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-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
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/9.0/sys/kern/kern_jail.c 225617 2011-09-16 13:58:51Z kmacy $");
   31 
   32 #include "opt_compat.h"
   33 #include "opt_ddb.h"
   34 #include "opt_inet.h"
   35 #include "opt_inet6.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/types.h>
   39 #include <sys/kernel.h>
   40 #include <sys/systm.h>
   41 #include <sys/errno.h>
   42 #include <sys/sysproto.h>
   43 #include <sys/malloc.h>
   44 #include <sys/osd.h>
   45 #include <sys/priv.h>
   46 #include <sys/proc.h>
   47 #include <sys/taskqueue.h>
   48 #include <sys/fcntl.h>
   49 #include <sys/jail.h>
   50 #include <sys/lock.h>
   51 #include <sys/mutex.h>
   52 #include <sys/racct.h>
   53 #include <sys/refcount.h>
   54 #include <sys/sx.h>
   55 #include <sys/sysent.h>
   56 #include <sys/namei.h>
   57 #include <sys/mount.h>
   58 #include <sys/queue.h>
   59 #include <sys/socket.h>
   60 #include <sys/syscallsubr.h>
   61 #include <sys/sysctl.h>
   62 #include <sys/vnode.h>
   63 
   64 #include <net/if.h>
   65 #include <net/vnet.h>
   66 
   67 #include <netinet/in.h>
   68 
   69 #ifdef DDB
   70 #include <ddb/ddb.h>
   71 #ifdef INET6
   72 #include <netinet6/in6_var.h>
   73 #endif /* INET6 */
   74 #endif /* DDB */
   75 
   76 #include <security/mac/mac_framework.h>
   77 
   78 #define DEFAULT_HOSTUUID        "00000000-0000-0000-0000-000000000000"
   79 
   80 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures");
   81 MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures");
   82 
   83 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */
   84 #ifdef INET
   85 #ifdef INET6
   86 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL
   87 #else
   88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL
   89 #endif
   90 #else /* !INET */
   91 #ifdef INET6
   92 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL
   93 #else
   94 #define _PR_IP_SADDRSEL 0
   95 #endif
   96 #endif
   97 
   98 /* prison0 describes what is "real" about the system. */
   99 struct prison prison0 = {
  100         .pr_id          = 0,
  101         .pr_name        = "",
  102         .pr_ref         = 1,
  103         .pr_uref        = 1,
  104         .pr_path        = "/",
  105         .pr_securelevel = -1,
  106         .pr_childmax    = JAIL_MAX,
  107         .pr_hostuuid    = DEFAULT_HOSTUUID,
  108         .pr_children    = LIST_HEAD_INITIALIZER(prison0.pr_children),
  109 #ifdef VIMAGE
  110         .pr_flags       = PR_HOST|PR_VNET|_PR_IP_SADDRSEL,
  111 #else
  112         .pr_flags       = PR_HOST|_PR_IP_SADDRSEL,
  113 #endif
  114         .pr_allow       = PR_ALLOW_ALL,
  115 };
  116 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF);
  117 
  118 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */
  119 struct  sx allprison_lock;
  120 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison");
  121 struct  prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison);
  122 LIST_HEAD(, prison_racct) allprison_racct;
  123 int     lastprid = 0;
  124 
  125 static int do_jail_attach(struct thread *td, struct prison *pr);
  126 static void prison_complete(void *context, int pending);
  127 static void prison_deref(struct prison *pr, int flags);
  128 static char *prison_path(struct prison *pr1, struct prison *pr2);
  129 static void prison_remove_one(struct prison *pr);
  130 #ifdef RACCT
  131 static void prison_racct_attach(struct prison *pr);
  132 static void prison_racct_detach(struct prison *pr);
  133 #endif
  134 #ifdef INET
  135 static int _prison_check_ip4(struct prison *pr, struct in_addr *ia);
  136 static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4);
  137 #endif
  138 #ifdef INET6
  139 static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6);
  140 static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6);
  141 #endif
  142 
  143 /* Flags for prison_deref */
  144 #define PD_DEREF        0x01
  145 #define PD_DEUREF       0x02
  146 #define PD_LOCKED       0x04
  147 #define PD_LIST_SLOCKED 0x08
  148 #define PD_LIST_XLOCKED 0x10
  149 
  150 /*
  151  * Parameter names corresponding to PR_* flag values.  Size values are for kvm
  152  * as we cannot figure out the size of a sparse array, or an array without a
  153  * terminating entry.
  154  */
  155 static char *pr_flag_names[] = {
  156         [0] = "persist",
  157 #ifdef INET
  158         [7] = "ip4.saddrsel",
  159 #endif
  160 #ifdef INET6
  161         [8] = "ip6.saddrsel",
  162 #endif
  163 };
  164 const size_t pr_flag_names_size = sizeof(pr_flag_names);
  165 
  166 static char *pr_flag_nonames[] = {
  167         [0] = "nopersist",
  168 #ifdef INET
  169         [7] = "ip4.nosaddrsel",
  170 #endif
  171 #ifdef INET6
  172         [8] = "ip6.nosaddrsel",
  173 #endif
  174 };
  175 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames);
  176 
  177 struct jailsys_flags {
  178         const char      *name;
  179         unsigned         disable;
  180         unsigned         new;
  181 } pr_flag_jailsys[] = {
  182         { "host", 0, PR_HOST },
  183 #ifdef VIMAGE
  184         { "vnet", 0, PR_VNET },
  185 #endif
  186 #ifdef INET
  187         { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER },
  188 #endif
  189 #ifdef INET6
  190         { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER },
  191 #endif
  192 };
  193 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys);
  194 
  195 static char *pr_allow_names[] = {
  196         "allow.set_hostname",
  197         "allow.sysvipc",
  198         "allow.raw_sockets",
  199         "allow.chflags",
  200         "allow.mount",
  201         "allow.quotas",
  202         "allow.socket_af",
  203 };
  204 const size_t pr_allow_names_size = sizeof(pr_allow_names);
  205 
  206 static char *pr_allow_nonames[] = {
  207         "allow.noset_hostname",
  208         "allow.nosysvipc",
  209         "allow.noraw_sockets",
  210         "allow.nochflags",
  211         "allow.nomount",
  212         "allow.noquotas",
  213         "allow.nosocket_af",
  214 };
  215 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames);
  216 
  217 #define JAIL_DEFAULT_ALLOW              PR_ALLOW_SET_HOSTNAME
  218 #define JAIL_DEFAULT_ENFORCE_STATFS     2
  219 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW;
  220 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
  221 #if defined(INET) || defined(INET6)
  222 static unsigned jail_max_af_ips = 255;
  223 #endif
  224 
  225 #ifdef INET
  226 static int
  227 qcmp_v4(const void *ip1, const void *ip2)
  228 {
  229         in_addr_t iaa, iab;
  230 
  231         /*
  232          * We need to compare in HBO here to get the list sorted as expected
  233          * by the result of the code.  Sorting NBO addresses gives you
  234          * interesting results.  If you do not understand, do not try.
  235          */
  236         iaa = ntohl(((const struct in_addr *)ip1)->s_addr);
  237         iab = ntohl(((const struct in_addr *)ip2)->s_addr);
  238 
  239         /*
  240          * Do not simply return the difference of the two numbers, the int is
  241          * not wide enough.
  242          */
  243         if (iaa > iab)
  244                 return (1);
  245         else if (iaa < iab)
  246                 return (-1);
  247         else
  248                 return (0);
  249 }
  250 #endif
  251 
  252 #ifdef INET6
  253 static int
  254 qcmp_v6(const void *ip1, const void *ip2)
  255 {
  256         const struct in6_addr *ia6a, *ia6b;
  257         int i, rc;
  258 
  259         ia6a = (const struct in6_addr *)ip1;
  260         ia6b = (const struct in6_addr *)ip2;
  261 
  262         rc = 0;
  263         for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) {
  264                 if (ia6a->s6_addr[i] > ia6b->s6_addr[i])
  265                         rc = 1;
  266                 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i])
  267                         rc = -1;
  268         }
  269         return (rc);
  270 }
  271 #endif
  272 
  273 /*
  274  * struct jail_args {
  275  *      struct jail *jail;
  276  * };
  277  */
  278 int
  279 sys_jail(struct thread *td, struct jail_args *uap)
  280 {
  281         uint32_t version;
  282         int error;
  283         struct jail j;
  284 
  285         error = copyin(uap->jail, &version, sizeof(uint32_t));
  286         if (error)
  287                 return (error);
  288 
  289         switch (version) {
  290         case 0:
  291         {
  292                 struct jail_v0 j0;
  293 
  294                 /* FreeBSD single IPv4 jails. */
  295                 bzero(&j, sizeof(struct jail));
  296                 error = copyin(uap->jail, &j0, sizeof(struct jail_v0));
  297                 if (error)
  298                         return (error);
  299                 j.version = j0.version;
  300                 j.path = j0.path;
  301                 j.hostname = j0.hostname;
  302                 j.ip4s = j0.ip_number;
  303                 break;
  304         }
  305 
  306         case 1:
  307                 /*
  308                  * Version 1 was used by multi-IPv4 jail implementations
  309                  * that never made it into the official kernel.
