The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_jail.c

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1999 Poul-Henning Kamp.
    3  * Copyright (c) 2008 Bjoern A. Zeeb.
    4  * Copyright (c) 2009 James Gritton.
    5  * All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  *
   16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   26  * SUCH DAMAGE.
   27  */
   28 
   29 #include <sys/cdefs.h>
   30 __FBSDID("$FreeBSD: releng/8.3/sys/kern/kern_jail.c 231346 2012-02-10 06:42:00Z bz $");
   31 
   32 #include "opt_compat.h"
   33 #include "opt_ddb.h"
   34 #include "opt_inet.h"
   35 #include "opt_inet6.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/types.h>
   39 #include <sys/kernel.h>
   40 #include <sys/systm.h>
   41 #include <sys/errno.h>
   42 #include <sys/sysproto.h>
   43 #include <sys/malloc.h>
   44 #include <sys/osd.h>
   45 #include <sys/priv.h>
   46 #include <sys/proc.h>
   47 #include <sys/taskqueue.h>
   48 #include <sys/fcntl.h>
   49 #include <sys/jail.h>
   50 #include <sys/lock.h>
   51 #include <sys/mutex.h>
   52 #include <sys/sx.h>
   53 #include <sys/sysent.h>
   54 #include <sys/namei.h>
   55 #include <sys/mount.h>
   56 #include <sys/queue.h>
   57 #include <sys/socket.h>
   58 #include <sys/syscallsubr.h>
   59 #include <sys/sysctl.h>
   60 #include <sys/vnode.h>
   61 
   62 #include <net/if.h>
   63 #include <net/vnet.h>
   64 
   65 #include <netinet/in.h>
   66 
   67 #ifdef DDB
   68 #include <ddb/ddb.h>
   69 #ifdef INET6
   70 #include <netinet6/in6_var.h>
   71 #endif /* INET6 */
   72 #endif /* DDB */
   73 
   74 #include <security/mac/mac_framework.h>
   75 
   76 #define DEFAULT_HOSTUUID        "00000000-0000-0000-0000-000000000000"
   77 
   78 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
   79 
   80 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
   81 #ifdef INET
   82 #ifdef INET6
   83 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
   84 #else
   85 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
   86 #endif
   87 #else /* !INET */
   88 #ifdef INET6
   89 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
   90 #else
   91 #define _PR_IP_SADDRSEL 0
   92 #endif
   93 #endif
   94 
   95 /* prison0 describes what is "real" about the system. */
   96 struct prison prison0 = {
   97         .pr_id          = 0,
   98         .pr_name        = "",
   99         .pr_ref         = 1,
  100         .pr_uref        = 1,
  101         .pr_path        = "/",
  102         .pr_securelevel = -1,
  103         .pr_childmax    = JAIL_MAX,
  104         .pr_hostuuid    = DEFAULT_HOSTUUID,
  105         .pr_children    = LIST_HEAD_INITIALIZER(prison0.pr_children),
  106 #ifdef VIMAGE
  107         .pr_flags       = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
  108 #else
  109         .pr_flags       = PR_HOST|_PR_IP_SADDRSEL,
  110 #endif
  111         .pr_allow       = PR_ALLOW_ALL,
  112 };
  113 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
  114 
  115 /* allprison and lastprid are protected by allprison_lock. */
  116 struct  sx allprison_lock;
  117 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
  118 struct  prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
  119 int     lastprid = 0;
  120 
  121 static int do_jail_attach(struct thread *td, struct prison *pr);
  122 static void prison_complete(void *context, int pending);
  123 static void prison_deref(struct prison *pr, int flags);
  124 static char *prison_path(struct prison *pr1, struct prison *pr2);
  125 static void prison_remove_one(struct prison *pr);
  126 #ifdef INET
  127 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
  128 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
  129 #endif
  130 #ifdef INET6
  131 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
  132 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
  133 #endif
  134 
  135 /* Flags for prison_deref */
  136 #define PD_DEREF        0x01
  137 #define PD_DEUREF       0x02
  138 #define PD_LOCKED       0x04
  139 #define PD_LIST_SLOCKED 0x08
  140 #define PD_LIST_XLOCKED 0x10
  141 
  142 /*
  143  * Parameter names corresponding to PR_* flag values.  Size values are for kvm
  144  * as we cannot figure out the size of a sparse array, or an array without a
  145  * terminating entry.
  146  */
  147 static char *pr_flag_names[] = {
  148         [0] = "persist",
  149 #ifdef INET
  150         [7] = "ip4.saddrsel",
  151 #endif
  152 #ifdef INET6
  153         [8] = "ip6.saddrsel",
  154 #endif
  155 };
  156 const size_t pr_flag_names_size = sizeof(pr_flag_names);
  157 
  158 static char *pr_flag_nonames[] = {
  159         [0] = "nopersist",
  160 #ifdef INET
  161         [7] = "ip4.nosaddrsel",
  162 #endif
  163 #ifdef INET6
  164         [8] = "ip6.nosaddrsel",
  165 #endif
  166 };
  167 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
  168 
  169 struct jailsys_flags {
  170         const char      *name;
  171         unsigned         disable;
  172         unsigned         new;
  173 } pr_flag_jailsys[] = {
  174         { "host", 0, PR_HOST },
  175 #ifdef VIMAGE
  176         { "vnet", 0, PR_VNET },
  177 #endif
  178 #ifdef INET
  179         { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
  180 #endif
  181 #ifdef INET6
  182         { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
  183 #endif
  184 };
  185 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
  186 
  187 static char *pr_allow_names[] = {
  188         "allow.set_hostname",
  189         "allow.sysvipc",
  190         "allow.raw_sockets",
  191         "allow.chflags",
  192         "allow.mount",
  193         "allow.quotas",
  194         "allow.socket_af",
  195 };
  196 const size_t pr_allow_names_size = sizeof(pr_allow_names);
  197 
  198 static char *pr_allow_nonames[] = {
  199         "allow.noset_hostname",
  200         "allow.nosysvipc",
  201         "allow.noraw_sockets",
  202         "allow.nochflags",
  203         "allow.nomount",
  204         "allow.noquotas",
  205         "allow.nosocket_af",
  206 };
  207 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
  208 
  209 #define JAIL_DEFAULT_ALLOW              PR_ALLOW_SET_HOSTNAME
  210 #define JAIL_DEFAULT_ENFORCE_STATFS     2
  211 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
  212 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
  213 #if defined(INET) || defined(INET6)
  214 static unsigned jail_max_af_ips = 255;
  215 #endif
  216 
  217 #ifdef INET
  218 static int
  219 qcmp_v4(const void *ip1, const void *ip2)
  220 {
  221         in_addr_t iaa, iab;
  222 
  223         /*
  224          * We need to compare in HBO here to get the list sorted as expected
  225          * by the result of the code.  Sorting NBO addresses gives you
  226          * interesting results.  If you do not understand, do not try.
  227          */
  228         iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
  229         iab = ntohl(((const struct in_addr *)ip2)->s_addr);
  230 
  231         /*
  232          * Do not simply return the difference of the two numbers, the int is
  233          * not wide enough.
  234          */
  235         if (iaa > iab)
  236                 return (1);
  237         else if (iaa < iab)
  238                 return (-1);
  239         else
  240                 return (0);
  241 }
  242 #endif
  243 
  244 #ifdef INET6
  245 static int
  246 qcmp_v6(const void *ip1, const void *ip2)
  247 {
  248         const struct in6_addr *ia6a, *ia6b;
  249         int i, rc;
  250 
  251         ia6a = (const struct in6_addr *)ip1;
  252         ia6b = (const struct in6_addr *)ip2;
  253 
  254         rc = 0;
  255         for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
  256                 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
  257                         rc = 1;
  258                 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
  259                         rc = -1;
  260         }
  261         return (rc);
  262 }
  263 #endif
  264 
  265 /*
  266  * struct jail_args {
  267  *      struct jail *jail;
  268  * };
  269  */
  270 int
  271 jail(struct thread *td, struct jail_args *uap)
  272 {
  273         uint32_t version;
  274         int error;
  275         struct jail j;
  276 
  277         error = copyin(uap->jail, &version, sizeof(uint32_t));
  278         if (error)
  279                 return (error);
  280 
  281         switch (version) {
  282         case 0:
  283         {
  284                 struct jail_v0 j0;
  285 
  286                 /* FreeBSD single IPv4 jails. */
  287                 bzero(&j, sizeof(struct jail));
  288                 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
  289                 if (error)
  290                         return (error);
  291                 j.version = j0.version;
  292                 j.path = j0.path;
  293                 j.hostname = j0.hostname;
  294                 j.ip4s = j0.ip_number;
  295                 break;
  296         }
  297 
  298         case 1:
  299                 /*
  300                  * Version 1 was used by multi-IPv4 jail implementations
  301                  * that never made it into the official kernel.
  302                  */
  303                 return (EINVAL);
  304 
  305         case 2: /* JAIL_API_VERSION */
  306                 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
  307                 error = copyin(uap->jail, &j, sizeof(struct jail));
  308                 if (error)
  309                         return (error);
  310                 break;
  311 
  312         default:
  313                 /* Sci-Fi jails are not supported, sorry. */
  314                 return (EINVAL);
  315         }
  316         return (kern_jail(td, &j));
  317 }
  318 
  319 int
  320 kern_jail(struct thread *td, struct jail *j)
  321 {
  322         struct iovec optiov[2 * (4
  323                             + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
  324 #ifdef INET
  325                             + 1
  326 #endif
  327 #ifdef INET6
  328                             + 1
  329 #endif
  330                             )];
  331         struct uio opt;
  332         char *u_path, *u_hostname, *u_name;
  333 #ifdef INET
  334         uint32_t ip4s;
  335         struct in_addr *u_ip4;
  336 #endif
  337 #ifdef INET6
  338         struct in6_addr *u_ip6;
  339 #endif
  340         size_t tmplen;
  341         int error, enforce_statfs, fi;
  342 
  343         bzero(&optiov, sizeof(optiov));
  344         opt.uio_iov = optiov;
  345         opt.uio_iovcnt = 0;
  346         opt.uio_offset = -1;
  347         opt.uio_resid = -1;
  348         opt.uio_segflg = UIO_SYSSPACE;
  349         opt.uio_rw = UIO_READ;
  350         opt.uio_td = td;
  351 
  352         /* Set permissions for top-level jails from sysctls. */
  353         if (!jailed(td->td_ucred)) {
  354                 for (fi = 0; fi < sizeof(pr_allow_names) /
  355                      sizeof(pr_allow_names[0]); fi++) {
  356                         optiov[opt.uio_iovcnt].iov_base =
  357                             (jail_default_allow & (1 << fi))
  358                             ? pr_allow_names[fi] : pr_allow_nonames[fi];
  359                         optiov[opt.uio_iovcnt].iov_len =
  360                             strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
  361                         opt.uio_iovcnt += 2;
  362                 }
  363                 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
  364                 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
  365                 opt.uio_iovcnt++;
  366                 enforce_statfs = jail_default_enforce_statfs;
  367                 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
  368                 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
  369                 opt.uio_iovcnt++;
  370         }
  371 
  372         tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
  373 #ifdef INET
  374         ip4s = (j->version == 0) ? 1 : j->ip4s;
  375         if (ip4s > jail_max_af_ips)
  376                 return (EINVAL);
  377         tmplen += ip4s * sizeof(struct in_addr);
  378 #else
  379         if (j->ip4s > 0)
  380                 return (EINVAL);
  381 #endif
  382 #ifdef INET6
  383         if (j->ip6s > jail_max_af_ips)
  384                 return (EINVAL);
  385         tmplen += j->ip6s * sizeof(struct in6_addr);
  386 #else
  387         if (j->ip6s > 0)
  388                 return (EINVAL);
  389 #endif
  390         u_path = malloc(tmplen, M_TEMP, M_WAITOK);
  391         u_hostname = u_path + MAXPATHLEN;
  392         u_name = u_hostname + MAXHOSTNAMELEN;
  393 #ifdef INET
  394         u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
  395 #endif
  396 #ifdef INET6
  397 #ifdef INET
  398         u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
  399 #else
  400         u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
  401 #endif
  402 #endif
  403         optiov[opt.uio_iovcnt].iov_base = "path";
  404         optiov[opt.uio_iovcnt].iov_len = sizeof("path");
  405         opt.uio_iovcnt++;
  406         optiov[opt.uio_iovcnt].iov_base = u_path;
  407         error = copyinstr(j->path, u_path, MAXPATHLEN,
  408             &optiov[opt.uio_iovcnt].iov_len);
  409         if (error) {
  410                 free(u_path, M_TEMP);
  411                 return (error);
  412         }
  413         opt.uio_iovcnt++;
  414         optiov[opt.uio_iovcnt].iov_base = "host.hostname";
  415         optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
  416         opt.uio_iovcnt++;
  417         optiov[opt.uio_iovcnt].iov_base = u_hostname;
  418         error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
  419             &optiov[opt.uio_iovcnt].iov_len);
  420         if (error) {
  421                 free(u_path, M_TEMP);
  422                 return (error);
  423         }
  424         opt.uio_iovcnt++;
  425         if (j->jailname != NULL) {
  426                 optiov[opt.uio_iovcnt].iov_base = "name";
  427                 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
  428                 opt.uio_iovcnt++;
  429                 optiov[opt.uio_iovcnt].iov_base = u_name;
  430                 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
  431                     &optiov[opt.uio_iovcnt].iov_len);
  432                 if (error) {
  433                         free(u_path, M_TEMP);
  434                         return (error);
  435                 }
  436                 opt.uio_iovcnt++;
  437         }
  438 #ifdef INET
  439         optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
  440         optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
  441         opt.uio_iovcnt++;
  442         optiov[opt.uio_iovcnt].iov_base = u_ip4;
  443         optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
  444         if (j->version == 0)
  445                 u_ip4->s_addr = j->ip4s;
  446         else {
  447                 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
  448                 if (error) {
  449                         free(u_path, M_TEMP);
  450                         return (error);
  451                 }
  452         }
  453         opt.uio_iovcnt++;
  454 #endif
  455 #ifdef INET6
  456         optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
  457         optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
  458         opt.uio_iovcnt++;
  459         optiov[opt.uio_iovcnt].iov_base = u_ip6;
  460         optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
  461         error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
  462         if (error) {
  463                 free(u_path, M_TEMP);
  464                 return (error);
  465         }
  466         opt.uio_iovcnt++;
  467 #endif
  468         KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
  469             ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
  470         error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
  471         free(u_path, M_TEMP);
  472         return (error);
  473 }
  474 
  475 
  476 /*
  477  * struct jail_set_args {
  478  *      struct iovec *iovp;
  479  *      unsigned int iovcnt;
  480  *      int flags;
  481  * };
  482  */
  483 int
  484 jail_set(struct thread *td, struct jail_set_args *uap)
  485 {
  486         struct uio *auio;
  487         int error;
  488 
  489         /* Check that we have an even number of iovecs. */
  490         if (uap->iovcnt & 1)
  491                 return (EINVAL);
  492 
  493         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
  494         if (error)
  495                 return (error);
  496         error = kern_jail_set(td, auio, uap->flags);
  497         free(auio, M_IOV);
  498         return (error);
  499 }
  500 
  501 int
  502 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
  503 {
  504         struct nameidata nd;
  505 #ifdef INET
  506         struct in_addr *ip4;
  507 #endif
  508 #ifdef INET6
  509         struct in6_addr *ip6;
  510 #endif
  511         struct vfsopt *opt;
  512         struct vfsoptlist *opts;
  513         struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
  514         struct vnode *root;
  515         char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
  516 #if defined(INET) || defined(INET6)
  517         struct prison *tppr;
  518         void *op;
  519 #endif
  520         unsigned long hid;
  521         size_t namelen, onamelen;
  522         int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
  523         int gotchildmax, gotenforce, gothid, gotslevel;
  524         int fi, jid, jsys, len, level;
  525         int childmax, slevel, vfslocked;
  526 #if defined(INET) || defined(INET6)
  527         int ii, ij;
  528 #endif
  529 #ifdef INET
  530         int ip4s, redo_ip4;
  531 #endif
  532 #ifdef INET6
  533         int ip6s, redo_ip6;
  534 #endif
  535         unsigned pr_flags, ch_flags;
  536         unsigned pr_allow, ch_allow, tallow;
  537         char numbuf[12];
  538 
  539         error = priv_check(td, PRIV_JAIL_SET);
  540         if (!error && (flags & JAIL_ATTACH))
  541                 error = priv_check(td, PRIV_JAIL_ATTACH);
  542         if (error)
  543                 return (error);
  544         mypr = ppr = td->td_ucred->cr_prison;
  545         if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
  546                 return (EPERM);
  547         if (flags & ~JAIL_SET_MASK)
  548                 return (EINVAL);
  549 
  550         /*
  551          * Check all the parameters before committing to anything.  Not all
  552          * errors can be caught early, but we may as well try.  Also, this
  553          * takes care of some expensive stuff (path lookup) before getting
  554          * the allprison lock.
  555          *
  556          * XXX Jails are not filesystems, and jail parameters are not mount
  557          *     options.  But it makes more sense to re-use the vfsopt code
  558          *     than duplicate it under a different name.
