FreeBSD kernel kern code
kern_jail.c
Go to the documentation of this file.
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("$BSDSUniX$");
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 = "0",
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 };
210 const size_t pr_allow_names_size = sizeof(pr_allow_names);
211 
212 static char *pr_allow_nonames[] = {
213  "allow.noset_hostname",
214  "allow.nosysvipc",
215  "allow.noraw_sockets",
216  "allow.nochflags",
217  "allow.nomount",
218  "allow.noquotas",
219  "allow.nosocket_af",
220  "allow.mount.nodevfs",
221  "allow.mount.nonullfs",
222  "allow.mount.nozfs",
223  "allow.mount.noprocfs",
224 };
225 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
226 
227 #define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME
228 #define JAIL_DEFAULT_ENFORCE_STATFS 2
229 #define JAIL_DEFAULT_DEVFS_RSNUM 0
230 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
231 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
232 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM;
233 #if defined(INET) || defined(INET6)
234 static unsigned jail_max_af_ips = 255;
235 #endif
236 
237 #ifdef INET
238 static int
239 qcmp_v4(const void *ip1, const void *ip2)
240 {
241  in_addr_t iaa, iab;
242 
243  /*
244  * We need to compare in HBO here to get the list sorted as expected
245  * by the result of the code. Sorting NBO addresses gives you
246  * interesting results. If you do not understand, do not try.
247  */
248  iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
249  iab = ntohl(((const struct in_addr *)ip2)->s_addr);
250 
251  /*
252  * Do not simply return the difference of the two numbers, the int is
253  * not wide enough.
254  */
255  if (iaa > iab)
256  return (1);
257  else if (iaa < iab)
258  return (-1);
259  else
260  return (0);
261 }
262 #endif
263 
264 #ifdef INET6
265 static int
266 qcmp_v6(const void *ip1, const void *ip2)
267 {
268  const struct in6_addr *ia6a, *ia6b;
269  int i, rc;
270 
271  ia6a = (const struct in6_addr *)ip1;
272  ia6b = (const struct in6_addr *)ip2;
273 
274  rc = 0;
275  for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
276  if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
277  rc = 1;
278  else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
279  rc = -1;
280  }
281  return (rc);
282 }
283 #endif
284 
285 /*
286  * struct jail_args {
287  * struct jail *jail;
288  * };
289  */
290 int
291 sys_jail(struct thread *td, struct jail_args *uap)
292 {
293  uint32_t version;
294  int error;
295  struct jail j;
296 
297  error = copyin(uap->jail, &version, sizeof(uint32_t));
298  if (error)
299  return (error);
300 
301  switch (version) {
302  case 0:
303  {
304  struct jail_v0 j0;
305 
306  /* FreeBSD single IPv4 jails. */
307  bzero(&j, sizeof(struct jail));
308  error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
309  if (error)
310  return (error);
311  j.version = j0.version;
312  j.path = j0.path;
313  j.hostname = j0.hostname;
314  j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */
315  break;
316  }
317 
318  case 1:
319  /*
320  * Version 1 was used by multi-IPv4 jail implementations
321  * that never made it into the official kernel.
322  */
323  return (EINVAL);
324 
325  case 2: /* JAIL_API_VERSION */
326  /* FreeBSD multi-IPv4/IPv6,noIP jails. */
327  error = copyin(uap->jail, &j, sizeof(struct jail));
328  if (error)
329  return (error);
330  break;
331 
332  default:
333  /* Sci-Fi jails are not supported, sorry. */
334  return (EINVAL);
335  }
336  return (kern_jail(td, &j));
337 }
338 
339 int
340 kern_jail(struct thread *td, struct jail *j)
341 {
342  struct iovec optiov[2 * (4
343  + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
344 #ifdef INET
345  + 1
346 #endif
347 #ifdef INET6
348  + 1
349 #endif
350  )];
351  struct uio opt;
352  char *u_path, *u_hostname, *u_name;
353 #ifdef INET
354  uint32_t ip4s;
355  struct in_addr *u_ip4;
356 #endif
357 #ifdef INET6
358  struct in6_addr *u_ip6;
359 #endif
360  size_t tmplen;
361  int error, enforce_statfs, fi;
362 
363  bzero(&optiov, sizeof(optiov));
364  opt.uio_iov = optiov;
365  opt.uio_iovcnt = 0;
366  opt.uio_offset = -1;
367  opt.uio_resid = -1;
368  opt.uio_segflg = UIO_SYSSPACE;
369  opt.uio_rw = UIO_READ;
370  opt.uio_td = td;
371 
372  /* Set permissions for top-level jails from sysctls. */
373  if (!jailed(td->td_ucred)) {
374  for (fi = 0; fi < sizeof(pr_allow_names) /
375  sizeof(pr_allow_names[0]); fi++) {
376  optiov[opt.uio_iovcnt].iov_base =
377  (jail_default_allow & (1 << fi))
378  ? pr_allow_names[fi] : pr_allow_nonames[fi];
379  optiov[opt.uio_iovcnt].iov_len =
380  strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
381  opt.uio_iovcnt += 2;
382  }
383  optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
384  optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
385  opt.uio_iovcnt++;
386  enforce_statfs = jail_default_enforce_statfs;
387  optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
388  optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
389  opt.uio_iovcnt++;
390  }
391 
392  tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
393 #ifdef INET
394  ip4s = (j->version == 0) ? 1 : j->ip4s;
395  if (ip4s > jail_max_af_ips)
396  return (EINVAL);
397  tmplen += ip4s * sizeof(struct in_addr);
398 #else
399  if (j->ip4s > 0)
400  return (EINVAL);
401 #endif
402 #ifdef INET6
403  if (j->ip6s > jail_max_af_ips)
404  return (EINVAL);
405  tmplen += j->ip6s * sizeof(struct in6_addr);
406 #else
407  if (j->ip6s > 0)
408  return (EINVAL);
409 #endif
410  u_path = malloc(tmplen, M_TEMP, M_WAITOK);
411  u_hostname = u_path + MAXPATHLEN;
412  u_name = u_hostname + MAXHOSTNAMELEN;
413 #ifdef INET
414  u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
415 #endif
416 #ifdef INET6
417 #ifdef INET
418  u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
419 #else
420  u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
421 #endif
422 #endif
423  optiov[opt.uio_iovcnt].iov_base = "path";
424  optiov[opt.uio_iovcnt].iov_len = sizeof("path");
425  opt.uio_iovcnt++;
426  optiov[opt.uio_iovcnt].iov_base = u_path;
427  error = copyinstr(j->path, u_path, MAXPATHLEN,
428  &optiov[opt.uio_iovcnt].iov_len);
429  if (error) {
430  free(u_path, M_TEMP);
431  return (error);
432  }
433  opt.uio_iovcnt++;
434  optiov[opt.uio_iovcnt].iov_base = "host.hostname";
435  optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
436  opt.uio_iovcnt++;
437  optiov[opt.uio_iovcnt].iov_base = u_hostname;
438  error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
439  &optiov[opt.uio_iovcnt].iov_len);
440  if (error) {
441  free(u_path, M_TEMP);
442  return (error);
443  }
444  opt.uio_iovcnt++;
445  if (j->jailname != NULL) {
446  optiov[opt.uio_iovcnt].iov_base = "name";
447  optiov[opt.uio_iovcnt].iov_len = sizeof("name");
448  opt.uio_iovcnt++;
449  optiov[opt.uio_iovcnt].iov_base = u_name;
450  error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
451  &optiov[opt.uio_iovcnt].iov_len);
452  if (error) {
453  free(u_path, M_TEMP);
454  return (error);
455  }
456  opt.uio_iovcnt++;
457  }
458 #ifdef INET
459  optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
460  optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
461  opt.uio_iovcnt++;
462  optiov[opt.uio_iovcnt].iov_base = u_ip4;
463  optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
464  if (j->version == 0)
465  u_ip4->s_addr = j->ip4s;
466  else {
467  error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
468  if (error) {
469  free(u_path, M_TEMP);
470  return (error);
471  }
472  }
473  opt.uio_iovcnt++;
474 #endif
475 #ifdef INET6
476  optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
477  optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
478  opt.uio_iovcnt++;
479  optiov[opt.uio_iovcnt].iov_base = u_ip6;
480  optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
481  error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
482  if (error) {
483  free(u_path, M_TEMP);
484  return (error);
485  }
486  opt.uio_iovcnt++;
487 #endif
488  KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
489  ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
490  error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
491  free(u_path, M_TEMP);
492  return (error);
493 }
494 
495 
496 /*
497  * struct jail_set_args {
498  * struct iovec *iovp;
499  * unsigned int iovcnt;
500  * int flags;
501  * };
502  */
503 int
504 sys_jail_set(struct thread *td, struct jail_set_args *uap)
505 {
506  struct uio *auio;
507  int error;
508 
509  /* Check that we have an even number of iovecs. */
510  if (uap->iovcnt & 1)
511  return (EINVAL);
512 
513  error = copyinuio(uap->iovp, uap->iovcnt, &auio);
514  if (error)
515  return (error);
516  error = kern_jail_set(td, auio, uap->flags);
517  free(auio, M_IOV);
518  return (error);
519 }
520 
521 int
522 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
523 {
524  struct nameidata nd;
525 #ifdef INET
526  struct in_addr *ip4;
527 #endif
528 #ifdef INET6
529  struct in6_addr *ip6;
530 #endif
531  struct vfsopt *opt;
532  struct vfsoptlist *opts;
533  struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
534  struct vnode *root;
535  char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
536  char *g_path;
537 #if defined(INET) || defined(INET6)
538  struct prison *tppr;
539  void *op;
540 #endif
541  unsigned long hid;
542  size_t namelen, onamelen;
543  int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
544  int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel;
545  int fi, jid, jsys, len, level;
546  int childmax, rsnum, slevel, vfslocked;
547  int fullpath_disabled;
548 #if defined(INET) || defined(INET6)
549  int ii, ij;
550 #endif
551 #ifdef INET
552  int ip4s, redo_ip4;
553 #endif
554 #ifdef INET6
555  int ip6s, redo_ip6;
556 #endif
557  uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
558  unsigned tallow;
559  char numbuf[12];
560 
561  error = priv_check(td, PRIV_JAIL_SET);
562  if (!error && (flags & JAIL_ATTACH))
563  error = priv_check(td, PRIV_JAIL_ATTACH);
564  if (error)
565  return (error);
566  mypr = ppr = td->td_ucred->cr_prison;
567  if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
568  return (EPERM);
569  if (flags & ~JAIL_SET_MASK)
570  return (EINVAL);
571 
572  /*
573  * Check all the parameters before committing to anything. Not all
574  * errors can be caught early, but we may as well try. Also, this
575  * takes care of some expensive stuff (path lookup) before getting
576  * the allprison lock.
577  *
578  * XXX Jails are not filesystems, and jail parameters are not mount
579  * options. But it makes more sense to re-use the vfsopt code
580  * than duplicate it under a different name.
581  */
582  error = vfs_buildopts(optuio, &opts);
583  if (error)
584  return (error);
585 #ifdef INET
586  ip4 = NULL;
587 #endif
588 #ifdef INET6
589  ip6 = NULL;
590 #endif
591  g_path = NULL;
592 
593  error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
594  if (error == ENOENT)
595  jid = 0;
596  else if (error != 0)
597  goto done_free;
598 
599  error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
600  if (error == ENOENT)
601  gotslevel = 0;
602  else if (error != 0)
603  goto done_free;
604  else
605  gotslevel = 1;
606 
607  error =
608  vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
609  if (error == ENOENT)
610  gotchildmax = 0;
611  else if (error != 0)
612  goto done_free;
613  else
614  gotchildmax = 1;
615 
616  error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
617  if (error == ENOENT)
618  gotenforce = 0;
619  else if (error != 0)
620  goto done_free;
621  else if (enforce < 0 || enforce > 2) {
622  error = EINVAL;
623  goto done_free;
624  } else
625  gotenforce = 1;
626 
627  error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum));
628  if (error == ENOENT)
629  gotrsnum = 0;
630  else if (error != 0)
631  goto done_free;
632  else
633  gotrsnum = 1;
634 
635  pr_flags = ch_flags = 0;
636  for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
637  fi++) {
638  if (pr_flag_names[fi] == NULL)
639  continue;
640  vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
641  vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
642  }
643  ch_flags |= pr_flags;
644  for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
645  fi++) {
646  error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
647  sizeof(jsys));
648  if (error == ENOENT)
649  continue;
650  if (error != 0)
651  goto done_free;
652  switch (jsys) {
653  case JAIL_SYS_DISABLE:
654  if (!pr_flag_jailsys[fi].disable) {
655  error = EINVAL;
656  goto done_free;
657  }
658  pr_flags |= pr_flag_jailsys[fi].disable;
659  break;
660  case JAIL_SYS_NEW:
661  pr_flags |= pr_flag_jailsys[fi].new;
662  break;
663  case JAIL_SYS_INHERIT:
664  break;
665  default:
666  error = EINVAL;
667  goto done_free;
668  }
669  ch_flags |=
670  pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
671  }
672  if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
673  && !(pr_flags & PR_PERSIST)) {
674  error = EINVAL;
675  vfs_opterror(opts, "new jail must persist or attach");
676  goto done_errmsg;
677  }
678 #ifdef VIMAGE
679  if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
680  error = EINVAL;
681  vfs_opterror(opts, "vnet cannot be changed after creation");
682  goto done_errmsg;
683  }
684 #endif
685 #ifdef INET
686  if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
687  error = EINVAL;
688  vfs_opterror(opts, "ip4 cannot be changed after creation");
689  goto done_errmsg;
690  }
691 #endif
692 #ifdef INET6
693  if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
694  error = EINVAL;
695  vfs_opterror(opts, "ip6 cannot be changed after creation");
696  goto done_errmsg;
697  }
698 #endif
699 
700  pr_allow = ch_allow = 0;
701  for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
702  fi++) {
703  vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
704  vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
705  }
706  ch_allow |= pr_allow;
707 
708  error = vfs_getopt(opts, "name", (void **)&name, &len);
709  if (error == ENOENT)
710  name = NULL;
711  else if (error != 0)
712  goto done_free;
713  else {
714  if (len == 0 || name[len - 1] != '\0') {
715  error = EINVAL;
716  goto done_free;
717  }
718  if (len > MAXHOSTNAMELEN) {
719  error = ENAMETOOLONG;
720  goto done_free;
721  }
722  }
723 
724  error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
725  if (error == ENOENT)
726  host = NULL;
727  else if (error != 0)
728  goto done_free;
729  else {
730  ch_flags |= PR_HOST;
731  pr_flags |= PR_HOST;
732  if (len == 0 || host[len - 1] != '\0') {
733  error = EINVAL;
734  goto done_free;
735  }
736  if (len > MAXHOSTNAMELEN) {
737  error = ENAMETOOLONG;
738  goto done_free;
739  }
740  }
741 
742  error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
743  if (error == ENOENT)
744  domain = NULL;
745  else if (error != 0)
746  goto done_free;
747  else {
748  ch_flags |= PR_HOST;
749  pr_flags |= PR_HOST;
750  if (len == 0 || domain[len - 1] != '\0') {
751  error = EINVAL;
752  goto done_free;
753  }
754  if (len > MAXHOSTNAMELEN) {
755  error = ENAMETOOLONG;
756  goto done_free;
757  }
758  }
759 
760  error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
761  if (error == ENOENT)
762  uuid = NULL;
763  else if (error != 0)
764  goto done_free;
765  else {
766  ch_flags |= PR_HOST;
767  pr_flags |= PR_HOST;
768  if (len == 0 || uuid[len - 1] != '\0') {
769  error = EINVAL;
770  goto done_free;
771  }
772  if (len > HOSTUUIDLEN) {
773  error = ENAMETOOLONG;
774  goto done_free;
775  }
776  }
777 
778 #ifdef COMPAT_32BIT
779  if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
780  uint32_t hid32;
781 
782  error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
783  hid = hid32;
784  } else
785 #endif
786  error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
787  if (error == ENOENT)
788  gothid = 0;
789  else if (error != 0)
790  goto done_free;
791  else {
792  gothid = 1;
793  ch_flags |= PR_HOST;
794  pr_flags |= PR_HOST;
795  }
796 
797 #ifdef INET
798  error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
799  if (error == ENOENT)
800  ip4s = 0;
801  else if (error != 0)
802  goto done_free;
803  else if (ip4s & (sizeof(*ip4) - 1)) {
804  error = EINVAL;
805  goto done_free;
806  } else {
807  ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
808  if (ip4s == 0)
809  pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
810  else {
811  pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
812  ip4s /= sizeof(*ip4);
813  if (ip4s > jail_max_af_ips) {
814  error = EINVAL;
815  vfs_opterror(opts, "too many IPv4 addresses");
816  goto done_errmsg;
817  }
818  ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
819  bcopy(op, ip4, ip4s * sizeof(*ip4));
820  /*
821  * IP addresses are all sorted but ip[0] to preserve
822  * the primary IP address as given from userland.
823  * This special IP is used for unbound outgoing
824  * connections as well for "loopback" traffic in case
825  * source address selection cannot find any more fitting
826  * address to connect from.
827  */
828  if (ip4s > 1)
829  qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
830  /*
831  * Check for duplicate addresses and do some simple
832  * zero and broadcast checks. If users give other bogus
833  * addresses it is their problem.
834  *
835  * We do not have to care about byte order for these
836  * checks so we will do them in NBO.
