2995 mtx_unlock(&pr->pr_mtx);
2996 return (0);
2997 }
2998 if (pr->pr_ip6 == NULL) {
2999 mtx_unlock(&pr->pr_mtx);
3000 return (EAFNOSUPPORT);
3001 }
3002
3003 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3004 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3005 mtx_unlock(&pr->pr_mtx);
3006 return (0);
3007 }
3008
3009 /*
3010 * Return success because nothing had to be changed.
3011 */
3012 mtx_unlock(&pr->pr_mtx);
3013 return (0);
3014}
3015
3016/*
3017 * Check if given address belongs to the jail referenced by cred/prison.
3018 *
3019 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3020 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3021 * doesn't allow IPv6.
3022 */
3023static int
3024_prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3025{
3026 int i, a, z, d;
3027
3028 /*
3029 * Check the primary IP.
3030 */
3031 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3032 return (0);
3033
3034 /*
3035 * All the other IPs are sorted so we can do a binary search.
3036 */
3037 a = 0;
3038 z = pr->pr_ip6s - 2;
3039 while (a <= z) {
3040 i = (a + z) / 2;
3041 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3042 if (d > 0)
3043 z = i - 1;
3044 else if (d < 0)
3045 a = i + 1;
3046 else
3047 return (0);
3048 }
3049
3050 return (EADDRNOTAVAIL);
3051}
3052
3053int
3054prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3055{
3056 struct prison *pr;
3057 int error;
3058
3059 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3060 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3061
3062 pr = cred->cr_prison;
3063 if (!(pr->pr_flags & PR_IP6))
3064 return (0);
3065 mtx_lock(&pr->pr_mtx);
3066 if (!(pr->pr_flags & PR_IP6)) {
3067 mtx_unlock(&pr->pr_mtx);
3068 return (0);
3069 }
3070 if (pr->pr_ip6 == NULL) {
3071 mtx_unlock(&pr->pr_mtx);
3072 return (EAFNOSUPPORT);
3073 }
3074
3075 error = _prison_check_ip6(pr, ia6);
3076 mtx_unlock(&pr->pr_mtx);
3077 return (error);
3078}
3079#endif
3080
3081/*
3082 * Check if a jail supports the given address family.
3083 *
3084 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3085 * if not.
3086 */
3087int
3088prison_check_af(struct ucred *cred, int af)
3089{
3090 struct prison *pr;
3091 int error;
3092
3093 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3094
3095 pr = cred->cr_prison;
3096 error = 0;
3097 switch (af)
3098 {
3099#ifdef INET
3100 case AF_INET:
3101 if (pr->pr_flags & PR_IP4)
3102 {
3103 mtx_lock(&pr->pr_mtx);
3104 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3105 error = EAFNOSUPPORT;
3106 mtx_unlock(&pr->pr_mtx);
3107 }
3108 break;
3109#endif
3110#ifdef INET6
3111 case AF_INET6:
3112 if (pr->pr_flags & PR_IP6)
3113 {
3114 mtx_lock(&pr->pr_mtx);
3115 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3116 error = EAFNOSUPPORT;
3117 mtx_unlock(&pr->pr_mtx);
3118 }
3119 break;
3120#endif
3121 case AF_LOCAL:
3122 case AF_ROUTE:
3123 break;
3124 default:
3125 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3126 error = EAFNOSUPPORT;
3127 }
3128 return (error);
3129}
3130
3131/*
3132 * Check if given address belongs to the jail referenced by cred (wrapper to
3133 * prison_check_ip[46]).
3134 *
3135 * Returns 0 if jail doesn't restrict the address family or if address belongs
3136 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3137 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3138 */
3139int
3140prison_if(struct ucred *cred, struct sockaddr *sa)
3141{
3142#ifdef INET
3143 struct sockaddr_in *sai;
3144#endif
3145#ifdef INET6
3146 struct sockaddr_in6 *sai6;
3147#endif
3148 int error;
3149
3150 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3151 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3152
3153 error = 0;
3154 switch (sa->sa_family)
3155 {
3156#ifdef INET
3157 case AF_INET:
3158 sai = (struct sockaddr_in *)sa;
3159 error = prison_check_ip4(cred, &sai->sin_addr);
3160 break;
3161#endif
3162#ifdef INET6
3163 case AF_INET6:
3164 sai6 = (struct sockaddr_in6 *)sa;
3165 error = prison_check_ip6(cred, &sai6->sin6_addr);
3166 break;
3167#endif
3168 default:
3169 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3170 error = EAFNOSUPPORT;
3171 }
3172 return (error);
3173}
3174
3175/*
3176 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3177 */
3178int
3179prison_check(struct ucred *cred1, struct ucred *cred2)
3180{
3181
3182#ifdef VIMAGE
3183 if (cred2->cr_vimage->v_procg != cred1->cr_vimage->v_procg)
3184 return (ESRCH);
3185#endif
3186 return ((cred1->cr_prison == cred2->cr_prison ||
3187 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3188}
3189
3190/*
3191 * Return 1 if p2 is a child of p1, otherwise 0.
3192 */
3193int
3194prison_ischild(struct prison *pr1, struct prison *pr2)
3195{
3196
3197 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3198 if (pr1 == pr2)
3199 return (1);
3200 return (0);
3201}
3202
3203/*
3204 * Return 1 if the passed credential is in a jail, otherwise 0.
3205 */
3206int
3207jailed(struct ucred *cred)
3208{
3209
3210 return (cred->cr_prison != &prison0);
3211}
3212
3213/*
3214 * Return the correct hostname (domainname, et al) for the passed credential.
3215 */
3216void
3217getcredhostname(struct ucred *cred, char *buf, size_t size)
3218{
3219 struct prison *pr;
3220
3221 /*
3222 * A NULL credential can be used to shortcut to the physical
3223 * system's hostname.
3224 */
3225 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3226 mtx_lock(&pr->pr_mtx);
3227 strlcpy(buf, pr->pr_hostname, size);
3228 mtx_unlock(&pr->pr_mtx);
3229}
3230
3231void
3232getcreddomainname(struct ucred *cred, char *buf, size_t size)
3233{
3234
3235 mtx_lock(&cred->cr_prison->pr_mtx);
3236 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3237 mtx_unlock(&cred->cr_prison->pr_mtx);
3238}
3239
3240void
3241getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3242{
3243
3244 mtx_lock(&cred->cr_prison->pr_mtx);
3245 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3246 mtx_unlock(&cred->cr_prison->pr_mtx);
3247}
3248
3249void
3250getcredhostid(struct ucred *cred, unsigned long *hostid)
3251{
3252
3253 mtx_lock(&cred->cr_prison->pr_mtx);
3254 *hostid = cred->cr_prison->pr_hostid;
3255 mtx_unlock(&cred->cr_prison->pr_mtx);
3256}
3257
3258/*
3259 * Determine whether the subject represented by cred can "see"
3260 * status of a mount point.
3261 * Returns: 0 for permitted, ENOENT otherwise.
3262 * XXX: This function should be called cr_canseemount() and should be
3263 * placed in kern_prot.c.
3264 */
3265int
3266prison_canseemount(struct ucred *cred, struct mount *mp)
3267{
3268 struct prison *pr;
3269 struct statfs *sp;
3270 size_t len;
3271
3272 pr = cred->cr_prison;
3273 if (pr->pr_enforce_statfs == 0)
3274 return (0);
3275 if (pr->pr_root->v_mount == mp)
3276 return (0);
3277 if (pr->pr_enforce_statfs == 2)
3278 return (ENOENT);
3279 /*
3280 * If jail's chroot directory is set to "/" we should be able to see
3281 * all mount-points from inside a jail.
3282 * This is ugly check, but this is the only situation when jail's
3283 * directory ends with '/'.
3284 */
3285 if (strcmp(pr->pr_path, "/") == 0)
3286 return (0);
3287 len = strlen(pr->pr_path);
3288 sp = &mp->mnt_stat;
3289 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3290 return (ENOENT);
3291 /*
3292 * Be sure that we don't have situation where jail's root directory
3293 * is "/some/path" and mount point is "/some/pathpath".
3294 */
3295 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3296 return (ENOENT);
3297 return (0);
3298}
3299
3300void
3301prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3302{
3303 char jpath[MAXPATHLEN];
3304 struct prison *pr;
3305 size_t len;
3306
3307 pr = cred->cr_prison;
3308 if (pr->pr_enforce_statfs == 0)
3309 return;
3310 if (prison_canseemount(cred, mp) != 0) {
3311 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3312 strlcpy(sp->f_mntonname, "[restricted]",
3313 sizeof(sp->f_mntonname));
3314 return;
3315 }
3316 if (pr->pr_root->v_mount == mp) {
3317 /*
3318 * Clear current buffer data, so we are sure nothing from
3319 * the valid path left there.
3320 */
3321 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3322 *sp->f_mntonname = '/';
3323 return;
3324 }
3325 /*
3326 * If jail's chroot directory is set to "/" we should be able to see
3327 * all mount-points from inside a jail.
3328 */
3329 if (strcmp(pr->pr_path, "/") == 0)
3330 return;
3331 len = strlen(pr->pr_path);
3332 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3333 /*
3334 * Clear current buffer data, so we are sure nothing from
3335 * the valid path left there.
3336 */
3337 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3338 if (*jpath == '\0') {
3339 /* Should never happen. */
3340 *sp->f_mntonname = '/';
3341 } else {
3342 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3343 }
3344}
3345
3346/*
3347 * Check with permission for a specific privilege is granted within jail. We
3348 * have a specific list of accepted privileges; the rest are denied.
3349 */
3350int
3351prison_priv_check(struct ucred *cred, int priv)
3352{
3353
3354 if (!jailed(cred))
3355 return (0);
3356
3357 switch (priv) {
3358
3359 /*
3360 * Allow ktrace privileges for root in jail.