  559          */
  560         error = vfs_buildopts(optuio, &opts);
  561         if (error)
  562                 return (error);
  563 #ifdef INET
  564         ip4 = NULL;
  565 #endif
  566 #ifdef INET6
  567         ip6 = NULL;
  568 #endif
  569 
  570         error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
  571         if (error == ENOENT)
  572                 jid = 0;
  573         else if (error != 0)
  574                 goto done_free;
  575 
  576         error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
  577         if (error == ENOENT)
  578                 gotslevel = 0;
  579         else if (error != 0)
  580                 goto done_free;
  581         else
  582                 gotslevel = 1;
  583 
  584         error =
  585             vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
  586         if (error == ENOENT)
  587                 gotchildmax = 0;
  588         else if (error != 0)
  589                 goto done_free;
  590         else
  591                 gotchildmax = 1;
  592 
  593         error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
  594         if (error == ENOENT)
  595                 gotenforce = 0;
  596         else if (error != 0)
  597                 goto done_free;
  598         else if (enforce < 0 || enforce > 2) {
  599                 error = EINVAL;
  600                 goto done_free;
  601         } else
  602                 gotenforce = 1;
  603 
  604         pr_flags = ch_flags = 0;
  605         for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
  606             fi++) {
  607                 if (pr_flag_names[fi] == NULL)
  608                         continue;
  609                 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
  610                 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
  611         }
  612         ch_flags |= pr_flags;
  613         for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
  614             fi++) {
  615                 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
  616                     sizeof(jsys));
  617                 if (error == ENOENT)
  618                         continue;
  619                 if (error != 0)
  620                         goto done_free;
  621                 switch (jsys) {
  622                 case JAIL_SYS_DISABLE:
  623                         if (!pr_flag_jailsys[fi].disable) {
  624                                 error = EINVAL;
  625                                 goto done_free;
  626                         }
  627                         pr_flags |= pr_flag_jailsys[fi].disable;
  628                         break;
  629                 case JAIL_SYS_NEW:
  630                         pr_flags |= pr_flag_jailsys[fi].new;
  631                         break;
  632                 case JAIL_SYS_INHERIT:
  633                         break;
  634                 default:
  635                         error = EINVAL;
  636                         goto done_free;
  637                 }
  638                 ch_flags |=
  639                     pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
  640         }
  641         if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
  642             && !(pr_flags & PR_PERSIST)) {
  643                 error = EINVAL;
  644                 vfs_opterror(opts, "new jail must persist or attach");
  645                 goto done_errmsg;
  646         }
  647 #ifdef VIMAGE
  648         if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
  649                 error = EINVAL;
  650                 vfs_opterror(opts, "vnet cannot be changed after creation");
  651                 goto done_errmsg;
  652         }
  653 #endif
  654 #ifdef INET
  655         if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
  656                 error = EINVAL;
  657                 vfs_opterror(opts, "ip4 cannot be changed after creation");
  658                 goto done_errmsg;
  659         }
  660 #endif
  661 #ifdef INET6
  662         if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
  663                 error = EINVAL;
  664                 vfs_opterror(opts, "ip6 cannot be changed after creation");
  665                 goto done_errmsg;
  666         }
  667 #endif
  668 
  669         pr_allow = ch_allow = 0;
  670         for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
  671             fi++) {
  672                 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
  673                 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
  674         }
  675         ch_allow |= pr_allow;
  676 
  677         error = vfs_getopt(opts, "name", (void **)&name, &len);
  678         if (error == ENOENT)
  679                 name = NULL;
  680         else if (error != 0)
  681                 goto done_free;
  682         else {
  683                 if (len == 0 || name[len - 1] != '\0') {
  684                         error = EINVAL;
  685                         goto done_free;
  686                 }
  687                 if (len > MAXHOSTNAMELEN) {
  688                         error = ENAMETOOLONG;
  689                         goto done_free;
  690                 }
  691         }
  692 
  693         error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
  694         if (error == ENOENT)
  695                 host = NULL;
  696         else if (error != 0)
  697                 goto done_free;
  698         else {
  699                 ch_flags |= PR_HOST;
  700                 pr_flags |= PR_HOST;
  701                 if (len == 0 || host[len - 1] != '\0') {
  702                         error = EINVAL;
  703                         goto done_free;
  704                 }
  705                 if (len > MAXHOSTNAMELEN) {
  706                         error = ENAMETOOLONG;
  707                         goto done_free;
  708                 }
  709         }
  710 
  711         error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
  712         if (error == ENOENT)
  713                 domain = NULL;
  714         else if (error != 0)
  715                 goto done_free;
  716         else {
  717                 ch_flags |= PR_HOST;
  718                 pr_flags |= PR_HOST;
  719                 if (len == 0 || domain[len - 1] != '\0') {
  720                         error = EINVAL;
  721                         goto done_free;
  722                 }
  723                 if (len > MAXHOSTNAMELEN) {
  724                         error = ENAMETOOLONG;
  725                         goto done_free;
  726                 }
  727         }
  728 
  729         error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
  730         if (error == ENOENT)
  731                 uuid = NULL;
  732         else if (error != 0)
  733                 goto done_free;
  734         else {
  735                 ch_flags |= PR_HOST;
  736                 pr_flags |= PR_HOST;
  737                 if (len == 0 || uuid[len - 1] != '\0') {
  738                         error = EINVAL;
  739                         goto done_free;
  740                 }
  741                 if (len > HOSTUUIDLEN) {
  742                         error = ENAMETOOLONG;
  743                         goto done_free;
  744                 }
  745         }
  746 
  747 #ifdef COMPAT_FREEBSD32
  748         if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
  749                 uint32_t hid32;
  750 
  751                 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
  752                 hid = hid32;
  753         } else
  754 #endif
  755                 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
  756         if (error == ENOENT)
  757                 gothid = 0;
  758         else if (error != 0)
  759                 goto done_free;
  760         else {
  761                 gothid = 1;
  762                 ch_flags |= PR_HOST;
  763                 pr_flags |= PR_HOST;
  764         }
  765 
  766 #ifdef INET
  767         error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
  768         if (error == ENOENT)
  769                 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
  770         else if (error != 0)
  771                 goto done_free;
  772         else if (ip4s & (sizeof(*ip4) - 1)) {
  773                 error = EINVAL;
  774                 goto done_free;
  775         } else {
  776                 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
  777                 if (ip4s == 0)
  778                         pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
  779                 else {
  780                         pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
  781                         ip4s /= sizeof(*ip4);
  782                         if (ip4s > jail_max_af_ips) {
  783                                 error = EINVAL;
  784                                 vfs_opterror(opts, "too many IPv4 addresses");
  785                                 goto done_errmsg;
  786                         }
  787                         ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
  788                         bcopy(op, ip4, ip4s * sizeof(*ip4));
  789                         /*
  790                          * IP addresses are all sorted but ip[0] to preserve
  791                          * the primary IP address as given from userland.
  792                          * This special IP is used for unbound outgoing
  793                          * connections as well for "loopback" traffic in case
  794                          * source address selection cannot find any more fitting
  795                          * address to connect from.
  796                          */
  797                         if (ip4s > 1)
  798                                 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
  799                         /*
  800                          * Check for duplicate addresses and do some simple
  801                          * zero and broadcast checks. If users give other bogus
  802                          * addresses it is their problem.
  803                          *
  804                          * We do not have to care about byte order for these
  805                          * checks so we will do them in NBO.
  806                          */
  807                         for (ii = 0; ii < ip4s; ii++) {
  808                                 if (ip4[ii].s_addr == INADDR_ANY ||
  809                                     ip4[ii].s_addr == INADDR_BROADCAST) {
  810                                         error = EINVAL;
  811                                         goto done_free;
  812                                 }
  813                                 if ((ii+1) < ip4s &&
  814                                     (ip4[0].s_addr == ip4[ii+1].s_addr ||
  815                                      ip4[ii].s_addr == ip4[ii+1].s_addr)) {
  816                                         error = EINVAL;
  817                                         goto done_free;
  818                                 }
  819                         }
  820                 }
  821         }
  822 #endif
  823 
  824 #ifdef INET6
  825         error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
  826         if (error == ENOENT)
  827                 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
  828         else if (error != 0)
  829                 goto done_free;
  830         else if (ip6s & (sizeof(*ip6) - 1)) {
  831                 error = EINVAL;
  832                 goto done_free;
  833         } else {
  834                 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
  835                 if (ip6s == 0)
  836                         pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
  837                 else {
  838                         pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
  839                         ip6s /= sizeof(*ip6);
  840                         if (ip6s > jail_max_af_ips) {
  841                                 error = EINVAL;
  842                                 vfs_opterror(opts, "too many IPv6 addresses");
  843                                 goto done_errmsg;
  844                         }
  845                         ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
  846                         bcopy(op, ip6, ip6s * sizeof(*ip6));
  847                         if (ip6s > 1)
  848                                 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
  849                         for (ii = 0; ii < ip6s; ii++) {
  850                                 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
  851                                         error = EINVAL;
  852                                         goto done_free;
  853                                 }
  854                                 if ((ii+1) < ip6s &&
  855                                     (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
  856                                      IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
  857                                 {
  858                                         error = EINVAL;
  859                                         goto done_free;
  860                                 }
  861                         }
  862                 }
  863         }
  864 #endif
  865 
  866 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
  867         if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
  868                 error = EINVAL;
  869                 vfs_opterror(opts,
  870                     "vnet jails cannot have IP address restrictions");
  871                 goto done_errmsg;
  872         }
  873 #endif
  874 
  875         root = NULL;
  876         error = vfs_getopt(opts, "path", (void **)&path, &len);
  877         if (error == ENOENT)
  878                 path = NULL;
  879         else if (error != 0)
  880                 goto done_free;
  881         else {
  882                 if (flags & JAIL_UPDATE) {
  883                         error = EINVAL;
  884                         vfs_opterror(opts,
  885                             "path cannot be changed after creation");
  886                         goto done_errmsg;
  887                 }
  888                 if (len == 0 || path[len - 1] != '\0') {
  889                         error = EINVAL;
  890                         goto done_free;
  891                 }
  892                 if (len < 2 || (len == 2 && path[0] == '/'))
  893                         path = NULL;
  894                 else {
  895                         /* Leave room for a real-root full pathname. */
  896                         if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
  897                             ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
  898                                 error = ENAMETOOLONG;
  899                                 goto done_free;
  900                         }
  901                         NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
  902                             path, td);
  903                         error = namei(&nd);
  904                         if (error)
  905                                 goto done_free;
  906                         vfslocked = NDHASGIANT(&nd);
  907                         root = nd.ni_vp;
  908                         NDFREE(&nd, NDF_ONLY_PNBUF);
  909                         if (root->v_type != VDIR) {
  910                                 error = ENOTDIR;
  911                                 vrele(root);
  912                                 VFS_UNLOCK_GIANT(vfslocked);
  913                                 goto done_free;
  914                         }
  915                         VFS_UNLOCK_GIANT(vfslocked);
  916                 }
  917         }
  918 
  919         /*
  920          * Grab the allprison lock before letting modules check their
  921          * parameters.  Once we have it, do not let go so we'll have a
  922          * consistent view of the OSD list.
  923          */
  924         sx_xlock(&allprison_lock);
  925         error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
  926         if (error)
  927                 goto done_unlock_list;
  928 
  929         /* By now, all parameters should have been noted. */
  930         TAILQ_FOREACH(opt, opts, link) {
  931                 if (!opt->seen && strcmp(opt->name, "errmsg")) {
  932                         error = EINVAL;
  933                         vfs_opterror(opts, "unknown parameter: %s", opt->name);
  934                         goto done_unlock_list;
  935                 }
  936         }
  937 
  938         /*
  939          * See if we are creating a new record or updating an existing one.
  940          * This abuses the file error codes ENOENT and EEXIST.
  941          */
  942         cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
  943         if (!cuflags) {
  944                 error = EINVAL;
  945                 vfs_opterror(opts, "no valid operation (create or update)");
  946                 goto done_unlock_list;
  947         }
  948         pr = NULL;
  949         namelc = NULL;
  950         if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
  951                 namelc = strrchr(name, '.');
  952                 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
  953                 if (*p != '\0')
  954                         jid = 0;
  955         }
  956         if (jid != 0) {
  957                 /*
  958                  * See if a requested jid already exists.  There is an
  959                  * information leak here if the jid exists but is not within
  960                  * the caller's jail hierarchy.  Jail creators will get EEXIST
  961                  * even though they cannot see the jail, and CREATE | UPDATE
  962                  * will return ENOENT which is not normally a valid error.
  963                  */
  964                 if (jid < 0) {
  965                         error = EINVAL;
  966                         vfs_opterror(opts, "negative jid");
  967                         goto done_unlock_list;
  968                 }
  969                 pr = prison_find(jid);
  970                 if (pr != NULL) {
  971                         ppr = pr->pr_parent;
  972                         /* Create: jid must not exist. */
  973                         if (cuflags == JAIL_CREATE) {
  974                                 mtx_unlock(&pr->pr_mtx);
  975                                 error = EEXIST;
  976                                 vfs_opterror(opts, "jail %d already exists",
  977                                     jid);
  978                                 goto done_unlock_list;
  979                         }
  980                         if (!prison_ischild(mypr, pr)) {
  981                                 mtx_unlock(&pr->pr_mtx);
  982                                 pr = NULL;
  983                         } else if (pr->pr_uref == 0) {
  984                                 if (!(flags & JAIL_DYING)) {
  985                                         mtx_unlock(&pr->pr_mtx);
  986                                         error = ENOENT;
  987                                         vfs_opterror(opts, "jail %d is dying",
  988                                             jid);
  989                                         goto done_unlock_list;
  990                                 } else if ((flags & JAIL_ATTACH) ||
  991                                     (pr_flags & PR_PERSIST)) {
  992                                         /*
  993                                          * A dying jail might be resurrected
  994                                          * (via attach or persist), but first
  995                                          * it must determine if another jail
  996                                          * has claimed its name.  Accomplish
  997                                          * this by implicitly re-setting the
  998                                          * name.
  999                                          */
 1000                                         if (name == NULL)
 1001                                                 name = prison_name(mypr, pr);
 1002                                 }
 1003                         }
 1004                 }
 1005                 if (pr == NULL) {
 1006                         /* Update: jid must exist. */
 1007                         if (cuflags == JAIL_UPDATE) {
 1008                                 error = ENOENT;
 1009                                 vfs_opterror(opts, "jail %d not found", jid);
 1010                                 goto done_unlock_list;
 1011                         }
 1012                 }
 1013         }
 1014         /*
 1015          * If the caller provided a name, look for a jail by that name.
 1016          * This has different semantics for creates and updates keyed by jid
 1017          * (where the name must not already exist in a different jail),
 1018          * and updates keyed by the name itself (where the name must exist
 1019          * because that is the jail being updated).
 1020          */
 1021         if (name != NULL) {
 1022                 namelc = strrchr(name, '.');
 1023                 if (namelc == NULL)
 1024                         namelc = name;
 1025                 else {
 1026                         /*
 1027                          * This is a hierarchical name.  Split it into the
 1028                          * parent and child names, and make sure the parent
 1029                          * exists or matches an already found jail.
 1030                          */
 1031                         *namelc = '\0';
 1032                         if (pr != NULL) {
 1033                                 if (strncmp(name, ppr->pr_name, namelc - name)
 1034                                     || ppr->pr_name[namelc - name] != '\0') {
 1035                                         mtx_unlock(&pr->pr_mtx);
 1036                                         error = EINVAL;
 1037                                         vfs_opterror(opts,
 1038                                             "cannot change jail's parent");
 1039                                         goto done_unlock_list;
 1040                                 }
 1041                         } else {
 1042                                 ppr = prison_find_name(mypr, name);
 1043                                 if (ppr == NULL) {
 1044                                         error = ENOENT;
 1045                                         vfs_opterror(opts,
 1046                                             "jail \"%s\" not found", name);
 1047                                         goto done_unlock_list;
 1048                                 }
 1049                                 mtx_unlock(&ppr->pr_mtx);
 1050                         }
 1051                         name = ++namelc;
 1052                 }
 1053                 if (name[0] != '\0') {
 1054                         namelen =
 1055                             (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
 1056  name_again:
 1057                         deadpr = NULL;
 1058                         FOREACH_PRISON_CHILD(ppr, tpr) {
 1059                                 if (tpr != pr && tpr->pr_ref > 0 &&
 1060                                     !strcmp(tpr->pr_name + namelen, name)) {
 1061                                         if (pr == NULL &&
 1062                                             cuflags != JAIL_CREATE) {
 1063                                                 mtx_lock(&tpr->pr_mtx);
 1064                                                 if (tpr->pr_ref > 0) {
 1065                                                         /*
 1066                                                          * Use this jail
 1067                                                          * for updates.
 1068                                                          */
 1069                                                         if (tpr->pr_uref > 0) {
 1070                                                                 pr = tpr;
 1071                                                                 break;
 1072                                                         }
 1073                                                         deadpr = tpr;
 1074                                                 }
 1075                                                 mtx_unlock(&tpr->pr_mtx);
 1076                                         } else if (tpr->pr_uref > 0) {
 1077                                                 /*
 1078                                                  * Create, or update(jid):
 1079                                                  * name must not exist in an
 1080                                                  * active sibling jail.
 1081                                                  */
 1082                                                 error = EEXIST;
 1083                                                 if (pr != NULL)
 1084                                                         mtx_unlock(&pr->pr_mtx);
 1085                                                 vfs_opterror(opts,
 1086                                                    "jail \"%s\" already exists",
 1087                                                    name);
 1088                                                 goto done_unlock_list;
 1089                                         }
 1090                                 }
 1091                         }
 1092                         /* If no active jail is found, use a dying one. */
 1093                         if (deadpr != NULL && pr == NULL) {
 1094                                 if (flags & JAIL_DYING) {
 1095                                         mtx_lock(&deadpr->pr_mtx);
 1096                                         if (deadpr->pr_ref == 0) {
 1097                                                 mtx_unlock(&deadpr->pr_mtx);
 1098                                                 goto name_again;
 1099                                         }
 1100                                         pr = deadpr;
 1101                                 } else if (cuflags == JAIL_UPDATE) {
 1102                                         error = ENOENT;
 1103                                         vfs_opterror(opts,
 1104                                             "jail \"%s\" is dying", name);
 1105                                         goto done_unlock_list;
 1106                                 }
 1107                         }
 1108                         /* Update: name must exist if no jid. */
 1109                         else if (cuflags == JAIL_UPDATE && pr == NULL) {
 1110                                 error = ENOENT;
 1111                                 vfs_opterror(opts, "jail \"%s\" not found",
 1112                                     name);
 1113                                 goto done_unlock_list;
 1114                         }
 1115                 }
 1116         }
 1117         /* Update: must provide a jid or name. */
 1118         else if (cuflags == JAIL_UPDATE && pr == NULL) {
 1119                 error = ENOENT;
 1120                 vfs_opterror(opts, "update specified no jail");
 1121                 goto done_unlock_list;
 1122         }
 1123 
 1124         /* If there's no prison to update, create a new one and link it in. */
 1125         if (pr == NULL) {
 1126                 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
 1127                         if (tpr->pr_childcount >= tpr->pr_childmax) {
 1128                                 error = EPERM;
 1129                                 vfs_opterror(opts, "prison limit exceeded");
 1130                                 goto done_unlock_list;
 1131                         }
 1132                 created = 1;
 1133                 mtx_lock(&ppr->pr_mtx);
 1134                 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
 1135                         mtx_unlock(&ppr->pr_mtx);
 1136                         error = ENOENT;
 1137                         vfs_opterror(opts, "parent jail went away!");
 1138                         goto done_unlock_list;
 1139                 }
 1140                 ppr->pr_ref++;
 1141                 ppr->pr_uref++;
 1142                 mtx_unlock(&ppr->pr_mtx);
 1143                 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
 1144                 if (jid == 0) {
 1145                         /* Find the next free jid. */
 1146                         jid = lastprid + 1;
 1147  findnext:
 1148                         if (jid == JAIL_MAX)
 1149                                 jid = 1;
 1150                         TAILQ_FOREACH(tpr, &allprison, pr_list) {
 1151                                 if (tpr->pr_id < jid)
 1152                                         continue;
 1153                                 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
 1154                                         TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
 1155                                         break;
 1156                                 }
 1157                                 if (jid == lastprid) {
 1158                                         error = EAGAIN;
 1159                                         vfs_opterror(opts,
 1160                                             "no available jail IDs");
 1161                                         free(pr, M_PRISON);
 1162                                         prison_deref(ppr, PD_DEREF |
 1163                                             PD_DEUREF | PD_LIST_XLOCKED);
 1164                                         goto done_releroot;
 1165                                 }
 1166                                 jid++;
 1167                                 goto findnext;
 1168                         }
 1169                         lastprid = jid;
 1170                 } else {
 1171                         /*
 1172                          * The jail already has a jid (that did not yet exist),
 1173                          * so just find where to insert it.
