FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_jail.c
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: releng/8.0/sys/kern/kern_jail.c 196989 2009-09-08 19:18:02Z 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
62 #include <net/if.h>
63 #include <net/vnet.h>
64
65 #include <netinet/in.h>
66
67 #ifdef DDB
68 #include <ddb/ddb.h>
69 #ifdef INET6
70 #include <netinet6/in6_var.h>
71 #endif /* INET6 */
72 #endif /* DDB */
73
74 #include <security/mac/mac_framework.h>
75
76 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
77
78 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
79
80 /* prison0 describes what is "real" about the system. */
81 struct prison prison0 = {
82 .pr_id = 0,
83 .pr_name = "",
84 .pr_ref = 1,
85 .pr_uref = 1,
86 .pr_path = "/",
87 .pr_securelevel = -1,
88 .pr_childmax = JAIL_MAX,
89 .pr_hostuuid = DEFAULT_HOSTUUID,
90 .pr_children = LIST_HEAD_INITIALIZER(&prison0.pr_children),
91 #ifdef VIMAGE
92 .pr_flags = PR_HOST|PR_VNET,
93 #else
94 .pr_flags = PR_HOST,
95 #endif
96 .pr_allow = PR_ALLOW_ALL,
97 };
98 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
99
100 /* allprison and lastprid are protected by allprison_lock. */
101 struct sx allprison_lock;
102 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
103 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
104 int lastprid = 0;
105
106 static int do_jail_attach(struct thread *td, struct prison *pr);
107 static void prison_complete(void *context, int pending);
108 static void prison_deref(struct prison *pr, int flags);
109 static char *prison_path(struct prison *pr1, struct prison *pr2);
110 static void prison_remove_one(struct prison *pr);
111 #ifdef INET
112 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
113 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
114 #endif
115 #ifdef INET6
116 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
117 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
118 #endif
119
120 /* Flags for prison_deref */
121 #define PD_DEREF 0x01
122 #define PD_DEUREF 0x02
123 #define PD_LOCKED 0x04
124 #define PD_LIST_SLOCKED 0x08
125 #define PD_LIST_XLOCKED 0x10
126
127 /*
128 * Parameter names corresponding to PR_* flag values
129 */
130 static char *pr_flag_names[] = {
131 [0] = "persist",
132 };
133
134 static char *pr_flag_nonames[] = {
135 [0] = "nopersist",
136 };
137
138 struct jailsys_flags {
139 const char *name;
140 unsigned disable;
141 unsigned new;
142 } pr_flag_jailsys[] = {
143 { "host", 0, PR_HOST },
144 #ifdef VIMAGE
145 { "vnet", 0, PR_VNET },
146 #endif
147 #ifdef INET
148 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
149 #endif
150 #ifdef INET6
151 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
152 #endif
153 };
154
155 static char *pr_allow_names[] = {
156 "allow.set_hostname",
157 "allow.sysvipc",
158 "allow.raw_sockets",
159 "allow.chflags",
160 "allow.mount",
161 "allow.quotas",
162 "allow.socket_af",
163 };
164
165 static char *pr_allow_nonames[] = {
166 "allow.noset_hostname",
167 "allow.nosysvipc",
168 "allow.noraw_sockets",
169 "allow.nochflags",
170 "allow.nomount",
171 "allow.noquotas",
172 "allow.nosocket_af",
173 };
174
175 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
176 #define JAIL_DEFAULT_ENFORCE_STATFS 2
177 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
178 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
179 #if defined(INET) || defined(INET6)
180 static unsigned jail_max_af_ips = 255;
181 #endif
182
183 #ifdef INET
184 static int
185 qcmp_v4(const void *ip1, const void *ip2)
186 {
187 in_addr_t iaa, iab;
188
189 /*
190 * We need to compare in HBO here to get the list sorted as expected
191 * by the result of the code. Sorting NBO addresses gives you
192 * interesting results. If you do not understand, do not try.
193 */
194 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
195 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
196
197 /*
198 * Do not simply return the difference of the two numbers, the int is
199 * not wide enough.
200 */
201 if (iaa > iab)
202 return (1);
203 else if (iaa < iab)
204 return (-1);
205 else
206 return (0);
207 }
208 #endif
209
210 #ifdef INET6
211 static int
212 qcmp_v6(const void *ip1, const void *ip2)
213 {
214 const struct in6_addr *ia6a, *ia6b;
215 int i, rc;
216
217 ia6a = (const struct in6_addr *)ip1;
218 ia6b = (const struct in6_addr *)ip2;
219
220 rc = 0;
221 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
222 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
223 rc = 1;
224 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
225 rc = -1;
226 }
227 return (rc);
228 }
229 #endif
230
231 /*
232 * struct jail_args {
233 * struct jail *jail;
234 * };
235 */
236 int
237 jail(struct thread *td, struct jail_args *uap)
238 {
239 uint32_t version;
240 int error;
241 struct jail j;
242
243 error = copyin(uap->jail, &version, sizeof(uint32_t));
244 if (error)
245 return (error);
246
247 switch (version) {
248 case 0:
249 {
250 struct jail_v0 j0;
251
252 /* FreeBSD single IPv4 jails. */
253 bzero(&j, sizeof(struct jail));
254 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
255 if (error)
256 return (error);
257 j.version = j0.version;
258 j.path = j0.path;
259 j.hostname = j0.hostname;
260 j.ip4s = j0.ip_number;
261 break;
262 }
263
264 case 1:
265 /*
266 * Version 1 was used by multi-IPv4 jail implementations
267 * that never made it into the official kernel.
268 */
269 return (EINVAL);
270
271 case 2: /* JAIL_API_VERSION */
272 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
273 error = copyin(uap->jail, &j, sizeof(struct jail));
274 if (error)
275 return (error);
276 break;
277
278 default:
279 /* Sci-Fi jails are not supported, sorry. */
280 return (EINVAL);
281 }
282 return (kern_jail(td, &j));
283 }
284
285 int
286 kern_jail(struct thread *td, struct jail *j)
287 {
288 struct iovec optiov[2 * (4
289 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
290 #ifdef INET
291 + 1
292 #endif
293 #ifdef INET6
294 + 1
295 #endif
296 )];
297 struct uio opt;
298 char *u_path, *u_hostname, *u_name;
299 #ifdef INET
300 uint32_t ip4s;
301 struct in_addr *u_ip4;
302 #endif
303 #ifdef INET6
304 struct in6_addr *u_ip6;
305 #endif
306 size_t tmplen;
307 int error, enforce_statfs, fi;
308
309 bzero(&optiov, sizeof(optiov));
310 opt.uio_iov = optiov;
311 opt.uio_iovcnt = 0;
312 opt.uio_offset = -1;
313 opt.uio_resid = -1;
314 opt.uio_segflg = UIO_SYSSPACE;
315 opt.uio_rw = UIO_READ;
316 opt.uio_td = td;
317
318 /* Set permissions for top-level jails from sysctls. */
319 if (!jailed(td->td_ucred)) {
320 for (fi = 0; fi < sizeof(pr_allow_names) /
321 sizeof(pr_allow_names[0]); fi++) {
322 optiov[opt.uio_iovcnt].iov_base =
323 (jail_default_allow & (1 << fi))
324 ? pr_allow_names[fi] : pr_allow_nonames[fi];
325 optiov[opt.uio_iovcnt].iov_len =
326 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
327 opt.uio_iovcnt += 2;
328 }
329 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
330 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
331 opt.uio_iovcnt++;
332 enforce_statfs = jail_default_enforce_statfs;
333 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
334 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
335 opt.uio_iovcnt++;
336 }
337
338 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
339 #ifdef INET
340 ip4s = (j->version == 0) ? 1 : j->ip4s;
341 if (ip4s > jail_max_af_ips)
342 return (EINVAL);
343 tmplen += ip4s * sizeof(struct in_addr);
344 #else
345 if (j->ip4s > 0)
346 return (EINVAL);
347 #endif
348 #ifdef INET6
349 if (j->ip6s > jail_max_af_ips)
350 return (EINVAL);
351 tmplen += j->ip6s * sizeof(struct in6_addr);
352 #else
353 if (j->ip6s > 0)
354 return (EINVAL);
355 #endif
356 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
357 u_hostname = u_path + MAXPATHLEN;
358 u_name = u_hostname + MAXHOSTNAMELEN;
359 #ifdef INET
360 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
361 #endif
362 #ifdef INET6
363 #ifdef INET
364 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
365 #else
366 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
367 #endif
368 #endif
369 optiov[opt.uio_iovcnt].iov_base = "path";
370 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
371 opt.uio_iovcnt++;
372 optiov[opt.uio_iovcnt].iov_base = u_path;
373 error = copyinstr(j->path, u_path, MAXPATHLEN,
374 &optiov[opt.uio_iovcnt].iov_len);
375 if (error) {
376 free(u_path, M_TEMP);
377 return (error);
378 }
379 opt.uio_iovcnt++;
380 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
381 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
382 opt.uio_iovcnt++;
383 optiov[opt.uio_iovcnt].iov_base = u_hostname;
384 error = copyinstr(j->hostname, u_hostname, 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 if (j->jailname != NULL) {
392 optiov[opt.uio_iovcnt].iov_base = "name";
393 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
394 opt.uio_iovcnt++;
395 optiov[opt.uio_iovcnt].iov_base = u_name;
396 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
397 &optiov[opt.uio_iovcnt].iov_len);
398 if (error) {
399 free(u_path, M_TEMP);
400 return (error);
401 }
402 opt.uio_iovcnt++;
403 }
404 #ifdef INET
405 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
406 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
407 opt.uio_iovcnt++;
408 optiov[opt.uio_iovcnt].iov_base = u_ip4;
409 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
410 if (j->version == 0)
411 u_ip4->s_addr = j->ip4s;
412 else {
413 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
414 if (error) {
415 free(u_path, M_TEMP);
416 return (error);
417 }
418 }
419 opt.uio_iovcnt++;
420 #endif
421 #ifdef INET6
422 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
423 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
424 opt.uio_iovcnt++;
425 optiov[opt.uio_iovcnt].iov_base = u_ip6;
426 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
427 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
428 if (error) {
429 free(u_path, M_TEMP);
430 return (error);
431 }
432 opt.uio_iovcnt++;
433 #endif
434 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
435 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
436 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
437 free(u_path, M_TEMP);
438 return (error);
439 }
440
441
442 /*
443 * struct jail_set_args {
444 * struct iovec *iovp;
445 * unsigned int iovcnt;
446 * int flags;
447 * };
448 */
449 int
450 jail_set(struct thread *td, struct jail_set_args *uap)
451 {
452 struct uio *auio;
453 int error;
454
455 /* Check that we have an even number of iovecs. */
456 if (uap->iovcnt & 1)
457 return (EINVAL);
458
459 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
460 if (error)
461 return (error);
462 error = kern_jail_set(td, auio, uap->flags);
463 free(auio, M_IOV);
464 return (error);
465 }
466
467 int
468 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
469 {
470 struct nameidata nd;
471 #ifdef INET
472 struct in_addr *ip4;
473 #endif
474 #ifdef INET6
475 struct in6_addr *ip6;
476 #endif
477 struct vfsopt *opt;
478 struct vfsoptlist *opts;
479 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
480 struct vnode *root;
481 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
482 #if defined(INET) || defined(INET6)
483 struct prison *tppr;
484 void *op;
485 #endif
486 unsigned long hid;
487 size_t namelen, onamelen;
488 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
489 int gotchildmax, gotenforce, gothid, gotslevel;
490 int fi, jid, jsys, len, level;
491 int childmax, slevel, vfslocked;
492 #if defined(INET) || defined(INET6)
493 int ii, ij;
494 #endif
495 #ifdef INET
496 int ip4s, redo_ip4;
497 #endif
498 #ifdef INET6
499 int ip6s, redo_ip6;
500 #endif
501 unsigned pr_flags, ch_flags;
502 unsigned pr_allow, ch_allow, tallow;
503 char numbuf[12];
504
505 error = priv_check(td, PRIV_JAIL_SET);
506 if (!error && (flags & JAIL_ATTACH))
507 error = priv_check(td, PRIV_JAIL_ATTACH);
508 if (error)
509 return (error);
510 mypr = ppr = td->td_ucred->cr_prison;
511 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
512 return (EPERM);
513 if (flags & ~JAIL_SET_MASK)
514 return (EINVAL);
515
516 /*
517 * Check all the parameters before committing to anything. Not all
518 * errors can be caught early, but we may as well try. Also, this
519 * takes care of some expensive stuff (path lookup) before getting
520 * the allprison lock.
521 *
522 * XXX Jails are not filesystems, and jail parameters are not mount
523 * options. But it makes more sense to re-use the vfsopt code
524 * than duplicate it under a different name.