837  */
838  for (ii = 0; ii < ip4s; ii++) {
839  if (ip4[ii].s_addr == INADDR_ANY ||
840  ip4[ii].s_addr == INADDR_BROADCAST) {
841  error = EINVAL;
842  goto done_free;
843  }
844  if ((ii+1) < ip4s &&
845  (ip4[0].s_addr == ip4[ii+1].s_addr ||
846  ip4[ii].s_addr == ip4[ii+1].s_addr)) {
847  error = EINVAL;
848  goto done_free;
849  }
850  }
851  }
852  }
853 #endif
854 
855 #ifdef INET6
856  error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
857  if (error == ENOENT)
858  ip6s = 0;
859  else if (error != 0)
860  goto done_free;
861  else if (ip6s & (sizeof(*ip6) - 1)) {
862  error = EINVAL;
863  goto done_free;
864  } else {
865  ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
866  if (ip6s == 0)
867  pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
868  else {
869  pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
870  ip6s /= sizeof(*ip6);
871  if (ip6s > jail_max_af_ips) {
872  error = EINVAL;
873  vfs_opterror(opts, "too many IPv6 addresses");
874  goto done_errmsg;
875  }
876  ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
877  bcopy(op, ip6, ip6s * sizeof(*ip6));
878  if (ip6s > 1)
879  qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
880  for (ii = 0; ii < ip6s; ii++) {
881  if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
882  error = EINVAL;
883  goto done_free;
884  }
885  if ((ii+1) < ip6s &&
886  (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
887  IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
888  {
889  error = EINVAL;
890  goto done_free;
891  }
892  }
893  }
894  }
895 #endif
896 
897 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
898  if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
899  error = EINVAL;
900  vfs_opterror(opts,
901  "vnet jails cannot have IP address restrictions");
902  goto done_errmsg;
903  }
904 #endif
905 
906  fullpath_disabled = 0;
907  root = NULL;
908  error = vfs_getopt(opts, "path", (void **)&path, &len);
909  if (error == ENOENT)
910  path = NULL;
911  else if (error != 0)
912  goto done_free;
913  else {
914  if (flags & JAIL_UPDATE) {
915  error = EINVAL;
916  vfs_opterror(opts,
917  "path cannot be changed after creation");
918  goto done_errmsg;
919  }
920  if (len == 0 || path[len - 1] != '\0') {
921  error = EINVAL;
922  goto done_free;
923  }
924  NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_SYSSPACE,
925  path, td);
926  error = namei(&nd);
927  if (error)
928  goto done_free;
929  vfslocked = NDHASGIANT(&nd);
930  root = nd.ni_vp;
931  NDFREE(&nd, NDF_ONLY_PNBUF);
932  g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
933  strlcpy(g_path, path, MAXPATHLEN);
934  error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN);
935  if (error == 0)
936  path = g_path;
937  else if (error == ENODEV) {
938  /* proceed if sysctl debug.disablefullpath == 1 */
939  fullpath_disabled = 1;
940  if (len < 2 || (len == 2 && path[0] == '/'))
941  path = NULL;
942  } else {
943  /* exit on other errors */
944  VFS_UNLOCK_GIANT(vfslocked);
945  goto done_free;
946  }
947  if (root->v_type != VDIR) {
948  error = ENOTDIR;
949  vput(root);
950  VFS_UNLOCK_GIANT(vfslocked);
951  goto done_free;
952  }
953  VOP_UNLOCK(root, 0);
954  VFS_UNLOCK_GIANT(vfslocked);
955  if (fullpath_disabled) {
956  /* Leave room for a real-root full pathname. */
957  if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
958  ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
959  error = ENAMETOOLONG;
960  goto done_free;
961  }
962  }
963  }
964 
965  /*
966  * Grab the allprison lock before letting modules check their
967  * parameters. Once we have it, do not let go so we'll have a
968  * consistent view of the OSD list.
969  */
970  sx_xlock(&allprison_lock);
971  error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
972  if (error)
973  goto done_unlock_list;
974 
975  /* By now, all parameters should have been noted. */
976  TAILQ_FOREACH(opt, opts, link) {
977  if (!opt->seen && strcmp(opt->name, "errmsg")) {
978  error = EINVAL;
979  vfs_opterror(opts, "unknown parameter: %s", opt->name);
980  goto done_unlock_list;
981  }
982  }
983 
984  /*
985  * See if we are creating a new record or updating an existing one.
986  * This abuses the file error codes ENOENT and EEXIST.
987  */
988  cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
989  if (!cuflags) {
990  error = EINVAL;
991  vfs_opterror(opts, "no valid operation (create or update)");
992  goto done_unlock_list;
993  }
994  pr = NULL;
995  namelc = NULL;
996  if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
997  namelc = strrchr(name, '.');
998  jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
999  if (*p != '\0')
1000  jid = 0;
1001  }
1002  if (jid != 0) {
1003  /*
1004  * See if a requested jid already exists. There is an
1005  * information leak here if the jid exists but is not within
1006  * the caller's jail hierarchy. Jail creators will get EEXIST
1007  * even though they cannot see the jail, and CREATE | UPDATE
1008  * will return ENOENT which is not normally a valid error.
1009  */
1010  if (jid < 0) {
1011  error = EINVAL;
1012  vfs_opterror(opts, "negative jid");
1013  goto done_unlock_list;
1014  }
1015  pr = prison_find(jid);
1016  if (pr != NULL) {
1017  ppr = pr->pr_parent;
1018  /* Create: jid must not exist. */
1019  if (cuflags == JAIL_CREATE) {
1020  mtx_unlock(&pr->pr_mtx);
1021  error = EEXIST;
1022  vfs_opterror(opts, "jail %d already exists",
1023  jid);
1024  goto done_unlock_list;
1025  }
1026  if (!prison_ischild(mypr, pr)) {
1027  mtx_unlock(&pr->pr_mtx);
1028  pr = NULL;
1029  } else if (pr->pr_uref == 0) {
1030  if (!(flags & JAIL_DYING)) {
1031  mtx_unlock(&pr->pr_mtx);
1032  error = ENOENT;
1033  vfs_opterror(opts, "jail %d is dying",
1034  jid);
1035  goto done_unlock_list;
1036  } else if ((flags & JAIL_ATTACH) ||
1037  (pr_flags & PR_PERSIST)) {
1038  /*
1039  * A dying jail might be resurrected
1040  * (via attach or persist), but first
1041  * it must determine if another jail
1042  * has claimed its name. Accomplish
1043  * this by implicitly re-setting the
1044  * name.
1045  */
1046  if (name == NULL)
1047  name = prison_name(mypr, pr);
1048  }
1049  }
1050  }
1051  if (pr == NULL) {
1052  /* Update: jid must exist. */
1053  if (cuflags == JAIL_UPDATE) {
1054  error = ENOENT;
1055  vfs_opterror(opts, "jail %d not found", jid);
1056  goto done_unlock_list;
1057  }
1058  }
1059  }
1060  /*
1061  * If the caller provided a name, look for a jail by that name.
1062  * This has different semantics for creates and updates keyed by jid
1063  * (where the name must not already exist in a different jail),
1064  * and updates keyed by the name itself (where the name must exist
1065  * because that is the jail being updated).
1066  */
1067  if (name != NULL) {
1068  namelc = strrchr(name, '.');
1069  if (namelc == NULL)
1070  namelc = name;
1071  else {
1072  /*
1073  * This is a hierarchical name. Split it into the
1074  * parent and child names, and make sure the parent
1075  * exists or matches an already found jail.
1076  */
1077  *namelc = '\0';
1078  if (pr != NULL) {
1079  if (strncmp(name, ppr->pr_name, namelc - name)
1080  || ppr->pr_name[namelc - name] != '\0') {
1081  mtx_unlock(&pr->pr_mtx);
1082  error = EINVAL;
1083  vfs_opterror(opts,
1084  "cannot change jail's parent");
1085  goto done_unlock_list;
1086  }
1087  } else {
1088  ppr = prison_find_name(mypr, name);
1089  if (ppr == NULL) {
1090  error = ENOENT;
1091  vfs_opterror(opts,
1092  "jail \"%s\" not found", name);
1093  goto done_unlock_list;
1094  }
1095  mtx_unlock(&ppr->pr_mtx);
1096  }
1097  name = ++namelc;
1098  }
1099  if (name[0] != '\0') {
1100  namelen =
1101  (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
1102  name_again:
1103  deadpr = NULL;
1104  FOREACH_PRISON_CHILD(ppr, tpr) {
1105  if (tpr != pr && tpr->pr_ref > 0 &&
1106  !strcmp(tpr->pr_name + namelen, name)) {
1107  if (pr == NULL &&
1108  cuflags != JAIL_CREATE) {
1109  mtx_lock(&tpr->pr_mtx);
1110  if (tpr->pr_ref > 0) {
1111  /*
1112  * Use this jail
1113  * for updates.
1114  */
1115  if (tpr->pr_uref > 0) {
1116  pr = tpr;
1117  break;
1118  }
1119  deadpr = tpr;
1120  }
1121  mtx_unlock(&tpr->pr_mtx);
1122  } else if (tpr->pr_uref > 0) {
1123  /*
1124  * Create, or update(jid):
1125  * name must not exist in an
1126  * active sibling jail.
1127  */
1128  error = EEXIST;
1129  if (pr != NULL)
1130  mtx_unlock(&pr->pr_mtx);
1131  vfs_opterror(opts,
1132  "jail \"%s\" already exists",
1133  name);
1134  goto done_unlock_list;
1135  }
1136  }
1137  }
1138  /* If no active jail is found, use a dying one. */
1139  if (deadpr != NULL && pr == NULL) {
1140  if (flags & JAIL_DYING) {
1141  mtx_lock(&deadpr->pr_mtx);
1142  if (deadpr->pr_ref == 0) {
1143  mtx_unlock(&deadpr->pr_mtx);
1144  goto name_again;
1145  }
1146  pr = deadpr;
1147  } else if (cuflags == JAIL_UPDATE) {
1148  error = ENOENT;
1149  vfs_opterror(opts,
1150  "jail \"%s\" is dying", name);
1151  goto done_unlock_list;
1152  }
1153  }
1154  /* Update: name must exist if no jid. */
1155  else if (cuflags == JAIL_UPDATE && pr == NULL) {
1156  error = ENOENT;
1157  vfs_opterror(opts, "jail \"%s\" not found",
1158  name);
1159  goto done_unlock_list;
1160  }
1161  }
1162  }
1163  /* Update: must provide a jid or name. */
1164  else if (cuflags == JAIL_UPDATE && pr == NULL) {
1165  error = ENOENT;
1166  vfs_opterror(opts, "update specified no jail");
1167  goto done_unlock_list;
1168  }
1169 
1170  /* If there's no prison to update, create a new one and link it in. */
1171  if (pr == NULL) {
1172  for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
1173  if (tpr->pr_childcount >= tpr->pr_childmax) {
1174  error = EPERM;
1175  vfs_opterror(opts, "prison limit exceeded");
1176  goto done_unlock_list;
1177  }
1178  created = 1;
1179  mtx_lock(&ppr->pr_mtx);
1180  if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
1181  mtx_unlock(&ppr->pr_mtx);
1182  error = ENOENT;
1183  vfs_opterror(opts, "parent jail went away!");
1184  goto done_unlock_list;
1185  }
1186  ppr->pr_ref++;
1187  ppr->pr_uref++;
1188  mtx_unlock(&ppr->pr_mtx);
1189  pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
1190  if (jid == 0) {
1191  /* Find the next free jid. */
1192  jid = lastprid + 1;
1193  findnext:
1194  if (jid == JAIL_MAX)
1195  jid = 1;
1196  TAILQ_FOREACH(tpr, &allprison, pr_list) {
1197  if (tpr->pr_id < jid)
1198  continue;
1199  if (tpr->pr_id > jid || tpr->pr_ref == 0) {
1200  TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1201  break;
1202  }
1203  if (jid == lastprid) {
1204  error = EAGAIN;
1205  vfs_opterror(opts,
1206  "no available jail IDs");
1207  free(pr, M_PRISON);
1208  prison_deref(ppr, PD_DEREF |
1209  PD_DEUREF | PD_LIST_XLOCKED);
1210  goto done_releroot;
1211  }
1212  jid++;
1213  goto findnext;
1214  }
1215  lastprid = jid;
1216  } else {
1217  /*
1218  * The jail already has a jid (that did not yet exist),
1219  * so just find where to insert it.
1220  */
1221  TAILQ_FOREACH(tpr, &allprison, pr_list)
1222  if (tpr->pr_id >= jid) {
1223  TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
1224  break;
1225  }
1226  }
1227  if (tpr == NULL)
1228  TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
1229  LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
1230  for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
1231  tpr->pr_childcount++;
1232 
1233  pr->pr_parent = ppr;
1234  pr->pr_id = jid;
1235 
1236  /* Set some default values, and inherit some from the parent. */
1237  if (name == NULL)
1238  name = "";
1239  if (path == NULL) {
1240  path = "/";
1241  root = mypr->pr_root;
1242  vref(root);
1243  }
1244  strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
1245  pr->pr_flags |= PR_HOST;
1246 #if defined(INET) || defined(INET6)
1247 #ifdef VIMAGE
1248  if (!(pr_flags & PR_VNET))
1249 #endif
1250  {
1251 #ifdef INET
1252  if (!(ch_flags & PR_IP4_USER))
1253  pr->pr_flags |=
1254  PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
1255  else if (!(pr_flags & PR_IP4_USER)) {
1256  pr->pr_flags |= ppr->pr_flags & PR_IP4;
1257  if (ppr->pr_ip4 != NULL) {
1258  pr->pr_ip4s = ppr->pr_ip4s;
1259  pr->pr_ip4 = malloc(pr->pr_ip4s *
1260  sizeof(struct in_addr), M_PRISON,
1261  M_WAITOK);
1262  bcopy(ppr->pr_ip4, pr->pr_ip4,
1263  pr->pr_ip4s * sizeof(*pr->pr_ip4));
1264  }
1265  }
1266 #endif
1267 #ifdef INET6
1268  if (!(ch_flags & PR_IP6_USER))
1269  pr->pr_flags |=
1270  PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
1271  else if (!(pr_flags & PR_IP6_USER)) {
1272  pr->pr_flags |= ppr->pr_flags & PR_IP6;
1273  if (ppr->pr_ip6 != NULL) {
1274  pr->pr_ip6s = ppr->pr_ip6s;
1275  pr->pr_ip6 = malloc(pr->pr_ip6s *
1276  sizeof(struct in6_addr), M_PRISON,
1277  M_WAITOK);
1278  bcopy(ppr->pr_ip6, pr->pr_ip6,
1279  pr->pr_ip6s * sizeof(*pr->pr_ip6));
1280  }
1281  }
1282 #endif
1283  }
1284 #endif
1285  /* Source address selection is always on by default. */
1286  pr->pr_flags |= _PR_IP_SADDRSEL;
1287 
1288  pr->pr_securelevel = ppr->pr_securelevel;
1289  pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
1290  pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
1291  pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum;
1292 
1293  LIST_INIT(&pr->pr_children);
1294  mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
1295 
1296 #ifdef VIMAGE
1297  /* Allocate a new vnet if specified. */
1298  pr->pr_vnet = (pr_flags & PR_VNET)
1299  ? vnet_alloc() : ppr->pr_vnet;
1300 #endif
1301  /*
1302  * Allocate a dedicated cpuset for each jail.
1303  * Unlike other initial settings, this may return an erorr.
1304  */
1305  error = cpuset_create_root(ppr, &pr->pr_cpuset);
1306  if (error) {
1307  prison_deref(pr, PD_LIST_XLOCKED);
1308  goto done_releroot;
1309  }
1310 
1311  mtx_lock(&pr->pr_mtx);
1312  /*
1313  * New prisons do not yet have a reference, because we do not
1314  * want other to see the incomplete prison once the
1315  * allprison_lock is downgraded.
1316  */
1317  } else {
1318  created = 0;
1319  /*
1320  * Grab a reference for existing prisons, to ensure they
1321  * continue to exist for the duration of the call.
1322  */
1323  pr->pr_ref++;
1324 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
1325  if ((pr->pr_flags & PR_VNET) &&
1326  (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
1327  error = EINVAL;
1328  vfs_opterror(opts,
1329  "vnet jails cannot have IP address restrictions");
1330  goto done_deref_locked;
1331  }
1332 #endif
1333 #ifdef INET
1334  if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1335  error = EINVAL;
1336  vfs_opterror(opts,
1337  "ip4 cannot be changed after creation");
1338  goto done_deref_locked;
1339  }
1340 #endif
1341 #ifdef INET6
1342  if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
1343  error = EINVAL;
1344  vfs_opterror(opts,
1345  "ip6 cannot be changed after creation");
1346  goto done_deref_locked;
1347  }
1348 #endif
1349  }
1350 
1351  /* Do final error checking before setting anything. */
1352  if (gotslevel) {
1353  if (slevel < ppr->pr_securelevel) {
1354  error = EPERM;
1355  goto done_deref_locked;
1356  }
1357  }
1358  if (gotchildmax) {
1359  if (childmax >= ppr->pr_childmax) {
1360  error = EPERM;
1361  goto done_deref_locked;
1362  }
1363  }
1364  if (gotenforce) {
1365  if (enforce < ppr->pr_enforce_statfs) {
1366  error = EPERM;
1367  goto done_deref_locked;
1368  }
1369  }
1370  if (gotrsnum) {
1371  /*
1372  * devfs_rsnum is a uint16_t
1373  */
1374  if (rsnum < 0 || rsnum > 65535) {
1375  error = EINVAL;
1376  goto done_deref_locked;
1377  }
1378  /*
1379  * Nested jails always inherit parent's devfs ruleset
1380  */
1381  if (jailed(td->td_ucred)) {
1382  if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) {
1383  error = EPERM;
1384  goto done_deref_locked;
1385  } else
1386  rsnum = ppr->pr_devfs_rsnum;
1387  }
1388  }
1389 #ifdef INET
1390  if (ip4s > 0) {
1391  if (ppr->pr_flags & PR_IP4) {
1392  /*
1393  * Make sure the new set of IP addresses is a
1394  * subset of the parent's list. Don't worry
1395  * about the parent being unlocked, as any
1396  * setting is done with allprison_lock held.
1397  */
1398  for (ij = 0; ij < ppr->pr_ip4s; ij++)
1399  if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
1400  break;
1401  if (ij == ppr->pr_ip4s) {
1402  error = EPERM;
1403  goto done_deref_locked;
1404  }
1405  if (ip4s > 1) {
1406  for (ii = ij = 1; ii < ip4s; ii++) {
1407  if (ip4[ii].s_addr ==
1408  ppr->pr_ip4[0].s_addr)
1409  continue;
1410  for (; ij < ppr->pr_ip4s; ij++)
1411  if (ip4[ii].s_addr ==
1412  ppr->pr_ip4[ij].s_addr)
1413  break;
1414  if (ij == ppr->pr_ip4s)
1415  break;
1416  }
1417  if (ij == ppr->pr_ip4s) {
1418  error = EPERM;
1419  goto done_deref_locked;
1420  }
1421  }
1422  }
1423  /*
1424  * Check for conflicting IP addresses. We permit them
1425  * if there is no more than one IP on each jail. If
1426  * there is a duplicate on a jail with more than one
1427  * IP stop checking and return error.
1428  */
1429  tppr = ppr;
1430 #ifdef VIMAGE
1431  for (; tppr != &prison0; tppr = tppr->pr_parent)
1432  if (tppr->pr_flags & PR_VNET)
1433  break;
1434 #endif
1435  FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1436  if (tpr == pr ||
1437 #ifdef VIMAGE
1438  (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1439 #endif
1440  tpr->pr_uref == 0) {
1441  descend = 0;
1442  continue;
1443  }
1444  if (!(tpr->pr_flags & PR_IP4_USER))
1445  continue;
1446  descend = 0;
1447  if (tpr->pr_ip4 == NULL ||
1448  (ip4s == 1 && tpr->pr_ip4s == 1))
1449  continue;
1450  for (ii = 0; ii < ip4s; ii++) {
1451  if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
1452  error = EADDRINUSE;
1453  vfs_opterror(opts,
1454  "IPv4 addresses clash");
1455  goto done_deref_locked;
1456  }
1457  }
1458  }
1459  }
1460 #endif
1461 #ifdef INET6
1462  if (ip6s > 0) {
1463  if (ppr->pr_flags & PR_IP6) {
1464  /*
1465  * Make sure the new set of IP addresses is a
1466  * subset of the parent's list.