 1174                          */
 1175                         TAILQ_FOREACH(tpr, &allprison, pr_list)
 1176                                 if (tpr->pr_id >= jid) {
 1177                                         TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
 1178                                         break;
 1179                                 }
 1180                 }
 1181                 if (tpr == NULL)
 1182                         TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
 1183                 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
 1184                 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
 1185                         tpr->pr_childcount++;
 1186 
 1187                 pr->pr_parent = ppr;
 1188                 pr->pr_id = jid;
 1189 
 1190                 /* Set some default values, and inherit some from the parent. */
 1191                 if (name == NULL)
 1192                         name = "";
 1193                 if (path == NULL) {
 1194                         path = "/";
 1195                         root = mypr->pr_root;
 1196                         vref(root);
 1197                 }
 1198                 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
 1199                 pr->pr_flags |= PR_HOST;
 1200 #if defined(INET) || defined(INET6)
 1201 #ifdef VIMAGE
 1202                 if (!(pr_flags & PR_VNET))
 1203 #endif
 1204                 {
 1205 #ifdef INET
 1206                         if (!(ch_flags & PR_IP4_USER))
 1207                                 pr->pr_flags |=
 1208                                     PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
 1209                         else if (!(pr_flags & PR_IP4_USER)) {
 1210                                 pr->pr_flags |= ppr->pr_flags & PR_IP4;
 1211                                 if (ppr->pr_ip4 != NULL) {
 1212                                         pr->pr_ip4s = ppr->pr_ip4s;
 1213                                         pr->pr_ip4 = malloc(pr->pr_ip4s *
 1214                                             sizeof(struct in_addr), M_PRISON,
 1215                                             M_WAITOK);
 1216                                         bcopy(ppr->pr_ip4, pr->pr_ip4,
 1217                                             pr->pr_ip4s * sizeof(*pr->pr_ip4));
 1218                                 }
 1219                         }
 1220 #endif
 1221 #ifdef INET6
 1222                         if (!(ch_flags & PR_IP6_USER))
 1223                                 pr->pr_flags |=
 1224                                     PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
 1225                         else if (!(pr_flags & PR_IP6_USER)) {
 1226                                 pr->pr_flags |= ppr->pr_flags & PR_IP6;
 1227                                 if (ppr->pr_ip6 != NULL) {
 1228                                         pr->pr_ip6s = ppr->pr_ip6s;
 1229                                         pr->pr_ip6 = malloc(pr->pr_ip6s *
 1230                                             sizeof(struct in6_addr), M_PRISON,
 1231                                             M_WAITOK);
 1232                                         bcopy(ppr->pr_ip6, pr->pr_ip6,
 1233                                             pr->pr_ip6s * sizeof(*pr->pr_ip6));
 1234                                 }
 1235                         }
 1236 #endif
 1237                 }
 1238 #endif
 1239                 /* Source address selection is always on by default. */
 1240                 pr->pr_flags |= _PR_IP_SADDRSEL;
 1241 
 1242                 pr->pr_securelevel = ppr->pr_securelevel;
 1243                 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
 1244                 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
 1245 
 1246                 LIST_INIT(&pr->pr_children);
 1247                 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
 1248 
 1249 #ifdef VIMAGE
 1250                 /* Allocate a new vnet if specified. */
 1251                 pr->pr_vnet = (pr_flags & PR_VNET)
 1252                     ? vnet_alloc() : ppr->pr_vnet;
 1253 #endif
 1254                 /*
 1255                  * Allocate a dedicated cpuset for each jail.
 1256                  * Unlike other initial settings, this may return an erorr.
 1257                  */
 1258                 error = cpuset_create_root(ppr, &pr->pr_cpuset);
 1259                 if (error) {
 1260                         prison_deref(pr, PD_LIST_XLOCKED);
 1261                         goto done_releroot;
 1262                 }
 1263 
 1264                 mtx_lock(&pr->pr_mtx);
 1265                 /*
 1266                  * New prisons do not yet have a reference, because we do not
 1267                  * want other to see the incomplete prison once the
 1268                  * allprison_lock is downgraded.
 1269                  */
 1270         } else {
 1271                 created = 0;
 1272                 /*
 1273                  * Grab a reference for existing prisons, to ensure they
 1274                  * continue to exist for the duration of the call.
 1275                  */
 1276                 pr->pr_ref++;
 1277 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
 1278                 if ((pr->pr_flags & PR_VNET) &&
 1279                     (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
 1280                         error = EINVAL;
 1281                         vfs_opterror(opts,
 1282                             "vnet jails cannot have IP address restrictions");
 1283                         goto done_deref_locked;
 1284                 }
 1285 #endif
 1286 #ifdef INET
 1287                 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
 1288                         error = EINVAL;
 1289                         vfs_opterror(opts,
 1290                             "ip4 cannot be changed after creation");
 1291                         goto done_deref_locked;
 1292                 }
 1293 #endif
 1294 #ifdef INET6
 1295                 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
 1296                         error = EINVAL;
 1297                         vfs_opterror(opts,
 1298                             "ip6 cannot be changed after creation");
 1299                         goto done_deref_locked;
 1300                 }
 1301 #endif
 1302         }
 1303 
 1304         /* Do final error checking before setting anything. */
 1305         if (gotslevel) {
 1306                 if (slevel < ppr->pr_securelevel) {
 1307                         error = EPERM;
 1308                         goto done_deref_locked;
 1309                 }
 1310         }
 1311         if (gotchildmax) {
 1312                 if (childmax >= ppr->pr_childmax) {
 1313                         error = EPERM;
 1314                         goto done_deref_locked;
 1315                 }
 1316         }
 1317         if (gotenforce) {
 1318                 if (enforce < ppr->pr_enforce_statfs) {
 1319                         error = EPERM;
 1320                         goto done_deref_locked;
 1321                 }
 1322         }
 1323 #ifdef INET
 1324         if (ip4s > 0) {
 1325                 if (ppr->pr_flags & PR_IP4) {
 1326                         /*
 1327                          * Make sure the new set of IP addresses is a
 1328                          * subset of the parent's list.  Don't worry
 1329                          * about the parent being unlocked, as any
 1330                          * setting is done with allprison_lock held.
 1331                          */
 1332                         for (ij = 0; ij < ppr->pr_ip4s; ij++)
 1333                                 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
 1334                                         break;
 1335                         if (ij == ppr->pr_ip4s) {
 1336                                 error = EPERM;
 1337                                 goto done_deref_locked;
 1338                         }
 1339                         if (ip4s > 1) {
 1340                                 for (ii = ij = 1; ii < ip4s; ii++) {
 1341                                         if (ip4[ii].s_addr ==
 1342                                             ppr->pr_ip4[0].s_addr)
 1343                                                 continue;
 1344                                         for (; ij < ppr->pr_ip4s; ij++)
 1345                                                 if (ip4[ii].s_addr ==
 1346                                                     ppr->pr_ip4[ij].s_addr)
 1347                                                         break;
 1348                                         if (ij == ppr->pr_ip4s)
 1349                                                 break;
 1350                                 }
 1351                                 if (ij == ppr->pr_ip4s) {
 1352                                         error = EPERM;
 1353                                         goto done_deref_locked;
 1354                                 }
 1355                         }
 1356                 }
 1357                 /*
 1358                  * Check for conflicting IP addresses.  We permit them
 1359                  * if there is no more than one IP on each jail.  If
 1360                  * there is a duplicate on a jail with more than one
 1361                  * IP stop checking and return error.
 1362                  */
 1363                 tppr = ppr;
 1364 #ifdef VIMAGE
 1365                 for (; tppr != &prison0; tppr = tppr->pr_parent)
 1366                         if (tppr->pr_flags & PR_VNET)
 1367                                 break;
 1368 #endif
 1369                 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
 1370                         if (tpr == pr ||
 1371 #ifdef VIMAGE
 1372                             (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
 1373 #endif
 1374                             tpr->pr_uref == 0) {
 1375                                 descend = 0;
 1376                                 continue;
 1377                         }
 1378                         if (!(tpr->pr_flags & PR_IP4_USER))
 1379                                 continue;
 1380                         descend = 0;
 1381                         if (tpr->pr_ip4 == NULL ||
 1382                             (ip4s == 1 && tpr->pr_ip4s == 1))
 1383                                 continue;
 1384                         for (ii = 0; ii < ip4s; ii++) {
 1385                                 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
 1386                                         error = EADDRINUSE;
 1387                                         vfs_opterror(opts,
 1388                                             "IPv4 addresses clash");
 1389                                         goto done_deref_locked;
 1390                                 }
 1391                         }
 1392                 }
 1393         }
 1394 #endif
 1395 #ifdef INET6
 1396         if (ip6s > 0) {
 1397                 if (ppr->pr_flags & PR_IP6) {
 1398                         /*
 1399                          * Make sure the new set of IP addresses is a
 1400                          * subset of the parent's list.
 1401                          */
 1402                         for (ij = 0; ij < ppr->pr_ip6s; ij++)
 1403                                 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
 1404                                     &ppr->pr_ip6[ij]))
 1405                                         break;
 1406                         if (ij == ppr->pr_ip6s) {
 1407                                 error = EPERM;
 1408                                 goto done_deref_locked;
 1409                         }
 1410                         if (ip6s > 1) {
 1411                                 for (ii = ij = 1; ii < ip6s; ii++) {
 1412                                         if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
 1413                                              &ppr->pr_ip6[0]))
 1414                                                 continue;
 1415                                         for (; ij < ppr->pr_ip6s; ij++)
 1416                                                 if (IN6_ARE_ADDR_EQUAL(
 1417                                                     &ip6[ii], &ppr->pr_ip6[ij]))
 1418                                                         break;
 1419                                         if (ij == ppr->pr_ip6s)
 1420                                                 break;
 1421                                 }
 1422                                 if (ij == ppr->pr_ip6s) {
 1423                                         error = EPERM;
 1424                                         goto done_deref_locked;
 1425                                 }
 1426                         }
 1427                 }
 1428                 /* Check for conflicting IP addresses. */
 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_IP6_USER))
 1445                                 continue;
 1446                         descend = 0;
 1447                         if (tpr->pr_ip6 == NULL ||
 1448                             (ip6s == 1 && tpr->pr_ip6s == 1))
 1449                                 continue;
 1450                         for (ii = 0; ii < ip6s; ii++) {
 1451                                 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
 1452                                         error = EADDRINUSE;
 1453                                         vfs_opterror(opts,
 1454                                             "IPv6 addresses clash");
 1455                                         goto done_deref_locked;
 1456                                 }
 1457                         }
 1458                 }
 1459         }
 1460 #endif
 1461         onamelen = namelen = 0;
 1462         if (name != NULL) {
 1463                 /* Give a default name of the jid. */
 1464                 if (name[0] == '\0')
 1465                         snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
 1466                 else if (*namelc == '' || (strtoul(namelc, &p, 10) != jid &&
 1467                     *p == '\0')) {
 1468                         error = EINVAL;
 1469                         vfs_opterror(opts,
 1470                             "name cannot be numeric (unless it is the jid)");
 1471                         goto done_deref_locked;
 1472                 }
 1473                 /*
 1474                  * Make sure the name isn't too long for the prison or its
 1475                  * children.
 1476                  */
 1477                 onamelen = strlen(pr->pr_name);
 1478                 namelen = strlen(name);
 1479                 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
 1480                         error = ENAMETOOLONG;
 1481                         goto done_deref_locked;
 1482                 }
 1483                 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
 1484                         if (strlen(tpr->pr_name) + (namelen - onamelen) >=
 1485                             sizeof(pr->pr_name)) {
 1486                                 error = ENAMETOOLONG;
 1487                                 goto done_deref_locked;
 1488                         }
 1489                 }
 1490         }
 1491         if (pr_allow & ~ppr->pr_allow) {
 1492                 error = EPERM;
 1493                 goto done_deref_locked;
 1494         }
 1495 
 1496         /* Set the parameters of the prison. */
 1497 #ifdef INET
 1498         redo_ip4 = 0;
 1499         if (pr_flags & PR_IP4_USER) {
 1500                 pr->pr_flags |= PR_IP4;
 1501                 free(pr->pr_ip4, M_PRISON);
 1502                 pr->pr_ip4s = ip4s;
 1503                 pr->pr_ip4 = ip4;
 1504                 ip4 = NULL;
 1505                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1506 #ifdef VIMAGE
 1507                         if (tpr->pr_flags & PR_VNET) {
 1508                                 descend = 0;
 1509                                 continue;
 1510                         }
 1511 #endif
 1512                         if (prison_restrict_ip4(tpr, NULL)) {
 1513                                 redo_ip4 = 1;
 1514                                 descend = 0;
 1515                         }
 1516                 }
 1517         }
 1518 #endif
 1519 #ifdef INET6
 1520         redo_ip6 = 0;
 1521         if (pr_flags & PR_IP6_USER) {
 1522                 pr->pr_flags |= PR_IP6;
 1523                 free(pr->pr_ip6, M_PRISON);
 1524                 pr->pr_ip6s = ip6s;
 1525                 pr->pr_ip6 = ip6;
 1526                 ip6 = NULL;
 1527                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1528 #ifdef VIMAGE
 1529                         if (tpr->pr_flags & PR_VNET) {
 1530                                 descend = 0;
 1531                                 continue;
 1532                         }
 1533 #endif
 1534                         if (prison_restrict_ip6(tpr, NULL)) {
 1535                                 redo_ip6 = 1;
 1536                                 descend = 0;
 1537                         }
 1538                 }
 1539         }
 1540 #endif
 1541         if (gotslevel) {
 1542                 pr->pr_securelevel = slevel;
 1543                 /* Set all child jails to be at least this level. */
 1544                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
 1545                         if (tpr->pr_securelevel < slevel)
 1546                                 tpr->pr_securelevel = slevel;
 1547         }
 1548         if (gotchildmax) {
 1549                 pr->pr_childmax = childmax;
 1550                 /* Set all child jails to under this limit. */
 1551                 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
 1552                         if (tpr->pr_childmax > childmax - level)
 1553                                 tpr->pr_childmax = childmax > level
 1554                                     ? childmax - level : 0;
 1555         }
 1556         if (gotenforce) {
 1557                 pr->pr_enforce_statfs = enforce;
 1558                 /* Pass this restriction on to the children. */
 1559                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
 1560                         if (tpr->pr_enforce_statfs < enforce)
 1561                                 tpr->pr_enforce_statfs = enforce;
 1562         }
 1563         if (name != NULL) {
 1564                 if (ppr == &prison0)
 1565                         strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
 1566                 else
 1567                         snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
 1568                             ppr->pr_name, name);
 1569                 /* Change this component of child names. */
 1570                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1571                         bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
 1572                             strlen(tpr->pr_name + onamelen) + 1);
 1573                         bcopy(pr->pr_name, tpr->pr_name, namelen);
 1574                 }
 1575         }
 1576         if (path != NULL) {
 1577                 /* Try to keep a real-rooted full pathname. */
 1578                 if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
 1579                         snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
 1580                             mypr->pr_path, path);
 1581                 else
 1582                         strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
 1583                 pr->pr_root = root;
 1584         }
 1585         if (PR_HOST & ch_flags & ~pr_flags) {
 1586                 if (pr->pr_flags & PR_HOST) {
 1587                         /*
 1588                          * Copy the parent's host info.  As with pr_ip4 above,
 1589                          * the lack of a lock on the parent is not a problem;
 1590                          * it is always set with allprison_lock at least
 1591                          * shared, and is held exclusively here.
 1592                          */
 1593                         strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
 1594                             sizeof(pr->pr_hostname));
 1595                         strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
 1596                             sizeof(pr->pr_domainname));
 1597                         strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
 1598                             sizeof(pr->pr_hostuuid));
 1599                         pr->pr_hostid = pr->pr_parent->pr_hostid;
 1600                 }
 1601         } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
 1602                 /* Set this prison, and any descendants without PR_HOST. */
 1603                 if (host != NULL)
 1604                         strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
 1605                 if (domain != NULL)
 1606                         strlcpy(pr->pr_domainname, domain, 
 1607                             sizeof(pr->pr_domainname));
 1608                 if (uuid != NULL)
 1609                         strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
 1610                 if (gothid)
 1611                         pr->pr_hostid = hid;
 1612                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1613                         if (tpr->pr_flags & PR_HOST)
 1614                                 descend = 0;
 1615                         else {
 1616                                 if (host != NULL)
 1617                                         strlcpy(tpr->pr_hostname,
 1618                                             pr->pr_hostname,
 1619                                             sizeof(tpr->pr_hostname));
 1620                                 if (domain != NULL)
 1621                                         strlcpy(tpr->pr_domainname, 
 1622                                             pr->pr_domainname,
 1623                                             sizeof(tpr->pr_domainname));
 1624                                 if (uuid != NULL)
 1625                                         strlcpy(tpr->pr_hostuuid,
 1626                                             pr->pr_hostuuid,
 1627                                             sizeof(tpr->pr_hostuuid));
 1628                                 if (gothid)
 1629                                         tpr->pr_hostid = hid;
 1630                         }
 1631                 }
 1632         }
 1633         if ((tallow = ch_allow & ~pr_allow)) {
 1634                 /* Clear allow bits in all children. */
 1635                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
 1636                         tpr->pr_allow &= ~tallow;
 1637         }
 1638         pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
 1639         /*
 1640          * Persistent prisons get an extra reference, and prisons losing their
 1641          * persist flag lose that reference.  Only do this for existing prisons
 1642          * for now, so new ones will remain unseen until after the module
 1643          * handlers have completed.
 1644          */
 1645         if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
 1646                 if (pr_flags & PR_PERSIST) {
 1647                         pr->pr_ref++;
 1648                         pr->pr_uref++;
 1649                 } else {
 1650                         pr->pr_ref--;
 1651                         pr->pr_uref--;
 1652                 }
 1653         }
 1654         pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
 1655         mtx_unlock(&pr->pr_mtx);
 1656 
 1657         /* Locks may have prevented a complete restriction of child IP
 1658          * addresses.  If so, allocate some more memory and try again.