3361 */
3362 case PRIV_KTRACE:
3363
3364#if 0
3365 /*
3366 * Allow jailed processes to configure audit identity and
3367 * submit audit records (login, etc). In the future we may
3368 * want to further refine the relationship between audit and
3369 * jail.
3370 */
3371 case PRIV_AUDIT_GETAUDIT:
3372 case PRIV_AUDIT_SETAUDIT:
3373 case PRIV_AUDIT_SUBMIT:
3374#endif
3375
3376 /*
3377 * Allow jailed processes to manipulate process UNIX
3378 * credentials in any way they see fit.
3379 */
3380 case PRIV_CRED_SETUID:
3381 case PRIV_CRED_SETEUID:
3382 case PRIV_CRED_SETGID:
3383 case PRIV_CRED_SETEGID:
3384 case PRIV_CRED_SETGROUPS:
3385 case PRIV_CRED_SETREUID:
3386 case PRIV_CRED_SETREGID:
3387 case PRIV_CRED_SETRESUID:
3388 case PRIV_CRED_SETRESGID:
3389
3390 /*
3391 * Jail implements visibility constraints already, so allow
3392 * jailed root to override uid/gid-based constraints.
3393 */
3394 case PRIV_SEEOTHERGIDS:
3395 case PRIV_SEEOTHERUIDS:
3396
3397 /*
3398 * Jail implements inter-process debugging limits already, so
3399 * allow jailed root various debugging privileges.
3400 */
3401 case PRIV_DEBUG_DIFFCRED:
3402 case PRIV_DEBUG_SUGID:
3403 case PRIV_DEBUG_UNPRIV:
3404
3405 /*
3406 * Allow jail to set various resource limits and login
3407 * properties, and for now, exceed process resource limits.
3408 */
3409 case PRIV_PROC_LIMIT:
3410 case PRIV_PROC_SETLOGIN:
3411 case PRIV_PROC_SETRLIMIT:
3412
3413 /*
3414 * System V and POSIX IPC privileges are granted in jail.
3415 */
3416 case PRIV_IPC_READ:
3417 case PRIV_IPC_WRITE:
3418 case PRIV_IPC_ADMIN:
3419 case PRIV_IPC_MSGSIZE:
3420 case PRIV_MQ_ADMIN:
3421
3422 /*
3423 * Jail operations within a jail work on child jails.
3424 */
3425 case PRIV_JAIL_ATTACH:
3426 case PRIV_JAIL_SET:
3427 case PRIV_JAIL_REMOVE:
3428
3429 /*
3430 * Jail implements its own inter-process limits, so allow
3431 * root processes in jail to change scheduling on other
3432 * processes in the same jail. Likewise for signalling.
3433 */
3434 case PRIV_SCHED_DIFFCRED:
3435 case PRIV_SCHED_CPUSET:
3436 case PRIV_SIGNAL_DIFFCRED:
3437 case PRIV_SIGNAL_SUGID:
3438
3439 /*
3440 * Allow jailed processes to write to sysctls marked as jail
3441 * writable.
3442 */
3443 case PRIV_SYSCTL_WRITEJAIL:
3444
3445 /*
3446 * Allow root in jail to manage a variety of quota
3447 * properties. These should likely be conditional on a
3448 * configuration option.
3449 */
3450 case PRIV_VFS_GETQUOTA:
3451 case PRIV_VFS_SETQUOTA:
3452
3453 /*
3454 * Since Jail relies on chroot() to implement file system
3455 * protections, grant many VFS privileges to root in jail.
3456 * Be careful to exclude mount-related and NFS-related
3457 * privileges.
3458 */
3459 case PRIV_VFS_READ:
3460 case PRIV_VFS_WRITE:
3461 case PRIV_VFS_ADMIN:
3462 case PRIV_VFS_EXEC:
3463 case PRIV_VFS_LOOKUP:
3464 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3465 case PRIV_VFS_CHFLAGS_DEV:
3466 case PRIV_VFS_CHOWN:
3467 case PRIV_VFS_CHROOT:
3468 case PRIV_VFS_RETAINSUGID:
3469 case PRIV_VFS_FCHROOT:
3470 case PRIV_VFS_LINK:
3471 case PRIV_VFS_SETGID:
3472 case PRIV_VFS_STAT:
3473 case PRIV_VFS_STICKYFILE:
3474 return (0);
3475
3476 /*
3477 * Depending on the global setting, allow privilege of
3478 * setting system flags.
3479 */
3480 case PRIV_VFS_SYSFLAGS:
3481 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3482 return (0);
3483 else
3484 return (EPERM);
3485
3486 /*
3487 * Depending on the global setting, allow privilege of
3488 * mounting/unmounting file systems.
3489 */
3490 case PRIV_VFS_MOUNT:
3491 case PRIV_VFS_UNMOUNT:
3492 case PRIV_VFS_MOUNT_NONUSER:
3493 case PRIV_VFS_MOUNT_OWNER:
3494 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT)
3495 return (0);
3496 else
3497 return (EPERM);
3498
3499 /*
3500 * Allow jailed root to bind reserved ports and reuse in-use
3501 * ports.
3502 */
3503 case PRIV_NETINET_RESERVEDPORT:
3504 case PRIV_NETINET_REUSEPORT:
3505 return (0);
3506
3507 /*
3508 * Allow jailed root to set certian IPv4/6 (option) headers.
3509 */
3510 case PRIV_NETINET_SETHDROPTS:
3511 return (0);
3512
3513 /*
3514 * Conditionally allow creating raw sockets in jail.
3515 */
3516 case PRIV_NETINET_RAW:
3517 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3518 return (0);
3519 else
3520 return (EPERM);
3521
3522 /*
3523 * Since jail implements its own visibility limits on netstat
3524 * sysctls, allow getcred. This allows identd to work in
3525 * jail.
3526 */
3527 case PRIV_NETINET_GETCRED:
3528 return (0);
3529
3530 default:
3531 /*
3532 * In all remaining cases, deny the privilege request. This
3533 * includes almost all network privileges, many system
3534 * configuration privileges.
3535 */
3536 return (EPERM);
3537 }
3538}
3539
3540/*
3541 * Return the part of pr2's name that is relative to pr1, or the whole name
3542 * if it does not directly follow.
3543 */
3544
3545char *
3546prison_name(struct prison *pr1, struct prison *pr2)
3547{
3548 char *name;
3549
3550 /* Jails see themselves as "0" (if they see themselves at all). */
3551 if (pr1 == pr2)
3552 return "0";
3553 name = pr2->pr_name;
3554 if (prison_ischild(pr1, pr2)) {
3555 /*
3556 * pr1 isn't locked (and allprison_lock may not be either)
3557 * so its length can't be counted on. But the number of dots
3558 * can be counted on - and counted.
3559 */
3560 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3561 name = strchr(name, '.') + 1;
3562 }
3563 return (name);
3564}
3565
3566/*
3567 * Return the part of pr2's path that is relative to pr1, or the whole path
3568 * if it does not directly follow.
3569 */
3570static char *
3571prison_path(struct prison *pr1, struct prison *pr2)
3572{
3573 char *path1, *path2;
3574 int len1;
3575
3576 path1 = pr1->pr_path;
3577 path2 = pr2->pr_path;
3578 if (!strcmp(path1, "/"))
3579 return (path2);
3580 len1 = strlen(path1);
3581 if (strncmp(path1, path2, len1))
3582 return (path2);
3583 if (path2[len1] == '\0')
3584 return "/";
3585 if (path2[len1] == '/')
3586 return (path2 + len1);
3587 return (path2);
3588}
3589
3590
3591/*
3592 * Jail-related sysctls.
3593 */
3594SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3595 "Jails");
3596
3597static int
3598sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3599{
3600 struct xprison *xp;
3601 struct prison *pr, *cpr;
3602#ifdef INET
3603 struct in_addr *ip4 = NULL;
3604 int ip4s = 0;
3605#endif
3606#ifdef INET6
3607 struct in_addr *ip6 = NULL;
3608 int ip6s = 0;
3609#endif
3610 int descend, error;
3611
3612 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3613 pr = req->td->td_ucred->cr_prison;
3614 error = 0;
3615 sx_slock(&allprison_lock);
3616 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3617#if defined(INET) || defined(INET6)
3618 again:
3619#endif
3620 mtx_lock(&cpr->pr_mtx);
3621#ifdef INET
3622 if (cpr->pr_ip4s > 0) {
3623 if (ip4s < cpr->pr_ip4s) {
3624 ip4s = cpr->pr_ip4s;
3625 mtx_unlock(&cpr->pr_mtx);
3626 ip4 = realloc(ip4, ip4s *
3627 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3628 goto again;
3629 }
3630 bcopy(cpr->pr_ip4, ip4,
3631 cpr->pr_ip4s * sizeof(struct in_addr));
3632 }
3633#endif
3634#ifdef INET6
3635 if (cpr->pr_ip6s > 0) {
3636 if (ip6s < cpr->pr_ip6s) {
3637 ip6s = cpr->pr_ip6s;
3638 mtx_unlock(&cpr->pr_mtx);
3639 ip6 = realloc(ip6, ip6s *
3640 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3641 goto again;
3642 }
3643 bcopy(cpr->pr_ip6, ip6,
3644 cpr->pr_ip6s * sizeof(struct in6_addr));
3645 }
3646#endif
3647 if (cpr->pr_ref == 0) {
3648 mtx_unlock(&cpr->pr_mtx);
3649 continue;
3650 }
3651 bzero(xp, sizeof(*xp));
3652 xp->pr_version = XPRISON_VERSION;
3653 xp->pr_id = cpr->pr_id;
3654 xp->pr_state = cpr->pr_uref > 0
3655 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3656 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3657 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3658 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3659#ifdef INET
3660 xp->pr_ip4s = cpr->pr_ip4s;
3661#endif
3662#ifdef INET6
3663 xp->pr_ip6s = cpr->pr_ip6s;
3664#endif
3665 mtx_unlock(&cpr->pr_mtx);
3666 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3667 if (error)
3668 break;
3669#ifdef INET
3670 if (xp->pr_ip4s > 0) {
3671 error = SYSCTL_OUT(req, ip4,
3672 xp->pr_ip4s * sizeof(struct in_addr));
3673 if (error)
3674 break;
3675 }
3676#endif
3677#ifdef INET6
3678 if (xp->pr_ip6s > 0) {
3679 error = SYSCTL_OUT(req, ip6,
3680 xp->pr_ip6s * sizeof(struct in6_addr));
3681 if (error)
3682 break;
3683 }
3684#endif
3685 }
3686 sx_sunlock(&allprison_lock);
3687 free(xp, M_TEMP);
3688#ifdef INET
3689 free(ip4, M_TEMP);
3690#endif
3691#ifdef INET6
3692 free(ip6, M_TEMP);
3693#endif
3694 return (error);
3695}
3696
3697SYSCTL_OID(_security_jail, OID_AUTO, list,
3698 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3699 sysctl_jail_list, "S", "List of active jails");
3700
3701static int
3702sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3703{
3704 int error, injail;
3705
3706 injail = jailed(req->td->td_ucred);
3707 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3708
3709 return (error);
3710}
3711
3712SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3713 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3714 sysctl_jail_jailed, "I", "Process in jail?");
3715
3716#if defined(INET) || defined(INET6)
3717SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3718 &jail_max_af_ips, 0,
3719 "Number of IP addresses a jail may have at most per address family");
3720#endif
3721
3722/*
3723 * Default parameters for jail(2) compatability. For historical reasons,
3724 * the sysctl names have varying similarity to the parameter names. Prisons
3725 * just see their own parameters, and can't change them.