1467  */
1468  for (ij = 0; ij < ppr->pr_ip6s; ij++)
1469  if (IN6_ARE_ADDR_EQUAL(&ip6[0],
1470  &ppr->pr_ip6[ij]))
1471  break;
1472  if (ij == ppr->pr_ip6s) {
1473  error = EPERM;
1474  goto done_deref_locked;
1475  }
1476  if (ip6s > 1) {
1477  for (ii = ij = 1; ii < ip6s; ii++) {
1478  if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
1479  &ppr->pr_ip6[0]))
1480  continue;
1481  for (; ij < ppr->pr_ip6s; ij++)
1482  if (IN6_ARE_ADDR_EQUAL(
1483  &ip6[ii], &ppr->pr_ip6[ij]))
1484  break;
1485  if (ij == ppr->pr_ip6s)
1486  break;
1487  }
1488  if (ij == ppr->pr_ip6s) {
1489  error = EPERM;
1490  goto done_deref_locked;
1491  }
1492  }
1493  }
1494  /* Check for conflicting IP addresses. */
1495  tppr = ppr;
1496 #ifdef VIMAGE
1497  for (; tppr != &prison0; tppr = tppr->pr_parent)
1498  if (tppr->pr_flags & PR_VNET)
1499  break;
1500 #endif
1501  FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
1502  if (tpr == pr ||
1503 #ifdef VIMAGE
1504  (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
1505 #endif
1506  tpr->pr_uref == 0) {
1507  descend = 0;
1508  continue;
1509  }
1510  if (!(tpr->pr_flags & PR_IP6_USER))
1511  continue;
1512  descend = 0;
1513  if (tpr->pr_ip6 == NULL ||
1514  (ip6s == 1 && tpr->pr_ip6s == 1))
1515  continue;
1516  for (ii = 0; ii < ip6s; ii++) {
1517  if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
1518  error = EADDRINUSE;
1519  vfs_opterror(opts,
1520  "IPv6 addresses clash");
1521  goto done_deref_locked;
1522  }
1523  }
1524  }
1525  }
1526 #endif
1527  onamelen = namelen = 0;
1528  if (name != NULL) {
1529  /* Give a default name of the jid. Also allow the name to be
1530  * explicitly the jid - but not any other number, and only in
1531  * normal form (no leading zero/etc).
1532  */
1533  if (name[0] == '\0')
1534  snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
1535  else if ((strtoul(namelc, &p, 10) != jid ||
1536  namelc[0] < '1' || namelc[0] > '9') && *p == '\0') {
1537  error = EINVAL;
1538  vfs_opterror(opts,
1539  "name cannot be numeric (unless it is the jid)");
1540  goto done_deref_locked;
1541  }
1542  /*
1543  * Make sure the name isn't too long for the prison or its
1544  * children.
1545  */
1546  onamelen = strlen(pr->pr_name);
1547  namelen = strlen(name);
1548  if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
1549  error = ENAMETOOLONG;
1550  goto done_deref_locked;
1551  }
1552  FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
1553  if (strlen(tpr->pr_name) + (namelen - onamelen) >=
1554  sizeof(pr->pr_name)) {
1555  error = ENAMETOOLONG;
1556  goto done_deref_locked;
1557  }
1558  }
1559  }
1560  if (pr_allow & ~ppr->pr_allow) {
1561  error = EPERM;
1562  goto done_deref_locked;
1563  }
1564 
1565  /* Set the parameters of the prison. */
1566 #ifdef INET
1567  redo_ip4 = 0;
1568  if (pr_flags & PR_IP4_USER) {
1569  pr->pr_flags |= PR_IP4;
1570  free(pr->pr_ip4, M_PRISON);
1571  pr->pr_ip4s = ip4s;
1572  pr->pr_ip4 = ip4;
1573  ip4 = NULL;
1574  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1575 #ifdef VIMAGE
1576  if (tpr->pr_flags & PR_VNET) {
1577  descend = 0;
1578  continue;
1579  }
1580 #endif
1581  if (prison_restrict_ip4(tpr, NULL)) {
1582  redo_ip4 = 1;
1583  descend = 0;
1584  }
1585  }
1586  }
1587 #endif
1588 #ifdef INET6
1589  redo_ip6 = 0;
1590  if (pr_flags & PR_IP6_USER) {
1591  pr->pr_flags |= PR_IP6;
1592  free(pr->pr_ip6, M_PRISON);
1593  pr->pr_ip6s = ip6s;
1594  pr->pr_ip6 = ip6;
1595  ip6 = NULL;
1596  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1597 #ifdef VIMAGE
1598  if (tpr->pr_flags & PR_VNET) {
1599  descend = 0;
1600  continue;
1601  }
1602 #endif
1603  if (prison_restrict_ip6(tpr, NULL)) {
1604  redo_ip6 = 1;
1605  descend = 0;
1606  }
1607  }
1608  }
1609 #endif
1610  if (gotslevel) {
1611  pr->pr_securelevel = slevel;
1612  /* Set all child jails to be at least this level. */
1613  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1614  if (tpr->pr_securelevel < slevel)
1615  tpr->pr_securelevel = slevel;
1616  }
1617  if (gotchildmax) {
1618  pr->pr_childmax = childmax;
1619  /* Set all child jails to under this limit. */
1620  FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
1621  if (tpr->pr_childmax > childmax - level)
1622  tpr->pr_childmax = childmax > level
1623  ? childmax - level : 0;
1624  }
1625  if (gotenforce) {
1626  pr->pr_enforce_statfs = enforce;
1627  /* Pass this restriction on to the children. */
1628  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1629  if (tpr->pr_enforce_statfs < enforce)
1630  tpr->pr_enforce_statfs = enforce;
1631  }
1632  if (gotrsnum) {
1633  pr->pr_devfs_rsnum = rsnum;
1634  /* Pass this restriction on to the children. */
1635  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1636  tpr->pr_devfs_rsnum = rsnum;
1637  }
1638  if (name != NULL) {
1639  if (ppr == &prison0)
1640  strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
1641  else
1642  snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
1643  ppr->pr_name, name);
1644  /* Change this component of child names. */
1645  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1646  bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
1647  strlen(tpr->pr_name + onamelen) + 1);
1648  bcopy(pr->pr_name, tpr->pr_name, namelen);
1649  }
1650  }
1651  if (path != NULL) {
1652  /* Try to keep a real-rooted full pathname. */
1653  if (fullpath_disabled && path[0] == '/' &&
1654  strcmp(mypr->pr_path, "/"))
1655  snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
1656  mypr->pr_path, path);
1657  else
1658  strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
1659  pr->pr_root = root;
1660  }
1661  if (PR_HOST & ch_flags & ~pr_flags) {
1662  if (pr->pr_flags & PR_HOST) {
1663  /*
1664  * Copy the parent's host info. As with pr_ip4 above,
1665  * the lack of a lock on the parent is not a problem;
1666  * it is always set with allprison_lock at least
1667  * shared, and is held exclusively here.
1668  */
1669  strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
1670  sizeof(pr->pr_hostname));
1671  strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
1672  sizeof(pr->pr_domainname));
1673  strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
1674  sizeof(pr->pr_hostuuid));
1675  pr->pr_hostid = pr->pr_parent->pr_hostid;
1676  }
1677  } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
1678  /* Set this prison, and any descendants without PR_HOST. */
1679  if (host != NULL)
1680  strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
1681  if (domain != NULL)
1682  strlcpy(pr->pr_domainname, domain,
1683  sizeof(pr->pr_domainname));
1684  if (uuid != NULL)
1685  strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
1686  if (gothid)
1687  pr->pr_hostid = hid;
1688  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1689  if (tpr->pr_flags & PR_HOST)
1690  descend = 0;
1691  else {
1692  if (host != NULL)
1693  strlcpy(tpr->pr_hostname,
1694  pr->pr_hostname,
1695  sizeof(tpr->pr_hostname));
1696  if (domain != NULL)
1697  strlcpy(tpr->pr_domainname,
1698  pr->pr_domainname,
1699  sizeof(tpr->pr_domainname));
1700  if (uuid != NULL)
1701  strlcpy(tpr->pr_hostuuid,
1702  pr->pr_hostuuid,
1703  sizeof(tpr->pr_hostuuid));
1704  if (gothid)
1705  tpr->pr_hostid = hid;
1706  }
1707  }
1708  }
1709  if ((tallow = ch_allow & ~pr_allow)) {
1710  /* Clear allow bits in all children. */
1711  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
1712  tpr->pr_allow &= ~tallow;
1713  }
1714  pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
1715  /*
1716  * Persistent prisons get an extra reference, and prisons losing their
1717  * persist flag lose that reference. Only do this for existing prisons
1718  * for now, so new ones will remain unseen until after the module
1719  * handlers have completed.
1720  */
1721  if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
1722  if (pr_flags & PR_PERSIST) {
1723  pr->pr_ref++;
1724  pr->pr_uref++;
1725  } else {
1726  pr->pr_ref--;
1727  pr->pr_uref--;
1728  }
1729  }
1730  pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
1731  mtx_unlock(&pr->pr_mtx);
1732 
1733 #ifdef RACCT
1734  if (created)
1735  prison_racct_attach(pr);
1736 #endif
1737 
1738  /* Locks may have prevented a complete restriction of child IP
1739  * addresses. If so, allocate some more memory and try again.
1740  */
1741 #ifdef INET
1742  while (redo_ip4) {
1743  ip4s = pr->pr_ip4s;
1744  ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
1745  mtx_lock(&pr->pr_mtx);
1746  redo_ip4 = 0;
1747  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1748 #ifdef VIMAGE
1749  if (tpr->pr_flags & PR_VNET) {
1750  descend = 0;
1751  continue;
1752  }
1753 #endif
1754  if (prison_restrict_ip4(tpr, ip4)) {
1755  if (ip4 != NULL)
1756  ip4 = NULL;
1757  else
1758  redo_ip4 = 1;
1759  }
1760  }
1761  mtx_unlock(&pr->pr_mtx);
1762  }
1763 #endif
1764 #ifdef INET6
1765  while (redo_ip6) {
1766  ip6s = pr->pr_ip6s;
1767  ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
1768  mtx_lock(&pr->pr_mtx);
1769  redo_ip6 = 0;
1770  FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
1771 #ifdef VIMAGE
1772  if (tpr->pr_flags & PR_VNET) {
1773  descend = 0;
1774  continue;
1775  }
1776 #endif
1777  if (prison_restrict_ip6(tpr, ip6)) {
1778  if (ip6 != NULL)
1779  ip6 = NULL;
1780  else
1781  redo_ip6 = 1;
1782  }
1783  }
1784  mtx_unlock(&pr->pr_mtx);
1785  }
1786 #endif
1787 
1788  /* Let the modules do their work. */
1789  sx_downgrade(&allprison_lock);
1790  if (created) {
1791  error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
1792  if (error) {
1793  prison_deref(pr, PD_LIST_SLOCKED);
1794  goto done_errmsg;
1795  }
1796  }
1797  error = osd_jail_call(pr, PR_METHOD_SET, opts);
1798  if (error) {
1799  prison_deref(pr, created
1800  ? PD_LIST_SLOCKED
1801  : PD_DEREF | PD_LIST_SLOCKED);
1802  goto done_errmsg;
1803  }
1804 
1805  /* Attach this process to the prison if requested. */
1806  if (flags & JAIL_ATTACH) {
1807  mtx_lock(&pr->pr_mtx);
1808  error = do_jail_attach(td, pr);
1809  if (error) {
1810  vfs_opterror(opts, "attach failed");
1811  if (!created)
1812  prison_deref(pr, PD_DEREF);
1813  goto done_errmsg;
1814  }
1815  }
1816 
1817 #ifdef RACCT
1818  if (!created) {
1819  sx_sunlock(&allprison_lock);
1820  prison_racct_modify(pr);
1821  sx_slock(&allprison_lock);
1822  }
1823 #endif
1824 
1825  td->td_retval[0] = pr->pr_id;
1826 
1827  /*
1828  * Now that it is all there, drop the temporary reference from existing
1829  * prisons. Or add a reference to newly created persistent prisons
1830  * (which was not done earlier so that the prison would not be publicly
1831  * visible).
1832  */
1833  if (!created) {
1834  prison_deref(pr, (flags & JAIL_ATTACH)
1835  ? PD_DEREF
1836  : PD_DEREF | PD_LIST_SLOCKED);
1837  } else {
1838  if (pr_flags & PR_PERSIST) {
1839  mtx_lock(&pr->pr_mtx);
1840  pr->pr_ref++;
1841  pr->pr_uref++;
1842  mtx_unlock(&pr->pr_mtx);
1843  }
1844  if (!(flags & JAIL_ATTACH))
1845  sx_sunlock(&allprison_lock);
1846  }
1847 
1848  goto done_errmsg;
1849 
1850  done_deref_locked:
1851  prison_deref(pr, created
1852  ? PD_LOCKED | PD_LIST_XLOCKED
1853  : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
1854  goto done_releroot;
1855  done_unlock_list:
1856  sx_xunlock(&allprison_lock);
1857  done_releroot:
1858  if (root != NULL) {
1859  vfslocked = VFS_LOCK_GIANT(root->v_mount);
1860  vrele(root);
1861  VFS_UNLOCK_GIANT(vfslocked);
1862  }
1863  done_errmsg:
1864  if (error) {
1865  vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
1866  if (errmsg_len > 0) {
1867  errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
1868  if (errmsg_pos > 0) {
1869  if (optuio->uio_segflg == UIO_SYSSPACE)
1870  bcopy(errmsg,
1871  optuio->uio_iov[errmsg_pos].iov_base,
1872  errmsg_len);
1873  else
1874  copyout(errmsg,
1875  optuio->uio_iov[errmsg_pos].iov_base,
1876  errmsg_len);
1877  }
1878  }
1879  }
1880  done_free:
1881 #ifdef INET
1882  free(ip4, M_PRISON);
1883 #endif
1884 #ifdef INET6
1885  free(ip6, M_PRISON);
1886 #endif
1887  if (g_path != NULL)
1888  free(g_path, M_TEMP);
1889  vfs_freeopts(opts);
1890  return (error);
1891 }
1892 
1893 
1894 /*
1895  * struct jail_get_args {
1896  * struct iovec *iovp;
1897  * unsigned int iovcnt;
1898  * int flags;
1899  * };
1900  */
1901 int
1902 sys_jail_get(struct thread *td, struct jail_get_args *uap)
1903 {
1904  struct uio *auio;
1905  int error;
1906 
1907  /* Check that we have an even number of iovecs. */
1908  if (uap->iovcnt & 1)
1909  return (EINVAL);
1910 
1911  error = copyinuio(uap->iovp, uap->iovcnt, &auio);
1912  if (error)
1913  return (error);
1914  error = kern_jail_get(td, auio, uap->flags);
1915  if (error == 0)
1916  error = copyout(auio->uio_iov, uap->iovp,
1917  uap->iovcnt * sizeof (struct iovec));
1918  free(auio, M_IOV);
1919  return (error);
1920 }
1921 
1922 int
1923 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
1924 {
1925  struct prison *pr, *mypr;
1926  struct vfsopt *opt;
1927  struct vfsoptlist *opts;
1928  char *errmsg, *name;
1929  int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
1930 
1931  if (flags & ~JAIL_GET_MASK)
1932  return (EINVAL);
1933 
1934  /* Get the parameter list. */
1935  error = vfs_buildopts(optuio, &opts);
1936  if (error)
1937  return (error);
1938  errmsg_pos = vfs_getopt_pos(opts, "errmsg");
1939  mypr = td->td_ucred->cr_prison;
1940 
1941  /*
1942  * Find the prison specified by one of: lastjid, jid, name.