  310                  */
  311                 return (EINVAL);
  312 
  313         case 2: /* JAIL_API_VERSION */
  314                 /* FreeBSD multi-IPv4/IPv6,noIP jails. */
  315                 error = copyin(uap->jail, &j, sizeof(struct jail));
  316                 if (error)
  317                         return (error);
  318                 break;
  319 
  320         default:
  321                 /* Sci-Fi jails are not supported, sorry. */
  322                 return (EINVAL);
  323         }
  324         return (kern_jail(td, &j));
  325 }
  326 
  327 int
  328 kern_jail(struct thread *td, struct jail *j)
  329 {
  330         struct iovec optiov[2 * (4
  331                             + sizeof(pr_allow_names) / sizeof(pr_allow_names[0])
  332 #ifdef INET
  333                             + 1
  334 #endif
  335 #ifdef INET6
  336                             + 1
  337 #endif
  338                             )];
  339         struct uio opt;
  340         char *u_path, *u_hostname, *u_name;
  341 #ifdef INET
  342         uint32_t ip4s;
  343         struct in_addr *u_ip4;
  344 #endif
  345 #ifdef INET6
  346         struct in6_addr *u_ip6;
  347 #endif
  348         size_t tmplen;
  349         int error, enforce_statfs, fi;
  350 
  351         bzero(&optiov, sizeof(optiov));
  352         opt.uio_iov = optiov;
  353         opt.uio_iovcnt = 0;
  354         opt.uio_offset = -1;
  355         opt.uio_resid = -1;
  356         opt.uio_segflg = UIO_SYSSPACE;
  357         opt.uio_rw = UIO_READ;
  358         opt.uio_td = td;
  359 
  360         /* Set permissions for top-level jails from sysctls. */
  361         if (!jailed(td->td_ucred)) {
  362                 for (fi = 0; fi < sizeof(pr_allow_names) /
  363                      sizeof(pr_allow_names[0]); fi++) {
  364                         optiov[opt.uio_iovcnt].iov_base =
  365                             (jail_default_allow & (1 << fi))
  366                             ? pr_allow_names[fi] : pr_allow_nonames[fi];
  367                         optiov[opt.uio_iovcnt].iov_len =
  368                             strlen(optiov[opt.uio_iovcnt].iov_base) + 1;
  369                         opt.uio_iovcnt += 2;
  370                 }
  371                 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs";
  372                 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs");
  373                 opt.uio_iovcnt++;
  374                 enforce_statfs = jail_default_enforce_statfs;
  375                 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs;
  376                 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs);
  377                 opt.uio_iovcnt++;
  378         }
  379 
  380         tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN;
  381 #ifdef INET
  382         ip4s = (j->version == 0) ? 1 : j->ip4s;
  383         if (ip4s > jail_max_af_ips)
  384                 return (EINVAL);
  385         tmplen += ip4s * sizeof(struct in_addr);
  386 #else
  387         if (j->ip4s > 0)
  388                 return (EINVAL);
  389 #endif
  390 #ifdef INET6
  391         if (j->ip6s > jail_max_af_ips)
  392                 return (EINVAL);
  393         tmplen += j->ip6s * sizeof(struct in6_addr);
  394 #else
  395         if (j->ip6s > 0)
  396                 return (EINVAL);
  397 #endif
  398         u_path = malloc(tmplen, M_TEMP, M_WAITOK);
  399         u_hostname = u_path + MAXPATHLEN;
  400         u_name = u_hostname + MAXHOSTNAMELEN;
  401 #ifdef INET
  402         u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN);
  403 #endif
  404 #ifdef INET6
  405 #ifdef INET
  406         u_ip6 = (struct in6_addr *)(u_ip4 + ip4s);
  407 #else
  408         u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN);
  409 #endif
  410 #endif
  411         optiov[opt.uio_iovcnt].iov_base = "path";
  412         optiov[opt.uio_iovcnt].iov_len = sizeof("path");
  413         opt.uio_iovcnt++;
  414         optiov[opt.uio_iovcnt].iov_base = u_path;
  415         error = copyinstr(j->path, u_path, MAXPATHLEN,
  416             &optiov[opt.uio_iovcnt].iov_len);
  417         if (error) {
  418                 free(u_path, M_TEMP);
  419                 return (error);
  420         }
  421         opt.uio_iovcnt++;
  422         optiov[opt.uio_iovcnt].iov_base = "host.hostname";
  423         optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname");
  424         opt.uio_iovcnt++;
  425         optiov[opt.uio_iovcnt].iov_base = u_hostname;
  426         error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN,
  427             &optiov[opt.uio_iovcnt].iov_len);
  428         if (error) {
  429                 free(u_path, M_TEMP);
  430                 return (error);
  431         }
  432         opt.uio_iovcnt++;
  433         if (j->jailname != NULL) {
  434                 optiov[opt.uio_iovcnt].iov_base = "name";
  435                 optiov[opt.uio_iovcnt].iov_len = sizeof("name");
  436                 opt.uio_iovcnt++;
  437                 optiov[opt.uio_iovcnt].iov_base = u_name;
  438                 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN,
  439                     &optiov[opt.uio_iovcnt].iov_len);
  440                 if (error) {
  441                         free(u_path, M_TEMP);
  442                         return (error);
  443                 }
  444                 opt.uio_iovcnt++;
  445         }
  446 #ifdef INET
  447         optiov[opt.uio_iovcnt].iov_base = "ip4.addr";
  448         optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr");
  449         opt.uio_iovcnt++;
  450         optiov[opt.uio_iovcnt].iov_base = u_ip4;
  451         optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr);
  452         if (j->version == 0)
  453                 u_ip4->s_addr = j->ip4s;
  454         else {
  455                 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len);
  456                 if (error) {
  457                         free(u_path, M_TEMP);
  458                         return (error);
  459                 }
  460         }
  461         opt.uio_iovcnt++;
  462 #endif
  463 #ifdef INET6
  464         optiov[opt.uio_iovcnt].iov_base = "ip6.addr";
  465         optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr");
  466         opt.uio_iovcnt++;
  467         optiov[opt.uio_iovcnt].iov_base = u_ip6;
  468         optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr);
  469         error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len);
  470         if (error) {
  471                 free(u_path, M_TEMP);
  472                 return (error);
  473         }
  474         opt.uio_iovcnt++;
  475 #endif
  476         KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]),
  477             ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt));
  478         error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH);
  479         free(u_path, M_TEMP);
  480         return (error);
  481 }
  482 
  483 
  484 /*
  485  * struct jail_set_args {
  486  *      struct iovec *iovp;
  487  *      unsigned int iovcnt;
  488  *      int flags;
  489  * };
  490  */
  491 int
  492 sys_jail_set(struct thread *td, struct jail_set_args *uap)
  493 {
  494         struct uio *auio;
  495         int error;
  496 
  497         /* Check that we have an even number of iovecs. */
  498         if (uap->iovcnt & 1)
  499                 return (EINVAL);
  500 
  501         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
  502         if (error)
  503                 return (error);
  504         error = kern_jail_set(td, auio, uap->flags);
  505         free(auio, M_IOV);
  506         return (error);
  507 }
  508 
  509 int
  510 kern_jail_set(struct thread *td, struct uio *optuio, int flags)
  511 {
  512         struct nameidata nd;
  513 #ifdef INET
  514         struct in_addr *ip4;
  515 #endif
  516 #ifdef INET6
  517         struct in6_addr *ip6;
  518 #endif
  519         struct vfsopt *opt;
  520         struct vfsoptlist *opts;
  521         struct prison *pr, *deadpr, *mypr, *ppr, *tpr;
  522         struct vnode *root;
  523         char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid;
  524 #if defined(INET) || defined(INET6)
  525         struct prison *tppr;
  526         void *op;
  527 #endif
  528         unsigned long hid;
  529         size_t namelen, onamelen;
  530         int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos;
  531         int gotchildmax, gotenforce, gothid, gotslevel;
  532         int fi, jid, jsys, len, level;
  533         int childmax, slevel, vfslocked;
  534 #if defined(INET) || defined(INET6)
  535         int ii, ij;
  536 #endif
  537 #ifdef INET
  538         int ip4s, redo_ip4;
  539 #endif
  540 #ifdef INET6
  541         int ip6s, redo_ip6;
  542 #endif
  543         uint64_t pr_allow, ch_allow, pr_flags, ch_flags;
  544         unsigned tallow;
  545         char numbuf[12];
  546 
  547         error = priv_check(td, PRIV_JAIL_SET);
  548         if (!error && (flags & JAIL_ATTACH))
  549                 error = priv_check(td, PRIV_JAIL_ATTACH);
  550         if (error)
  551                 return (error);
  552         mypr = ppr = td->td_ucred->cr_prison;
  553         if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0)
  554                 return (EPERM);
  555         if (flags & ~JAIL_SET_MASK)
  556                 return (EINVAL);
  557 
  558         /*
  559          * Check all the parameters before committing to anything.  Not all
  560          * errors can be caught early, but we may as well try.  Also, this
  561          * takes care of some expensive stuff (path lookup) before getting
  562          * the allprison lock.
  563          *
  564          * XXX Jails are not filesystems, and jail parameters are not mount
  565          *     options.  But it makes more sense to re-use the vfsopt code
  566          *     than duplicate it under a different name.