3726 */
3727static int
3728sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3729{
3730 struct prison *pr;
3731 int allow, error, i;
3732
3733 pr = req->td->td_ucred->cr_prison;
3734 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3735
3736 /* Get the current flag value, and convert it to a boolean. */
3737 i = (allow & arg2) ? 1 : 0;
3738 if (arg1 != NULL)
3739 i = !i;
3740 error = sysctl_handle_int(oidp, &i, 0, req);
3741 if (error || !req->newptr)
3742 return (error);
3743 i = i ? arg2 : 0;
3744 if (arg1 != NULL)
3745 i ^= arg2;
3746 /*
3747 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3748 * for writing.
3749 */
3750 mtx_lock(&prison0.pr_mtx);
3751 jail_default_allow = (jail_default_allow & ~arg2) | i;
3752 mtx_unlock(&prison0.pr_mtx);
3753 return (0);
3754}
3755
3756SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3757 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3758 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3759 "Processes in jail can set their hostnames");
3760SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3761 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3762 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3763 "Processes in jail are limited to creating UNIX/IP/route sockets only");
3764SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3765 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3766 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3767 "Processes in jail can use System V IPC primitives");
3768SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3769 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3770 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3771 "Prison root can create raw sockets");
3772SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3773 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3774 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3775 "Processes in jail can alter system file flags");
3776SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3777 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3778 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3779 "Processes in jail can mount/unmount jail-friendly file systems");
3780
3781static int
3782sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3783{
3784 struct prison *pr;
3785 int level, error;
3786
3787 pr = req->td->td_ucred->cr_prison;
3788 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3789 error = sysctl_handle_int(oidp, &level, 0, req);
3790 if (error || !req->newptr)
3791 return (error);
3792 *(int *)arg1 = level;
3793 return (0);
3794}
3795
3796SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3797 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3798 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3799 sysctl_jail_default_level, "I",
3800 "Processes in jail cannot see all mounted file systems");
3801
3802/*
3803 * Nodes to describe jail parameters. Maximum length of string parameters
3804 * is returned in the string itself, and the other parameters exist merely
3805 * to make themselves and their types known.
3806 */
3807SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
3808 "Jail parameters");
3809
3810int
3811sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3812{
3813 int i;
3814 long l;
3815 size_t s;
3816 char numbuf[12];
3817
3818 switch (oidp->oid_kind & CTLTYPE)
3819 {
3820 case CTLTYPE_LONG:
3821 case CTLTYPE_ULONG:
3822 l = 0;
3823#ifdef SCTL_MASK32
3824 if (!(req->flags & SCTL_MASK32))
3825#endif
3826 return (SYSCTL_OUT(req, &l, sizeof(l)));
3827 case CTLTYPE_INT:
3828 case CTLTYPE_UINT:
3829 i = 0;
3830 return (SYSCTL_OUT(req, &i, sizeof(i)));
3831 case CTLTYPE_STRING:
3832 snprintf(numbuf, sizeof(numbuf), "%d", arg2);
3833 return
3834 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3835 case CTLTYPE_STRUCT:
3836 s = (size_t)arg2;
3837 return (SYSCTL_OUT(req, &s, sizeof(s)));
3838 }
3839 return (0);
3840}
3841
3842SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3843SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3844SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3845SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3846SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3847 "I", "Jail secure level");
3848SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3849 "I", "Jail cannot see all mounted file systems");
3850SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3851 "B", "Jail persistence");
| 2435 if (pr->pr_root != NULL) {
2436 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2437 vrele(pr->pr_root);
2438 VFS_UNLOCK_GIANT(vfslocked);
2439 }
2440 mtx_destroy(&pr->pr_mtx);
2441#ifdef INET
2442 free(pr->pr_ip4, M_PRISON);
2443#endif
2444#ifdef INET6
2445 free(pr->pr_ip6, M_PRISON);
2446#endif
2447 if (pr->pr_cpuset != NULL)
2448 cpuset_rel(pr->pr_cpuset);
2449 osd_jail_exit(pr);
2450 free(pr, M_PRISON);
2451
2452 /* Removing a prison frees a reference on its parent. */
2453 pr = ppr;
2454 mtx_lock(&pr->pr_mtx);
2455 flags = PD_DEREF | PD_LIST_SLOCKED;
2456 }
2457}
2458
2459void
2460prison_hold_locked(struct prison *pr)
2461{
2462
2463 mtx_assert(&pr->pr_mtx, MA_OWNED);
2464 KASSERT(pr->pr_ref > 0,
2465 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2466 pr->pr_ref++;
2467}
2468
2469void
2470prison_hold(struct prison *pr)
2471{
2472
2473 mtx_lock(&pr->pr_mtx);
2474 prison_hold_locked(pr);
2475 mtx_unlock(&pr->pr_mtx);
2476}
2477
2478void
2479prison_proc_hold(struct prison *pr)
2480{
2481
2482 mtx_lock(&pr->pr_mtx);
2483 KASSERT(pr->pr_uref > 0,
2484 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2485 pr->pr_uref++;
2486 mtx_unlock(&pr->pr_mtx);
2487}
2488
2489void
2490prison_proc_free(struct prison *pr)
2491{
2492
2493 mtx_lock(&pr->pr_mtx);
2494 KASSERT(pr->pr_uref > 0,
2495 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2496 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2497}
2498
2499
2500#ifdef INET
2501/*
2502 * Restrict a prison's IP address list with its parent's, possibly replacing
2503 * it. Return true if the replacement buffer was used (or would have been).
2504 */
2505static int
2506prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2507{
2508 int ii, ij, used;
2509 struct prison *ppr;
2510
2511 ppr = pr->pr_parent;
2512 if (!(pr->pr_flags & PR_IP4_USER)) {
2513 /* This has no user settings, so just copy the parent's list. */
2514 if (pr->pr_ip4s < ppr->pr_ip4s) {
2515 /*
2516 * There's no room for the parent's list. Use the
2517 * new list buffer, which is assumed to be big enough
2518 * (if it was passed). If there's no buffer, try to
2519 * allocate one.
2520 */
2521 used = 1;
2522 if (newip4 == NULL) {
2523 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2524 M_PRISON, M_NOWAIT);
2525 if (newip4 != NULL)
2526 used = 0;
2527 }
2528 if (newip4 != NULL) {
2529 bcopy(ppr->pr_ip4, newip4,
2530 ppr->pr_ip4s * sizeof(*newip4));
2531 free(pr->pr_ip4, M_PRISON);
2532 pr->pr_ip4 = newip4;
2533 pr->pr_ip4s = ppr->pr_ip4s;
2534 pr->pr_flags |= PR_IP4;
2535 }
2536 return (used);
2537 }
2538 pr->pr_ip4s = ppr->pr_ip4s;
2539 if (pr->pr_ip4s > 0)
2540 bcopy(ppr->pr_ip4, pr->pr_ip4,
2541 pr->pr_ip4s * sizeof(*newip4));
2542 else if (pr->pr_ip4 != NULL) {
2543 free(pr->pr_ip4, M_PRISON);
2544 pr->pr_ip4 = NULL;
2545 }
2546 pr->pr_flags =
2547 (pr->pr_flags & ~PR_IP4) | (ppr->pr_flags & PR_IP4);
2548 } else if (pr->pr_ip4s > 0 && (ppr->pr_flags & PR_IP4)) {
2549 /* Remove addresses that aren't in the parent. */
2550 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2551 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2552 break;
2553 if (ij < ppr->pr_ip4s)
2554 ii = 1;
2555 else {
2556 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2557 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2558 ii = 0;
2559 }
2560 for (ij = 1; ii < pr->pr_ip4s; ) {
2561 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2562 ii++;
2563 continue;
2564 }
2565 switch (ij >= ppr->pr_ip4s ? -1 :
2566 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2567 case -1:
2568 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2569 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2570 break;
2571 case 0:
2572 ii++;
2573 ij++;
2574 break;
2575 case 1:
2576 ij++;
2577 break;
2578 }
2579 }
2580 if (pr->pr_ip4s == 0) {
2581 free(pr->pr_ip4, M_PRISON);
2582 pr->pr_ip4 = NULL;
2583 }
2584 }
2585 return (0);
2586}
2587
2588/*
2589 * Pass back primary IPv4 address of this jail.