 1659          */
 1660 #ifdef INET
 1661         while (redo_ip4) {
 1662                 ip4s = pr->pr_ip4s;
 1663                 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
 1664                 mtx_lock(&pr->pr_mtx);
 1665                 redo_ip4 = 0;
 1666                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1667 #ifdef VIMAGE
 1668                         if (tpr->pr_flags & PR_VNET) {
 1669                                 descend = 0;
 1670                                 continue;
 1671                         }
 1672 #endif
 1673                         if (prison_restrict_ip4(tpr, ip4)) {
 1674                                 if (ip4 != NULL)
 1675                                         ip4 = NULL;
 1676                                 else
 1677                                         redo_ip4 = 1;
 1678                         }
 1679                 }
 1680                 mtx_unlock(&pr->pr_mtx);
 1681         }
 1682 #endif
 1683 #ifdef INET6
 1684         while (redo_ip6) {
 1685                 ip6s = pr->pr_ip6s;
 1686                 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
 1687                 mtx_lock(&pr->pr_mtx);
 1688                 redo_ip6 = 0;
 1689                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1690 #ifdef VIMAGE
 1691                         if (tpr->pr_flags & PR_VNET) {
 1692                                 descend = 0;
 1693                                 continue;
 1694                         }
 1695 #endif
 1696                         if (prison_restrict_ip6(tpr, ip6)) {
 1697                                 if (ip6 != NULL)
 1698                                         ip6 = NULL;
 1699                                 else
 1700                                         redo_ip6 = 1;
 1701                         }
 1702                 }
 1703                 mtx_unlock(&pr->pr_mtx);
 1704         }
 1705 #endif
 1706 
 1707         /* Let the modules do their work. */
 1708         sx_downgrade(&allprison_lock);
 1709         if (created) {
 1710                 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
 1711                 if (error) {
 1712                         prison_deref(pr, PD_LIST_SLOCKED);
 1713                         goto done_errmsg;
 1714                 }
 1715         }
 1716         error = osd_jail_call(pr, PR_METHOD_SET, opts);
 1717         if (error) {
 1718                 prison_deref(pr, created
 1719                     ? PD_LIST_SLOCKED
 1720                     : PD_DEREF | PD_LIST_SLOCKED);
 1721                 goto done_errmsg;
 1722         }
 1723 
 1724         /* Attach this process to the prison if requested. */
 1725         if (flags & JAIL_ATTACH) {
 1726                 mtx_lock(&pr->pr_mtx);
 1727                 error = do_jail_attach(td, pr);
 1728                 if (error) {
 1729                         vfs_opterror(opts, "attach failed");
 1730                         if (!created)
 1731                                 prison_deref(pr, PD_DEREF);
 1732                         goto done_errmsg;
 1733                 }
 1734         }
 1735 
 1736         /*
 1737          * Now that it is all there, drop the temporary reference from existing
 1738          * prisons.  Or add a reference to newly created persistent prisons
 1739          * (which was not done earlier so that the prison would not be publicly
 1740          * visible).
 1741          */
 1742         if (!created) {
 1743                 prison_deref(pr, (flags & JAIL_ATTACH)
 1744                     ? PD_DEREF
 1745                     : PD_DEREF | PD_LIST_SLOCKED);
 1746         } else {
 1747                 if (pr_flags & PR_PERSIST) {
 1748                         mtx_lock(&pr->pr_mtx);
 1749                         pr->pr_ref++;
 1750                         pr->pr_uref++;
 1751                         mtx_unlock(&pr->pr_mtx);
 1752                 }
 1753                 if (!(flags & JAIL_ATTACH))
 1754                         sx_sunlock(&allprison_lock);
 1755         }
 1756         td->td_retval[0] = pr->pr_id;
 1757         goto done_errmsg;
 1758 
 1759  done_deref_locked:
 1760         prison_deref(pr, created
 1761             ? PD_LOCKED | PD_LIST_XLOCKED
 1762             : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
 1763         goto done_releroot;
 1764  done_unlock_list:
 1765         sx_xunlock(&allprison_lock);
 1766  done_releroot:
 1767         if (root != NULL) {
 1768                 vfslocked = VFS_LOCK_GIANT(root->v_mount);
 1769                 vrele(root);
 1770                 VFS_UNLOCK_GIANT(vfslocked);
 1771         }
 1772  done_errmsg:
 1773         if (error) {
 1774                 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
 1775                 if (errmsg_len > 0) {
 1776                         errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
 1777                         if (errmsg_pos > 0) {
 1778                                 if (optuio->uio_segflg == UIO_SYSSPACE)
 1779                                         bcopy(errmsg,
 1780                                            optuio->uio_iov[errmsg_pos].iov_base,
 1781                                            errmsg_len);
 1782                                 else
 1783                                         copyout(errmsg,
 1784                                            optuio->uio_iov[errmsg_pos].iov_base,
 1785                                            errmsg_len);
 1786                         }
 1787                 }
 1788         }
 1789  done_free:
 1790 #ifdef INET
 1791         free(ip4, M_PRISON);
 1792 #endif
 1793 #ifdef INET6
 1794         free(ip6, M_PRISON);
 1795 #endif
 1796         vfs_freeopts(opts);
 1797         return (error);
 1798 }
 1799 
 1800 
 1801 /*
 1802  * struct jail_get_args {
 1803  *      struct iovec *iovp;
 1804  *      unsigned int iovcnt;
 1805  *      int flags;
 1806  * };
 1807  */
 1808 int
 1809 jail_get(struct thread *td, struct jail_get_args *uap)
 1810 {
 1811         struct uio *auio;
 1812         int error;
 1813 
 1814         /* Check that we have an even number of iovecs. */
 1815         if (uap->iovcnt & 1)
 1816                 return (EINVAL);
 1817 
 1818         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
 1819         if (error)
 1820                 return (error);
 1821         error = kern_jail_get(td, auio, uap->flags);
 1822         if (error == 0)
 1823                 error = copyout(auio->uio_iov, uap->iovp,
 1824                     uap->iovcnt * sizeof (struct iovec));
 1825         free(auio, M_IOV);
 1826         return (error);
 1827 }
 1828 
 1829 int
 1830 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
 1831 {
 1832         struct prison *pr, *mypr;
 1833         struct vfsopt *opt;
 1834         struct vfsoptlist *opts;
 1835         char *errmsg, *name;
 1836         int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
 1837 
 1838         if (flags & ~JAIL_GET_MASK)
 1839                 return (EINVAL);
 1840 
 1841         /* Get the parameter list. */
 1842         error = vfs_buildopts(optuio, &opts);
 1843         if (error)
 1844                 return (error);
 1845         errmsg_pos = vfs_getopt_pos(opts, "errmsg");
 1846         mypr = td->td_ucred->cr_prison;
 1847 
 1848         /*
 1849          * Find the prison specified by one of: lastjid, jid, name.
 1850          */
 1851         sx_slock(&allprison_lock);
 1852         error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
 1853         if (error == 0) {
 1854                 TAILQ_FOREACH(pr, &allprison, pr_list) {
 1855                         if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
 1856                                 mtx_lock(&pr->pr_mtx);
 1857                                 if (pr->pr_ref > 0 &&
 1858                                     (pr->pr_uref > 0 || (flags & JAIL_DYING)))
 1859                                         break;
 1860                                 mtx_unlock(&pr->pr_mtx);
 1861                         }
 1862                 }
 1863                 if (pr != NULL)
 1864                         goto found_prison;
 1865                 error = ENOENT;
 1866                 vfs_opterror(opts, "no jail after %d", jid);
 1867                 goto done_unlock_list;
 1868         } else if (error != ENOENT)
 1869                 goto done_unlock_list;
 1870 
 1871         error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
 1872         if (error == 0) {
 1873                 if (jid != 0) {
 1874                         pr = prison_find_child(mypr, jid);
 1875                         if (pr != NULL) {
 1876                                 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
 1877                                         mtx_unlock(&pr->pr_mtx);
 1878                                         error = ENOENT;
 1879                                         vfs_opterror(opts, "jail %d is dying",
 1880                                             jid);
 1881                                         goto done_unlock_list;
 1882                                 }
 1883                                 goto found_prison;
 1884                         }
 1885                         error = ENOENT;
 1886                         vfs_opterror(opts, "jail %d not found", jid);
 1887                         goto done_unlock_list;
 1888                 }
 1889         } else if (error != ENOENT)
 1890                 goto done_unlock_list;
 1891 
 1892         error = vfs_getopt(opts, "name", (void **)&name, &len);
 1893         if (error == 0) {
 1894                 if (len == 0 || name[len - 1] != '\0') {
 1895                         error = EINVAL;
 1896                         goto done_unlock_list;
 1897                 }
 1898                 pr = prison_find_name(mypr, name);
 1899                 if (pr != NULL) {
 1900                         if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
 1901                                 mtx_unlock(&pr->pr_mtx);
 1902                                 error = ENOENT;
 1903                                 vfs_opterror(opts, "jail \"%s\" is dying",
 1904                                     name);
 1905                                 goto done_unlock_list;
 1906                         }
 1907                         goto found_prison;
 1908                 }
 1909                 error = ENOENT;
 1910                 vfs_opterror(opts, "jail \"%s\" not found", name);
 1911                 goto done_unlock_list;
 1912         } else if (error != ENOENT)
 1913                 goto done_unlock_list;
 1914 
 1915         vfs_opterror(opts, "no jail specified");
 1916         error = ENOENT;
 1917         goto done_unlock_list;
 1918 
 1919  found_prison:
 1920         /* Get the parameters of the prison. */
 1921         pr->pr_ref++;
 1922         locked = PD_LOCKED;
 1923         td->td_retval[0] = pr->pr_id;
 1924         error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
 1925         if (error != 0 && error != ENOENT)
 1926                 goto done_deref;
 1927         i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
 1928         error = vfs_setopt(opts, "parent", &i, sizeof(i));
 1929         if (error != 0 && error != ENOENT)
 1930                 goto done_deref;
 1931         error = vfs_setopts(opts, "name", prison_name(mypr, pr));
 1932         if (error != 0 && error != ENOENT)
 1933                 goto done_deref;
 1934         error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
 1935             sizeof(pr->pr_cpuset->cs_id));
 1936         if (error != 0 && error != ENOENT)
 1937                 goto done_deref;
 1938         error = vfs_setopts(opts, "path", prison_path(mypr, pr));
 1939         if (error != 0 && error != ENOENT)
 1940                 goto done_deref;
 1941 #ifdef INET
 1942         error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
 1943             pr->pr_ip4s * sizeof(*pr->pr_ip4));
 1944         if (error != 0 && error != ENOENT)
 1945                 goto done_deref;
 1946 #endif
 1947 #ifdef INET6
 1948         error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
 1949             pr->pr_ip6s * sizeof(*pr->pr_ip6));
 1950         if (error != 0 && error != ENOENT)
 1951                 goto done_deref;
 1952 #endif
 1953         error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
 1954             sizeof(pr->pr_securelevel));
 1955         if (error != 0 && error != ENOENT)
 1956                 goto done_deref;
 1957         error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
 1958             sizeof(pr->pr_childcount));
 1959         if (error != 0 && error != ENOENT)
 1960                 goto done_deref;
 1961         error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
 1962             sizeof(pr->pr_childmax));
 1963         if (error != 0 && error != ENOENT)
 1964                 goto done_deref;
 1965         error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
 1966         if (error != 0 && error != ENOENT)
 1967                 goto done_deref;
 1968         error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
 1969         if (error != 0 && error != ENOENT)
 1970                 goto done_deref;
 1971         error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
 1972         if (error != 0 && error != ENOENT)
 1973                 goto done_deref;
 1974 #ifdef COMPAT_FREEBSD32
 1975         if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
 1976                 uint32_t hid32 = pr->pr_hostid;
 1977 
 1978                 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
 1979         } else
 1980 #endif
 1981         error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
 1982             sizeof(pr->pr_hostid));
 1983         if (error != 0 && error != ENOENT)
 1984                 goto done_deref;
 1985         error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
 1986             sizeof(pr->pr_enforce_statfs));
 1987         if (error != 0 && error != ENOENT)
 1988                 goto done_deref;
 1989         for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
 1990             fi++) {
 1991                 if (pr_flag_names[fi] == NULL)
 1992                         continue;
 1993                 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
 1994                 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
 1995                 if (error != 0 && error != ENOENT)
 1996                         goto done_deref;
 1997                 i = !i;
 1998                 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
 1999                 if (error != 0 && error != ENOENT)
 2000                         goto done_deref;
 2001         }
 2002         for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
 2003             fi++) {
 2004                 i = pr->pr_flags &
 2005                     (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
 2006                 i = pr_flag_jailsys[fi].disable &&
 2007                       (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
 2008                     : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
 2009                     : JAIL_SYS_INHERIT;
 2010                 error =
 2011                     vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
 2012                 if (error != 0 && error != ENOENT)
 2013                         goto done_deref;
 2014         }
 2015         for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
 2016             fi++) {
 2017                 if (pr_allow_names[fi] == NULL)
 2018                         continue;
 2019                 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
 2020                 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
 2021                 if (error != 0 && error != ENOENT)
 2022                         goto done_deref;
 2023                 i = !i;
 2024                 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
 2025                 if (error != 0 && error != ENOENT)
 2026                         goto done_deref;
 2027         }
 2028         i = (pr->pr_uref == 0);
 2029         error = vfs_setopt(opts, "dying", &i, sizeof(i));
 2030         if (error != 0 && error != ENOENT)
 2031                 goto done_deref;
 2032         i = !i;
 2033         error = vfs_setopt(opts, "nodying", &i, sizeof(i));
 2034         if (error != 0 && error != ENOENT)
 2035                 goto done_deref;
 2036 
 2037         /* Get the module parameters. */
 2038         mtx_unlock(&pr->pr_mtx);
 2039         locked = 0;
 2040         error = osd_jail_call(pr, PR_METHOD_GET, opts);
 2041         if (error)
 2042                 goto done_deref;
 2043         prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
 2044 
 2045         /* By now, all parameters should have been noted. */
 2046         TAILQ_FOREACH(opt, opts, link) {
 2047                 if (!opt->seen && strcmp(opt->name, "errmsg")) {
 2048                         error = EINVAL;
 2049                         vfs_opterror(opts, "unknown parameter: %s", opt->name);
 2050                         goto done_errmsg;
 2051                 }
 2052         }
 2053 
 2054         /* Write the fetched parameters back to userspace. */
 2055         error = 0;
 2056         TAILQ_FOREACH(opt, opts, link) {
 2057                 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
 2058                         pos = 2 * opt->pos + 1;
 2059                         optuio->uio_iov[pos].iov_len = opt->len;
 2060                         if (opt->value != NULL) {
 2061                                 if (optuio->uio_segflg == UIO_SYSSPACE) {
 2062                                         bcopy(opt->value,
 2063                                             optuio->uio_iov[pos].iov_base,
 2064                                             opt->len);
 2065                                 } else {
 2066                                         error = copyout(opt->value,
 2067                                             optuio->uio_iov[pos].iov_base,
 2068                                             opt->len);
 2069                                         if (error)
 2070                                                 break;
 2071                                 }
 2072                         }
 2073                 }
 2074         }
 2075         goto done_errmsg;
 2076 
 2077  done_deref:
 2078         prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
 2079         goto done_errmsg;
 2080 
 2081  done_unlock_list:
 2082         sx_sunlock(&allprison_lock);
 2083  done_errmsg:
 2084         if (error && errmsg_pos >= 0) {
 2085                 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
 2086                 errmsg_pos = 2 * errmsg_pos + 1;
 2087                 if (errmsg_len > 0) {
 2088                         if (optuio->uio_segflg == UIO_SYSSPACE)
 2089                                 bcopy(errmsg,
 2090                                     optuio->uio_iov[errmsg_pos].iov_base,
 2091                                     errmsg_len);
 2092                         else
 2093                                 copyout(errmsg,
 2094                                     optuio->uio_iov[errmsg_pos].iov_base,
 2095                                     errmsg_len);
 2096                 }
 2097         }
 2098         vfs_freeopts(opts);
 2099         return (error);
 2100 }
 2101 
 2102 
 2103 /*
 2104  * struct jail_remove_args {
 2105  *      int jid;
 2106  * };
 2107  */
 2108 int
 2109 jail_remove(struct thread *td, struct jail_remove_args *uap)
 2110 {
 2111         struct prison *pr, *cpr, *lpr, *tpr;
 2112         int descend, error;
 2113 
 2114         error = priv_check(td, PRIV_JAIL_REMOVE);
 2115         if (error)
 2116                 return (error);
 2117 
 2118         sx_xlock(&allprison_lock);
 2119         pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
 2120         if (pr == NULL) {
 2121                 sx_xunlock(&allprison_lock);
 2122                 return (EINVAL);
 2123         }
 2124 
 2125         /* Remove all descendants of this prison, then remove this prison. */
 2126         pr->pr_ref++;
 2127         pr->pr_flags |= PR_REMOVE;
 2128         if (!LIST_EMPTY(&pr->pr_children)) {
 2129                 mtx_unlock(&pr->pr_mtx);
 2130                 lpr = NULL;
 2131                 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
 2132                         mtx_lock(&cpr->pr_mtx);
 2133                         if (cpr->pr_ref > 0) {
 2134                                 tpr = cpr;
 2135                                 cpr->pr_ref++;
 2136                                 cpr->pr_flags |= PR_REMOVE;
 2137                         } else {
 2138                                 /* Already removed - do not do it again. */
 2139                                 tpr = NULL;
 2140                         }
 2141                         mtx_unlock(&cpr->pr_mtx);
 2142                         if (lpr != NULL) {
 2143                                 mtx_lock(&lpr->pr_mtx);
 2144                                 prison_remove_one(lpr);
 2145                                 sx_xlock(&allprison_lock);
 2146                         }
 2147                         lpr = tpr;
 2148                 }
 2149                 if (lpr != NULL) {
 2150                         mtx_lock(&lpr->pr_mtx);
 2151                         prison_remove_one(lpr);
 2152                         sx_xlock(&allprison_lock);
 2153                 }
 2154                 mtx_lock(&pr->pr_mtx);
 2155         }
 2156         prison_remove_one(pr);
 2157         return (0);
 2158 }
 2159 
 2160 static void
 2161 prison_remove_one(struct prison *pr)
 2162 {
 2163         struct proc *p;
 2164         int deuref;
 2165 
 2166         /* If the prison was persistent, it is not anymore. */
 2167         deuref = 0;
 2168         if (pr->pr_flags & PR_PERSIST) {
 2169                 pr->pr_ref--;
 2170                 deuref = PD_DEUREF;
 2171                 pr->pr_flags &= ~PR_PERSIST;
 2172         }
 2173 
 2174         /*
 2175          * jail_remove added a reference.  If that's the only one, remove
 2176          * the prison now.
 2177          */
 2178         KASSERT(pr->pr_ref > 0,
 2179             ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
 2180         if (pr->pr_ref == 1) {
 2181                 prison_deref(pr,
 2182                     deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
 2183                 return;
 2184         }
 2185 
 2186         mtx_unlock(&pr->pr_mtx);
 2187         sx_xunlock(&allprison_lock);
 2188         /*
 2189          * Kill all processes unfortunate enough to be attached to this prison.