525 */
526 error = vfs_buildopts(optuio, &opts);
527 if (error)
528 return (error);
529 #ifdef INET
530 ip4 = NULL;
531 #endif
532 #ifdef INET6
533 ip6 = NULL;
534 #endif
535
536 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
537 if (error == ENOENT)
538 jid = 0;
539 else if (error != 0)
540 goto done_free;
541
542 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
543 if (error == ENOENT)
544 gotslevel = 0;
545 else if (error != 0)
546 goto done_free;
547 else
548 gotslevel = 1;
549
550 error =
551 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
552 if (error == ENOENT)
553 gotchildmax = 0;
554 else if (error != 0)
555 goto done_free;
556 else
557 gotchildmax = 1;
558
559 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
560 gotenforce = (error == 0);
561 if (gotenforce) {
562 if (enforce < 0 || enforce > 2)
563 return (EINVAL);
564 } else if (error != ENOENT)
565 goto done_free;
566
567 pr_flags = ch_flags = 0;
568 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
569 fi++) {
570 if (pr_flag_names[fi] == NULL)
571 continue;
572 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
573 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
574 }
575 ch_flags |= pr_flags;
576 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
577 fi++) {
578 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
579 sizeof(jsys));
580 if (error == ENOENT)
581 continue;
582 if (error != 0)
583 goto done_free;
584 switch (jsys) {
585 case JAIL_SYS_DISABLE:
586 if (!pr_flag_jailsys[fi].disable) {
587 error = EINVAL;
588 goto done_free;
589 }
590 pr_flags |= pr_flag_jailsys[fi].disable;
591 break;
592 case JAIL_SYS_NEW:
593 pr_flags |= pr_flag_jailsys[fi].new;
594 break;
595 case JAIL_SYS_INHERIT:
596 break;
597 default:
598 error = EINVAL;
599 goto done_free;
600 }
601 ch_flags |=
602 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
603 }
604 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
605 && !(pr_flags & PR_PERSIST)) {
606 error = EINVAL;
607 vfs_opterror(opts, "new jail must persist or attach");
608 goto done_errmsg;
609 }
610 #ifdef VIMAGE
611 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
612 error = EINVAL;
613 vfs_opterror(opts, "vnet cannot be changed after creation");
614 goto done_errmsg;
615 }
616 #endif
617 #ifdef INET
618 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
619 error = EINVAL;
620 vfs_opterror(opts, "ip4 cannot be changed after creation");
621 goto done_errmsg;
622 }
623 #endif
624 #ifdef INET6
625 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
626 error = EINVAL;
627 vfs_opterror(opts, "ip6 cannot be changed after creation");
628 goto done_errmsg;
629 }
630 #endif
631
632 pr_allow = ch_allow = 0;
633 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
634 fi++) {
635 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
636 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
637 }
638 ch_allow |= pr_allow;
639
640 error = vfs_getopt(opts, "name", (void **)&name, &len);
641 if (error == ENOENT)
642 name = NULL;
643 else if (error != 0)
644 goto done_free;
645 else {
646 if (len == 0 || name[len - 1] != '\0') {
647 error = EINVAL;
648 goto done_free;
649 }
650 if (len > MAXHOSTNAMELEN) {
651 error = ENAMETOOLONG;
652 goto done_free;
653 }
654 }
655
656 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
657 if (error == ENOENT)
658 host = NULL;
659 else if (error != 0)
660 goto done_free;
661 else {
662 ch_flags |= PR_HOST;
663 pr_flags |= PR_HOST;
664 if (len == 0 || host[len - 1] != '\0') {
665 error = EINVAL;
666 goto done_free;
667 }
668 if (len > MAXHOSTNAMELEN) {
669 error = ENAMETOOLONG;
670 goto done_free;
671 }
672 }
673
674 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
675 if (error == ENOENT)
676 domain = NULL;
677 else if (error != 0)
678 goto done_free;
679 else {
680 ch_flags |= PR_HOST;
681 pr_flags |= PR_HOST;
682 if (len == 0 || domain[len - 1] != '\0') {
683 error = EINVAL;
684 goto done_free;
685 }
686 if (len > MAXHOSTNAMELEN) {
687 error = ENAMETOOLONG;
688 goto done_free;
689 }
690 }
691
692 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
693 if (error == ENOENT)
694 uuid = NULL;
695 else if (error != 0)
696 goto done_free;
697 else {
698 ch_flags |= PR_HOST;
699 pr_flags |= PR_HOST;
700 if (len == 0 || uuid[len - 1] != '\0') {
701 error = EINVAL;
702 goto done_free;
703 }
704 if (len > HOSTUUIDLEN) {
705 error = ENAMETOOLONG;
706 goto done_free;
707 }
708 }
709
710 #ifdef COMPAT_IA32
711 if (td->td_proc->p_sysent->sv_flags & SV_IA32) {
712 uint32_t hid32;
713
714 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
715 hid = hid32;
716 } else
717 #endif
718 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
719 if (error == ENOENT)
720 gothid = 0;
721 else if (error != 0)
722 goto done_free;
723 else {
724 gothid = 1;
725 ch_flags |= PR_HOST;
726 pr_flags |= PR_HOST;
727 }
728
729 #ifdef INET
730 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
731 if (error == ENOENT)
732 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
733 else if (error != 0)
734 goto done_free;
735 else if (ip4s & (sizeof(*ip4) - 1)) {
736 error = EINVAL;
737 goto done_free;
738 } else {
739 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
740 if (ip4s == 0)
741 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
742 else {
743 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
744 ip4s /= sizeof(*ip4);
745 if (ip4s > jail_max_af_ips) {
746 error = EINVAL;
747 vfs_opterror(opts, "too many IPv4 addresses");
748 goto done_errmsg;
749 }
750 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
751 bcopy(op, ip4, ip4s * sizeof(*ip4));
752 /*
753 * IP addresses are all sorted but ip[0] to preserve
754 * the primary IP address as given from userland.
755 * This special IP is used for unbound outgoing
756 * connections as well for "loopback" traffic.
757 */
758 if (ip4s > 1)
759 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
760 /*
761 * Check for duplicate addresses and do some simple
762 * zero and broadcast checks. If users give other bogus
763 * addresses it is their problem.
764 *
765 * We do not have to care about byte order for these
766 * checks so we will do them in NBO.
767 */
768 for (ii = 0; ii < ip4s; ii++) {
769 if (ip4[ii].s_addr == INADDR_ANY ||
770 ip4[ii].s_addr == INADDR_BROADCAST) {
771 error = EINVAL;
772 goto done_free;
773 }
774 if ((ii+1) < ip4s &&
775 (ip4[0].s_addr == ip4[ii+1].s_addr ||
776 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
777 error = EINVAL;
778 goto done_free;
779 }
780 }
781 }
782 }
783 #endif
784
785 #ifdef INET6
786 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
787 if (error == ENOENT)
788 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
789 else if (error != 0)
790 goto done_free;
791 else if (ip6s & (sizeof(*ip6) - 1)) {
792 error = EINVAL;
793 goto done_free;
794 } else {
795 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
796 if (ip6s == 0)
797 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
798 else {
799 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
800 ip6s /= sizeof(*ip6);
801 if (ip6s > jail_max_af_ips) {
802 error = EINVAL;
803 vfs_opterror(opts, "too many IPv6 addresses");
804 goto done_errmsg;
805 }
806 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
807 bcopy(op, ip6, ip6s * sizeof(*ip6));
808 if (ip6s > 1)
809 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
810 for (ii = 0; ii < ip6s; ii++) {
811 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
812 error = EINVAL;
813 goto done_free;
814 }
815 if ((ii+1) < ip6s &&
816 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
817 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
818 {
819 error = EINVAL;
820 goto done_free;
821 }
822 }
823 }
824 }
825 #endif
826
827 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
828 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
829 error = EINVAL;
830 vfs_opterror(opts,
831 "vnet jails cannot have IP address restrictions");
832 goto done_errmsg;
833 }
834 #endif
835
836 root = NULL;
837 error = vfs_getopt(opts, "path", (void **)&path, &len);
838 if (error == ENOENT)
839 path = NULL;
840 else if (error != 0)
841 goto done_free;
842 else {
843 if (flags & JAIL_UPDATE) {
844 error = EINVAL;
845 vfs_opterror(opts,
846 "path cannot be changed after creation");
847 goto done_errmsg;
848 }
849 if (len == 0 || path[len - 1] != '\0') {
850 error = EINVAL;
851 goto done_free;
852 }
853 if (len < 2 || (len == 2 && path[0] == '/'))
854 path = NULL;
855 else {
856 /* Leave room for a real-root full pathname. */
857 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
858 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
859 error = ENAMETOOLONG;
860 goto done_free;
861 }
862 NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
863 path, td);
864 error = namei(&nd);
865 if (error)
866 goto done_free;
867 vfslocked = NDHASGIANT(&nd);
868 root = nd.ni_vp;
869 NDFREE(&nd, NDF_ONLY_PNBUF);
870 if (root->v_type != VDIR) {
871 error = ENOTDIR;
872 vrele(root);
873 VFS_UNLOCK_GIANT(vfslocked);
874 goto done_free;
875 }
876 VFS_UNLOCK_GIANT(vfslocked);
877 }
878 }
879
880 /*
881 * Grab the allprison lock before letting modules check their
882 * parameters. Once we have it, do not let go so we'll have a
883 * consistent view of the OSD list.
884 */
885 sx_xlock(&allprison_lock);
886 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
887 if (error)
888 goto done_unlock_list;
889
890 /* By now, all parameters should have been noted. */
891 TAILQ_FOREACH(opt, opts, link) {
892 if (!opt->seen && strcmp(opt->name, "errmsg")) {
893 error = EINVAL;
894 vfs_opterror(opts, "unknown parameter: %s", opt->name);
895 goto done_unlock_list;
896 }
897 }
898
899 /*
900 * See if we are creating a new record or updating an existing one.
901 * This abuses the file error codes ENOENT and EEXIST.
902 */
903 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
904 if (!cuflags) {
905 error = EINVAL;
906 vfs_opterror(opts, "no valid operation (create or update)");
907 goto done_unlock_list;
908 }
909 pr = NULL;
910 namelc = NULL;
911 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
912 namelc = strrchr(name, '.');
913 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
914 if (*p != '\0')
915 jid = 0;
916 }
917 if (jid != 0) {
918 /*
919 * See if a requested jid already exists. There is an
920 * information leak here if the jid exists but is not within
921 * the caller's jail hierarchy. Jail creators will get EEXIST
922 * even though they cannot see the jail, and CREATE | UPDATE
923 * will return ENOENT which is not normally a valid error.
924 */
925 if (jid < 0) {
926 error = EINVAL;
927 vfs_opterror(opts, "negative jid");
928 goto done_unlock_list;
929 }
930 pr = prison_find(jid);
931 if (pr != NULL) {
932 ppr = pr->pr_parent;
933 /* Create: jid must not exist. */
934 if (cuflags == JAIL_CREATE) {
935 mtx_unlock(&pr->pr_mtx);
936 error = EEXIST;
937 vfs_opterror(opts, "jail %d already exists",
938 jid);
939 goto done_unlock_list;
940 }
941 if (!prison_ischild(mypr, pr)) {
942 mtx_unlock(&pr->pr_mtx);
943 pr = NULL;
944 } else if (pr->pr_uref == 0) {
945 if (!(flags & JAIL_DYING)) {
946 mtx_unlock(&pr->pr_mtx);
947 error = ENOENT;
948 vfs_opterror(opts, "jail %d is dying",
949 jid);
950 goto done_unlock_list;
951 } else if ((flags & JAIL_ATTACH) ||
952 (pr_flags & PR_PERSIST)) {
953 /*
954 * A dying jail might be resurrected
955 * (via attach or persist), but first
956 * it must determine if another jail
957 * has claimed its name. Accomplish
958 * this by implicitly re-setting the
959 * name.
960 */
961 if (name == NULL)
962 name = prison_name(mypr, pr);
963 }
964 }
965 }
966 if (pr == NULL) {
967 /* Update: jid must exist. */
968 if (cuflags == JAIL_UPDATE) {
969 error = ENOENT;
970 vfs_opterror(opts, "jail %d not found", jid);
971 goto done_unlock_list;
972 }
973 }
974 }
975 /*
976 * If the caller provided a name, look for a jail by that name.
977 * This has different semantics for creates and updates keyed by jid
978 * (where the name must not already exist in a different jail),
979 * and updates keyed by the name itself (where the name must exist
980 * because that is the jail being updated).
981 */
982 if (name != NULL) {
983 namelc = strrchr(name, '.');
984 if (namelc == NULL)
985 namelc = name;
986 else {
987 /*
988 * This is a hierarchical name. Split it into the
989 * parent and child names, and make sure the parent
990 * exists or matches an already found jail.
991 */
992 *namelc = '\0';
993 if (pr != NULL) {
994 if (strncmp(name, ppr->pr_name, namelc - name)
995 || ppr->pr_name[namelc - name] != '\0') {
996 mtx_unlock(&pr->pr_mtx);
997 error = EINVAL;
998 vfs_opterror(opts,
999 "cannot change jail's parent");
1000 goto done_unlock_list;
1001 }
1002 } else {
1003 ppr = prison_find_name(mypr, name);
1004 if (ppr == NULL) {
1005 error = ENOENT;
1006 vfs_opterror(opts,
1007 "jail \"%s\" not found", name);
1008 goto done_unlock_list;
1009 }
1010 mtx_unlock(&ppr->pr_mtx);
1011 }
1012 name = ++namelc;
1013 }
1014 if (name[0] != '\0') {
1015 namelen =
1016 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1017 name_again:
1018 deadpr = NULL;
1019 FOREACH_PRISON_CHILD(ppr, tpr) {
1020 if (tpr != pr && tpr->pr_ref > 0 &&
1021 !strcmp(tpr->pr_name + namelen, name)) {
1022 if (pr == NULL &&
1023 cuflags != JAIL_CREATE) {
1024 mtx_lock(&tpr->pr_mtx);
1025 if (tpr->pr_ref > 0) {
1026 /*
1027 * Use this jail
1028 * for updates.
1029 */
1030 if (tpr->pr_uref > 0) {
1031 pr = tpr;
1032 break;
1033 }
1034 deadpr = tpr;
1035 }
1036 mtx_unlock(&tpr->pr_mtx);
1037 } else if (tpr->pr_uref > 0) {
1038 /*
1039 * Create, or update(jid):
1040 * name must not exist in an
1041 * active sibling jail.
1042 */
1043 error = EEXIST;
1044 if (pr != NULL)
1045 mtx_unlock(&pr->pr_mtx);
1046 vfs_opterror(opts,
1047 "jail \"%s\" already exists",
1048 name);
1049 goto done_unlock_list;
1050 }
1051 }
1052 }
1053 /* If no active jail is found, use a dying one. */
1054 if (deadpr != NULL && pr == NULL) {
1055 if (flags & JAIL_DYING) {
1056 mtx_lock(&deadpr->pr_mtx);
1057 if (deadpr->pr_ref == 0) {
1058 mtx_unlock(&deadpr->pr_mtx);
1059 goto name_again;
1060 }
1061 pr = deadpr;
1062 } else if (cuflags == JAIL_UPDATE) {
1063 error = ENOENT;
1064 vfs_opterror(opts,
1065 "jail \"%s\" is dying", name);
1066 goto done_unlock_list;
1067 }
1068 }
1069 /* Update: name must exist if no jid. */
1070 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1071 error = ENOENT;
1072 vfs_opterror(opts, "jail \"%s\" not found",
1073 name);
1074 goto done_unlock_list;
1075 }
1076 }
1077 }
1078 /* Update: must provide a jid or name. */
1079 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1080 error = ENOENT;
1081 vfs_opterror(opts, "update specified no jail");
1082 goto done_unlock_list;
1083 }
1084
1085 /* If there's no prison to update, create a new one and link it in. */
1086 if (pr == NULL) {
1087 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1088 if (tpr->pr_childcount >= tpr->pr_childmax) {
1089 error = EPERM;
1090 vfs_opterror(opts, "prison limit exceeded");
1091 goto done_unlock_list;
1092 }
1093 created = 1;
1094 mtx_lock(&ppr->pr_mtx);
1095 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1096 mtx_unlock(&ppr->pr_mtx);
1097 error = ENOENT;
1098 vfs_opterror(opts, "parent jail went away!");
1099 goto done_unlock_list;
1100 }
1101 ppr->pr_ref++;
1102 ppr->pr_uref++;
1103 mtx_unlock(&ppr->pr_mtx);
1104 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1105 if (jid == 0) {
1106 /* Find the next free jid. */
1107 jid = lastprid + 1;
1108 findnext:
1109 if (jid == JAIL_MAX)
1110 jid = 1;
1111 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1112 if (tpr->pr_id < jid)
1113 continue;
1114 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1115 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1116 break;
1117 }
1118 if (jid == lastprid) {
1119 error = EAGAIN;
1120 vfs_opterror(opts,
1121 "no available jail IDs");
1122 free(pr, M_PRISON);
1123 prison_deref(ppr, PD_DEREF |
1124 PD_DEUREF | PD_LIST_XLOCKED);
1125 goto done_releroot;
1126 }
1127 jid++;
1128 goto findnext;
1129 }
1130 lastprid = jid;
1131 } else {
1132 /*
1133 * The jail already has a jid (that did not yet exist),
1134 * so just find where to insert it.