2590 *
2591 * If not restricted return success but do not alter the address. Caller has
2592 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2593 *
2594 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2595 * Address returned in NBO.
2596 */
2597int
2598prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2599{
2600 struct prison *pr;
2601
2602 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2603 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2604
2605 pr = cred->cr_prison;
2606 if (!(pr->pr_flags & PR_IP4))
2607 return (0);
2608 mtx_lock(&pr->pr_mtx);
2609 if (!(pr->pr_flags & PR_IP4)) {
2610 mtx_unlock(&pr->pr_mtx);
2611 return (0);
2612 }
2613 if (pr->pr_ip4 == NULL) {
2614 mtx_unlock(&pr->pr_mtx);
2615 return (EAFNOSUPPORT);
2616 }
2617
2618 ia->s_addr = pr->pr_ip4[0].s_addr;
2619 mtx_unlock(&pr->pr_mtx);
2620 return (0);
2621}
2622
2623/*
2624 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2625 */
2626int
2627prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2628{
2629
2630 if (pr1 == pr2)
2631 return (1);
2632
2633 /*
2634 * jail_set maintains an exclusive hold on allprison_lock while it
2635 * changes the IP addresses, so only a shared hold is needed. This is
2636 * easier than locking the two prisons which would require finding the
2637 * proper locking order and end up needing allprison_lock anyway.
2638 */
2639 sx_slock(&allprison_lock);
2640 while (pr1 != &prison0 && !(pr1->pr_flags & PR_IP4_USER))
2641 pr1 = pr1->pr_parent;
2642 while (pr2 != &prison0 && !(pr2->pr_flags & PR_IP4_USER))
2643 pr2 = pr2->pr_parent;
2644 sx_sunlock(&allprison_lock);
2645 return (pr1 == pr2);
2646}
2647
2648/*
2649 * Make sure our (source) address is set to something meaningful to this
2650 * jail.
2651 *
2652 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2653 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2654 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2655 */
2656int
2657prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2658{
2659 struct prison *pr;
2660 struct in_addr ia0;
2661 int error;
2662
2663 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2664 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2665
2666 pr = cred->cr_prison;
2667 if (!(pr->pr_flags & PR_IP4))
2668 return (0);
2669 mtx_lock(&pr->pr_mtx);
2670 if (!(pr->pr_flags & PR_IP4)) {
2671 mtx_unlock(&pr->pr_mtx);
2672 return (0);
2673 }
2674 if (pr->pr_ip4 == NULL) {
2675 mtx_unlock(&pr->pr_mtx);
2676 return (EAFNOSUPPORT);
2677 }
2678
2679 ia0.s_addr = ntohl(ia->s_addr);
2680 if (ia0.s_addr == INADDR_LOOPBACK) {
2681 ia->s_addr = pr->pr_ip4[0].s_addr;
2682 mtx_unlock(&pr->pr_mtx);
2683 return (0);
2684 }
2685
2686 if (ia0.s_addr == INADDR_ANY) {
2687 /*
2688 * In case there is only 1 IPv4 address, bind directly.
2689 */
2690 if (pr->pr_ip4s == 1)
2691 ia->s_addr = pr->pr_ip4[0].s_addr;
2692 mtx_unlock(&pr->pr_mtx);
2693 return (0);
2694 }
2695
2696 error = _prison_check_ip4(pr, ia);
2697 mtx_unlock(&pr->pr_mtx);
2698 return (error);
2699}
2700
2701/*
2702 * Rewrite destination address in case we will connect to loopback address.
2703 *
2704 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2705 * Address passed in in NBO and returned in NBO.
2706 */
2707int
2708prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2709{
2710 struct prison *pr;
2711
2712 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2713 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2714
2715 pr = cred->cr_prison;
2716 if (!(pr->pr_flags & PR_IP4))
2717 return (0);
2718 mtx_lock(&pr->pr_mtx);
2719 if (!(pr->pr_flags & PR_IP4)) {
2720 mtx_unlock(&pr->pr_mtx);
2721 return (0);
2722 }
2723 if (pr->pr_ip4 == NULL) {
2724 mtx_unlock(&pr->pr_mtx);
2725 return (EAFNOSUPPORT);
2726 }
2727
2728 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2729 ia->s_addr = pr->pr_ip4[0].s_addr;
2730 mtx_unlock(&pr->pr_mtx);
2731 return (0);
2732 }
2733
2734 /*
2735 * Return success because nothing had to be changed.
2736 */
2737 mtx_unlock(&pr->pr_mtx);
2738 return (0);
2739}
2740
2741/*
2742 * Check if given address belongs to the jail referenced by cred/prison.
2743 *
2744 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2745 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2746 * doesn't allow IPv4. Address passed in in NBO.
2747 */
2748static int
2749_prison_check_ip4(struct prison *pr, struct in_addr *ia)
2750{
2751 int i, a, z, d;
2752
2753 /*
2754 * Check the primary IP.
2755 */
2756 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2757 return (0);
2758
2759 /*
2760 * All the other IPs are sorted so we can do a binary search.
2761 */
2762 a = 0;
2763 z = pr->pr_ip4s - 2;
2764 while (a <= z) {
2765 i = (a + z) / 2;
2766 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2767 if (d > 0)
2768 z = i - 1;
2769 else if (d < 0)
2770 a = i + 1;
2771 else
2772 return (0);
2773 }
2774
2775 return (EADDRNOTAVAIL);
2776}
2777
2778int
2779prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2780{
2781 struct prison *pr;
2782 int error;
2783
2784 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2785 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2786
2787 pr = cred->cr_prison;
2788 if (!(pr->pr_flags & PR_IP4))
2789 return (0);
2790 mtx_lock(&pr->pr_mtx);
2791 if (!(pr->pr_flags & PR_IP4)) {
2792 mtx_unlock(&pr->pr_mtx);
2793 return (0);
2794 }
2795 if (pr->pr_ip4 == NULL) {
2796 mtx_unlock(&pr->pr_mtx);
2797 return (EAFNOSUPPORT);
2798 }
2799
2800 error = _prison_check_ip4(pr, ia);
2801 mtx_unlock(&pr->pr_mtx);
2802 return (error);
2803}
2804#endif
2805
2806#ifdef INET6
2807static int
2808prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2809{
2810 int ii, ij, used;
2811 struct prison *ppr;
2812
2813 ppr = pr->pr_parent;
2814 if (!(pr->pr_flags & PR_IP6_USER)) {
2815 /* This has no user settings, so just copy the parent's list. */
2816 if (pr->pr_ip6s < ppr->pr_ip6s) {
2817 /*
2818 * There's no room for the parent's list. Use the
2819 * new list buffer, which is assumed to be big enough
2820 * (if it was passed). If there's no buffer, try to
2821 * allocate one.
2822 */
2823 used = 1;
2824 if (newip6 == NULL) {
2825 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
2826 M_PRISON, M_NOWAIT);
2827 if (newip6 != NULL)
2828 used = 0;
2829 }
2830 if (newip6 != NULL) {
2831 bcopy(ppr->pr_ip6, newip6,
2832 ppr->pr_ip6s * sizeof(*newip6));
2833 free(pr->pr_ip6, M_PRISON);
2834 pr->pr_ip6 = newip6;
2835 pr->pr_ip6s = ppr->pr_ip6s;
2836 pr->pr_flags |= PR_IP6;
2837 }
2838 return (used);
2839 }
2840 pr->pr_ip6s = ppr->pr_ip6s;
2841 if (pr->pr_ip6s > 0)
2842 bcopy(ppr->pr_ip6, pr->pr_ip6,
2843 pr->pr_ip6s * sizeof(*newip6));
2844 else if (pr->pr_ip6 != NULL) {
2845 free(pr->pr_ip6, M_PRISON);
2846 pr->pr_ip6 = NULL;
2847 }
2848 pr->pr_flags =
2849 (pr->pr_flags & ~PR_IP6) | (ppr->pr_flags & PR_IP6);
2850 } else if (pr->pr_ip6s > 0 && (ppr->pr_flags & PR_IP6)) {
2851 /* Remove addresses that aren't in the parent. */
2852 for (ij = 0; ij < ppr->pr_ip6s; ij++)
2853 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
2854 &ppr->pr_ip6[ij]))
2855 break;
2856 if (ij < ppr->pr_ip6s)
2857 ii = 1;
2858 else {
2859 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
2860 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
2861 ii = 0;
2862 }
2863 for (ij = 1; ii < pr->pr_ip6s; ) {
2864 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
2865 &ppr->pr_ip6[0])) {
2866 ii++;
2867 continue;
2868 }
2869 switch (ij >= ppr->pr_ip4s ? -1 :
2870 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
2871 case -1:
2872 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
2873 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
2874 break;
2875 case 0:
2876 ii++;
2877 ij++;
2878 break;
2879 case 1:
2880 ij++;
2881 break;
2882 }
2883 }
2884 if (pr->pr_ip6s == 0) {
2885 free(pr->pr_ip6, M_PRISON);
2886 pr->pr_ip6 = NULL;
2887 }
2888 }
2889 return 0;
2890}
2891
2892/*
2893 * Pass back primary IPv6 address for this jail.