 2190          */
 2191         sx_slock(&allproc_lock);
 2192         LIST_FOREACH(p, &allproc, p_list) {
 2193                 PROC_LOCK(p);
 2194                 if (p->p_state != PRS_NEW && p->p_ucred &&
 2195                     p->p_ucred->cr_prison == pr)
 2196                         psignal(p, SIGKILL);
 2197                 PROC_UNLOCK(p);
 2198         }
 2199         sx_sunlock(&allproc_lock);
 2200         /* Remove the temporary reference added by jail_remove. */
 2201         prison_deref(pr, deuref | PD_DEREF);
 2202 }
 2203 
 2204 
 2205 /*
 2206  * struct jail_attach_args {
 2207  *      int jid;
 2208  * };
 2209  */
 2210 int
 2211 jail_attach(struct thread *td, struct jail_attach_args *uap)
 2212 {
 2213         struct prison *pr;
 2214         int error;
 2215 
 2216         error = priv_check(td, PRIV_JAIL_ATTACH);
 2217         if (error)
 2218                 return (error);
 2219 
 2220         sx_slock(&allprison_lock);
 2221         pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
 2222         if (pr == NULL) {
 2223                 sx_sunlock(&allprison_lock);
 2224                 return (EINVAL);
 2225         }
 2226 
 2227         /*
 2228          * Do not allow a process to attach to a prison that is not
 2229          * considered to be "alive".
 2230          */
 2231         if (pr->pr_uref == 0) {
 2232                 mtx_unlock(&pr->pr_mtx);
 2233                 sx_sunlock(&allprison_lock);
 2234                 return (EINVAL);
 2235         }
 2236 
 2237         return (do_jail_attach(td, pr));
 2238 }
 2239 
 2240 static int
 2241 do_jail_attach(struct thread *td, struct prison *pr)
 2242 {
 2243         struct prison *ppr;
 2244         struct proc *p;
 2245         struct ucred *newcred, *oldcred;
 2246         int vfslocked, error;
 2247 
 2248         /*
 2249          * XXX: Note that there is a slight race here if two threads
 2250          * in the same privileged process attempt to attach to two
 2251          * different jails at the same time.  It is important for
 2252          * user processes not to do this, or they might end up with
 2253          * a process root from one prison, but attached to the jail
 2254          * of another.
 2255          */
 2256         pr->pr_ref++;
 2257         pr->pr_uref++;
 2258         mtx_unlock(&pr->pr_mtx);
 2259 
 2260         /* Let modules do whatever they need to prepare for attaching. */
 2261         error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
 2262         if (error) {
 2263                 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
 2264                 return (error);
 2265         }
 2266         sx_sunlock(&allprison_lock);
 2267 
 2268         /*
 2269          * Reparent the newly attached process to this jail.
 2270          */
 2271         ppr = td->td_ucred->cr_prison;
 2272         p = td->td_proc;
 2273         error = cpuset_setproc_update_set(p, pr->pr_cpuset);
 2274         if (error)
 2275                 goto e_revert_osd;
 2276 
 2277         vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
 2278         vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
 2279         if ((error = change_dir(pr->pr_root, td)) != 0)
 2280                 goto e_unlock;
 2281 #ifdef MAC
 2282         if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
 2283                 goto e_unlock;
 2284 #endif
 2285         VOP_UNLOCK(pr->pr_root, 0);
 2286         if ((error = change_root(pr->pr_root, td)))
 2287                 goto e_unlock_giant;
 2288         VFS_UNLOCK_GIANT(vfslocked);
 2289 
 2290         newcred = crget();
 2291         PROC_LOCK(p);
 2292         oldcred = p->p_ucred;
 2293         setsugid(p);
 2294         crcopy(newcred, oldcred);
 2295         newcred->cr_prison = pr;
 2296         p->p_ucred = newcred;
 2297         PROC_UNLOCK(p);
 2298         crfree(oldcred);
 2299         prison_deref(ppr, PD_DEREF | PD_DEUREF);
 2300         return (0);
 2301  e_unlock:
 2302         VOP_UNLOCK(pr->pr_root, 0);
 2303  e_unlock_giant:
 2304         VFS_UNLOCK_GIANT(vfslocked);
 2305  e_revert_osd:
 2306         /* Tell modules this thread is still in its old jail after all. */
 2307         (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
 2308         prison_deref(pr, PD_DEREF | PD_DEUREF);
 2309         return (error);
 2310 }
 2311 
 2312 
 2313 /*
 2314  * Returns a locked prison instance, or NULL on failure.
 2315  */
 2316 struct prison *
 2317 prison_find(int prid)
 2318 {
 2319         struct prison *pr;
 2320 
 2321         sx_assert(&allprison_lock, SX_LOCKED);
 2322         TAILQ_FOREACH(pr, &allprison, pr_list) {
 2323                 if (pr->pr_id == prid) {
 2324                         mtx_lock(&pr->pr_mtx);
 2325                         if (pr->pr_ref > 0)
 2326                                 return (pr);
 2327                         mtx_unlock(&pr->pr_mtx);
 2328                 }
 2329         }
 2330         return (NULL);
 2331 }
 2332 
 2333 /*
 2334  * Find a prison that is a descendant of mypr.  Returns a locked prison or NULL.
 2335  */
 2336 struct prison *
 2337 prison_find_child(struct prison *mypr, int prid)
 2338 {
 2339         struct prison *pr;
 2340         int descend;
 2341 
 2342         sx_assert(&allprison_lock, SX_LOCKED);
 2343         FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
 2344                 if (pr->pr_id == prid) {
 2345                         mtx_lock(&pr->pr_mtx);
 2346                         if (pr->pr_ref > 0)
 2347                                 return (pr);
 2348                         mtx_unlock(&pr->pr_mtx);
 2349                 }
 2350         }
 2351         return (NULL);
 2352 }
 2353 
 2354 /*
 2355  * Look for the name relative to mypr.  Returns a locked prison or NULL.
 2356  */
 2357 struct prison *
 2358 prison_find_name(struct prison *mypr, const char *name)
 2359 {
 2360         struct prison *pr, *deadpr;
 2361         size_t mylen;
 2362         int descend;
 2363 
 2364         sx_assert(&allprison_lock, SX_LOCKED);
 2365         mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
 2366  again:
 2367         deadpr = NULL;
 2368         FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
 2369                 if (!strcmp(pr->pr_name + mylen, name)) {
 2370                         mtx_lock(&pr->pr_mtx);
 2371                         if (pr->pr_ref > 0) {
 2372                                 if (pr->pr_uref > 0)
 2373                                         return (pr);
 2374                                 deadpr = pr;
 2375                         }
 2376                         mtx_unlock(&pr->pr_mtx);
 2377                 }
 2378         }
 2379         /* There was no valid prison - perhaps there was a dying one. */
 2380         if (deadpr != NULL) {
 2381                 mtx_lock(&deadpr->pr_mtx);
 2382                 if (deadpr->pr_ref == 0) {
 2383                         mtx_unlock(&deadpr->pr_mtx);
 2384                         goto again;
 2385                 }
 2386         }
 2387         return (deadpr);
 2388 }
 2389 
 2390 /*
 2391  * See if a prison has the specific flag set.
 2392  */
 2393 int
 2394 prison_flag(struct ucred *cred, unsigned flag)
 2395 {
 2396 
 2397         /* This is an atomic read, so no locking is necessary. */
 2398         return (cred->cr_prison->pr_flags & flag);
 2399 }
 2400 
 2401 int
 2402 prison_allow(struct ucred *cred, unsigned flag)
 2403 {
 2404 
 2405         /* This is an atomic read, so no locking is necessary. */
 2406         return (cred->cr_prison->pr_allow & flag);
 2407 }
 2408 
 2409 /*
 2410  * Remove a prison reference.  If that was the last reference, remove the
 2411  * prison itself - but not in this context in case there are locks held.
 2412  */
 2413 void
 2414 prison_free_locked(struct prison *pr)
 2415 {
 2416 
 2417         mtx_assert(&pr->pr_mtx, MA_OWNED);
 2418         pr->pr_ref--;
 2419         if (pr->pr_ref == 0) {
 2420                 mtx_unlock(&pr->pr_mtx);
 2421                 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
 2422                 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
 2423                 return;
 2424         }
 2425         mtx_unlock(&pr->pr_mtx);
 2426 }
 2427 
 2428 void
 2429 prison_free(struct prison *pr)
 2430 {
 2431 
 2432         mtx_lock(&pr->pr_mtx);
 2433         prison_free_locked(pr);
 2434 }
 2435 
 2436 static void
 2437 prison_complete(void *context, int pending)
 2438 {
 2439 
 2440         prison_deref((struct prison *)context, 0);
 2441 }
 2442 
 2443 /*
 2444  * Remove a prison reference (usually).  This internal version assumes no
 2445  * mutexes are held, except perhaps the prison itself.  If there are no more
 2446  * references, release and delist the prison.  On completion, the prison lock
 2447  * and the allprison lock are both unlocked.
 2448  */
 2449 static void
 2450 prison_deref(struct prison *pr, int flags)
 2451 {
 2452         struct prison *ppr, *tpr;
 2453         int vfslocked;
 2454 
 2455         if (!(flags & PD_LOCKED))
 2456                 mtx_lock(&pr->pr_mtx);
 2457         /* Decrement the user references in a separate loop. */
 2458         if (flags & PD_DEUREF) {
 2459                 for (tpr = pr;; tpr = tpr->pr_parent) {
 2460                         if (tpr != pr)
 2461                                 mtx_lock(&tpr->pr_mtx);
 2462                         if (--tpr->pr_uref > 0)
 2463                                 break;
 2464                         KASSERT(tpr != &prison0, ("prison0 pr_uref=0"));
 2465                         mtx_unlock(&tpr->pr_mtx);
 2466                 }
 2467                 /* Done if there were only user references to remove. */
 2468                 if (!(flags & PD_DEREF)) {
 2469                         mtx_unlock(&tpr->pr_mtx);
 2470                         if (flags & PD_LIST_SLOCKED)
 2471                                 sx_sunlock(&allprison_lock);
 2472                         else if (flags & PD_LIST_XLOCKED)
 2473                                 sx_xunlock(&allprison_lock);
 2474                         return;
 2475                 }
 2476                 if (tpr != pr) {
 2477                         mtx_unlock(&tpr->pr_mtx);
 2478                         mtx_lock(&pr->pr_mtx);
 2479                 }
 2480         }
 2481 
 2482         for (;;) {
 2483                 if (flags & PD_DEREF)
 2484                         pr->pr_ref--;
 2485                 /* If the prison still has references, nothing else to do. */
 2486                 if (pr->pr_ref > 0) {
 2487                         mtx_unlock(&pr->pr_mtx);
 2488                         if (flags & PD_LIST_SLOCKED)
 2489                                 sx_sunlock(&allprison_lock);
 2490                         else if (flags & PD_LIST_XLOCKED)
 2491                                 sx_xunlock(&allprison_lock);
 2492                         return;
 2493                 }
 2494 
 2495                 mtx_unlock(&pr->pr_mtx);
 2496                 if (flags & PD_LIST_SLOCKED) {
 2497                         if (!sx_try_upgrade(&allprison_lock)) {
 2498                                 sx_sunlock(&allprison_lock);
 2499                                 sx_xlock(&allprison_lock);
 2500                         }
 2501                 } else if (!(flags & PD_LIST_XLOCKED))
 2502                         sx_xlock(&allprison_lock);
 2503 
 2504                 TAILQ_REMOVE(&allprison, pr, pr_list);
 2505                 LIST_REMOVE(pr, pr_sibling);
 2506                 ppr = pr->pr_parent;
 2507                 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
 2508                         tpr->pr_childcount--;
 2509                 sx_xunlock(&allprison_lock);
 2510 
 2511 #ifdef VIMAGE
 2512                 if (pr->pr_vnet != ppr->pr_vnet)
 2513                         vnet_destroy(pr->pr_vnet);
 2514 #endif
 2515                 if (pr->pr_root != NULL) {
 2516                         vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
 2517                         vrele(pr->pr_root);
 2518                         VFS_UNLOCK_GIANT(vfslocked);
 2519                 }
 2520                 mtx_destroy(&pr->pr_mtx);
 2521 #ifdef INET
 2522                 free(pr->pr_ip4, M_PRISON);
 2523 #endif
 2524 #ifdef INET6
 2525                 free(pr->pr_ip6, M_PRISON);
 2526 #endif
 2527                 if (pr->pr_cpuset != NULL)
 2528                         cpuset_rel(pr->pr_cpuset);
 2529                 osd_jail_exit(pr);
 2530                 free(pr, M_PRISON);
 2531 
 2532                 /* Removing a prison frees a reference on its parent. */
 2533                 pr = ppr;
 2534                 mtx_lock(&pr->pr_mtx);
 2535                 flags = PD_DEREF;
 2536         }
 2537 }
 2538 
 2539 void
 2540 prison_hold_locked(struct prison *pr)
 2541 {
 2542 
 2543         mtx_assert(&pr->pr_mtx, MA_OWNED);
 2544         KASSERT(pr->pr_ref > 0,
 2545             ("Trying to hold dead prison (jid=%d).", pr->pr_id));
 2546         pr->pr_ref++;
 2547 }
 2548 
 2549 void
 2550 prison_hold(struct prison *pr)
 2551 {
 2552 
 2553         mtx_lock(&pr->pr_mtx);
 2554         prison_hold_locked(pr);
 2555         mtx_unlock(&pr->pr_mtx);
 2556 }
 2557 
 2558 void
 2559 prison_proc_hold(struct prison *pr)
 2560 {
 2561 
 2562         mtx_lock(&pr->pr_mtx);
 2563         KASSERT(pr->pr_uref > 0,
 2564             ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
 2565         pr->pr_uref++;
 2566         mtx_unlock(&pr->pr_mtx);
 2567 }
 2568 
 2569 void
 2570 prison_proc_free(struct prison *pr)
 2571 {
 2572 
 2573         mtx_lock(&pr->pr_mtx);
 2574         KASSERT(pr->pr_uref > 0,
 2575             ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
 2576         prison_deref(pr, PD_DEUREF | PD_LOCKED);
 2577 }
 2578 
 2579 
 2580 #ifdef INET
 2581 /*
 2582  * Restrict a prison's IP address list with its parent's, possibly replacing
 2583  * it.  Return true if the replacement buffer was used (or would have been).
 2584  */
 2585 static int
 2586 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
 2587 {
 2588         int ii, ij, used;
 2589         struct prison *ppr;
 2590 
 2591         ppr = pr->pr_parent;
 2592         if (!(pr->pr_flags & PR_IP4_USER)) {
 2593                 /* This has no user settings, so just copy the parent's list. */
 2594                 if (pr->pr_ip4s < ppr->pr_ip4s) {
 2595                         /*
 2596                          * There's no room for the parent's list.  Use the
 2597                          * new list buffer, which is assumed to be big enough
 2598                          * (if it was passed).  If there's no buffer, try to
 2599                          * allocate one.
 2600                          */
 2601                         used = 1;
 2602                         if (newip4 == NULL) {
 2603                                 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
 2604                                     M_PRISON, M_NOWAIT);
 2605                                 if (newip4 != NULL)
 2606                                         used = 0;
 2607                         }
 2608                         if (newip4 != NULL) {
 2609                                 bcopy(ppr->pr_ip4, newip4,
 2610                                     ppr->pr_ip4s * sizeof(*newip4));
 2611                                 free(pr->pr_ip4, M_PRISON);
 2612                                 pr->pr_ip4 = newip4;
 2613                                 pr->pr_ip4s = ppr->pr_ip4s;
 2614                         }
 2615                         return (used);
 2616                 }
 2617                 pr->pr_ip4s = ppr->pr_ip4s;
 2618                 if (pr->pr_ip4s > 0)
 2619                         bcopy(ppr->pr_ip4, pr->pr_ip4,
 2620                             pr->pr_ip4s * sizeof(*newip4));
 2621                 else if (pr->pr_ip4 != NULL) {
 2622                         free(pr->pr_ip4, M_PRISON);
 2623                         pr->pr_ip4 = NULL;
 2624                 }
 2625         } else if (pr->pr_ip4s > 0) {
 2626                 /* Remove addresses that aren't in the parent. */
 2627                 for (ij = 0; ij < ppr->pr_ip4s; ij++)
 2628                         if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
 2629                                 break;
 2630                 if (ij < ppr->pr_ip4s)
 2631                         ii = 1;
 2632                 else {
 2633                         bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
 2634                             --pr->pr_ip4s * sizeof(*pr->pr_ip4));
 2635                         ii = 0;
 2636                 }
 2637                 for (ij = 1; ii < pr->pr_ip4s; ) {
 2638                         if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
 2639                                 ii++;
 2640                                 continue;
 2641                         }
 2642                         switch (ij >= ppr->pr_ip4s ? -1 :
 2643                                 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
 2644                         case -1:
 2645                                 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
 2646                                     (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
 2647                                 break;
 2648                         case 0:
 2649                                 ii++;
 2650                                 ij++;
 2651                                 break;
 2652                         case 1:
 2653                                 ij++;
 2654                                 break;
 2655                         }
 2656                 }
 2657                 if (pr->pr_ip4s == 0) {
 2658                         pr->pr_flags |= PR_IP4_DISABLE;
 2659                         free(pr->pr_ip4, M_PRISON);
 2660                         pr->pr_ip4 = NULL;
 2661                 }
 2662         }
 2663         return (0);
 2664 }
 2665 
 2666 /*
 2667  * Pass back primary IPv4 address of this jail.
 2668  *
 2669  * If not restricted return success but do not alter the address.  Caller has
 2670  * to make sure to initialize it correctly (e.g. INADDR_ANY).
 2671  *
 2672  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
 2673  * Address returned in NBO.
 2674  */
 2675 int
 2676 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
 2677 {
 2678         struct prison *pr;
 2679 
 2680         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2681         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2682 
 2683         pr = cred->cr_prison;
 2684         if (!(pr->pr_flags & PR_IP4))
 2685                 return (0);
 2686         mtx_lock(&pr->pr_mtx);
 2687         if (!(pr->pr_flags & PR_IP4)) {
 2688                 mtx_unlock(&pr->pr_mtx);
 2689                 return (0);
 2690         }
 2691         if (pr->pr_ip4 == NULL) {
 2692                 mtx_unlock(&pr->pr_mtx);
 2693                 return (EAFNOSUPPORT);
 2694         }
 2695 
 2696         ia->s_addr = pr->pr_ip4[0].s_addr;
 2697         mtx_unlock(&pr->pr_mtx);
 2698         return (0);
 2699 }
 2700 
 2701 /*
 2702  * Return 1 if we should do proper source address selection or are not jailed.
 2703  * We will return 0 if we should bypass source address selection in favour
 2704  * of the primary jail IPv4 address. Only in this case *ia will be updated and
 2705  * returned in NBO.
