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/10.3/sys/kern/kern_jail.c 295951 2016-02-24 02:34:11Z araujo $");
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/racct.h>
53 #include <sys/refcount.h>
54 #include <sys/sx.h>
55 #include <sys/sysent.h>
56 #include <sys/namei.h>
57 #include <sys/mount.h>
58 #include <sys/queue.h>
59 #include <sys/socket.h>
60 #include <sys/syscallsubr.h>
61 #include <sys/sysctl.h>
62 #include <sys/vnode.h>
63
64 #include <net/if.h>
65 #include <net/vnet.h>
66
67 #include <netinet/in.h>
68
69 #ifdef DDB
70 #include <ddb/ddb.h>
71 #ifdef INET6
72 #include <netinet6/in6_var.h>
73 #endif /* INET6 */
74 #endif /* DDB */
75
76 #include <security/mac/mac_framework.h>
77
78 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000"
79
80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
81 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
82
83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
84 #ifdef INET
85 #ifdef INET6
86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
87 #else
88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
89 #endif
90 #else /* !INET */
91 #ifdef INET6
92 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
93 #else
94 #define _PR_IP_SADDRSEL 0
95 #endif
96 #endif
97
98 /* prison0 describes what is "real" about the system. */
99 struct prison prison0 = {
100 .pr_id = 0,
101 .pr_name = "",
102 .pr_ref = 1,
103 .pr_uref = 1,
104 .pr_path = "/",
105 .pr_securelevel = -1,
106 .pr_devfs_rsnum = 0,
107 .pr_childmax = JAIL_MAX,
108 .pr_hostuuid = DEFAULT_HOSTUUID,
109 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children),
110 #ifdef VIMAGE
111 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
112 #else
113 .pr_flags = PR_HOST|_PR_IP_SADDRSEL,
114 #endif
115 .pr_allow = PR_ALLOW_ALL,
116 };
117 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
118
119 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
120 struct sx allprison_lock;
121 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
122 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
123 LIST_HEAD(, prison_racct) allprison_racct;
124 int lastprid = 0;
125
126 static int do_jail_attach(struct thread *td, struct prison *pr);
127 static void prison_complete(void *context, int pending);
128 static void prison_deref(struct prison *pr, int flags);
129 static char *prison_path(struct prison *pr1, struct prison *pr2);
130 static void prison_remove_one(struct prison *pr);
131 #ifdef RACCT
132 static void prison_racct_attach(struct prison *pr);
133 static void prison_racct_modify(struct prison *pr);
134 static void prison_racct_detach(struct prison *pr);
135 #endif
136 #ifdef INET
137 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
138 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
139 #endif
140 #ifdef INET6
141 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
142 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
143 #endif
144
145 /* Flags for prison_deref */
146 #define PD_DEREF 0x01
147 #define PD_DEUREF 0x02
148 #define PD_LOCKED 0x04
149 #define PD_LIST_SLOCKED 0x08
150 #define PD_LIST_XLOCKED 0x10
151
152 /*
153 * Parameter names corresponding to PR_* flag values. Size values are for kvm
154 * as we cannot figure out the size of a sparse array, or an array without a
155 * terminating entry.
156 */
157 static char *pr_flag_names[] = {
158 [0] = "persist",
159 #ifdef INET
160 [7] = "ip4.saddrsel",
161 #endif
162 #ifdef INET6
163 [8] = "ip6.saddrsel",
164 #endif
165 };
166 const size_t pr_flag_names_size = sizeof(pr_flag_names);
167
168 static char *pr_flag_nonames[] = {
169 [0] = "nopersist",
170 #ifdef INET
171 [7] = "ip4.nosaddrsel",
172 #endif
173 #ifdef INET6
174 [8] = "ip6.nosaddrsel",
175 #endif
176 };
177 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
178
179 struct jailsys_flags {
180 const char *name;
181 unsigned disable;
182 unsigned new;
183 } pr_flag_jailsys[] = {
184 { "host", 0, PR_HOST },
185 #ifdef VIMAGE
186 { "vnet", 0, PR_VNET },
187 #endif
188 #ifdef INET
189 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
190 #endif
191 #ifdef INET6
192 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
193 #endif
194 };
195 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
196
197 static char *pr_allow_names[] = {
198 "allow.set_hostname",
199 "allow.sysvipc",
200 "allow.raw_sockets",
201 "allow.chflags",
202 "allow.mount",
203 "allow.quotas",
204 "allow.socket_af",
205 "allow.mount.devfs",
206 "allow.mount.nullfs",
207 "allow.mount.zfs",
208 "allow.mount.procfs",
209 "allow.mount.tmpfs",
210 "allow.mount.fdescfs",
211 "allow.mount.linprocfs",
212 "allow.mount.linsysfs",
213 };
214 const size_t pr_allow_names_size = sizeof(pr_allow_names);
215
216 static char *pr_allow_nonames[] = {
217 "allow.noset_hostname",
218 "allow.nosysvipc",
219 "allow.noraw_sockets",
220 "allow.nochflags",
221 "allow.nomount",
222 "allow.noquotas",
223 "allow.nosocket_af",
224 "allow.mount.nodevfs",
225 "allow.mount.nonullfs",
226 "allow.mount.nozfs",
227 "allow.mount.noprocfs",
228 "allow.mount.notmpfs",
229 "allow.mount.nofdescfs",
230 "allow.mount.nolinprocfs",
231 "allow.mount.nolinsysfs",
232 };
233 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
234
235 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
236 #define JAIL_DEFAULT_ENFORCE_STATFS 2
237 #define JAIL_DEFAULT_DEVFS_RSNUM 0
238 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
239 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
240 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
241 #if defined(INET) || defined(INET6)
242 static unsigned jail_max_af_ips = 255;
243 #endif
244
245 /*
246 * Initialize the parts of prison0 that can't be static-initialized with
247 * constants. This is called from proc0_init() after creating thread0 cpuset.
248 */
249 void
250 prison0_init(void)
251 {
252
253 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset);
254 prison0.pr_osreldate = osreldate;
255 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease));
256 }
257
258 #ifdef INET
259 static int
260 qcmp_v4(const void *ip1, const void *ip2)
261 {
262 in_addr_t iaa, iab;
263
264 /*
265 * We need to compare in HBO here to get the list sorted as expected
266 * by the result of the code. Sorting NBO addresses gives you
267 * interesting results. If you do not understand, do not try.
268 */
269 iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
270 iab = ntohl(((const struct in_addr *)ip2)->s_addr);
271
272 /*
273 * Do not simply return the difference of the two numbers, the int is
274 * not wide enough.
275 */
276 if (iaa > iab)
277 return (1);
278 else if (iaa < iab)
279 return (-1);
280 else
281 return (0);
282 }
283 #endif
284
285 #ifdef INET6
286 static int
287 qcmp_v6(const void *ip1, const void *ip2)
288 {
289 const struct in6_addr *ia6a, *ia6b;
290 int i, rc;
291
292 ia6a = (const struct in6_addr *)ip1;
293 ia6b = (const struct in6_addr *)ip2;
294
295 rc = 0;
296 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
297 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
298 rc = 1;
299 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
300 rc = -1;
301 }
302 return (rc);
303 }
304 #endif
305
306 /*
307 * struct jail_args {
308 * struct jail *jail;
309 * };
310 */
311 int
312 sys_jail(struct thread *td, struct jail_args *uap)
313 {
314 uint32_t version;
315 int error;
316 struct jail j;
317
318 error = copyin(uap->jail, &version, sizeof(uint32_t));
319 if (error)
320 return (error);
321
322 switch (version) {
323 case 0:
324 {
325 struct jail_v0 j0;
326
327 /* FreeBSD single IPv4 jails. */
328 bzero(&j, sizeof(struct jail));
329 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
330 if (error)
331 return (error);
332 j.version = j0.version;
333 j.path = j0.path;
334 j.hostname = j0.hostname;
335 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
336 break;
337 }
338
339 case 1:
340 /*
341 * Version 1 was used by multi-IPv4 jail implementations
342 * that never made it into the official kernel.
343 */
344 return (EINVAL);
345
346 case 2: /* JAIL_API_VERSION */
347 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
348 error = copyin(uap->jail, &j, sizeof(struct jail));
349 if (error)
350 return (error);
351 break;
352
353 default:
354 /* Sci-Fi jails are not supported, sorry. */
355 return (EINVAL);
356 }
357 return (kern_jail(td, &j));
358 }
359
360 int
361 kern_jail(struct thread *td, struct jail *j)
362 {
363 struct iovec optiov[2 * (4
364 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
365 #ifdef INET
366 + 1
367 #endif
368 #ifdef INET6
369 + 1
370 #endif
371 )];
372 struct uio opt;
373 char *u_path, *u_hostname, *u_name;
374 #ifdef INET
375 uint32_t ip4s;
376 struct in_addr *u_ip4;
377 #endif
378 #ifdef INET6
379 struct in6_addr *u_ip6;
380 #endif
381 size_t tmplen;
382 int error, enforce_statfs, fi;
383
384 bzero(&optiov, sizeof(optiov));
385 opt.uio_iov = optiov;
386 opt.uio_iovcnt = 0;
387 opt.uio_offset = -1;
388 opt.uio_resid = -1;
389 opt.uio_segflg = UIO_SYSSPACE;
390 opt.uio_rw = UIO_READ;
391 opt.uio_td = td;
392
393 /* Set permissions for top-level jails from sysctls. */
394 if (!jailed(td->td_ucred)) {
395 for (fi = 0; fi < sizeof(pr_allow_names) /
396 sizeof(pr_allow_names[0]); fi++) {
397 optiov[opt.uio_iovcnt].iov_base =
398 (jail_default_allow & (1 << fi))
399 ? pr_allow_names[fi] : pr_allow_nonames[fi];
400 optiov[opt.uio_iovcnt].iov_len =
401 strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
402 opt.uio_iovcnt += 2;
403 }
404 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
405 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
406 opt.uio_iovcnt++;
407 enforce_statfs = jail_default_enforce_statfs;
408 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
409 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
410 opt.uio_iovcnt++;
411 }
412
413 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
414 #ifdef INET
415 ip4s = (j->version == 0) ? 1 : j->ip4s;
416 if (ip4s > jail_max_af_ips)
417 return (EINVAL);
418 tmplen += ip4s * sizeof(struct in_addr);
419 #else
420 if (j->ip4s > 0)
421 return (EINVAL);
422 #endif
423 #ifdef INET6
424 if (j->ip6s > jail_max_af_ips)
425 return (EINVAL);
426 tmplen += j->ip6s * sizeof(struct in6_addr);
427 #else
428 if (j->ip6s > 0)
429 return (EINVAL);
430 #endif
431 u_path = malloc(tmplen, M_TEMP, M_WAITOK);
432 u_hostname = u_path + MAXPATHLEN;
433 u_name = u_hostname + MAXHOSTNAMELEN;
434 #ifdef INET
435 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
436 #endif
437 #ifdef INET6
438 #ifdef INET
439 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
440 #else
441 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
442 #endif
443 #endif
444 optiov[opt.uio_iovcnt].iov_base = "path";
445 optiov[opt.uio_iovcnt].iov_len = sizeof("path");
446 opt.uio_iovcnt++;
447 optiov[opt.uio_iovcnt].iov_base = u_path;
448 error = copyinstr(j->path, u_path, MAXPATHLEN,
449 &optiov[opt.uio_iovcnt].iov_len);
450 if (error) {
451 free(u_path, M_TEMP);
452 return (error);
453 }
454 opt.uio_iovcnt++;
455 optiov[opt.uio_iovcnt].iov_base = "host.hostname";
456 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
457 opt.uio_iovcnt++;
458 optiov[opt.uio_iovcnt].iov_base = u_hostname;
459 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
460 &optiov[opt.uio_iovcnt].iov_len);
461 if (error) {
462 free(u_path, M_TEMP);
463 return (error);
464 }
465 opt.uio_iovcnt++;
466 if (j->jailname != NULL) {
467 optiov[opt.uio_iovcnt].iov_base = "name";
468 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
469 opt.uio_iovcnt++;
470 optiov[opt.uio_iovcnt].iov_base = u_name;
471 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
472 &optiov[opt.uio_iovcnt].iov_len);
473 if (error) {
474 free(u_path, M_TEMP);
475 return (error);
476 }
477 opt.uio_iovcnt++;
478 }
479 #ifdef INET
480 optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
481 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
482 opt.uio_iovcnt++;
483 optiov[opt.uio_iovcnt].iov_base = u_ip4;
484 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
485 if (j->version == 0)
486 u_ip4->s_addr = j->ip4s;
487 else {
488 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
489 if (error) {
490 free(u_path, M_TEMP);
491 return (error);
492 }
493 }
494 opt.uio_iovcnt++;
495 #endif
496 #ifdef INET6
497 optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
498 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
499 opt.uio_iovcnt++;
500 optiov[opt.uio_iovcnt].iov_base = u_ip6;
501 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
502 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
503 if (error) {
504 free(u_path, M_TEMP);
505 return (error);
506 }
507 opt.uio_iovcnt++;
508 #endif
509 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
510 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
511 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
512 free(u_path, M_TEMP);
513 return (error);
514 }
515
516
517 /*
518 * struct jail_set_args {
519 * struct iovec *iovp;
520 * unsigned int iovcnt;
521 * int flags;
522 * };
523 */
524 int
525 sys_jail_set(struct thread *td, struct jail_set_args *uap)
526 {
527 struct uio *auio;
528 int error;
529
530 /* Check that we have an even number of iovecs. */
531 if (uap->iovcnt & 1)
532 return (EINVAL);
533
534 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
535 if (error)
536 return (error);
537 error = kern_jail_set(td, auio, uap->flags);
538 free(auio, M_IOV);
539 return (error);
540 }
541
542 int
543 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
544 {
545 struct nameidata nd;
546 #ifdef INET
547 struct in_addr *ip4;
548 #endif
549 #ifdef INET6
550 struct in6_addr *ip6;
551 #endif
552 struct vfsopt *opt;
553 struct vfsoptlist *opts;
554 struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
555 struct vnode *root;
556 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
557 char *g_path, *osrelstr;
558 #if defined(INET) || defined(INET6)
559 struct prison *tppr;
560 void *op;
561 #endif
562 unsigned long hid;
563 size_t namelen, onamelen;
564 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
565 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
566 int fi, jid, jsys, len, level;
567 int childmax, osreldt, rsnum, slevel;
568 int fullpath_disabled;
569 #if defined(INET) || defined(INET6)
570 int ii, ij;
571 #endif
572 #ifdef INET
573 int ip4s, redo_ip4;
574 #endif
575 #ifdef INET6
576 int ip6s, redo_ip6;
577 #endif
578 uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
579 unsigned tallow;
580 char numbuf[12];
581
582 error = priv_check(td, PRIV_JAIL_SET);
583 if (!error && (flags & JAIL_ATTACH))
584 error = priv_check(td, PRIV_JAIL_ATTACH);
585 if (error)
586 return (error);
587 mypr = ppr = td->td_ucred->cr_prison;
588 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
589 return (EPERM);
590 if (flags & ~JAIL_SET_MASK)
591 return (EINVAL);
592
593 /*
594 * Check all the parameters before committing to anything. Not all
595 * errors can be caught early, but we may as well try. Also, this
596 * takes care of some expensive stuff (path lookup) before getting
597 * the allprison lock.
598 *
599 * XXX Jails are not filesystems, and jail parameters are not mount
600 * options. But it makes more sense to re-use the vfsopt code
601 * than duplicate it under a different name.