1943  */
1944  sx_slock(&allprison_lock);
1945  error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
1946  if (error == 0) {
1947  TAILQ_FOREACH(pr, &allprison, pr_list) {
1948  if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
1949  mtx_lock(&pr->pr_mtx);
1950  if (pr->pr_ref > 0 &&
1951  (pr->pr_uref > 0 || (flags & JAIL_DYING)))
1952  break;
1953  mtx_unlock(&pr->pr_mtx);
1954  }
1955  }
1956  if (pr != NULL)
1957  goto found_prison;
1958  error = ENOENT;
1959  vfs_opterror(opts, "no jail after %d", jid);
1960  goto done_unlock_list;
1961  } else if (error != ENOENT)
1962  goto done_unlock_list;
1963 
1964  error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
1965  if (error == 0) {
1966  if (jid != 0) {
1967  pr = prison_find_child(mypr, jid);
1968  if (pr != NULL) {
1969  if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1970  mtx_unlock(&pr->pr_mtx);
1971  error = ENOENT;
1972  vfs_opterror(opts, "jail %d is dying",
1973  jid);
1974  goto done_unlock_list;
1975  }
1976  goto found_prison;
1977  }
1978  error = ENOENT;
1979  vfs_opterror(opts, "jail %d not found", jid);
1980  goto done_unlock_list;
1981  }
1982  } else if (error != ENOENT)
1983  goto done_unlock_list;
1984 
1985  error = vfs_getopt(opts, "name", (void **)&name, &len);
1986  if (error == 0) {
1987  if (len == 0 || name[len - 1] != '\0') {
1988  error = EINVAL;
1989  goto done_unlock_list;
1990  }
1991  pr = prison_find_name(mypr, name);
1992  if (pr != NULL) {
1993  if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
1994  mtx_unlock(&pr->pr_mtx);
1995  error = ENOENT;
1996  vfs_opterror(opts, "jail \"%s\" is dying",
1997  name);
1998  goto done_unlock_list;
1999  }
2000  goto found_prison;
2001  }
2002  error = ENOENT;
2003  vfs_opterror(opts, "jail \"%s\" not found", name);
2004  goto done_unlock_list;
2005  } else if (error != ENOENT)
2006  goto done_unlock_list;
2007 
2008  vfs_opterror(opts, "no jail specified");
2009  error = ENOENT;
2010  goto done_unlock_list;
2011 
2012  found_prison:
2013  /* Get the parameters of the prison. */
2014  pr->pr_ref++;
2015  locked = PD_LOCKED;
2016  td->td_retval[0] = pr->pr_id;
2017  error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
2018  if (error != 0 && error != ENOENT)
2019  goto done_deref;
2020  i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
2021  error = vfs_setopt(opts, "parent", &i, sizeof(i));
2022  if (error != 0 && error != ENOENT)
2023  goto done_deref;
2024  error = vfs_setopts(opts, "name", prison_name(mypr, pr));
2025  if (error != 0 && error != ENOENT)
2026  goto done_deref;
2027  error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
2028  sizeof(pr->pr_cpuset->cs_id));
2029  if (error != 0 && error != ENOENT)
2030  goto done_deref;
2031  error = vfs_setopts(opts, "path", prison_path(mypr, pr));
2032  if (error != 0 && error != ENOENT)
2033  goto done_deref;
2034 #ifdef INET
2035  error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
2036  pr->pr_ip4s * sizeof(*pr->pr_ip4));
2037  if (error != 0 && error != ENOENT)
2038  goto done_deref;
2039 #endif
2040 #ifdef INET6
2041  error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
2042  pr->pr_ip6s * sizeof(*pr->pr_ip6));
2043  if (error != 0 && error != ENOENT)
2044  goto done_deref;
2045 #endif
2046  error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
2047  sizeof(pr->pr_securelevel));
2048  if (error != 0 && error != ENOENT)
2049  goto done_deref;
2050  error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
2051  sizeof(pr->pr_childcount));
2052  if (error != 0 && error != ENOENT)
2053  goto done_deref;
2054  error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
2055  sizeof(pr->pr_childmax));
2056  if (error != 0 && error != ENOENT)
2057  goto done_deref;
2058  error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
2059  if (error != 0 && error != ENOENT)
2060  goto done_deref;
2061  error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
2062  if (error != 0 && error != ENOENT)
2063  goto done_deref;
2064  error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
2065  if (error != 0 && error != ENOENT)
2066  goto done_deref;
2067 #ifdef COMPAT_32BIT
2068  if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
2069  uint32_t hid32 = pr->pr_hostid;
2070 
2071  error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
2072  } else
2073 #endif
2074  error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
2075  sizeof(pr->pr_hostid));
2076  if (error != 0 && error != ENOENT)
2077  goto done_deref;
2078  error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
2079  sizeof(pr->pr_enforce_statfs));
2080  if (error != 0 && error != ENOENT)
2081  goto done_deref;
2082  error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum,
2083  sizeof(pr->pr_devfs_rsnum));
2084  if (error != 0 && error != ENOENT)
2085  goto done_deref;
2086  for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
2087  fi++) {
2088  if (pr_flag_names[fi] == NULL)
2089  continue;
2090  i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
2091  error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
2092  if (error != 0 && error != ENOENT)
2093  goto done_deref;
2094  i = !i;
2095  error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
2096  if (error != 0 && error != ENOENT)
2097  goto done_deref;
2098  }
2099  for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
2100  fi++) {
2101  i = pr->pr_flags &
2102  (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
2103  i = pr_flag_jailsys[fi].disable &&
2104  (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
2105  : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
2106  : JAIL_SYS_INHERIT;
2107  error =
2108  vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
2109  if (error != 0 && error != ENOENT)
2110  goto done_deref;
2111  }
2112  for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
2113  fi++) {
2114  if (pr_allow_names[fi] == NULL)
2115  continue;
2116  i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
2117  error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
2118  if (error != 0 && error != ENOENT)
2119  goto done_deref;
2120  i = !i;
2121  error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
2122  if (error != 0 && error != ENOENT)
2123  goto done_deref;
2124  }
2125  i = (pr->pr_uref == 0);
2126  error = vfs_setopt(opts, "dying", &i, sizeof(i));
2127  if (error != 0 && error != ENOENT)
2128  goto done_deref;
2129  i = !i;
2130  error = vfs_setopt(opts, "nodying", &i, sizeof(i));
2131  if (error != 0 && error != ENOENT)
2132  goto done_deref;
2133 
2134  /* Get the module parameters. */
2135  mtx_unlock(&pr->pr_mtx);
2136  locked = 0;
2137  error = osd_jail_call(pr, PR_METHOD_GET, opts);
2138  if (error)
2139  goto done_deref;
2140  prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
2141 
2142  /* By now, all parameters should have been noted. */
2143  TAILQ_FOREACH(opt, opts, link) {
2144  if (!opt->seen && strcmp(opt->name, "errmsg")) {
2145  error = EINVAL;
2146  vfs_opterror(opts, "unknown parameter: %s", opt->name);
2147  goto done_errmsg;
2148  }
2149  }
2150 
2151  /* Write the fetched parameters back to userspace. */
2152  error = 0;
2153  TAILQ_FOREACH(opt, opts, link) {
2154  if (opt->pos >= 0 && opt->pos != errmsg_pos) {
2155  pos = 2 * opt->pos + 1;
2156  optuio->uio_iov[pos].iov_len = opt->len;
2157  if (opt->value != NULL) {
2158  if (optuio->uio_segflg == UIO_SYSSPACE) {
2159  bcopy(opt->value,
2160  optuio->uio_iov[pos].iov_base,
2161  opt->len);
2162  } else {
2163  error = copyout(opt->value,
2164  optuio->uio_iov[pos].iov_base,
2165  opt->len);
2166  if (error)
2167  break;
2168  }
2169  }
2170  }
2171  }
2172  goto done_errmsg;
2173 
2174  done_deref:
2175  prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
2176  goto done_errmsg;
2177 
2178  done_unlock_list:
2179  sx_sunlock(&allprison_lock);
2180  done_errmsg:
2181  if (error && errmsg_pos >= 0) {
2182  vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
2183  errmsg_pos = 2 * errmsg_pos + 1;
2184  if (errmsg_len > 0) {
2185  if (optuio->uio_segflg == UIO_SYSSPACE)
2186  bcopy(errmsg,
2187  optuio->uio_iov[errmsg_pos].iov_base,
2188  errmsg_len);
2189  else
2190  copyout(errmsg,
2191  optuio->uio_iov[errmsg_pos].iov_base,
2192  errmsg_len);
2193  }
2194  }
2195  vfs_freeopts(opts);
2196  return (error);
2197 }
2198 
2199 
2200 /*
2201  * struct jail_remove_args {
2202  * int jid;
2203  * };
2204  */
2205 int
2206 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
2207 {
2208  struct prison *pr, *cpr, *lpr, *tpr;
2209  int descend, error;
2210 
2211  error = priv_check(td, PRIV_JAIL_REMOVE);
2212  if (error)
2213  return (error);
2214 
2215  sx_xlock(&allprison_lock);
2216  pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2217  if (pr == NULL) {
2218  sx_xunlock(&allprison_lock);
2219  return (EINVAL);
2220  }
2221 
2222  /* Remove all descendants of this prison, then remove this prison. */
2223  pr->pr_ref++;
2224  pr->pr_flags |= PR_REMOVE;
2225  if (!LIST_EMPTY(&pr->pr_children)) {
2226  mtx_unlock(&pr->pr_mtx);
2227  lpr = NULL;
2228  FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
2229  mtx_lock(&cpr->pr_mtx);
2230  if (cpr->pr_ref > 0) {
2231  tpr = cpr;
2232  cpr->pr_ref++;
2233  cpr->pr_flags |= PR_REMOVE;
2234  } else {
2235  /* Already removed - do not do it again. */
2236  tpr = NULL;
2237  }
2238  mtx_unlock(&cpr->pr_mtx);
2239  if (lpr != NULL) {
2240  mtx_lock(&lpr->pr_mtx);
2241  prison_remove_one(lpr);
2242  sx_xlock(&allprison_lock);
2243  }
2244  lpr = tpr;
2245  }
2246  if (lpr != NULL) {
2247  mtx_lock(&lpr->pr_mtx);
2248  prison_remove_one(lpr);
2249  sx_xlock(&allprison_lock);
2250  }
2251  mtx_lock(&pr->pr_mtx);
2252  }
2253  prison_remove_one(pr);
2254  return (0);
2255 }
2256 
2257 static void
2258 prison_remove_one(struct prison *pr)
2259 {
2260  struct proc *p;
2261  int deuref;
2262 
2263  /* If the prison was persistent, it is not anymore. */
2264  deuref = 0;
2265  if (pr->pr_flags & PR_PERSIST) {
2266  pr->pr_ref--;
2267  deuref = PD_DEUREF;
2268  pr->pr_flags &= ~PR_PERSIST;
2269  }
2270 
2271  /*
2272  * jail_remove added a reference. If that's the only one, remove
2273  * the prison now.
2274  */
2275  KASSERT(pr->pr_ref > 0,
2276  ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
2277  if (pr->pr_ref == 1) {
2278  prison_deref(pr,
2279  deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
2280  return;
2281  }
2282 
2283  mtx_unlock(&pr->pr_mtx);
2284  sx_xunlock(&allprison_lock);
2285  /*
2286  * Kill all processes unfortunate enough to be attached to this prison.
2287  */
2288  sx_slock(&allproc_lock);
2289  LIST_FOREACH(p, &allproc, p_list) {
2290  PROC_LOCK(p);
2291  if (p->p_state != PRS_NEW && p->p_ucred &&
2292  p->p_ucred->cr_prison == pr)
2293  kern_psignal(p, SIGKILL);
2294  PROC_UNLOCK(p);
2295  }
2296  sx_sunlock(&allproc_lock);
2297  /* Remove the temporary reference added by jail_remove. */
2298  prison_deref(pr, deuref | PD_DEREF);
2299 }
2300 
2301 
2302 /*
2303  * struct jail_attach_args {
2304  * int jid;
2305  * };
2306  */
2307 int
2308 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
2309 {
2310  struct prison *pr;
2311  int error;
2312 
2313  error = priv_check(td, PRIV_JAIL_ATTACH);
2314  if (error)
2315  return (error);
2316 
2317  sx_slock(&allprison_lock);
2318  pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
2319  if (pr == NULL) {
2320  sx_sunlock(&allprison_lock);
2321  return (EINVAL);
2322  }
2323 
2324  /*
2325  * Do not allow a process to attach to a prison that is not
2326  * considered to be "alive".
2327  */
2328  if (pr->pr_uref == 0) {
2329  mtx_unlock(&pr->pr_mtx);
2330  sx_sunlock(&allprison_lock);
2331  return (EINVAL);
2332  }
2333 
2334  return (do_jail_attach(td, pr));
2335 }
2336 
2337 static int
2338 do_jail_attach(struct thread *td, struct prison *pr)
2339 {
2340  struct prison *ppr;
2341  struct proc *p;
2342  struct ucred *newcred, *oldcred;
2343  int vfslocked, error;
2344 
2345  /*
2346  * XXX: Note that there is a slight race here if two threads
2347  * in the same privileged process attempt to attach to two
2348  * different jails at the same time. It is important for
2349  * user processes not to do this, or they might end up with
2350  * a process root from one prison, but attached to the jail
2351  * of another.
2352  */
2353  pr->pr_ref++;
2354  pr->pr_uref++;
2355  mtx_unlock(&pr->pr_mtx);
2356 
2357  /* Let modules do whatever they need to prepare for attaching. */
2358  error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
2359  if (error) {
2360  prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
2361  return (error);
2362  }
2363  sx_sunlock(&allprison_lock);
2364 
2365  /*
2366  * Reparent the newly attached process to this jail.
2367  */
2368  ppr = td->td_ucred->cr_prison;
2369  p = td->td_proc;
2370  error = cpuset_setproc_update_set(p, pr->pr_cpuset);
2371  if (error)
2372  goto e_revert_osd;
2373 
2374  vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2375  vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
2376  if ((error = change_dir(pr->pr_root, td)) != 0)
2377  goto e_unlock;
2378 #ifdef MAC
2379  if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
2380  goto e_unlock;
2381 #endif
2382  VOP_UNLOCK(pr->pr_root, 0);
2383  if ((error = change_root(pr->pr_root, td)))
2384  goto e_unlock_giant;
2385  VFS_UNLOCK_GIANT(vfslocked);
2386 
2387  newcred = crget();
2388  PROC_LOCK(p);
2389  oldcred = p->p_ucred;
2390  setsugid(p);
2391  crcopy(newcred, oldcred);
2392  newcred->cr_prison = pr;
2393  p->p_ucred = newcred;
2394  PROC_UNLOCK(p);
2395 #ifdef RACCT
2396  racct_proc_ucred_changed(p, oldcred, newcred);
2397 #endif
2398  crfree(oldcred);
2399  prison_deref(ppr, PD_DEREF | PD_DEUREF);
2400  return (0);
2401  e_unlock:
2402  VOP_UNLOCK(pr->pr_root, 0);
2403  e_unlock_giant:
2404  VFS_UNLOCK_GIANT(vfslocked);
2405  e_revert_osd:
2406  /* Tell modules this thread is still in its old jail after all. */
2407  (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
2408  prison_deref(pr, PD_DEREF | PD_DEUREF);
2409  return (error);
2410 }
2411 
2412 
2413 /*
2414  * Returns a locked prison instance, or NULL on failure.
2415  */
2416 struct prison *
2417 prison_find(int prid)
2418 {
2419  struct prison *pr;
2420 
2421  sx_assert(&allprison_lock, SX_LOCKED);
2422  TAILQ_FOREACH(pr, &allprison, pr_list) {
2423  if (pr->pr_id == prid) {
2424  mtx_lock(&pr->pr_mtx);
2425  if (pr->pr_ref > 0)
2426  return (pr);
2427  mtx_unlock(&pr->pr_mtx);
2428  }
2429  }
2430  return (NULL);
2431 }
2432 
2433 /*
2434  * Find a prison that is a descendant of mypr. Returns a locked prison or NULL.
2435  */
2436 struct prison *
2437 prison_find_child(struct prison *mypr, int prid)
2438 {
2439  struct prison *pr;
2440  int descend;
2441 
2442  sx_assert(&allprison_lock, SX_LOCKED);
2443  FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2444  if (pr->pr_id == prid) {
2445  mtx_lock(&pr->pr_mtx);
2446  if (pr->pr_ref > 0)
2447  return (pr);
2448  mtx_unlock(&pr->pr_mtx);
2449  }
2450  }
2451  return (NULL);
2452 }
2453 
2454 /*
2455  * Look for the name relative to mypr. Returns a locked prison or NULL.
2456  */
2457 struct prison *
2458 prison_find_name(struct prison *mypr, const char *name)
2459 {
2460  struct prison *pr, *deadpr;
2461  size_t mylen;
2462  int descend;
2463 
2464  sx_assert(&allprison_lock, SX_LOCKED);
2465  mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
2466  again:
2467  deadpr = NULL;
2468  FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
2469  if (!strcmp(pr->pr_name + mylen, name)) {
2470  mtx_lock(&pr->pr_mtx);
2471  if (pr->pr_ref > 0) {
2472  if (pr->pr_uref > 0)
2473  return (pr);
2474  deadpr = pr;
2475  }
2476  mtx_unlock(&pr->pr_mtx);
2477  }
2478  }
2479  /* There was no valid prison - perhaps there was a dying one. */
2480  if (deadpr != NULL) {
2481  mtx_lock(&deadpr->pr_mtx);
2482  if (deadpr->pr_ref == 0) {
2483  mtx_unlock(&deadpr->pr_mtx);
2484  goto again;
2485  }
2486  }
2487  return (deadpr);
2488 }
2489 
2490 /*
2491  * See if a prison has the specific flag set.
2492  */
2493 int
2494 prison_flag(struct ucred *cred, unsigned flag)
2495 {
2496 
2497  /* This is an atomic read, so no locking is necessary. */
2498  return (cred->cr_prison->pr_flags & flag);
2499 }
2500 
2501 int
2502 prison_allow(struct ucred *cred, unsigned flag)
2503 {
2504 
2505  /* This is an atomic read, so no locking is necessary. */
2506  return (cred->cr_prison->pr_allow & flag);
2507 }
2508 
2509 /*
2510  * Remove a prison reference. If that was the last reference, remove the
2511  * prison itself - but not in this context in case there are locks held.
2512  */
2513 void
2514 prison_free_locked(struct prison *pr)
2515 {
2516 
2517  mtx_assert(&pr->pr_mtx, MA_OWNED);
2518  pr->pr_ref--;
2519  if (pr->pr_ref == 0) {
2520  mtx_unlock(&pr->pr_mtx);
2521  TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
2522  taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
2523  return;
2524  }
2525  mtx_unlock(&pr->pr_mtx);
2526 }
2527 
2528 void
2529 prison_free(struct prison *pr)
2530 {
2531 
2532  mtx_lock(&pr->pr_mtx);
2533  prison_free_locked(pr);
2534 }
2535 
2536 static void
2537 prison_complete(void *context, int pending)
2538 {
2539 
2540  prison_deref((struct prison *)context, 0);
2541 }
2542 
2543 /*
2544  * Remove a prison reference (usually). This internal version assumes no
2545  * mutexes are held, except perhaps the prison itself. If there are no more
2546  * references, release and delist the prison. On completion, the prison lock
2547  * and the allprison lock are both unlocked.
2548  */
2549 static void
2550 prison_deref(struct prison *pr, int flags)
2551 {
2552  struct prison *ppr, *tpr;
2553  int vfslocked;
2554 
2555  if (!(flags & PD_LOCKED))
2556  mtx_lock(&pr->pr_mtx);
2557  for (;;) {
2558  if (flags & PD_DEUREF) {
2559  pr->pr_uref--;
2560  KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
2561  }
2562  if (flags & PD_DEREF)
2563  pr->pr_ref--;
2564  /* If the prison still has references, nothing else to do. */
2565  if (pr->pr_ref > 0) {
2566  mtx_unlock(&pr->pr_mtx);
2567  if (flags & PD_LIST_SLOCKED)
2568  sx_sunlock(&allprison_lock);
2569  else if (flags & PD_LIST_XLOCKED)
2570  sx_xunlock(&allprison_lock);
2571  return;
2572  }
2573 
2574  mtx_unlock(&pr->pr_mtx);
2575  if (flags & PD_LIST_SLOCKED) {
2576  if (!sx_try_upgrade(&allprison_lock)) {
2577  sx_sunlock(&allprison_lock);
2578  sx_xlock(&allprison_lock);
2579  }
2580  } else if (!(flags & PD_LIST_XLOCKED))
2581  sx_xlock(&allprison_lock);
2582 
2583  TAILQ_REMOVE(&allprison, pr, pr_list);
2584  LIST_REMOVE(pr, pr_sibling);
2585  ppr = pr->pr_parent;
2586  for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
2587  tpr->pr_childcount--;
2588  sx_xunlock(&allprison_lock);
2589 
2590 #ifdef VIMAGE
2591  if (pr->pr_vnet != ppr->pr_vnet)
2592  vnet_destroy(pr->pr_vnet);
2593 #endif
2594  if (pr->pr_root != NULL) {
2595  vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
2596  vrele(pr->pr_root);
2597  VFS_UNLOCK_GIANT(vfslocked);
2598  }
2599  mtx_destroy(&pr->pr_mtx);
2600 #ifdef INET
2601  free(pr->pr_ip4, M_PRISON);
2602 #endif
2603 #ifdef INET6
2604  free(pr->pr_ip6, M_PRISON);
2605 #endif
2606  if (pr->pr_cpuset != NULL)
2607  cpuset_rel(pr->pr_cpuset);
2608  osd_jail_exit(pr);
2609 #ifdef RACCT
2610  prison_racct_detach(pr);
2611 #endif
2612  free(pr, M_PRISON);
2613 
2614  /* Removing a prison frees a reference on its parent. */
2615  pr = ppr;
2616  mtx_lock(&pr->pr_mtx);
2617  flags = PD_DEREF | PD_DEUREF;
2618  }
2619 }
2620 
2621 void
2622 prison_hold_locked(struct prison *pr)
2623 {
2624 
2625  mtx_assert(&pr->pr_mtx, MA_OWNED);
2626  KASSERT(pr->pr_ref > 0,
2627  ("Trying to hold dead prison (jid=%d).", pr->pr_id));
2628  pr->pr_ref++;
2629 }
2630 
2631 void
2632 prison_hold(struct prison *pr)
2633 {
2634 
2635  mtx_lock(&pr->pr_mtx);
2636  prison_hold_locked(pr);
2637  mtx_unlock(&pr->pr_mtx);
2638 }
2639 
2640 void
2641 prison_proc_hold(struct prison *pr)
2642 {
2643 
2644  mtx_lock(&pr->pr_mtx);
2645  KASSERT(pr->pr_uref > 0,
2646  ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
2647  pr->pr_uref++;
2648  mtx_unlock(&pr->pr_mtx);
2649 }
2650 
2651 void
2652 prison_proc_free(struct prison *pr)
2653 {
2654 
2655  mtx_lock(&pr->pr_mtx);
2656  KASSERT(pr->pr_uref > 0,
2657  ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
2658  prison_deref(pr, PD_DEUREF | PD_LOCKED);
2659 }
2660 
2661 
2662 #ifdef INET
2663 /*
2664  * Restrict a prison's IP address list with its parent's, possibly replacing
2665  * it. Return true if the replacement buffer was used (or would have been).