1135 */
1136 TAILQ_FOREACH(tpr, &allprison, pr_list)
1137 if (tpr->pr_id >= jid) {
1138 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1139 break;
1140 }
1141 }
1142 if (tpr == NULL)
1143 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1144 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1145 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1146 tpr->pr_childcount++;
1147
1148 pr->pr_parent = ppr;
1149 pr->pr_id = jid;
1150
1151 /* Set some default values, and inherit some from the parent. */
1152 if (name == NULL)
1153 name = "";
1154 if (path == NULL) {
1155 path = "/";
1156 root = mypr->pr_root;
1157 vref(root);
1158 }
1159 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1160 pr->pr_flags |= PR_HOST;
1161 #if defined(INET) || defined(INET6)
1162 #ifdef VIMAGE
1163 if (!(pr_flags & PR_VNET))
1164 #endif
1165 {
1166 #ifdef INET
1167 if (!(ch_flags & PR_IP4_USER))
1168 pr->pr_flags |=
1169 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1170 else if (!(pr_flags & PR_IP4_USER)) {
1171 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1172 if (ppr->pr_ip4 != NULL) {
1173 pr->pr_ip4s = ppr->pr_ip4s;
1174 pr->pr_ip4 = malloc(pr->pr_ip4s *
1175 sizeof(struct in_addr), M_PRISON,
1176 M_WAITOK);
1177 bcopy(ppr->pr_ip4, pr->pr_ip4,
1178 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1179 }
1180 }
1181 #endif
1182 #ifdef INET6
1183 if (!(ch_flags & PR_IP6_USER))
1184 pr->pr_flags |=
1185 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1186 else if (!(pr_flags & PR_IP6_USER)) {
1187 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1188 if (ppr->pr_ip6 != NULL) {
1189 pr->pr_ip6s = ppr->pr_ip6s;
1190 pr->pr_ip6 = malloc(pr->pr_ip6s *
1191 sizeof(struct in6_addr), M_PRISON,
1192 M_WAITOK);
1193 bcopy(ppr->pr_ip6, pr->pr_ip6,
1194 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1195 }
1196 }
1197 #endif
1198 }
1199 #endif
1200 pr->pr_securelevel = ppr->pr_securelevel;
1201 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1202 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1203
1204 LIST_INIT(&pr->pr_children);
1205 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1206
1207 #ifdef VIMAGE
1208 /* Allocate a new vnet if specified. */
1209 pr->pr_vnet = (pr_flags & PR_VNET)
1210 ? vnet_alloc() : ppr->pr_vnet;
1211 #endif
1212 /*
1213 * Allocate a dedicated cpuset for each jail.
1214 * Unlike other initial settings, this may return an erorr.
1215 */
1216 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1217 if (error) {
1218 prison_deref(pr, PD_LIST_XLOCKED);
1219 goto done_releroot;
1220 }
1221
1222 mtx_lock(&pr->pr_mtx);
1223 /*
1224 * New prisons do not yet have a reference, because we do not
1225 * want other to see the incomplete prison once the
1226 * allprison_lock is downgraded.
1227 */
1228 } else {
1229 created = 0;
1230 /*
1231 * Grab a reference for existing prisons, to ensure they
1232 * continue to exist for the duration of the call.
1233 */
1234 pr->pr_ref++;
1235 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1236 if ((pr->pr_flags & PR_VNET) &&
1237 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1238 error = EINVAL;
1239 vfs_opterror(opts,
1240 "vnet jails cannot have IP address restrictions");
1241 goto done_deref_locked;
1242 }
1243 #endif
1244 #ifdef INET
1245 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1246 error = EINVAL;
1247 vfs_opterror(opts,
1248 "ip4 cannot be changed after creation");
1249 goto done_deref_locked;
1250 }
1251 #endif
1252 #ifdef INET6
1253 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1254 error = EINVAL;
1255 vfs_opterror(opts,
1256 "ip6 cannot be changed after creation");
1257 goto done_deref_locked;
1258 }
1259 #endif
1260 }
1261
1262 /* Do final error checking before setting anything. */
1263 if (gotslevel) {
1264 if (slevel < ppr->pr_securelevel) {
1265 error = EPERM;
1266 goto done_deref_locked;
1267 }
1268 }
1269 if (gotchildmax) {
1270 if (childmax >= ppr->pr_childmax) {
1271 error = EPERM;
1272 goto done_deref_locked;
1273 }
1274 }
1275 if (gotenforce) {
1276 if (enforce < ppr->pr_enforce_statfs) {
1277 error = EPERM;
1278 goto done_deref_locked;
1279 }
1280 }
1281 #ifdef INET
1282 if (ip4s > 0) {
1283 if (ppr->pr_flags & PR_IP4) {
1284 /*
1285 * Make sure the new set of IP addresses is a
1286 * subset of the parent's list. Don't worry
1287 * about the parent being unlocked, as any
1288 * setting is done with allprison_lock held.
1289 */
1290 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1291 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1292 break;
1293 if (ij == ppr->pr_ip4s) {
1294 error = EPERM;
1295 goto done_deref_locked;
1296 }
1297 if (ip4s > 1) {
1298 for (ii = ij = 1; ii < ip4s; ii++) {
1299 if (ip4[ii].s_addr ==
1300 ppr->pr_ip4[0].s_addr)
1301 continue;
1302 for (; ij < ppr->pr_ip4s; ij++)
1303 if (ip4[ii].s_addr ==
1304 ppr->pr_ip4[ij].s_addr)
1305 break;
1306 if (ij == ppr->pr_ip4s)
1307 break;
1308 }
1309 if (ij == ppr->pr_ip4s) {
1310 error = EPERM;
1311 goto done_deref_locked;
1312 }
1313 }
1314 }
1315 /*
1316 * Check for conflicting IP addresses. We permit them
1317 * if there is no more than one IP on each jail. If
1318 * there is a duplicate on a jail with more than one
1319 * IP stop checking and return error.
1320 */
1321 tppr = ppr;
1322 #ifdef VIMAGE
1323 for (; tppr != &prison0; tppr = tppr->pr_parent)
1324 if (tppr->pr_flags & PR_VNET)
1325 break;
1326 #endif
1327 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1328 if (tpr == pr ||
1329 #ifdef VIMAGE
1330 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1331 #endif
1332 tpr->pr_uref == 0) {
1333 descend = 0;
1334 continue;
1335 }
1336 if (!(tpr->pr_flags & PR_IP4_USER))
1337 continue;
1338 descend = 0;
1339 if (tpr->pr_ip4 == NULL ||
1340 (ip4s == 1 && tpr->pr_ip4s == 1))
1341 continue;
1342 for (ii = 0; ii < ip4s; ii++) {
1343 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1344 error = EADDRINUSE;
1345 vfs_opterror(opts,
1346 "IPv4 addresses clash");
1347 goto done_deref_locked;
1348 }
1349 }
1350 }
1351 }
1352 #endif
1353 #ifdef INET6
1354 if (ip6s > 0) {
1355 if (ppr->pr_flags & PR_IP6) {
1356 /*
1357 * Make sure the new set of IP addresses is a
1358 * subset of the parent's list.
1359 */
1360 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1361 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1362 &ppr->pr_ip6[ij]))
1363 break;
1364 if (ij == ppr->pr_ip6s) {
1365 error = EPERM;
1366 goto done_deref_locked;
1367 }
1368 if (ip6s > 1) {
1369 for (ii = ij = 1; ii < ip6s; ii++) {
1370 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1371 &ppr->pr_ip6[0]))
1372 continue;
1373 for (; ij < ppr->pr_ip6s; ij++)
1374 if (IN6_ARE_ADDR_EQUAL(
1375 &ip6[ii], &ppr->pr_ip6[ij]))
1376 break;
1377 if (ij == ppr->pr_ip6s)
1378 break;
1379 }
1380 if (ij == ppr->pr_ip6s) {
1381 error = EPERM;
1382 goto done_deref_locked;
1383 }
1384 }
1385 }
1386 /* Check for conflicting IP addresses. */
1387 tppr = ppr;
1388 #ifdef VIMAGE
1389 for (; tppr != &prison0; tppr = tppr->pr_parent)
1390 if (tppr->pr_flags & PR_VNET)
1391 break;
1392 #endif
1393 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1394 if (tpr == pr ||
1395 #ifdef VIMAGE
1396 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1397 #endif
1398 tpr->pr_uref == 0) {
1399 descend = 0;
1400 continue;
1401 }
1402 if (!(tpr->pr_flags & PR_IP6_USER))
1403 continue;
1404 descend = 0;
1405 if (tpr->pr_ip6 == NULL ||
1406 (ip6s == 1 && tpr->pr_ip6s == 1))
1407 continue;
1408 for (ii = 0; ii < ip6s; ii++) {
1409 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1410 error = EADDRINUSE;
1411 vfs_opterror(opts,
1412 "IPv6 addresses clash");
1413 goto done_deref_locked;
1414 }
1415 }
1416 }
1417 }
1418 #endif
1419 onamelen = namelen = 0;
1420 if (name != NULL) {
1421 /* Give a default name of the jid. */
1422 if (name[0] == '\0')
1423 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1424 else if (*namelc == '' || (strtoul(namelc, &p, 10) != jid &&
1425 *p == '\0')) {
1426 error = EINVAL;
1427 vfs_opterror(opts,
1428 "name cannot be numeric (unless it is the jid)");
1429 goto done_deref_locked;
1430 }
1431 /*
1432 * Make sure the name isn't too long for the prison or its
1433 * children.
1434 */
1435 onamelen = strlen(pr->pr_name);
1436 namelen = strlen(name);
1437 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1438 error = ENAMETOOLONG;
1439 goto done_deref_locked;
1440 }
1441 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1442 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1443 sizeof(pr->pr_name)) {
1444 error = ENAMETOOLONG;
1445 goto done_deref_locked;
1446 }
1447 }
1448 }
1449 if (pr_allow & ~ppr->pr_allow) {
1450 error = EPERM;
1451 goto done_deref_locked;
1452 }
1453
1454 /* Set the parameters of the prison. */
1455 #ifdef INET
1456 redo_ip4 = 0;
1457 if (pr_flags & PR_IP4_USER) {
1458 pr->pr_flags |= PR_IP4;
1459 free(pr->pr_ip4, M_PRISON);
1460 pr->pr_ip4s = ip4s;
1461 pr->pr_ip4 = ip4;
1462 ip4 = NULL;
1463 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1464 #ifdef VIMAGE
1465 if (tpr->pr_flags & PR_VNET) {
1466 descend = 0;
1467 continue;
1468 }
1469 #endif
1470 if (prison_restrict_ip4(tpr, NULL)) {
1471 redo_ip4 = 1;
1472 descend = 0;
1473 }
1474 }
1475 }
1476 #endif
1477 #ifdef INET6
1478 redo_ip6 = 0;
1479 if (pr_flags & PR_IP6_USER) {
1480 pr->pr_flags |= PR_IP6;
1481 free(pr->pr_ip6, M_PRISON);
1482 pr->pr_ip6s = ip6s;
1483 pr->pr_ip6 = ip6;
1484 ip6 = NULL;
1485 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1486 #ifdef VIMAGE
1487 if (tpr->pr_flags & PR_VNET) {
1488 descend = 0;
1489 continue;
1490 }
1491 #endif
1492 if (prison_restrict_ip6(tpr, NULL)) {
1493 redo_ip6 = 1;
1494 descend = 0;
1495 }
1496 }
1497 }
1498 #endif
1499 if (gotslevel) {
1500 pr->pr_securelevel = slevel;
1501 /* Set all child jails to be at least this level. */
1502 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1503 if (tpr->pr_securelevel < slevel)
1504 tpr->pr_securelevel = slevel;
1505 }
1506 if (gotchildmax) {
1507 pr->pr_childmax = childmax;
1508 /* Set all child jails to under this limit. */
1509 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1510 if (tpr->pr_childmax > childmax - level)
1511 tpr->pr_childmax = childmax > level
1512 ? childmax - level : 0;
1513 }
1514 if (gotenforce) {
1515 pr->pr_enforce_statfs = enforce;
1516 /* Pass this restriction on to the children. */
1517 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1518 if (tpr->pr_enforce_statfs < enforce)
1519 tpr->pr_enforce_statfs = enforce;
1520 }
1521 if (name != NULL) {
1522 if (ppr == &prison0)
1523 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1524 else
1525 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1526 ppr->pr_name, name);
1527 /* Change this component of child names. */
1528 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1529 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1530 strlen(tpr->pr_name + onamelen) + 1);
1531 bcopy(pr->pr_name, tpr->pr_name, namelen);
1532 }
1533 }
1534 if (path != NULL) {
1535 /* Try to keep a real-rooted full pathname. */
1536 if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
1537 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1538 mypr->pr_path, path);
1539 else
1540 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1541 pr->pr_root = root;
1542 }
1543 if (PR_HOST & ch_flags & ~pr_flags) {
1544 if (pr->pr_flags & PR_HOST) {
1545 /*
1546 * Copy the parent's host info. As with pr_ip4 above,
1547 * the lack of a lock on the parent is not a problem;
1548 * it is always set with allprison_lock at least
1549 * shared, and is held exclusively here.
1550 */
1551 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1552 sizeof(pr->pr_hostname));
1553 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1554 sizeof(pr->pr_domainname));
1555 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1556 sizeof(pr->pr_hostuuid));
1557 pr->pr_hostid = pr->pr_parent->pr_hostid;
1558 }
1559 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1560 /* Set this prison, and any descendants without PR_HOST. */
1561 if (host != NULL)
1562 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1563 if (domain != NULL)
1564 strlcpy(pr->pr_domainname, domain,
1565 sizeof(pr->pr_domainname));
1566 if (uuid != NULL)
1567 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1568 if (gothid)
1569 pr->pr_hostid = hid;
1570 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1571 if (tpr->pr_flags & PR_HOST)
1572 descend = 0;
1573 else {
1574 if (host != NULL)
1575 strlcpy(tpr->pr_hostname,
1576 pr->pr_hostname,
1577 sizeof(tpr->pr_hostname));
1578 if (domain != NULL)
1579 strlcpy(tpr->pr_domainname,
1580 pr->pr_domainname,
1581 sizeof(tpr->pr_domainname));
1582 if (uuid != NULL)
1583 strlcpy(tpr->pr_hostuuid,
1584 pr->pr_hostuuid,
1585 sizeof(tpr->pr_hostuuid));
1586 if (gothid)
1587 tpr->pr_hostid = hid;
1588 }
1589 }
1590 }
1591 if ((tallow = ch_allow & ~pr_allow)) {
1592 /* Clear allow bits in all children. */
1593 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1594 tpr->pr_allow &= ~tallow;
1595 }
1596 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1597 /*
1598 * Persistent prisons get an extra reference, and prisons losing their
1599 * persist flag lose that reference. Only do this for existing prisons
1600 * for now, so new ones will remain unseen until after the module
1601 * handlers have completed.
1602 */
1603 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1604 if (pr_flags & PR_PERSIST) {
1605 pr->pr_ref++;
1606 pr->pr_uref++;
1607 } else {
1608 pr->pr_ref--;
1609 pr->pr_uref--;
1610 }
1611 }
1612 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1613 mtx_unlock(&pr->pr_mtx);
1614
1615 /* Locks may have prevented a complete restriction of child IP
1616 * addresses. If so, allocate some more memory and try again.