602 */
603 error = vfs_buildopts(optuio, &opts);
604 if (error)
605 return (error);
606 #ifdef INET
607 ip4 = NULL;
608 #endif
609 #ifdef INET6
610 ip6 = NULL;
611 #endif
612 g_path = NULL;
613
614 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
615 if (error == ENOENT)
616 jid = 0;
617 else if (error != 0)
618 goto done_free;
619
620 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
621 if (error == ENOENT)
622 gotslevel = 0;
623 else if (error != 0)
624 goto done_free;
625 else
626 gotslevel = 1;
627
628 error =
629 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
630 if (error == ENOENT)
631 gotchildmax = 0;
632 else if (error != 0)
633 goto done_free;
634 else
635 gotchildmax = 1;
636
637 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
638 if (error == ENOENT)
639 gotenforce = 0;
640 else if (error != 0)
641 goto done_free;
642 else if (enforce < 0 || enforce > 2) {
643 error = EINVAL;
644 goto done_free;
645 } else
646 gotenforce = 1;
647
648 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
649 if (error == ENOENT)
650 gotrsnum = 0;
651 else if (error != 0)
652 goto done_free;
653 else
654 gotrsnum = 1;
655
656 pr_flags = ch_flags = 0;
657 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
658 fi++) {
659 if (pr_flag_names[fi] == NULL)
660 continue;
661 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
662 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
663 }
664 ch_flags |= pr_flags;
665 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
666 fi++) {
667 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
668 sizeof(jsys));
669 if (error == ENOENT)
670 continue;
671 if (error != 0)
672 goto done_free;
673 switch (jsys) {
674 case JAIL_SYS_DISABLE:
675 if (!pr_flag_jailsys[fi].disable) {
676 error = EINVAL;
677 goto done_free;
678 }
679 pr_flags |= pr_flag_jailsys[fi].disable;
680 break;
681 case JAIL_SYS_NEW:
682 pr_flags |= pr_flag_jailsys[fi].new;
683 break;
684 case JAIL_SYS_INHERIT:
685 break;
686 default:
687 error = EINVAL;
688 goto done_free;
689 }
690 ch_flags |=
691 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
692 }
693 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
694 && !(pr_flags & PR_PERSIST)) {
695 error = EINVAL;
696 vfs_opterror(opts, "new jail must persist or attach");
697 goto done_errmsg;
698 }
699 #ifdef VIMAGE
700 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
701 error = EINVAL;
702 vfs_opterror(opts, "vnet cannot be changed after creation");
703 goto done_errmsg;
704 }
705 #endif
706 #ifdef INET
707 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
708 error = EINVAL;
709 vfs_opterror(opts, "ip4 cannot be changed after creation");
710 goto done_errmsg;
711 }
712 #endif
713 #ifdef INET6
714 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
715 error = EINVAL;
716 vfs_opterror(opts, "ip6 cannot be changed after creation");
717 goto done_errmsg;
718 }
719 #endif
720
721 pr_allow = ch_allow = 0;
722 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
723 fi++) {
724 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
725 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
726 }
727 ch_allow |= pr_allow;
728
729 error = vfs_getopt(opts, "name", (void **)&name, &len);
730 if (error == ENOENT)
731 name = NULL;
732 else if (error != 0)
733 goto done_free;
734 else {
735 if (len == 0 || name[len - 1] != '\0') {
736 error = EINVAL;
737 goto done_free;
738 }
739 if (len > MAXHOSTNAMELEN) {
740 error = ENAMETOOLONG;
741 goto done_free;
742 }
743 }
744
745 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
746 if (error == ENOENT)
747 host = NULL;
748 else if (error != 0)
749 goto done_free;
750 else {
751 ch_flags |= PR_HOST;
752 pr_flags |= PR_HOST;
753 if (len == 0 || host[len - 1] != '\0') {
754 error = EINVAL;
755 goto done_free;
756 }
757 if (len > MAXHOSTNAMELEN) {
758 error = ENAMETOOLONG;
759 goto done_free;
760 }
761 }
762
763 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
764 if (error == ENOENT)
765 domain = NULL;
766 else if (error != 0)
767 goto done_free;
768 else {
769 ch_flags |= PR_HOST;
770 pr_flags |= PR_HOST;
771 if (len == 0 || domain[len - 1] != '\0') {
772 error = EINVAL;
773 goto done_free;
774 }
775 if (len > MAXHOSTNAMELEN) {
776 error = ENAMETOOLONG;
777 goto done_free;
778 }
779 }
780
781 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
782 if (error == ENOENT)
783 uuid = NULL;
784 else if (error != 0)
785 goto done_free;
786 else {
787 ch_flags |= PR_HOST;
788 pr_flags |= PR_HOST;
789 if (len == 0 || uuid[len - 1] != '\0') {
790 error = EINVAL;
791 goto done_free;
792 }
793 if (len > HOSTUUIDLEN) {
794 error = ENAMETOOLONG;
795 goto done_free;
796 }
797 }
798
799 #ifdef COMPAT_FREEBSD32
800 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
801 uint32_t hid32;
802
803 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
804 hid = hid32;
805 } else
806 #endif
807 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
808 if (error == ENOENT)
809 gothid = 0;
810 else if (error != 0)
811 goto done_free;
812 else {
813 gothid = 1;
814 ch_flags |= PR_HOST;
815 pr_flags |= PR_HOST;
816 }
817
818 #ifdef INET
819 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
820 if (error == ENOENT)
821 ip4s = 0;
822 else if (error != 0)
823 goto done_free;
824 else if (ip4s & (sizeof(*ip4) - 1)) {
825 error = EINVAL;
826 goto done_free;
827 } else {
828 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
829 if (ip4s == 0)
830 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
831 else {
832 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
833 ip4s /= sizeof(*ip4);
834 if (ip4s > jail_max_af_ips) {
835 error = EINVAL;
836 vfs_opterror(opts, "too many IPv4 addresses");
837 goto done_errmsg;
838 }
839 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
840 bcopy(op, ip4, ip4s * sizeof(*ip4));
841 /*
842 * IP addresses are all sorted but ip[0] to preserve
843 * the primary IP address as given from userland.
844 * This special IP is used for unbound outgoing
845 * connections as well for "loopback" traffic in case
846 * source address selection cannot find any more fitting
847 * address to connect from.
848 */
849 if (ip4s > 1)
850 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
851 /*
852 * Check for duplicate addresses and do some simple
853 * zero and broadcast checks. If users give other bogus
854 * addresses it is their problem.
855 *
856 * We do not have to care about byte order for these
857 * checks so we will do them in NBO.
858 */
859 for (ii = 0; ii < ip4s; ii++) {
860 if (ip4[ii].s_addr == INADDR_ANY ||
861 ip4[ii].s_addr == INADDR_BROADCAST) {
862 error = EINVAL;
863 goto done_free;
864 }
865 if ((ii+1) < ip4s &&
866 (ip4[0].s_addr == ip4[ii+1].s_addr ||
867 ip4[ii].s_addr == ip4[ii+1].s_addr)) {
868 error = EINVAL;
869 goto done_free;
870 }
871 }
872 }
873 }
874 #endif
875
876 #ifdef INET6
877 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
878 if (error == ENOENT)
879 ip6s = 0;
880 else if (error != 0)
881 goto done_free;
882 else if (ip6s & (sizeof(*ip6) - 1)) {
883 error = EINVAL;
884 goto done_free;
885 } else {
886 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
887 if (ip6s == 0)
888 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
889 else {
890 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
891 ip6s /= sizeof(*ip6);
892 if (ip6s > jail_max_af_ips) {
893 error = EINVAL;
894 vfs_opterror(opts, "too many IPv6 addresses");
895 goto done_errmsg;
896 }
897 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
898 bcopy(op, ip6, ip6s * sizeof(*ip6));
899 if (ip6s > 1)
900 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
901 for (ii = 0; ii < ip6s; ii++) {
902 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
903 error = EINVAL;
904 goto done_free;
905 }
906 if ((ii+1) < ip6s &&
907 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
908 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
909 {
910 error = EINVAL;
911 goto done_free;
912 }
913 }
914 }
915 }
916 #endif
917
918 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
919 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
920 error = EINVAL;
921 vfs_opterror(opts,
922 "vnet jails cannot have IP address restrictions");
923 goto done_errmsg;
924 }
925 #endif
926
927 fullpath_disabled = 0;
928 root = NULL;
929 error = vfs_getopt(opts, "path", (void **)&path, &len);
930 if (error == ENOENT)
931 path = NULL;
932 else if (error != 0)
933 goto done_free;
934 else {
935 if (flags & JAIL_UPDATE) {
936 error = EINVAL;
937 vfs_opterror(opts,
938 "path cannot be changed after creation");
939 goto done_errmsg;
940 }
941 if (len == 0 || path[len - 1] != '\0') {
942 error = EINVAL;
943 goto done_free;
944 }
945 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
946 path, td);
947 error = namei(&nd);
948 if (error)
949 goto done_free;
950 root = nd.ni_vp;
951 NDFREE(&nd, NDF_ONLY_PNBUF);
952 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
953 strlcpy(g_path, path, MAXPATHLEN);
954 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
955 if (error == 0)
956 path = g_path;
957 else if (error == ENODEV) {
958 /* proceed if sysctl debug.disablefullpath == 1 */
959 fullpath_disabled = 1;
960 if (len < 2 || (len == 2 && path[0] == '/'))
961 path = NULL;
962 } else {
963 /* exit on other errors */
964 goto done_free;
965 }
966 if (root->v_type != VDIR) {
967 error = ENOTDIR;
968 vput(root);
969 goto done_free;
970 }
971 VOP_UNLOCK(root, 0);
972 if (fullpath_disabled) {
973 /* Leave room for a real-root full pathname. */
974 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
975 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
976 error = ENAMETOOLONG;
977 goto done_free;
978 }
979 }
980 }
981
982 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len);
983 if (error == ENOENT)
984 osrelstr = NULL;
985 else if (error != 0)
986 goto done_free;
987 else {
988 if (flags & JAIL_UPDATE) {
989 error = EINVAL;
990 vfs_opterror(opts,
991 "osrelease cannot be changed after creation");
992 goto done_errmsg;
993 }
994 if (len == 0 || len >= OSRELEASELEN) {
995 error = EINVAL;
996 vfs_opterror(opts,
997 "osrelease string must be 1-%d bytes long",
998 OSRELEASELEN - 1);
999 goto done_errmsg;
1000 }
1001 }
1002
1003 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt));
1004 if (error == ENOENT)
1005 osreldt = 0;
1006 else if (error != 0)
1007 goto done_free;
1008 else {
1009 if (flags & JAIL_UPDATE) {
1010 error = EINVAL;
1011 vfs_opterror(opts,
1012 "osreldate cannot be changed after creation");
1013 goto done_errmsg;
1014 }
1015 if (osreldt == 0) {
1016 error = EINVAL;
1017 vfs_opterror(opts, "osreldate cannot be 0");
1018 goto done_errmsg;
1019 }
1020 }
1021
1022 /*
1023 * Grab the allprison lock before letting modules check their
1024 * parameters. Once we have it, do not let go so we'll have a
1025 * consistent view of the OSD list.
1026 */
1027 sx_xlock(&allprison_lock);
1028 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
1029 if (error)
1030 goto done_unlock_list;
1031
1032 /* By now, all parameters should have been noted. */
1033 TAILQ_FOREACH(opt, opts, link) {
1034 if (!opt->seen && strcmp(opt->name, "errmsg")) {
1035 error = EINVAL;
1036 vfs_opterror(opts, "unknown parameter: %s", opt->name);
1037 goto done_unlock_list;
1038 }
1039 }
1040
1041 /*
1042 * See if we are creating a new record or updating an existing one.
1043 * This abuses the file error codes ENOENT and EEXIST.
1044 */
1045 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
1046 if (!cuflags) {
1047 error = EINVAL;
1048 vfs_opterror(opts, "no valid operation (create or update)");
1049 goto done_unlock_list;
1050 }
1051 pr = NULL;
1052 namelc = NULL;
1053 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
1054 namelc = strrchr(name, '.');
1055 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
1056 if (*p != '\0')
1057 jid = 0;
1058 }
1059 if (jid != 0) {
1060 /*
1061 * See if a requested jid already exists. There is an
1062 * information leak here if the jid exists but is not within
1063 * the caller's jail hierarchy. Jail creators will get EEXIST
1064 * even though they cannot see the jail, and CREATE | UPDATE
1065 * will return ENOENT which is not normally a valid error.
1066 */
1067 if (jid < 0) {
1068 error = EINVAL;
1069 vfs_opterror(opts, "negative jid");
1070 goto done_unlock_list;
1071 }
1072 pr = prison_find(jid);
1073 if (pr != NULL) {
1074 ppr = pr->pr_parent;
1075 /* Create: jid must not exist. */
1076 if (cuflags == JAIL_CREATE) {
1077 mtx_unlock(&pr->pr_mtx);
1078 error = EEXIST;
1079 vfs_opterror(opts, "jail %d already exists",
1080 jid);
1081 goto done_unlock_list;
1082 }
1083 if (!prison_ischild(mypr, pr)) {
1084 mtx_unlock(&pr->pr_mtx);
1085 pr = NULL;
1086 } else if (pr->pr_uref == 0) {
1087 if (!(flags & JAIL_DYING)) {
1088 mtx_unlock(&pr->pr_mtx);
1089 error = ENOENT;
1090 vfs_opterror(opts, "jail %d is dying",
1091 jid);
1092 goto done_unlock_list;
1093 } else if ((flags & JAIL_ATTACH) ||
1094 (pr_flags & PR_PERSIST)) {
1095 /*
1096 * A dying jail might be resurrected
1097 * (via attach or persist), but first
1098 * it must determine if another jail
1099 * has claimed its name. Accomplish
1100 * this by implicitly re-setting the
1101 * name.
1102 */
1103 if (name == NULL)
1104 name = prison_name(mypr, pr);
1105 }
1106 }
1107 }
1108 if (pr == NULL) {
1109 /* Update: jid must exist. */
1110 if (cuflags == JAIL_UPDATE) {
1111 error = ENOENT;
1112 vfs_opterror(opts, "jail %d not found", jid);
1113 goto done_unlock_list;
1114 }
1115 }
1116 }
1117 /*
1118 * If the caller provided a name, look for a jail by that name.
1119 * This has different semantics for creates and updates keyed by jid
1120 * (where the name must not already exist in a different jail),
1121 * and updates keyed by the name itself (where the name must exist
1122 * because that is the jail being updated).
1123 */
1124 if (name != NULL) {
1125 namelc = strrchr(name, '.');
1126 if (namelc == NULL)
1127 namelc = name;
1128 else {
1129 /*
1130 * This is a hierarchical name. Split it into the
1131 * parent and child names, and make sure the parent
1132 * exists or matches an already found jail.
1133 */
1134 *namelc = '\0';
1135 if (pr != NULL) {
1136 if (strncmp(name, ppr->pr_name, namelc - name)
1137 || ppr->pr_name[namelc - name] != '\0') {
1138 mtx_unlock(&pr->pr_mtx);
1139 error = EINVAL;
1140 vfs_opterror(opts,
1141 "cannot change jail's parent");
1142 goto done_unlock_list;
1143 }
1144 } else {
1145 ppr = prison_find_name(mypr, name);
1146 if (ppr == NULL) {
1147 error = ENOENT;
1148 vfs_opterror(opts,
1149 "jail \"%s\" not found", name);
1150 goto done_unlock_list;
1151 }
1152 mtx_unlock(&ppr->pr_mtx);
1153 }
1154 name = ++namelc;
1155 }
1156 if (name[0] != '\0') {
1157 namelen =
1158 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1159 name_again:
1160 deadpr = NULL;
1161 FOREACH_PRISON_CHILD(ppr, tpr) {
1162 if (tpr != pr && tpr->pr_ref > 0 &&
1163 !strcmp(tpr->pr_name + namelen, name)) {
1164 if (pr == NULL &&
1165 cuflags != JAIL_CREATE) {
1166 mtx_lock(&tpr->pr_mtx);
1167 if (tpr->pr_ref > 0) {
1168 /*
1169 * Use this jail
1170 * for updates.
1171 */
1172 if (tpr->pr_uref > 0) {
1173 pr = tpr;
1174 break;
1175 }
1176 deadpr = tpr;
1177 }
1178 mtx_unlock(&tpr->pr_mtx);
1179 } else if (tpr->pr_uref > 0) {
1180 /*
1181 * Create, or update(jid):
1182 * name must not exist in an
1183 * active sibling jail.
1184 */
1185 error = EEXIST;
1186 if (pr != NULL)
1187 mtx_unlock(&pr->pr_mtx);
1188 vfs_opterror(opts,
1189 "jail \"%s\" already exists",
1190 name);
1191 goto done_unlock_list;
1192 }
1193 }
1194 }
1195 /* If no active jail is found, use a dying one. */
1196 if (deadpr != NULL && pr == NULL) {
1197 if (flags & JAIL_DYING) {
1198 mtx_lock(&deadpr->pr_mtx);
1199 if (deadpr->pr_ref == 0) {
1200 mtx_unlock(&deadpr->pr_mtx);
1201 goto name_again;
1202 }
1203 pr = deadpr;
1204 } else if (cuflags == JAIL_UPDATE) {
1205 error = ENOENT;
1206 vfs_opterror(opts,
1207 "jail \"%s\" is dying", name);
1208 goto done_unlock_list;
1209 }
1210 }
1211 /* Update: name must exist if no jid. */
1212 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1213 error = ENOENT;
1214 vfs_opterror(opts, "jail \"%s\" not found",
1215 name);
1216 goto done_unlock_list;
1217 }
1218 }
1219 }
1220 /* Update: must provide a jid or name. */
1221 else if (cuflags == JAIL_UPDATE && pr == NULL) {
1222 error = ENOENT;
1223 vfs_opterror(opts, "update specified no jail");
1224 goto done_unlock_list;
1225 }
1226
1227 /* If there's no prison to update, create a new one and link it in. */
1228 if (pr == NULL) {
1229 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1230 if (tpr->pr_childcount >= tpr->pr_childmax) {
1231 error = EPERM;
1232 vfs_opterror(opts, "prison limit exceeded");
1233 goto done_unlock_list;
1234 }
1235 created = 1;
1236 mtx_lock(&ppr->pr_mtx);
1237 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1238 mtx_unlock(&ppr->pr_mtx);
1239 error = ENOENT;
1240 vfs_opterror(opts, "parent jail went away!");
1241 goto done_unlock_list;
1242 }
1243 ppr->pr_ref++;
1244 ppr->pr_uref++;
1245 mtx_unlock(&ppr->pr_mtx);
1246 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1247 if (jid == 0) {
1248 /* Find the next free jid. */
1249 jid = lastprid + 1;
1250 findnext:
1251 if (jid == JAIL_MAX)
1252 jid = 1;
1253 TAILQ_FOREACH(tpr, &allprison, pr_list) {
1254 if (tpr->pr_id < jid)
1255 continue;
1256 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1257 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1258 break;
1259 }
1260 if (jid == lastprid) {
1261 error = EAGAIN;
1262 vfs_opterror(opts,
1263 "no available jail IDs");
1264 free(pr, M_PRISON);
1265 prison_deref(ppr, PD_DEREF |
1266 PD_DEUREF | PD_LIST_XLOCKED);
1267 goto done_releroot;
1268 }
1269 jid++;
1270 goto findnext;
1271 }
1272 lastprid = jid;
1273 } else {
1274 /*
1275 * The jail already has a jid (that did not yet exist),
1276 * so just find where to insert it.