  567          */
  568         error = vfs_buildopts(optuio, &opts);
  569         if (error)
  570                 return (error);
  571 #ifdef INET
  572         ip4 = NULL;
  573 #endif
  574 #ifdef INET6
  575         ip6 = NULL;
  576 #endif
  577 
  578         error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
  579         if (error == ENOENT)
  580                 jid = 0;
  581         else if (error != 0)
  582                 goto done_free;
  583 
  584         error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel));
  585         if (error == ENOENT)
  586                 gotslevel = 0;
  587         else if (error != 0)
  588                 goto done_free;
  589         else
  590                 gotslevel = 1;
  591 
  592         error =
  593             vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax));
  594         if (error == ENOENT)
  595                 gotchildmax = 0;
  596         else if (error != 0)
  597                 goto done_free;
  598         else
  599                 gotchildmax = 1;
  600 
  601         error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce));
  602         if (error == ENOENT)
  603                 gotenforce = 0;
  604         else if (error != 0)
  605                 goto done_free;
  606         else if (enforce < 0 || enforce > 2) {
  607                 error = EINVAL;
  608                 goto done_free;
  609         } else
  610                 gotenforce = 1;
  611 
  612         pr_flags = ch_flags = 0;
  613         for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
  614             fi++) {
  615                 if (pr_flag_names[fi] == NULL)
  616                         continue;
  617                 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi);
  618                 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi);
  619         }
  620         ch_flags |= pr_flags;
  621         for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
  622             fi++) {
  623                 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys,
  624                     sizeof(jsys));
  625                 if (error == ENOENT)
  626                         continue;
  627                 if (error != 0)
  628                         goto done_free;
  629                 switch (jsys) {
  630                 case JAIL_SYS_DISABLE:
  631                         if (!pr_flag_jailsys[fi].disable) {
  632                                 error = EINVAL;
  633                                 goto done_free;
  634                         }
  635                         pr_flags |= pr_flag_jailsys[fi].disable;
  636                         break;
  637                 case JAIL_SYS_NEW:
  638                         pr_flags |= pr_flag_jailsys[fi].new;
  639                         break;
  640                 case JAIL_SYS_INHERIT:
  641                         break;
  642                 default:
  643                         error = EINVAL;
  644                         goto done_free;
  645                 }
  646                 ch_flags |=
  647                     pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable;
  648         }
  649         if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE
  650             && !(pr_flags & PR_PERSIST)) {
  651                 error = EINVAL;
  652                 vfs_opterror(opts, "new jail must persist or attach");
  653                 goto done_errmsg;
  654         }
  655 #ifdef VIMAGE
  656         if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) {
  657                 error = EINVAL;
  658                 vfs_opterror(opts, "vnet cannot be changed after creation");
  659                 goto done_errmsg;
  660         }
  661 #endif
  662 #ifdef INET
  663         if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) {
  664                 error = EINVAL;
  665                 vfs_opterror(opts, "ip4 cannot be changed after creation");
  666                 goto done_errmsg;
  667         }
  668 #endif
  669 #ifdef INET6
  670         if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) {
  671                 error = EINVAL;
  672                 vfs_opterror(opts, "ip6 cannot be changed after creation");
  673                 goto done_errmsg;
  674         }
  675 #endif
  676 
  677         pr_allow = ch_allow = 0;
  678         for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
  679             fi++) {
  680                 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi);
  681                 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi);
  682         }
  683         ch_allow |= pr_allow;
  684 
  685         error = vfs_getopt(opts, "name", (void **)&name, &len);
  686         if (error == ENOENT)
  687                 name = NULL;
  688         else if (error != 0)
  689                 goto done_free;
  690         else {
  691                 if (len == 0 || name[len - 1] != '\0') {
  692                         error = EINVAL;
  693                         goto done_free;
  694                 }
  695                 if (len > MAXHOSTNAMELEN) {
  696                         error = ENAMETOOLONG;
  697                         goto done_free;
  698                 }
  699         }
  700 
  701         error = vfs_getopt(opts, "host.hostname", (void **)&host, &len);
  702         if (error == ENOENT)
  703                 host = NULL;
  704         else if (error != 0)
  705                 goto done_free;
  706         else {
  707                 ch_flags |= PR_HOST;
  708                 pr_flags |= PR_HOST;
  709                 if (len == 0 || host[len - 1] != '\0') {
  710                         error = EINVAL;
  711                         goto done_free;
  712                 }
  713                 if (len > MAXHOSTNAMELEN) {
  714                         error = ENAMETOOLONG;
  715                         goto done_free;
  716                 }
  717         }
  718 
  719         error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len);
  720         if (error == ENOENT)
  721                 domain = NULL;
  722         else if (error != 0)
  723                 goto done_free;
  724         else {
  725                 ch_flags |= PR_HOST;
  726                 pr_flags |= PR_HOST;
  727                 if (len == 0 || domain[len - 1] != '\0') {
  728                         error = EINVAL;
  729                         goto done_free;
  730                 }
  731                 if (len > MAXHOSTNAMELEN) {
  732                         error = ENAMETOOLONG;
  733                         goto done_free;
  734                 }
  735         }
  736 
  737         error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len);
  738         if (error == ENOENT)
  739                 uuid = NULL;
  740         else if (error != 0)
  741                 goto done_free;
  742         else {
  743                 ch_flags |= PR_HOST;
  744                 pr_flags |= PR_HOST;
  745                 if (len == 0 || uuid[len - 1] != '\0') {
  746                         error = EINVAL;
  747                         goto done_free;
  748                 }
  749                 if (len > HOSTUUIDLEN) {
  750                         error = ENAMETOOLONG;
  751                         goto done_free;
  752                 }
  753         }
  754 
  755 #ifdef COMPAT_FREEBSD32
  756         if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
  757                 uint32_t hid32;
  758 
  759                 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32));
  760                 hid = hid32;
  761         } else
  762 #endif
  763                 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid));
  764         if (error == ENOENT)
  765                 gothid = 0;
  766         else if (error != 0)
  767                 goto done_free;
  768         else {
  769                 gothid = 1;
  770                 ch_flags |= PR_HOST;
  771                 pr_flags |= PR_HOST;
  772         }
  773 
  774 #ifdef INET
  775         error = vfs_getopt(opts, "ip4.addr", &op, &ip4s);
  776         if (error == ENOENT)
  777                 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1;
  778         else if (error != 0)
  779                 goto done_free;
  780         else if (ip4s & (sizeof(*ip4) - 1)) {
  781                 error = EINVAL;
  782                 goto done_free;
  783         } else {
  784                 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE;
  785                 if (ip4s == 0)
  786                         pr_flags |= PR_IP4_USER | PR_IP4_DISABLE;
  787                 else {
  788                         pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER;
  789                         ip4s /= sizeof(*ip4);
  790                         if (ip4s > jail_max_af_ips) {
  791                                 error = EINVAL;
  792                                 vfs_opterror(opts, "too many IPv4 addresses");
  793                                 goto done_errmsg;
  794                         }
  795                         ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
  796                         bcopy(op, ip4, ip4s * sizeof(*ip4));
  797                         /*
  798                          * IP addresses are all sorted but ip[0] to preserve
  799                          * the primary IP address as given from userland.
  800                          * This special IP is used for unbound outgoing
  801                          * connections as well for "loopback" traffic in case
  802                          * source address selection cannot find any more fitting
  803                          * address to connect from.
  804                          */
  805                         if (ip4s > 1)
  806                                 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4);
  807                         /*
  808                          * Check for duplicate addresses and do some simple
  809                          * zero and broadcast checks. If users give other bogus
  810                          * addresses it is their problem.
  811                          *
  812                          * We do not have to care about byte order for these
  813                          * checks so we will do them in NBO.