1617 */
1618 #ifdef INET
1619 while (redo_ip4) {
1620 ip4s = pr->pr_ip4s;
1621 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1622 mtx_lock(&pr->pr_mtx);
1623 redo_ip4 = 0;
1624 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1625 #ifdef VIMAGE
1626 if (tpr->pr_flags & PR_VNET) {
1627 descend = 0;
1628 continue;
1629 }
1630 #endif
1631 if (prison_restrict_ip4(tpr, ip4)) {
1632 if (ip4 != NULL)
1633 ip4 = NULL;
1634 else
1635 redo_ip4 = 1;
1636 }
1637 }
1638 mtx_unlock(&pr->pr_mtx);
1639 }
1640 #endif
1641 #ifdef INET6
1642 while (redo_ip6) {
1643 ip6s = pr->pr_ip6s;
1644 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1645 mtx_lock(&pr->pr_mtx);
1646 redo_ip6 = 0;
1647 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1648 #ifdef VIMAGE
1649 if (tpr->pr_flags & PR_VNET) {
1650 descend = 0;
1651 continue;
1652 }
1653 #endif
1654 if (prison_restrict_ip6(tpr, ip6)) {
1655 if (ip6 != NULL)
1656 ip6 = NULL;
1657 else
1658 redo_ip6 = 1;
1659 }
1660 }
1661 mtx_unlock(&pr->pr_mtx);
1662 }
1663 #endif
1664
1665 /* Let the modules do their work. */
1666 sx_downgrade(&allprison_lock);
1667 if (created) {
1668 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1669 if (error) {
1670 prison_deref(pr, PD_LIST_SLOCKED);
1671 goto done_errmsg;
1672 }
1673 }
1674 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1675 if (error) {
1676 prison_deref(pr, created
1677 ? PD_LIST_SLOCKED
1678 : PD_DEREF | PD_LIST_SLOCKED);
1679 goto done_errmsg;
1680 }
1681
1682 /* Attach this process to the prison if requested. */
1683 if (flags & JAIL_ATTACH) {
1684 mtx_lock(&pr->pr_mtx);
1685 error = do_jail_attach(td, pr);
1686 if (error) {
1687 vfs_opterror(opts, "attach failed");
1688 if (!created)
1689 prison_deref(pr, PD_DEREF);
1690 goto done_errmsg;
1691 }
1692 }
1693
1694 /*
1695 * Now that it is all there, drop the temporary reference from existing
1696 * prisons. Or add a reference to newly created persistent prisons
1697 * (which was not done earlier so that the prison would not be publicly
1698 * visible).
1699 */
1700 if (!created) {
1701 prison_deref(pr, (flags & JAIL_ATTACH)
1702 ? PD_DEREF
1703 : PD_DEREF | PD_LIST_SLOCKED);
1704 } else {
1705 if (pr_flags & PR_PERSIST) {
1706 mtx_lock(&pr->pr_mtx);
1707 pr->pr_ref++;
1708 pr->pr_uref++;
1709 mtx_unlock(&pr->pr_mtx);
1710 }
1711 if (!(flags & JAIL_ATTACH))
1712 sx_sunlock(&allprison_lock);
1713 }
1714 td->td_retval[0] = pr->pr_id;
1715 goto done_errmsg;
1716
1717 done_deref_locked:
1718 prison_deref(pr, created
1719 ? PD_LOCKED | PD_LIST_XLOCKED
1720 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1721 goto done_releroot;
1722 done_unlock_list:
1723 sx_xunlock(&allprison_lock);
1724 done_releroot:
1725 if (root != NULL) {
1726 vfslocked = VFS_LOCK_GIANT(root->v_mount);
1727 vrele(root);
1728 VFS_UNLOCK_GIANT(vfslocked);
1729 }
1730 done_errmsg:
1731 if (error) {
1732 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1733 if (errmsg_len > 0) {
1734 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1735 if (errmsg_pos > 0) {
1736 if (optuio->uio_segflg == UIO_SYSSPACE)
1737 bcopy(errmsg,
1738 optuio->uio_iov[errmsg_pos].iov_base,
1739 errmsg_len);
1740 else
1741 copyout(errmsg,
1742 optuio->uio_iov[errmsg_pos].iov_base,
1743 errmsg_len);
1744 }
1745 }
1746 }
1747 done_free:
1748 #ifdef INET
1749 free(ip4, M_PRISON);
1750 #endif
1751 #ifdef INET6
1752 free(ip6, M_PRISON);
1753 #endif
1754 vfs_freeopts(opts);
1755 return (error);
1756 }
1757
1758
1759 /*
1760 * struct jail_get_args {
1761 * struct iovec *iovp;
1762 * unsigned int iovcnt;
1763 * int flags;
1764 * };
1765 */
1766 int
1767 jail_get(struct thread *td, struct jail_get_args *uap)
1768 {
1769 struct uio *auio;
1770 int error;
1771
1772 /* Check that we have an even number of iovecs. */
1773 if (uap->iovcnt & 1)
1774 return (EINVAL);
1775
1776 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1777 if (error)
1778 return (error);
1779 error = kern_jail_get(td, auio, uap->flags);
1780 if (error == 0)
1781 error = copyout(auio->uio_iov, uap->iovp,
1782 uap->iovcnt * sizeof (struct iovec));
1783 free(auio, M_IOV);
1784 return (error);
1785 }
1786
1787 int
1788 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1789 {
1790 struct prison *pr, *mypr;
1791 struct vfsopt *opt;
1792 struct vfsoptlist *opts;
1793 char *errmsg, *name;
1794 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1795
1796 if (flags & ~JAIL_GET_MASK)
1797 return (EINVAL);
1798
1799 /* Get the parameter list. */
1800 error = vfs_buildopts(optuio, &opts);
1801 if (error)
1802 return (error);
1803 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1804 mypr = td->td_ucred->cr_prison;
1805
1806 /*
1807 * Find the prison specified by one of: lastjid, jid, name.
1808 */
1809 sx_slock(&allprison_lock);
1810 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1811 if (error == 0) {
1812 TAILQ_FOREACH(pr, &allprison, pr_list) {
1813 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1814 mtx_lock(&pr->pr_mtx);
1815 if (pr->pr_ref > 0 &&
1816 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1817 break;
1818 mtx_unlock(&pr->pr_mtx);
1819 }
1820 }
1821 if (pr != NULL)
1822 goto found_prison;
1823 error = ENOENT;
1824 vfs_opterror(opts, "no jail after %d", jid);
1825 goto done_unlock_list;
1826 } else if (error != ENOENT)
1827 goto done_unlock_list;
1828
1829 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1830 if (error == 0) {
1831 if (jid != 0) {
1832 pr = prison_find_child(mypr, jid);
1833 if (pr != NULL) {
1834 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1835 mtx_unlock(&pr->pr_mtx);
1836 error = ENOENT;
1837 vfs_opterror(opts, "jail %d is dying",
1838 jid);
1839 goto done_unlock_list;
1840 }
1841 goto found_prison;
1842 }
1843 error = ENOENT;
1844 vfs_opterror(opts, "jail %d not found", jid);
1845 goto done_unlock_list;
1846 }
1847 } else if (error != ENOENT)
1848 goto done_unlock_list;
1849
1850 error = vfs_getopt(opts, "name", (void **)&name, &len);
1851 if (error == 0) {
1852 if (len == 0 || name[len - 1] != '\0') {
1853 error = EINVAL;
1854 goto done_unlock_list;
1855 }
1856 pr = prison_find_name(mypr, name);
1857 if (pr != NULL) {
1858 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1859 mtx_unlock(&pr->pr_mtx);
1860 error = ENOENT;
1861 vfs_opterror(opts, "jail \"%s\" is dying",
1862 name);
1863 goto done_unlock_list;
1864 }
1865 goto found_prison;
1866 }
1867 error = ENOENT;
1868 vfs_opterror(opts, "jail \"%s\" not found", name);
1869 goto done_unlock_list;
1870 } else if (error != ENOENT)
1871 goto done_unlock_list;
1872
1873 vfs_opterror(opts, "no jail specified");
1874 error = ENOENT;
1875 goto done_unlock_list;
1876
1877 found_prison:
1878 /* Get the parameters of the prison. */
1879 pr->pr_ref++;
1880 locked = PD_LOCKED;
1881 td->td_retval[0] = pr->pr_id;
1882 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
1883 if (error != 0 && error != ENOENT)
1884 goto done_deref;
1885 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
1886 error = vfs_setopt(opts, "parent", &i, sizeof(i));
1887 if (error != 0 && error != ENOENT)
1888 goto done_deref;
1889 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
1890 if (error != 0 && error != ENOENT)
1891 goto done_deref;
1892 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
1893 sizeof(pr->pr_cpuset->cs_id));
1894 if (error != 0 && error != ENOENT)
1895 goto done_deref;
1896 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
1897 if (error != 0 && error != ENOENT)
1898 goto done_deref;
1899 #ifdef INET
1900 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
1901 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1902 if (error != 0 && error != ENOENT)
1903 goto done_deref;
1904 #endif
1905 #ifdef INET6
1906 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
1907 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1908 if (error != 0 && error != ENOENT)
1909 goto done_deref;
1910 #endif
1911 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
1912 sizeof(pr->pr_securelevel));
1913 if (error != 0 && error != ENOENT)
1914 goto done_deref;
1915 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
1916 sizeof(pr->pr_childcount));
1917 if (error != 0 && error != ENOENT)
1918 goto done_deref;
1919 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
1920 sizeof(pr->pr_childmax));
1921 if (error != 0 && error != ENOENT)
1922 goto done_deref;
1923 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
1924 if (error != 0 && error != ENOENT)
1925 goto done_deref;
1926 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
1927 if (error != 0 && error != ENOENT)
1928 goto done_deref;
1929 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
1930 if (error != 0 && error != ENOENT)
1931 goto done_deref;
1932 #ifdef COMPAT_IA32
1933 if (td->td_proc->p_sysent->sv_flags & SV_IA32) {
1934 uint32_t hid32 = pr->pr_hostid;
1935
1936 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
1937 } else
1938 #endif
1939 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
1940 sizeof(pr->pr_hostid));
1941 if (error != 0 && error != ENOENT)
1942 goto done_deref;
1943 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
1944 sizeof(pr->pr_enforce_statfs));
1945 if (error != 0 && error != ENOENT)
1946 goto done_deref;
1947 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
1948 fi++) {
1949 if (pr_flag_names[fi] == NULL)
1950 continue;
1951 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
1952 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
1953 if (error != 0 && error != ENOENT)
1954 goto done_deref;
1955 i = !i;
1956 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
1957 if (error != 0 && error != ENOENT)
1958 goto done_deref;
1959 }
1960 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
1961 fi++) {
1962 i = pr->pr_flags &
1963 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
1964 i = pr_flag_jailsys[fi].disable &&
1965 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
1966 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
1967 : JAIL_SYS_INHERIT;
1968 error =
1969 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
1970 if (error != 0 && error != ENOENT)
1971 goto done_deref;
1972 }
1973 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
1974 fi++) {
1975 if (pr_allow_names[fi] == NULL)
1976 continue;
1977 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
1978 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
1979 if (error != 0 && error != ENOENT)
1980 goto done_deref;
1981 i = !i;
1982 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
1983 if (error != 0 && error != ENOENT)
1984 goto done_deref;
1985 }
1986 i = (pr->pr_uref == 0);
1987 error = vfs_setopt(opts, "dying", &i, sizeof(i));
1988 if (error != 0 && error != ENOENT)
1989 goto done_deref;
1990 i = !i;
1991 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
1992 if (error != 0 && error != ENOENT)
1993 goto done_deref;
1994
1995 /* Get the module parameters. */
1996 mtx_unlock(&pr->pr_mtx);
1997 locked = 0;
1998 error = osd_jail_call(pr, PR_METHOD_GET, opts);
1999 if (error)
2000 goto done_deref;
2001 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2002
2003 /* By now, all parameters should have been noted. */
2004 TAILQ_FOREACH(opt, opts, link) {
2005 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2006 error = EINVAL;
2007 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2008 goto done_errmsg;
2009 }
2010 }
2011
2012 /* Write the fetched parameters back to userspace. */
2013 error = 0;
2014 TAILQ_FOREACH(opt, opts, link) {
2015 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2016 pos = 2 * opt->pos + 1;
2017 optuio->uio_iov[pos].iov_len = opt->len;
2018 if (opt->value != NULL) {
2019 if (optuio->uio_segflg == UIO_SYSSPACE) {
2020 bcopy(opt->value,
2021 optuio->uio_iov[pos].iov_base,
2022 opt->len);
2023 } else {
2024 error = copyout(opt->value,
2025 optuio->uio_iov[pos].iov_base,
2026 opt->len);
2027 if (error)
2028 break;
2029 }
2030 }
2031 }
2032 }
2033 goto done_errmsg;
2034
2035 done_deref:
2036 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2037 goto done_errmsg;
2038
2039 done_unlock_list:
2040 sx_sunlock(&allprison_lock);
2041 done_errmsg:
2042 if (error && errmsg_pos >= 0) {
2043 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2044 errmsg_pos = 2 * errmsg_pos + 1;
2045 if (errmsg_len > 0) {
2046 if (optuio->uio_segflg == UIO_SYSSPACE)
2047 bcopy(errmsg,
2048 optuio->uio_iov[errmsg_pos].iov_base,
2049 errmsg_len);
2050 else
2051 copyout(errmsg,
2052 optuio->uio_iov[errmsg_pos].iov_base,
2053 errmsg_len);
2054 }
2055 }
2056 vfs_freeopts(opts);
2057 return (error);
2058 }
2059
2060
2061 /*
2062 * struct jail_remove_args {
2063 * int jid;
2064 * };
2065 */
2066 int
2067 jail_remove(struct thread *td, struct jail_remove_args *uap)
2068 {
2069 struct prison *pr, *cpr, *lpr, *tpr;
2070 int descend, error;
2071
2072 error = priv_check(td, PRIV_JAIL_REMOVE);
2073 if (error)
2074 return (error);
2075
2076 sx_xlock(&allprison_lock);
2077 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2078 if (pr == NULL) {
2079 sx_xunlock(&allprison_lock);
2080 return (EINVAL);
2081 }
2082
2083 /* Remove all descendants of this prison, then remove this prison. */
2084 pr->pr_ref++;
2085 pr->pr_flags |= PR_REMOVE;
2086 if (!LIST_EMPTY(&pr->pr_children)) {
2087 mtx_unlock(&pr->pr_mtx);
2088 lpr = NULL;
2089 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2090 mtx_lock(&cpr->pr_mtx);
2091 if (cpr->pr_ref > 0) {
2092 tpr = cpr;
2093 cpr->pr_ref++;
2094 cpr->pr_flags |= PR_REMOVE;
2095 } else {
2096 /* Already removed - do not do it again. */
2097 tpr = NULL;
2098 }
2099 mtx_unlock(&cpr->pr_mtx);
2100 if (lpr != NULL) {
2101 mtx_lock(&lpr->pr_mtx);
2102 prison_remove_one(lpr);
2103 sx_xlock(&allprison_lock);
2104 }
2105 lpr = tpr;
2106 }
2107 if (lpr != NULL) {
2108 mtx_lock(&lpr->pr_mtx);
2109 prison_remove_one(lpr);
2110 sx_xlock(&allprison_lock);
2111 }
2112 mtx_lock(&pr->pr_mtx);
2113 }
2114 prison_remove_one(pr);
2115 return (0);
2116 }
2117
2118 static void
2119 prison_remove_one(struct prison *pr)
2120 {
2121 struct proc *p;
2122 int deuref;
2123
2124 /* If the prison was persistent, it is not anymore. */
2125 deuref = 0;
2126 if (pr->pr_flags & PR_PERSIST) {
2127 pr->pr_ref--;
2128 deuref = PD_DEUREF;
2129 pr->pr_flags &= ~PR_PERSIST;
2130 }
2131
2132 /*
2133 * jail_remove added a reference. If that's the only one, remove
2134 * the prison now.
2135 */
2136 KASSERT(pr->pr_ref > 0,
2137 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2138 if (pr->pr_ref == 1) {
2139 prison_deref(pr,
2140 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2141 return;
2142 }
2143
2144 mtx_unlock(&pr->pr_mtx);
2145 sx_xunlock(&allprison_lock);
2146 /*
2147 * Kill all processes unfortunate enough to be attached to this prison.