2894 *
2895 * If not restricted return success but do not alter the address. Caller has
2896 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
2897 *
2898 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
2899 */
2900int
2901prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
2902{
2903 struct prison *pr;
2904
2905 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2906 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
2907
2908 pr = cred->cr_prison;
2909 if (!(pr->pr_flags & PR_IP6))
2910 return (0);
2911 mtx_lock(&pr->pr_mtx);
2912 if (!(pr->pr_flags & PR_IP6)) {
2913 mtx_unlock(&pr->pr_mtx);
2914 return (0);
2915 }
2916 if (pr->pr_ip6 == NULL) {
2917 mtx_unlock(&pr->pr_mtx);
2918 return (EAFNOSUPPORT);
2919 }
2920
2921 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2922 mtx_unlock(&pr->pr_mtx);
2923 return (0);
2924}
2925
2926/*
2927 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
2928 */
2929int
2930prison_equal_ip6(struct prison *pr1, struct prison *pr2)
2931{
2932
2933 if (pr1 == pr2)
2934 return (1);
2935
2936 sx_slock(&allprison_lock);
2937 while (pr1 != &prison0 && !(pr1->pr_flags & PR_IP6_USER))
2938 pr1 = pr1->pr_parent;
2939 while (pr2 != &prison0 && !(pr2->pr_flags & PR_IP6_USER))
2940 pr2 = pr2->pr_parent;
2941 sx_sunlock(&allprison_lock);
2942 return (pr1 == pr2);
2943}
2944
2945/*
2946 * Make sure our (source) address is set to something meaningful to this jail.
2947 *
2948 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
2949 * when needed while binding.
2950 *
2951 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
2952 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2953 * doesn't allow IPv6.
2954 */
2955int
2956prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
2957{
2958 struct prison *pr;
2959 int error;
2960
2961 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2962 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
2963
2964 pr = cred->cr_prison;
2965 if (!(pr->pr_flags & PR_IP6))
2966 return (0);
2967 mtx_lock(&pr->pr_mtx);
2968 if (!(pr->pr_flags & PR_IP6)) {
2969 mtx_unlock(&pr->pr_mtx);
2970 return (0);
2971 }
2972 if (pr->pr_ip6 == NULL) {
2973 mtx_unlock(&pr->pr_mtx);
2974 return (EAFNOSUPPORT);
2975 }
2976
2977 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
2978 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2979 mtx_unlock(&pr->pr_mtx);
2980 return (0);
2981 }
2982
2983 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
2984 /*
2985 * In case there is only 1 IPv6 address, and v6only is true,
2986 * then bind directly.
2987 */
2988 if (v6only != 0 && pr->pr_ip6s == 1)
2989 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2990 mtx_unlock(&pr->pr_mtx);
2991 return (0);
2992 }
2993
2994 error = _prison_check_ip6(pr, ia6);
2995 mtx_unlock(&pr->pr_mtx);
2996 return (error);
2997}
2998
2999/*
3000 * Rewrite destination address in case we will connect to loopback address.
3001 *
3002 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3003 */
3004int
3005prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3006{
3007 struct prison *pr;
3008
3009 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3010 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3011
3012 pr = cred->cr_prison;
3013 if (!(pr->pr_flags & PR_IP6))
3014 return (0);
3015 mtx_lock(&pr->pr_mtx);
3016 if (!(pr->pr_flags & PR_IP6)) {
3017 mtx_unlock(&pr->pr_mtx);
3018 return (0);
3019 }
3020 if (pr->pr_ip6 == NULL) {
3021 mtx_unlock(&pr->pr_mtx);
3022 return (EAFNOSUPPORT);
3023 }
3024
3025 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3026 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3027 mtx_unlock(&pr->pr_mtx);
3028 return (0);
3029 }
3030
3031 /*
3032 * Return success because nothing had to be changed.
3033 */
3034 mtx_unlock(&pr->pr_mtx);
3035 return (0);
3036}
3037
3038/*
3039 * Check if given address belongs to the jail referenced by cred/prison.
3040 *
3041 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3042 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3043 * doesn't allow IPv6.
3044 */
3045static int
3046_prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3047{
3048 int i, a, z, d;
3049
3050 /*
3051 * Check the primary IP.
3052 */
3053 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3054 return (0);
3055
3056 /*
3057 * All the other IPs are sorted so we can do a binary search.
3058 */
3059 a = 0;
3060 z = pr->pr_ip6s - 2;
3061 while (a <= z) {
3062 i = (a + z) / 2;
3063 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3064 if (d > 0)
3065 z = i - 1;
3066 else if (d < 0)
3067 a = i + 1;
3068 else
3069 return (0);
3070 }
3071
3072 return (EADDRNOTAVAIL);
3073}
3074
3075int
3076prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3077{
3078 struct prison *pr;
3079 int error;
3080
3081 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3082 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3083
3084 pr = cred->cr_prison;
3085 if (!(pr->pr_flags & PR_IP6))
3086 return (0);
3087 mtx_lock(&pr->pr_mtx);
3088 if (!(pr->pr_flags & PR_IP6)) {
3089 mtx_unlock(&pr->pr_mtx);
3090 return (0);
3091 }
3092 if (pr->pr_ip6 == NULL) {
3093 mtx_unlock(&pr->pr_mtx);
3094 return (EAFNOSUPPORT);
3095 }
3096
3097 error = _prison_check_ip6(pr, ia6);
3098 mtx_unlock(&pr->pr_mtx);
3099 return (error);
3100}
3101#endif
3102
3103/*
3104 * Check if a jail supports the given address family.
3105 *
3106 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3107 * if not.
3108 */
3109int
3110prison_check_af(struct ucred *cred, int af)
3111{
3112 struct prison *pr;
3113 int error;
3114
3115 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3116
3117 pr = cred->cr_prison;
3118 error = 0;
3119 switch (af)
3120 {
3121#ifdef INET
3122 case AF_INET:
3123 if (pr->pr_flags & PR_IP4)
3124 {
3125 mtx_lock(&pr->pr_mtx);
3126 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3127 error = EAFNOSUPPORT;
3128 mtx_unlock(&pr->pr_mtx);
3129 }
3130 break;
3131#endif
3132#ifdef INET6
3133 case AF_INET6:
3134 if (pr->pr_flags & PR_IP6)
3135 {
3136 mtx_lock(&pr->pr_mtx);
3137 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3138 error = EAFNOSUPPORT;
3139 mtx_unlock(&pr->pr_mtx);
3140 }
3141 break;
3142#endif
3143 case AF_LOCAL:
3144 case AF_ROUTE:
3145 break;
3146 default:
3147 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3148 error = EAFNOSUPPORT;
3149 }
3150 return (error);
3151}
3152
3153/*
3154 * Check if given address belongs to the jail referenced by cred (wrapper to
3155 * prison_check_ip[46]).
3156 *
3157 * Returns 0 if jail doesn't restrict the address family or if address belongs
3158 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3159 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3160 */
3161int
3162prison_if(struct ucred *cred, struct sockaddr *sa)
3163{
3164#ifdef INET
3165 struct sockaddr_in *sai;
3166#endif
3167#ifdef INET6
3168 struct sockaddr_in6 *sai6;
3169#endif
3170 int error;
3171
3172 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3173 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3174
3175 error = 0;
3176 switch (sa->sa_family)
3177 {
3178#ifdef INET
3179 case AF_INET:
3180 sai = (struct sockaddr_in *)sa;
3181 error = prison_check_ip4(cred, &sai->sin_addr);
3182 break;
3183#endif
3184#ifdef INET6
3185 case AF_INET6:
3186 sai6 = (struct sockaddr_in6 *)sa;
3187 error = prison_check_ip6(cred, &sai6->sin6_addr);
3188 break;
3189#endif
3190 default:
3191 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3192 error = EAFNOSUPPORT;
3193 }
3194 return (error);
3195}
3196
3197/*
3198 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3199 */
3200int
3201prison_check(struct ucred *cred1, struct ucred *cred2)
3202{
3203
3204#ifdef VIMAGE
3205 if (cred2->cr_vimage->v_procg != cred1->cr_vimage->v_procg)
3206 return (ESRCH);
3207#endif
3208 return ((cred1->cr_prison == cred2->cr_prison ||
3209 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3210}
3211
3212/*
3213 * Return 1 if p2 is a child of p1, otherwise 0.
3214 */
3215int
3216prison_ischild(struct prison *pr1, struct prison *pr2)
3217{
3218
3219 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3220 if (pr1 == pr2)
3221 return (1);
3222 return (0);
3223}
3224
3225/*
3226 * Return 1 if the passed credential is in a jail, otherwise 0.
3227 */
3228int
3229jailed(struct ucred *cred)
3230{
3231
3232 return (cred->cr_prison != &prison0);
3233}
3234
3235/*
3236 * Return the correct hostname (domainname, et al) for the passed credential.
3237 */
3238void
3239getcredhostname(struct ucred *cred, char *buf, size_t size)
3240{
3241 struct prison *pr;
3242
3243 /*
3244 * A NULL credential can be used to shortcut to the physical
3245 * system's hostname.
3246 */
3247 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3248 mtx_lock(&pr->pr_mtx);
3249 strlcpy(buf, pr->pr_hostname, size);
3250 mtx_unlock(&pr->pr_mtx);
3251}
3252
3253void
3254getcreddomainname(struct ucred *cred, char *buf, size_t size)
3255{
3256
3257 mtx_lock(&cred->cr_prison->pr_mtx);
3258 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3259 mtx_unlock(&cred->cr_prison->pr_mtx);
3260}
3261
3262void
3263getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3264{
3265
3266 mtx_lock(&cred->cr_prison->pr_mtx);
3267 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3268 mtx_unlock(&cred->cr_prison->pr_mtx);
3269}
3270
3271void
3272getcredhostid(struct ucred *cred, unsigned long *hostid)
3273{
3274
3275 mtx_lock(&cred->cr_prison->pr_mtx);
3276 *hostid = cred->cr_prison->pr_hostid;
3277 mtx_unlock(&cred->cr_prison->pr_mtx);
3278}
3279
3280/*
3281 * Determine whether the subject represented by cred can "see"
3282 * status of a mount point.