  814                          */
  815                         for (ii = 0; ii < ip4s; ii++) {
  816                                 if (ip4[ii].s_addr == INADDR_ANY ||
  817                                     ip4[ii].s_addr == INADDR_BROADCAST) {
  818                                         error = EINVAL;
  819                                         goto done_free;
  820                                 }
  821                                 if ((ii+1) < ip4s &&
  822                                     (ip4[0].s_addr == ip4[ii+1].s_addr ||
  823                                      ip4[ii].s_addr == ip4[ii+1].s_addr)) {
  824                                         error = EINVAL;
  825                                         goto done_free;
  826                                 }
  827                         }
  828                 }
  829         }
  830 #endif
  831 
  832 #ifdef INET6
  833         error = vfs_getopt(opts, "ip6.addr", &op, &ip6s);
  834         if (error == ENOENT)
  835                 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1;
  836         else if (error != 0)
  837                 goto done_free;
  838         else if (ip6s & (sizeof(*ip6) - 1)) {
  839                 error = EINVAL;
  840                 goto done_free;
  841         } else {
  842                 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE;
  843                 if (ip6s == 0)
  844                         pr_flags |= PR_IP6_USER | PR_IP6_DISABLE;
  845                 else {
  846                         pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER;
  847                         ip6s /= sizeof(*ip6);
  848                         if (ip6s > jail_max_af_ips) {
  849                                 error = EINVAL;
  850                                 vfs_opterror(opts, "too many IPv6 addresses");
  851                                 goto done_errmsg;
  852                         }
  853                         ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
  854                         bcopy(op, ip6, ip6s * sizeof(*ip6));
  855                         if (ip6s > 1)
  856                                 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6);
  857                         for (ii = 0; ii < ip6s; ii++) {
  858                                 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) {
  859                                         error = EINVAL;
  860                                         goto done_free;
  861                                 }
  862                                 if ((ii+1) < ip6s &&
  863                                     (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) ||
  864                                      IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1])))
  865                                 {
  866                                         error = EINVAL;
  867                                         goto done_free;
  868                                 }
  869                         }
  870                 }
  871         }
  872 #endif
  873 
  874 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
  875         if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
  876                 error = EINVAL;
  877                 vfs_opterror(opts,
  878                     "vnet jails cannot have IP address restrictions");
  879                 goto done_errmsg;
  880         }
  881 #endif
  882 
  883         root = NULL;
  884         error = vfs_getopt(opts, "path", (void **)&path, &len);
  885         if (error == ENOENT)
  886                 path = NULL;
  887         else if (error != 0)
  888                 goto done_free;
  889         else {
  890                 if (flags & JAIL_UPDATE) {
  891                         error = EINVAL;
  892                         vfs_opterror(opts,
  893                             "path cannot be changed after creation");
  894                         goto done_errmsg;
  895                 }
  896                 if (len == 0 || path[len - 1] != '\0') {
  897                         error = EINVAL;
  898                         goto done_free;
  899                 }
  900                 if (len < 2 || (len == 2 && path[0] == '/'))
  901                         path = NULL;
  902                 else {
  903                         /* Leave room for a real-root full pathname. */
  904                         if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/")
  905                             ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) {
  906                                 error = ENAMETOOLONG;
  907                                 goto done_free;
  908                         }
  909                         NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE,
  910                             path, td);
  911                         error = namei(&nd);
  912                         if (error)
  913                                 goto done_free;
  914                         vfslocked = NDHASGIANT(&nd);
  915                         root = nd.ni_vp;
  916                         NDFREE(&nd, NDF_ONLY_PNBUF);
  917                         if (root->v_type != VDIR) {
  918                                 error = ENOTDIR;
  919                                 vrele(root);
  920                                 VFS_UNLOCK_GIANT(vfslocked);
  921                                 goto done_free;
  922                         }
  923                         VFS_UNLOCK_GIANT(vfslocked);
  924                 }
  925         }
  926 
  927         /*
  928          * Grab the allprison lock before letting modules check their
  929          * parameters.  Once we have it, do not let go so we'll have a
  930          * consistent view of the OSD list.
  931          */
  932         sx_xlock(&allprison_lock);
  933         error = osd_jail_call(NULL, PR_METHOD_CHECK, opts);
  934         if (error)
  935                 goto done_unlock_list;
  936 
  937         /* By now, all parameters should have been noted. */
  938         TAILQ_FOREACH(opt, opts, link) {
  939                 if (!opt->seen && strcmp(opt->name, "errmsg")) {
  940                         error = EINVAL;
  941                         vfs_opterror(opts, "unknown parameter: %s", opt->name);
  942                         goto done_unlock_list;
  943                 }
  944         }
  945 
  946         /*
  947          * See if we are creating a new record or updating an existing one.
  948          * This abuses the file error codes ENOENT and EEXIST.
  949          */
  950         cuflags = flags & (JAIL_CREATE | JAIL_UPDATE);
  951         if (!cuflags) {
  952                 error = EINVAL;
  953                 vfs_opterror(opts, "no valid operation (create or update)");
  954                 goto done_unlock_list;
  955         }
  956         pr = NULL;
  957         namelc = NULL;
  958         if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) {
  959                 namelc = strrchr(name, '.');
  960                 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10);
  961                 if (*p != '\0')
  962                         jid = 0;
  963         }
  964         if (jid != 0) {
  965                 /*
  966                  * See if a requested jid already exists.  There is an
  967                  * information leak here if the jid exists but is not within
  968                  * the caller's jail hierarchy.  Jail creators will get EEXIST
  969                  * even though they cannot see the jail, and CREATE | UPDATE
  970                  * will return ENOENT which is not normally a valid error.
  971                  */
  972                 if (jid < 0) {
  973                         error = EINVAL;
  974                         vfs_opterror(opts, "negative jid");
  975                         goto done_unlock_list;
  976                 }
  977                 pr = prison_find(jid);
  978                 if (pr != NULL) {
  979                         ppr = pr->pr_parent;
  980                         /* Create: jid must not exist. */
  981                         if (cuflags == JAIL_CREATE) {
  982                                 mtx_unlock(&pr->pr_mtx);
  983                                 error = EEXIST;
  984                                 vfs_opterror(opts, "jail %d already exists",
  985                                     jid);
  986                                 goto done_unlock_list;
  987                         }
  988                         if (!prison_ischild(mypr, pr)) {
  989                                 mtx_unlock(&pr->pr_mtx);
  990                                 pr = NULL;
  991                         } else if (pr->pr_uref == 0) {
  992                                 if (!(flags & JAIL_DYING)) {
  993                                         mtx_unlock(&pr->pr_mtx);
  994                                         error = ENOENT;
  995                                         vfs_opterror(opts, "jail %d is dying",
  996                                             jid);
  997                                         goto done_unlock_list;
  998                                 } else if ((flags & JAIL_ATTACH) ||
  999                                     (pr_flags & PR_PERSIST)) {
 1000                                         /*
 1001                                          * A dying jail might be resurrected
 1002                                          * (via attach or persist), but first
 1003                                          * it must determine if another jail
 1004                                          * has claimed its name.  Accomplish
 1005                                          * this by implicitly re-setting the
 1006                                          * name.
 1007                                          */
 1008                                         if (name == NULL)
 1009                                                 name = prison_name(mypr, pr);
 1010                                 }
 1011                         }
 1012                 }
 1013                 if (pr == NULL) {
 1014                         /* Update: jid must exist. */
 1015                         if (cuflags == JAIL_UPDATE) {
 1016                                 error = ENOENT;
 1017                                 vfs_opterror(opts, "jail %d not found", jid);
 1018                                 goto done_unlock_list;
 1019                         }
 1020                 }
 1021         }
 1022         /*
 1023          * If the caller provided a name, look for a jail by that name.
 1024          * This has different semantics for creates and updates keyed by jid
 1025          * (where the name must not already exist in a different jail),
 1026          * and updates keyed by the name itself (where the name must exist
 1027          * because that is the jail being updated).
 1028          */
 1029         if (name != NULL) {
 1030                 namelc = strrchr(name, '.');
 1031                 if (namelc == NULL)
 1032                         namelc = name;
 1033                 else {
 1034                         /*
 1035                          * This is a hierarchical name.  Split it into the
 1036                          * parent and child names, and make sure the parent
 1037                          * exists or matches an already found jail.
 1038                          */
 1039                         *namelc = '\0';
 1040                         if (pr != NULL) {
 1041                                 if (strncmp(name, ppr->pr_name, namelc - name)
 1042                                     || ppr->pr_name[namelc - name] != '\0') {
 1043                                         mtx_unlock(&pr->pr_mtx);
 1044                                         error = EINVAL;
 1045                                         vfs_opterror(opts,
 1046                                             "cannot change jail's parent");
 1047                                         goto done_unlock_list;
 1048                                 }
 1049                         } else {
 1050                                 ppr = prison_find_name(mypr, name);
 1051                                 if (ppr == NULL) {
 1052                                         error = ENOENT;
 1053                                         vfs_opterror(opts,
 1054                                             "jail \"%s\" not found", name);
 1055                                         goto done_unlock_list;
 1056                                 }
 1057                                 mtx_unlock(&ppr->pr_mtx);
 1058                         }
 1059                         name = ++namelc;
 1060                 }
 1061                 if (name[0] != '\0') {
 1062                         namelen =
 1063                             (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1;
 1064  name_again:
 1065                         deadpr = NULL;
 1066                         FOREACH_PRISON_CHILD(ppr, tpr) {
 1067                                 if (tpr != pr && tpr->pr_ref > 0 &&
 1068                                     !strcmp(tpr->pr_name + namelen, name)) {
 1069                                         if (pr == NULL &&
 1070                                             cuflags != JAIL_CREATE) {
 1071                                                 mtx_lock(&tpr->pr_mtx);
 1072                                                 if (tpr->pr_ref > 0) {
 1073                                                         /*
 1074                                                          * Use this jail
 1075                                                          * for updates.
 1076                                                          */
 1077                                                         if (tpr->pr_uref > 0) {
 1078                                                                 pr = tpr;
 1079                                                                 break;
 1080                                                         }
 1081                                                         deadpr = tpr;
 1082                                                 }
 1083                                                 mtx_unlock(&tpr->pr_mtx);
 1084                                         } else if (tpr->pr_uref > 0) {
 1085                                                 /*
 1086                                                  * Create, or update(jid):
 1087                                                  * name must not exist in an
 1088                                                  * active sibling jail.