2666  */
2667 static int
2668 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
2669 {
2670  int ii, ij, used;
2671  struct prison *ppr;
2672 
2673  ppr = pr->pr_parent;
2674  if (!(pr->pr_flags & PR_IP4_USER)) {
2675  /* This has no user settings, so just copy the parent's list. */
2676  if (pr->pr_ip4s < ppr->pr_ip4s) {
2677  /*
2678  * There's no room for the parent's list. Use the
2679  * new list buffer, which is assumed to be big enough
2680  * (if it was passed). If there's no buffer, try to
2681  * allocate one.
2682  */
2683  used = 1;
2684  if (newip4 == NULL) {
2685  newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
2686  M_PRISON, M_NOWAIT);
2687  if (newip4 != NULL)
2688  used = 0;
2689  }
2690  if (newip4 != NULL) {
2691  bcopy(ppr->pr_ip4, newip4,
2692  ppr->pr_ip4s * sizeof(*newip4));
2693  free(pr->pr_ip4, M_PRISON);
2694  pr->pr_ip4 = newip4;
2695  pr->pr_ip4s = ppr->pr_ip4s;
2696  }
2697  return (used);
2698  }
2699  pr->pr_ip4s = ppr->pr_ip4s;
2700  if (pr->pr_ip4s > 0)
2701  bcopy(ppr->pr_ip4, pr->pr_ip4,
2702  pr->pr_ip4s * sizeof(*newip4));
2703  else if (pr->pr_ip4 != NULL) {
2704  free(pr->pr_ip4, M_PRISON);
2705  pr->pr_ip4 = NULL;
2706  }
2707  } else if (pr->pr_ip4s > 0) {
2708  /* Remove addresses that aren't in the parent. */
2709  for (ij = 0; ij < ppr->pr_ip4s; ij++)
2710  if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
2711  break;
2712  if (ij < ppr->pr_ip4s)
2713  ii = 1;
2714  else {
2715  bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
2716  --pr->pr_ip4s * sizeof(*pr->pr_ip4));
2717  ii = 0;
2718  }
2719  for (ij = 1; ii < pr->pr_ip4s; ) {
2720  if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
2721  ii++;
2722  continue;
2723  }
2724  switch (ij >= ppr->pr_ip4s ? -1 :
2725  qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
2726  case -1:
2727  bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
2728  (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
2729  break;
2730  case 0:
2731  ii++;
2732  ij++;
2733  break;
2734  case 1:
2735  ij++;
2736  break;
2737  }
2738  }
2739  if (pr->pr_ip4s == 0) {
2740  pr->pr_flags |= PR_IP4_DISABLE;
2741  free(pr->pr_ip4, M_PRISON);
2742  pr->pr_ip4 = NULL;
2743  }
2744  }
2745  return (0);
2746 }
2747 
2748 /*
2749  * Pass back primary IPv4 address of this jail.
2750  *
2751  * If not restricted return success but do not alter the address. Caller has
2752  * to make sure to initialize it correctly (e.g. INADDR_ANY).
2753  *
2754  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2755  * Address returned in NBO.
2756  */
2757 int
2758 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
2759 {
2760  struct prison *pr;
2761 
2762  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2763  KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2764 
2765  pr = cred->cr_prison;
2766  if (!(pr->pr_flags & PR_IP4))
2767  return (0);
2768  mtx_lock(&pr->pr_mtx);
2769  if (!(pr->pr_flags & PR_IP4)) {
2770  mtx_unlock(&pr->pr_mtx);
2771  return (0);
2772  }
2773  if (pr->pr_ip4 == NULL) {
2774  mtx_unlock(&pr->pr_mtx);
2775  return (EAFNOSUPPORT);
2776  }
2777 
2778  ia->s_addr = pr->pr_ip4[0].s_addr;
2779  mtx_unlock(&pr->pr_mtx);
2780  return (0);
2781 }
2782 
2783 /*
2784  * Return 1 if we should do proper source address selection or are not jailed.
2785  * We will return 0 if we should bypass source address selection in favour
2786  * of the primary jail IPv4 address. Only in this case *ia will be updated and
2787  * returned in NBO.
2788  * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
2789  */
2790 int
2791 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
2792 {
2793  struct prison *pr;
2794  struct in_addr lia;
2795  int error;
2796 
2797  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2798  KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2799 
2800  if (!jailed(cred))
2801  return (1);
2802 
2803  pr = cred->cr_prison;
2804  if (pr->pr_flags & PR_IP4_SADDRSEL)
2805  return (1);
2806 
2807  lia.s_addr = INADDR_ANY;
2808  error = prison_get_ip4(cred, &lia);
2809  if (error)
2810  return (error);
2811  if (lia.s_addr == INADDR_ANY)
2812  return (1);
2813 
2814  ia->s_addr = lia.s_addr;
2815  return (0);
2816 }
2817 
2818 /*
2819  * Return true if pr1 and pr2 have the same IPv4 address restrictions.
2820  */
2821 int
2822 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
2823 {
2824 
2825  if (pr1 == pr2)
2826  return (1);
2827 
2828  /*
2829  * No need to lock since the PR_IP4_USER flag can't be altered for
2830  * existing prisons.
2831  */
2832  while (pr1 != &prison0 &&
2833 #ifdef VIMAGE
2834  !(pr1->pr_flags & PR_VNET) &&
2835 #endif
2836  !(pr1->pr_flags & PR_IP4_USER))
2837  pr1 = pr1->pr_parent;
2838  while (pr2 != &prison0 &&
2839 #ifdef VIMAGE
2840  !(pr2->pr_flags & PR_VNET) &&
2841 #endif
2842  !(pr2->pr_flags & PR_IP4_USER))
2843  pr2 = pr2->pr_parent;
2844  return (pr1 == pr2);
2845 }
2846 
2847 /*
2848  * Make sure our (source) address is set to something meaningful to this
2849  * jail.
2850  *
2851  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2852  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2853  * doesn't allow IPv4. Address passed in in NBO and returned in NBO.
2854  */
2855 int
2856 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
2857 {
2858  struct prison *pr;
2859  struct in_addr ia0;
2860  int error;
2861 
2862  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2863  KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2864 
2865  pr = cred->cr_prison;
2866  if (!(pr->pr_flags & PR_IP4))
2867  return (0);
2868  mtx_lock(&pr->pr_mtx);
2869  if (!(pr->pr_flags & PR_IP4)) {
2870  mtx_unlock(&pr->pr_mtx);
2871  return (0);
2872  }
2873  if (pr->pr_ip4 == NULL) {
2874  mtx_unlock(&pr->pr_mtx);
2875  return (EAFNOSUPPORT);
2876  }
2877 
2878  ia0.s_addr = ntohl(ia->s_addr);
2879  if (ia0.s_addr == INADDR_LOOPBACK) {
2880  ia->s_addr = pr->pr_ip4[0].s_addr;
2881  mtx_unlock(&pr->pr_mtx);
2882  return (0);
2883  }
2884 
2885  if (ia0.s_addr == INADDR_ANY) {
2886  /*
2887  * In case there is only 1 IPv4 address, bind directly.
2888  */
2889  if (pr->pr_ip4s == 1)
2890  ia->s_addr = pr->pr_ip4[0].s_addr;
2891  mtx_unlock(&pr->pr_mtx);
2892  return (0);
2893  }
2894 
2895  error = _prison_check_ip4(pr, ia);
2896  mtx_unlock(&pr->pr_mtx);
2897  return (error);
2898 }
2899 
2900 /*
2901  * Rewrite destination address in case we will connect to loopback address.
2902  *
2903  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
2904  * Address passed in in NBO and returned in NBO.
2905  */
2906 int
2907 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
2908 {
2909  struct prison *pr;
2910 
2911  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2912  KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2913 
2914  pr = cred->cr_prison;
2915  if (!(pr->pr_flags & PR_IP4))
2916  return (0);
2917  mtx_lock(&pr->pr_mtx);
2918  if (!(pr->pr_flags & PR_IP4)) {
2919  mtx_unlock(&pr->pr_mtx);
2920  return (0);
2921  }
2922  if (pr->pr_ip4 == NULL) {
2923  mtx_unlock(&pr->pr_mtx);
2924  return (EAFNOSUPPORT);
2925  }
2926 
2927  if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
2928  ia->s_addr = pr->pr_ip4[0].s_addr;
2929  mtx_unlock(&pr->pr_mtx);
2930  return (0);
2931  }
2932 
2933  /*
2934  * Return success because nothing had to be changed.
2935  */
2936  mtx_unlock(&pr->pr_mtx);
2937  return (0);
2938 }
2939 
2940 /*
2941  * Check if given address belongs to the jail referenced by cred/prison.
2942  *
2943  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
2944  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
2945  * doesn't allow IPv4. Address passed in in NBO.
2946  */
2947 static int
2948 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
2949 {
2950  int i, a, z, d;
2951 
2952  /*
2953  * Check the primary IP.
2954  */
2955  if (pr->pr_ip4[0].s_addr == ia->s_addr)
2956  return (0);
2957 
2958  /*
2959  * All the other IPs are sorted so we can do a binary search.
2960  */
2961  a = 0;
2962  z = pr->pr_ip4s - 2;
2963  while (a <= z) {
2964  i = (a + z) / 2;
2965  d = qcmp_v4(&pr->pr_ip4[i+1], ia);
2966  if (d > 0)
2967  z = i - 1;
2968  else if (d < 0)
2969  a = i + 1;
2970  else
2971  return (0);
2972  }
2973 
2974  return (EADDRNOTAVAIL);
2975 }
2976 
2977 int
2978 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
2979 {
2980  struct prison *pr;
2981  int error;
2982 
2983  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
2984  KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
2985 
2986  pr = cred->cr_prison;
2987  if (!(pr->pr_flags & PR_IP4))
2988  return (0);
2989  mtx_lock(&pr->pr_mtx);
2990  if (!(pr->pr_flags & PR_IP4)) {
2991  mtx_unlock(&pr->pr_mtx);
2992  return (0);
2993  }
2994  if (pr->pr_ip4 == NULL) {
2995  mtx_unlock(&pr->pr_mtx);
2996  return (EAFNOSUPPORT);
2997  }
2998 
2999  error = _prison_check_ip4(pr, ia);
3000  mtx_unlock(&pr->pr_mtx);
3001  return (error);
3002 }
3003 #endif
3004 
3005 #ifdef INET6
3006 static int
3007 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
3008 {
3009  int ii, ij, used;
3010  struct prison *ppr;
3011 
3012  ppr = pr->pr_parent;
3013  if (!(pr->pr_flags & PR_IP6_USER)) {
3014  /* This has no user settings, so just copy the parent's list. */
3015  if (pr->pr_ip6s < ppr->pr_ip6s) {
3016  /*
3017  * There's no room for the parent's list. Use the
3018  * new list buffer, which is assumed to be big enough
3019  * (if it was passed). If there's no buffer, try to
3020  * allocate one.
3021  */
3022  used = 1;
3023  if (newip6 == NULL) {
3024  newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
3025  M_PRISON, M_NOWAIT);
3026  if (newip6 != NULL)
3027  used = 0;
3028  }
3029  if (newip6 != NULL) {
3030  bcopy(ppr->pr_ip6, newip6,
3031  ppr->pr_ip6s * sizeof(*newip6));
3032  free(pr->pr_ip6, M_PRISON);
3033  pr->pr_ip6 = newip6;
3034  pr->pr_ip6s = ppr->pr_ip6s;
3035  }
3036  return (used);
3037  }
3038  pr->pr_ip6s = ppr->pr_ip6s;
3039  if (pr->pr_ip6s > 0)
3040  bcopy(ppr->pr_ip6, pr->pr_ip6,
3041  pr->pr_ip6s * sizeof(*newip6));
3042  else if (pr->pr_ip6 != NULL) {
3043  free(pr->pr_ip6, M_PRISON);
3044  pr->pr_ip6 = NULL;
3045  }
3046  } else if (pr->pr_ip6s > 0) {
3047  /* Remove addresses that aren't in the parent. */
3048  for (ij = 0; ij < ppr->pr_ip6s; ij++)
3049  if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
3050  &ppr->pr_ip6[ij]))
3051  break;
3052  if (ij < ppr->pr_ip6s)
3053  ii = 1;
3054  else {
3055  bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
3056  --pr->pr_ip6s * sizeof(*pr->pr_ip6));
3057  ii = 0;
3058  }
3059  for (ij = 1; ii < pr->pr_ip6s; ) {
3060  if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
3061  &ppr->pr_ip6[0])) {
3062  ii++;
3063  continue;
3064  }
3065  switch (ij >= ppr->pr_ip6s ? -1 :
3066  qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
3067  case -1:
3068  bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
3069  (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
3070  break;
3071  case 0:
3072  ii++;
3073  ij++;
3074  break;
3075  case 1:
3076  ij++;
3077  break;
3078  }
3079  }
3080  if (pr->pr_ip6s == 0) {
3081  pr->pr_flags |= PR_IP6_DISABLE;
3082  free(pr->pr_ip6, M_PRISON);
3083  pr->pr_ip6 = NULL;
3084  }
3085  }
3086  return 0;
3087 }
3088 
3089 /*
3090  * Pass back primary IPv6 address for this jail.
3091  *
3092  * If not restricted return success but do not alter the address. Caller has
3093  * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
3094  *
3095  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3096  */
3097 int
3098 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
3099 {
3100  struct prison *pr;
3101 
3102  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3103  KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3104 
3105  pr = cred->cr_prison;
3106  if (!(pr->pr_flags & PR_IP6))
3107  return (0);
3108  mtx_lock(&pr->pr_mtx);
3109  if (!(pr->pr_flags & PR_IP6)) {
3110  mtx_unlock(&pr->pr_mtx);
3111  return (0);
3112  }
3113  if (pr->pr_ip6 == NULL) {
3114  mtx_unlock(&pr->pr_mtx);
3115  return (EAFNOSUPPORT);
3116  }
3117 
3118  bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3119  mtx_unlock(&pr->pr_mtx);
3120  return (0);
3121 }
3122 
3123 /*
3124  * Return 1 if we should do proper source address selection or are not jailed.
3125  * We will return 0 if we should bypass source address selection in favour
3126  * of the primary jail IPv6 address. Only in this case *ia will be updated and
3127  * returned in NBO.
3128  * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
3129  */
3130 int
3131 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
3132 {
3133  struct prison *pr;
3134  struct in6_addr lia6;
3135  int error;
3136 
3137  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3138  KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3139 
3140  if (!jailed(cred))
3141  return (1);
3142 
3143  pr = cred->cr_prison;
3144  if (pr->pr_flags & PR_IP6_SADDRSEL)
3145  return (1);
3146 
3147  lia6 = in6addr_any;
3148  error = prison_get_ip6(cred, &lia6);
3149  if (error)
3150  return (error);
3151  if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
3152  return (1);
3153 
3154  bcopy(&lia6, ia6, sizeof(struct in6_addr));
3155  return (0);
3156 }
3157 
3158 /*
3159  * Return true if pr1 and pr2 have the same IPv6 address restrictions.
3160  */
3161 int
3162 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
3163 {
3164 
3165  if (pr1 == pr2)
3166  return (1);
3167 
3168  while (pr1 != &prison0 &&
3169 #ifdef VIMAGE
3170  !(pr1->pr_flags & PR_VNET) &&
3171 #endif
3172  !(pr1->pr_flags & PR_IP6_USER))
3173  pr1 = pr1->pr_parent;
3174  while (pr2 != &prison0 &&
3175 #ifdef VIMAGE
3176  !(pr2->pr_flags & PR_VNET) &&
3177 #endif
3178  !(pr2->pr_flags & PR_IP6_USER))
3179  pr2 = pr2->pr_parent;
3180  return (pr1 == pr2);
3181 }
3182 
3183 /*
3184  * Make sure our (source) address is set to something meaningful to this jail.
3185  *
3186  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
3187  * when needed while binding.
3188  *
3189  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3190  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3191  * doesn't allow IPv6.
3192  */
3193 int
3194 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
3195 {
3196  struct prison *pr;
3197  int error;
3198 
3199  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3200  KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3201 
3202  pr = cred->cr_prison;
3203  if (!(pr->pr_flags & PR_IP6))
3204  return (0);
3205  mtx_lock(&pr->pr_mtx);
3206  if (!(pr->pr_flags & PR_IP6)) {
3207  mtx_unlock(&pr->pr_mtx);
3208  return (0);
3209  }
3210  if (pr->pr_ip6 == NULL) {
3211  mtx_unlock(&pr->pr_mtx);
3212  return (EAFNOSUPPORT);
3213  }
3214 
3215  if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3216  bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3217  mtx_unlock(&pr->pr_mtx);
3218  return (0);
3219  }
3220 
3221  if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
3222  /*
3223  * In case there is only 1 IPv6 address, and v6only is true,
3224  * then bind directly.
3225  */
3226  if (v6only != 0 && pr->pr_ip6s == 1)
3227  bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3228  mtx_unlock(&pr->pr_mtx);
3229  return (0);
3230  }
3231 
3232  error = _prison_check_ip6(pr, ia6);
3233  mtx_unlock(&pr->pr_mtx);
3234  return (error);
3235 }
3236 
3237 /*
3238  * Rewrite destination address in case we will connect to loopback address.
3239  *
3240  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
3241  */
3242 int
3243 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
3244 {
3245  struct prison *pr;
3246 
3247  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3248  KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3249 
3250  pr = cred->cr_prison;
3251  if (!(pr->pr_flags & PR_IP6))
3252  return (0);
3253  mtx_lock(&pr->pr_mtx);
3254  if (!(pr->pr_flags & PR_IP6)) {
3255  mtx_unlock(&pr->pr_mtx);
3256  return (0);
3257  }
3258  if (pr->pr_ip6 == NULL) {
3259  mtx_unlock(&pr->pr_mtx);
3260  return (EAFNOSUPPORT);
3261  }
3262 
3263  if (IN6_IS_ADDR_LOOPBACK(ia6)) {
3264  bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
3265  mtx_unlock(&pr->pr_mtx);
3266  return (0);
3267  }
3268 
3269  /*
3270  * Return success because nothing had to be changed.
3271  */
3272  mtx_unlock(&pr->pr_mtx);
3273  return (0);
3274 }
3275 
3276 /*
3277  * Check if given address belongs to the jail referenced by cred/prison.
3278  *
3279  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
3280  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
3281  * doesn't allow IPv6.
3282  */
3283 static int
3284 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
3285 {
3286  int i, a, z, d;
3287 
3288  /*
3289  * Check the primary IP.
3290  */
3291  if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
3292  return (0);
3293 
3294  /*
3295  * All the other IPs are sorted so we can do a binary search.