 2706  * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
 2707  */
 2708 int
 2709 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
 2710 {
 2711         struct prison *pr;
 2712         struct in_addr lia;
 2713         int error;
 2714 
 2715         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2716         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2717 
 2718         if (!jailed(cred))
 2719                 return (1);
 2720 
 2721         pr = cred->cr_prison;
 2722         if (pr->pr_flags & PR_IP4_SADDRSEL)
 2723                 return (1);
 2724 
 2725         lia.s_addr = INADDR_ANY;
 2726         error = prison_get_ip4(cred, &lia);
 2727         if (error)
 2728                 return (error);
 2729         if (lia.s_addr == INADDR_ANY)
 2730                 return (1);
 2731 
 2732         ia->s_addr = lia.s_addr;
 2733         return (0);
 2734 }
 2735 
 2736 /*
 2737  * Return true if pr1 and pr2 have the same IPv4 address restrictions.
 2738  */
 2739 int
 2740 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
 2741 {
 2742 
 2743         if (pr1 == pr2)
 2744                 return (1);
 2745 
 2746         /*
 2747          * No need to lock since the PR_IP4_USER flag can't be altered for
 2748          * existing prisons.
 2749          */
 2750         while (pr1 != &prison0 &&
 2751 #ifdef VIMAGE
 2752                !(pr1->pr_flags & PR_VNET) &&
 2753 #endif
 2754                !(pr1->pr_flags & PR_IP4_USER))
 2755                 pr1 = pr1->pr_parent;
 2756         while (pr2 != &prison0 &&
 2757 #ifdef VIMAGE
 2758                !(pr2->pr_flags & PR_VNET) &&
 2759 #endif
 2760                !(pr2->pr_flags & PR_IP4_USER))
 2761                 pr2 = pr2->pr_parent;
 2762         return (pr1 == pr2);
 2763 }
 2764 
 2765 /*
 2766  * Make sure our (source) address is set to something meaningful to this
 2767  * jail.
 2768  *
 2769  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
 2770  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 2771  * doesn't allow IPv4.  Address passed in in NBO and returned in NBO.
 2772  */
 2773 int
 2774 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
 2775 {
 2776         struct prison *pr;
 2777         struct in_addr ia0;
 2778         int error;
 2779 
 2780         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2781         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2782 
 2783         pr = cred->cr_prison;
 2784         if (!(pr->pr_flags & PR_IP4))
 2785                 return (0);
 2786         mtx_lock(&pr->pr_mtx);
 2787         if (!(pr->pr_flags & PR_IP4)) {
 2788                 mtx_unlock(&pr->pr_mtx);
 2789                 return (0);
 2790         }
 2791         if (pr->pr_ip4 == NULL) {
 2792                 mtx_unlock(&pr->pr_mtx);
 2793                 return (EAFNOSUPPORT);
 2794         }
 2795 
 2796         ia0.s_addr = ntohl(ia->s_addr);
 2797         if (ia0.s_addr == INADDR_LOOPBACK) {
 2798                 ia->s_addr = pr->pr_ip4[0].s_addr;
 2799                 mtx_unlock(&pr->pr_mtx);
 2800                 return (0);
 2801         }
 2802 
 2803         if (ia0.s_addr == INADDR_ANY) {
 2804                 /*
 2805                  * In case there is only 1 IPv4 address, bind directly.
 2806                  */
 2807                 if (pr->pr_ip4s == 1)
 2808                         ia->s_addr = pr->pr_ip4[0].s_addr;
 2809                 mtx_unlock(&pr->pr_mtx);
 2810                 return (0);
 2811         }
 2812 
 2813         error = _prison_check_ip4(pr, ia);
 2814         mtx_unlock(&pr->pr_mtx);
 2815         return (error);
 2816 }
 2817 
 2818 /*
 2819  * Rewrite destination address in case we will connect to loopback address.
 2820  *
 2821  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
 2822  * Address passed in in NBO and returned in NBO.
 2823  */
 2824 int
 2825 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
 2826 {
 2827         struct prison *pr;
 2828 
 2829         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2830         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2831 
 2832         pr = cred->cr_prison;
 2833         if (!(pr->pr_flags & PR_IP4))
 2834                 return (0);
 2835         mtx_lock(&pr->pr_mtx);
 2836         if (!(pr->pr_flags & PR_IP4)) {
 2837                 mtx_unlock(&pr->pr_mtx);
 2838                 return (0);
 2839         }
 2840         if (pr->pr_ip4 == NULL) {
 2841                 mtx_unlock(&pr->pr_mtx);
 2842                 return (EAFNOSUPPORT);
 2843         }
 2844 
 2845         if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
 2846                 ia->s_addr = pr->pr_ip4[0].s_addr;
 2847                 mtx_unlock(&pr->pr_mtx);
 2848                 return (0);
 2849         }
 2850 
 2851         /*
 2852          * Return success because nothing had to be changed.
 2853          */
 2854         mtx_unlock(&pr->pr_mtx);
 2855         return (0);
 2856 }
 2857 
 2858 /*
 2859  * Check if given address belongs to the jail referenced by cred/prison.
 2860  *
 2861  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
 2862  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 2863  * doesn't allow IPv4.  Address passed in in NBO.
 2864  */
 2865 static int
 2866 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
 2867 {
 2868         int i, a, z, d;
 2869 
 2870         /*
 2871          * Check the primary IP.
 2872          */
 2873         if (pr->pr_ip4[0].s_addr == ia->s_addr)
 2874                 return (0);
 2875 
 2876         /*
 2877          * All the other IPs are sorted so we can do a binary search.
 2878          */
 2879         a = 0;
 2880         z = pr->pr_ip4s - 2;
 2881         while (a <= z) {
 2882                 i = (a + z) / 2;
 2883                 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
 2884                 if (d > 0)
 2885                         z = i - 1;
 2886                 else if (d < 0)
 2887                         a = i + 1;
 2888                 else
 2889                         return (0);
 2890         }
 2891 
 2892         return (EADDRNOTAVAIL);
 2893 }
 2894 
 2895 int
 2896 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
 2897 {
 2898         struct prison *pr;
 2899         int error;
 2900 
 2901         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2902         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2903 
 2904         pr = cred->cr_prison;
 2905         if (!(pr->pr_flags & PR_IP4))
 2906                 return (0);
 2907         mtx_lock(&pr->pr_mtx);
 2908         if (!(pr->pr_flags & PR_IP4)) {
 2909                 mtx_unlock(&pr->pr_mtx);
 2910                 return (0);
 2911         }
 2912         if (pr->pr_ip4 == NULL) {
 2913                 mtx_unlock(&pr->pr_mtx);
 2914                 return (EAFNOSUPPORT);
 2915         }
 2916 
 2917         error = _prison_check_ip4(pr, ia);
 2918         mtx_unlock(&pr->pr_mtx);
 2919         return (error);
 2920 }
 2921 #endif
 2922 
 2923 #ifdef INET6
 2924 static int
 2925 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
 2926 {
 2927         int ii, ij, used;
 2928         struct prison *ppr;
 2929 
 2930         ppr = pr->pr_parent;
 2931         if (!(pr->pr_flags & PR_IP6_USER)) {
 2932                 /* This has no user settings, so just copy the parent's list. */
 2933                 if (pr->pr_ip6s < ppr->pr_ip6s) {
 2934                         /*
 2935                          * There's no room for the parent's list.  Use the
 2936                          * new list buffer, which is assumed to be big enough
 2937                          * (if it was passed).  If there's no buffer, try to
 2938                          * allocate one.
 2939                          */
 2940                         used = 1;
 2941                         if (newip6 == NULL) {
 2942                                 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
 2943                                     M_PRISON, M_NOWAIT);
 2944                                 if (newip6 != NULL)
 2945                                         used = 0;
 2946                         }
 2947                         if (newip6 != NULL) {
 2948                                 bcopy(ppr->pr_ip6, newip6,
 2949                                     ppr->pr_ip6s * sizeof(*newip6));
 2950                                 free(pr->pr_ip6, M_PRISON);
 2951                                 pr->pr_ip6 = newip6;
 2952                                 pr->pr_ip6s = ppr->pr_ip6s;
 2953                         }
 2954                         return (used);
 2955                 }
 2956                 pr->pr_ip6s = ppr->pr_ip6s;
 2957                 if (pr->pr_ip6s > 0)
 2958                         bcopy(ppr->pr_ip6, pr->pr_ip6,
 2959                             pr->pr_ip6s * sizeof(*newip6));
 2960                 else if (pr->pr_ip6 != NULL) {
 2961                         free(pr->pr_ip6, M_PRISON);
 2962                         pr->pr_ip6 = NULL;
 2963                 }
 2964         } else if (pr->pr_ip6s > 0) {
 2965                 /* Remove addresses that aren't in the parent. */
 2966                 for (ij = 0; ij < ppr->pr_ip6s; ij++)
 2967                         if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
 2968                             &ppr->pr_ip6[ij]))
 2969                                 break;
 2970                 if (ij < ppr->pr_ip6s)
 2971                         ii = 1;
 2972                 else {
 2973                         bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
 2974                             --pr->pr_ip6s * sizeof(*pr->pr_ip6));
 2975                         ii = 0;
 2976                 }
 2977                 for (ij = 1; ii < pr->pr_ip6s; ) {
 2978                         if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
 2979                             &ppr->pr_ip6[0])) {
 2980                                 ii++;
 2981                                 continue;
 2982                         }
 2983                         switch (ij >= ppr->pr_ip4s ? -1 :
 2984                                 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
 2985                         case -1:
 2986                                 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
 2987                                     (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
 2988                                 break;
 2989                         case 0:
 2990                                 ii++;
 2991                                 ij++;
 2992                                 break;
 2993                         case 1:
 2994                                 ij++;
 2995                                 break;
 2996                         }
 2997                 }
 2998                 if (pr->pr_ip6s == 0) {
 2999                         pr->pr_flags |= PR_IP6_DISABLE;
 3000                         free(pr->pr_ip6, M_PRISON);
 3001                         pr->pr_ip6 = NULL;
 3002                 }
 3003         }
 3004         return 0;
 3005 }
 3006 
 3007 /*
 3008  * Pass back primary IPv6 address for this jail.
 3009  *
 3010  * If not restricted return success but do not alter the address.  Caller has
 3011  * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
 3012  *
 3013  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
 3014  */
 3015 int
 3016 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
 3017 {
 3018         struct prison *pr;
 3019 
 3020         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3021         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3022 
 3023         pr = cred->cr_prison;
 3024         if (!(pr->pr_flags & PR_IP6))
 3025                 return (0);
 3026         mtx_lock(&pr->pr_mtx);
 3027         if (!(pr->pr_flags & PR_IP6)) {
 3028                 mtx_unlock(&pr->pr_mtx);
 3029                 return (0);
 3030         }
 3031         if (pr->pr_ip6 == NULL) {
 3032                 mtx_unlock(&pr->pr_mtx);
 3033                 return (EAFNOSUPPORT);
 3034         }
 3035 
 3036         bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3037         mtx_unlock(&pr->pr_mtx);
 3038         return (0);
 3039 }
 3040 
 3041 /*
 3042  * Return 1 if we should do proper source address selection or are not jailed.
 3043  * We will return 0 if we should bypass source address selection in favour
 3044  * of the primary jail IPv6 address. Only in this case *ia will be updated and
 3045  * returned in NBO.
 3046  * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
 3047  */
 3048 int
 3049 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
 3050 {
 3051         struct prison *pr;
 3052         struct in6_addr lia6;
 3053         int error;
 3054 
 3055         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3056         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3057 
 3058         if (!jailed(cred))
 3059                 return (1);
 3060 
 3061         pr = cred->cr_prison;
 3062         if (pr->pr_flags & PR_IP6_SADDRSEL)
 3063                 return (1);
 3064 
 3065         lia6 = in6addr_any;
 3066         error = prison_get_ip6(cred, &lia6);
 3067         if (error)
 3068                 return (error);
 3069         if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
 3070                 return (1);
 3071 
 3072         bcopy(&lia6, ia6, sizeof(struct in6_addr));
 3073         return (0);
 3074 }
 3075 
 3076 /*
 3077  * Return true if pr1 and pr2 have the same IPv6 address restrictions.
 3078  */
 3079 int
 3080 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
 3081 {
 3082 
 3083         if (pr1 == pr2)
 3084                 return (1);
 3085 
 3086         while (pr1 != &prison0 &&
 3087 #ifdef VIMAGE
 3088                !(pr1->pr_flags & PR_VNET) &&
 3089 #endif
 3090                !(pr1->pr_flags & PR_IP6_USER))
 3091                 pr1 = pr1->pr_parent;
 3092         while (pr2 != &prison0 &&
 3093 #ifdef VIMAGE
 3094                !(pr2->pr_flags & PR_VNET) &&
 3095 #endif
 3096                !(pr2->pr_flags & PR_IP6_USER))
 3097                 pr2 = pr2->pr_parent;
 3098         return (pr1 == pr2);
 3099 }
 3100 
 3101 /*
 3102  * Make sure our (source) address is set to something meaningful to this jail.
 3103  *
 3104  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
 3105  * when needed while binding.
 3106  *
 3107  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
 3108  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 3109  * doesn't allow IPv6.
 3110  */
 3111 int
 3112 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
 3113 {
 3114         struct prison *pr;
 3115         int error;
 3116 
 3117         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3118         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3119 
 3120         pr = cred->cr_prison;
 3121         if (!(pr->pr_flags & PR_IP6))
 3122                 return (0);
 3123         mtx_lock(&pr->pr_mtx);
 3124         if (!(pr->pr_flags & PR_IP6)) {
 3125                 mtx_unlock(&pr->pr_mtx);
 3126                 return (0);
 3127         }
 3128         if (pr->pr_ip6 == NULL) {
 3129                 mtx_unlock(&pr->pr_mtx);
 3130                 return (EAFNOSUPPORT);
 3131         }
 3132 
 3133         if (IN6_IS_ADDR_LOOPBACK(ia6)) {
 3134                 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3135                 mtx_unlock(&pr->pr_mtx);
 3136                 return (0);
 3137         }
 3138 
 3139         if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
 3140                 /*
 3141                  * In case there is only 1 IPv6 address, and v6only is true,
 3142                  * then bind directly.
 3143                  */
 3144                 if (v6only != 0 && pr->pr_ip6s == 1)
 3145                         bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3146                 mtx_unlock(&pr->pr_mtx);
 3147                 return (0);
 3148         }
 3149 
 3150         error = _prison_check_ip6(pr, ia6);
 3151         mtx_unlock(&pr->pr_mtx);
 3152         return (error);
 3153 }
 3154 
 3155 /*
 3156  * Rewrite destination address in case we will connect to loopback address.
 3157  *
 3158  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
 3159  */
 3160 int
 3161 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
 3162 {
 3163         struct prison *pr;
 3164 
 3165         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3166         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3167 
 3168         pr = cred->cr_prison;
 3169         if (!(pr->pr_flags & PR_IP6))
 3170                 return (0);
 3171         mtx_lock(&pr->pr_mtx);
 3172         if (!(pr->pr_flags & PR_IP6)) {
 3173                 mtx_unlock(&pr->pr_mtx);
 3174                 return (0);
 3175         }
 3176         if (pr->pr_ip6 == NULL) {
 3177                 mtx_unlock(&pr->pr_mtx);
 3178                 return (EAFNOSUPPORT);
 3179         }
 3180 
 3181         if (IN6_IS_ADDR_LOOPBACK(ia6)) {
 3182                 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3183                 mtx_unlock(&pr->pr_mtx);
 3184                 return (0);
 3185         }
 3186 
 3187         /*
 3188          * Return success because nothing had to be changed.
 3189          */
 3190         mtx_unlock(&pr->pr_mtx);
 3191         return (0);
 3192 }
 3193 
 3194 /*
 3195  * Check if given address belongs to the jail referenced by cred/prison.
 3196  *
 3197  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
 3198  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 3199  * doesn't allow IPv6.
 3200  */
 3201 static int
 3202 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
 3203 {
 3204         int i, a, z, d;
 3205 
 3206         /*
 3207          * Check the primary IP.
 3208          */
 3209         if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
 3210                 return (0);
 3211 
 3212         /*
 3213          * All the other IPs are sorted so we can do a binary search.
 3214          */
 3215         a = 0;
 3216         z = pr->pr_ip6s - 2;
 3217         while (a <= z) {
 3218                 i = (a + z) / 2;
 3219                 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
 3220                 if (d > 0)
 3221                         z = i - 1;
 3222                 else if (d < 0)
 3223                         a = i + 1;
 3224                 else
 3225                         return (0);
 3226         }
 3227 
 3228         return (EADDRNOTAVAIL);
 3229 }
 3230 
 3231 int
 3232 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
 3233 {
 3234         struct prison *pr;
 3235         int error;
 3236 
 3237         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3238         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3239 
 3240         pr = cred->cr_prison;
 3241         if (!(pr->pr_flags & PR_IP6))
 3242                 return (0);
 3243         mtx_lock(&pr->pr_mtx);
 3244         if (!(pr->pr_flags & PR_IP6)) {
 3245                 mtx_unlock(&pr->pr_mtx);
 3246                 return (0);
 3247         }
 3248         if (pr->pr_ip6 == NULL) {
 3249                 mtx_unlock(&pr->pr_mtx);
 3250                 return (EAFNOSUPPORT);
 3251         }
 3252 
 3253         error = _prison_check_ip6(pr, ia6);
 3254         mtx_unlock(&pr->pr_mtx);
 3255         return (error);
 3256 }
 3257 #endif
 3258 
 3259 /*
 3260  * Check if a jail supports the given address family.
 3261  *
 3262  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
 3263  * if not.
 3264  */
 3265 int
 3266 prison_check_af(struct ucred *cred, int af)
 3267 {
 3268         struct prison *pr;
 3269         int error;
 3270 
 3271         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3272 
 3273         pr = cred->cr_prison;
 3274 #ifdef VIMAGE
 3275         /* Prisons with their own network stack are not limited. */
 3276         if (prison_owns_vnet(cred))
 3277                 return (0);
 3278 #endif
 3279 
 3280         error = 0;
 3281         switch (af)
 3282         {
 3283 #ifdef INET
 3284         case AF_INET:
 3285                 if (pr->pr_flags & PR_IP4)
 3286                 {
 3287                         mtx_lock(&pr->pr_mtx);
 3288                         if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
 3289                                 error = EAFNOSUPPORT;
 3290                         mtx_unlock(&pr->pr_mtx);
 3291                 }
 3292                 break;
 3293 #endif
 3294 #ifdef INET6
 3295         case AF_INET6:
 3296                 if (pr->pr_flags & PR_IP6)
 3297                 {
 3298                         mtx_lock(&pr->pr_mtx);
 3299                         if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
 3300                                 error = EAFNOSUPPORT;
 3301                         mtx_unlock(&pr->pr_mtx);
 3302                 }
 3303                 break;
 3304 #endif
 3305         case AF_LOCAL:
 3306         case AF_ROUTE:
 3307                 break;
 3308         default:
 3309                 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
 3310                         error = EAFNOSUPPORT;
 3311         }
 3312         return (error);
 3313 }
 3314 
 3315 /*
 3316  * Check if given address belongs to the jail referenced by cred (wrapper to
 3317  * prison_check_ip[46]).