3283 * Returns: 0 for permitted, ENOENT otherwise.
3284 * XXX: This function should be called cr_canseemount() and should be
3285 * placed in kern_prot.c.
3286 */
3287int
3288prison_canseemount(struct ucred *cred, struct mount *mp)
3289{
3290 struct prison *pr;
3291 struct statfs *sp;
3292 size_t len;
3293
3294 pr = cred->cr_prison;
3295 if (pr->pr_enforce_statfs == 0)
3296 return (0);
3297 if (pr->pr_root->v_mount == mp)
3298 return (0);
3299 if (pr->pr_enforce_statfs == 2)
3300 return (ENOENT);
3301 /*
3302 * If jail's chroot directory is set to "/" we should be able to see
3303 * all mount-points from inside a jail.
3304 * This is ugly check, but this is the only situation when jail's
3305 * directory ends with '/'.
3306 */
3307 if (strcmp(pr->pr_path, "/") == 0)
3308 return (0);
3309 len = strlen(pr->pr_path);
3310 sp = &mp->mnt_stat;
3311 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3312 return (ENOENT);
3313 /*
3314 * Be sure that we don't have situation where jail's root directory
3315 * is "/some/path" and mount point is "/some/pathpath".
3316 */
3317 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3318 return (ENOENT);
3319 return (0);
3320}
3321
3322void
3323prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3324{
3325 char jpath[MAXPATHLEN];
3326 struct prison *pr;
3327 size_t len;
3328
3329 pr = cred->cr_prison;
3330 if (pr->pr_enforce_statfs == 0)
3331 return;
3332 if (prison_canseemount(cred, mp) != 0) {
3333 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3334 strlcpy(sp->f_mntonname, "[restricted]",
3335 sizeof(sp->f_mntonname));
3336 return;
3337 }
3338 if (pr->pr_root->v_mount == mp) {
3339 /*
3340 * Clear current buffer data, so we are sure nothing from
3341 * the valid path left there.
3342 */
3343 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3344 *sp->f_mntonname = '/';
3345 return;
3346 }
3347 /*
3348 * If jail's chroot directory is set to "/" we should be able to see
3349 * all mount-points from inside a jail.
3350 */
3351 if (strcmp(pr->pr_path, "/") == 0)
3352 return;
3353 len = strlen(pr->pr_path);
3354 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3355 /*
3356 * Clear current buffer data, so we are sure nothing from
3357 * the valid path left there.
3358 */
3359 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3360 if (*jpath == '\0') {
3361 /* Should never happen. */
3362 *sp->f_mntonname = '/';
3363 } else {
3364 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3365 }
3366}
3367
3368/*
3369 * Check with permission for a specific privilege is granted within jail. We
3370 * have a specific list of accepted privileges; the rest are denied.
3371 */
3372int
3373prison_priv_check(struct ucred *cred, int priv)
3374{
3375
3376 if (!jailed(cred))
3377 return (0);
3378
3379 switch (priv) {
3380
3381 /*
3382 * Allow ktrace privileges for root in jail.
3383 */
3384 case PRIV_KTRACE:
3385
3386#if 0
3387 /*
3388 * Allow jailed processes to configure audit identity and
3389 * submit audit records (login, etc). In the future we may
3390 * want to further refine the relationship between audit and
3391 * jail.
3392 */
3393 case PRIV_AUDIT_GETAUDIT:
3394 case PRIV_AUDIT_SETAUDIT:
3395 case PRIV_AUDIT_SUBMIT:
3396#endif
3397
3398 /*
3399 * Allow jailed processes to manipulate process UNIX
3400 * credentials in any way they see fit.
3401 */
3402 case PRIV_CRED_SETUID:
3403 case PRIV_CRED_SETEUID:
3404 case PRIV_CRED_SETGID:
3405 case PRIV_CRED_SETEGID:
3406 case PRIV_CRED_SETGROUPS:
3407 case PRIV_CRED_SETREUID:
3408 case PRIV_CRED_SETREGID:
3409 case PRIV_CRED_SETRESUID:
3410 case PRIV_CRED_SETRESGID:
3411
3412 /*
3413 * Jail implements visibility constraints already, so allow
3414 * jailed root to override uid/gid-based constraints.
3415 */
3416 case PRIV_SEEOTHERGIDS:
3417 case PRIV_SEEOTHERUIDS:
3418
3419 /*
3420 * Jail implements inter-process debugging limits already, so
3421 * allow jailed root various debugging privileges.
3422 */
3423 case PRIV_DEBUG_DIFFCRED:
3424 case PRIV_DEBUG_SUGID:
3425 case PRIV_DEBUG_UNPRIV:
3426
3427 /*
3428 * Allow jail to set various resource limits and login
3429 * properties, and for now, exceed process resource limits.
3430 */
3431 case PRIV_PROC_LIMIT:
3432 case PRIV_PROC_SETLOGIN:
3433 case PRIV_PROC_SETRLIMIT:
3434
3435 /*
3436 * System V and POSIX IPC privileges are granted in jail.
3437 */
3438 case PRIV_IPC_READ:
3439 case PRIV_IPC_WRITE:
3440 case PRIV_IPC_ADMIN:
3441 case PRIV_IPC_MSGSIZE:
3442 case PRIV_MQ_ADMIN:
3443
3444 /*
3445 * Jail operations within a jail work on child jails.
3446 */
3447 case PRIV_JAIL_ATTACH:
3448 case PRIV_JAIL_SET:
3449 case PRIV_JAIL_REMOVE:
3450
3451 /*
3452 * Jail implements its own inter-process limits, so allow
3453 * root processes in jail to change scheduling on other
3454 * processes in the same jail. Likewise for signalling.
3455 */
3456 case PRIV_SCHED_DIFFCRED:
3457 case PRIV_SCHED_CPUSET:
3458 case PRIV_SIGNAL_DIFFCRED:
3459 case PRIV_SIGNAL_SUGID:
3460
3461 /*
3462 * Allow jailed processes to write to sysctls marked as jail
3463 * writable.
3464 */
3465 case PRIV_SYSCTL_WRITEJAIL:
3466
3467 /*
3468 * Allow root in jail to manage a variety of quota
3469 * properties. These should likely be conditional on a
3470 * configuration option.
3471 */
3472 case PRIV_VFS_GETQUOTA:
3473 case PRIV_VFS_SETQUOTA:
3474
3475 /*
3476 * Since Jail relies on chroot() to implement file system
3477 * protections, grant many VFS privileges to root in jail.
3478 * Be careful to exclude mount-related and NFS-related
3479 * privileges.
3480 */
3481 case PRIV_VFS_READ:
3482 case PRIV_VFS_WRITE:
3483 case PRIV_VFS_ADMIN:
3484 case PRIV_VFS_EXEC:
3485 case PRIV_VFS_LOOKUP:
3486 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3487 case PRIV_VFS_CHFLAGS_DEV:
3488 case PRIV_VFS_CHOWN:
3489 case PRIV_VFS_CHROOT:
3490 case PRIV_VFS_RETAINSUGID:
3491 case PRIV_VFS_FCHROOT:
3492 case PRIV_VFS_LINK:
3493 case PRIV_VFS_SETGID:
3494 case PRIV_VFS_STAT:
3495 case PRIV_VFS_STICKYFILE:
3496 return (0);
3497
3498 /*
3499 * Depending on the global setting, allow privilege of
3500 * setting system flags.
3501 */
3502 case PRIV_VFS_SYSFLAGS:
3503 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3504 return (0);
3505 else
3506 return (EPERM);
3507
3508 /*
3509 * Depending on the global setting, allow privilege of
3510 * mounting/unmounting file systems.
3511 */
3512 case PRIV_VFS_MOUNT:
3513 case PRIV_VFS_UNMOUNT:
3514 case PRIV_VFS_MOUNT_NONUSER:
3515 case PRIV_VFS_MOUNT_OWNER:
3516 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT)
3517 return (0);
3518 else
3519 return (EPERM);
3520
3521 /*
3522 * Allow jailed root to bind reserved ports and reuse in-use
3523 * ports.
3524 */
3525 case PRIV_NETINET_RESERVEDPORT:
3526 case PRIV_NETINET_REUSEPORT:
3527 return (0);
3528
3529 /*
3530 * Allow jailed root to set certian IPv4/6 (option) headers.
3531 */
3532 case PRIV_NETINET_SETHDROPTS:
3533 return (0);
3534
3535 /*
3536 * Conditionally allow creating raw sockets in jail.
3537 */
3538 case PRIV_NETINET_RAW:
3539 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3540 return (0);
3541 else
3542 return (EPERM);
3543
3544 /*
3545 * Since jail implements its own visibility limits on netstat
3546 * sysctls, allow getcred. This allows identd to work in
3547 * jail.
3548 */
3549 case PRIV_NETINET_GETCRED:
3550 return (0);
3551
3552 default:
3553 /*
3554 * In all remaining cases, deny the privilege request. This
3555 * includes almost all network privileges, many system
3556 * configuration privileges.
3557 */
3558 return (EPERM);
3559 }
3560}
3561
3562/*
3563 * Return the part of pr2's name that is relative to pr1, or the whole name
3564 * if it does not directly follow.
3565 */
3566
3567char *
3568prison_name(struct prison *pr1, struct prison *pr2)
3569{
3570 char *name;
3571
3572 /* Jails see themselves as "0" (if they see themselves at all). */
3573 if (pr1 == pr2)
3574 return "0";
3575 name = pr2->pr_name;
3576 if (prison_ischild(pr1, pr2)) {
3577 /*
3578 * pr1 isn't locked (and allprison_lock may not be either)
3579 * so its length can't be counted on. But the number of dots
3580 * can be counted on - and counted.