3296  */
3297  a = 0;
3298  z = pr->pr_ip6s - 2;
3299  while (a <= z) {
3300  i = (a + z) / 2;
3301  d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
3302  if (d > 0)
3303  z = i - 1;
3304  else if (d < 0)
3305  a = i + 1;
3306  else
3307  return (0);
3308  }
3309 
3310  return (EADDRNOTAVAIL);
3311 }
3312 
3313 int
3314 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
3315 {
3316  struct prison *pr;
3317  int error;
3318 
3319  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3320  KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
3321 
3322  pr = cred->cr_prison;
3323  if (!(pr->pr_flags & PR_IP6))
3324  return (0);
3325  mtx_lock(&pr->pr_mtx);
3326  if (!(pr->pr_flags & PR_IP6)) {
3327  mtx_unlock(&pr->pr_mtx);
3328  return (0);
3329  }
3330  if (pr->pr_ip6 == NULL) {
3331  mtx_unlock(&pr->pr_mtx);
3332  return (EAFNOSUPPORT);
3333  }
3334 
3335  error = _prison_check_ip6(pr, ia6);
3336  mtx_unlock(&pr->pr_mtx);
3337  return (error);
3338 }
3339 #endif
3340 
3341 /*
3342  * Check if a jail supports the given address family.
3343  *
3344  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
3345  * if not.
3346  */
3347 int
3348 prison_check_af(struct ucred *cred, int af)
3349 {
3350  struct prison *pr;
3351  int error;
3352 
3353  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3354 
3355  pr = cred->cr_prison;
3356 #ifdef VIMAGE
3357  /* Prisons with their own network stack are not limited. */
3358  if (prison_owns_vnet(cred))
3359  return (0);
3360 #endif
3361 
3362  error = 0;
3363  switch (af)
3364  {
3365 #ifdef INET
3366  case AF_INET:
3367  if (pr->pr_flags & PR_IP4)
3368  {
3369  mtx_lock(&pr->pr_mtx);
3370  if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
3371  error = EAFNOSUPPORT;
3372  mtx_unlock(&pr->pr_mtx);
3373  }
3374  break;
3375 #endif
3376 #ifdef INET6
3377  case AF_INET6:
3378  if (pr->pr_flags & PR_IP6)
3379  {
3380  mtx_lock(&pr->pr_mtx);
3381  if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
3382  error = EAFNOSUPPORT;
3383  mtx_unlock(&pr->pr_mtx);
3384  }
3385  break;
3386 #endif
3387  case AF_LOCAL:
3388  case AF_ROUTE:
3389  break;
3390  default:
3391  if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
3392  error = EAFNOSUPPORT;
3393  }
3394  return (error);
3395 }
3396 
3397 /*
3398  * Check if given address belongs to the jail referenced by cred (wrapper to
3399  * prison_check_ip[46]).
3400  *
3401  * Returns 0 if jail doesn't restrict the address family or if address belongs
3402  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
3403  * the jail doesn't allow the address family. IPv4 Address passed in in NBO.
3404  */
3405 int
3406 prison_if(struct ucred *cred, struct sockaddr *sa)
3407 {
3408 #ifdef INET
3409  struct sockaddr_in *sai;
3410 #endif
3411 #ifdef INET6
3412  struct sockaddr_in6 *sai6;
3413 #endif
3414  int error;
3415 
3416  KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
3417  KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
3418 
3419 #ifdef VIMAGE
3420  if (prison_owns_vnet(cred))
3421  return (0);
3422 #endif
3423 
3424  error = 0;
3425  switch (sa->sa_family)
3426  {
3427 #ifdef INET
3428  case AF_INET:
3429  sai = (struct sockaddr_in *)sa;
3430  error = prison_check_ip4(cred, &sai->sin_addr);
3431  break;
3432 #endif
3433 #ifdef INET6
3434  case AF_INET6:
3435  sai6 = (struct sockaddr_in6 *)sa;
3436  error = prison_check_ip6(cred, &sai6->sin6_addr);
3437  break;
3438 #endif
3439  default:
3440  if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
3441  error = EAFNOSUPPORT;
3442  }
3443  return (error);
3444 }
3445 
3446 /*
3447  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
3448  */
3449 int
3450 prison_check(struct ucred *cred1, struct ucred *cred2)
3451 {
3452 
3453  return ((cred1->cr_prison == cred2->cr_prison ||
3454  prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
3455 }
3456 
3457 /*
3458  * Return 1 if p2 is a child of p1, otherwise 0.
3459  */
3460 int
3461 prison_ischild(struct prison *pr1, struct prison *pr2)
3462 {
3463 
3464  for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
3465  if (pr1 == pr2)
3466  return (1);
3467  return (0);
3468 }
3469 
3470 /*
3471  * Return 1 if the passed credential is in a jail, otherwise 0.
3472  */
3473 int
3474 jailed(struct ucred *cred)
3475 {
3476 
3477  return (cred->cr_prison != &prison0);
3478 }
3479 
3480 /*
3481  * Return 1 if the passed credential is in a jail and that jail does not
3482  * have its own virtual network stack, otherwise 0.
3483  */
3484 int
3485 jailed_without_vnet(struct ucred *cred)
3486 {
3487 
3488  if (!jailed(cred))
3489  return (0);
3490 #ifdef VIMAGE
3491  if (prison_owns_vnet(cred))
3492  return (0);
3493 #endif
3494 
3495  return (1);
3496 }
3497 
3498 /*
3499  * Return the correct hostname (domainname, et al) for the passed credential.
3500  */
3501 void
3502 getcredhostname(struct ucred *cred, char *buf, size_t size)
3503 {
3504  struct prison *pr;
3505 
3506  /*
3507  * A NULL credential can be used to shortcut to the physical
3508  * system's hostname.
3509  */
3510  pr = (cred != NULL) ? cred->cr_prison : &prison0;
3511  mtx_lock(&pr->pr_mtx);
3512  strlcpy(buf, pr->pr_hostname, size);
3513  mtx_unlock(&pr->pr_mtx);
3514 }
3515 
3516 void
3517 getcreddomainname(struct ucred *cred, char *buf, size_t size)
3518 {
3519 
3520  mtx_lock(&cred->cr_prison->pr_mtx);
3521  strlcpy(buf, cred->cr_prison->pr_domainname, size);
3522  mtx_unlock(&cred->cr_prison->pr_mtx);
3523 }
3524 
3525 void
3526 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
3527 {
3528 
3529  mtx_lock(&cred->cr_prison->pr_mtx);
3530  strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
3531  mtx_unlock(&cred->cr_prison->pr_mtx);
3532 }
3533 
3534 void
3535 getcredhostid(struct ucred *cred, unsigned long *hostid)
3536 {
3537 
3538  mtx_lock(&cred->cr_prison->pr_mtx);
3539  *hostid = cred->cr_prison->pr_hostid;
3540  mtx_unlock(&cred->cr_prison->pr_mtx);
3541 }
3542 
3543 #ifdef VIMAGE
3544 /*
3545  * Determine whether the prison represented by cred owns
3546  * its vnet rather than having it inherited.
3547  *
3548  * Returns 1 in case the prison owns the vnet, 0 otherwise.
3549  */
3550 int
3551 prison_owns_vnet(struct ucred *cred)
3552 {
3553 
3554  /*
3555  * vnets cannot be added/removed after jail creation,
3556  * so no need to lock here.
3557  */
3558  return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
3559 }
3560 #endif
3561 
3562 /*
3563  * Determine whether the subject represented by cred can "see"
3564  * status of a mount point.
3565  * Returns: 0 for permitted, ENOENT otherwise.
3566  * XXX: This function should be called cr_canseemount() and should be
3567  * placed in kern_prot.c.
3568  */
3569 int
3570 prison_canseemount(struct ucred *cred, struct mount *mp)
3571 {
3572  struct prison *pr;
3573  struct statfs *sp;
3574  size_t len;
3575 
3576  pr = cred->cr_prison;
3577  if (pr->pr_enforce_statfs == 0)
3578  return (0);
3579  if (pr->pr_root->v_mount == mp)
3580  return (0);
3581  if (pr->pr_enforce_statfs == 2)
3582  return (ENOENT);
3583  /*
3584  * If jail's chroot directory is set to "/" we should be able to see
3585  * all mount-points from inside a jail.
3586  * This is ugly check, but this is the only situation when jail's
3587  * directory ends with '/'.
3588  */
3589  if (strcmp(pr->pr_path, "/") == 0)
3590  return (0);
3591  len = strlen(pr->pr_path);
3592  sp = &mp->mnt_stat;
3593  if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
3594  return (ENOENT);
3595  /*
3596  * Be sure that we don't have situation where jail's root directory
3597  * is "/some/path" and mount point is "/some/pathpath".
3598  */
3599  if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
3600  return (ENOENT);
3601  return (0);
3602 }
3603 
3604 void
3605 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
3606 {
3607  char jpath[MAXPATHLEN];
3608  struct prison *pr;
3609  size_t len;
3610 
3611  pr = cred->cr_prison;
3612  if (pr->pr_enforce_statfs == 0)
3613  return;
3614  if (prison_canseemount(cred, mp) != 0) {
3615  bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3616  strlcpy(sp->f_mntonname, "[restricted]",
3617  sizeof(sp->f_mntonname));
3618  return;
3619  }
3620  if (pr->pr_root->v_mount == mp) {
3621  /*
3622  * Clear current buffer data, so we are sure nothing from
3623  * the valid path left there.
3624  */
3625  bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3626  *sp->f_mntonname = '/';
3627  return;
3628  }
3629  /*
3630  * If jail's chroot directory is set to "/" we should be able to see
3631  * all mount-points from inside a jail.
3632  */
3633  if (strcmp(pr->pr_path, "/") == 0)
3634  return;
3635  len = strlen(pr->pr_path);
3636  strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
3637  /*
3638  * Clear current buffer data, so we are sure nothing from
3639  * the valid path left there.
3640  */
3641  bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
3642  if (*jpath == '\0') {
3643  /* Should never happen. */
3644  *sp->f_mntonname = '/';
3645  } else {
3646  strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
3647  }
3648 }
3649 
3650 /*
3651  * Check with permission for a specific privilege is granted within jail. We
3652  * have a specific list of accepted privileges; the rest are denied.
3653  */
3654 int
3655 prison_priv_check(struct ucred *cred, int priv)
3656 {
3657 
3658  if (!jailed(cred))
3659  return (0);
3660 
3661 #ifdef VIMAGE
3662  /*
3663  * Privileges specific to prisons with a virtual network stack.
3664  * There might be a duplicate entry here in case the privilege
3665  * is only granted conditionally in the legacy jail case.
3666  */
3667  switch (priv) {
3668 #ifdef notyet
3669  /*
3670  * NFS-specific privileges.
3671  */
3672  case PRIV_NFS_DAEMON:
3673  case PRIV_NFS_LOCKD:
3674 #endif
3675  /*
3676  * Network stack privileges.
3677  */
3678  case PRIV_NET_BRIDGE:
3679  case PRIV_NET_GRE:
3680  case PRIV_NET_BPF:
3681  case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */
3682  case PRIV_NET_ROUTE:
3683  case PRIV_NET_TAP:
3684  case PRIV_NET_SETIFMTU:
3685  case PRIV_NET_SETIFFLAGS:
3686  case PRIV_NET_SETIFCAP:
3687  case PRIV_NET_SETIFDESCR:
3688  case PRIV_NET_SETIFNAME :
3689  case PRIV_NET_SETIFMETRIC:
3690  case PRIV_NET_SETIFPHYS:
3691  case PRIV_NET_SETIFMAC:
3692  case PRIV_NET_ADDMULTI:
3693  case PRIV_NET_DELMULTI:
3694  case PRIV_NET_HWIOCTL:
3695  case PRIV_NET_SETLLADDR:
3696  case PRIV_NET_ADDIFGROUP:
3697  case PRIV_NET_DELIFGROUP:
3698  case PRIV_NET_IFCREATE:
3699  case PRIV_NET_IFDESTROY:
3700  case PRIV_NET_ADDIFADDR:
3701  case PRIV_NET_DELIFADDR:
3702  case PRIV_NET_LAGG:
3703  case PRIV_NET_GIF:
3704  case PRIV_NET_SETIFVNET:
3705  case PRIV_NET_SETIFFIB:
3706 
3707  /*
3708  * 802.11-related privileges.
3709  */
3710  case PRIV_NET80211_GETKEY:
3711 #ifdef notyet
3712  case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */
3713 #endif
3714 
3715 #ifdef notyet
3716  /*
3717  * AppleTalk privileges.
3718  */
3719  case PRIV_NETATALK_RESERVEDPORT:
3720 
3721  /*
3722  * ATM privileges.
3723  */
3724  case PRIV_NETATM_CFG:
3725  case PRIV_NETATM_ADD:
3726  case PRIV_NETATM_DEL:
3727  case PRIV_NETATM_SET:
3728 
3729  /*
3730  * Bluetooth privileges.
3731  */
3732  case PRIV_NETBLUETOOTH_RAW:
3733 #endif
3734 
3735  /*
3736  * Netgraph and netgraph module privileges.
3737  */
3738  case PRIV_NETGRAPH_CONTROL:
3739 #ifdef notyet
3740  case PRIV_NETGRAPH_TTY:
3741 #endif
3742 
3743  /*
3744  * IPv4 and IPv6 privileges.
3745  */
3746  case PRIV_NETINET_IPFW:
3747  case PRIV_NETINET_DIVERT:
3748  case PRIV_NETINET_PF:
3749  case PRIV_NETINET_DUMMYNET:
3750  case PRIV_NETINET_CARP:
3751  case PRIV_NETINET_MROUTE:
3752  case PRIV_NETINET_RAW:
3753  case PRIV_NETINET_ADDRCTRL6:
3754  case PRIV_NETINET_ND6:
3755  case PRIV_NETINET_SCOPE6:
3756  case PRIV_NETINET_ALIFETIME6:
3757  case PRIV_NETINET_IPSEC:
3758  case PRIV_NETINET_BINDANY:
3759 
3760 #ifdef notyet
3761  /*
3762  * IPX/SPX privileges.
3763  */
3764  case PRIV_NETIPX_RESERVEDPORT:
3765  case PRIV_NETIPX_RAW:
3766 
3767  /*
3768  * NCP privileges.
3769  */
3770  case PRIV_NETNCP:
3771 
3772  /*
3773  * SMB privileges.
3774  */
3775  case PRIV_NETSMB:
3776 #endif
3777 
3778  /*
3779  * No default: or deny here.
3780  * In case of no permit fall through to next switch().
3781  */
3782  if (cred->cr_prison->pr_flags & PR_VNET)
3783  return (0);
3784  }
3785 #endif /* VIMAGE */
3786 
3787  switch (priv) {
3788 
3789  /*
3790  * Allow ktrace privileges for root in jail.
3791  */
3792  case PRIV_KTRACE:
3793 
3794 #if 0
3795  /*
3796  * Allow jailed processes to configure audit identity and
3797  * submit audit records (login, etc). In the future we may
3798  * want to further refine the relationship between audit and
3799  * jail.
3800  */
3801  case PRIV_AUDIT_GETAUDIT:
3802  case PRIV_AUDIT_SETAUDIT:
3803  case PRIV_AUDIT_SUBMIT:
3804 #endif
3805 
3806  /*
3807  * Allow jailed processes to manipulate process UNIX
3808  * credentials in any way they see fit.
3809  */
3810  case PRIV_CRED_SETUID:
3811  case PRIV_CRED_SETEUID:
3812  case PRIV_CRED_SETGID:
3813  case PRIV_CRED_SETEGID:
3814  case PRIV_CRED_SETGROUPS:
3815  case PRIV_CRED_SETREUID:
3816  case PRIV_CRED_SETREGID:
3817  case PRIV_CRED_SETRESUID:
3818  case PRIV_CRED_SETRESGID:
3819 
3820  /*
3821  * Jail implements visibility constraints already, so allow
3822  * jailed root to override uid/gid-based constraints.
3823  */
3824  case PRIV_SEEOTHERGIDS:
3825  case PRIV_SEEOTHERUIDS:
3826 
3827  /*
3828  * Jail implements inter-process debugging limits already, so
3829  * allow jailed root various debugging privileges.
3830  */
3831  case PRIV_DEBUG_DIFFCRED:
3832  case PRIV_DEBUG_SUGID:
3833  case PRIV_DEBUG_UNPRIV:
3834 
3835  /*
3836  * Allow jail to set various resource limits and login
3837  * properties, and for now, exceed process resource limits.
3838  */
3839  case PRIV_PROC_LIMIT:
3840  case PRIV_PROC_SETLOGIN:
3841  case PRIV_PROC_SETRLIMIT:
3842 
3843  /*
3844  * System V and POSIX IPC privileges are granted in jail.
3845  */
3846  case PRIV_IPC_READ:
3847  case PRIV_IPC_WRITE:
3848  case PRIV_IPC_ADMIN:
3849  case PRIV_IPC_MSGSIZE:
3850  case PRIV_MQ_ADMIN:
3851 
3852  /*
3853  * Jail operations within a jail work on child jails.
3854  */
3855  case PRIV_JAIL_ATTACH:
3856  case PRIV_JAIL_SET:
3857  case PRIV_JAIL_REMOVE:
3858 
3859  /*
3860  * Jail implements its own inter-process limits, so allow
3861  * root processes in jail to change scheduling on other
3862  * processes in the same jail. Likewise for signalling.
3863  */
3864  case PRIV_SCHED_DIFFCRED:
3865  case PRIV_SCHED_CPUSET:
3866  case PRIV_SIGNAL_DIFFCRED:
3867  case PRIV_SIGNAL_SUGID:
3868 
3869  /*
3870  * Allow jailed processes to write to sysctls marked as jail
3871  * writable.
3872  */
3873  case PRIV_SYSCTL_WRITEJAIL:
3874 
3875  /*
3876  * Allow root in jail to manage a variety of quota
3877  * properties. These should likely be conditional on a
3878  * configuration option.
3879  */
3880  case PRIV_VFS_GETQUOTA:
3881  case PRIV_VFS_SETQUOTA:
3882 
3883  /*
3884  * Since Jail relies on chroot() to implement file system
3885  * protections, grant many VFS privileges to root in jail.
3886  * Be careful to exclude mount-related and NFS-related
3887  * privileges.
3888  */
3889  case PRIV_VFS_READ:
3890  case PRIV_VFS_WRITE:
3891  case PRIV_VFS_ADMIN:
3892  case PRIV_VFS_EXEC:
3893  case PRIV_VFS_LOOKUP:
3894  case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */
3895  case PRIV_VFS_CHFLAGS_DEV:
3896  case PRIV_VFS_CHOWN:
3897  case PRIV_VFS_CHROOT:
3898  case PRIV_VFS_RETAINSUGID:
3899  case PRIV_VFS_FCHROOT:
3900  case PRIV_VFS_LINK:
3901  case PRIV_VFS_SETGID:
3902  case PRIV_VFS_STAT:
3903  case PRIV_VFS_STICKYFILE:
3904  return (0);
3905 
3906  /*
3907  * Depending on the global setting, allow privilege of
3908  * setting system flags.