 3318  *
 3319  * Returns 0 if jail doesn't restrict the address family or if address belongs
 3320  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
 3321  * the jail doesn't allow the address family.  IPv4 Address passed in in NBO.
 3322  */
 3323 int
 3324 prison_if(struct ucred *cred, struct sockaddr *sa)
 3325 {
 3326 #ifdef INET
 3327         struct sockaddr_in *sai;
 3328 #endif
 3329 #ifdef INET6
 3330         struct sockaddr_in6 *sai6;
 3331 #endif
 3332         int error;
 3333 
 3334         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3335         KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
 3336 
 3337 #ifdef VIMAGE
 3338         if (prison_owns_vnet(cred))
 3339                 return (0);
 3340 #endif
 3341 
 3342         error = 0;
 3343         switch (sa->sa_family)
 3344         {
 3345 #ifdef INET
 3346         case AF_INET:
 3347                 sai = (struct sockaddr_in *)sa;
 3348                 error = prison_check_ip4(cred, &sai->sin_addr);
 3349                 break;
 3350 #endif
 3351 #ifdef INET6
 3352         case AF_INET6:
 3353                 sai6 = (struct sockaddr_in6 *)sa;
 3354                 error = prison_check_ip6(cred, &sai6->sin6_addr);
 3355                 break;
 3356 #endif
 3357         default:
 3358                 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
 3359                         error = EAFNOSUPPORT;
 3360         }
 3361         return (error);
 3362 }
 3363 
 3364 /*
 3365  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
 3366  */
 3367 int
 3368 prison_check(struct ucred *cred1, struct ucred *cred2)
 3369 {
 3370 
 3371         return ((cred1->cr_prison == cred2->cr_prison ||
 3372             prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
 3373 }
 3374 
 3375 /*
 3376  * Return 1 if p2 is a child of p1, otherwise 0.
 3377  */
 3378 int
 3379 prison_ischild(struct prison *pr1, struct prison *pr2)
 3380 {
 3381 
 3382         for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
 3383                 if (pr1 == pr2)
 3384                         return (1);
 3385         return (0);
 3386 }
 3387 
 3388 /*
 3389  * Return 1 if the passed credential is in a jail, otherwise 0.
 3390  */
 3391 int
 3392 jailed(struct ucred *cred)
 3393 {
 3394 
 3395         return (cred->cr_prison != &prison0);
 3396 }
 3397 
 3398 /*
 3399  * Return 1 if the passed credential is in a jail and that jail does not
 3400  * have its own virtual network stack, otherwise 0.
 3401  */
 3402 int
 3403 jailed_without_vnet(struct ucred *cred)
 3404 {
 3405 
 3406         if (!jailed(cred))
 3407                 return (0);
 3408 #ifdef VIMAGE
 3409         if (prison_owns_vnet(cred))
 3410                 return (0);
 3411 #endif
 3412 
 3413         return (1);
 3414 }
 3415 
 3416 /*
 3417  * Return the correct hostname (domainname, et al) for the passed credential.
 3418  */
 3419 void
 3420 getcredhostname(struct ucred *cred, char *buf, size_t size)
 3421 {
 3422         struct prison *pr;
 3423 
 3424         /*
 3425          * A NULL credential can be used to shortcut to the physical
 3426          * system's hostname.
 3427          */
 3428         pr = (cred != NULL) ? cred->cr_prison : &prison0;
 3429         mtx_lock(&pr->pr_mtx);
 3430         strlcpy(buf, pr->pr_hostname, size);
 3431         mtx_unlock(&pr->pr_mtx);
 3432 }
 3433 
 3434 void
 3435 getcreddomainname(struct ucred *cred, char *buf, size_t size)
 3436 {
 3437 
 3438         mtx_lock(&cred->cr_prison->pr_mtx);
 3439         strlcpy(buf, cred->cr_prison->pr_domainname, size);
 3440         mtx_unlock(&cred->cr_prison->pr_mtx);
 3441 }
 3442 
 3443 void
 3444 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
 3445 {
 3446 
 3447         mtx_lock(&cred->cr_prison->pr_mtx);
 3448         strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
 3449         mtx_unlock(&cred->cr_prison->pr_mtx);
 3450 }
 3451 
 3452 void
 3453 getcredhostid(struct ucred *cred, unsigned long *hostid)
 3454 {
 3455 
 3456         mtx_lock(&cred->cr_prison->pr_mtx);
 3457         *hostid = cred->cr_prison->pr_hostid;
 3458         mtx_unlock(&cred->cr_prison->pr_mtx);
 3459 }
 3460 
 3461 #ifdef VIMAGE
 3462 /*
 3463  * Determine whether the prison represented by cred owns
 3464  * its vnet rather than having it inherited.
 3465  *
 3466  * Returns 1 in case the prison owns the vnet, 0 otherwise.
 3467  */
 3468 int
 3469 prison_owns_vnet(struct ucred *cred)
 3470 {
 3471 
 3472         /*
 3473          * vnets cannot be added/removed after jail creation,
 3474          * so no need to lock here.
 3475          */
 3476         return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
 3477 }
 3478 #endif
 3479 
 3480 /*
 3481  * Determine whether the subject represented by cred can "see"
 3482  * status of a mount point.
 3483  * Returns: 0 for permitted, ENOENT otherwise.
 3484  * XXX: This function should be called cr_canseemount() and should be
 3485  *      placed in kern_prot.c.
 3486  */
 3487 int
 3488 prison_canseemount(struct ucred *cred, struct mount *mp)
 3489 {
 3490         struct prison *pr;
 3491         struct statfs *sp;
 3492         size_t len;
 3493 
 3494         pr = cred->cr_prison;
 3495         if (pr->pr_enforce_statfs == 0)
 3496                 return (0);
 3497         if (pr->pr_root->v_mount == mp)
 3498                 return (0);
 3499         if (pr->pr_enforce_statfs == 2)
 3500                 return (ENOENT);
 3501         /*
 3502          * If jail's chroot directory is set to "/" we should be able to see
 3503          * all mount-points from inside a jail.
 3504          * This is ugly check, but this is the only situation when jail's
 3505          * directory ends with '/'.
 3506          */
 3507         if (strcmp(pr->pr_path, "/") == 0)
 3508                 return (0);
 3509         len = strlen(pr->pr_path);
 3510         sp = &mp->mnt_stat;
 3511         if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
 3512                 return (ENOENT);
 3513         /*
 3514          * Be sure that we don't have situation where jail's root directory
 3515          * is "/some/path" and mount point is "/some/pathpath".
 3516          */
 3517         if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
 3518                 return (ENOENT);
 3519         return (0);
 3520 }
 3521 
 3522 void
 3523 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
 3524 {
 3525         char jpath[MAXPATHLEN];
 3526         struct prison *pr;
 3527         size_t len;
 3528 
 3529         pr = cred->cr_prison;
 3530         if (pr->pr_enforce_statfs == 0)
 3531                 return;
 3532         if (prison_canseemount(cred, mp) != 0) {
 3533                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
 3534                 strlcpy(sp->f_mntonname, "[restricted]",
 3535                     sizeof(sp->f_mntonname));
 3536                 return;
 3537         }
 3538         if (pr->pr_root->v_mount == mp) {
 3539                 /*
 3540                  * Clear current buffer data, so we are sure nothing from
 3541                  * the valid path left there.
 3542                  */
 3543                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
 3544                 *sp->f_mntonname = '/';
 3545                 return;
 3546         }
 3547         /*
 3548          * If jail's chroot directory is set to "/" we should be able to see
 3549          * all mount-points from inside a jail.
 3550          */
 3551         if (strcmp(pr->pr_path, "/") == 0)
 3552                 return;
 3553         len = strlen(pr->pr_path);
 3554         strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
 3555         /*
 3556          * Clear current buffer data, so we are sure nothing from
 3557          * the valid path left there.
 3558          */
 3559         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
 3560         if (*jpath == '\0') {
 3561                 /* Should never happen. */
 3562                 *sp->f_mntonname = '/';
 3563         } else {
 3564                 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
 3565         }
 3566 }
 3567 
 3568 /*
 3569  * Check with permission for a specific privilege is granted within jail.  We
 3570  * have a specific list of accepted privileges; the rest are denied.
 3571  */
 3572 int
 3573 prison_priv_check(struct ucred *cred, int priv)
 3574 {
 3575 
 3576         if (!jailed(cred))
 3577                 return (0);
 3578 
 3579 #ifdef VIMAGE
 3580         /*
 3581          * Privileges specific to prisons with a virtual network stack.
 3582          * There might be a duplicate entry here in case the privilege
 3583          * is only granted conditionally in the legacy jail case.
 3584          */
 3585         switch (priv) {
 3586 #ifdef notyet
 3587                 /*
 3588                  * NFS-specific privileges.
 3589                  */
 3590         case PRIV_NFS_DAEMON:
 3591         case PRIV_NFS_LOCKD:
 3592 #endif
 3593                 /*
 3594                  * Network stack privileges.
 3595                  */
 3596         case PRIV_NET_BRIDGE:
 3597         case PRIV_NET_GRE:
 3598         case PRIV_NET_BPF:
 3599         case PRIV_NET_RAW:              /* Dup, cond. in legacy jail case. */
 3600         case PRIV_NET_ROUTE:
 3601         case PRIV_NET_TAP:
 3602         case PRIV_NET_SETIFMTU:
 3603         case PRIV_NET_SETIFFLAGS:
 3604         case PRIV_NET_SETIFCAP:
 3605         case PRIV_NET_SETIFDESCR:
 3606         case PRIV_NET_SETIFNAME :
 3607         case PRIV_NET_SETIFMETRIC:
 3608         case PRIV_NET_SETIFPHYS:
 3609         case PRIV_NET_SETIFMAC:
 3610         case PRIV_NET_ADDMULTI:
 3611         case PRIV_NET_DELMULTI:
 3612         case PRIV_NET_HWIOCTL:
 3613         case PRIV_NET_SETLLADDR:
 3614         case PRIV_NET_ADDIFGROUP:
 3615         case PRIV_NET_DELIFGROUP:
 3616         case PRIV_NET_IFCREATE:
 3617         case PRIV_NET_IFDESTROY:
 3618         case PRIV_NET_ADDIFADDR:
 3619         case PRIV_NET_DELIFADDR:
 3620         case PRIV_NET_LAGG:
 3621         case PRIV_NET_GIF:
 3622         case PRIV_NET_SETIFVNET:
 3623         case PRIV_NET_SETIFFIB:
 3624 
 3625                 /*
 3626                  * 802.11-related privileges.
 3627                  */
 3628         case PRIV_NET80211_GETKEY:
 3629 #ifdef notyet
 3630         case PRIV_NET80211_MANAGE:              /* XXX-BZ discuss with sam@ */
 3631 #endif
 3632 
 3633 #ifdef notyet
 3634                 /*
 3635                  * AppleTalk privileges.
 3636                  */
 3637         case PRIV_NETATALK_RESERVEDPORT:
 3638 
 3639                 /*
 3640                  * ATM privileges.
 3641                  */
 3642         case PRIV_NETATM_CFG:
 3643         case PRIV_NETATM_ADD:
 3644         case PRIV_NETATM_DEL:
 3645         case PRIV_NETATM_SET:
 3646 
 3647                 /*
 3648                  * Bluetooth privileges.
 3649                  */
 3650         case PRIV_NETBLUETOOTH_RAW:
 3651 #endif
 3652 
 3653                 /*
 3654                  * Netgraph and netgraph module privileges.
 3655                  */
 3656         case PRIV_NETGRAPH_CONTROL:
 3657 #ifdef notyet
 3658         case PRIV_NETGRAPH_TTY:
 3659 #endif
 3660 
 3661                 /*
 3662                  * IPv4 and IPv6 privileges.
 3663                  */
 3664         case PRIV_NETINET_IPFW:
 3665         case PRIV_NETINET_DIVERT:
 3666         case PRIV_NETINET_PF:
 3667         case PRIV_NETINET_DUMMYNET:
 3668         case PRIV_NETINET_CARP:
 3669         case PRIV_NETINET_MROUTE:
 3670         case PRIV_NETINET_RAW:
 3671         case PRIV_NETINET_ADDRCTRL6:
 3672         case PRIV_NETINET_ND6:
 3673         case PRIV_NETINET_SCOPE6:
 3674         case PRIV_NETINET_ALIFETIME6:
 3675         case PRIV_NETINET_IPSEC:
 3676         case PRIV_NETINET_BINDANY:
 3677 
 3678 #ifdef notyet
 3679                 /*
 3680                  * IPX/SPX privileges.
 3681                  */
 3682         case PRIV_NETIPX_RESERVEDPORT:
 3683         case PRIV_NETIPX_RAW:
 3684 
 3685                 /*
 3686                  * NCP privileges.
 3687                  */
 3688         case PRIV_NETNCP:
 3689 
 3690                 /*
 3691                  * SMB privileges.
 3692                  */
 3693         case PRIV_NETSMB:
 3694 #endif
 3695 
 3696         /*
 3697          * No default: or deny here.
 3698          * In case of no permit fall through to next switch().
 3699          */
 3700                 if (cred->cr_prison->pr_flags & PR_VNET)
 3701                         return (0);
 3702         }
 3703 #endif /* VIMAGE */
 3704 
 3705         switch (priv) {
 3706 
 3707                 /*
 3708                  * Allow ktrace privileges for root in jail.
 3709                  */
 3710         case PRIV_KTRACE:
 3711 
 3712 #if 0
 3713                 /*
 3714                  * Allow jailed processes to configure audit identity and
 3715                  * submit audit records (login, etc).  In the future we may
 3716                  * want to further refine the relationship between audit and
 3717                  * jail.
 3718                  */
 3719         case PRIV_AUDIT_GETAUDIT:
 3720         case PRIV_AUDIT_SETAUDIT:
 3721         case PRIV_AUDIT_SUBMIT:
 3722 #endif
 3723 
 3724                 /*
 3725                  * Allow jailed processes to manipulate process UNIX
 3726                  * credentials in any way they see fit.
 3727                  */
 3728         case PRIV_CRED_SETUID:
 3729         case PRIV_CRED_SETEUID:
 3730         case PRIV_CRED_SETGID:
 3731         case PRIV_CRED_SETEGID:
 3732         case PRIV_CRED_SETGROUPS:
 3733         case PRIV_CRED_SETREUID:
 3734         case PRIV_CRED_SETREGID:
 3735         case PRIV_CRED_SETRESUID:
 3736         case PRIV_CRED_SETRESGID:
 3737 
 3738                 /*
 3739                  * Jail implements visibility constraints already, so allow
 3740                  * jailed root to override uid/gid-based constraints.
 3741                  */
 3742         case PRIV_SEEOTHERGIDS:
 3743         case PRIV_SEEOTHERUIDS:
 3744 
 3745                 /*
 3746                  * Jail implements inter-process debugging limits already, so
 3747                  * allow jailed root various debugging privileges.
 3748                  */
 3749         case PRIV_DEBUG_DIFFCRED:
 3750         case PRIV_DEBUG_SUGID:
 3751         case PRIV_DEBUG_UNPRIV:
 3752 
 3753                 /*
 3754                  * Allow jail to set various resource limits and login
 3755                  * properties, and for now, exceed process resource limits.
 3756                  */
 3757         case PRIV_PROC_LIMIT:
 3758         case PRIV_PROC_SETLOGIN:
 3759         case PRIV_PROC_SETRLIMIT:
 3760 
 3761                 /*
 3762                  * System V and POSIX IPC privileges are granted in jail.
 3763                  */
 3764         case PRIV_IPC_READ:
 3765         case PRIV_IPC_WRITE:
 3766         case PRIV_IPC_ADMIN:
 3767         case PRIV_IPC_MSGSIZE:
 3768         case PRIV_MQ_ADMIN:
 3769 
 3770                 /*
 3771                  * Jail operations within a jail work on child jails.
 3772                  */
 3773         case PRIV_JAIL_ATTACH:
 3774         case PRIV_JAIL_SET:
 3775         case PRIV_JAIL_REMOVE:
 3776 
 3777                 /*
 3778                  * Jail implements its own inter-process limits, so allow
 3779                  * root processes in jail to change scheduling on other
 3780                  * processes in the same jail.  Likewise for signalling.
 3781                  */
 3782         case PRIV_SCHED_DIFFCRED:
 3783         case PRIV_SCHED_CPUSET:
 3784         case PRIV_SIGNAL_DIFFCRED:
 3785         case PRIV_SIGNAL_SUGID:
 3786 
 3787                 /*
 3788                  * Allow jailed processes to write to sysctls marked as jail
 3789                  * writable.
 3790                  */
 3791         case PRIV_SYSCTL_WRITEJAIL:
 3792 
 3793                 /*
 3794                  * Allow root in jail to manage a variety of quota
 3795                  * properties.  These should likely be conditional on a
 3796                  * configuration option.
 3797                  */
 3798         case PRIV_VFS_GETQUOTA:
 3799         case PRIV_VFS_SETQUOTA:
 3800 
 3801                 /*
 3802                  * Since Jail relies on chroot() to implement file system
 3803                  * protections, grant many VFS privileges to root in jail.
 3804                  * Be careful to exclude mount-related and NFS-related
 3805                  * privileges.
 3806                  */
 3807         case PRIV_VFS_READ:
 3808         case PRIV_VFS_WRITE:
 3809         case PRIV_VFS_ADMIN:
 3810         case PRIV_VFS_EXEC:
 3811         case PRIV_VFS_LOOKUP:
 3812         case PRIV_VFS_BLOCKRESERVE:     /* XXXRW: Slightly surprising. */
 3813         case PRIV_VFS_CHFLAGS_DEV:
 3814         case PRIV_VFS_CHOWN:
 3815         case PRIV_VFS_CHROOT:
 3816         case PRIV_VFS_RETAINSUGID:
 3817         case PRIV_VFS_FCHROOT:
 3818         case PRIV_VFS_LINK:
 3819         case PRIV_VFS_SETGID:
 3820         case PRIV_VFS_STAT:
 3821         case PRIV_VFS_STICKYFILE:
 3822                 return (0);
 3823 
 3824                 /*
 3825                  * Depending on the global setting, allow privilege of
 3826                  * setting system flags.
 3827                  */
 3828         case PRIV_VFS_SYSFLAGS:
 3829                 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
 3830                         return (0);
 3831                 else
 3832                         return (EPERM);
 3833 
 3834                 /*
 3835                  * Depending on the global setting, allow privilege of
 3836                  * mounting/unmounting file systems.