 1089                                                  */
 1090                                                 error = EEXIST;
 1091                                                 if (pr != NULL)
 1092                                                         mtx_unlock(&pr->pr_mtx);
 1093                                                 vfs_opterror(opts,
 1094                                                    "jail \"%s\" already exists",
 1095                                                    name);
 1096                                                 goto done_unlock_list;
 1097                                         }
 1098                                 }
 1099                         }
 1100                         /* If no active jail is found, use a dying one. */
 1101                         if (deadpr != NULL && pr == NULL) {
 1102                                 if (flags & JAIL_DYING) {
 1103                                         mtx_lock(&deadpr->pr_mtx);
 1104                                         if (deadpr->pr_ref == 0) {
 1105                                                 mtx_unlock(&deadpr->pr_mtx);
 1106                                                 goto name_again;
 1107                                         }
 1108                                         pr = deadpr;
 1109                                 } else if (cuflags == JAIL_UPDATE) {
 1110                                         error = ENOENT;
 1111                                         vfs_opterror(opts,
 1112                                             "jail \"%s\" is dying", name);
 1113                                         goto done_unlock_list;
 1114                                 }
 1115                         }
 1116                         /* Update: name must exist if no jid. */
 1117                         else if (cuflags == JAIL_UPDATE && pr == NULL) {
 1118                                 error = ENOENT;
 1119                                 vfs_opterror(opts, "jail \"%s\" not found",
 1120                                     name);
 1121                                 goto done_unlock_list;
 1122                         }
 1123                 }
 1124         }
 1125         /* Update: must provide a jid or name. */
 1126         else if (cuflags == JAIL_UPDATE && pr == NULL) {
 1127                 error = ENOENT;
 1128                 vfs_opterror(opts, "update specified no jail");
 1129                 goto done_unlock_list;
 1130         }
 1131 
 1132         /* If there's no prison to update, create a new one and link it in. */
 1133         if (pr == NULL) {
 1134                 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent)
 1135                         if (tpr->pr_childcount >= tpr->pr_childmax) {
 1136                                 error = EPERM;
 1137                                 vfs_opterror(opts, "prison limit exceeded");
 1138                                 goto done_unlock_list;
 1139                         }
 1140                 created = 1;
 1141                 mtx_lock(&ppr->pr_mtx);
 1142                 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) {
 1143                         mtx_unlock(&ppr->pr_mtx);
 1144                         error = ENOENT;
 1145                         vfs_opterror(opts, "parent jail went away!");
 1146                         goto done_unlock_list;
 1147                 }
 1148                 ppr->pr_ref++;
 1149                 ppr->pr_uref++;
 1150                 mtx_unlock(&ppr->pr_mtx);
 1151                 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
 1152                 if (jid == 0) {
 1153                         /* Find the next free jid. */
 1154                         jid = lastprid + 1;
 1155  findnext:
 1156                         if (jid == JAIL_MAX)
 1157                                 jid = 1;
 1158                         TAILQ_FOREACH(tpr, &allprison, pr_list) {
 1159                                 if (tpr->pr_id < jid)
 1160                                         continue;
 1161                                 if (tpr->pr_id > jid || tpr->pr_ref == 0) {
 1162                                         TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
 1163                                         break;
 1164                                 }
 1165                                 if (jid == lastprid) {
 1166                                         error = EAGAIN;
 1167                                         vfs_opterror(opts,
 1168                                             "no available jail IDs");
 1169                                         free(pr, M_PRISON);
 1170                                         prison_deref(ppr, PD_DEREF |
 1171                                             PD_DEUREF | PD_LIST_XLOCKED);
 1172                                         goto done_releroot;
 1173                                 }
 1174                                 jid++;
 1175                                 goto findnext;
 1176                         }
 1177                         lastprid = jid;
 1178                 } else {
 1179                         /*
 1180                          * The jail already has a jid (that did not yet exist),
 1181                          * so just find where to insert it.
 1182                          */
 1183                         TAILQ_FOREACH(tpr, &allprison, pr_list)
 1184                                 if (tpr->pr_id >= jid) {
 1185                                         TAILQ_INSERT_BEFORE(tpr, pr, pr_list);
 1186                                         break;
 1187                                 }
 1188                 }
 1189                 if (tpr == NULL)
 1190                         TAILQ_INSERT_TAIL(&allprison, pr, pr_list);
 1191                 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling);
 1192                 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
 1193                         tpr->pr_childcount++;
 1194 
 1195                 pr->pr_parent = ppr;
 1196                 pr->pr_id = jid;
 1197 
 1198                 /* Set some default values, and inherit some from the parent. */
 1199                 if (name == NULL)
 1200                         name = "";
 1201                 if (path == NULL) {
 1202                         path = "/";
 1203                         root = mypr->pr_root;
 1204                         vref(root);
 1205                 }
 1206                 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN);
 1207                 pr->pr_flags |= PR_HOST;
 1208 #if defined(INET) || defined(INET6)
 1209 #ifdef VIMAGE
 1210                 if (!(pr_flags & PR_VNET))
 1211 #endif
 1212                 {
 1213 #ifdef INET
 1214                         if (!(ch_flags & PR_IP4_USER))
 1215                                 pr->pr_flags |=
 1216                                     PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE;
 1217                         else if (!(pr_flags & PR_IP4_USER)) {
 1218                                 pr->pr_flags |= ppr->pr_flags & PR_IP4;
 1219                                 if (ppr->pr_ip4 != NULL) {
 1220                                         pr->pr_ip4s = ppr->pr_ip4s;
 1221                                         pr->pr_ip4 = malloc(pr->pr_ip4s *
 1222                                             sizeof(struct in_addr), M_PRISON,
 1223                                             M_WAITOK);
 1224                                         bcopy(ppr->pr_ip4, pr->pr_ip4,
 1225                                             pr->pr_ip4s * sizeof(*pr->pr_ip4));
 1226                                 }
 1227                         }
 1228 #endif
 1229 #ifdef INET6
 1230                         if (!(ch_flags & PR_IP6_USER))
 1231                                 pr->pr_flags |=
 1232                                     PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE;
 1233                         else if (!(pr_flags & PR_IP6_USER)) {
 1234                                 pr->pr_flags |= ppr->pr_flags & PR_IP6;
 1235                                 if (ppr->pr_ip6 != NULL) {
 1236                                         pr->pr_ip6s = ppr->pr_ip6s;
 1237                                         pr->pr_ip6 = malloc(pr->pr_ip6s *
 1238                                             sizeof(struct in6_addr), M_PRISON,
 1239                                             M_WAITOK);
 1240                                         bcopy(ppr->pr_ip6, pr->pr_ip6,
 1241                                             pr->pr_ip6s * sizeof(*pr->pr_ip6));
 1242                                 }
 1243                         }
 1244 #endif
 1245                 }
 1246 #endif
 1247                 /* Source address selection is always on by default. */
 1248                 pr->pr_flags |= _PR_IP_SADDRSEL;
 1249 
 1250                 pr->pr_securelevel = ppr->pr_securelevel;
 1251                 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow;
 1252                 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS;
 1253 
 1254                 LIST_INIT(&pr->pr_children);
 1255                 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK);
 1256 
 1257 #ifdef VIMAGE
 1258                 /* Allocate a new vnet if specified. */
 1259                 pr->pr_vnet = (pr_flags & PR_VNET)
 1260                     ? vnet_alloc() : ppr->pr_vnet;
 1261 #endif
 1262                 /*
 1263                  * Allocate a dedicated cpuset for each jail.
 1264                  * Unlike other initial settings, this may return an erorr.
 1265                  */
 1266                 error = cpuset_create_root(ppr, &pr->pr_cpuset);
 1267                 if (error) {
 1268                         prison_deref(pr, PD_LIST_XLOCKED);
 1269                         goto done_releroot;
 1270                 }
 1271 
 1272                 mtx_lock(&pr->pr_mtx);
 1273                 /*
 1274                  * New prisons do not yet have a reference, because we do not
 1275                  * want other to see the incomplete prison once the
 1276                  * allprison_lock is downgraded.
 1277                  */
 1278         } else {
 1279                 created = 0;
 1280                 /*
 1281                  * Grab a reference for existing prisons, to ensure they
 1282                  * continue to exist for the duration of the call.
 1283                  */
 1284                 pr->pr_ref++;
 1285 #if defined(VIMAGE) && (defined(INET) || defined(INET6))
 1286                 if ((pr->pr_flags & PR_VNET) &&
 1287                     (ch_flags & (PR_IP4_USER | PR_IP6_USER))) {
 1288                         error = EINVAL;
 1289                         vfs_opterror(opts,
 1290                             "vnet jails cannot have IP address restrictions");
 1291                         goto done_deref_locked;
 1292                 }
 1293 #endif
 1294 #ifdef INET
 1295                 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
 1296                         error = EINVAL;
 1297                         vfs_opterror(opts,
 1298                             "ip4 cannot be changed after creation");
 1299                         goto done_deref_locked;
 1300                 }
 1301 #endif
 1302 #ifdef INET6
 1303                 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) {
 1304                         error = EINVAL;
 1305                         vfs_opterror(opts,
 1306                             "ip6 cannot be changed after creation");
 1307                         goto done_deref_locked;
 1308                 }
 1309 #endif
 1310         }
 1311 
 1312         /* Do final error checking before setting anything. */
 1313         if (gotslevel) {
 1314                 if (slevel < ppr->pr_securelevel) {
 1315                         error = EPERM;
 1316                         goto done_deref_locked;
 1317                 }
 1318         }
 1319         if (gotchildmax) {
 1320                 if (childmax >= ppr->pr_childmax) {
 1321                         error = EPERM;
 1322                         goto done_deref_locked;
 1323                 }
 1324         }
 1325         if (gotenforce) {
 1326                 if (enforce < ppr->pr_enforce_statfs) {
 1327                         error = EPERM;
 1328                         goto done_deref_locked;
 1329                 }
 1330         }
 1331 #ifdef INET
 1332         if (ip4s > 0) {
 1333                 if (ppr->pr_flags & PR_IP4) {
 1334                         /*
 1335                          * Make sure the new set of IP addresses is a
 1336                          * subset of the parent's list.  Don't worry
 1337                          * about the parent being unlocked, as any
 1338                          * setting is done with allprison_lock held.