3909  */
3910  case PRIV_VFS_SYSFLAGS:
3911  if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
3912  return (0);
3913  else
3914  return (EPERM);
3915 
3916  /*
3917  * Depending on the global setting, allow privilege of
3918  * mounting/unmounting file systems.
3919  */
3920  case PRIV_VFS_MOUNT:
3921  case PRIV_VFS_UNMOUNT:
3922  case PRIV_VFS_MOUNT_NONUSER:
3923  case PRIV_VFS_MOUNT_OWNER:
3924  if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
3925  cred->cr_prison->pr_enforce_statfs < 2)
3926  return (0);
3927  else
3928  return (EPERM);
3929 
3930  /*
3931  * Allow jailed root to bind reserved ports and reuse in-use
3932  * ports.
3933  */
3934  case PRIV_NETINET_RESERVEDPORT:
3935  case PRIV_NETINET_REUSEPORT:
3936  return (0);
3937 
3938  /*
3939  * Allow jailed root to set certian IPv4/6 (option) headers.
3940  */
3941  case PRIV_NETINET_SETHDROPTS:
3942  return (0);
3943 
3944  /*
3945  * Conditionally allow creating raw sockets in jail.
3946  */
3947  case PRIV_NETINET_RAW:
3948  if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
3949  return (0);
3950  else
3951  return (EPERM);
3952 
3953  /*
3954  * Since jail implements its own visibility limits on netstat
3955  * sysctls, allow getcred. This allows identd to work in
3956  * jail.
3957  */
3958  case PRIV_NETINET_GETCRED:
3959  return (0);
3960 
3961  /*
3962  * Allow jailed root to set loginclass.
3963  */
3964  case PRIV_PROC_SETLOGINCLASS:
3965  return (0);
3966 
3967  default:
3968  /*
3969  * In all remaining cases, deny the privilege request. This
3970  * includes almost all network privileges, many system
3971  * configuration privileges.
3972  */
3973  return (EPERM);
3974  }
3975 }
3976 
3977 /*
3978  * Return the part of pr2's name that is relative to pr1, or the whole name
3979  * if it does not directly follow.
3980  */
3981 
3982 char *
3983 prison_name(struct prison *pr1, struct prison *pr2)
3984 {
3985  char *name;
3986 
3987  /* Jails see themselves as "0" (if they see themselves at all). */
3988  if (pr1 == pr2)
3989  return "0";
3990  name = pr2->pr_name;
3991  if (prison_ischild(pr1, pr2)) {
3992  /*
3993  * pr1 isn't locked (and allprison_lock may not be either)
3994  * so its length can't be counted on. But the number of dots
3995  * can be counted on - and counted.
3996  */
3997  for (; pr1 != &prison0; pr1 = pr1->pr_parent)
3998  name = strchr(name, '.') + 1;
3999  }
4000  return (name);
4001 }
4002 
4003 /*
4004  * Return the part of pr2's path that is relative to pr1, or the whole path
4005  * if it does not directly follow.
4006  */
4007 static char *
4008 prison_path(struct prison *pr1, struct prison *pr2)
4009 {
4010  char *path1, *path2;
4011  int len1;
4012 
4013  path1 = pr1->pr_path;
4014  path2 = pr2->pr_path;
4015  if (!strcmp(path1, "/"))
4016  return (path2);
4017  len1 = strlen(path1);
4018  if (strncmp(path1, path2, len1))
4019  return (path2);
4020  if (path2[len1] == '\0')
4021  return "/";
4022  if (path2[len1] == '/')
4023  return (path2 + len1);
4024  return (path2);
4025 }
4026 
4027 
4028 /*
4029  * Jail-related sysctls.
4030  */
4031 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
4032  "Jails");
4033 
4034 static int
4035 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
4036 {
4037  struct xprison *xp;
4038  struct prison *pr, *cpr;
4039 #ifdef INET
4040  struct in_addr *ip4 = NULL;
4041  int ip4s = 0;
4042 #endif
4043 #ifdef INET6
4044  struct in6_addr *ip6 = NULL;
4045  int ip6s = 0;
4046 #endif
4047  int descend, error;
4048 
4049  xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
4050  pr = req->td->td_ucred->cr_prison;
4051  error = 0;
4052  sx_slock(&allprison_lock);
4053  FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
4054 #if defined(INET) || defined(INET6)
4055  again:
4056 #endif
4057  mtx_lock(&cpr->pr_mtx);
4058 #ifdef INET
4059  if (cpr->pr_ip4s > 0) {
4060  if (ip4s < cpr->pr_ip4s) {
4061  ip4s = cpr->pr_ip4s;
4062  mtx_unlock(&cpr->pr_mtx);
4063  ip4 = realloc(ip4, ip4s *
4064  sizeof(struct in_addr), M_TEMP, M_WAITOK);
4065  goto again;
4066  }
4067  bcopy(cpr->pr_ip4, ip4,
4068  cpr->pr_ip4s * sizeof(struct in_addr));
4069  }
4070 #endif
4071 #ifdef INET6
4072  if (cpr->pr_ip6s > 0) {
4073  if (ip6s < cpr->pr_ip6s) {
4074  ip6s = cpr->pr_ip6s;
4075  mtx_unlock(&cpr->pr_mtx);
4076  ip6 = realloc(ip6, ip6s *
4077  sizeof(struct in6_addr), M_TEMP, M_WAITOK);
4078  goto again;
4079  }
4080  bcopy(cpr->pr_ip6, ip6,
4081  cpr->pr_ip6s * sizeof(struct in6_addr));
4082  }
4083 #endif
4084  if (cpr->pr_ref == 0) {
4085  mtx_unlock(&cpr->pr_mtx);
4086  continue;
4087  }
4088  bzero(xp, sizeof(*xp));
4089  xp->pr_version = XPRISON_VERSION;
4090  xp->pr_id = cpr->pr_id;
4091  xp->pr_state = cpr->pr_uref > 0
4092  ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
4093  strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
4094  strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
4095  strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
4096 #ifdef INET
4097  xp->pr_ip4s = cpr->pr_ip4s;
4098 #endif
4099 #ifdef INET6
4100  xp->pr_ip6s = cpr->pr_ip6s;
4101 #endif
4102  mtx_unlock(&cpr->pr_mtx);
4103  error = SYSCTL_OUT(req, xp, sizeof(*xp));
4104  if (error)
4105  break;
4106 #ifdef INET
4107  if (xp->pr_ip4s > 0) {
4108  error = SYSCTL_OUT(req, ip4,
4109  xp->pr_ip4s * sizeof(struct in_addr));
4110  if (error)
4111  break;
4112  }
4113 #endif
4114 #ifdef INET6
4115  if (xp->pr_ip6s > 0) {
4116  error = SYSCTL_OUT(req, ip6,
4117  xp->pr_ip6s * sizeof(struct in6_addr));
4118  if (error)
4119  break;
4120  }
4121 #endif
4122  }
4123  sx_sunlock(&allprison_lock);
4124  free(xp, M_TEMP);
4125 #ifdef INET
4126  free(ip4, M_TEMP);
4127 #endif
4128 #ifdef INET6
4129  free(ip6, M_TEMP);
4130 #endif
4131  return (error);
4132 }
4133 
4134 SYSCTL_OID(_security_jail, OID_AUTO, list,
4135  CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4136  sysctl_jail_list, "S", "List of active jails");
4137 
4138 static int
4139 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
4140 {
4141  int error, injail;
4142 
4143  injail = jailed(req->td->td_ucred);
4144  error = SYSCTL_OUT(req, &injail, sizeof(injail));
4145 
4146  return (error);
4147 }
4148 
4149 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
4150  CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4151  sysctl_jail_jailed, "I", "Process in jail?");
4152 
4153 static int
4154 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
4155 {
4156  int error, havevnet;
4157 #ifdef VIMAGE
4158  struct ucred *cred = req->td->td_ucred;
4159 
4160  havevnet = jailed(cred) && prison_owns_vnet(cred);
4161 #else
4162  havevnet = 0;
4163 #endif
4164  error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet));
4165 
4166  return (error);
4167 }
4168 
4169 SYSCTL_PROC(_security_jail, OID_AUTO, vnet,
4170  CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
4171  sysctl_jail_vnet, "I", "Jail owns VNET?");
4172 
4173 #if defined(INET) || defined(INET6)
4174 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
4175  &jail_max_af_ips, 0,
4176  "Number of IP addresses a jail may have at most per address family");
4177 #endif
4178 
4179 /*
4180  * Default parameters for jail(2) compatability. For historical reasons,
4181  * the sysctl names have varying similarity to the parameter names. Prisons
4182  * just see their own parameters, and can't change them.
4183  */
4184 static int
4185 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
4186 {
4187  struct prison *pr;
4188  int allow, error, i;
4189 
4190  pr = req->td->td_ucred->cr_prison;
4191  allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
4192 
4193  /* Get the current flag value, and convert it to a boolean. */
4194  i = (allow & arg2) ? 1 : 0;
4195  if (arg1 != NULL)
4196  i = !i;
4197  error = sysctl_handle_int(oidp, &i, 0, req);
4198  if (error || !req->newptr)
4199  return (error);
4200  i = i ? arg2 : 0;
4201  if (arg1 != NULL)
4202  i ^= arg2;
4203  /*
4204  * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
4205  * for writing.
4206  */
4207  mtx_lock(&prison0.pr_mtx);
4208  jail_default_allow = (jail_default_allow & ~arg2) | i;
4209  mtx_unlock(&prison0.pr_mtx);
4210  return (0);
4211 }
4212 
4213 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
4214  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4215  NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
4216  "Processes in jail can set their hostnames");
4217 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
4218  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4219  (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
4220  "Processes in jail are limited to creating UNIX/IP/route sockets only");
4221 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
4222  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4223  NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
4224  "Processes in jail can use System V IPC primitives");
4225 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
4226  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4227  NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
4228  "Prison root can create raw sockets");
4229 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
4230  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4231  NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
4232  "Processes in jail can alter system file flags");
4233 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
4234  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4235  NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
4236  "Processes in jail can mount/unmount jail-friendly file systems");
4237 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed,
4238  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4239  NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I",
4240  "Processes in jail can mount the devfs file system");
4241 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed,
4242  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4243  NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I",
4244  "Processes in jail can mount the nullfs file system");
4245 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed,
4246  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4247  NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I",
4248  "Processes in jail can mount the procfs file system");
4249 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed,
4250  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4251  NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I",
4252  "Processes in jail can mount the zfs file system");
4253 
4254 static int
4255 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
4256 {
4257  struct prison *pr;
4258  int level, error;
4259 
4260  pr = req->td->td_ucred->cr_prison;
4261  level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
4262  error = sysctl_handle_int(oidp, &level, 0, req);
4263  if (error || !req->newptr)
4264  return (error);
4265  *(int *)arg1 = level;
4266  return (0);
4267 }
4268 
4269 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
4270  CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
4271  &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
4272  sysctl_jail_default_level, "I",
4273  "Processes in jail cannot see all mounted file systems");
4274 
4275 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset,
4276  CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
4277  &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum),
4278  sysctl_jail_default_level, "I",
4279  "Ruleset for the devfs filesystem in jail");
4280 
4281 /*
4282  * Nodes to describe jail parameters. Maximum length of string parameters
4283  * is returned in the string itself, and the other parameters exist merely
4284  * to make themselves and their types known.
4285  */
4286 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
4287  "Jail parameters");
4288 
4289 int
4290 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
4291 {
4292  int i;
4293  long l;
4294  size_t s;
4295  char numbuf[12];
4296 
4297  switch (oidp->oid_kind & CTLTYPE)
4298  {
4299  case CTLTYPE_LONG:
4300  case CTLTYPE_ULONG:
4301  l = 0;
4302 #ifdef SCTL_MASK32
4303  if (!(req->flags & SCTL_MASK32))
4304 #endif
4305  return (SYSCTL_OUT(req, &l, sizeof(l)));
4306  case CTLTYPE_INT:
4307  case CTLTYPE_UINT:
4308  i = 0;
4309  return (SYSCTL_OUT(req, &i, sizeof(i)));
4310  case CTLTYPE_STRING:
4311  snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
4312  return
4313  (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
4314  case CTLTYPE_STRUCT:
4315  s = (size_t)arg2;
4316  return (SYSCTL_OUT(req, &s, sizeof(s)));
4317  }
4318  return (0);
4319 }
4320 
4321 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
4322 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
4323 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
4324 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
4325 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
4326  "I", "Jail secure level");
4327 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
4328  "I", "Jail cannot see all mounted file systems");
4329 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW,
4330  "I", "Ruleset for in-jail devfs mounts");
4331 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
4332  "B", "Jail persistence");
4333 #ifdef VIMAGE
4334 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
4335  "E,jailsys", "Virtual network stack");
4336 #endif
4337 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
4338  "B", "Jail is in the process of shutting down");
4339 
4340 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
4341 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
4342  "I", "Current number of child jails");
4343 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
4344  "I", "Maximum number of child jails");
4345 
4346 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
4347 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
4348  "Jail hostname");
4349 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
4350  "Jail NIS domainname");
4351 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
4352  "Jail host UUID");
4353 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
4354  "LU", "Jail host ID");
4355 
4356 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
4357 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
4358 
4359 #ifdef INET
4360 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
4361  "Jail IPv4 address virtualization");
4362 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
4363  "S,in_addr,a", "Jail IPv4 addresses");
4364 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4365  "B", "Do (not) use IPv4 source address selection rather than the "
4366  "primary jail IPv4 address.");
4367 #endif
4368 #ifdef INET6
4369 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
4370  "Jail IPv6 address virtualization");
4371 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
4372  "S,in6_addr,a", "Jail IPv6 addresses");
4373 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
4374  "B", "Do (not) use IPv6 source address selection rather than the "
4375  "primary jail IPv6 address.");
4376 #endif
4377 
4378 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
4379 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
4380  "B", "Jail may set hostname");
4381 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
4382  "B", "Jail may use SYSV IPC");
4383 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
4384  "B", "Jail may create raw sockets");
4385 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
4386  "B", "Jail may alter system file flags");
4387 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
4388  "B", "Jail may set file quotas");
4389 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
4390  "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
4391 
4392 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags");
4393 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW,
4394  "B", "Jail may mount/unmount jail-friendly file systems in general");
4395 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW,
4396  "B", "Jail may mount the devfs file system");
4397 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW,
4398  "B", "Jail may mount the nullfs file system");
4399 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW,
4400  "B", "Jail may mount the procfs file system");
4401 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW,
4402  "B", "Jail may mount the zfs file system");
4403 
4404 void
4405 prison_racct_foreach(void (*callback)(struct racct *racct,
4406  void *arg2, void *arg3), void *arg2, void *arg3)
4407 {
4408  struct prison_racct *prr;
4409 
4410  sx_slock(&allprison_lock);
4411  LIST_FOREACH(prr, &allprison_racct, prr_next)
4412  (callback)(prr->prr_racct, arg2, arg3);
4413  sx_sunlock(&allprison_lock);
4414 }
4415 
4416 static struct prison_racct *
4417 prison_racct_find_locked(const char *name)
4418 {
4419  struct prison_racct *prr;
4420 
4421  sx_assert(&allprison_lock, SA_XLOCKED);
4422 
4423  if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
4424  return (NULL);
4425 
4426  LIST_FOREACH(prr, &allprison_racct, prr_next) {
4427  if (strcmp(name, prr->prr_name) != 0)
4428  continue;
4429 
4430  /* Found prison_racct with a matching name? */
4431  prison_racct_hold(prr);
4432  return (prr);
4433  }
4434 
4435  /* Add new prison_racct. */
4436  prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
4437  racct_create(&prr->prr_racct);
4438 
4439  strcpy(prr->prr_name, name);
4440  refcount_init(&prr->prr_refcount, 1);
4441  LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
4442 
4443  return (prr);
4444 }
4445 
4446 struct prison_racct *
4447 prison_racct_find(const char *name)
4448 {
4449  struct prison_racct *prr;
4450 
4451  sx_xlock(&allprison_lock);
4452  prr = prison_racct_find_locked(name);
4453  sx_xunlock(&allprison_lock);
4454  return (prr);
4455 }
4456 
4457 void
4458 prison_racct_hold(struct prison_racct *prr)
4459 {
4460 
4461  refcount_acquire(&prr->prr_refcount);
4462 }
4463 
4464 static void
4465 prison_racct_free_locked(struct prison_racct *prr)
4466 {
4467 
4468  sx_assert(&allprison_lock, SA_XLOCKED);
4469 
4470  if (refcount_release(&prr->prr_refcount)) {
4471  racct_destroy(&prr->prr_racct);
4472  LIST_REMOVE(prr, prr_next);
4473  free(prr, M_PRISON_RACCT);
4474  }
4475 }
4476 
4477 void
4478 prison_racct_free(struct prison_racct *prr)
4479 {
4480  int old;
4481 
4482  sx_assert(&allprison_lock, SA_UNLOCKED);
4483 
4484  old = prr->prr_refcount;
4485  if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
4486  return;
4487 
4488  sx_xlock(&allprison_lock);
4489  prison_racct_free_locked(prr);
4490  sx_xunlock(&allprison_lock);
4491 }
4492 
4493 #ifdef RACCT
4494 static void
4495 prison_racct_attach(struct prison *pr)
4496 {
4497  struct prison_racct *prr;
4498 
4499  sx_assert(&allprison_lock, SA_XLOCKED);
4500 
4501  prr = prison_racct_find_locked(pr->pr_name);
4502  KASSERT(prr != NULL, ("cannot find prison_racct"));
4503 
4504  pr->pr_prison_racct = prr;
4505 }
4506 
4507 /*
4508  * Handle jail renaming. From the racct point of view, renaming means
4509  * moving from one prison_racct to another.
4510  */
4511 static void
4512 prison_racct_modify(struct prison *pr)
4513 {
4514  struct proc *p;
4515  struct ucred *cred;
4516  struct prison_racct *oldprr;
4517 
4518  sx_slock(&allproc_lock);
4519  sx_xlock(&allprison_lock);
4520 
4521  if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) {
4522  sx_xunlock(&allprison_lock);
4523  sx_sunlock(&allproc_lock);
4524  return;
4525  }
4526 
4527  oldprr = pr->pr_prison_racct;
4528  pr->pr_prison_racct = NULL;
4529 
4530  prison_racct_attach(pr);
4531 
4532  /*
4533  * Move resource utilisation records.
4534  */
4535  racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct);
4536 
4537  /*
4538  * Force rctl to reattach rules to processes.