2148 */
2149 sx_slock(&allproc_lock);
2150 LIST_FOREACH(p, &allproc, p_list) {
2151 PROC_LOCK(p);
2152 if (p->p_state != PRS_NEW && p->p_ucred &&
2153 p->p_ucred->cr_prison == pr)
2154 psignal(p, SIGKILL);
2155 PROC_UNLOCK(p);
2156 }
2157 sx_sunlock(&allproc_lock);
2158 /* Remove the temporary reference added by jail_remove. */
2159 prison_deref(pr, deuref | PD_DEREF);
2160 }
2161
2162
2163 /*
2164 * struct jail_attach_args {
2165 * int jid;
2166 * };
2167 */
2168 int
2169 jail_attach(struct thread *td, struct jail_attach_args *uap)
2170 {
2171 struct prison *pr;
2172 int error;
2173
2174 error = priv_check(td, PRIV_JAIL_ATTACH);
2175 if (error)
2176 return (error);
2177
2178 sx_slock(&allprison_lock);
2179 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2180 if (pr == NULL) {
2181 sx_sunlock(&allprison_lock);
2182 return (EINVAL);
2183 }
2184
2185 /*
2186 * Do not allow a process to attach to a prison that is not
2187 * considered to be "alive".
2188 */
2189 if (pr->pr_uref == 0) {
2190 mtx_unlock(&pr->pr_mtx);
2191 sx_sunlock(&allprison_lock);
2192 return (EINVAL);
2193 }
2194
2195 return (do_jail_attach(td, pr));
2196 }
2197
2198 static int
2199 do_jail_attach(struct thread *td, struct prison *pr)
2200 {
2201 struct prison *ppr;
2202 struct proc *p;
2203 struct ucred *newcred, *oldcred;
2204 int vfslocked, error;
2205
2206 /*
2207 * XXX: Note that there is a slight race here if two threads
2208 * in the same privileged process attempt to attach to two
2209 * different jails at the same time. It is important for
2210 * user processes not to do this, or they might end up with
2211 * a process root from one prison, but attached to the jail
2212 * of another.
2213 */
2214 pr->pr_ref++;
2215 pr->pr_uref++;
2216 mtx_unlock(&pr->pr_mtx);
2217
2218 /* Let modules do whatever they need to prepare for attaching. */
2219 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2220 if (error) {
2221 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2222 return (error);
2223 }
2224 sx_sunlock(&allprison_lock);
2225
2226 /*
2227 * Reparent the newly attached process to this jail.
2228 */
2229 ppr = td->td_ucred->cr_prison;
2230 p = td->td_proc;
2231 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2232 if (error)
2233 goto e_revert_osd;
2234
2235 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2236 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2237 if ((error = change_dir(pr->pr_root, td)) != 0)
2238 goto e_unlock;
2239 #ifdef MAC
2240 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2241 goto e_unlock;
2242 #endif
2243 VOP_UNLOCK(pr->pr_root, 0);
2244 if ((error = change_root(pr->pr_root, td)))
2245 goto e_unlock_giant;
2246 VFS_UNLOCK_GIANT(vfslocked);
2247
2248 newcred = crget();
2249 PROC_LOCK(p);
2250 oldcred = p->p_ucred;
2251 setsugid(p);
2252 crcopy(newcred, oldcred);
2253 newcred->cr_prison = pr;
2254 p->p_ucred = newcred;
2255 PROC_UNLOCK(p);
2256 crfree(oldcred);
2257 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2258 return (0);
2259 e_unlock:
2260 VOP_UNLOCK(pr->pr_root, 0);
2261 e_unlock_giant:
2262 VFS_UNLOCK_GIANT(vfslocked);
2263 e_revert_osd:
2264 /* Tell modules this thread is still in its old jail after all. */
2265 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2266 prison_deref(pr, PD_DEREF | PD_DEUREF);
2267 return (error);
2268 }
2269
2270
2271 /*
2272 * Returns a locked prison instance, or NULL on failure.
2273 */
2274 struct prison *
2275 prison_find(int prid)
2276 {
2277 struct prison *pr;
2278
2279 sx_assert(&allprison_lock, SX_LOCKED);
2280 TAILQ_FOREACH(pr, &allprison, pr_list) {
2281 if (pr->pr_id == prid) {
2282 mtx_lock(&pr->pr_mtx);
2283 if (pr->pr_ref > 0)
2284 return (pr);
2285 mtx_unlock(&pr->pr_mtx);
2286 }
2287 }
2288 return (NULL);
2289 }
2290
2291 /*
2292 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2293 */
2294 struct prison *
2295 prison_find_child(struct prison *mypr, int prid)
2296 {
2297 struct prison *pr;
2298 int descend;
2299
2300 sx_assert(&allprison_lock, SX_LOCKED);
2301 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2302 if (pr->pr_id == prid) {
2303 mtx_lock(&pr->pr_mtx);
2304 if (pr->pr_ref > 0)
2305 return (pr);
2306 mtx_unlock(&pr->pr_mtx);
2307 }
2308 }
2309 return (NULL);
2310 }
2311
2312 /*
2313 * Look for the name relative to mypr. Returns a locked prison or NULL.
2314 */
2315 struct prison *
2316 prison_find_name(struct prison *mypr, const char *name)
2317 {
2318 struct prison *pr, *deadpr;
2319 size_t mylen;
2320 int descend;
2321
2322 sx_assert(&allprison_lock, SX_LOCKED);
2323 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2324 again:
2325 deadpr = NULL;
2326 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2327 if (!strcmp(pr->pr_name + mylen, name)) {
2328 mtx_lock(&pr->pr_mtx);
2329 if (pr->pr_ref > 0) {
2330 if (pr->pr_uref > 0)
2331 return (pr);
2332 deadpr = pr;
2333 }
2334 mtx_unlock(&pr->pr_mtx);
2335 }
2336 }
2337 /* There was no valid prison - perhaps there was a dying one. */
2338 if (deadpr != NULL) {
2339 mtx_lock(&deadpr->pr_mtx);
2340 if (deadpr->pr_ref == 0) {
2341 mtx_unlock(&deadpr->pr_mtx);
2342 goto again;
2343 }
2344 }
2345 return (deadpr);
2346 }
2347
2348 /*
2349 * See if a prison has the specific flag set.
2350 */
2351 int
2352 prison_flag(struct ucred *cred, unsigned flag)
2353 {
2354
2355 /* This is an atomic read, so no locking is necessary. */
2356 return (cred->cr_prison->pr_flags & flag);
2357 }
2358
2359 int
2360 prison_allow(struct ucred *cred, unsigned flag)
2361 {
2362
2363 /* This is an atomic read, so no locking is necessary. */
2364 return (cred->cr_prison->pr_allow & flag);
2365 }
2366
2367 /*
2368 * Remove a prison reference. If that was the last reference, remove the
2369 * prison itself - but not in this context in case there are locks held.
2370 */
2371 void
2372 prison_free_locked(struct prison *pr)
2373 {
2374
2375 mtx_assert(&pr->pr_mtx, MA_OWNED);
2376 pr->pr_ref--;
2377 if (pr->pr_ref == 0) {
2378 mtx_unlock(&pr->pr_mtx);
2379 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2380 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2381 return;
2382 }
2383 mtx_unlock(&pr->pr_mtx);
2384 }
2385
2386 void
2387 prison_free(struct prison *pr)
2388 {
2389
2390 mtx_lock(&pr->pr_mtx);
2391 prison_free_locked(pr);
2392 }
2393
2394 static void
2395 prison_complete(void *context, int pending)
2396 {
2397
2398 prison_deref((struct prison *)context, 0);
2399 }
2400
2401 /*
2402 * Remove a prison reference (usually). This internal version assumes no
2403 * mutexes are held, except perhaps the prison itself. If there are no more
2404 * references, release and delist the prison. On completion, the prison lock
2405 * and the allprison lock are both unlocked.
2406 */
2407 static void
2408 prison_deref(struct prison *pr, int flags)
2409 {
2410 struct prison *ppr, *tpr;
2411 int vfslocked;
2412
2413 if (!(flags & PD_LOCKED))
2414 mtx_lock(&pr->pr_mtx);
2415 /* Decrement the user references in a separate loop. */
2416 if (flags & PD_DEUREF) {
2417 for (tpr = pr;; tpr = tpr->pr_parent) {
2418 if (tpr != pr)
2419 mtx_lock(&tpr->pr_mtx);
2420 if (--tpr->pr_uref > 0)
2421 break;
2422 KASSERT(tpr != &prison0, ("prison0 pr_uref=0"));
2423 mtx_unlock(&tpr->pr_mtx);
2424 }
2425 /* Done if there were only user references to remove. */
2426 if (!(flags & PD_DEREF)) {
2427 mtx_unlock(&tpr->pr_mtx);
2428 if (flags & PD_LIST_SLOCKED)
2429 sx_sunlock(&allprison_lock);
2430 else if (flags & PD_LIST_XLOCKED)
2431 sx_xunlock(&allprison_lock);
2432 return;
2433 }
2434 if (tpr != pr) {
2435 mtx_unlock(&tpr->pr_mtx);
2436 mtx_lock(&pr->pr_mtx);
2437 }
2438 }
2439
2440 for (;;) {
2441 if (flags & PD_DEREF)
2442 pr->pr_ref--;
2443 /* If the prison still has references, nothing else to do. */
2444 if (pr->pr_ref > 0) {
2445 mtx_unlock(&pr->pr_mtx);
2446 if (flags & PD_LIST_SLOCKED)
2447 sx_sunlock(&allprison_lock);
2448 else if (flags & PD_LIST_XLOCKED)
2449 sx_xunlock(&allprison_lock);
2450 return;
2451 }
2452
2453 mtx_unlock(&pr->pr_mtx);
2454 if (flags & PD_LIST_SLOCKED) {
2455 if (!sx_try_upgrade(&allprison_lock)) {
2456 sx_sunlock(&allprison_lock);
2457 sx_xlock(&allprison_lock);
2458 }
2459 } else if (!(flags & PD_LIST_XLOCKED))
2460 sx_xlock(&allprison_lock);
2461
2462 TAILQ_REMOVE(&allprison, pr, pr_list);
2463 LIST_REMOVE(pr, pr_sibling);
2464 ppr = pr->pr_parent;
2465 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2466 tpr->pr_childcount--;
2467 sx_xunlock(&allprison_lock);
2468
2469 #ifdef VIMAGE
2470 if (pr->pr_vnet != ppr->pr_vnet)
2471 vnet_destroy(pr->pr_vnet);
2472 #endif
2473 if (pr->pr_root != NULL) {
2474 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2475 vrele(pr->pr_root);
2476 VFS_UNLOCK_GIANT(vfslocked);
2477 }
2478 mtx_destroy(&pr->pr_mtx);
2479 #ifdef INET
2480 free(pr->pr_ip4, M_PRISON);
2481 #endif
2482 #ifdef INET6
2483 free(pr->pr_ip6, M_PRISON);
2484 #endif
2485 if (pr->pr_cpuset != NULL)
2486 cpuset_rel(pr->pr_cpuset);
2487 osd_jail_exit(pr);
2488 free(pr, M_PRISON);
2489
2490 /* Removing a prison frees a reference on its parent. */
2491 pr = ppr;
2492 mtx_lock(&pr->pr_mtx);
2493 flags = PD_DEREF;
2494 }
2495 }
2496
2497 void
2498 prison_hold_locked(struct prison *pr)
2499 {
2500
2501 mtx_assert(&pr->pr_mtx, MA_OWNED);
2502 KASSERT(pr->pr_ref > 0,
2503 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2504 pr->pr_ref++;
2505 }
2506
2507 void
2508 prison_hold(struct prison *pr)
2509 {
2510
2511 mtx_lock(&pr->pr_mtx);
2512 prison_hold_locked(pr);
2513 mtx_unlock(&pr->pr_mtx);
2514 }
2515
2516 void
2517 prison_proc_hold(struct prison *pr)
2518 {
2519
2520 mtx_lock(&pr->pr_mtx);
2521 KASSERT(pr->pr_uref > 0,
2522 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2523 pr->pr_uref++;
2524 mtx_unlock(&pr->pr_mtx);
2525 }
2526
2527 void
2528 prison_proc_free(struct prison *pr)
2529 {
2530
2531 mtx_lock(&pr->pr_mtx);
2532 KASSERT(pr->pr_uref > 0,
2533 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2534 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2535 }
2536
2537
2538 #ifdef INET
2539 /*
2540 * Restrict a prison's IP address list with its parent's, possibly replacing
2541 * it. Return true if the replacement buffer was used (or would have been).
2542 */
2543 static int
2544 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2545 {
2546 int ii, ij, used;
2547 struct prison *ppr;
2548
2549 ppr = pr->pr_parent;
2550 if (!(pr->pr_flags & PR_IP4_USER)) {
2551 /* This has no user settings, so just copy the parent's list. */
2552 if (pr->pr_ip4s < ppr->pr_ip4s) {
2553 /*
2554 * There's no room for the parent's list. Use the
2555 * new list buffer, which is assumed to be big enough
2556 * (if it was passed). If there's no buffer, try to
2557 * allocate one.
2558 */
2559 used = 1;
2560 if (newip4 == NULL) {
2561 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2562 M_PRISON, M_NOWAIT);
2563 if (newip4 != NULL)
2564 used = 0;
2565 }
2566 if (newip4 != NULL) {
2567 bcopy(ppr->pr_ip4, newip4,
2568 ppr->pr_ip4s * sizeof(*newip4));
2569 free(pr->pr_ip4, M_PRISON);
2570 pr->pr_ip4 = newip4;
2571 pr->pr_ip4s = ppr->pr_ip4s;
2572 }
2573 return (used);
2574 }
2575 pr->pr_ip4s = ppr->pr_ip4s;
2576 if (pr->pr_ip4s > 0)
2577 bcopy(ppr->pr_ip4, pr->pr_ip4,
2578 pr->pr_ip4s * sizeof(*newip4));
2579 else if (pr->pr_ip4 != NULL) {
2580 free(pr->pr_ip4, M_PRISON);
2581 pr->pr_ip4 = NULL;
2582 }
2583 } else if (pr->pr_ip4s > 0) {
2584 /* Remove addresses that aren't in the parent. */
2585 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2586 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2587 break;
2588 if (ij < ppr->pr_ip4s)
2589 ii = 1;
2590 else {
2591 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2592 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2593 ii = 0;
2594 }
2595 for (ij = 1; ii < pr->pr_ip4s; ) {
2596 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2597 ii++;
2598 continue;
2599 }
2600 switch (ij >= ppr->pr_ip4s ? -1 :
2601 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2602 case -1:
2603 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2604 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2605 break;
2606 case 0:
2607 ii++;
2608 ij++;
2609 break;
2610 case 1:
2611 ij++;
2612 break;
2613 }
2614 }
2615 if (pr->pr_ip4s == 0) {
2616 pr->pr_flags |= PR_IP4_DISABLE;
2617 free(pr->pr_ip4, M_PRISON);
2618 pr->pr_ip4 = NULL;
2619 }
2620 }
2621 return (0);
2622 }
2623
2624 /*
2625 * Pass back primary IPv4 address of this jail.
2626 *
2627 * If not restricted return success but do not alter the address. Caller has
2628 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2629 *
2630 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2631 * Address returned in NBO.