 1339                          */
 1340                         for (ij = 0; ij < ppr->pr_ip4s; ij++)
 1341                                 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
 1342                                         break;
 1343                         if (ij == ppr->pr_ip4s) {
 1344                                 error = EPERM;
 1345                                 goto done_deref_locked;
 1346                         }
 1347                         if (ip4s > 1) {
 1348                                 for (ii = ij = 1; ii < ip4s; ii++) {
 1349                                         if (ip4[ii].s_addr ==
 1350                                             ppr->pr_ip4[0].s_addr)
 1351                                                 continue;
 1352                                         for (; ij < ppr->pr_ip4s; ij++)
 1353                                                 if (ip4[ii].s_addr ==
 1354                                                     ppr->pr_ip4[ij].s_addr)
 1355                                                         break;
 1356                                         if (ij == ppr->pr_ip4s)
 1357                                                 break;
 1358                                 }
 1359                                 if (ij == ppr->pr_ip4s) {
 1360                                         error = EPERM;
 1361                                         goto done_deref_locked;
 1362                                 }
 1363                         }
 1364                 }
 1365                 /*
 1366                  * Check for conflicting IP addresses.  We permit them
 1367                  * if there is no more than one IP on each jail.  If
 1368                  * there is a duplicate on a jail with more than one
 1369                  * IP stop checking and return error.
 1370                  */
 1371                 tppr = ppr;
 1372 #ifdef VIMAGE
 1373                 for (; tppr != &prison0; tppr = tppr->pr_parent)
 1374                         if (tppr->pr_flags & PR_VNET)
 1375                                 break;
 1376 #endif
 1377                 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
 1378                         if (tpr == pr ||
 1379 #ifdef VIMAGE
 1380                             (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
 1381 #endif
 1382                             tpr->pr_uref == 0) {
 1383                                 descend = 0;
 1384                                 continue;
 1385                         }
 1386                         if (!(tpr->pr_flags & PR_IP4_USER))
 1387                                 continue;
 1388                         descend = 0;
 1389                         if (tpr->pr_ip4 == NULL ||
 1390                             (ip4s == 1 && tpr->pr_ip4s == 1))
 1391                                 continue;
 1392                         for (ii = 0; ii < ip4s; ii++) {
 1393                                 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) {
 1394                                         error = EADDRINUSE;
 1395                                         vfs_opterror(opts,
 1396                                             "IPv4 addresses clash");
 1397                                         goto done_deref_locked;
 1398                                 }
 1399                         }
 1400                 }
 1401         }
 1402 #endif
 1403 #ifdef INET6
 1404         if (ip6s > 0) {
 1405                 if (ppr->pr_flags & PR_IP6) {
 1406                         /*
 1407                          * Make sure the new set of IP addresses is a
 1408                          * subset of the parent's list.
 1409                          */
 1410                         for (ij = 0; ij < ppr->pr_ip6s; ij++)
 1411                                 if (IN6_ARE_ADDR_EQUAL(&ip6[0],
 1412                                     &ppr->pr_ip6[ij]))
 1413                                         break;
 1414                         if (ij == ppr->pr_ip6s) {
 1415                                 error = EPERM;
 1416                                 goto done_deref_locked;
 1417                         }
 1418                         if (ip6s > 1) {
 1419                                 for (ii = ij = 1; ii < ip6s; ii++) {
 1420                                         if (IN6_ARE_ADDR_EQUAL(&ip6[ii],
 1421                                              &ppr->pr_ip6[0]))
 1422                                                 continue;
 1423                                         for (; ij < ppr->pr_ip6s; ij++)
 1424                                                 if (IN6_ARE_ADDR_EQUAL(
 1425                                                     &ip6[ii], &ppr->pr_ip6[ij]))
 1426                                                         break;
 1427                                         if (ij == ppr->pr_ip6s)
 1428                                                 break;
 1429                                 }
 1430                                 if (ij == ppr->pr_ip6s) {
 1431                                         error = EPERM;
 1432                                         goto done_deref_locked;
 1433                                 }
 1434                         }
 1435                 }
 1436                 /* Check for conflicting IP addresses. */
 1437                 tppr = ppr;
 1438 #ifdef VIMAGE
 1439                 for (; tppr != &prison0; tppr = tppr->pr_parent)
 1440                         if (tppr->pr_flags & PR_VNET)
 1441                                 break;
 1442 #endif
 1443                 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) {
 1444                         if (tpr == pr ||
 1445 #ifdef VIMAGE
 1446                             (tpr != tppr && (tpr->pr_flags & PR_VNET)) ||
 1447 #endif
 1448                             tpr->pr_uref == 0) {
 1449                                 descend = 0;
 1450                                 continue;
 1451                         }
 1452                         if (!(tpr->pr_flags & PR_IP6_USER))
 1453                                 continue;
 1454                         descend = 0;
 1455                         if (tpr->pr_ip6 == NULL ||
 1456                             (ip6s == 1 && tpr->pr_ip6s == 1))
 1457                                 continue;
 1458                         for (ii = 0; ii < ip6s; ii++) {
 1459                                 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) {
 1460                                         error = EADDRINUSE;
 1461                                         vfs_opterror(opts,
 1462                                             "IPv6 addresses clash");
 1463                                         goto done_deref_locked;
 1464                                 }
 1465                         }
 1466                 }
 1467         }
 1468 #endif
 1469         onamelen = namelen = 0;
 1470         if (name != NULL) {
 1471                 /* Give a default name of the jid. */
 1472                 if (name[0] == '\0')
 1473                         snprintf(name = numbuf, sizeof(numbuf), "%d", jid);
 1474                 else if (*namelc == '' || (strtoul(namelc, &p, 10) != jid &&
 1475                     *p == '\0')) {
 1476                         error = EINVAL;
 1477                         vfs_opterror(opts,
 1478                             "name cannot be numeric (unless it is the jid)");
 1479                         goto done_deref_locked;
 1480                 }
 1481                 /*
 1482                  * Make sure the name isn't too long for the prison or its
 1483                  * children.
 1484                  */
 1485                 onamelen = strlen(pr->pr_name);
 1486                 namelen = strlen(name);
 1487                 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) {
 1488                         error = ENAMETOOLONG;
 1489                         goto done_deref_locked;
 1490                 }
 1491                 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) {
 1492                         if (strlen(tpr->pr_name) + (namelen - onamelen) >=
 1493                             sizeof(pr->pr_name)) {
 1494                                 error = ENAMETOOLONG;
 1495                                 goto done_deref_locked;
 1496                         }
 1497                 }
 1498         }
 1499         if (pr_allow & ~ppr->pr_allow) {
 1500                 error = EPERM;
 1501                 goto done_deref_locked;
 1502         }
 1503 
 1504         /* Set the parameters of the prison. */
 1505 #ifdef INET
 1506         redo_ip4 = 0;
 1507         if (pr_flags & PR_IP4_USER) {
 1508                 pr->pr_flags |= PR_IP4;
 1509                 free(pr->pr_ip4, M_PRISON);
 1510                 pr->pr_ip4s = ip4s;
 1511                 pr->pr_ip4 = ip4;
 1512                 ip4 = NULL;
 1513                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1514 #ifdef VIMAGE
 1515                         if (tpr->pr_flags & PR_VNET) {
 1516                                 descend = 0;
 1517                                 continue;
 1518                         }
 1519 #endif
 1520                         if (prison_restrict_ip4(tpr, NULL)) {
 1521                                 redo_ip4 = 1;
 1522                                 descend = 0;
 1523                         }
 1524                 }
 1525         }
 1526 #endif
 1527 #ifdef INET6
 1528         redo_ip6 = 0;
 1529         if (pr_flags & PR_IP6_USER) {
 1530                 pr->pr_flags |= PR_IP6;
 1531                 free(pr->pr_ip6, M_PRISON);
 1532                 pr->pr_ip6s = ip6s;
 1533                 pr->pr_ip6 = ip6;
 1534                 ip6 = NULL;
 1535                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1536 #ifdef VIMAGE
 1537                         if (tpr->pr_flags & PR_VNET) {
 1538                                 descend = 0;
 1539                                 continue;
 1540                         }
 1541 #endif
 1542                         if (prison_restrict_ip6(tpr, NULL)) {
 1543                                 redo_ip6 = 1;
 1544                                 descend = 0;
 1545                         }
 1546                 }
 1547         }
 1548 #endif
 1549         if (gotslevel) {
 1550                 pr->pr_securelevel = slevel;
 1551                 /* Set all child jails to be at least this level. */
 1552                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
 1553                         if (tpr->pr_securelevel < slevel)
 1554                                 tpr->pr_securelevel = slevel;
 1555         }
 1556         if (gotchildmax) {
 1557                 pr->pr_childmax = childmax;
 1558                 /* Set all child jails to under this limit. */
 1559                 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level)
 1560                         if (tpr->pr_childmax > childmax - level)
 1561                                 tpr->pr_childmax = childmax > level
 1562                                     ? childmax - level : 0;
 1563         }
 1564         if (gotenforce) {
 1565                 pr->pr_enforce_statfs = enforce;
 1566                 /* Pass this restriction on to the children. */
 1567                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
 1568                         if (tpr->pr_enforce_statfs < enforce)
 1569                                 tpr->pr_enforce_statfs = enforce;
 1570         }
 1571         if (name != NULL) {
 1572                 if (ppr == &prison0)
 1573                         strlcpy(pr->pr_name, name, sizeof(pr->pr_name));
 1574                 else
 1575                         snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s",
 1576                             ppr->pr_name, name);
 1577                 /* Change this component of child names. */
 1578                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1579                         bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen,
 1580                             strlen(tpr->pr_name + onamelen) + 1);
 1581                         bcopy(pr->pr_name, tpr->pr_name, namelen);
 1582                 }
 1583         }
 1584         if (path != NULL) {
 1585                 /* Try to keep a real-rooted full pathname. */
 1586                 if (path[0] == '/' && strcmp(mypr->pr_path, "/"))
 1587                         snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s",
 1588                             mypr->pr_path, path);
 1589                 else
 1590                         strlcpy(pr->pr_path, path, sizeof(pr->pr_path));
 1591                 pr->pr_root = root;
 1592         }
 1593         if (PR_HOST & ch_flags & ~pr_flags) {
 1594                 if (pr->pr_flags & PR_HOST) {
 1595                         /*
 1596                          * Copy the parent's host info.  As with pr_ip4 above,
 1597                          * the lack of a lock on the parent is not a problem;
 1598                          * it is always set with allprison_lock at least
 1599                          * shared, and is held exclusively here.