1277 */
1278 TAILQ_FOREACH(tpr, &allprison, pr_list)
1279 if (tpr->pr_id >= jid) {
1280 TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1281 break;
1282 }
1283 }
1284 if (tpr == NULL)
1285 TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1286 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1287 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1288 tpr->pr_childcount++;
1289
1290 pr->pr_parent = ppr;
1291 pr->pr_id = jid;
1292
1293 /* Set some default values, and inherit some from the parent. */
1294 if (name == NULL)
1295 name = "";
1296 if (path == NULL) {
1297 path = "/";
1298 root = mypr->pr_root;
1299 vref(root);
1300 }
1301 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1302 pr->pr_flags |= PR_HOST;
1303 #if defined(INET) || defined(INET6)
1304 #ifdef VIMAGE
1305 if (!(pr_flags & PR_VNET))
1306 #endif
1307 {
1308 #ifdef INET
1309 if (!(ch_flags & PR_IP4_USER))
1310 pr->pr_flags |=
1311 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1312 else if (!(pr_flags & PR_IP4_USER)) {
1313 pr->pr_flags |= ppr->pr_flags & PR_IP4;
1314 if (ppr->pr_ip4 != NULL) {
1315 pr->pr_ip4s = ppr->pr_ip4s;
1316 pr->pr_ip4 = malloc(pr->pr_ip4s *
1317 sizeof(struct in_addr), M_PRISON,
1318 M_WAITOK);
1319 bcopy(ppr->pr_ip4, pr->pr_ip4,
1320 pr->pr_ip4s * sizeof(*pr->pr_ip4));
1321 }
1322 }
1323 #endif
1324 #ifdef INET6
1325 if (!(ch_flags & PR_IP6_USER))
1326 pr->pr_flags |=
1327 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1328 else if (!(pr_flags & PR_IP6_USER)) {
1329 pr->pr_flags |= ppr->pr_flags & PR_IP6;
1330 if (ppr->pr_ip6 != NULL) {
1331 pr->pr_ip6s = ppr->pr_ip6s;
1332 pr->pr_ip6 = malloc(pr->pr_ip6s *
1333 sizeof(struct in6_addr), M_PRISON,
1334 M_WAITOK);
1335 bcopy(ppr->pr_ip6, pr->pr_ip6,
1336 pr->pr_ip6s * sizeof(*pr->pr_ip6));
1337 }
1338 }
1339 #endif
1340 }
1341 #endif
1342 /* Source address selection is always on by default. */
1343 pr->pr_flags |= _PR_IP_SADDRSEL;
1344
1345 pr->pr_securelevel = ppr->pr_securelevel;
1346 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1347 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1348 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1349
1350 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate;
1351 if (osrelstr == NULL)
1352 strcpy(pr->pr_osrelease, ppr->pr_osrelease);
1353 else
1354 strcpy(pr->pr_osrelease, osrelstr);
1355
1356 LIST_INIT(&pr->pr_children);
1357 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1358
1359 #ifdef VIMAGE
1360 /* Allocate a new vnet if specified. */
1361 pr->pr_vnet = (pr_flags & PR_VNET)
1362 ? vnet_alloc() : ppr->pr_vnet;
1363 #endif
1364 /*
1365 * Allocate a dedicated cpuset for each jail.
1366 * Unlike other initial settings, this may return an erorr.
1367 */
1368 error = cpuset_create_root(ppr, &pr->pr_cpuset);
1369 if (error) {
1370 prison_deref(pr, PD_LIST_XLOCKED);
1371 goto done_releroot;
1372 }
1373
1374 mtx_lock(&pr->pr_mtx);
1375 /*
1376 * New prisons do not yet have a reference, because we do not
1377 * want other to see the incomplete prison once the
1378 * allprison_lock is downgraded.
1379 */
1380 } else {
1381 created = 0;
1382 /*
1383 * Grab a reference for existing prisons, to ensure they
1384 * continue to exist for the duration of the call.
1385 */
1386 pr->pr_ref++;
1387 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1388 if ((pr->pr_flags & PR_VNET) &&
1389 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1390 error = EINVAL;
1391 vfs_opterror(opts,
1392 "vnet jails cannot have IP address restrictions");
1393 goto done_deref_locked;
1394 }
1395 #endif
1396 #ifdef INET
1397 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1398 error = EINVAL;
1399 vfs_opterror(opts,
1400 "ip4 cannot be changed after creation");
1401 goto done_deref_locked;
1402 }
1403 #endif
1404 #ifdef INET6
1405 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1406 error = EINVAL;
1407 vfs_opterror(opts,
1408 "ip6 cannot be changed after creation");
1409 goto done_deref_locked;
1410 }
1411 #endif
1412 }
1413
1414 /* Do final error checking before setting anything. */
1415 if (gotslevel) {
1416 if (slevel < ppr->pr_securelevel) {
1417 error = EPERM;
1418 goto done_deref_locked;
1419 }
1420 }
1421 if (gotchildmax) {
1422 if (childmax >= ppr->pr_childmax) {
1423 error = EPERM;
1424 goto done_deref_locked;
1425 }
1426 }
1427 if (gotenforce) {
1428 if (enforce < ppr->pr_enforce_statfs) {
1429 error = EPERM;
1430 goto done_deref_locked;
1431 }
1432 }
1433 if (gotrsnum) {
1434 /*
1435 * devfs_rsnum is a uint16_t
1436 */
1437 if (rsnum < 0 || rsnum > 65535) {
1438 error = EINVAL;
1439 goto done_deref_locked;
1440 }
1441 /*
1442 * Nested jails always inherit parent's devfs ruleset
1443 */
1444 if (jailed(td->td_ucred)) {
1445 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1446 error = EPERM;
1447 goto done_deref_locked;
1448 } else
1449 rsnum = ppr->pr_devfs_rsnum;
1450 }
1451 }
1452 #ifdef INET
1453 if (ip4s > 0) {
1454 if (ppr->pr_flags & PR_IP4) {
1455 /*
1456 * Make sure the new set of IP addresses is a
1457 * subset of the parent's list. Don't worry
1458 * about the parent being unlocked, as any
1459 * setting is done with allprison_lock held.
1460 */
1461 for (ij = 0; ij < ppr->pr_ip4s; ij++)
1462 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1463 break;
1464 if (ij == ppr->pr_ip4s) {
1465 error = EPERM;
1466 goto done_deref_locked;
1467 }
1468 if (ip4s > 1) {
1469 for (ii = ij = 1; ii < ip4s; ii++) {
1470 if (ip4[ii].s_addr ==
1471 ppr->pr_ip4[0].s_addr)
1472 continue;
1473 for (; ij < ppr->pr_ip4s; ij++)
1474 if (ip4[ii].s_addr ==
1475 ppr->pr_ip4[ij].s_addr)
1476 break;
1477 if (ij == ppr->pr_ip4s)
1478 break;
1479 }
1480 if (ij == ppr->pr_ip4s) {
1481 error = EPERM;
1482 goto done_deref_locked;
1483 }
1484 }
1485 }
1486 /*
1487 * Check for conflicting IP addresses. We permit them
1488 * if there is no more than one IP on each jail. If
1489 * there is a duplicate on a jail with more than one
1490 * IP stop checking and return error.
1491 */
1492 tppr = ppr;
1493 #ifdef VIMAGE
1494 for (; tppr != &prison0; tppr = tppr->pr_parent)
1495 if (tppr->pr_flags & PR_VNET)
1496 break;
1497 #endif
1498 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1499 if (tpr == pr ||
1500 #ifdef VIMAGE
1501 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1502 #endif
1503 tpr->pr_uref == 0) {
1504 descend = 0;
1505 continue;
1506 }
1507 if (!(tpr->pr_flags & PR_IP4_USER))
1508 continue;
1509 descend = 0;
1510 if (tpr->pr_ip4 == NULL ||
1511 (ip4s == 1 && tpr->pr_ip4s == 1))
1512 continue;
1513 for (ii = 0; ii < ip4s; ii++) {
1514 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1515 error = EADDRINUSE;
1516 vfs_opterror(opts,
1517 "IPv4 addresses clash");
1518 goto done_deref_locked;
1519 }
1520 }
1521 }
1522 }
1523 #endif
1524 #ifdef INET6
1525 if (ip6s > 0) {
1526 if (ppr->pr_flags & PR_IP6) {
1527 /*
1528 * Make sure the new set of IP addresses is a
1529 * subset of the parent's list.
1530 */
1531 for (ij = 0; ij < ppr->pr_ip6s; ij++)
1532 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1533 &ppr->pr_ip6[ij]))
1534 break;
1535 if (ij == ppr->pr_ip6s) {
1536 error = EPERM;
1537 goto done_deref_locked;
1538 }
1539 if (ip6s > 1) {
1540 for (ii = ij = 1; ii < ip6s; ii++) {
1541 if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1542 &ppr->pr_ip6[0]))
1543 continue;
1544 for (; ij < ppr->pr_ip6s; ij++)
1545 if (IN6_ARE_ADDR_EQUAL(
1546 &ip6[ii], &ppr->pr_ip6[ij]))
1547 break;
1548 if (ij == ppr->pr_ip6s)
1549 break;
1550 }
1551 if (ij == ppr->pr_ip6s) {
1552 error = EPERM;
1553 goto done_deref_locked;
1554 }
1555 }
1556 }
1557 /* Check for conflicting IP addresses. */
1558 tppr = ppr;
1559 #ifdef VIMAGE
1560 for (; tppr != &prison0; tppr = tppr->pr_parent)
1561 if (tppr->pr_flags & PR_VNET)
1562 break;
1563 #endif
1564 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1565 if (tpr == pr ||
1566 #ifdef VIMAGE
1567 (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1568 #endif
1569 tpr->pr_uref == 0) {
1570 descend = 0;
1571 continue;
1572 }
1573 if (!(tpr->pr_flags & PR_IP6_USER))
1574 continue;
1575 descend = 0;
1576 if (tpr->pr_ip6 == NULL ||
1577 (ip6s == 1 && tpr->pr_ip6s == 1))
1578 continue;
1579 for (ii = 0; ii < ip6s; ii++) {
1580 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1581 error = EADDRINUSE;
1582 vfs_opterror(opts,
1583 "IPv6 addresses clash");
1584 goto done_deref_locked;
1585 }
1586 }
1587 }
1588 }
1589 #endif
1590 onamelen = namelen = 0;
1591 if (name != NULL) {
1592 /* Give a default name of the jid. Also allow the name to be
1593 * explicitly the jid - but not any other number, and only in
1594 * normal form (no leading zero/etc).
1595 */
1596 if (name[0] == '\0')
1597 snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1598 else if ((strtoul(namelc, &p, 10) != jid ||
1599 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1600 error = EINVAL;
1601 vfs_opterror(opts,
1602 "name cannot be numeric (unless it is the jid)");
1603 goto done_deref_locked;
1604 }
1605 /*
1606 * Make sure the name isn't too long for the prison or its
1607 * children.
1608 */
1609 onamelen = strlen(pr->pr_name);
1610 namelen = strlen(name);
1611 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1612 error = ENAMETOOLONG;
1613 goto done_deref_locked;
1614 }
1615 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1616 if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1617 sizeof(pr->pr_name)) {
1618 error = ENAMETOOLONG;
1619 goto done_deref_locked;
1620 }
1621 }
1622 }
1623 if (pr_allow & ~ppr->pr_allow) {
1624 error = EPERM;
1625 goto done_deref_locked;
1626 }
1627
1628 /* Set the parameters of the prison. */
1629 #ifdef INET
1630 redo_ip4 = 0;
1631 if (pr_flags & PR_IP4_USER) {
1632 pr->pr_flags |= PR_IP4;
1633 free(pr->pr_ip4, M_PRISON);
1634 pr->pr_ip4s = ip4s;
1635 pr->pr_ip4 = ip4;
1636 ip4 = NULL;
1637 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1638 #ifdef VIMAGE
1639 if (tpr->pr_flags & PR_VNET) {
1640 descend = 0;
1641 continue;
1642 }
1643 #endif
1644 if (prison_restrict_ip4(tpr, NULL)) {
1645 redo_ip4 = 1;
1646 descend = 0;
1647 }
1648 }
1649 }
1650 #endif
1651 #ifdef INET6
1652 redo_ip6 = 0;
1653 if (pr_flags & PR_IP6_USER) {
1654 pr->pr_flags |= PR_IP6;
1655 free(pr->pr_ip6, M_PRISON);
1656 pr->pr_ip6s = ip6s;
1657 pr->pr_ip6 = ip6;
1658 ip6 = NULL;
1659 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1660 #ifdef VIMAGE
1661 if (tpr->pr_flags & PR_VNET) {
1662 descend = 0;
1663 continue;
1664 }
1665 #endif
1666 if (prison_restrict_ip6(tpr, NULL)) {
1667 redo_ip6 = 1;
1668 descend = 0;
1669 }
1670 }
1671 }
1672 #endif
1673 if (gotslevel) {
1674 pr->pr_securelevel = slevel;
1675 /* Set all child jails to be at least this level. */
1676 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1677 if (tpr->pr_securelevel < slevel)
1678 tpr->pr_securelevel = slevel;
1679 }
1680 if (gotchildmax) {
1681 pr->pr_childmax = childmax;
1682 /* Set all child jails to under this limit. */
1683 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1684 if (tpr->pr_childmax > childmax - level)
1685 tpr->pr_childmax = childmax > level
1686 ? childmax - level : 0;
1687 }
1688 if (gotenforce) {
1689 pr->pr_enforce_statfs = enforce;
1690 /* Pass this restriction on to the children. */
1691 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1692 if (tpr->pr_enforce_statfs < enforce)
1693 tpr->pr_enforce_statfs = enforce;
1694 }
1695 if (gotrsnum) {
1696 pr->pr_devfs_rsnum = rsnum;
1697 /* Pass this restriction on to the children. */
1698 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1699 tpr->pr_devfs_rsnum = rsnum;
1700 }
1701 if (name != NULL) {
1702 if (ppr == &prison0)
1703 strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1704 else
1705 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1706 ppr->pr_name, name);
1707 /* Change this component of child names. */
1708 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1709 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1710 strlen(tpr->pr_name + onamelen) + 1);
1711 bcopy(pr->pr_name, tpr->pr_name, namelen);
1712 }
1713 }
1714 if (path != NULL) {
1715 /* Try to keep a real-rooted full pathname. */
1716 if (fullpath_disabled && path[0] == '/' &&
1717 strcmp(mypr->pr_path, "/"))
1718 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1719 mypr->pr_path, path);
1720 else
1721 strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1722 pr->pr_root = root;
1723 }
1724 if (PR_HOST & ch_flags & ~pr_flags) {
1725 if (pr->pr_flags & PR_HOST) {
1726 /*
1727 * Copy the parent's host info. As with pr_ip4 above,
1728 * the lack of a lock on the parent is not a problem;
1729 * it is always set with allprison_lock at least
1730 * shared, and is held exclusively here.
1731 */
1732 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1733 sizeof(pr->pr_hostname));
1734 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1735 sizeof(pr->pr_domainname));
1736 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1737 sizeof(pr->pr_hostuuid));
1738 pr->pr_hostid = pr->pr_parent->pr_hostid;
1739 }
1740 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1741 /* Set this prison, and any descendants without PR_HOST. */
1742 if (host != NULL)
1743 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1744 if (domain != NULL)
1745 strlcpy(pr->pr_domainname, domain,
1746 sizeof(pr->pr_domainname));
1747 if (uuid != NULL)
1748 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1749 if (gothid)
1750 pr->pr_hostid = hid;
1751 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1752 if (tpr->pr_flags & PR_HOST)
1753 descend = 0;
1754 else {
1755 if (host != NULL)
1756 strlcpy(tpr->pr_hostname,
1757 pr->pr_hostname,
1758 sizeof(tpr->pr_hostname));
1759 if (domain != NULL)
1760 strlcpy(tpr->pr_domainname,
1761 pr->pr_domainname,
1762 sizeof(tpr->pr_domainname));
1763 if (uuid != NULL)
1764 strlcpy(tpr->pr_hostuuid,
1765 pr->pr_hostuuid,
1766 sizeof(tpr->pr_hostuuid));
1767 if (gothid)
1768 tpr->pr_hostid = hid;
1769 }
1770 }
1771 }
1772 if ((tallow = ch_allow & ~pr_allow)) {
1773 /* Clear allow bits in all children. */
1774 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1775 tpr->pr_allow &= ~tallow;
1776 }
1777 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1778 /*
1779 * Persistent prisons get an extra reference, and prisons losing their
1780 * persist flag lose that reference. Only do this for existing prisons
1781 * for now, so new ones will remain unseen until after the module
1782 * handlers have completed.
1783 */
1784 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1785 if (pr_flags & PR_PERSIST) {
1786 pr->pr_ref++;
1787 pr->pr_uref++;
1788 } else {
1789 pr->pr_ref--;
1790 pr->pr_uref--;
1791 }
1792 }
1793 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1794 mtx_unlock(&pr->pr_mtx);
1795
1796 #ifdef RACCT
1797 if (racct_enable && created)
1798 prison_racct_attach(pr);
1799 #endif
1800
1801 /* Locks may have prevented a complete restriction of child IP
1802 * addresses. If so, allocate some more memory and try again.