2632 */
2633 int
2634 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2635 {
2636 struct prison *pr;
2637
2638 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2639 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2640
2641 pr = cred->cr_prison;
2642 if (!(pr->pr_flags & PR_IP4))
2643 return (0);
2644 mtx_lock(&pr->pr_mtx);
2645 if (!(pr->pr_flags & PR_IP4)) {
2646 mtx_unlock(&pr->pr_mtx);
2647 return (0);
2648 }
2649 if (pr->pr_ip4 == NULL) {
2650 mtx_unlock(&pr->pr_mtx);
2651 return (EAFNOSUPPORT);
2652 }
2653
2654 ia->s_addr = pr->pr_ip4[0].s_addr;
2655 mtx_unlock(&pr->pr_mtx);
2656 return (0);
2657 }
2658
2659 /*
2660 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2661 */
2662 int
2663 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2664 {
2665
2666 if (pr1 == pr2)
2667 return (1);
2668
2669 /*
2670 * No need to lock since the PR_IP4_USER flag can't be altered for
2671 * existing prisons.
2672 */
2673 while (pr1 != &prison0 &&
2674 #ifdef VIMAGE
2675 !(pr1->pr_flags & PR_VNET) &&
2676 #endif
2677 !(pr1->pr_flags & PR_IP4_USER))
2678 pr1 = pr1->pr_parent;
2679 while (pr2 != &prison0 &&
2680 #ifdef VIMAGE
2681 !(pr2->pr_flags & PR_VNET) &&
2682 #endif
2683 !(pr2->pr_flags & PR_IP4_USER))
2684 pr2 = pr2->pr_parent;
2685 return (pr1 == pr2);
2686 }
2687
2688 /*
2689 * Make sure our (source) address is set to something meaningful to this
2690 * jail.
2691 *
2692 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2693 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2694 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2695 */
2696 int
2697 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2698 {
2699 struct prison *pr;
2700 struct in_addr ia0;
2701 int error;
2702
2703 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2704 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2705
2706 pr = cred->cr_prison;
2707 if (!(pr->pr_flags & PR_IP4))
2708 return (0);
2709 mtx_lock(&pr->pr_mtx);
2710 if (!(pr->pr_flags & PR_IP4)) {
2711 mtx_unlock(&pr->pr_mtx);
2712 return (0);
2713 }
2714 if (pr->pr_ip4 == NULL) {
2715 mtx_unlock(&pr->pr_mtx);
2716 return (EAFNOSUPPORT);
2717 }
2718
2719 ia0.s_addr = ntohl(ia->s_addr);
2720 if (ia0.s_addr == INADDR_LOOPBACK) {
2721 ia->s_addr = pr->pr_ip4[0].s_addr;
2722 mtx_unlock(&pr->pr_mtx);
2723 return (0);
2724 }
2725
2726 if (ia0.s_addr == INADDR_ANY) {
2727 /*
2728 * In case there is only 1 IPv4 address, bind directly.
2729 */
2730 if (pr->pr_ip4s == 1)
2731 ia->s_addr = pr->pr_ip4[0].s_addr;
2732 mtx_unlock(&pr->pr_mtx);
2733 return (0);
2734 }
2735
2736 error = _prison_check_ip4(pr, ia);
2737 mtx_unlock(&pr->pr_mtx);
2738 return (error);
2739 }
2740
2741 /*
2742 * Rewrite destination address in case we will connect to loopback address.
2743 *
2744 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2745 * Address passed in in NBO and returned in NBO.
2746 */
2747 int
2748 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2749 {
2750 struct prison *pr;
2751
2752 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2753 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2754
2755 pr = cred->cr_prison;
2756 if (!(pr->pr_flags & PR_IP4))
2757 return (0);
2758 mtx_lock(&pr->pr_mtx);
2759 if (!(pr->pr_flags & PR_IP4)) {
2760 mtx_unlock(&pr->pr_mtx);
2761 return (0);
2762 }
2763 if (pr->pr_ip4 == NULL) {
2764 mtx_unlock(&pr->pr_mtx);
2765 return (EAFNOSUPPORT);
2766 }
2767
2768 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2769 ia->s_addr = pr->pr_ip4[0].s_addr;
2770 mtx_unlock(&pr->pr_mtx);
2771 return (0);
2772 }
2773
2774 /*
2775 * Return success because nothing had to be changed.
2776 */
2777 mtx_unlock(&pr->pr_mtx);
2778 return (0);
2779 }
2780
2781 /*
2782 * Check if given address belongs to the jail referenced by cred/prison.
2783 *
2784 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2785 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2786 * doesn't allow IPv4. Address passed in in NBO.
2787 */
2788 static int
2789 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2790 {
2791 int i, a, z, d;
2792
2793 /*
2794 * Check the primary IP.
2795 */
2796 if (pr->pr_ip4[0].s_addr == ia->s_addr)
2797 return (0);
2798
2799 /*
2800 * All the other IPs are sorted so we can do a binary search.
2801 */
2802 a = 0;
2803 z = pr->pr_ip4s - 2;
2804 while (a <= z) {
2805 i = (a + z) / 2;
2806 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2807 if (d > 0)
2808 z = i - 1;
2809 else if (d < 0)
2810 a = i + 1;
2811 else
2812 return (0);
2813 }
2814
2815 return (EADDRNOTAVAIL);
2816 }
2817
2818 int
2819 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2820 {
2821 struct prison *pr;
2822 int error;
2823
2824 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2825 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2826
2827 pr = cred->cr_prison;
2828 if (!(pr->pr_flags & PR_IP4))
2829 return (0);
2830 mtx_lock(&pr->pr_mtx);
2831 if (!(pr->pr_flags & PR_IP4)) {
2832 mtx_unlock(&pr->pr_mtx);
2833 return (0);
2834 }
2835 if (pr->pr_ip4 == NULL) {
2836 mtx_unlock(&pr->pr_mtx);
2837 return (EAFNOSUPPORT);
2838 }
2839
2840 error = _prison_check_ip4(pr, ia);
2841 mtx_unlock(&pr->pr_mtx);
2842 return (error);
2843 }
2844 #endif
2845
2846 #ifdef INET6
2847 static int
2848 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
2849 {
2850 int ii, ij, used;
2851 struct prison *ppr;
2852
2853 ppr = pr->pr_parent;
2854 if (!(pr->pr_flags & PR_IP6_USER)) {
2855 /* This has no user settings, so just copy the parent's list. */
2856 if (pr->pr_ip6s < ppr->pr_ip6s) {
2857 /*
2858 * There's no room for the parent's list. Use the
2859 * new list buffer, which is assumed to be big enough
2860 * (if it was passed). If there's no buffer, try to
2861 * allocate one.
2862 */
2863 used = 1;
2864 if (newip6 == NULL) {
2865 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
2866 M_PRISON, M_NOWAIT);
2867 if (newip6 != NULL)
2868 used = 0;
2869 }
2870 if (newip6 != NULL) {
2871 bcopy(ppr->pr_ip6, newip6,
2872 ppr->pr_ip6s * sizeof(*newip6));
2873 free(pr->pr_ip6, M_PRISON);
2874 pr->pr_ip6 = newip6;
2875 pr->pr_ip6s = ppr->pr_ip6s;
2876 }
2877 return (used);
2878 }
2879 pr->pr_ip6s = ppr->pr_ip6s;
2880 if (pr->pr_ip6s > 0)
2881 bcopy(ppr->pr_ip6, pr->pr_ip6,
2882 pr->pr_ip6s * sizeof(*newip6));
2883 else if (pr->pr_ip6 != NULL) {
2884 free(pr->pr_ip6, M_PRISON);
2885 pr->pr_ip6 = NULL;
2886 }
2887 } else if (pr->pr_ip6s > 0) {
2888 /* Remove addresses that aren't in the parent. */
2889 for (ij = 0; ij < ppr->pr_ip6s; ij++)
2890 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
2891 &ppr->pr_ip6[ij]))
2892 break;
2893 if (ij < ppr->pr_ip6s)
2894 ii = 1;
2895 else {
2896 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
2897 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
2898 ii = 0;
2899 }
2900 for (ij = 1; ii < pr->pr_ip6s; ) {
2901 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
2902 &ppr->pr_ip6[0])) {
2903 ii++;
2904 continue;
2905 }
2906 switch (ij >= ppr->pr_ip4s ? -1 :
2907 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
2908 case -1:
2909 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
2910 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
2911 break;
2912 case 0:
2913 ii++;
2914 ij++;
2915 break;
2916 case 1:
2917 ij++;
2918 break;
2919 }
2920 }
2921 if (pr->pr_ip6s == 0) {
2922 pr->pr_flags |= PR_IP6_DISABLE;
2923 free(pr->pr_ip6, M_PRISON);
2924 pr->pr_ip6 = NULL;
2925 }
2926 }
2927 return 0;
2928 }
2929
2930 /*
2931 * Pass back primary IPv6 address for this jail.
2932 *
2933 * If not restricted return success but do not alter the address. Caller has
2934 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
2935 *
2936 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
2937 */
2938 int
2939 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
2940 {
2941 struct prison *pr;
2942
2943 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2944 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
2945
2946 pr = cred->cr_prison;
2947 if (!(pr->pr_flags & PR_IP6))
2948 return (0);
2949 mtx_lock(&pr->pr_mtx);
2950 if (!(pr->pr_flags & PR_IP6)) {
2951 mtx_unlock(&pr->pr_mtx);
2952 return (0);
2953 }
2954 if (pr->pr_ip6 == NULL) {
2955 mtx_unlock(&pr->pr_mtx);
2956 return (EAFNOSUPPORT);
2957 }
2958
2959 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
2960 mtx_unlock(&pr->pr_mtx);
2961 return (0);
2962 }
2963
2964 /*
2965 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
2966 */
2967 int
2968 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
2969 {
2970
2971 if (pr1 == pr2)
2972 return (1);
2973
2974 while (pr1 != &prison0 &&
2975 #ifdef VIMAGE
2976 !(pr1->pr_flags & PR_VNET) &&
2977 #endif
2978 !(pr1->pr_flags & PR_IP6_USER))
2979 pr1 = pr1->pr_parent;
2980 while (pr2 != &prison0 &&
2981 #ifdef VIMAGE
2982 !(pr2->pr_flags & PR_VNET) &&
2983 #endif
2984 !(pr2->pr_flags & PR_IP6_USER))
2985 pr2 = pr2->pr_parent;
2986 return (pr1 == pr2);
2987 }
2988
2989 /*
2990 * Make sure our (source) address is set to something meaningful to this jail.
2991 *
2992 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
2993 * when needed while binding.
2994 *
2995 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
2996 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2997 * doesn't allow IPv6.
2998 */
2999 int
3000 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3001 {
3002 struct prison *pr;
3003 int error;
3004
3005 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3006 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3007
3008 pr = cred->cr_prison;
3009 if (!(pr->pr_flags & PR_IP6))
3010 return (0);
3011 mtx_lock(&pr->pr_mtx);
3012 if (!(pr->pr_flags & PR_IP6)) {
3013 mtx_unlock(&pr->pr_mtx);
3014 return (0);
3015 }
3016 if (pr->pr_ip6 == NULL) {
3017 mtx_unlock(&pr->pr_mtx);
3018 return (EAFNOSUPPORT);
3019 }
3020
3021 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3022 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3023 mtx_unlock(&pr->pr_mtx);
3024 return (0);
3025 }
3026
3027 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3028 /*
3029 * In case there is only 1 IPv6 address, and v6only is true,
3030 * then bind directly.
3031 */
3032 if (v6only != 0 && pr->pr_ip6s == 1)
3033 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3034 mtx_unlock(&pr->pr_mtx);
3035 return (0);
3036 }
3037
3038 error = _prison_check_ip6(pr, ia6);
3039 mtx_unlock(&pr->pr_mtx);
3040 return (error);
3041 }
3042
3043 /*
3044 * Rewrite destination address in case we will connect to loopback address.
3045 *
3046 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3047 */
3048 int
3049 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3050 {
3051 struct prison *pr;
3052
3053 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3054 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3055
3056 pr = cred->cr_prison;
3057 if (!(pr->pr_flags & PR_IP6))
3058 return (0);
3059 mtx_lock(&pr->pr_mtx);
3060 if (!(pr->pr_flags & PR_IP6)) {
3061 mtx_unlock(&pr->pr_mtx);
3062 return (0);
3063 }
3064 if (pr->pr_ip6 == NULL) {
3065 mtx_unlock(&pr->pr_mtx);
3066 return (EAFNOSUPPORT);
3067 }
3068
3069 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3070 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3071 mtx_unlock(&pr->pr_mtx);
3072 return (0);
3073 }
3074
3075 /*
3076 * Return success because nothing had to be changed.
3077 */
3078 mtx_unlock(&pr->pr_mtx);
3079 return (0);
3080 }
3081
3082 /*
3083 * Check if given address belongs to the jail referenced by cred/prison.
3084 *
3085 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3086 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3087 * doesn't allow IPv6.
3088 */
3089 static int
3090 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3091 {
3092 int i, a, z, d;
3093
3094 /*
3095 * Check the primary IP.
3096 */
3097 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3098 return (0);
3099
3100 /*
3101 * All the other IPs are sorted so we can do a binary search.
3102 */
3103 a = 0;
3104 z = pr->pr_ip6s - 2;
3105 while (a <= z) {
3106 i = (a + z) / 2;
3107 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3108 if (d > 0)
3109 z = i - 1;
3110 else if (d < 0)
3111 a = i + 1;
3112 else
3113 return (0);
3114 }
3115
3116 return (EADDRNOTAVAIL);
3117 }
3118
3119 int
3120 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3121 {
3122 struct prison *pr;
3123 int error;
3124
3125 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3126 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3127
3128 pr = cred->cr_prison;
3129 if (!(pr->pr_flags & PR_IP6))
3130 return (0);
3131 mtx_lock(&pr->pr_mtx);
3132 if (!(pr->pr_flags & PR_IP6)) {
3133 mtx_unlock(&pr->pr_mtx);
3134 return (0);
3135 }
3136 if (pr->pr_ip6 == NULL) {
3137 mtx_unlock(&pr->pr_mtx);
3138 return (EAFNOSUPPORT);
3139 }
3140
3141 error = _prison_check_ip6(pr, ia6);
3142 mtx_unlock(&pr->pr_mtx);
3143 return (error);
3144 }
3145 #endif
3146
3147 /*
3148 * Check if a jail supports the given address family.
3149 *
3150 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3151 * if not.
3152 */
3153 int
3154 prison_check_af(struct ucred *cred, int af)
3155 {
3156 struct prison *pr;
3157 int error;
3158
3159 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3160
3161 pr = cred->cr_prison;
3162 #ifdef VIMAGE
3163 /* Prisons with their own network stack are not limited. */
3164 if (pr->pr_flags & PR_VNET)
3165 return (0);
3166 #endif
3167
3168 error = 0;
3169 switch (af)
3170 {
3171 #ifdef INET
3172 case AF_INET:
3173 if (pr->pr_flags & PR_IP4)
3174 {
3175 mtx_lock(&pr->pr_mtx);
3176 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3177 error = EAFNOSUPPORT;
3178 mtx_unlock(&pr->pr_mtx);
3179 }
3180 break;
3181 #endif
3182 #ifdef INET6
3183 case AF_INET6:
3184 if (pr->pr_flags & PR_IP6)
3185 {
3186 mtx_lock(&pr->pr_mtx);
3187 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3188 error = EAFNOSUPPORT;
3189 mtx_unlock(&pr->pr_mtx);
3190 }
3191 break;
3192 #endif
3193 case AF_LOCAL:
3194 case AF_ROUTE:
3195 break;
3196 default:
3197 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3198 error = EAFNOSUPPORT;
3199 }
3200 return (error);
3201 }
3202
3203 /*
3204 * Check if given address belongs to the jail referenced by cred (wrapper to
3205 * prison_check_ip[46]).