 1600                          */
 1601                         strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname,
 1602                             sizeof(pr->pr_hostname));
 1603                         strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname,
 1604                             sizeof(pr->pr_domainname));
 1605                         strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid,
 1606                             sizeof(pr->pr_hostuuid));
 1607                         pr->pr_hostid = pr->pr_parent->pr_hostid;
 1608                 }
 1609         } else if (host != NULL || domain != NULL || uuid != NULL || gothid) {
 1610                 /* Set this prison, and any descendants without PR_HOST. */
 1611                 if (host != NULL)
 1612                         strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname));
 1613                 if (domain != NULL)
 1614                         strlcpy(pr->pr_domainname, domain, 
 1615                             sizeof(pr->pr_domainname));
 1616                 if (uuid != NULL)
 1617                         strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid));
 1618                 if (gothid)
 1619                         pr->pr_hostid = hid;
 1620                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1621                         if (tpr->pr_flags & PR_HOST)
 1622                                 descend = 0;
 1623                         else {
 1624                                 if (host != NULL)
 1625                                         strlcpy(tpr->pr_hostname,
 1626                                             pr->pr_hostname,
 1627                                             sizeof(tpr->pr_hostname));
 1628                                 if (domain != NULL)
 1629                                         strlcpy(tpr->pr_domainname, 
 1630                                             pr->pr_domainname,
 1631                                             sizeof(tpr->pr_domainname));
 1632                                 if (uuid != NULL)
 1633                                         strlcpy(tpr->pr_hostuuid,
 1634                                             pr->pr_hostuuid,
 1635                                             sizeof(tpr->pr_hostuuid));
 1636                                 if (gothid)
 1637                                         tpr->pr_hostid = hid;
 1638                         }
 1639                 }
 1640         }
 1641         if ((tallow = ch_allow & ~pr_allow)) {
 1642                 /* Clear allow bits in all children. */
 1643                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend)
 1644                         tpr->pr_allow &= ~tallow;
 1645         }
 1646         pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow;
 1647         /*
 1648          * Persistent prisons get an extra reference, and prisons losing their
 1649          * persist flag lose that reference.  Only do this for existing prisons
 1650          * for now, so new ones will remain unseen until after the module
 1651          * handlers have completed.
 1652          */
 1653         if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) {
 1654                 if (pr_flags & PR_PERSIST) {
 1655                         pr->pr_ref++;
 1656                         pr->pr_uref++;
 1657                 } else {
 1658                         pr->pr_ref--;
 1659                         pr->pr_uref--;
 1660                 }
 1661         }
 1662         pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags;
 1663         mtx_unlock(&pr->pr_mtx);
 1664 
 1665 #ifdef RACCT
 1666         if (created)
 1667                 prison_racct_attach(pr);
 1668 #endif
 1669 
 1670         /* Locks may have prevented a complete restriction of child IP
 1671          * addresses.  If so, allocate some more memory and try again.
 1672          */
 1673 #ifdef INET
 1674         while (redo_ip4) {
 1675                 ip4s = pr->pr_ip4s;
 1676                 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK);
 1677                 mtx_lock(&pr->pr_mtx);
 1678                 redo_ip4 = 0;
 1679                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1680 #ifdef VIMAGE
 1681                         if (tpr->pr_flags & PR_VNET) {
 1682                                 descend = 0;
 1683                                 continue;
 1684                         }
 1685 #endif
 1686                         if (prison_restrict_ip4(tpr, ip4)) {
 1687                                 if (ip4 != NULL)
 1688                                         ip4 = NULL;
 1689                                 else
 1690                                         redo_ip4 = 1;
 1691                         }
 1692                 }
 1693                 mtx_unlock(&pr->pr_mtx);
 1694         }
 1695 #endif
 1696 #ifdef INET6
 1697         while (redo_ip6) {
 1698                 ip6s = pr->pr_ip6s;
 1699                 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK);
 1700                 mtx_lock(&pr->pr_mtx);
 1701                 redo_ip6 = 0;
 1702                 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) {
 1703 #ifdef VIMAGE
 1704                         if (tpr->pr_flags & PR_VNET) {
 1705                                 descend = 0;
 1706                                 continue;
 1707                         }
 1708 #endif
 1709                         if (prison_restrict_ip6(tpr, ip6)) {
 1710                                 if (ip6 != NULL)
 1711                                         ip6 = NULL;
 1712                                 else
 1713                                         redo_ip6 = 1;
 1714                         }
 1715                 }
 1716                 mtx_unlock(&pr->pr_mtx);
 1717         }
 1718 #endif
 1719 
 1720         /* Let the modules do their work. */
 1721         sx_downgrade(&allprison_lock);
 1722         if (created) {
 1723                 error = osd_jail_call(pr, PR_METHOD_CREATE, opts);
 1724                 if (error) {
 1725                         prison_deref(pr, PD_LIST_SLOCKED);
 1726                         goto done_errmsg;
 1727                 }
 1728         }
 1729         error = osd_jail_call(pr, PR_METHOD_SET, opts);
 1730         if (error) {
 1731                 prison_deref(pr, created
 1732                     ? PD_LIST_SLOCKED
 1733                     : PD_DEREF | PD_LIST_SLOCKED);
 1734                 goto done_errmsg;
 1735         }
 1736 
 1737         /* Attach this process to the prison if requested. */
 1738         if (flags & JAIL_ATTACH) {
 1739                 mtx_lock(&pr->pr_mtx);
 1740                 error = do_jail_attach(td, pr);
 1741                 if (error) {
 1742                         vfs_opterror(opts, "attach failed");
 1743                         if (!created)
 1744                                 prison_deref(pr, PD_DEREF);
 1745                         goto done_errmsg;
 1746                 }
 1747         }
 1748 
 1749         /*
 1750          * Now that it is all there, drop the temporary reference from existing
 1751          * prisons.  Or add a reference to newly created persistent prisons
 1752          * (which was not done earlier so that the prison would not be publicly
 1753          * visible).