1803 */
1804 #ifdef INET
1805 while (redo_ip4) {
1806 ip4s = pr->pr_ip4s;
1807 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1808 mtx_lock(&pr->pr_mtx);
1809 redo_ip4 = 0;
1810 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1811 #ifdef VIMAGE
1812 if (tpr->pr_flags & PR_VNET) {
1813 descend = 0;
1814 continue;
1815 }
1816 #endif
1817 if (prison_restrict_ip4(tpr, ip4)) {
1818 if (ip4 != NULL)
1819 ip4 = NULL;
1820 else
1821 redo_ip4 = 1;
1822 }
1823 }
1824 mtx_unlock(&pr->pr_mtx);
1825 }
1826 #endif
1827 #ifdef INET6
1828 while (redo_ip6) {
1829 ip6s = pr->pr_ip6s;
1830 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1831 mtx_lock(&pr->pr_mtx);
1832 redo_ip6 = 0;
1833 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1834 #ifdef VIMAGE
1835 if (tpr->pr_flags & PR_VNET) {
1836 descend = 0;
1837 continue;
1838 }
1839 #endif
1840 if (prison_restrict_ip6(tpr, ip6)) {
1841 if (ip6 != NULL)
1842 ip6 = NULL;
1843 else
1844 redo_ip6 = 1;
1845 }
1846 }
1847 mtx_unlock(&pr->pr_mtx);
1848 }
1849 #endif
1850
1851 /* Let the modules do their work. */
1852 sx_downgrade(&allprison_lock);
1853 if (created) {
1854 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1855 if (error) {
1856 prison_deref(pr, PD_LIST_SLOCKED);
1857 goto done_errmsg;
1858 }
1859 }
1860 error = osd_jail_call(pr, PR_METHOD_SET, opts);
1861 if (error) {
1862 prison_deref(pr, created
1863 ? PD_LIST_SLOCKED
1864 : PD_DEREF | PD_LIST_SLOCKED);
1865 goto done_errmsg;
1866 }
1867
1868 /* Attach this process to the prison if requested. */
1869 if (flags & JAIL_ATTACH) {
1870 mtx_lock(&pr->pr_mtx);
1871 error = do_jail_attach(td, pr);
1872 if (error) {
1873 vfs_opterror(opts, "attach failed");
1874 if (!created)
1875 prison_deref(pr, PD_DEREF);
1876 goto done_errmsg;
1877 }
1878 }
1879
1880 #ifdef RACCT
1881 if (racct_enable && !created) {
1882 if (!(flags & JAIL_ATTACH))
1883 sx_sunlock(&allprison_lock);
1884 prison_racct_modify(pr);
1885 if (!(flags & JAIL_ATTACH))
1886 sx_slock(&allprison_lock);
1887 }
1888 #endif
1889
1890 td->td_retval[0] = pr->pr_id;
1891
1892 /*
1893 * Now that it is all there, drop the temporary reference from existing
1894 * prisons. Or add a reference to newly created persistent prisons
1895 * (which was not done earlier so that the prison would not be publicly
1896 * visible).
1897 */
1898 if (!created) {
1899 prison_deref(pr, (flags & JAIL_ATTACH)
1900 ? PD_DEREF
1901 : PD_DEREF | PD_LIST_SLOCKED);
1902 } else {
1903 if (pr_flags & PR_PERSIST) {
1904 mtx_lock(&pr->pr_mtx);
1905 pr->pr_ref++;
1906 pr->pr_uref++;
1907 mtx_unlock(&pr->pr_mtx);
1908 }
1909 if (!(flags & JAIL_ATTACH))
1910 sx_sunlock(&allprison_lock);
1911 }
1912
1913 goto done_errmsg;
1914
1915 done_deref_locked:
1916 prison_deref(pr, created
1917 ? PD_LOCKED | PD_LIST_XLOCKED
1918 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1919 goto done_releroot;
1920 done_unlock_list:
1921 sx_xunlock(&allprison_lock);
1922 done_releroot:
1923 if (root != NULL)
1924 vrele(root);
1925 done_errmsg:
1926 if (error) {
1927 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1928 if (errmsg_len > 0) {
1929 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1930 if (errmsg_pos > 0) {
1931 if (optuio->uio_segflg == UIO_SYSSPACE)
1932 bcopy(errmsg,
1933 optuio->uio_iov[errmsg_pos].iov_base,
1934 errmsg_len);
1935 else
1936 copyout(errmsg,
1937 optuio->uio_iov[errmsg_pos].iov_base,
1938 errmsg_len);
1939 }
1940 }
1941 }
1942 done_free:
1943 #ifdef INET
1944 free(ip4, M_PRISON);
1945 #endif
1946 #ifdef INET6
1947 free(ip6, M_PRISON);
1948 #endif
1949 if (g_path != NULL)
1950 free(g_path, M_TEMP);
1951 vfs_freeopts(opts);
1952 return (error);
1953 }
1954
1955
1956 /*
1957 * struct jail_get_args {
1958 * struct iovec *iovp;
1959 * unsigned int iovcnt;
1960 * int flags;
1961 * };
1962 */
1963 int
1964 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1965 {
1966 struct uio *auio;
1967 int error;
1968
1969 /* Check that we have an even number of iovecs. */
1970 if (uap->iovcnt & 1)
1971 return (EINVAL);
1972
1973 error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1974 if (error)
1975 return (error);
1976 error = kern_jail_get(td, auio, uap->flags);
1977 if (error == 0)
1978 error = copyout(auio->uio_iov, uap->iovp,
1979 uap->iovcnt * sizeof (struct iovec));
1980 free(auio, M_IOV);
1981 return (error);
1982 }
1983
1984 int
1985 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1986 {
1987 struct prison *pr, *mypr;
1988 struct vfsopt *opt;
1989 struct vfsoptlist *opts;
1990 char *errmsg, *name;
1991 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1992
1993 if (flags & ~JAIL_GET_MASK)
1994 return (EINVAL);
1995
1996 /* Get the parameter list. */
1997 error = vfs_buildopts(optuio, &opts);
1998 if (error)
1999 return (error);
2000 errmsg_pos = vfs_getopt_pos(opts, "errmsg");
2001 mypr = td->td_ucred->cr_prison;
2002
2003 /*
2004 * Find the prison specified by one of: lastjid, jid, name.
2005 */
2006 sx_slock(&allprison_lock);
2007 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
2008 if (error == 0) {
2009 TAILQ_FOREACH(pr, &allprison, pr_list) {
2010 if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
2011 mtx_lock(&pr->pr_mtx);
2012 if (pr->pr_ref > 0 &&
2013 (pr->pr_uref > 0 || (flags & JAIL_DYING)))
2014 break;
2015 mtx_unlock(&pr->pr_mtx);
2016 }
2017 }
2018 if (pr != NULL)
2019 goto found_prison;
2020 error = ENOENT;
2021 vfs_opterror(opts, "no jail after %d", jid);
2022 goto done_unlock_list;
2023 } else if (error != ENOENT)
2024 goto done_unlock_list;
2025
2026 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
2027 if (error == 0) {
2028 if (jid != 0) {
2029 pr = prison_find_child(mypr, jid);
2030 if (pr != NULL) {
2031 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2032 mtx_unlock(&pr->pr_mtx);
2033 error = ENOENT;
2034 vfs_opterror(opts, "jail %d is dying",
2035 jid);
2036 goto done_unlock_list;
2037 }
2038 goto found_prison;
2039 }
2040 error = ENOENT;
2041 vfs_opterror(opts, "jail %d not found", jid);
2042 goto done_unlock_list;
2043 }
2044 } else if (error != ENOENT)
2045 goto done_unlock_list;
2046
2047 error = vfs_getopt(opts, "name", (void **)&name, &len);
2048 if (error == 0) {
2049 if (len == 0 || name[len - 1] != '\0') {
2050 error = EINVAL;
2051 goto done_unlock_list;
2052 }
2053 pr = prison_find_name(mypr, name);
2054 if (pr != NULL) {
2055 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
2056 mtx_unlock(&pr->pr_mtx);
2057 error = ENOENT;
2058 vfs_opterror(opts, "jail \"%s\" is dying",
2059 name);
2060 goto done_unlock_list;
2061 }
2062 goto found_prison;
2063 }
2064 error = ENOENT;
2065 vfs_opterror(opts, "jail \"%s\" not found", name);
2066 goto done_unlock_list;
2067 } else if (error != ENOENT)
2068 goto done_unlock_list;
2069
2070 vfs_opterror(opts, "no jail specified");
2071 error = ENOENT;
2072 goto done_unlock_list;
2073
2074 found_prison:
2075 /* Get the parameters of the prison. */
2076 pr->pr_ref++;
2077 locked = PD_LOCKED;
2078 td->td_retval[0] = pr->pr_id;
2079 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2080 if (error != 0 && error != ENOENT)
2081 goto done_deref;
2082 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2083 error = vfs_setopt(opts, "parent", &i, sizeof(i));
2084 if (error != 0 && error != ENOENT)
2085 goto done_deref;
2086 error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2087 if (error != 0 && error != ENOENT)
2088 goto done_deref;
2089 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2090 sizeof(pr->pr_cpuset->cs_id));
2091 if (error != 0 && error != ENOENT)
2092 goto done_deref;
2093 error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2094 if (error != 0 && error != ENOENT)
2095 goto done_deref;
2096 #ifdef INET
2097 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2098 pr->pr_ip4s * sizeof(*pr->pr_ip4));
2099 if (error != 0 && error != ENOENT)
2100 goto done_deref;
2101 #endif
2102 #ifdef INET6
2103 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2104 pr->pr_ip6s * sizeof(*pr->pr_ip6));
2105 if (error != 0 && error != ENOENT)
2106 goto done_deref;
2107 #endif
2108 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2109 sizeof(pr->pr_securelevel));
2110 if (error != 0 && error != ENOENT)
2111 goto done_deref;
2112 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2113 sizeof(pr->pr_childcount));
2114 if (error != 0 && error != ENOENT)
2115 goto done_deref;
2116 error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2117 sizeof(pr->pr_childmax));
2118 if (error != 0 && error != ENOENT)
2119 goto done_deref;
2120 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2121 if (error != 0 && error != ENOENT)
2122 goto done_deref;
2123 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2124 if (error != 0 && error != ENOENT)
2125 goto done_deref;
2126 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2127 if (error != 0 && error != ENOENT)
2128 goto done_deref;
2129 #ifdef COMPAT_FREEBSD32
2130 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2131 uint32_t hid32 = pr->pr_hostid;
2132
2133 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2134 } else
2135 #endif
2136 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2137 sizeof(pr->pr_hostid));
2138 if (error != 0 && error != ENOENT)
2139 goto done_deref;
2140 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2141 sizeof(pr->pr_enforce_statfs));
2142 if (error != 0 && error != ENOENT)
2143 goto done_deref;
2144 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2145 sizeof(pr->pr_devfs_rsnum));
2146 if (error != 0 && error != ENOENT)
2147 goto done_deref;
2148 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2149 fi++) {
2150 if (pr_flag_names[fi] == NULL)
2151 continue;
2152 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2153 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2154 if (error != 0 && error != ENOENT)
2155 goto done_deref;
2156 i = !i;
2157 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2158 if (error != 0 && error != ENOENT)
2159 goto done_deref;
2160 }
2161 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2162 fi++) {
2163 i = pr->pr_flags &
2164 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2165 i = pr_flag_jailsys[fi].disable &&
2166 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2167 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2168 : JAIL_SYS_INHERIT;
2169 error =
2170 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2171 if (error != 0 && error != ENOENT)
2172 goto done_deref;
2173 }
2174 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2175 fi++) {
2176 if (pr_allow_names[fi] == NULL)
2177 continue;
2178 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2179 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2180 if (error != 0 && error != ENOENT)
2181 goto done_deref;
2182 i = !i;
2183 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2184 if (error != 0 && error != ENOENT)
2185 goto done_deref;
2186 }
2187 i = (pr->pr_uref == 0);
2188 error = vfs_setopt(opts, "dying", &i, sizeof(i));
2189 if (error != 0 && error != ENOENT)
2190 goto done_deref;
2191 i = !i;
2192 error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2193 if (error != 0 && error != ENOENT)
2194 goto done_deref;
2195 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate,
2196 sizeof(pr->pr_osreldate));
2197 if (error != 0 && error != ENOENT)
2198 goto done_deref;
2199 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease);
2200 if (error != 0 && error != ENOENT)
2201 goto done_deref;
2202
2203 /* Get the module parameters. */
2204 mtx_unlock(&pr->pr_mtx);
2205 locked = 0;
2206 error = osd_jail_call(pr, PR_METHOD_GET, opts);
2207 if (error)
2208 goto done_deref;
2209 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2210
2211 /* By now, all parameters should have been noted. */
2212 TAILQ_FOREACH(opt, opts, link) {
2213 if (!opt->seen && strcmp(opt->name, "errmsg")) {
2214 error = EINVAL;
2215 vfs_opterror(opts, "unknown parameter: %s", opt->name);
2216 goto done_errmsg;
2217 }
2218 }
2219
2220 /* Write the fetched parameters back to userspace. */
2221 error = 0;
2222 TAILQ_FOREACH(opt, opts, link) {
2223 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2224 pos = 2 * opt->pos + 1;
2225 optuio->uio_iov[pos].iov_len = opt->len;
2226 if (opt->value != NULL) {
2227 if (optuio->uio_segflg == UIO_SYSSPACE) {
2228 bcopy(opt->value,
2229 optuio->uio_iov[pos].iov_base,
2230 opt->len);
2231 } else {
2232 error = copyout(opt->value,
2233 optuio->uio_iov[pos].iov_base,
2234 opt->len);
2235 if (error)
2236 break;
2237 }
2238 }
2239 }
2240 }
2241 goto done_errmsg;
2242
2243 done_deref:
2244 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2245 goto done_errmsg;
2246
2247 done_unlock_list:
2248 sx_sunlock(&allprison_lock);
2249 done_errmsg:
2250 if (error && errmsg_pos >= 0) {
2251 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2252 errmsg_pos = 2 * errmsg_pos + 1;
2253 if (errmsg_len > 0) {
2254 if (optuio->uio_segflg == UIO_SYSSPACE)
2255 bcopy(errmsg,
2256 optuio->uio_iov[errmsg_pos].iov_base,
2257 errmsg_len);
2258 else
2259 copyout(errmsg,
2260 optuio->uio_iov[errmsg_pos].iov_base,
2261 errmsg_len);
2262 }
2263 }
2264 vfs_freeopts(opts);
2265 return (error);
2266 }
2267
2268
2269 /*
2270 * struct jail_remove_args {
2271 * int jid;
2272 * };
2273 */
2274 int
2275 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2276 {
2277 struct prison *pr, *cpr, *lpr, *tpr;
2278 int descend, error;
2279
2280 error = priv_check(td, PRIV_JAIL_REMOVE);
2281 if (error)
2282 return (error);
2283
2284 sx_xlock(&allprison_lock);
2285 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2286 if (pr == NULL) {
2287 sx_xunlock(&allprison_lock);
2288 return (EINVAL);
2289 }
2290
2291 /* Remove all descendants of this prison, then remove this prison. */
2292 pr->pr_ref++;
2293 pr->pr_flags |= PR_REMOVE;
2294 if (!LIST_EMPTY(&pr->pr_children)) {
2295 mtx_unlock(&pr->pr_mtx);
2296 lpr = NULL;
2297 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2298 mtx_lock(&cpr->pr_mtx);
2299 if (cpr->pr_ref > 0) {
2300 tpr = cpr;
2301 cpr->pr_ref++;
2302 cpr->pr_flags |= PR_REMOVE;
2303 } else {
2304 /* Already removed - do not do it again. */
2305 tpr = NULL;
2306 }
2307 mtx_unlock(&cpr->pr_mtx);
2308 if (lpr != NULL) {
2309 mtx_lock(&lpr->pr_mtx);
2310 prison_remove_one(lpr);
2311 sx_xlock(&allprison_lock);
2312 }
2313 lpr = tpr;
2314 }
2315 if (lpr != NULL) {
2316 mtx_lock(&lpr->pr_mtx);
2317 prison_remove_one(lpr);
2318 sx_xlock(&allprison_lock);
2319 }
2320 mtx_lock(&pr->pr_mtx);
2321 }
2322 prison_remove_one(pr);
2323 return (0);
2324 }
2325
2326 static void
2327 prison_remove_one(struct prison *pr)
2328 {
2329 struct proc *p;
2330 int deuref;
2331
2332 /* If the prison was persistent, it is not anymore. */
2333 deuref = 0;
2334 if (pr->pr_flags & PR_PERSIST) {
2335 pr->pr_ref--;
2336 deuref = PD_DEUREF;
2337 pr->pr_flags &= ~PR_PERSIST;
2338 }
2339
2340 /*
2341 * jail_remove added a reference. If that's the only one, remove
2342 * the prison now.
2343 */
2344 KASSERT(pr->pr_ref > 0,
2345 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2346 if (pr->pr_ref == 1) {
2347 prison_deref(pr,
2348 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2349 return;
2350 }
2351
2352 mtx_unlock(&pr->pr_mtx);
2353 sx_xunlock(&allprison_lock);
2354 /*
2355 * Kill all processes unfortunate enough to be attached to this prison.
2356 */
2357 sx_slock(&allproc_lock);
2358 LIST_FOREACH(p, &allproc, p_list) {
2359 PROC_LOCK(p);
2360 if (p->p_state != PRS_NEW && p->p_ucred &&
2361 p->p_ucred->cr_prison == pr)
2362 kern_psignal(p, SIGKILL);
2363 PROC_UNLOCK(p);
2364 }
2365 sx_sunlock(&allproc_lock);
2366 /* Remove the temporary reference added by jail_remove. */
2367 prison_deref(pr, deuref | PD_DEREF);
2368 }
2369
2370
2371 /*
2372 * struct jail_attach_args {
2373 * int jid;
2374 * };
2375 */
2376 int
2377 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2378 {
2379 struct prison *pr;
2380 int error;
2381
2382 error = priv_check(td, PRIV_JAIL_ATTACH);
2383 if (error)
2384 return (error);
2385
2386 sx_slock(&allprison_lock);
2387 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2388 if (pr == NULL) {
2389 sx_sunlock(&allprison_lock);
2390 return (EINVAL);
2391 }
2392
2393 /*
2394 * Do not allow a process to attach to a prison that is not
2395 * considered to be "alive".