3581 */
3582 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3583 name = strchr(name, '.') + 1;
3584 }
3585 return (name);
3586}
3587
3588/*
3589 * Return the part of pr2's path that is relative to pr1, or the whole path
3590 * if it does not directly follow.
3591 */
3592static char *
3593prison_path(struct prison *pr1, struct prison *pr2)
3594{
3595 char *path1, *path2;
3596 int len1;
3597
3598 path1 = pr1->pr_path;
3599 path2 = pr2->pr_path;
3600 if (!strcmp(path1, "/"))
3601 return (path2);
3602 len1 = strlen(path1);
3603 if (strncmp(path1, path2, len1))
3604 return (path2);
3605 if (path2[len1] == '\0')
3606 return "/";
3607 if (path2[len1] == '/')
3608 return (path2 + len1);
3609 return (path2);
3610}
3611
3612
3613/*
3614 * Jail-related sysctls.
3615 */
3616SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3617 "Jails");
3618
3619static int
3620sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3621{
3622 struct xprison *xp;
3623 struct prison *pr, *cpr;
3624#ifdef INET
3625 struct in_addr *ip4 = NULL;
3626 int ip4s = 0;
3627#endif
3628#ifdef INET6
3629 struct in_addr *ip6 = NULL;
3630 int ip6s = 0;
3631#endif
3632 int descend, error;
3633
3634 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3635 pr = req->td->td_ucred->cr_prison;
3636 error = 0;
3637 sx_slock(&allprison_lock);
3638 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3639#if defined(INET) || defined(INET6)
3640 again:
3641#endif
3642 mtx_lock(&cpr->pr_mtx);
3643#ifdef INET
3644 if (cpr->pr_ip4s > 0) {
3645 if (ip4s < cpr->pr_ip4s) {
3646 ip4s = cpr->pr_ip4s;
3647 mtx_unlock(&cpr->pr_mtx);
3648 ip4 = realloc(ip4, ip4s *
3649 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3650 goto again;
3651 }
3652 bcopy(cpr->pr_ip4, ip4,
3653 cpr->pr_ip4s * sizeof(struct in_addr));
3654 }
3655#endif
3656#ifdef INET6
3657 if (cpr->pr_ip6s > 0) {
3658 if (ip6s < cpr->pr_ip6s) {
3659 ip6s = cpr->pr_ip6s;
3660 mtx_unlock(&cpr->pr_mtx);
3661 ip6 = realloc(ip6, ip6s *
3662 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3663 goto again;
3664 }
3665 bcopy(cpr->pr_ip6, ip6,
3666 cpr->pr_ip6s * sizeof(struct in6_addr));
3667 }
3668#endif
3669 if (cpr->pr_ref == 0) {
3670 mtx_unlock(&cpr->pr_mtx);
3671 continue;
3672 }
3673 bzero(xp, sizeof(*xp));
3674 xp->pr_version = XPRISON_VERSION;
3675 xp->pr_id = cpr->pr_id;
3676 xp->pr_state = cpr->pr_uref > 0
3677 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3678 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3679 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3680 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3681#ifdef INET
3682 xp->pr_ip4s = cpr->pr_ip4s;
3683#endif
3684#ifdef INET6
3685 xp->pr_ip6s = cpr->pr_ip6s;
3686#endif
3687 mtx_unlock(&cpr->pr_mtx);
3688 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3689 if (error)
3690 break;
3691#ifdef INET
3692 if (xp->pr_ip4s > 0) {
3693 error = SYSCTL_OUT(req, ip4,
3694 xp->pr_ip4s * sizeof(struct in_addr));
3695 if (error)
3696 break;
3697 }
3698#endif
3699#ifdef INET6
3700 if (xp->pr_ip6s > 0) {
3701 error = SYSCTL_OUT(req, ip6,
3702 xp->pr_ip6s * sizeof(struct in6_addr));
3703 if (error)
3704 break;
3705 }
3706#endif
3707 }
3708 sx_sunlock(&allprison_lock);
3709 free(xp, M_TEMP);
3710#ifdef INET
3711 free(ip4, M_TEMP);
3712#endif
3713#ifdef INET6
3714 free(ip6, M_TEMP);
3715#endif
3716 return (error);
3717}
3718
3719SYSCTL_OID(_security_jail, OID_AUTO, list,
3720 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3721 sysctl_jail_list, "S", "List of active jails");
3722
3723static int
3724sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3725{
3726 int error, injail;
3727
3728 injail = jailed(req->td->td_ucred);
3729 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3730
3731 return (error);
3732}
3733
3734SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3735 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3736 sysctl_jail_jailed, "I", "Process in jail?");
3737
3738#if defined(INET) || defined(INET6)
3739SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3740 &jail_max_af_ips, 0,
3741 "Number of IP addresses a jail may have at most per address family");
3742#endif
3743
3744/*
3745 * Default parameters for jail(2) compatability. For historical reasons,
3746 * the sysctl names have varying similarity to the parameter names. Prisons
3747 * just see their own parameters, and can't change them.
3748 */
3749static int
3750sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3751{
3752 struct prison *pr;
3753 int allow, error, i;
3754
3755 pr = req->td->td_ucred->cr_prison;
3756 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3757
3758 /* Get the current flag value, and convert it to a boolean. */
3759 i = (allow & arg2) ? 1 : 0;
3760 if (arg1 != NULL)
3761 i = !i;
3762 error = sysctl_handle_int(oidp, &i, 0, req);
3763 if (error || !req->newptr)
3764 return (error);
3765 i = i ? arg2 : 0;
3766 if (arg1 != NULL)
3767 i ^= arg2;
3768 /*
3769 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3770 * for writing.
3771 */
3772 mtx_lock(&prison0.pr_mtx);
3773 jail_default_allow = (jail_default_allow & ~arg2) | i;
3774 mtx_unlock(&prison0.pr_mtx);
3775 return (0);
3776}
3777
3778SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3779 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3780 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3781 "Processes in jail can set their hostnames");
3782SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3783 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3784 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3785 "Processes in jail are limited to creating UNIX/IP/route sockets only");
3786SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3787 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3788 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3789 "Processes in jail can use System V IPC primitives");
3790SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3791 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3792 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3793 "Prison root can create raw sockets");
3794SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3795 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3796 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3797 "Processes in jail can alter system file flags");
3798SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3799 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3800 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3801 "Processes in jail can mount/unmount jail-friendly file systems");
3802
3803static int
3804sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3805{
3806 struct prison *pr;
3807 int level, error;
3808
3809 pr = req->td->td_ucred->cr_prison;
3810 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
3811 error = sysctl_handle_int(oidp, &level, 0, req);
3812 if (error || !req->newptr)
3813 return (error);
3814 *(int *)arg1 = level;
3815 return (0);
3816}
3817
3818SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
3819 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3820 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
3821 sysctl_jail_default_level, "I",
3822 "Processes in jail cannot see all mounted file systems");
3823
3824/*
3825 * Nodes to describe jail parameters. Maximum length of string parameters
3826 * is returned in the string itself, and the other parameters exist merely
3827 * to make themselves and their types known.