3206 *
3207 * Returns 0 if jail doesn't restrict the address family or if address belongs
3208 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3209 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3210 */
3211 int
3212 prison_if(struct ucred *cred, struct sockaddr *sa)
3213 {
3214 #ifdef INET
3215 struct sockaddr_in *sai;
3216 #endif
3217 #ifdef INET6
3218 struct sockaddr_in6 *sai6;
3219 #endif
3220 int error;
3221
3222 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3223 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3224
3225 error = 0;
3226 switch (sa->sa_family)
3227 {
3228 #ifdef INET
3229 case AF_INET:
3230 sai = (struct sockaddr_in *)sa;
3231 error = prison_check_ip4(cred, &sai->sin_addr);
3232 break;
3233 #endif
3234 #ifdef INET6
3235 case AF_INET6:
3236 sai6 = (struct sockaddr_in6 *)sa;
3237 error = prison_check_ip6(cred, &sai6->sin6_addr);
3238 break;
3239 #endif
3240 default:
3241 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3242 error = EAFNOSUPPORT;
3243 }
3244 return (error);
3245 }
3246
3247 /*
3248 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3249 */
3250 int
3251 prison_check(struct ucred *cred1, struct ucred *cred2)
3252 {
3253
3254 return ((cred1->cr_prison == cred2->cr_prison ||
3255 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3256 }
3257
3258 /*
3259 * Return 1 if p2 is a child of p1, otherwise 0.
3260 */
3261 int
3262 prison_ischild(struct prison *pr1, struct prison *pr2)
3263 {
3264
3265 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3266 if (pr1 == pr2)
3267 return (1);
3268 return (0);
3269 }
3270
3271 /*
3272 * Return 1 if the passed credential is in a jail, otherwise 0.
3273 */
3274 int
3275 jailed(struct ucred *cred)
3276 {
3277
3278 return (cred->cr_prison != &prison0);
3279 }
3280
3281 /*
3282 * Return the correct hostname (domainname, et al) for the passed credential.
3283 */
3284 void
3285 getcredhostname(struct ucred *cred, char *buf, size_t size)
3286 {
3287 struct prison *pr;
3288
3289 /*
3290 * A NULL credential can be used to shortcut to the physical
3291 * system's hostname.
3292 */
3293 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3294 mtx_lock(&pr->pr_mtx);
3295 strlcpy(buf, pr->pr_hostname, size);
3296 mtx_unlock(&pr->pr_mtx);
3297 }
3298
3299 void
3300 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3301 {
3302
3303 mtx_lock(&cred->cr_prison->pr_mtx);
3304 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3305 mtx_unlock(&cred->cr_prison->pr_mtx);
3306 }
3307
3308 void
3309 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3310 {
3311
3312 mtx_lock(&cred->cr_prison->pr_mtx);
3313 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3314 mtx_unlock(&cred->cr_prison->pr_mtx);
3315 }
3316
3317 void
3318 getcredhostid(struct ucred *cred, unsigned long *hostid)
3319 {
3320
3321 mtx_lock(&cred->cr_prison->pr_mtx);
3322 *hostid = cred->cr_prison->pr_hostid;
3323 mtx_unlock(&cred->cr_prison->pr_mtx);
3324 }
3325
3326 #ifdef VIMAGE
3327 /*
3328 * Determine whether the prison represented by cred owns
3329 * its vnet rather than having it inherited.
3330 *
3331 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3332 */
3333 int
3334 prison_owns_vnet(struct ucred *cred)
3335 {
3336
3337 /*
3338 * vnets cannot be added/removed after jail creation,
3339 * so no need to lock here.
3340 */
3341 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3342 }
3343 #endif
3344
3345 /*
3346 * Determine whether the subject represented by cred can "see"
3347 * status of a mount point.
3348 * Returns: 0 for permitted, ENOENT otherwise.
3349 * XXX: This function should be called cr_canseemount() and should be
3350 * placed in kern_prot.c.
3351 */
3352 int
3353 prison_canseemount(struct ucred *cred, struct mount *mp)
3354 {
3355 struct prison *pr;
3356 struct statfs *sp;
3357 size_t len;
3358
3359 pr = cred->cr_prison;
3360 if (pr->pr_enforce_statfs == 0)
3361 return (0);
3362 if (pr->pr_root->v_mount == mp)
3363 return (0);
3364 if (pr->pr_enforce_statfs == 2)
3365 return (ENOENT);
3366 /*
3367 * If jail's chroot directory is set to "/" we should be able to see
3368 * all mount-points from inside a jail.
3369 * This is ugly check, but this is the only situation when jail's
3370 * directory ends with '/'.
3371 */
3372 if (strcmp(pr->pr_path, "/") == 0)
3373 return (0);
3374 len = strlen(pr->pr_path);
3375 sp = &mp->mnt_stat;
3376 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3377 return (ENOENT);
3378 /*
3379 * Be sure that we don't have situation where jail's root directory
3380 * is "/some/path" and mount point is "/some/pathpath".
3381 */
3382 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3383 return (ENOENT);
3384 return (0);
3385 }
3386
3387 void
3388 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3389 {
3390 char jpath[MAXPATHLEN];
3391 struct prison *pr;
3392 size_t len;
3393
3394 pr = cred->cr_prison;
3395 if (pr->pr_enforce_statfs == 0)
3396 return;
3397 if (prison_canseemount(cred, mp) != 0) {
3398 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3399 strlcpy(sp->f_mntonname, "[restricted]",
3400 sizeof(sp->f_mntonname));
3401 return;
3402 }
3403 if (pr->pr_root->v_mount == mp) {
3404 /*
3405 * Clear current buffer data, so we are sure nothing from
3406 * the valid path left there.
3407 */
3408 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3409 *sp->f_mntonname = '/';
3410 return;
3411 }
3412 /*
3413 * If jail's chroot directory is set to "/" we should be able to see
3414 * all mount-points from inside a jail.
3415 */
3416 if (strcmp(pr->pr_path, "/") == 0)
3417 return;
3418 len = strlen(pr->pr_path);
3419 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3420 /*
3421 * Clear current buffer data, so we are sure nothing from
3422 * the valid path left there.
3423 */
3424 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3425 if (*jpath == '\0') {
3426 /* Should never happen. */
3427 *sp->f_mntonname = '/';
3428 } else {
3429 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3430 }
3431 }
3432
3433 /*
3434 * Check with permission for a specific privilege is granted within jail. We
3435 * have a specific list of accepted privileges; the rest are denied.
3436 */
3437 int
3438 prison_priv_check(struct ucred *cred, int priv)
3439 {
3440
3441 if (!jailed(cred))
3442 return (0);
3443
3444 #ifdef VIMAGE
3445 /*
3446 * Privileges specific to prisons with a virtual network stack.
3447 * There might be a duplicate entry here in case the privilege
3448 * is only granted conditionally in the legacy jail case.
3449 */
3450 switch (priv) {
3451 #ifdef notyet
3452 /*
3453 * NFS-specific privileges.
3454 */
3455 case PRIV_NFS_DAEMON:
3456 case PRIV_NFS_LOCKD:
3457 #endif
3458 /*
3459 * Network stack privileges.
3460 */
3461 case PRIV_NET_BRIDGE:
3462 case PRIV_NET_GRE:
3463 case PRIV_NET_BPF:
3464 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3465 case PRIV_NET_ROUTE:
3466 case PRIV_NET_TAP:
3467 case PRIV_NET_SETIFMTU:
3468 case PRIV_NET_SETIFFLAGS:
3469 case PRIV_NET_SETIFCAP:
3470 case PRIV_NET_SETIFNAME :
3471 case PRIV_NET_SETIFMETRIC:
3472 case PRIV_NET_SETIFPHYS:
3473 case PRIV_NET_SETIFMAC:
3474 case PRIV_NET_ADDMULTI:
3475 case PRIV_NET_DELMULTI:
3476 case PRIV_NET_HWIOCTL:
3477 case PRIV_NET_SETLLADDR:
3478 case PRIV_NET_ADDIFGROUP:
3479 case PRIV_NET_DELIFGROUP:
3480 case PRIV_NET_IFCREATE:
3481 case PRIV_NET_IFDESTROY:
3482 case PRIV_NET_ADDIFADDR:
3483 case PRIV_NET_DELIFADDR:
3484 case PRIV_NET_LAGG:
3485 case PRIV_NET_GIF:
3486 case PRIV_NET_SETIFVNET:
3487
3488 /*
3489 * 802.11-related privileges.
3490 */
3491 case PRIV_NET80211_GETKEY:
3492 #ifdef notyet
3493 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3494 #endif
3495
3496 #ifdef notyet
3497 /*
3498 * AppleTalk privileges.
3499 */
3500 case PRIV_NETATALK_RESERVEDPORT:
3501
3502 /*
3503 * ATM privileges.
3504 */
3505 case PRIV_NETATM_CFG:
3506 case PRIV_NETATM_ADD:
3507 case PRIV_NETATM_DEL:
3508 case PRIV_NETATM_SET:
3509
3510 /*
3511 * Bluetooth privileges.
3512 */
3513 case PRIV_NETBLUETOOTH_RAW:
3514 #endif
3515
3516 /*
3517 * Netgraph and netgraph module privileges.
3518 */
3519 case PRIV_NETGRAPH_CONTROL:
3520 #ifdef notyet
3521 case PRIV_NETGRAPH_TTY:
3522 #endif
3523
3524 /*
3525 * IPv4 and IPv6 privileges.
3526 */
3527 case PRIV_NETINET_IPFW:
3528 case PRIV_NETINET_DIVERT:
3529 case PRIV_NETINET_PF:
3530 case PRIV_NETINET_DUMMYNET:
3531 case PRIV_NETINET_CARP:
3532 case PRIV_NETINET_MROUTE:
3533 case PRIV_NETINET_RAW:
3534 case PRIV_NETINET_ADDRCTRL6:
3535 case PRIV_NETINET_ND6:
3536 case PRIV_NETINET_SCOPE6:
3537 case PRIV_NETINET_ALIFETIME6:
3538 case PRIV_NETINET_IPSEC:
3539 case PRIV_NETINET_BINDANY:
3540
3541 #ifdef notyet
3542 /*
3543 * IPX/SPX privileges.
3544 */
3545 case PRIV_NETIPX_RESERVEDPORT:
3546 case PRIV_NETIPX_RAW:
3547
3548 /*
3549 * NCP privileges.
3550 */
3551 case PRIV_NETNCP:
3552
3553 /*
3554 * SMB privileges.
3555 */
3556 case PRIV_NETSMB:
3557 #endif
3558
3559 /*
3560 * No default: or deny here.
3561 * In case of no permit fall through to next switch().
3562 */
3563 if (cred->cr_prison->pr_flags & PR_VNET)
3564 return (0);
3565 }
3566 #endif /* VIMAGE */
3567
3568 switch (priv) {
3569
3570 /*
3571 * Allow ktrace privileges for root in jail.
3572 */
3573 case PRIV_KTRACE:
3574
3575 #if 0
3576 /*
3577 * Allow jailed processes to configure audit identity and
3578 * submit audit records (login, etc). In the future we may
3579 * want to further refine the relationship between audit and
3580 * jail.
3581 */
3582 case PRIV_AUDIT_GETAUDIT:
3583 case PRIV_AUDIT_SETAUDIT:
3584 case PRIV_AUDIT_SUBMIT:
3585 #endif
3586
3587 /*
3588 * Allow jailed processes to manipulate process UNIX
3589 * credentials in any way they see fit.
3590 */
3591 case PRIV_CRED_SETUID:
3592 case PRIV_CRED_SETEUID:
3593 case PRIV_CRED_SETGID:
3594 case PRIV_CRED_SETEGID:
3595 case PRIV_CRED_SETGROUPS:
3596 case PRIV_CRED_SETREUID:
3597 case PRIV_CRED_SETREGID:
3598 case PRIV_CRED_SETRESUID:
3599 case PRIV_CRED_SETRESGID:
3600
3601 /*
3602 * Jail implements visibility constraints already, so allow
3603 * jailed root to override uid/gid-based constraints.
3604 */
3605 case PRIV_SEEOTHERGIDS:
3606 case PRIV_SEEOTHERUIDS:
3607
3608 /*
3609 * Jail implements inter-process debugging limits already, so
3610 * allow jailed root various debugging privileges.
3611 */
3612 case PRIV_DEBUG_DIFFCRED:
3613 case PRIV_DEBUG_SUGID:
3614 case PRIV_DEBUG_UNPRIV:
3615
3616 /*
3617 * Allow jail to set various resource limits and login
3618 * properties, and for now, exceed process resource limits.
3619 */
3620 case PRIV_PROC_LIMIT:
3621 case PRIV_PROC_SETLOGIN:
3622 case PRIV_PROC_SETRLIMIT:
3623
3624 /*
3625 * System V and POSIX IPC privileges are granted in jail.
3626 */
3627 case PRIV_IPC_READ:
3628 case PRIV_IPC_WRITE:
3629 case PRIV_IPC_ADMIN:
3630 case PRIV_IPC_MSGSIZE:
3631 case PRIV_MQ_ADMIN:
3632
3633 /*
3634 * Jail operations within a jail work on child jails.
3635 */
3636 case PRIV_JAIL_ATTACH:
3637 case PRIV_JAIL_SET:
3638 case PRIV_JAIL_REMOVE:
3639
3640 /*
3641 * Jail implements its own inter-process limits, so allow
3642 * root processes in jail to change scheduling on other
3643 * processes in the same jail. Likewise for signalling.
3644 */
3645 case PRIV_SCHED_DIFFCRED:
3646 case PRIV_SCHED_CPUSET:
3647 case PRIV_SIGNAL_DIFFCRED:
3648 case PRIV_SIGNAL_SUGID:
3649
3650 /*
3651 * Allow jailed processes to write to sysctls marked as jail
3652 * writable.
3653 */
3654 case PRIV_SYSCTL_WRITEJAIL:
3655
3656 /*
3657 * Allow root in jail to manage a variety of quota
3658 * properties. These should likely be conditional on a
3659 * configuration option.
3660 */
3661 case PRIV_VFS_GETQUOTA:
3662 case PRIV_VFS_SETQUOTA:
3663
3664 /*
3665 * Since Jail relies on chroot() to implement file system
3666 * protections, grant many VFS privileges to root in jail.
3667 * Be careful to exclude mount-related and NFS-related
3668 * privileges.
3669 */
3670 case PRIV_VFS_READ:
3671 case PRIV_VFS_WRITE:
3672 case PRIV_VFS_ADMIN:
3673 case PRIV_VFS_EXEC:
3674 case PRIV_VFS_LOOKUP:
3675 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3676 case PRIV_VFS_CHFLAGS_DEV:
3677 case PRIV_VFS_CHOWN:
3678 case PRIV_VFS_CHROOT:
3679 case PRIV_VFS_RETAINSUGID:
3680 case PRIV_VFS_FCHROOT:
3681 case PRIV_VFS_LINK:
3682 case PRIV_VFS_SETGID:
3683 case PRIV_VFS_STAT:
3684 case PRIV_VFS_STICKYFILE:
3685 return (0);
3686
3687 /*
3688 * Depending on the global setting, allow privilege of
3689 * setting system flags.
3690 */
3691 case PRIV_VFS_SYSFLAGS:
3692 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3693 return (0);
3694 else
3695 return (EPERM);
3696
3697 /*
3698 * Depending on the global setting, allow privilege of
3699 * mounting/unmounting file systems.