2396 */
2397 if (pr->pr_uref == 0) {
2398 mtx_unlock(&pr->pr_mtx);
2399 sx_sunlock(&allprison_lock);
2400 return (EINVAL);
2401 }
2402
2403 return (do_jail_attach(td, pr));
2404 }
2405
2406 static int
2407 do_jail_attach(struct thread *td, struct prison *pr)
2408 {
2409 struct prison *ppr;
2410 struct proc *p;
2411 struct ucred *newcred, *oldcred;
2412 int error;
2413
2414 /*
2415 * XXX: Note that there is a slight race here if two threads
2416 * in the same privileged process attempt to attach to two
2417 * different jails at the same time. It is important for
2418 * user processes not to do this, or they might end up with
2419 * a process root from one prison, but attached to the jail
2420 * of another.
2421 */
2422 pr->pr_ref++;
2423 pr->pr_uref++;
2424 mtx_unlock(&pr->pr_mtx);
2425
2426 /* Let modules do whatever they need to prepare for attaching. */
2427 error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2428 if (error) {
2429 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2430 return (error);
2431 }
2432 sx_sunlock(&allprison_lock);
2433
2434 /*
2435 * Reparent the newly attached process to this jail.
2436 */
2437 ppr = td->td_ucred->cr_prison;
2438 p = td->td_proc;
2439 error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2440 if (error)
2441 goto e_revert_osd;
2442
2443 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2444 if ((error = change_dir(pr->pr_root, td)) != 0)
2445 goto e_unlock;
2446 #ifdef MAC
2447 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2448 goto e_unlock;
2449 #endif
2450 VOP_UNLOCK(pr->pr_root, 0);
2451 if ((error = change_root(pr->pr_root, td)))
2452 goto e_revert_osd;
2453
2454 newcred = crget();
2455 PROC_LOCK(p);
2456 oldcred = p->p_ucred;
2457 setsugid(p);
2458 crcopy(newcred, oldcred);
2459 newcred->cr_prison = pr;
2460 p->p_ucred = newcred;
2461 PROC_UNLOCK(p);
2462 #ifdef RACCT
2463 racct_proc_ucred_changed(p, oldcred, newcred);
2464 #endif
2465 crfree(oldcred);
2466 prison_deref(ppr, PD_DEREF | PD_DEUREF);
2467 return (0);
2468 e_unlock:
2469 VOP_UNLOCK(pr->pr_root, 0);
2470 e_revert_osd:
2471 /* Tell modules this thread is still in its old jail after all. */
2472 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2473 prison_deref(pr, PD_DEREF | PD_DEUREF);
2474 return (error);
2475 }
2476
2477
2478 /*
2479 * Returns a locked prison instance, or NULL on failure.
2480 */
2481 struct prison *
2482 prison_find(int prid)
2483 {
2484 struct prison *pr;
2485
2486 sx_assert(&allprison_lock, SX_LOCKED);
2487 TAILQ_FOREACH(pr, &allprison, pr_list) {
2488 if (pr->pr_id == prid) {
2489 mtx_lock(&pr->pr_mtx);
2490 if (pr->pr_ref > 0)
2491 return (pr);
2492 mtx_unlock(&pr->pr_mtx);
2493 }
2494 }
2495 return (NULL);
2496 }
2497
2498 /*
2499 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2500 */
2501 struct prison *
2502 prison_find_child(struct prison *mypr, int prid)
2503 {
2504 struct prison *pr;
2505 int descend;
2506
2507 sx_assert(&allprison_lock, SX_LOCKED);
2508 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2509 if (pr->pr_id == prid) {
2510 mtx_lock(&pr->pr_mtx);
2511 if (pr->pr_ref > 0)
2512 return (pr);
2513 mtx_unlock(&pr->pr_mtx);
2514 }
2515 }
2516 return (NULL);
2517 }
2518
2519 /*
2520 * Look for the name relative to mypr. Returns a locked prison or NULL.
2521 */
2522 struct prison *
2523 prison_find_name(struct prison *mypr, const char *name)
2524 {
2525 struct prison *pr, *deadpr;
2526 size_t mylen;
2527 int descend;
2528
2529 sx_assert(&allprison_lock, SX_LOCKED);
2530 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2531 again:
2532 deadpr = NULL;
2533 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2534 if (!strcmp(pr->pr_name + mylen, name)) {
2535 mtx_lock(&pr->pr_mtx);
2536 if (pr->pr_ref > 0) {
2537 if (pr->pr_uref > 0)
2538 return (pr);
2539 deadpr = pr;
2540 }
2541 mtx_unlock(&pr->pr_mtx);
2542 }
2543 }
2544 /* There was no valid prison - perhaps there was a dying one. */
2545 if (deadpr != NULL) {
2546 mtx_lock(&deadpr->pr_mtx);
2547 if (deadpr->pr_ref == 0) {
2548 mtx_unlock(&deadpr->pr_mtx);
2549 goto again;
2550 }
2551 }
2552 return (deadpr);
2553 }
2554
2555 /*
2556 * See if a prison has the specific flag set.
2557 */
2558 int
2559 prison_flag(struct ucred *cred, unsigned flag)
2560 {
2561
2562 /* This is an atomic read, so no locking is necessary. */
2563 return (cred->cr_prison->pr_flags & flag);
2564 }
2565
2566 int
2567 prison_allow(struct ucred *cred, unsigned flag)
2568 {
2569
2570 /* This is an atomic read, so no locking is necessary. */
2571 return (cred->cr_prison->pr_allow & flag);
2572 }
2573
2574 /*
2575 * Remove a prison reference. If that was the last reference, remove the
2576 * prison itself - but not in this context in case there are locks held.
2577 */
2578 void
2579 prison_free_locked(struct prison *pr)
2580 {
2581
2582 mtx_assert(&pr->pr_mtx, MA_OWNED);
2583 pr->pr_ref--;
2584 if (pr->pr_ref == 0) {
2585 mtx_unlock(&pr->pr_mtx);
2586 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2587 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2588 return;
2589 }
2590 mtx_unlock(&pr->pr_mtx);
2591 }
2592
2593 void
2594 prison_free(struct prison *pr)
2595 {
2596
2597 mtx_lock(&pr->pr_mtx);
2598 prison_free_locked(pr);
2599 }
2600
2601 static void
2602 prison_complete(void *context, int pending)
2603 {
2604
2605 prison_deref((struct prison *)context, 0);
2606 }
2607
2608 /*
2609 * Remove a prison reference (usually). This internal version assumes no
2610 * mutexes are held, except perhaps the prison itself. If there are no more
2611 * references, release and delist the prison. On completion, the prison lock
2612 * and the allprison lock are both unlocked.
2613 */
2614 static void
2615 prison_deref(struct prison *pr, int flags)
2616 {
2617 struct prison *ppr, *tpr;
2618
2619 if (!(flags & PD_LOCKED))
2620 mtx_lock(&pr->pr_mtx);
2621 for (;;) {
2622 if (flags & PD_DEUREF) {
2623 pr->pr_uref--;
2624 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2625 }
2626 if (flags & PD_DEREF)
2627 pr->pr_ref--;
2628 /* If the prison still has references, nothing else to do. */
2629 if (pr->pr_ref > 0) {
2630 mtx_unlock(&pr->pr_mtx);
2631 if (flags & PD_LIST_SLOCKED)
2632 sx_sunlock(&allprison_lock);
2633 else if (flags & PD_LIST_XLOCKED)
2634 sx_xunlock(&allprison_lock);
2635 return;
2636 }
2637
2638 mtx_unlock(&pr->pr_mtx);
2639 if (flags & PD_LIST_SLOCKED) {
2640 if (!sx_try_upgrade(&allprison_lock)) {
2641 sx_sunlock(&allprison_lock);
2642 sx_xlock(&allprison_lock);
2643 }
2644 } else if (!(flags & PD_LIST_XLOCKED))
2645 sx_xlock(&allprison_lock);
2646
2647 TAILQ_REMOVE(&allprison, pr, pr_list);
2648 LIST_REMOVE(pr, pr_sibling);
2649 ppr = pr->pr_parent;
2650 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2651 tpr->pr_childcount--;
2652 sx_xunlock(&allprison_lock);
2653
2654 #ifdef VIMAGE
2655 if (pr->pr_vnet != ppr->pr_vnet)
2656 vnet_destroy(pr->pr_vnet);
2657 #endif
2658 if (pr->pr_root != NULL)
2659 vrele(pr->pr_root);
2660 mtx_destroy(&pr->pr_mtx);
2661 #ifdef INET
2662 free(pr->pr_ip4, M_PRISON);
2663 #endif
2664 #ifdef INET6
2665 free(pr->pr_ip6, M_PRISON);
2666 #endif
2667 if (pr->pr_cpuset != NULL)
2668 cpuset_rel(pr->pr_cpuset);
2669 osd_jail_exit(pr);
2670 #ifdef RACCT
2671 if (racct_enable)
2672 prison_racct_detach(pr);
2673 #endif
2674 free(pr, M_PRISON);
2675
2676 /* Removing a prison frees a reference on its parent. */
2677 pr = ppr;
2678 mtx_lock(&pr->pr_mtx);
2679 flags = PD_DEREF | PD_DEUREF;
2680 }
2681 }
2682
2683 void
2684 prison_hold_locked(struct prison *pr)
2685 {
2686
2687 mtx_assert(&pr->pr_mtx, MA_OWNED);
2688 KASSERT(pr->pr_ref > 0,
2689 ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2690 pr->pr_ref++;
2691 }
2692
2693 void
2694 prison_hold(struct prison *pr)
2695 {
2696
2697 mtx_lock(&pr->pr_mtx);
2698 prison_hold_locked(pr);
2699 mtx_unlock(&pr->pr_mtx);
2700 }
2701
2702 void
2703 prison_proc_hold(struct prison *pr)
2704 {
2705
2706 mtx_lock(&pr->pr_mtx);
2707 KASSERT(pr->pr_uref > 0,
2708 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2709 pr->pr_uref++;
2710 mtx_unlock(&pr->pr_mtx);
2711 }
2712
2713 void
2714 prison_proc_free(struct prison *pr)
2715 {
2716
2717 mtx_lock(&pr->pr_mtx);
2718 KASSERT(pr->pr_uref > 0,
2719 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2720 prison_deref(pr, PD_DEUREF | PD_LOCKED);
2721 }
2722
2723
2724 #ifdef INET
2725 /*
2726 * Restrict a prison's IP address list with its parent's, possibly replacing
2727 * it. Return true if the replacement buffer was used (or would have been).
2728 */
2729 static int
2730 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2731 {
2732 int ii, ij, used;
2733 struct prison *ppr;
2734
2735 ppr = pr->pr_parent;
2736 if (!(pr->pr_flags & PR_IP4_USER)) {
2737 /* This has no user settings, so just copy the parent's list. */
2738 if (pr->pr_ip4s < ppr->pr_ip4s) {
2739 /*
2740 * There's no room for the parent's list. Use the
2741 * new list buffer, which is assumed to be big enough
2742 * (if it was passed). If there's no buffer, try to
2743 * allocate one.
2744 */
2745 used = 1;
2746 if (newip4 == NULL) {
2747 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2748 M_PRISON, M_NOWAIT);
2749 if (newip4 != NULL)
2750 used = 0;
2751 }
2752 if (newip4 != NULL) {
2753 bcopy(ppr->pr_ip4, newip4,
2754 ppr->pr_ip4s * sizeof(*newip4));
2755 free(pr->pr_ip4, M_PRISON);
2756 pr->pr_ip4 = newip4;
2757 pr->pr_ip4s = ppr->pr_ip4s;
2758 }
2759 return (used);
2760 }
2761 pr->pr_ip4s = ppr->pr_ip4s;
2762 if (pr->pr_ip4s > 0)
2763 bcopy(ppr->pr_ip4, pr->pr_ip4,
2764 pr->pr_ip4s * sizeof(*newip4));
2765 else if (pr->pr_ip4 != NULL) {
2766 free(pr->pr_ip4, M_PRISON);
2767 pr->pr_ip4 = NULL;
2768 }
2769 } else if (pr->pr_ip4s > 0) {
2770 /* Remove addresses that aren't in the parent. */
2771 for (ij = 0; ij < ppr->pr_ip4s; ij++)
2772 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2773 break;
2774 if (ij < ppr->pr_ip4s)
2775 ii = 1;
2776 else {
2777 bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2778 --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2779 ii = 0;
2780 }
2781 for (ij = 1; ii < pr->pr_ip4s; ) {
2782 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2783 ii++;
2784 continue;
2785 }
2786 switch (ij >= ppr->pr_ip4s ? -1 :
2787 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2788 case -1:
2789 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2790 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2791 break;
2792 case 0:
2793 ii++;
2794 ij++;
2795 break;
2796 case 1:
2797 ij++;
2798 break;
2799 }
2800 }
2801 if (pr->pr_ip4s == 0) {
2802 pr->pr_flags |= PR_IP4_DISABLE;
2803 free(pr->pr_ip4, M_PRISON);
2804 pr->pr_ip4 = NULL;
2805 }
2806 }
2807 return (0);
2808 }
2809
2810 /*
2811 * Pass back primary IPv4 address of this jail.
2812 *
2813 * If not restricted return success but do not alter the address. Caller has
2814 * to make sure to initialize it correctly (e.g. INADDR_ANY).
2815 *
2816 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2817 * Address returned in NBO.
2818 */
2819 int
2820 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2821 {
2822 struct prison *pr;
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 ia->s_addr = pr->pr_ip4[0].s_addr;
2841 mtx_unlock(&pr->pr_mtx);
2842 return (0);
2843 }
2844
2845 /*
2846 * Return 1 if we should do proper source address selection or are not jailed.
2847 * We will return 0 if we should bypass source address selection in favour
2848 * of the primary jail IPv4 address. Only in this case *ia will be updated and
2849 * returned in NBO.
2850 * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2851 */
2852 int
2853 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2854 {
2855 struct prison *pr;
2856 struct in_addr lia;
2857 int error;
2858
2859 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2860 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2861
2862 if (!jailed(cred))
2863 return (1);
2864
2865 pr = cred->cr_prison;
2866 if (pr->pr_flags & PR_IP4_SADDRSEL)
2867 return (1);
2868
2869 lia.s_addr = INADDR_ANY;
2870 error = prison_get_ip4(cred, &lia);
2871 if (error)
2872 return (error);
2873 if (lia.s_addr == INADDR_ANY)
2874 return (1);
2875
2876 ia->s_addr = lia.s_addr;
2877 return (0);
2878 }
2879
2880 /*
2881 * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2882 */
2883 int
2884 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2885 {
2886
2887 if (pr1 == pr2)
2888 return (1);
2889
2890 /*
2891 * No need to lock since the PR_IP4_USER flag can't be altered for
2892 * existing prisons.
2893 */
2894 while (pr1 != &prison0 &&
2895 #ifdef VIMAGE
2896 !(pr1->pr_flags & PR_VNET) &&
2897 #endif
2898 !(pr1->pr_flags & PR_IP4_USER))
2899 pr1 = pr1->pr_parent;
2900 while (pr2 != &prison0 &&
2901 #ifdef VIMAGE
2902 !(pr2->pr_flags & PR_VNET) &&
2903 #endif
2904 !(pr2->pr_flags & PR_IP4_USER))
2905 pr2 = pr2->pr_parent;
2906 return (pr1 == pr2);
2907 }
2908
2909 /*
2910 * Make sure our (source) address is set to something meaningful to this
2911 * jail.
2912 *
2913 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2914 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2915 * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2916 */
2917 int
2918 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2919 {
2920 struct prison *pr;
2921 struct in_addr ia0;
2922 int error;
2923
2924 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2925 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2926
2927 pr = cred->cr_prison;
2928 if (!(pr->pr_flags & PR_IP4))
2929 return (0);
2930 mtx_lock(&pr->pr_mtx);
2931 if (!(pr->pr_flags & PR_IP4)) {
2932 mtx_unlock(&pr->pr_mtx);
2933 return (0);
2934 }
2935 if (pr->pr_ip4 == NULL) {
2936 mtx_unlock(&pr->pr_mtx);
2937 return (EAFNOSUPPORT);
2938 }
2939
2940 ia0.s_addr = ntohl(ia->s_addr);
2941 if (ia0.s_addr == INADDR_LOOPBACK) {
2942 ia->s_addr = pr->pr_ip4[0].s_addr;
2943 mtx_unlock(&pr->pr_mtx);
2944 return (0);
2945 }
2946
2947 if (ia0.s_addr == INADDR_ANY) {
2948 /*
2949 * In case there is only 1 IPv4 address, bind directly.
2950 */
2951 if (pr->pr_ip4s == 1)
2952 ia->s_addr = pr->pr_ip4[0].s_addr;
2953 mtx_unlock(&pr->pr_mtx);
2954 return (0);
2955 }
2956
2957 error = _prison_check_ip4(pr, ia);
2958 mtx_unlock(&pr->pr_mtx);
2959 return (error);
2960 }
2961
2962 /*
2963 * Rewrite destination address in case we will connect to loopback address.
2964 *
2965 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2966 * Address passed in in NBO and returned in NBO.
2967 */
2968 int
2969 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2970 {
2971 struct prison *pr;
2972
2973 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2974 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2975
2976 pr = cred->cr_prison;
2977 if (!(pr->pr_flags & PR_IP4))
2978 return (0);
2979 mtx_lock(&pr->pr_mtx);
2980 if (!(pr->pr_flags & PR_IP4)) {
2981 mtx_unlock(&pr->pr_mtx);
2982 return (0);
2983 }
2984 if (pr->pr_ip4 == NULL) {
2985 mtx_unlock(&pr->pr_mtx);
2986 return (EAFNOSUPPORT);
2987 }
2988
2989 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2990 ia->s_addr = pr->pr_ip4[0].s_addr;
2991 mtx_unlock(&pr->pr_mtx);
2992 return (0);
2993 }
2994
2995 /*
2996 * Return success because nothing had to be changed.