 3837                  */
 3838         case PRIV_VFS_MOUNT:
 3839         case PRIV_VFS_UNMOUNT:
 3840         case PRIV_VFS_MOUNT_NONUSER:
 3841         case PRIV_VFS_MOUNT_OWNER:
 3842                 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT)
 3843                         return (0);
 3844                 else
 3845                         return (EPERM);
 3846 
 3847                 /*
 3848                  * Allow jailed root to bind reserved ports and reuse in-use
 3849                  * ports.
 3850                  */
 3851         case PRIV_NETINET_RESERVEDPORT:
 3852         case PRIV_NETINET_REUSEPORT:
 3853                 return (0);
 3854 
 3855                 /*
 3856                  * Allow jailed root to set certian IPv4/6 (option) headers.
 3857                  */
 3858         case PRIV_NETINET_SETHDROPTS:
 3859                 return (0);
 3860 
 3861                 /*
 3862                  * Conditionally allow creating raw sockets in jail.
 3863                  */
 3864         case PRIV_NETINET_RAW:
 3865                 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
 3866                         return (0);
 3867                 else
 3868                         return (EPERM);
 3869 
 3870                 /*
 3871                  * Since jail implements its own visibility limits on netstat
 3872                  * sysctls, allow getcred.  This allows identd to work in
 3873                  * jail.
 3874                  */
 3875         case PRIV_NETINET_GETCRED:
 3876                 return (0);
 3877 
 3878         default:
 3879                 /*
 3880                  * In all remaining cases, deny the privilege request.  This
 3881                  * includes almost all network privileges, many system
 3882                  * configuration privileges.
 3883                  */
 3884                 return (EPERM);
 3885         }
 3886 }
 3887 
 3888 /*
 3889  * Return the part of pr2's name that is relative to pr1, or the whole name
 3890  * if it does not directly follow.
 3891  */
 3892 
 3893 char *
 3894 prison_name(struct prison *pr1, struct prison *pr2)
 3895 {
 3896         char *name;
 3897 
 3898         /* Jails see themselves as "" (if they see themselves at all). */
 3899         if (pr1 == pr2)
 3900                 return "";
 3901         name = pr2->pr_name;
 3902         if (prison_ischild(pr1, pr2)) {
 3903                 /*
 3904                  * pr1 isn't locked (and allprison_lock may not be either)
 3905                  * so its length can't be counted on.  But the number of dots
 3906                  * can be counted on - and counted.
 3907                  */
 3908                 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
 3909                         name = strchr(name, '.') + 1;
 3910         }
 3911         return (name);
 3912 }
 3913 
 3914 /*
 3915  * Return the part of pr2's path that is relative to pr1, or the whole path
 3916  * if it does not directly follow.
 3917  */
 3918 static char *
 3919 prison_path(struct prison *pr1, struct prison *pr2)
 3920 {
 3921         char *path1, *path2;
 3922         int len1;
 3923 
 3924         path1 = pr1->pr_path;
 3925         path2 = pr2->pr_path;
 3926         if (!strcmp(path1, "/"))
 3927                 return (path2);
 3928         len1 = strlen(path1);
 3929         if (strncmp(path1, path2, len1))
 3930                 return (path2);
 3931         if (path2[len1] == '\0')
 3932                 return "/";
 3933         if (path2[len1] == '/')
 3934                 return (path2 + len1);
 3935         return (path2);
 3936 }
 3937 
 3938 
 3939 /*
 3940  * Jail-related sysctls.
 3941  */
 3942 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
 3943     "Jails");
 3944 
 3945 static int
 3946 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
 3947 {
 3948         struct xprison *xp;
 3949         struct prison *pr, *cpr;
 3950 #ifdef INET
 3951         struct in_addr *ip4 = NULL;
 3952         int ip4s = 0;
 3953 #endif
 3954 #ifdef INET6
 3955         struct in_addr *ip6 = NULL;
 3956         int ip6s = 0;
 3957 #endif
 3958         int descend, error;
 3959 
 3960         xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
 3961         pr = req->td->td_ucred->cr_prison;
 3962         error = 0;
 3963         sx_slock(&allprison_lock);
 3964         FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
 3965 #if defined(INET) || defined(INET6)
 3966  again:
 3967 #endif
 3968                 mtx_lock(&cpr->pr_mtx);
 3969 #ifdef INET
 3970                 if (cpr->pr_ip4s > 0) {
 3971                         if (ip4s < cpr->pr_ip4s) {
 3972                                 ip4s = cpr->pr_ip4s;
 3973                                 mtx_unlock(&cpr->pr_mtx);
 3974                                 ip4 = realloc(ip4, ip4s *
 3975                                     sizeof(struct in_addr), M_TEMP, M_WAITOK);
 3976                                 goto again;
 3977                         }
 3978                         bcopy(cpr->pr_ip4, ip4,
 3979                             cpr->pr_ip4s * sizeof(struct in_addr));
 3980                 }
 3981 #endif
 3982 #ifdef INET6
 3983                 if (cpr->pr_ip6s > 0) {
 3984                         if (ip6s < cpr->pr_ip6s) {
 3985                                 ip6s = cpr->pr_ip6s;
 3986                                 mtx_unlock(&cpr->pr_mtx);
 3987                                 ip6 = realloc(ip6, ip6s *
 3988                                     sizeof(struct in6_addr), M_TEMP, M_WAITOK);
 3989                                 goto again;
 3990                         }
 3991                         bcopy(cpr->pr_ip6, ip6,
 3992                             cpr->pr_ip6s * sizeof(struct in6_addr));
 3993                 }
 3994 #endif
 3995                 if (cpr->pr_ref == 0) {
 3996                         mtx_unlock(&cpr->pr_mtx);
 3997                         continue;
 3998                 }
 3999                 bzero(xp, sizeof(*xp));
 4000                 xp->pr_version = XPRISON_VERSION;
 4001                 xp->pr_id = cpr->pr_id;
 4002                 xp->pr_state = cpr->pr_uref > 0
 4003                     ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
 4004                 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
 4005                 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
 4006                 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
 4007 #ifdef INET
 4008                 xp->pr_ip4s = cpr->pr_ip4s;
 4009 #endif
 4010 #ifdef INET6
 4011                 xp->pr_ip6s = cpr->pr_ip6s;
 4012 #endif
 4013                 mtx_unlock(&cpr->pr_mtx);
 4014                 error = SYSCTL_OUT(req, xp, sizeof(*xp));
 4015                 if (error)
 4016                         break;
 4017 #ifdef INET
 4018                 if (xp->pr_ip4s > 0) {
 4019                         error = SYSCTL_OUT(req, ip4,
 4020                             xp->pr_ip4s * sizeof(struct in_addr));
 4021                         if (error)
 4022                                 break;
 4023                 }
 4024 #endif
 4025 #ifdef INET6
 4026                 if (xp->pr_ip6s > 0) {
 4027                         error = SYSCTL_OUT(req, ip6,
 4028                             xp->pr_ip6s * sizeof(struct in6_addr));
 4029                         if (error)
 4030                                 break;
 4031                 }
 4032 #endif
 4033         }
 4034         sx_sunlock(&allprison_lock);
 4035         free(xp, M_TEMP);
 4036 #ifdef INET
 4037         free(ip4, M_TEMP);
 4038 #endif
 4039 #ifdef INET6
 4040         free(ip6, M_TEMP);
 4041 #endif
 4042         return (error);
 4043 }
 4044 
 4045 SYSCTL_OID(_security_jail, OID_AUTO, list,
 4046     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
 4047     sysctl_jail_list, "S", "List of active jails");
 4048 
 4049 static int
 4050 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
 4051 {
 4052         int error, injail;
 4053 
 4054         injail = jailed(req->td->td_ucred);
 4055         error = SYSCTL_OUT(req, &injail, sizeof(injail));
 4056 
 4057         return (error);
 4058 }
 4059 
 4060 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
 4061     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
 4062     sysctl_jail_jailed, "I", "Process in jail?");
 4063 
 4064 #if defined(INET) || defined(INET6)
 4065 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
 4066     &jail_max_af_ips, 0,
 4067     "Number of IP addresses a jail may have at most per address family");
 4068 #endif
 4069 
 4070 /*
 4071  * Default parameters for jail(2) compatability.  For historical reasons,
 4072  * the sysctl names have varying similarity to the parameter names.  Prisons
 4073  * just see their own parameters, and can't change them.
 4074  */
 4075 static int
 4076 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
 4077 {
 4078         struct prison *pr;
 4079         int allow, error, i;
 4080 
 4081         pr = req->td->td_ucred->cr_prison;
 4082         allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
 4083 
 4084         /* Get the current flag value, and convert it to a boolean. */
 4085         i = (allow & arg2) ? 1 : 0;
 4086         if (arg1 != NULL)
 4087                 i = !i;
 4088         error = sysctl_handle_int(oidp, &i, 0, req);
 4089         if (error || !req->newptr)
 4090                 return (error);
 4091         i = i ? arg2 : 0;
 4092         if (arg1 != NULL)
 4093                 i ^= arg2;
 4094         /*
 4095          * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
 4096          * for writing.
 4097          */
 4098         mtx_lock(&prison0.pr_mtx);
 4099         jail_default_allow = (jail_default_allow & ~arg2) | i;
 4100         mtx_unlock(&prison0.pr_mtx);
 4101         return (0);
 4102 }
 4103 
 4104 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
 4105     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4106     NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
 4107     "Processes in jail can set their hostnames");
 4108 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
 4109     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4110     (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
 4111     "Processes in jail are limited to creating UNIX/IP/route sockets only");
 4112 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
 4113     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4114     NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
 4115     "Processes in jail can use System V IPC primitives");
 4116 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
 4117     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4118     NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
 4119     "Prison root can create raw sockets");
 4120 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
 4121     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4122     NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
 4123     "Processes in jail can alter system file flags");
 4124 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
 4125     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4126     NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
 4127     "Processes in jail can mount/unmount jail-friendly file systems");
 4128 
 4129 static int
 4130 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
 4131 {
 4132         struct prison *pr;
 4133         int level, error;
 4134 
 4135         pr = req->td->td_ucred->cr_prison;
 4136         level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
 4137         error = sysctl_handle_int(oidp, &level, 0, req);
 4138         if (error || !req->newptr)
 4139                 return (error);
 4140         *(int *)arg1 = level;
 4141         return (0);
 4142 }
 4143 
 4144 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
 4145     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4146     &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
 4147     sysctl_jail_default_level, "I",
 4148     "Processes in jail cannot see all mounted file systems");
 4149 
 4150 /*
 4151  * Nodes to describe jail parameters.  Maximum length of string parameters
 4152  * is returned in the string itself, and the other parameters exist merely
 4153  * to make themselves and their types known.
 4154  */
 4155 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
 4156     "Jail parameters");
 4157 
 4158 int
 4159 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
 4160 {
 4161         int i;
 4162         long l;
 4163         size_t s;
 4164         char numbuf[12];
 4165 
 4166         switch (oidp->oid_kind & CTLTYPE)
 4167         {
 4168         case CTLTYPE_LONG:
 4169         case CTLTYPE_ULONG:
 4170                 l = 0;
 4171 #ifdef SCTL_MASK32
 4172                 if (!(req->flags & SCTL_MASK32))
 4173 #endif
 4174                         return (SYSCTL_OUT(req, &l, sizeof(l)));
 4175         case CTLTYPE_INT:
 4176         case CTLTYPE_UINT:
 4177                 i = 0;
 4178                 return (SYSCTL_OUT(req, &i, sizeof(i)));
 4179         case CTLTYPE_STRING:
 4180                 snprintf(numbuf, sizeof(numbuf), "%d", arg2);
 4181                 return
 4182                     (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
 4183         case CTLTYPE_STRUCT:
 4184                 s = (size_t)arg2;
 4185                 return (SYSCTL_OUT(req, &s, sizeof(s)));
 4186         }
 4187         return (0);
 4188 }
 4189 
 4190 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
 4191 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
 4192 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
 4193 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
 4194 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
 4195     "I", "Jail secure level");
 4196 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
 4197     "I", "Jail cannot see all mounted file systems");
 4198 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
 4199     "B", "Jail persistence");
 4200 #ifdef VIMAGE
 4201 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
 4202     "E,jailsys", "Virtual network stack");
 4203 #endif
 4204 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
 4205     "B", "Jail is in the process of shutting down");
 4206 
 4207 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
 4208 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
 4209     "I", "Current number of child jails");
 4210 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
 4211     "I", "Maximum number of child jails");
 4212 
 4213 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
 4214 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
 4215     "Jail hostname");
 4216 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
 4217     "Jail NIS domainname");
 4218 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
 4219     "Jail host UUID");
 4220 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
 4221     "LU", "Jail host ID");
 4222 
 4223 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
 4224 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
 4225 
 4226 #ifdef INET
 4227 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
 4228     "Jail IPv4 address virtualization");
 4229 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
 4230     "S,in_addr,a", "Jail IPv4 addresses");
 4231 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
 4232     "B", "Do (not) use IPv4 source address selection rather than the "
 4233     "primary jail IPv4 address.");
 4234 #endif
 4235 #ifdef INET6
 4236 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
 4237     "Jail IPv6 address virtualization");
 4238 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
 4239     "S,in6_addr,a", "Jail IPv6 addresses");
 4240 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
 4241     "B", "Do (not) use IPv6 source address selection rather than the "
 4242     "primary jail IPv6 address.");
 4243 #endif
 4244 
 4245 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
 4246 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
 4247     "B", "Jail may set hostname");
 4248 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
 4249     "B", "Jail may use SYSV IPC");
 4250 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
 4251     "B", "Jail may create raw sockets");
 4252 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
 4253     "B", "Jail may alter system file flags");
 4254 SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
 4255     "B", "Jail may mount/unmount jail-friendly file systems");
 4256 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
 4257     "B", "Jail may set file quotas");
 4258 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
 4259     "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
 4260 
 4261 
 4262 #ifdef DDB
 4263 
 4264 static void
 4265 db_show_prison(struct prison *pr)
 4266 {
 4267         int fi;
 4268 #if defined(INET) || defined(INET6)
 4269         int ii;
 4270 #endif
 4271         unsigned jsf;
 4272 #ifdef INET6
 4273         char ip6buf[INET6_ADDRSTRLEN];
 4274 #endif
 4275 
 4276         db_printf("prison %p:\n", pr);
 4277         db_printf(" jid             = %d\n", pr->pr_id);
 4278         db_printf(" name            = %s\n", pr->pr_name);
 4279         db_printf(" parent          = %p\n", pr->pr_parent);
 4280         db_printf(" ref             = %d\n", pr->pr_ref);
 4281         db_printf(" uref            = %d\n", pr->pr_uref);
 4282         db_printf(" path            = %s\n", pr->pr_path);
 4283         db_printf(" cpuset          = %d\n", pr->pr_cpuset
 4284             ? pr->pr_cpuset->cs_id : -1);
 4285 #ifdef VIMAGE
 4286         db_printf(" vnet            = %p\n", pr->pr_vnet);
 4287 #endif
 4288         db_printf(" root            = %p\n", pr->pr_root);
 4289         db_printf(" securelevel     = %d\n", pr->pr_securelevel);
 4290         db_printf(" children.max    = %d\n", pr->pr_childmax);
 4291         db_printf(" children.cur    = %d\n", pr->pr_childcount);
 4292         db_printf(" child           = %p\n", LIST_FIRST(&pr->pr_children));
 4293         db_printf(" sibling         = %p\n", LIST_NEXT(pr, pr_sibling));
 4294         db_printf(" flags           = 0x%x", pr->pr_flags);
 4295         for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
 4296             fi++)
 4297                 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
 4298                         db_printf(" %s", pr_flag_names[fi]);
 4299         for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
 4300             fi++) {
 4301                 jsf = pr->pr_flags &
 4302                     (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
 4303                 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
 4304                     pr_flag_jailsys[fi].disable && 
 4305                       (jsf == pr_flag_jailsys[fi].disable) ? "disable"
 4306                     : (jsf == pr_flag_jailsys[fi].new) ? "new"
 4307                     : "inherit");
 4308         }
 4309         db_printf(" allow           = 0x%x", pr->pr_allow);
 4310         for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
 4311             fi++)
 4312                 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
 4313                         db_printf(" %s", pr_allow_names[fi]);
 4314         db_printf("\n");
 4315         db_printf(" enforce_statfs  = %d\n", pr->pr_enforce_statfs);
 4316         db_printf(" host.hostname   = %s\n", pr->pr_hostname);
 4317         db_printf(" host.domainname = %s\n", pr->pr_domainname);
 4318         db_printf(" host.hostuuid   = %s\n", pr->pr_hostuuid);
 4319         db_printf(" host.hostid     = %lu\n", pr->pr_hostid);
 4320 #ifdef INET
 4321         db_printf(" ip4s            = %d\n", pr->pr_ip4s);
 4322         for (ii = 0; ii < pr->pr_ip4s; ii++)
 4323                 db_printf(" %s %s\n",
 4324                     ii == 0 ? "ip4.addr        =" : "                 ",
 4325                     inet_ntoa(pr->pr_ip4[ii]));
 4326 #endif
 4327 #ifdef INET6
 4328         db_printf(" ip6s            = %d\n", pr->pr_ip6s);
 4329         for (ii = 0; ii < pr->pr_ip6s; ii++)
 4330                 db_printf(" %s %s\n",
 4331                     ii == 0 ? "ip6.addr        =" : "                 ",
 4332                     ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
 4333 #endif
 4334 }
 4335 
 4336 DB_SHOW_COMMAND(prison, db_show_prison_command)
 4337 {
 4338         struct prison *pr;
 4339 
 4340         if (!have_addr) {
 4341                 /*
 4342                  * Show all prisons in the list, and prison0 which is not
 4343                  * listed.
 4344                  */
 4345                 db_show_prison(&prison0);
 4346                 if (!db_pager_quit) {
 4347                         TAILQ_FOREACH(pr, &allprison, pr_list) {
 4348                                 db_show_prison(pr);
 4349                                 if (db_pager_quit)
 4350                                         break;
 4351                         }
 4352                 }
 4353                 return;
 4354         }
 4355 
 4356         if (addr == 0)
 4357                 pr = &prison0;
 4358         else {
 4359                 /* Look for a prison with the ID and with references. */
 4360                 TAILQ_FOREACH(pr, &allprison, pr_list)
 4361                         if (pr->pr_id == addr && pr->pr_ref > 0)
 4362                                 break;
 4363                 if (pr == NULL)
 4364                         /* Look again, without requiring a reference. */
 4365                         TAILQ_FOREACH(pr, &allprison, pr_list)
 4366                                 if (pr->pr_id == addr)
 4367                                         break;
 4368                 if (pr == NULL)
 4369                         /* Assume address points to a valid prison. */
 4370                         pr = (struct prison *)addr;
 4371         }
 4372         db_show_prison(pr);
 4373 }
 4374 
 4375 #endif /* DDB */

Cache object: 5f0b53bc1880b1a9403a22473bb12b91


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.