4539  */
4540  FOREACH_PROC_IN_SYSTEM(p) {
4541  PROC_LOCK(p);
4542  cred = crhold(p->p_ucred);
4543  PROC_UNLOCK(p);
4544  racct_proc_ucred_changed(p, cred, cred);
4545  crfree(cred);
4546  }
4547 
4548  sx_sunlock(&allproc_lock);
4549  prison_racct_free_locked(oldprr);
4550  sx_xunlock(&allprison_lock);
4551 }
4552 
4553 static void
4554 prison_racct_detach(struct prison *pr)
4555 {
4556 
4557  sx_assert(&allprison_lock, SA_UNLOCKED);
4558 
4559  if (pr->pr_prison_racct == NULL)
4560  return;
4561  prison_racct_free(pr->pr_prison_racct);
4562  pr->pr_prison_racct = NULL;
4563 }
4564 #endif /* RACCT */
4565 
4566 #ifdef DDB
4567 
4568 static void
4569 db_show_prison(struct prison *pr)
4570 {
4571  int fi;
4572 #if defined(INET) || defined(INET6)
4573  int ii;
4574 #endif
4575  unsigned jsf;
4576 #ifdef INET6
4577  char ip6buf[INET6_ADDRSTRLEN];
4578 #endif
4579 
4580  db_printf("prison %p:\n", pr);
4581  db_printf(" jid = %d\n", pr->pr_id);
4582  db_printf(" name = %s\n", pr->pr_name);
4583  db_printf(" parent = %p\n", pr->pr_parent);
4584  db_printf(" ref = %d\n", pr->pr_ref);
4585  db_printf(" uref = %d\n", pr->pr_uref);
4586  db_printf(" path = %s\n", pr->pr_path);
4587  db_printf(" cpuset = %d\n", pr->pr_cpuset
4588  ? pr->pr_cpuset->cs_id : -1);
4589 #ifdef VIMAGE
4590  db_printf(" vnet = %p\n", pr->pr_vnet);
4591 #endif
4592  db_printf(" root = %p\n", pr->pr_root);
4593  db_printf(" securelevel = %d\n", pr->pr_securelevel);
4594  db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum);
4595  db_printf(" children.max = %d\n", pr->pr_childmax);
4596  db_printf(" children.cur = %d\n", pr->pr_childcount);
4597  db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children));
4598  db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling));
4599  db_printf(" flags = 0x%x", pr->pr_flags);
4600  for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
4601  fi++)
4602  if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
4603  db_printf(" %s", pr_flag_names[fi]);
4604  for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
4605  fi++) {
4606  jsf = pr->pr_flags &
4607  (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
4608  db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
4609  pr_flag_jailsys[fi].disable &&
4610  (jsf == pr_flag_jailsys[fi].disable) ? "disable"
4611  : (jsf == pr_flag_jailsys[fi].new) ? "new"
4612  : "inherit");
4613  }
4614  db_printf(" allow = 0x%x", pr->pr_allow);
4615  for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
4616  fi++)
4617  if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
4618  db_printf(" %s", pr_allow_names[fi]);
4619  db_printf("\n");
4620  db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs);
4621  db_printf(" host.hostname = %s\n", pr->pr_hostname);
4622  db_printf(" host.domainname = %s\n", pr->pr_domainname);
4623  db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid);
4624  db_printf(" host.hostid = %lu\n", pr->pr_hostid);
4625 #ifdef INET
4626  db_printf(" ip4s = %d\n", pr->pr_ip4s);
4627  for (ii = 0; ii < pr->pr_ip4s; ii++)
4628  db_printf(" %s %s\n",
4629  ii == 0 ? "ip4.addr =" : " ",
4630  inet_ntoa(pr->pr_ip4[ii]));
4631 #endif
4632 #ifdef INET6
4633  db_printf(" ip6s = %d\n", pr->pr_ip6s);
4634  for (ii = 0; ii < pr->pr_ip6s; ii++)
4635  db_printf(" %s %s\n",
4636  ii == 0 ? "ip6.addr =" : " ",
4637  ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
4638 #endif
4639 }
4640 
4641 DB_SHOW_COMMAND(prison, db_show_prison_command)
4642 {
4643  struct prison *pr;
4644 
4645  if (!have_addr) {
4646  /*
4647  * Show all prisons in the list, and prison0 which is not
4648  * listed.
4649  */
4650  db_show_prison(&prison0);
4651  if (!db_pager_quit) {
4652  TAILQ_FOREACH(pr, &allprison, pr_list) {
4653  db_show_prison(pr);
4654  if (db_pager_quit)
4655  break;
4656  }
4657  }
4658  return;
4659  }
4660 
4661  if (addr == 0)
4662  pr = &prison0;
4663  else {
4664  /* Look for a prison with the ID and with references. */
4665  TAILQ_FOREACH(pr, &allprison, pr_list)
4666  if (pr->pr_id == addr && pr->pr_ref > 0)
4667  break;
4668  if (pr == NULL)
4669  /* Look again, without requiring a reference. */
4670  TAILQ_FOREACH(pr, &allprison, pr_list)
4671  if (pr->pr_id == addr)
4672  break;
4673  if (pr == NULL)
4674  /* Assume address points to a valid prison. */
4675  pr = (struct prison *)addr;
4676  }
4677  db_show_prison(pr);
4678 }
4679 
4680 #endif /* DDB */
prison_allow
int prison_allow(struct ucred *cred, unsigned flag)
Definition: kern_jail.c:2502
allproc
struct proclist allproc
Definition: kern_proc.c:134
prison_enforce_statfs
void prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
Definition: kern_jail.c:3605
getcredhostid
void getcredhostid(struct ucred *cred, unsigned long *hostid)
Definition: kern_jail.c:3535
sysctl_jail_list
static int sysctl_jail_list(SYSCTL_HANDLER_ARGS)
Definition: kern_jail.c:4035
kern_jail_set
int kern_jail_set(struct thread *td, struct uio *optuio, int flags)
Definition: kern_jail.c:522
realloc
void * realloc(void *addr, unsigned long size, struct malloc_type *mtp, int flags)
Definition: kern_malloc.c:616
pr_allow_nonames_size
const size_t pr_allow_nonames_size
Definition: kern_jail.c:225
buf
struct buf * buf
Definition: vfs_bio.c:97
prison_check_af
int prison_check_af(struct ucred *cred, int af)
Definition: kern_jail.c:3348
prison_deref
static void prison_deref(struct prison *pr, int flags)
Definition: kern_jail.c:2550
path
char * path
Definition: vfs_syscalls.c:2157
NDFREE
void NDFREE(struct nameidata *ndp, const u_int flags)
Definition: vfs_lookup.c:1091
pr_flag_jailsys
struct jailsys_flags pr_flag_jailsys[]
vfs_setopts
int vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
Definition: vfs_mount.c:1623
snprintf
int snprintf(char *str, size_t size, const char *format,...)
Definition: subr_prf.c:509
prison_find
struct prison * prison_find(int prid)
Definition: kern_jail.c:2417
vfs_freeopts
void vfs_freeopts(struct vfsoptlist *opts)
Definition: vfs_mount.c:154
jailsys_flags::name
const char * name
Definition: kern_jail.c:180
vfs_flagopt
int vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w, uint64_t val)
Definition: vfs_mount.c:1536
sysctl_jail_default_level
static int sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
Definition: kern_jail.c:4255
vfs_opterror
void vfs_opterror(struct vfsoptlist *opts, const char *fmt,...)
Definition: vfs_mount.c:1392
jail_default_allow
static unsigned jail_default_allow
Definition: kern_jail.c:230
malloc
void * malloc(unsigned long size, struct malloc_type *mtp, int flags)
Definition: kern_malloc.c:454
prison_racct_find_locked
static struct prison_racct * prison_racct_find_locked(const char *name)
Definition: kern_jail.c:4417
sys_jail_set
int sys_jail_set(struct thread *td, struct jail_set_args *uap)
Definition: kern_jail.c:504
vn_path_to_global_path
int vn_path_to_global_path(struct thread *td, struct vnode *vp, char *path, u_int pathlen)
Definition: vfs_cache.c:1463
vfs_getopt
int vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
Definition: vfs_mount.c:1474
pr_flag_nonames_size
const size_t pr_flag_nonames_size
Definition: kern_jail.c:177
SX_SYSINIT
SX_SYSINIT(allprison_lock,&allprison_lock,"allprison")
prison_proc_hold
void prison_proc_hold(struct prison *pr)
Definition: kern_jail.c:2641
parent
device_t parent
Definition: device_if.m:171
vfs_copyopt
int vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
Definition: vfs_mount.c:1649
prison_check
int prison_check(struct ucred *cred1, struct ucred *cred2)
Definition: kern_jail.c:3450
prison_racct_free
void prison_racct_free(struct prison_racct *prr)
Definition: kern_jail.c:4478
SYSCTL_JAIL_PARAM
SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT|CTLFLAG_RDTUN,"I","Jail ID")
name
const char * name
Definition: kern_fail.c:97
vfs_setopt
int vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
Definition: vfs_mount.c:1580
sysctl_jail_jailed
static int sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
Definition: kern_jail.c:4139
SYSCTL_JAIL_PARAM_NODE
SYSCTL_JAIL_PARAM_NODE(children,"Number of child jails")
setsugid
void setsugid(struct proc *p)
Definition: kern_prot.c:2127
SYSCTL_UINT
SYSCTL_UINT(_kern_eventtimer, OID_AUTO, idletick, CTLFLAG_RW,&idletick, 0,"Run periodic events when idle")
sys_jail_get
int sys_jail_get(struct thread *td, struct jail_get_args *uap)
Definition: kern_jail.c:1902
callback
linker_function_name_callback_t callback
Definition: linker_if.m:62
kern_psignal
void kern_psignal(struct proc *p, int sig)
Definition: kern_sig.c:1975
MALLOC_DEFINE
MALLOC_DEFINE(M_PRISON,"prison","Prison structures")
sysctl_handle_string
int sysctl_handle_string(SYSCTL_HANDLER_ARGS)
Definition: kern_sysctl.c:1121
sys_jail_attach
int sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
Definition: kern_jail.c:2308
pr_flag_names_size
const size_t pr_flag_names_size
Definition: kern_jail.c:166
priv_check
int priv_check(struct thread *td, int priv)
Definition: kern_priv.c:170
vfs_getopt_pos
int vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
Definition: vfs_mount.c:1498
vref
void vref(struct vnode *vp)
Definition: vfs_subr.c:2302
SYSCTL_JAIL_PARAM_STRING
SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN,"Jail name")
allproc_lock
struct sx allproc_lock
Definition: kern_proc.c:136
PD_DEUREF
#define PD_DEUREF
vput
void vput(struct vnode *vp)
Definition: vfs_subr.c:2428
SYSCTL_OID
SYSCTL_OID(_security_jail, OID_AUTO, list, CTLTYPE_STRUCT|CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, 0, sysctl_jail_list,"S","List of active jails")
prison_find_child
struct prison * prison_find_child(struct prison *mypr, int prid)
Definition: kern_jail.c:2437
jail_default_enforce_statfs
static int jail_default_enforce_statfs
Definition: kern_jail.c:231
pr_allow_names
static char * pr_allow_names[]
Definition: kern_jail.c:197
allprison
struct prisonlist allprison
Definition: kern_jail.c:122
JAIL_DEFAULT_DEVFS_RSNUM
#define JAIL_DEFAULT_DEVFS_RSNUM
Definition: kern_jail.c:229
PD_LOCKED
#define PD_LOCKED
jailed
int jailed(struct ucred *cred)
Definition: kern_jail.c:3474
pr_flag_jailsys_size
const size_t pr_flag_jailsys_size
Definition: kern_jail.c:195
cpuset_rel
void cpuset_rel(struct cpuset *set)
Definition: kern_cpuset.c:164
sysctl_jail_default_allow
static int sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
Definition: kern_jail.c:4185
cpuset_setproc_update_set
int cpuset_setproc_update_set(struct proc *p, struct cpuset *set)
Definition: kern_cpuset.c:817
cpuset_create_root
int cpuset_create_root(struct prison *pr, struct cpuset **setp)
Definition: kern_cpuset.c:794
prison_remove_one
static void prison_remove_one(struct prison *pr)
Definition: kern_jail.c:2258
jailed_without_vnet
int jailed_without_vnet(struct ucred *cred)
Definition: kern_jail.c:3485
namei
int namei(struct nameidata *ndp)
Definition: vfs_lookup.c:135
getcreddomainname
void getcreddomainname(struct ucred *cred, char *buf, size_t size)
Definition: kern_jail.c:3517
prison0
struct prison prison0
Definition: kern_jail.c:99
crfree
void crfree(struct ucred *cr)
Definition: kern_prot.c:1835
SYSCTL_JAIL_PARAM_SUBNODE
SYSCTL_JAIL_PARAM_SUBNODE(allow, mount,"Jail mount/unmount permission flags")
jailsys_flags::new
unsigned new
Definition: kern_jail.c:182
prison_hold
void prison_hold(struct prison *pr)
Definition: kern_jail.c:2632
sysctl_handle_int
int sysctl_handle_int(SYSCTL_HANDLER_ARGS)
Definition: kern_sysctl.c:986
pr_allow_nonames
static char * pr_allow_nonames[]
Definition: kern_jail.c:212
prison_if
int prison_if(struct ucred *cred, struct sockaddr *sa)
Definition: kern_jail.c:3406
prison_racct_free_locked
static void prison_racct_free_locked(struct prison_racct *prr)
Definition: kern_jail.c:4465
kern_jail
int kern_jail(struct thread *td, struct jail *j)
Definition: kern_jail.c:340
prison_proc_free
void prison_proc_free(struct prison *pr)
Definition: kern_jail.c:2652
vfs_setopt_part
int vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
Definition: vfs_mount.c:1601
JAIL_DEFAULT_ALLOW
#define JAIL_DEFAULT_ALLOW
Definition: kern_jail.c:227
jailsys_flags::disable
unsigned disable
Definition: kern_jail.c:181
racct_destroy
void racct_destroy(struct racct **racctp)
Definition: kern_racct.c:1268
vfs_buildopts
int vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
Definition: vfs_mount.c:260
sys_jail
int sys_jail(struct thread *td, struct jail_args *uap)
Definition: kern_jail.c:291
crhold
struct ucred * crhold(struct ucred *cr)
Definition: kern_prot.c:1824
prison_canseemount
int prison_canseemount(struct ucred *cred, struct mount *mp)
Definition: kern_jail.c:3570
allprison_lock
struct sx allprison_lock
Definition: kern_jail.c:120
getcredhostuuid
void getcredhostuuid(struct ucred *cred, char *buf, size_t size)
Definition: kern_jail.c:3526
SYSCTL_PROC
SYSCTL_PROC(_security_jail, OID_AUTO, jailed, CTLTYPE_INT|CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, 0, sysctl_jail_jailed,"I","Process in jail?")
prison_free
void prison_free(struct prison *pr)
Definition: kern_jail.c:2529
free
void free(void *addr, struct malloc_type *mtp)
Definition: kern_malloc.c:554
PD_LIST_SLOCKED
#define PD_LIST_SLOCKED
SYSCTL_JAIL_PARAM_SYS_NODE
SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW,"Jail host info")
taskqueue_enqueue
int taskqueue_enqueue(struct taskqueue *queue, struct task *task)
Definition: subr_taskqueue.c:210
prison_racct_hold
void prison_racct_hold(struct prison_racct *prr)
Definition: kern_jail.c:4458
change_dir
int change_dir(struct vnode *vp, struct thread *td)
Definition: vfs_syscalls.c:941
sysctl_jail_param
int sysctl_jail_param(SYSCTL_HANDLER_ARGS)
Definition: kern_jail.c:4290
prison_racct_foreach
void prison_racct_foreach(void(*callback)(struct racct *racct, void *arg2, void *arg3), void *arg2, void *arg3)
Definition: kern_jail.c:4405
__FBSDID
__FBSDID("$BSDSUniX$")
prison_complete
static void prison_complete(void *context, int pending)
Definition: kern_jail.c:2537
change_root
int change_root(struct vnode *vp, struct thread *td)
Definition: vfs_syscalls.c:965
copyinuio
int copyinuio(struct iovec *iovp, u_int iovcnt, struct uio **uiop)
Definition: subr_uio.c:508
prison_path
static char * prison_path(struct prison *pr1, struct prison *pr2)
Definition: kern_jail.c:4008
mtx_init
void mtx_init(struct mtx *m, const char *name, const char *type, int opts)
Definition: kern_mutex.c:837
MTX_SYSINIT
MTX_SYSINIT(prison0,&prison0.pr_mtx,"jail mutex", MTX_DEF)
SYSCTL_NODE
static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,"Jails")
vrele
void vrele(struct vnode *vp)
Definition: vfs_subr.c:2416
DEFAULT_HOSTUUID
#define DEFAULT_HOSTUUID
Definition: kern_jail.c:78
JAIL_DEFAULT_ENFORCE_STATFS
#define JAIL_DEFAULT_ENFORCE_STATFS
Definition: kern_jail.c:228
jail_default_devfs_rsnum
static int jail_default_devfs_rsnum
Definition: kern_jail.c:232
prison_priv_check
int prison_priv_check(struct ucred *cred, int priv)
Definition: kern_jail.c:3655
PD_LIST_XLOCKED
#define PD_LIST_XLOCKED
kern_jail_get
int kern_jail_get(struct thread *td, struct uio *optuio, int flags)
Definition: kern_jail.c:1923
prison_ischild
int prison_ischild(struct prison *pr1, struct prison *pr2)
Definition: kern_jail.c:3461
PD_DEREF
#define PD_DEREF
_PR_IP_SADDRSEL
#define _PR_IP_SADDRSEL
Definition: kern_jail.c:94
crcopy
void crcopy(struct ucred *dest, struct ucred *src)
Definition: kern_prot.c:1882
prison_flag
int prison_flag(struct ucred *cred, unsigned flag)
Definition: kern_jail.c:2494
mtx_destroy
void mtx_destroy(struct mtx *m)
Definition: kern_mutex.c:884
jailsys_flags
Definition: kern_jail.c:179
prison_hold_locked
void prison_hold_locked(struct prison *pr)
Definition: kern_jail.c:2622
pr_allow_names_size
const size_t pr_allow_names_size
Definition: kern_jail.c:210
pr_flag_nonames
static char * pr_flag_nonames[]
Definition: kern_jail.c:168
prison_racct_find
struct prison_racct * prison_racct_find(const char *name)
Definition: kern_jail.c:4447
crget
struct ucred * crget(void)
Definition: kern_prot.c:1804
racct_create
void racct_create(struct racct **racctp)
Definition: kern_racct.c:1263
pr
static struct pollrec pr[POLL_LIST_LEN]
Definition: kern_poll.c:254
prison_name
char * prison_name(struct prison *pr1, struct prison *pr2)
Definition: kern_jail.c:3983
sys_jail_remove
int sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
Definition: kern_jail.c:2206
prison_find_name
struct prison * prison_find_name(struct prison *mypr, const char *name)
Definition: kern_jail.c:2458
sysctl_jail_vnet
static int sysctl_jail_vnet(SYSCTL_HANDLER_ARGS)
Definition: kern_jail.c:4154
prison_free_locked
void prison_free_locked(struct prison *pr)
Definition: kern_jail.c:2514
flag
int flag
Definition: vfs_syscalls.c:2159
getcredhostname
void getcredhostname(struct ucred *cred, char *buf, size_t size)
Definition: kern_jail.c:3502
do_jail_attach
static int do_jail_attach(struct thread *td, struct prison *pr)
Definition: kern_jail.c:2338
level
const struct cf_level * level
Definition: cpufreq_if.m:45
LIST_HEAD
LIST_HEAD(prison_racct)
Definition: kern_jail.c:123