2997 */
2998 mtx_unlock(&pr->pr_mtx);
2999 return (0);
3000 }
3001
3002 /*
3003 * Check if given address belongs to the jail referenced by cred/prison.
3004 *
3005 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
3006 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3007 * doesn't allow IPv4. Address passed in in NBO.
3008 */
3009 static int
3010 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
3011 {
3012 int i, a, z, d;
3013
3014 /*
3015 * Check the primary IP.
3016 */
3017 if (pr->pr_ip4[0].s_addr == ia->s_addr)
3018 return (0);
3019
3020 /*
3021 * All the other IPs are sorted so we can do a binary search.
3022 */
3023 a = 0;
3024 z = pr->pr_ip4s - 2;
3025 while (a <= z) {
3026 i = (a + z) / 2;
3027 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
3028 if (d > 0)
3029 z = i - 1;
3030 else if (d < 0)
3031 a = i + 1;
3032 else
3033 return (0);
3034 }
3035
3036 return (EADDRNOTAVAIL);
3037 }
3038
3039 int
3040 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
3041 {
3042 struct prison *pr;
3043 int error;
3044
3045 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3046 KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
3047
3048 pr = cred->cr_prison;
3049 if (!(pr->pr_flags & PR_IP4))
3050 return (0);
3051 mtx_lock(&pr->pr_mtx);
3052 if (!(pr->pr_flags & PR_IP4)) {
3053 mtx_unlock(&pr->pr_mtx);
3054 return (0);
3055 }
3056 if (pr->pr_ip4 == NULL) {
3057 mtx_unlock(&pr->pr_mtx);
3058 return (EAFNOSUPPORT);
3059 }
3060
3061 error = _prison_check_ip4(pr, ia);
3062 mtx_unlock(&pr->pr_mtx);
3063 return (error);
3064 }
3065 #endif
3066
3067 #ifdef INET6
3068 static int
3069 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3070 {
3071 int ii, ij, used;
3072 struct prison *ppr;
3073
3074 ppr = pr->pr_parent;
3075 if (!(pr->pr_flags & PR_IP6_USER)) {
3076 /* This has no user settings, so just copy the parent's list. */
3077 if (pr->pr_ip6s < ppr->pr_ip6s) {
3078 /*
3079 * There's no room for the parent's list. Use the
3080 * new list buffer, which is assumed to be big enough
3081 * (if it was passed). If there's no buffer, try to
3082 * allocate one.
3083 */
3084 used = 1;
3085 if (newip6 == NULL) {
3086 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3087 M_PRISON, M_NOWAIT);
3088 if (newip6 != NULL)
3089 used = 0;
3090 }
3091 if (newip6 != NULL) {
3092 bcopy(ppr->pr_ip6, newip6,
3093 ppr->pr_ip6s * sizeof(*newip6));
3094 free(pr->pr_ip6, M_PRISON);
3095 pr->pr_ip6 = newip6;
3096 pr->pr_ip6s = ppr->pr_ip6s;
3097 }
3098 return (used);
3099 }
3100 pr->pr_ip6s = ppr->pr_ip6s;
3101 if (pr->pr_ip6s > 0)
3102 bcopy(ppr->pr_ip6, pr->pr_ip6,
3103 pr->pr_ip6s * sizeof(*newip6));
3104 else if (pr->pr_ip6 != NULL) {
3105 free(pr->pr_ip6, M_PRISON);
3106 pr->pr_ip6 = NULL;
3107 }
3108 } else if (pr->pr_ip6s > 0) {
3109 /* Remove addresses that aren't in the parent. */
3110 for (ij = 0; ij < ppr->pr_ip6s; ij++)
3111 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3112 &ppr->pr_ip6[ij]))
3113 break;
3114 if (ij < ppr->pr_ip6s)
3115 ii = 1;
3116 else {
3117 bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3118 --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3119 ii = 0;
3120 }
3121 for (ij = 1; ii < pr->pr_ip6s; ) {
3122 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3123 &ppr->pr_ip6[0])) {
3124 ii++;
3125 continue;
3126 }
3127 switch (ij >= ppr->pr_ip6s ? -1 :
3128 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3129 case -1:
3130 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3131 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3132 break;
3133 case 0:
3134 ii++;
3135 ij++;
3136 break;
3137 case 1:
3138 ij++;
3139 break;
3140 }
3141 }
3142 if (pr->pr_ip6s == 0) {
3143 pr->pr_flags |= PR_IP6_DISABLE;
3144 free(pr->pr_ip6, M_PRISON);
3145 pr->pr_ip6 = NULL;
3146 }
3147 }
3148 return 0;
3149 }
3150
3151 /*
3152 * Pass back primary IPv6 address for this jail.
3153 *
3154 * If not restricted return success but do not alter the address. Caller has
3155 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3156 *
3157 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3158 */
3159 int
3160 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3161 {
3162 struct prison *pr;
3163
3164 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3165 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3166
3167 pr = cred->cr_prison;
3168 if (!(pr->pr_flags & PR_IP6))
3169 return (0);
3170 mtx_lock(&pr->pr_mtx);
3171 if (!(pr->pr_flags & PR_IP6)) {
3172 mtx_unlock(&pr->pr_mtx);
3173 return (0);
3174 }
3175 if (pr->pr_ip6 == NULL) {
3176 mtx_unlock(&pr->pr_mtx);
3177 return (EAFNOSUPPORT);
3178 }
3179
3180 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3181 mtx_unlock(&pr->pr_mtx);
3182 return (0);
3183 }
3184
3185 /*
3186 * Return 1 if we should do proper source address selection or are not jailed.
3187 * We will return 0 if we should bypass source address selection in favour
3188 * of the primary jail IPv6 address. Only in this case *ia will be updated and
3189 * returned in NBO.
3190 * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3191 */
3192 int
3193 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3194 {
3195 struct prison *pr;
3196 struct in6_addr lia6;
3197 int error;
3198
3199 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3200 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3201
3202 if (!jailed(cred))
3203 return (1);
3204
3205 pr = cred->cr_prison;
3206 if (pr->pr_flags & PR_IP6_SADDRSEL)
3207 return (1);
3208
3209 lia6 = in6addr_any;
3210 error = prison_get_ip6(cred, &lia6);
3211 if (error)
3212 return (error);
3213 if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3214 return (1);
3215
3216 bcopy(&lia6, ia6, sizeof(struct in6_addr));
3217 return (0);
3218 }
3219
3220 /*
3221 * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3222 */
3223 int
3224 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3225 {
3226
3227 if (pr1 == pr2)
3228 return (1);
3229
3230 while (pr1 != &prison0 &&
3231 #ifdef VIMAGE
3232 !(pr1->pr_flags & PR_VNET) &&
3233 #endif
3234 !(pr1->pr_flags & PR_IP6_USER))
3235 pr1 = pr1->pr_parent;
3236 while (pr2 != &prison0 &&
3237 #ifdef VIMAGE
3238 !(pr2->pr_flags & PR_VNET) &&
3239 #endif
3240 !(pr2->pr_flags & PR_IP6_USER))
3241 pr2 = pr2->pr_parent;
3242 return (pr1 == pr2);
3243 }
3244
3245 /*
3246 * Make sure our (source) address is set to something meaningful to this jail.
3247 *
3248 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3249 * when needed while binding.
3250 *
3251 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3252 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3253 * doesn't allow IPv6.
3254 */
3255 int
3256 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3257 {
3258 struct prison *pr;
3259 int error;
3260
3261 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3262 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3263
3264 pr = cred->cr_prison;
3265 if (!(pr->pr_flags & PR_IP6))
3266 return (0);
3267 mtx_lock(&pr->pr_mtx);
3268 if (!(pr->pr_flags & PR_IP6)) {
3269 mtx_unlock(&pr->pr_mtx);
3270 return (0);
3271 }
3272 if (pr->pr_ip6 == NULL) {
3273 mtx_unlock(&pr->pr_mtx);
3274 return (EAFNOSUPPORT);
3275 }
3276
3277 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3278 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3279 mtx_unlock(&pr->pr_mtx);
3280 return (0);
3281 }
3282
3283 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3284 /*
3285 * In case there is only 1 IPv6 address, and v6only is true,
3286 * then bind directly.
3287 */
3288 if (v6only != 0 && pr->pr_ip6s == 1)
3289 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3290 mtx_unlock(&pr->pr_mtx);
3291 return (0);
3292 }
3293
3294 error = _prison_check_ip6(pr, ia6);
3295 mtx_unlock(&pr->pr_mtx);
3296 return (error);
3297 }
3298
3299 /*
3300 * Rewrite destination address in case we will connect to loopback address.
3301 *
3302 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3303 */
3304 int
3305 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3306 {
3307 struct prison *pr;
3308
3309 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3310 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3311
3312 pr = cred->cr_prison;
3313 if (!(pr->pr_flags & PR_IP6))
3314 return (0);
3315 mtx_lock(&pr->pr_mtx);
3316 if (!(pr->pr_flags & PR_IP6)) {
3317 mtx_unlock(&pr->pr_mtx);
3318 return (0);
3319 }
3320 if (pr->pr_ip6 == NULL) {
3321 mtx_unlock(&pr->pr_mtx);
3322 return (EAFNOSUPPORT);
3323 }
3324
3325 if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3326 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3327 mtx_unlock(&pr->pr_mtx);
3328 return (0);
3329 }
3330
3331 /*
3332 * Return success because nothing had to be changed.
3333 */
3334 mtx_unlock(&pr->pr_mtx);
3335 return (0);
3336 }
3337
3338 /*
3339 * Check if given address belongs to the jail referenced by cred/prison.
3340 *
3341 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3342 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3343 * doesn't allow IPv6.
3344 */
3345 static int
3346 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3347 {
3348 int i, a, z, d;
3349
3350 /*
3351 * Check the primary IP.
3352 */
3353 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3354 return (0);
3355
3356 /*
3357 * All the other IPs are sorted so we can do a binary search.
3358 */
3359 a = 0;
3360 z = pr->pr_ip6s - 2;
3361 while (a <= z) {
3362 i = (a + z) / 2;
3363 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3364 if (d > 0)
3365 z = i - 1;
3366 else if (d < 0)
3367 a = i + 1;
3368 else
3369 return (0);
3370 }
3371
3372 return (EADDRNOTAVAIL);
3373 }
3374
3375 int
3376 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3377 {
3378 struct prison *pr;
3379 int error;
3380
3381 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3382 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3383
3384 pr = cred->cr_prison;
3385 if (!(pr->pr_flags & PR_IP6))
3386 return (0);
3387 mtx_lock(&pr->pr_mtx);
3388 if (!(pr->pr_flags & PR_IP6)) {
3389 mtx_unlock(&pr->pr_mtx);
3390 return (0);
3391 }
3392 if (pr->pr_ip6 == NULL) {
3393 mtx_unlock(&pr->pr_mtx);
3394 return (EAFNOSUPPORT);
3395 }
3396
3397 error = _prison_check_ip6(pr, ia6);
3398 mtx_unlock(&pr->pr_mtx);
3399 return (error);
3400 }
3401 #endif
3402
3403 /*
3404 * Check if a jail supports the given address family.
3405 *
3406 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3407 * if not.
3408 */
3409 int
3410 prison_check_af(struct ucred *cred, int af)
3411 {
3412 struct prison *pr;
3413 int error;
3414
3415 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3416
3417 pr = cred->cr_prison;
3418 #ifdef VIMAGE
3419 /* Prisons with their own network stack are not limited. */
3420 if (prison_owns_vnet(cred))
3421 return (0);
3422 #endif
3423
3424 error = 0;
3425 switch (af)
3426 {
3427 #ifdef INET
3428 case AF_INET:
3429 if (pr->pr_flags & PR_IP4)
3430 {
3431 mtx_lock(&pr->pr_mtx);
3432 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3433 error = EAFNOSUPPORT;
3434 mtx_unlock(&pr->pr_mtx);
3435 }
3436 break;
3437 #endif
3438 #ifdef INET6
3439 case AF_INET6:
3440 if (pr->pr_flags & PR_IP6)
3441 {
3442 mtx_lock(&pr->pr_mtx);
3443 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3444 error = EAFNOSUPPORT;
3445 mtx_unlock(&pr->pr_mtx);
3446 }
3447 break;
3448 #endif
3449 case AF_LOCAL:
3450 case AF_ROUTE:
3451 break;
3452 default:
3453 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3454 error = EAFNOSUPPORT;
3455 }
3456 return (error);
3457 }
3458
3459 /*
3460 * Check if given address belongs to the jail referenced by cred (wrapper to
3461 * prison_check_ip[46]).
3462 *
3463 * Returns 0 if jail doesn't restrict the address family or if address belongs
3464 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3465 * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3466 */
3467 int
3468 prison_if(struct ucred *cred, struct sockaddr *sa)
3469 {
3470 #ifdef INET
3471 struct sockaddr_in *sai;
3472 #endif
3473 #ifdef INET6
3474 struct sockaddr_in6 *sai6;
3475 #endif
3476 int error;
3477
3478 KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3479 KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3480
3481 #ifdef VIMAGE
3482 if (prison_owns_vnet(cred))
3483 return (0);
3484 #endif
3485
3486 error = 0;
3487 switch (sa->sa_family)
3488 {
3489 #ifdef INET
3490 case AF_INET:
3491 sai = (struct sockaddr_in *)sa;
3492 error = prison_check_ip4(cred, &sai->sin_addr);
3493 break;
3494 #endif
3495 #ifdef INET6
3496 case AF_INET6:
3497 sai6 = (struct sockaddr_in6 *)sa;
3498 error = prison_check_ip6(cred, &sai6->sin6_addr);
3499 break;
3500 #endif
3501 default:
3502 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3503 error = EAFNOSUPPORT;
3504 }
3505 return (error);
3506 }
3507
3508 /*
3509 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3510 */
3511 int
3512 prison_check(struct ucred *cred1, struct ucred *cred2)
3513 {
3514
3515 return ((cred1->cr_prison == cred2->cr_prison ||
3516 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3517 }
3518
3519 /*
3520 * Return 1 if p2 is a child of p1, otherwise 0.
3521 */
3522 int
3523 prison_ischild(struct prison *pr1, struct prison *pr2)
3524 {
3525
3526 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3527 if (pr1 == pr2)
3528 return (1);
3529 return (0);
3530 }
3531
3532 /*
3533 * Return 1 if the passed credential is in a jail, otherwise 0.
3534 */
3535 int
3536 jailed(struct ucred *cred)
3537 {
3538
3539 return (cred->cr_prison != &prison0);
3540 }
3541
3542 /*
3543 * Return 1 if the passed credential is in a jail and that jail does not
3544 * have its own virtual network stack, otherwise 0.
3545 */
3546 int
3547 jailed_without_vnet(struct ucred *cred)
3548 {
3549
3550 if (!jailed(cred))
3551 return (0);
3552 #ifdef VIMAGE
3553 if (prison_owns_vnet(cred))
3554 return (0);
3555 #endif
3556
3557 return (1);
3558 }
3559
3560 /*
3561 * Return the correct hostname (domainname, et al) for the passed credential.
3562 */
3563 void
3564 getcredhostname(struct ucred *cred, char *buf, size_t size)
3565 {
3566 struct prison *pr;
3567
3568 /*
3569 * A NULL credential can be used to shortcut to the physical
3570 * system's hostname.
3571 */
3572 pr = (cred != NULL) ? cred->cr_prison : &prison0;
3573 mtx_lock(&pr->pr_mtx);
3574 strlcpy(buf, pr->pr_hostname, size);
3575 mtx_unlock(&pr->pr_mtx);
3576 }
3577
3578 void
3579 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3580 {
3581
3582 mtx_lock(&cred->cr_prison->pr_mtx);
3583 strlcpy(buf, cred->cr_prison->pr_domainname, size);
3584 mtx_unlock(&cred->cr_prison->pr_mtx);
3585 }
3586
3587 void
3588 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3589 {
3590
3591 mtx_lock(&cred->cr_prison->pr_mtx);
3592 strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3593 mtx_unlock(&cred->cr_prison->pr_mtx);
3594 }
3595
3596 void
3597 getcredhostid(struct ucred *cred, unsigned long *hostid)
3598 {
3599
3600 mtx_lock(&cred->cr_prison->pr_mtx);
3601 *hostid = cred->cr_prison->pr_hostid;
3602 mtx_unlock(&cred->cr_prison->pr_mtx);
3603 }
3604
3605 #ifdef VIMAGE
3606 /*
3607 * Determine whether the prison represented by cred owns
3608 * its vnet rather than having it inherited.
3609 *
3610 * Returns 1 in case the prison owns the vnet, 0 otherwise.
3611 */
3612 int
3613 prison_owns_vnet(struct ucred *cred)
3614 {
3615
3616 /*
3617 * vnets cannot be added/removed after jail creation,
3618 * so no need to lock here.
3619 */
3620 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3621 }
3622 #endif
3623
3624 /*
3625 * Determine whether the subject represented by cred can "see"
3626 * status of a mount point.
3627 * Returns: 0 for permitted, ENOENT otherwise.
3628 * XXX: This function should be called cr_canseemount() and should be
3629 * placed in kern_prot.c.
3630 */
3631 int
3632 prison_canseemount(struct ucred *cred, struct mount *mp)
3633 {
3634 struct prison *pr;
3635 struct statfs *sp;
3636 size_t len;
3637
3638 pr = cred->cr_prison;
3639 if (pr->pr_enforce_statfs == 0)
3640 return (0);
3641 if (pr->pr_root->v_mount == mp)
3642 return (0);
3643 if (pr->pr_enforce_statfs == 2)
3644 return (ENOENT);
3645 /*
3646 * If jail's chroot directory is set to "/" we should be able to see
3647 * all mount-points from inside a jail.