3828 */
3829SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
3830 "Jail parameters");
3831
3832int
3833sysctl_jail_param(SYSCTL_HANDLER_ARGS)
3834{
3835 int i;
3836 long l;
3837 size_t s;
3838 char numbuf[12];
3839
3840 switch (oidp->oid_kind & CTLTYPE)
3841 {
3842 case CTLTYPE_LONG:
3843 case CTLTYPE_ULONG:
3844 l = 0;
3845#ifdef SCTL_MASK32
3846 if (!(req->flags & SCTL_MASK32))
3847#endif
3848 return (SYSCTL_OUT(req, &l, sizeof(l)));
3849 case CTLTYPE_INT:
3850 case CTLTYPE_UINT:
3851 i = 0;
3852 return (SYSCTL_OUT(req, &i, sizeof(i)));
3853 case CTLTYPE_STRING:
3854 snprintf(numbuf, sizeof(numbuf), "%d", arg2);
3855 return
3856 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
3857 case CTLTYPE_STRUCT:
3858 s = (size_t)arg2;
3859 return (SYSCTL_OUT(req, &s, sizeof(s)));
3860 }
3861 return (0);
3862}
3863
3864SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
3865SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
3866SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
3867SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
3868SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
3869 "I", "Jail secure level");
3870SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
3871 "I", "Jail cannot see all mounted file systems");
3872SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
3873 "B", "Jail persistence");
|
| 3874#ifdef VIMAGE
3875SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
3876 "B", "Virtual network stack");
3877#endif
|
3852SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3853 "B", "Jail is in the process of shutting down");
3854
3855SYSCTL_JAIL_PARAM_NODE(host, "Jail host info");
3856SYSCTL_JAIL_PARAM(, nohost, CTLTYPE_INT | CTLFLAG_RW,
3857 "BN", "Jail w/ no host info");
3858SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3859 "Jail hostname");
3860SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3861 "Jail NIS domainname");
3862SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3863 "Jail host UUID");
3864SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3865 "LU", "Jail host ID");
3866
3867SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3868SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3869
3870#ifdef INET
3871SYSCTL_JAIL_PARAM_NODE(ip4, "Jail IPv4 address virtualization");
3872SYSCTL_JAIL_PARAM(, noip4, CTLTYPE_INT | CTLFLAG_RW,
3873 "BN", "Jail w/ no IP address virtualization");
3874SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3875 "S,in_addr,a", "Jail IPv4 addresses");
3876#endif
3877#ifdef INET6
3878SYSCTL_JAIL_PARAM_NODE(ip6, "Jail IPv6 address virtualization");
3879SYSCTL_JAIL_PARAM(, noip6, CTLTYPE_INT | CTLFLAG_RW,
3880 "BN", "Jail w/ no IP address virtualization");
3881SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3882 "S,in6_addr,a", "Jail IPv6 addresses");
3883#endif
3884
3885SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3886SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3887 "B", "Jail may set hostname");
3888SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3889 "B", "Jail may use SYSV IPC");
3890SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3891 "B", "Jail may create raw sockets");
3892SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3893 "B", "Jail may alter system file flags");
3894SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
3895 "B", "Jail may mount/unmount jail-friendly file systems");
3896SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3897 "B", "Jail may set file quotas");
3898SYSCTL_JAIL_PARAM(_allow, jails, CTLTYPE_INT | CTLFLAG_RW,
3899 "B", "Jail may create child jails");
3900SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3901 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3902
3903
3904#ifdef DDB
3905
3906static void
3907db_show_prison(struct prison *pr)
3908{
3909 int fi;
3910#if defined(INET) || defined(INET6)
3911 int ii;
3912#endif
3913#ifdef INET6
3914 char ip6buf[INET6_ADDRSTRLEN];
3915#endif
3916
3917 db_printf("prison %p:\n", pr);
3918 db_printf(" jid = %d\n", pr->pr_id);
3919 db_printf(" name = %s\n", pr->pr_name);
3920 db_printf(" parent = %p\n", pr->pr_parent);
3921 db_printf(" ref = %d\n", pr->pr_ref);
3922 db_printf(" uref = %d\n", pr->pr_uref);
3923 db_printf(" path = %s\n", pr->pr_path);
3924 db_printf(" cpuset = %d\n", pr->pr_cpuset
3925 ? pr->pr_cpuset->cs_id : -1);
| 3878SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
3879 "B", "Jail is in the process of shutting down");
3880
3881SYSCTL_JAIL_PARAM_NODE(host, "Jail host info");
3882SYSCTL_JAIL_PARAM(, nohost, CTLTYPE_INT | CTLFLAG_RW,
3883 "BN", "Jail w/ no host info");
3884SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
3885 "Jail hostname");
3886SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
3887 "Jail NIS domainname");
3888SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
3889 "Jail host UUID");
3890SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
3891 "LU", "Jail host ID");
3892
3893SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
3894SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
3895
3896#ifdef INET
3897SYSCTL_JAIL_PARAM_NODE(ip4, "Jail IPv4 address virtualization");
3898SYSCTL_JAIL_PARAM(, noip4, CTLTYPE_INT | CTLFLAG_RW,
3899 "BN", "Jail w/ no IP address virtualization");
3900SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
3901 "S,in_addr,a", "Jail IPv4 addresses");
3902#endif
3903#ifdef INET6
3904SYSCTL_JAIL_PARAM_NODE(ip6, "Jail IPv6 address virtualization");
3905SYSCTL_JAIL_PARAM(, noip6, CTLTYPE_INT | CTLFLAG_RW,
3906 "BN", "Jail w/ no IP address virtualization");
3907SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
3908 "S,in6_addr,a", "Jail IPv6 addresses");
3909#endif
3910
3911SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
3912SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
3913 "B", "Jail may set hostname");
3914SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
3915 "B", "Jail may use SYSV IPC");
3916SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
3917 "B", "Jail may create raw sockets");
3918SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
3919 "B", "Jail may alter system file flags");
3920SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
3921 "B", "Jail may mount/unmount jail-friendly file systems");
3922SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
3923 "B", "Jail may set file quotas");
3924SYSCTL_JAIL_PARAM(_allow, jails, CTLTYPE_INT | CTLFLAG_RW,
3925 "B", "Jail may create child jails");
3926SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
3927 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
3928
3929
3930#ifdef DDB
3931
3932static void
3933db_show_prison(struct prison *pr)
3934{
3935 int fi;
3936#if defined(INET) || defined(INET6)
3937 int ii;
3938#endif
3939#ifdef INET6
3940 char ip6buf[INET6_ADDRSTRLEN];
3941#endif
3942
3943 db_printf("prison %p:\n", pr);
3944 db_printf(" jid = %d\n", pr->pr_id);
3945 db_printf(" name = %s\n", pr->pr_name);
3946 db_printf(" parent = %p\n", pr->pr_parent);
3947 db_printf(" ref = %d\n", pr->pr_ref);
3948 db_printf(" uref = %d\n", pr->pr_uref);
3949 db_printf(" path = %s\n", pr->pr_path);
3950 db_printf(" cpuset = %d\n", pr->pr_cpuset
3951 ? pr->pr_cpuset->cs_id : -1);
|
| 3952#ifdef VIMAGE
3953 db_printf(" vnet = %p\n", pr->pr_vnet);
3954#endif
|
3926 db_printf(" root = %p\n", pr->pr_root);
3927 db_printf(" securelevel = %d\n", pr->pr_securelevel);
3928 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
3929 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
3930 db_printf(" flags = %x", pr->pr_flags);
3931 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
3932 fi++)
3933 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
3934 db_printf(" %s", pr_flag_names[fi]);
3935 db_printf(" allow = %x", pr->pr_allow);
3936 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
3937 fi++)
3938 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
3939 db_printf(" %s", pr_allow_names[fi]);
3940 db_printf("\n");
3941 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
3942 db_printf(" host.hostname = %s\n", pr->pr_hostname);
3943 db_printf(" host.domainname = %s\n", pr->pr_domainname);
3944 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
3945 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
3946#ifdef INET
3947 db_printf(" ip4s = %d\n", pr->pr_ip4s);
3948 for (ii = 0; ii < pr->pr_ip4s; ii++)
3949 db_printf(" %s %s\n",
3950 ii == 0 ? "ip4 =" : " ",
3951 inet_ntoa(pr->pr_ip4[ii]));
3952#endif
3953#ifdef INET6
3954 db_printf(" ip6s = %d\n", pr->pr_ip6s);
3955 for (ii = 0; ii < pr->pr_ip6s; ii++)
3956 db_printf(" %s %s\n",
3957 ii == 0 ? "ip6 =" : " ",
3958 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
3959#endif
3960}
3961
3962DB_SHOW_COMMAND(prison, db_show_prison_command)
3963{
3964 struct prison *pr;
3965
3966 if (!have_addr) {
3967 /*
3968 * Show all prisons in the list, and prison0 which is not
3969 * listed.
3970 */
3971 db_show_prison(&prison0);
3972 if (!db_pager_quit) {
3973 TAILQ_FOREACH(pr, &allprison, pr_list) {
3974 db_show_prison(pr);
3975 if (db_pager_quit)
3976 break;
3977 }
3978 }
3979 return;
3980 }
3981
3982 if (addr == 0)
3983 pr = &prison0;
3984 else {
3985 /* Look for a prison with the ID and with references. */
3986 TAILQ_FOREACH(pr, &allprison, pr_list)
3987 if (pr->pr_id == addr && pr->pr_ref > 0)
3988 break;
3989 if (pr == NULL)
3990 /* Look again, without requiring a reference. */
3991 TAILQ_FOREACH(pr, &allprison, pr_list)
3992 if (pr->pr_id == addr)
3993 break;
3994 if (pr == NULL)
3995 /* Assume address points to a valid prison. */
3996 pr = (struct prison *)addr;
3997 }
3998 db_show_prison(pr);
3999}
4000
4001#endif /* DDB */
| 3955 db_printf(" root = %p\n", pr->pr_root);
3956 db_printf(" securelevel = %d\n", pr->pr_securelevel);
3957 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
3958 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
3959 db_printf(" flags = %x", pr->pr_flags);
3960 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
3961 fi++)
3962 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
3963 db_printf(" %s", pr_flag_names[fi]);
3964 db_printf(" allow = %x", pr->pr_allow);
3965 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
3966 fi++)
3967 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
3968 db_printf(" %s", pr_allow_names[fi]);
3969 db_printf("\n");
3970 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
3971 db_printf(" host.hostname = %s\n", pr->pr_hostname);
3972 db_printf(" host.domainname = %s\n", pr->pr_domainname);
3973 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
3974 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
3975#ifdef INET
3976 db_printf(" ip4s = %d\n", pr->pr_ip4s);
3977 for (ii = 0; ii < pr->pr_ip4s; ii++)
3978 db_printf(" %s %s\n",
3979 ii == 0 ? "ip4 =" : " ",
3980 inet_ntoa(pr->pr_ip4[ii]));
3981#endif
3982#ifdef INET6
3983 db_printf(" ip6s = %d\n", pr->pr_ip6s);
3984 for (ii = 0; ii < pr->pr_ip6s; ii++)
3985 db_printf(" %s %s\n",
3986 ii == 0 ? "ip6 =" : " ",
3987 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
3988#endif
3989}
3990
3991DB_SHOW_COMMAND(prison, db_show_prison_command)
3992{
3993 struct prison *pr;
3994
3995 if (!have_addr) {
3996 /*
3997 * Show all prisons in the list, and prison0 which is not
3998 * listed.
3999 */
4000 db_show_prison(&prison0);
4001 if (!db_pager_quit) {
4002 TAILQ_FOREACH(pr, &allprison, pr_list) {
4003 db_show_prison(pr);
4004 if (db_pager_quit)
4005 break;
4006 }
4007 }
4008 return;
4009 }
4010
4011 if (addr == 0)
4012 pr = &prison0;
4013 else {
4014 /* Look for a prison with the ID and with references. */
4015 TAILQ_FOREACH(pr, &allprison, pr_list)
4016 if (pr->pr_id == addr && pr->pr_ref > 0)
4017 break;
4018 if (pr == NULL)
4019 /* Look again, without requiring a reference. */
4020 TAILQ_FOREACH(pr, &allprison, pr_list)
4021 if (pr->pr_id == addr)
4022 break;
4023 if (pr == NULL)
4024 /* Assume address points to a valid prison. */
4025 pr = (struct prison *)addr;
4026 }
4027 db_show_prison(pr);
4028}
4029
4030#endif /* DDB */
|