3700 */
3701 case PRIV_VFS_MOUNT:
3702 case PRIV_VFS_UNMOUNT:
3703 case PRIV_VFS_MOUNT_NONUSER:
3704 case PRIV_VFS_MOUNT_OWNER:
3705 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT)
3706 return (0);
3707 else
3708 return (EPERM);
3709
3710 /*
3711 * Allow jailed root to bind reserved ports and reuse in-use
3712 * ports.
3713 */
3714 case PRIV_NETINET_RESERVEDPORT:
3715 case PRIV_NETINET_REUSEPORT:
3716 return (0);
3717
3718 /*
3719 * Allow jailed root to set certian IPv4/6 (option) headers.
3720 */
3721 case PRIV_NETINET_SETHDROPTS:
3722 return (0);
3723
3724 /*
3725 * Conditionally allow creating raw sockets in jail.
3726 */
3727 case PRIV_NETINET_RAW:
3728 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3729 return (0);
3730 else
3731 return (EPERM);
3732
3733 /*
3734 * Since jail implements its own visibility limits on netstat
3735 * sysctls, allow getcred. This allows identd to work in
3736 * jail.
3737 */
3738 case PRIV_NETINET_GETCRED:
3739 return (0);
3740
3741 default:
3742 /*
3743 * In all remaining cases, deny the privilege request. This
3744 * includes almost all network privileges, many system
3745 * configuration privileges.
3746 */
3747 return (EPERM);
3748 }
3749 }
3750
3751 /*
3752 * Return the part of pr2's name that is relative to pr1, or the whole name
3753 * if it does not directly follow.
3754 */
3755
3756 char *
3757 prison_name(struct prison *pr1, struct prison *pr2)
3758 {
3759 char *name;
3760
3761 /* Jails see themselves as "" (if they see themselves at all). */
3762 if (pr1 == pr2)
3763 return "";
3764 name = pr2->pr_name;
3765 if (prison_ischild(pr1, pr2)) {
3766 /*
3767 * pr1 isn't locked (and allprison_lock may not be either)
3768 * so its length can't be counted on. But the number of dots
3769 * can be counted on - and counted.
3770 */
3771 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3772 name = strchr(name, '.') + 1;
3773 }
3774 return (name);
3775 }
3776
3777 /*
3778 * Return the part of pr2's path that is relative to pr1, or the whole path
3779 * if it does not directly follow.
3780 */
3781 static char *
3782 prison_path(struct prison *pr1, struct prison *pr2)
3783 {
3784 char *path1, *path2;
3785 int len1;
3786
3787 path1 = pr1->pr_path;
3788 path2 = pr2->pr_path;
3789 if (!strcmp(path1, "/"))
3790 return (path2);
3791 len1 = strlen(path1);
3792 if (strncmp(path1, path2, len1))
3793 return (path2);
3794 if (path2[len1] == '\0')
3795 return "/";
3796 if (path2[len1] == '/')
3797 return (path2 + len1);
3798 return (path2);
3799 }
3800
3801
3802 /*
3803 * Jail-related sysctls.
3804 */
3805 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
3806 "Jails");
3807
3808 static int
3809 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
3810 {
3811 struct xprison *xp;
3812 struct prison *pr, *cpr;
3813 #ifdef INET
3814 struct in_addr *ip4 = NULL;
3815 int ip4s = 0;
3816 #endif
3817 #ifdef INET6
3818 struct in_addr *ip6 = NULL;
3819 int ip6s = 0;
3820 #endif
3821 int descend, error;
3822
3823 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
3824 pr = req->td->td_ucred->cr_prison;
3825 error = 0;
3826 sx_slock(&allprison_lock);
3827 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
3828 #if defined(INET) || defined(INET6)
3829 again:
3830 #endif
3831 mtx_lock(&cpr->pr_mtx);
3832 #ifdef INET
3833 if (cpr->pr_ip4s > 0) {
3834 if (ip4s < cpr->pr_ip4s) {
3835 ip4s = cpr->pr_ip4s;
3836 mtx_unlock(&cpr->pr_mtx);
3837 ip4 = realloc(ip4, ip4s *
3838 sizeof(struct in_addr), M_TEMP, M_WAITOK);
3839 goto again;
3840 }
3841 bcopy(cpr->pr_ip4, ip4,
3842 cpr->pr_ip4s * sizeof(struct in_addr));
3843 }
3844 #endif
3845 #ifdef INET6
3846 if (cpr->pr_ip6s > 0) {
3847 if (ip6s < cpr->pr_ip6s) {
3848 ip6s = cpr->pr_ip6s;
3849 mtx_unlock(&cpr->pr_mtx);
3850 ip6 = realloc(ip6, ip6s *
3851 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
3852 goto again;
3853 }
3854 bcopy(cpr->pr_ip6, ip6,
3855 cpr->pr_ip6s * sizeof(struct in6_addr));
3856 }
3857 #endif
3858 if (cpr->pr_ref == 0) {
3859 mtx_unlock(&cpr->pr_mtx);
3860 continue;
3861 }
3862 bzero(xp, sizeof(*xp));
3863 xp->pr_version = XPRISON_VERSION;
3864 xp->pr_id = cpr->pr_id;
3865 xp->pr_state = cpr->pr_uref > 0
3866 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
3867 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
3868 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
3869 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
3870 #ifdef INET
3871 xp->pr_ip4s = cpr->pr_ip4s;
3872 #endif
3873 #ifdef INET6
3874 xp->pr_ip6s = cpr->pr_ip6s;
3875 #endif
3876 mtx_unlock(&cpr->pr_mtx);
3877 error = SYSCTL_OUT(req, xp, sizeof(*xp));
3878 if (error)
3879 break;
3880 #ifdef INET
3881 if (xp->pr_ip4s > 0) {
3882 error = SYSCTL_OUT(req, ip4,
3883 xp->pr_ip4s * sizeof(struct in_addr));
3884 if (error)
3885 break;
3886 }
3887 #endif
3888 #ifdef INET6
3889 if (xp->pr_ip6s > 0) {
3890 error = SYSCTL_OUT(req, ip6,
3891 xp->pr_ip6s * sizeof(struct in6_addr));
3892 if (error)
3893 break;
3894 }
3895 #endif
3896 }
3897 sx_sunlock(&allprison_lock);
3898 free(xp, M_TEMP);
3899 #ifdef INET
3900 free(ip4, M_TEMP);
3901 #endif
3902 #ifdef INET6
3903 free(ip6, M_TEMP);
3904 #endif
3905 return (error);
3906 }
3907
3908 SYSCTL_OID(_security_jail, OID_AUTO, list,
3909 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3910 sysctl_jail_list, "S", "List of active jails");
3911
3912 static int
3913 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
3914 {
3915 int error, injail;
3916
3917 injail = jailed(req->td->td_ucred);
3918 error = SYSCTL_OUT(req, &injail, sizeof(injail));
3919
3920 return (error);
3921 }
3922
3923 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
3924 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
3925 sysctl_jail_jailed, "I", "Process in jail?");
3926
3927 #if defined(INET) || defined(INET6)
3928 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
3929 &jail_max_af_ips, 0,
3930 "Number of IP addresses a jail may have at most per address family");
3931 #endif
3932
3933 /*
3934 * Default parameters for jail(2) compatability. For historical reasons,
3935 * the sysctl names have varying similarity to the parameter names. Prisons
3936 * just see their own parameters, and can't change them.
3937 */
3938 static int
3939 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
3940 {
3941 struct prison *pr;
3942 int allow, error, i;
3943
3944 pr = req->td->td_ucred->cr_prison;
3945 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
3946
3947 /* Get the current flag value, and convert it to a boolean. */
3948 i = (allow & arg2) ? 1 : 0;
3949 if (arg1 != NULL)
3950 i = !i;
3951 error = sysctl_handle_int(oidp, &i, 0, req);
3952 if (error || !req->newptr)
3953 return (error);
3954 i = i ? arg2 : 0;
3955 if (arg1 != NULL)
3956 i ^= arg2;
3957 /*
3958 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
3959 * for writing.
3960 */
3961 mtx_lock(&prison0.pr_mtx);
3962 jail_default_allow = (jail_default_allow & ~arg2) | i;
3963 mtx_unlock(&prison0.pr_mtx);
3964 return (0);
3965 }
3966
3967 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
3968 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3969 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
3970 "Processes in jail can set their hostnames");
3971 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
3972 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3973 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
3974 "Processes in jail are limited to creating UNIX/IP/route sockets only");
3975 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
3976 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3977 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
3978 "Processes in jail can use System V IPC primitives");
3979 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
3980 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3981 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
3982 "Prison root can create raw sockets");
3983 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
3984 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3985 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
3986 "Processes in jail can alter system file flags");
3987 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
3988 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
3989 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
3990 "Processes in jail can mount/unmount jail-friendly file systems");
3991
3992 static int
3993 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
3994 {
3995 struct prison *pr;
3996 int level, error;
3997
3998 pr = req->td->td_ucred->cr_prison;
3999 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4000 error = sysctl_handle_int(oidp, &level, 0, req);
4001 if (error || !req->newptr)
4002 return (error);
4003 *(int *)arg1 = level;
4004 return (0);
4005 }
4006
4007 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4008 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4009 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4010 sysctl_jail_default_level, "I",
4011 "Processes in jail cannot see all mounted file systems");
4012
4013 /*
4014 * Nodes to describe jail parameters. Maximum length of string parameters
4015 * is returned in the string itself, and the other parameters exist merely
4016 * to make themselves and their types known.
4017 */
4018 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4019 "Jail parameters");
4020
4021 int
4022 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4023 {
4024 int i;
4025 long l;
4026 size_t s;
4027 char numbuf[12];
4028
4029 switch (oidp->oid_kind & CTLTYPE)
4030 {
4031 case CTLTYPE_LONG:
4032 case CTLTYPE_ULONG:
4033 l = 0;
4034 #ifdef SCTL_MASK32
4035 if (!(req->flags & SCTL_MASK32))
4036 #endif
4037 return (SYSCTL_OUT(req, &l, sizeof(l)));
4038 case CTLTYPE_INT:
4039 case CTLTYPE_UINT:
4040 i = 0;
4041 return (SYSCTL_OUT(req, &i, sizeof(i)));
4042 case CTLTYPE_STRING:
4043 snprintf(numbuf, sizeof(numbuf), "%d", arg2);
4044 return
4045 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4046 case CTLTYPE_STRUCT:
4047 s = (size_t)arg2;
4048 return (SYSCTL_OUT(req, &s, sizeof(s)));
4049 }
4050 return (0);
4051 }
4052
4053 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4054 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4055 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4056 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4057 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4058 "I", "Jail secure level");
4059 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4060 "I", "Jail cannot see all mounted file systems");
4061 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4062 "B", "Jail persistence");
4063 #ifdef VIMAGE
4064 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4065 "E,jailsys", "Virtual network stack");
4066 #endif
4067 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4068 "B", "Jail is in the process of shutting down");
4069
4070 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4071 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4072 "I", "Current number of child jails");
4073 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4074 "I", "Maximum number of child jails");
4075
4076 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4077 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4078 "Jail hostname");
4079 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4080 "Jail NIS domainname");
4081 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4082 "Jail host UUID");
4083 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4084 "LU", "Jail host ID");
4085
4086 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4087 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4088
4089 #ifdef INET
4090 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4091 "Jail IPv4 address virtualization");
4092 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4093 "S,in_addr,a", "Jail IPv4 addresses");
4094 #endif
4095 #ifdef INET6
4096 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4097 "Jail IPv6 address virtualization");
4098 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4099 "S,in6_addr,a", "Jail IPv6 addresses");
4100 #endif
4101
4102 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4103 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4104 "B", "Jail may set hostname");
4105 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4106 "B", "Jail may use SYSV IPC");
4107 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4108 "B", "Jail may create raw sockets");
4109 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4110 "B", "Jail may alter system file flags");
4111 SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
4112 "B", "Jail may mount/unmount jail-friendly file systems");
4113 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4114 "B", "Jail may set file quotas");
4115 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4116 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4117
4118
4119 #ifdef DDB
4120
4121 static void
4122 db_show_prison(struct prison *pr)
4123 {
4124 int fi;
4125 #if defined(INET) || defined(INET6)
4126 int ii;
4127 #endif
4128 unsigned jsf;
4129 #ifdef INET6
4130 char ip6buf[INET6_ADDRSTRLEN];
4131 #endif
4132
4133 db_printf("prison %p:\n", pr);
4134 db_printf(" jid = %d\n", pr->pr_id);
4135 db_printf(" name = %s\n", pr->pr_name);
4136 db_printf(" parent = %p\n", pr->pr_parent);
4137 db_printf(" ref = %d\n", pr->pr_ref);
4138 db_printf(" uref = %d\n", pr->pr_uref);
4139 db_printf(" path = %s\n", pr->pr_path);
4140 db_printf(" cpuset = %d\n", pr->pr_cpuset
4141 ? pr->pr_cpuset->cs_id : -1);
4142 #ifdef VIMAGE
4143 db_printf(" vnet = %p\n", pr->pr_vnet);
4144 #endif
4145 db_printf(" root = %p\n", pr->pr_root);
4146 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4147 db_printf(" childcount = %d\n", pr->pr_childcount);
4148 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4149 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4150 db_printf(" flags = %x", pr->pr_flags);
4151 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4152 fi++)
4153 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4154 db_printf(" %s", pr_flag_names[fi]);
4155 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4156 fi++) {
4157 jsf = pr->pr_flags &
4158 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4159 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4160 pr_flag_jailsys[fi].disable &&
4161 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4162 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4163 : "inherit");
4164 }
4165 db_printf(" allow = %x", pr->pr_allow);
4166 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4167 fi++)
4168 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4169 db_printf(" %s", pr_allow_names[fi]);
4170 db_printf("\n");
4171 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4172 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4173 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4174 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4175 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4176 #ifdef INET
4177 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4178 for (ii = 0; ii < pr->pr_ip4s; ii++)
4179 db_printf(" %s %s\n",
4180 ii == 0 ? "ip4 =" : " ",
4181 inet_ntoa(pr->pr_ip4[ii]));
4182 #endif
4183 #ifdef INET6
4184 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4185 for (ii = 0; ii < pr->pr_ip6s; ii++)
4186 db_printf(" %s %s\n",
4187 ii == 0 ? "ip6 =" : " ",
4188 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4189 #endif
4190 }
4191
4192 DB_SHOW_COMMAND(prison, db_show_prison_command)
4193 {
4194 struct prison *pr;
4195
4196 if (!have_addr) {
4197 /*
4198 * Show all prisons in the list, and prison0 which is not
4199 * listed.
4200 */
4201 db_show_prison(&prison0);
4202 if (!db_pager_quit) {
4203 TAILQ_FOREACH(pr, &allprison, pr_list) {
4204 db_show_prison(pr);
4205 if (db_pager_quit)
4206 break;
4207 }
4208 }
4209 return;
4210 }
4211
4212 if (addr == 0)
4213 pr = &prison0;
4214 else {
4215 /* Look for a prison with the ID and with references. */
4216 TAILQ_FOREACH(pr, &allprison, pr_list)
4217 if (pr->pr_id == addr && pr->pr_ref > 0)
4218 break;
4219 if (pr == NULL)
4220 /* Look again, without requiring a reference. */
4221 TAILQ_FOREACH(pr, &allprison, pr_list)
4222 if (pr->pr_id == addr)
4223 break;
4224 if (pr == NULL)
4225 /* Assume address points to a valid prison. */
4226 pr = (struct prison *)addr;
4227 }
4228 db_show_prison(pr);
4229 }
4230
4231 #endif /* DDB */
Cache object: 29c21d0317246fc2b917d34940d7ee9d
|