3648 * This is ugly check, but this is the only situation when jail's
3649 * directory ends with '/'.
3650 */
3651 if (strcmp(pr->pr_path, "/") == 0)
3652 return (0);
3653 len = strlen(pr->pr_path);
3654 sp = &mp->mnt_stat;
3655 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3656 return (ENOENT);
3657 /*
3658 * Be sure that we don't have situation where jail's root directory
3659 * is "/some/path" and mount point is "/some/pathpath".
3660 */
3661 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3662 return (ENOENT);
3663 return (0);
3664 }
3665
3666 void
3667 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3668 {
3669 char jpath[MAXPATHLEN];
3670 struct prison *pr;
3671 size_t len;
3672
3673 pr = cred->cr_prison;
3674 if (pr->pr_enforce_statfs == 0)
3675 return;
3676 if (prison_canseemount(cred, mp) != 0) {
3677 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3678 strlcpy(sp->f_mntonname, "[restricted]",
3679 sizeof(sp->f_mntonname));
3680 return;
3681 }
3682 if (pr->pr_root->v_mount == mp) {
3683 /*
3684 * Clear current buffer data, so we are sure nothing from
3685 * the valid path left there.
3686 */
3687 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3688 *sp->f_mntonname = '/';
3689 return;
3690 }
3691 /*
3692 * If jail's chroot directory is set to "/" we should be able to see
3693 * all mount-points from inside a jail.
3694 */
3695 if (strcmp(pr->pr_path, "/") == 0)
3696 return;
3697 len = strlen(pr->pr_path);
3698 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3699 /*
3700 * Clear current buffer data, so we are sure nothing from
3701 * the valid path left there.
3702 */
3703 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3704 if (*jpath == '\0') {
3705 /* Should never happen. */
3706 *sp->f_mntonname = '/';
3707 } else {
3708 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3709 }
3710 }
3711
3712 /*
3713 * Check with permission for a specific privilege is granted within jail. We
3714 * have a specific list of accepted privileges; the rest are denied.
3715 */
3716 int
3717 prison_priv_check(struct ucred *cred, int priv)
3718 {
3719
3720 if (!jailed(cred))
3721 return (0);
3722
3723 #ifdef VIMAGE
3724 /*
3725 * Privileges specific to prisons with a virtual network stack.
3726 * There might be a duplicate entry here in case the privilege
3727 * is only granted conditionally in the legacy jail case.
3728 */
3729 switch (priv) {
3730 #ifdef notyet
3731 /*
3732 * NFS-specific privileges.
3733 */
3734 case PRIV_NFS_DAEMON:
3735 case PRIV_NFS_LOCKD:
3736 #endif
3737 /*
3738 * Network stack privileges.
3739 */
3740 case PRIV_NET_BRIDGE:
3741 case PRIV_NET_GRE:
3742 case PRIV_NET_BPF:
3743 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3744 case PRIV_NET_ROUTE:
3745 case PRIV_NET_TAP:
3746 case PRIV_NET_SETIFMTU:
3747 case PRIV_NET_SETIFFLAGS:
3748 case PRIV_NET_SETIFCAP:
3749 case PRIV_NET_SETIFDESCR:
3750 case PRIV_NET_SETIFNAME :
3751 case PRIV_NET_SETIFMETRIC:
3752 case PRIV_NET_SETIFPHYS:
3753 case PRIV_NET_SETIFMAC:
3754 case PRIV_NET_ADDMULTI:
3755 case PRIV_NET_DELMULTI:
3756 case PRIV_NET_HWIOCTL:
3757 case PRIV_NET_SETLLADDR:
3758 case PRIV_NET_ADDIFGROUP:
3759 case PRIV_NET_DELIFGROUP:
3760 case PRIV_NET_IFCREATE:
3761 case PRIV_NET_IFDESTROY:
3762 case PRIV_NET_ADDIFADDR:
3763 case PRIV_NET_DELIFADDR:
3764 case PRIV_NET_LAGG:
3765 case PRIV_NET_GIF:
3766 case PRIV_NET_SETIFVNET:
3767 case PRIV_NET_SETIFFIB:
3768
3769 /*
3770 * 802.11-related privileges.
3771 */
3772 case PRIV_NET80211_GETKEY:
3773 #ifdef notyet
3774 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3775 #endif
3776
3777 #ifdef notyet
3778 /*
3779 * AppleTalk privileges.
3780 */
3781 case PRIV_NETATALK_RESERVEDPORT:
3782
3783 /*
3784 * ATM privileges.
3785 */
3786 case PRIV_NETATM_CFG:
3787 case PRIV_NETATM_ADD:
3788 case PRIV_NETATM_DEL:
3789 case PRIV_NETATM_SET:
3790
3791 /*
3792 * Bluetooth privileges.
3793 */
3794 case PRIV_NETBLUETOOTH_RAW:
3795 #endif
3796
3797 /*
3798 * Netgraph and netgraph module privileges.
3799 */
3800 case PRIV_NETGRAPH_CONTROL:
3801 #ifdef notyet
3802 case PRIV_NETGRAPH_TTY:
3803 #endif
3804
3805 /*
3806 * IPv4 and IPv6 privileges.
3807 */
3808 case PRIV_NETINET_IPFW:
3809 case PRIV_NETINET_DIVERT:
3810 case PRIV_NETINET_PF:
3811 case PRIV_NETINET_DUMMYNET:
3812 case PRIV_NETINET_CARP:
3813 case PRIV_NETINET_MROUTE:
3814 case PRIV_NETINET_RAW:
3815 case PRIV_NETINET_ADDRCTRL6:
3816 case PRIV_NETINET_ND6:
3817 case PRIV_NETINET_SCOPE6:
3818 case PRIV_NETINET_ALIFETIME6:
3819 case PRIV_NETINET_IPSEC:
3820 case PRIV_NETINET_BINDANY:
3821
3822 #ifdef notyet
3823 /*
3824 * IPX/SPX privileges.
3825 */
3826 case PRIV_NETIPX_RESERVEDPORT:
3827 case PRIV_NETIPX_RAW:
3828
3829 /*
3830 * NCP privileges.
3831 */
3832 case PRIV_NETNCP:
3833
3834 /*
3835 * SMB privileges.
3836 */
3837 case PRIV_NETSMB:
3838 #endif
3839
3840 /*
3841 * No default: or deny here.
3842 * In case of no permit fall through to next switch().
3843 */
3844 if (cred->cr_prison->pr_flags & PR_VNET)
3845 return (0);
3846 }
3847 #endif /* VIMAGE */
3848
3849 switch (priv) {
3850
3851 /*
3852 * Allow ktrace privileges for root in jail.
3853 */
3854 case PRIV_KTRACE:
3855
3856 #if 0
3857 /*
3858 * Allow jailed processes to configure audit identity and
3859 * submit audit records (login, etc). In the future we may
3860 * want to further refine the relationship between audit and
3861 * jail.
3862 */
3863 case PRIV_AUDIT_GETAUDIT:
3864 case PRIV_AUDIT_SETAUDIT:
3865 case PRIV_AUDIT_SUBMIT:
3866 #endif
3867
3868 /*
3869 * Allow jailed processes to manipulate process UNIX
3870 * credentials in any way they see fit.
3871 */
3872 case PRIV_CRED_SETUID:
3873 case PRIV_CRED_SETEUID:
3874 case PRIV_CRED_SETGID:
3875 case PRIV_CRED_SETEGID:
3876 case PRIV_CRED_SETGROUPS:
3877 case PRIV_CRED_SETREUID:
3878 case PRIV_CRED_SETREGID:
3879 case PRIV_CRED_SETRESUID:
3880 case PRIV_CRED_SETRESGID:
3881
3882 /*
3883 * Jail implements visibility constraints already, so allow
3884 * jailed root to override uid/gid-based constraints.
3885 */
3886 case PRIV_SEEOTHERGIDS:
3887 case PRIV_SEEOTHERUIDS:
3888
3889 /*
3890 * Jail implements inter-process debugging limits already, so
3891 * allow jailed root various debugging privileges.
3892 */
3893 case PRIV_DEBUG_DIFFCRED:
3894 case PRIV_DEBUG_SUGID:
3895 case PRIV_DEBUG_UNPRIV:
3896
3897 /*
3898 * Allow jail to set various resource limits and login
3899 * properties, and for now, exceed process resource limits.
3900 */
3901 case PRIV_PROC_LIMIT:
3902 case PRIV_PROC_SETLOGIN:
3903 case PRIV_PROC_SETRLIMIT:
3904
3905 /*
3906 * System V and POSIX IPC privileges are granted in jail.
3907 */
3908 case PRIV_IPC_READ:
3909 case PRIV_IPC_WRITE:
3910 case PRIV_IPC_ADMIN:
3911 case PRIV_IPC_MSGSIZE:
3912 case PRIV_MQ_ADMIN:
3913
3914 /*
3915 * Jail operations within a jail work on child jails.
3916 */
3917 case PRIV_JAIL_ATTACH:
3918 case PRIV_JAIL_SET:
3919 case PRIV_JAIL_REMOVE:
3920
3921 /*
3922 * Jail implements its own inter-process limits, so allow
3923 * root processes in jail to change scheduling on other
3924 * processes in the same jail. Likewise for signalling.
3925 */
3926 case PRIV_SCHED_DIFFCRED:
3927 case PRIV_SCHED_CPUSET:
3928 case PRIV_SIGNAL_DIFFCRED:
3929 case PRIV_SIGNAL_SUGID:
3930
3931 /*
3932 * Allow jailed processes to write to sysctls marked as jail
3933 * writable.
3934 */
3935 case PRIV_SYSCTL_WRITEJAIL:
3936
3937 /*
3938 * Allow root in jail to manage a variety of quota
3939 * properties. These should likely be conditional on a
3940 * configuration option.
3941 */
3942 case PRIV_VFS_GETQUOTA:
3943 case PRIV_VFS_SETQUOTA:
3944
3945 /*
3946 * Since Jail relies on chroot() to implement file system
3947 * protections, grant many VFS privileges to root in jail.
3948 * Be careful to exclude mount-related and NFS-related
3949 * privileges.
3950 */
3951 case PRIV_VFS_READ:
3952 case PRIV_VFS_WRITE:
3953 case PRIV_VFS_ADMIN:
3954 case PRIV_VFS_EXEC:
3955 case PRIV_VFS_LOOKUP:
3956 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3957 case PRIV_VFS_CHFLAGS_DEV:
3958 case PRIV_VFS_CHOWN:
3959 case PRIV_VFS_CHROOT:
3960 case PRIV_VFS_RETAINSUGID:
3961 case PRIV_VFS_FCHROOT:
3962 case PRIV_VFS_LINK:
3963 case PRIV_VFS_SETGID:
3964 case PRIV_VFS_STAT:
3965 case PRIV_VFS_STICKYFILE:
3966
3967 /*
3968 * As in the non-jail case, non-root users are expected to be
3969 * able to read kernel/phyiscal memory (provided /dev/[k]mem
3970 * exists in the jail and they have permission to access it).
3971 */
3972 case PRIV_KMEM_READ:
3973 return (0);
3974
3975 /*
3976 * Depending on the global setting, allow privilege of
3977 * setting system flags.
3978 */
3979 case PRIV_VFS_SYSFLAGS:
3980 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3981 return (0);
3982 else
3983 return (EPERM);
3984
3985 /*
3986 * Depending on the global setting, allow privilege of
3987 * mounting/unmounting file systems.
3988 */
3989 case PRIV_VFS_MOUNT:
3990 case PRIV_VFS_UNMOUNT:
3991 case PRIV_VFS_MOUNT_NONUSER:
3992 case PRIV_VFS_MOUNT_OWNER:
3993 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3994 cred->cr_prison->pr_enforce_statfs < 2)
3995 return (0);
3996 else
3997 return (EPERM);
3998
3999 /*
4000 * Allow jailed root to bind reserved ports and reuse in-use
4001 * ports.
4002 */
4003 case PRIV_NETINET_RESERVEDPORT:
4004 case PRIV_NETINET_REUSEPORT:
4005 return (0);
4006
4007 /*
4008 * Allow jailed root to set certian IPv4/6 (option) headers.
4009 */
4010 case PRIV_NETINET_SETHDROPTS:
4011 return (0);
4012
4013 /*
4014 * Conditionally allow creating raw sockets in jail.
4015 */
4016 case PRIV_NETINET_RAW:
4017 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
4018 return (0);
4019 else
4020 return (EPERM);
4021
4022 /*
4023 * Since jail implements its own visibility limits on netstat
4024 * sysctls, allow getcred. This allows identd to work in
4025 * jail.
4026 */
4027 case PRIV_NETINET_GETCRED:
4028 return (0);
4029
4030 /*
4031 * Allow jailed root to set loginclass.
4032 */
4033 case PRIV_PROC_SETLOGINCLASS:
4034 return (0);
4035
4036 default:
4037 /*
4038 * In all remaining cases, deny the privilege request. This
4039 * includes almost all network privileges, many system
4040 * configuration privileges.
4041 */
4042 return (EPERM);
4043 }
4044 }
4045
4046 /*
4047 * Return the part of pr2's name that is relative to pr1, or the whole name
4048 * if it does not directly follow.
4049 */
4050
4051 char *
4052 prison_name(struct prison *pr1, struct prison *pr2)
4053 {
4054 char *name;
4055
4056 /* Jails see themselves as "" (if they see themselves at all). */
4057 if (pr1 == pr2)
4058 return "";
4059 name = pr2->pr_name;
4060 if (prison_ischild(pr1, pr2)) {
4061 /*
4062 * pr1 isn't locked (and allprison_lock may not be either)
4063 * so its length can't be counted on. But the number of dots
4064 * can be counted on - and counted.
4065 */
4066 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
4067 name = strchr(name, '.') + 1;
4068 }
4069 return (name);
4070 }
4071
4072 /*
4073 * Return the part of pr2's path that is relative to pr1, or the whole path
4074 * if it does not directly follow.
4075 */
4076 static char *
4077 prison_path(struct prison *pr1, struct prison *pr2)
4078 {
4079 char *path1, *path2;
4080 int len1;
4081
4082 path1 = pr1->pr_path;
4083 path2 = pr2->pr_path;
4084 if (!strcmp(path1, "/"))
4085 return (path2);
4086 len1 = strlen(path1);
4087 if (strncmp(path1, path2, len1))
4088 return (path2);
4089 if (path2[len1] == '\0')
4090 return "/";
4091 if (path2[len1] == '/')
4092 return (path2 + len1);
4093 return (path2);
4094 }
4095
4096
4097 /*
4098 * Jail-related sysctls.
4099 */
4100 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4101 "Jails");
4102
4103 static int
4104 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4105 {
4106 struct xprison *xp;
4107 struct prison *pr, *cpr;
4108 #ifdef INET
4109 struct in_addr *ip4 = NULL;
4110 int ip4s = 0;
4111 #endif
4112 #ifdef INET6
4113 struct in6_addr *ip6 = NULL;
4114 int ip6s = 0;
4115 #endif
4116 int descend, error;
4117
4118 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4119 pr = req->td->td_ucred->cr_prison;
4120 error = 0;
4121 sx_slock(&allprison_lock);
4122 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4123 #if defined(INET) || defined(INET6)
4124 again:
4125 #endif
4126 mtx_lock(&cpr->pr_mtx);
4127 #ifdef INET
4128 if (cpr->pr_ip4s > 0) {
4129 if (ip4s < cpr->pr_ip4s) {
4130 ip4s = cpr->pr_ip4s;
4131 mtx_unlock(&cpr->pr_mtx);
4132 ip4 = realloc(ip4, ip4s *
4133 sizeof(struct in_addr), M_TEMP, M_WAITOK);
4134 goto again;
4135 }
4136 bcopy(cpr->pr_ip4, ip4,
4137 cpr->pr_ip4s * sizeof(struct in_addr));
4138 }
4139 #endif
4140 #ifdef INET6
4141 if (cpr->pr_ip6s > 0) {
4142 if (ip6s < cpr->pr_ip6s) {
4143 ip6s = cpr->pr_ip6s;
4144 mtx_unlock(&cpr->pr_mtx);
4145 ip6 = realloc(ip6, ip6s *
4146 sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4147 goto again;
4148 }
4149 bcopy(cpr->pr_ip6, ip6,
4150 cpr->pr_ip6s * sizeof(struct in6_addr));
4151 }
4152 #endif
4153 if (cpr->pr_ref == 0) {
4154 mtx_unlock(&cpr->pr_mtx);
4155 continue;
4156 }
4157 bzero(xp, sizeof(*xp));
4158 xp->pr_version = XPRISON_VERSION;
4159 xp->pr_id = cpr->pr_id;
4160 xp->pr_state = cpr->pr_uref > 0
4161 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4162 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4163 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4164 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4165 #ifdef INET
4166 xp->pr_ip4s = cpr->pr_ip4s;
4167 #endif
4168 #ifdef INET6
4169 xp->pr_ip6s = cpr->pr_ip6s;
4170 #endif
4171 mtx_unlock(&cpr->pr_mtx);
4172 error = SYSCTL_OUT(req, xp, sizeof(*xp));
4173 if (error)
4174 break;
4175 #ifdef INET
4176 if (xp->pr_ip4s > 0) {
4177 error = SYSCTL_OUT(req, ip4,
4178 xp->pr_ip4s * sizeof(struct in_addr));
4179 if (error)
4180 break;
4181 }
4182 #endif
4183 #ifdef INET6
4184 if (xp->pr_ip6s > 0) {
4185 error = SYSCTL_OUT(req, ip6,
4186 xp->pr_ip6s * sizeof(struct in6_addr));
4187 if (error)
4188 break;
4189 }
4190 #endif
4191 }
4192 sx_sunlock(&allprison_lock);
4193 free(xp, M_TEMP);
4194 #ifdef INET
4195 free(ip4, M_TEMP);
4196 #endif
4197 #ifdef INET6
4198 free(ip6, M_TEMP);
4199 #endif
4200 return (error);
4201 }
4202
4203 SYSCTL_OID(_security_jail, OID_AUTO, list,
4204 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4205 sysctl_jail_list, "S", "List of active jails");
4206
4207 static int
4208 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4209 {
4210 int error, injail;
4211
4212 injail = jailed(req->td->td_ucred);
4213 error = SYSCTL_OUT(req, &injail, sizeof(injail));
4214
4215 return (error);
4216 }
4217
4218 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4219 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4220 sysctl_jail_jailed, "I", "Process in jail?");
4221
4222 static int
4223 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4224 {
4225 int error, havevnet;
4226 #ifdef VIMAGE
4227 struct ucred *cred = req->td->td_ucred;
4228
4229 havevnet = jailed(cred) && prison_owns_vnet(cred);
4230 #else
4231 havevnet = 0;
4232 #endif
4233 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4234
4235 return (error);
4236 }
4237
4238 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4239 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4240 sysctl_jail_vnet, "I", "Jail owns VNET?");
4241
4242 #if defined(INET) || defined(INET6)
4243 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4244 &jail_max_af_ips, 0,
4245 "Number of IP addresses a jail may have at most per address family");
4246 #endif
4247
4248 /*
4249 * Default parameters for jail(2) compatability. For historical reasons,
4250 * the sysctl names have varying similarity to the parameter names. Prisons
4251 * just see their own parameters, and can't change them.
4252 */
4253 static int
4254 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4255 {
4256 struct prison *pr;
4257 int allow, error, i;
4258
4259 pr = req->td->td_ucred->cr_prison;
4260 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4261
4262 /* Get the current flag value, and convert it to a boolean. */
4263 i = (allow & arg2) ? 1 : 0;
4264 if (arg1 != NULL)
4265 i = !i;
4266 error = sysctl_handle_int(oidp, &i, 0, req);
4267 if (error || !req->newptr)
4268 return (error);
4269 i = i ? arg2 : 0;
4270 if (arg1 != NULL)
4271 i ^= arg2;
4272 /*
4273 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4274 * for writing.
4275 */
4276 mtx_lock(&prison0.pr_mtx);
4277 jail_default_allow = (jail_default_allow & ~arg2) | i;
4278 mtx_unlock(&prison0.pr_mtx);
4279 return (0);
4280 }
4281
4282 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4283 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4284 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4285 "Processes in jail can set their hostnames");
4286 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4287 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4288 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4289 "Processes in jail are limited to creating UNIX/IP/route sockets only");
4290 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4291 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4292 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4293 "Processes in jail can use System V IPC primitives");
4294 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4295 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4296 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4297 "Prison root can create raw sockets");
4298 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4299 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4300 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4301 "Processes in jail can alter system file flags");
4302 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4303 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4304 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4305 "Processes in jail can mount/unmount jail-friendly file systems");
4306 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4307 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4308 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4309 "Processes in jail can mount the devfs file system");
4310 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed,
4311 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4312 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I",
4313 "Processes in jail can mount the fdescfs file system");
4314 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4315 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4316 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4317 "Processes in jail can mount the nullfs file system");
4318 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4319 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4320 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4321 "Processes in jail can mount the procfs file system");
4322 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed,
4323 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4324 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I",
4325 "Processes in jail can mount the linprocfs file system");
4326 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed,
4327 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4328 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I",
4329 "Processes in jail can mount the linsysfs file system");
4330 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed,
4331 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4332 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I",
4333 "Processes in jail can mount the tmpfs file system");
4334 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4335 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4336 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4337 "Processes in jail can mount the zfs file system");
4338
4339 static int
4340 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4341 {
4342 struct prison *pr;
4343 int level, error;
4344
4345 pr = req->td->td_ucred->cr_prison;
4346 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4347 error = sysctl_handle_int(oidp, &level, 0, req);
4348 if (error || !req->newptr)
4349 return (error);
4350 *(int *)arg1 = level;
4351 return (0);
4352 }
4353
4354 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4355 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4356 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4357 sysctl_jail_default_level, "I",
4358 "Processes in jail cannot see all mounted file systems");
4359
4360 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4361 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4362 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4363 sysctl_jail_default_level, "I",
4364 "Ruleset for the devfs filesystem in jail");
4365
4366 /*
4367 * Nodes to describe jail parameters. Maximum length of string parameters
4368 * is returned in the string itself, and the other parameters exist merely
4369 * to make themselves and their types known.
4370 */
4371 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4372 "Jail parameters");
4373
4374 int
4375 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4376 {
4377 int i;
4378 long l;
4379 size_t s;
4380 char numbuf[12];
4381
4382 switch (oidp->oid_kind & CTLTYPE)
4383 {
4384 case CTLTYPE_LONG:
4385 case CTLTYPE_ULONG:
4386 l = 0;
4387 #ifdef SCTL_MASK32
4388 if (!(req->flags & SCTL_MASK32))
4389 #endif
4390 return (SYSCTL_OUT(req, &l, sizeof(l)));
4391 case CTLTYPE_INT:
4392 case CTLTYPE_UINT:
4393 i = 0;
4394 return (SYSCTL_OUT(req, &i, sizeof(i)));
4395 case CTLTYPE_STRING:
4396 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4397 return
4398 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4399 case CTLTYPE_STRUCT:
4400 s = (size_t)arg2;
4401 return (SYSCTL_OUT(req, &s, sizeof(s)));
4402 }
4403 return (0);
4404 }
4405
4406 /*
4407 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at
4408 * jail creation time but cannot be changed in an existing jail.
4409 */
4410 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4411 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4412 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4413 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4414 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4415 "I", "Jail secure level");
4416 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I",
4417 "Jail value for kern.osreldate and uname -K");
4418 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN,
4419 "Jail value for kern.osrelease and uname -r");
4420 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4421 "I", "Jail cannot see all mounted file systems");
4422 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4423 "I", "Ruleset for in-jail devfs mounts");
4424 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4425 "B", "Jail persistence");
4426 #ifdef VIMAGE
4427 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4428 "E,jailsys", "Virtual network stack");
4429 #endif
4430 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4431 "B", "Jail is in the process of shutting down");
4432
4433 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4434 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4435 "I", "Current number of child jails");
4436 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4437 "I", "Maximum number of child jails");
4438
4439 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4440 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4441 "Jail hostname");
4442 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4443 "Jail NIS domainname");
4444 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4445 "Jail host UUID");
4446 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4447 "LU", "Jail host ID");
4448
4449 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4450 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4451
4452 #ifdef INET
4453 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4454 "Jail IPv4 address virtualization");
4455 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4456 "S,in_addr,a", "Jail IPv4 addresses");
4457 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4458 "B", "Do (not) use IPv4 source address selection rather than the "
4459 "primary jail IPv4 address.");
4460 #endif
4461 #ifdef INET6
4462 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4463 "Jail IPv6 address virtualization");
4464 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4465 "S,in6_addr,a", "Jail IPv6 addresses");
4466 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4467 "B", "Do (not) use IPv6 source address selection rather than the "
4468 "primary jail IPv6 address.");
4469 #endif
4470
4471 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4472 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4473 "B", "Jail may set hostname");
4474 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4475 "B", "Jail may use SYSV IPC");
4476 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4477 "B", "Jail may create raw sockets");
4478 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4479 "B", "Jail may alter system file flags");
4480 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4481 "B", "Jail may set file quotas");
4482 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4483 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4484
4485 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4486 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4487 "B", "Jail may mount/unmount jail-friendly file systems in general");
4488 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4489 "B", "Jail may mount the devfs file system");
4490 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW,
4491 "B", "Jail may mount the fdescfs file system");
4492 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4493 "B", "Jail may mount the nullfs file system");
4494 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4495 "B", "Jail may mount the procfs file system");
4496 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW,
4497 "B", "Jail may mount the linprocfs file system");
4498 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW,
4499 "B", "Jail may mount the linsysfs file system");
4500 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW,
4501 "B", "Jail may mount the tmpfs file system");
4502 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4503 "B", "Jail may mount the zfs file system");
4504
4505 #ifdef RACCT
4506 void
4507 prison_racct_foreach(void (*callback)(struct racct *racct,
4508 void *arg2, void *arg3), void *arg2, void *arg3)
4509 {
4510 struct prison_racct *prr;
4511
4512 ASSERT_RACCT_ENABLED();
4513
4514 sx_slock(&allprison_lock);
4515 LIST_FOREACH(prr, &allprison_racct, prr_next)
4516 (callback)(prr->prr_racct, arg2, arg3);
4517 sx_sunlock(&allprison_lock);
4518 }
4519
4520 static struct prison_racct *
4521 prison_racct_find_locked(const char *name)
4522 {
4523 struct prison_racct *prr;
4524
4525 ASSERT_RACCT_ENABLED();
4526 sx_assert(&allprison_lock, SA_XLOCKED);
4527
4528 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4529 return (NULL);
4530
4531 LIST_FOREACH(prr, &allprison_racct, prr_next) {
4532 if (strcmp(name, prr->prr_name) != 0)
4533 continue;
4534
4535 /* Found prison_racct with a matching name? */
4536 prison_racct_hold(prr);
4537 return (prr);
4538 }
4539
4540 /* Add new prison_racct. */
4541 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4542 racct_create(&prr->prr_racct);
4543
4544 strcpy(prr->prr_name, name);
4545 refcount_init(&prr->prr_refcount, 1);
4546 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4547
4548 return (prr);
4549 }
4550
4551 struct prison_racct *
4552 prison_racct_find(const char *name)
4553 {
4554 struct prison_racct *prr;
4555
4556 ASSERT_RACCT_ENABLED();
4557
4558 sx_xlock(&allprison_lock);
4559 prr = prison_racct_find_locked(name);
4560 sx_xunlock(&allprison_lock);
4561 return (prr);
4562 }
4563
4564 void
4565 prison_racct_hold(struct prison_racct *prr)
4566 {
4567
4568 ASSERT_RACCT_ENABLED();
4569
4570 refcount_acquire(&prr->prr_refcount);
4571 }
4572
4573 static void
4574 prison_racct_free_locked(struct prison_racct *prr)
4575 {
4576
4577 ASSERT_RACCT_ENABLED();
4578 sx_assert(&allprison_lock, SA_XLOCKED);
4579
4580 if (refcount_release(&prr->prr_refcount)) {
4581 racct_destroy(&prr->prr_racct);
4582 LIST_REMOVE(prr, prr_next);
4583 free(prr, M_PRISON_RACCT);
4584 }
4585 }
4586
4587 void
4588 prison_racct_free(struct prison_racct *prr)
4589 {
4590 int old;
4591
4592 ASSERT_RACCT_ENABLED();
4593 sx_assert(&allprison_lock, SA_UNLOCKED);
4594
4595 old = prr->prr_refcount;
4596 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4597 return;
4598
4599 sx_xlock(&allprison_lock);
4600 prison_racct_free_locked(prr);
4601 sx_xunlock(&allprison_lock);
4602 }
4603
4604 static void
4605 prison_racct_attach(struct prison *pr)
4606 {
4607 struct prison_racct *prr;
4608
4609 ASSERT_RACCT_ENABLED();
4610 sx_assert(&allprison_lock, SA_XLOCKED);
4611
4612 prr = prison_racct_find_locked(pr->pr_name);
4613 KASSERT(prr != NULL, ("cannot find prison_racct"));
4614
4615 pr->pr_prison_racct = prr;
4616 }
4617
4618 /*
4619 * Handle jail renaming. From the racct point of view, renaming means
4620 * moving from one prison_racct to another.
4621 */
4622 static void
4623 prison_racct_modify(struct prison *pr)
4624 {
4625 struct proc *p;
4626 struct ucred *cred;
4627 struct prison_racct *oldprr;
4628
4629 ASSERT_RACCT_ENABLED();
4630
4631 sx_slock(&allproc_lock);
4632 sx_xlock(&allprison_lock);
4633
4634 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4635 sx_xunlock(&allprison_lock);
4636 sx_sunlock(&allproc_lock);
4637 return;
4638 }
4639
4640 oldprr = pr->pr_prison_racct;
4641 pr->pr_prison_racct = NULL;
4642
4643 prison_racct_attach(pr);
4644
4645 /*
4646 * Move resource utilisation records.
4647 */
4648 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4649
4650 /*
4651 * Force rctl to reattach rules to processes.
4652 */
4653 FOREACH_PROC_IN_SYSTEM(p) {
4654 PROC_LOCK(p);
4655 cred = crhold(p->p_ucred);
4656 PROC_UNLOCK(p);
4657 racct_proc_ucred_changed(p, cred, cred);
4658 crfree(cred);
4659 }
4660
4661 sx_sunlock(&allproc_lock);
4662 prison_racct_free_locked(oldprr);
4663 sx_xunlock(&allprison_lock);
4664 }
4665
4666 static void
4667 prison_racct_detach(struct prison *pr)
4668 {
4669
4670 ASSERT_RACCT_ENABLED();
4671 sx_assert(&allprison_lock, SA_UNLOCKED);
4672
4673 if (pr->pr_prison_racct == NULL)
4674 return;
4675 prison_racct_free(pr->pr_prison_racct);
4676 pr->pr_prison_racct = NULL;
4677 }
4678 #endif /* RACCT */
4679
4680 #ifdef DDB
4681
4682 static void
4683 db_show_prison(struct prison *pr)
4684 {
4685 int fi;
4686 #if defined(INET) || defined(INET6)
4687 int ii;
4688 #endif
4689 unsigned jsf;
4690 #ifdef INET6
4691 char ip6buf[INET6_ADDRSTRLEN];
4692 #endif
4693
4694 db_printf("prison %p:\n", pr);
4695 db_printf(" jid = %d\n", pr->pr_id);
4696 db_printf(" name = %s\n", pr->pr_name);
4697 db_printf(" parent = %p\n", pr->pr_parent);
4698 db_printf(" ref = %d\n", pr->pr_ref);
4699 db_printf(" uref = %d\n", pr->pr_uref);
4700 db_printf(" path = %s\n", pr->pr_path);
4701 db_printf(" cpuset = %d\n", pr->pr_cpuset
4702 ? pr->pr_cpuset->cs_id : -1);
4703 #ifdef VIMAGE
4704 db_printf(" vnet = %p\n", pr->pr_vnet);
4705 #endif
4706 db_printf(" root = %p\n", pr->pr_root);
4707 db_printf(" securelevel = %d\n", pr->pr_securelevel);
4708 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4709 db_printf(" children.max = %d\n", pr->pr_childmax);
4710 db_printf(" children.cur = %d\n", pr->pr_childcount);
4711 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4712 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4713 db_printf(" flags = 0x%x", pr->pr_flags);
4714 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4715 fi++)
4716 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4717 db_printf(" %s", pr_flag_names[fi]);
4718 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4719 fi++) {
4720 jsf = pr->pr_flags &
4721 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4722 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4723 pr_flag_jailsys[fi].disable &&
4724 (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4725 : (jsf == pr_flag_jailsys[fi].new) ? "new"
4726 : "inherit");
4727 }
4728 db_printf(" allow = 0x%x", pr->pr_allow);
4729 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4730 fi++)
4731 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4732 db_printf(" %s", pr_allow_names[fi]);
4733 db_printf("\n");
4734 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4735 db_printf(" host.hostname = %s\n", pr->pr_hostname);
4736 db_printf(" host.domainname = %s\n", pr->pr_domainname);
4737 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4738 db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4739 #ifdef INET
4740 db_printf(" ip4s = %d\n", pr->pr_ip4s);
4741 for (ii = 0; ii < pr->pr_ip4s; ii++)
4742 db_printf(" %s %s\n",
4743 ii == 0 ? "ip4.addr =" : " ",
4744 inet_ntoa(pr->pr_ip4[ii]));
4745 #endif
4746 #ifdef INET6
4747 db_printf(" ip6s = %d\n", pr->pr_ip6s);
4748 for (ii = 0; ii < pr->pr_ip6s; ii++)
4749 db_printf(" %s %s\n",
4750 ii == 0 ? "ip6.addr =" : " ",
4751 ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4752 #endif
4753 }
4754
4755 DB_SHOW_COMMAND(prison, db_show_prison_command)
4756 {
4757 struct prison *pr;
4758
4759 if (!have_addr) {
4760 /*
4761 * Show all prisons in the list, and prison0 which is not
4762 * listed.
4763 */
4764 db_show_prison(&prison0);
4765 if (!db_pager_quit) {
4766 TAILQ_FOREACH(pr, &allprison, pr_list) {
4767 db_show_prison(pr);
4768 if (db_pager_quit)
4769 break;
4770 }
4771 }
4772 return;
4773 }
4774
4775 if (addr == 0)
4776 pr = &prison0;
4777 else {
4778 /* Look for a prison with the ID and with references. */
4779 TAILQ_FOREACH(pr, &allprison, pr_list)
4780 if (pr->pr_id == addr && pr->pr_ref > 0)
4781 break;
4782 if (pr == NULL)
4783 /* Look again, without requiring a reference. */
4784 TAILQ_FOREACH(pr, &allprison, pr_list)
4785 if (pr->pr_id == addr)
4786 break;
4787 if (pr == NULL)
4788 /* Assume address points to a valid prison. */
4789 pr = (struct prison *)addr;
4790 }
4791 db_show_prison(pr);
4792 }
4793
4794 #endif /* DDB */
Cache object: 7123a9d3098b5a0dab48b321db1b0fb3
|