 1754          */
 1755         if (!created) {
 1756                 prison_deref(pr, (flags & JAIL_ATTACH)
 1757                     ? PD_DEREF
 1758                     : PD_DEREF | PD_LIST_SLOCKED);
 1759         } else {
 1760                 if (pr_flags & PR_PERSIST) {
 1761                         mtx_lock(&pr->pr_mtx);
 1762                         pr->pr_ref++;
 1763                         pr->pr_uref++;
 1764                         mtx_unlock(&pr->pr_mtx);
 1765                 }
 1766                 if (!(flags & JAIL_ATTACH))
 1767                         sx_sunlock(&allprison_lock);
 1768         }
 1769         td->td_retval[0] = pr->pr_id;
 1770         goto done_errmsg;
 1771 
 1772  done_deref_locked:
 1773         prison_deref(pr, created
 1774             ? PD_LOCKED | PD_LIST_XLOCKED
 1775             : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
 1776         goto done_releroot;
 1777  done_unlock_list:
 1778         sx_xunlock(&allprison_lock);
 1779  done_releroot:
 1780         if (root != NULL) {
 1781                 vfslocked = VFS_LOCK_GIANT(root->v_mount);
 1782                 vrele(root);
 1783                 VFS_UNLOCK_GIANT(vfslocked);
 1784         }
 1785  done_errmsg:
 1786         if (error) {
 1787                 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
 1788                 if (errmsg_len > 0) {
 1789                         errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1;
 1790                         if (errmsg_pos > 0) {
 1791                                 if (optuio->uio_segflg == UIO_SYSSPACE)
 1792                                         bcopy(errmsg,
 1793                                            optuio->uio_iov[errmsg_pos].iov_base,
 1794                                            errmsg_len);
 1795                                 else
 1796                                         copyout(errmsg,
 1797                                            optuio->uio_iov[errmsg_pos].iov_base,
 1798                                            errmsg_len);
 1799                         }
 1800                 }
 1801         }
 1802  done_free:
 1803 #ifdef INET
 1804         free(ip4, M_PRISON);
 1805 #endif
 1806 #ifdef INET6
 1807         free(ip6, M_PRISON);
 1808 #endif
 1809         vfs_freeopts(opts);
 1810         return (error);
 1811 }
 1812 
 1813 
 1814 /*
 1815  * struct jail_get_args {
 1816  *      struct iovec *iovp;
 1817  *      unsigned int iovcnt;
 1818  *      int flags;
 1819  * };
 1820  */
 1821 int
 1822 sys_jail_get(struct thread *td, struct jail_get_args *uap)
 1823 {
 1824         struct uio *auio;
 1825         int error;
 1826 
 1827         /* Check that we have an even number of iovecs. */
 1828         if (uap->iovcnt & 1)
 1829                 return (EINVAL);
 1830 
 1831         error = copyinuio(uap->iovp, uap->iovcnt, &auio);
 1832         if (error)
 1833                 return (error);
 1834         error = kern_jail_get(td, auio, uap->flags);
 1835         if (error == 0)
 1836                 error = copyout(auio->uio_iov, uap->iovp,
 1837                     uap->iovcnt * sizeof (struct iovec));
 1838         free(auio, M_IOV);
 1839         return (error);
 1840 }
 1841 
 1842 int
 1843 kern_jail_get(struct thread *td, struct uio *optuio, int flags)
 1844 {
 1845         struct prison *pr, *mypr;
 1846         struct vfsopt *opt;
 1847         struct vfsoptlist *opts;
 1848         char *errmsg, *name;
 1849         int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos;
 1850 
 1851         if (flags & ~JAIL_GET_MASK)
 1852                 return (EINVAL);
 1853 
 1854         /* Get the parameter list. */
 1855         error = vfs_buildopts(optuio, &opts);
 1856         if (error)
 1857                 return (error);
 1858         errmsg_pos = vfs_getopt_pos(opts, "errmsg");
 1859         mypr = td->td_ucred->cr_prison;
 1860 
 1861         /*
 1862          * Find the prison specified by one of: lastjid, jid, name.
 1863          */
 1864         sx_slock(&allprison_lock);
 1865         error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid));
 1866         if (error == 0) {
 1867                 TAILQ_FOREACH(pr, &allprison, pr_list) {
 1868                         if (pr->pr_id > jid && prison_ischild(mypr, pr)) {
 1869                                 mtx_lock(&pr->pr_mtx);
 1870                                 if (pr->pr_ref > 0 &&
 1871                                     (pr->pr_uref > 0 || (flags & JAIL_DYING)))
 1872                                         break;
 1873                                 mtx_unlock(&pr->pr_mtx);
 1874                         }
 1875                 }
 1876                 if (pr != NULL)
 1877                         goto found_prison;
 1878                 error = ENOENT;
 1879                 vfs_opterror(opts, "no jail after %d", jid);
 1880                 goto done_unlock_list;
 1881         } else if (error != ENOENT)
 1882                 goto done_unlock_list;
 1883 
 1884         error = vfs_copyopt(opts, "jid", &jid, sizeof(jid));
 1885         if (error == 0) {
 1886                 if (jid != 0) {
 1887                         pr = prison_find_child(mypr, jid);
 1888                         if (pr != NULL) {
 1889                                 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
 1890                                         mtx_unlock(&pr->pr_mtx);
 1891                                         error = ENOENT;
 1892                                         vfs_opterror(opts, "jail %d is dying",
 1893                                             jid);
 1894                                         goto done_unlock_list;
 1895                                 }
 1896                                 goto found_prison;
 1897                         }
 1898                         error = ENOENT;
 1899                         vfs_opterror(opts, "jail %d not found", jid);
 1900                         goto done_unlock_list;
 1901                 }
 1902         } else if (error != ENOENT)
 1903                 goto done_unlock_list;
 1904 
 1905         error = vfs_getopt(opts, "name", (void **)&name, &len);
 1906         if (error == 0) {
 1907                 if (len == 0 || name[len - 1] != '\0') {
 1908                         error = EINVAL;
 1909                         goto done_unlock_list;
 1910                 }
 1911                 pr = prison_find_name(mypr, name);
 1912                 if (pr != NULL) {
 1913                         if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) {
 1914                                 mtx_unlock(&pr->pr_mtx);
 1915                                 error = ENOENT;
 1916                                 vfs_opterror(opts, "jail \"%s\" is dying",
 1917                                     name);
 1918                                 goto done_unlock_list;
 1919                         }
 1920                         goto found_prison;
 1921                 }
 1922                 error = ENOENT;
 1923                 vfs_opterror(opts, "jail \"%s\" not found", name);
 1924                 goto done_unlock_list;
 1925         } else if (error != ENOENT)
 1926                 goto done_unlock_list;
 1927 
 1928         vfs_opterror(opts, "no jail specified");
 1929         error = ENOENT;
 1930         goto done_unlock_list;
 1931 
 1932  found_prison:
 1933         /* Get the parameters of the prison. */
 1934         pr->pr_ref++;
 1935         locked = PD_LOCKED;
 1936         td->td_retval[0] = pr->pr_id;
 1937         error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id));
 1938         if (error != 0 && error != ENOENT)
 1939                 goto done_deref;
 1940         i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id;
 1941         error = vfs_setopt(opts, "parent", &i, sizeof(i));
 1942         if (error != 0 && error != ENOENT)
 1943                 goto done_deref;
 1944         error = vfs_setopts(opts, "name", prison_name(mypr, pr));
 1945         if (error != 0 && error != ENOENT)
 1946                 goto done_deref;
 1947         error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id,
 1948             sizeof(pr->pr_cpuset->cs_id));
 1949         if (error != 0 && error != ENOENT)
 1950                 goto done_deref;
 1951         error = vfs_setopts(opts, "path", prison_path(mypr, pr));
 1952         if (error != 0 && error != ENOENT)
 1953                 goto done_deref;
 1954 #ifdef INET
 1955         error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4,
 1956             pr->pr_ip4s * sizeof(*pr->pr_ip4));
 1957         if (error != 0 && error != ENOENT)
 1958                 goto done_deref;
 1959 #endif
 1960 #ifdef INET6
 1961         error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6,
 1962             pr->pr_ip6s * sizeof(*pr->pr_ip6));
 1963         if (error != 0 && error != ENOENT)
 1964                 goto done_deref;
 1965 #endif
 1966         error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel,
 1967             sizeof(pr->pr_securelevel));
 1968         if (error != 0 && error != ENOENT)
 1969                 goto done_deref;
 1970         error = vfs_setopt(opts, "children.cur", &pr->pr_childcount,
 1971             sizeof(pr->pr_childcount));
 1972         if (error != 0 && error != ENOENT)
 1973                 goto done_deref;
 1974         error = vfs_setopt(opts, "children.max", &pr->pr_childmax,
 1975             sizeof(pr->pr_childmax));
 1976         if (error != 0 && error != ENOENT)
 1977                 goto done_deref;
 1978         error = vfs_setopts(opts, "host.hostname", pr->pr_hostname);
 1979         if (error != 0 && error != ENOENT)
 1980                 goto done_deref;
 1981         error = vfs_setopts(opts, "host.domainname", pr->pr_domainname);
 1982         if (error != 0 && error != ENOENT)
 1983                 goto done_deref;
 1984         error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid);
 1985         if (error != 0 && error != ENOENT)
 1986                 goto done_deref;
 1987 #ifdef COMPAT_FREEBSD32
 1988         if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) {
 1989                 uint32_t hid32 = pr->pr_hostid;
 1990 
 1991                 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32));
 1992         } else
 1993 #endif
 1994         error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid,
 1995             sizeof(pr->pr_hostid));
 1996         if (error != 0 && error != ENOENT)
 1997                 goto done_deref;
 1998         error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs,
 1999             sizeof(pr->pr_enforce_statfs));
 2000         if (error != 0 && error != ENOENT)
 2001                 goto done_deref;
 2002         for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
 2003             fi++) {
 2004                 if (pr_flag_names[fi] == NULL)
 2005                         continue;
 2006                 i = (pr->pr_flags & (1 << fi)) ? 1 : 0;
 2007                 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i));
 2008                 if (error != 0 && error != ENOENT)
 2009                         goto done_deref;
 2010                 i = !i;
 2011                 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i));
 2012                 if (error != 0 && error != ENOENT)
 2013                         goto done_deref;
 2014         }
 2015         for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
 2016             fi++) {
 2017                 i = pr->pr_flags &
 2018                     (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
 2019                 i = pr_flag_jailsys[fi].disable &&
 2020                       (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE
 2021                     : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW
 2022                     : JAIL_SYS_INHERIT;
 2023                 error =
 2024                     vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i));
 2025                 if (error != 0 && error != ENOENT)
 2026                         goto done_deref;
 2027         }
 2028         for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
 2029             fi++) {
 2030                 if (pr_allow_names[fi] == NULL)
 2031                         continue;
 2032                 i = (pr->pr_allow & (1 << fi)) ? 1 : 0;
 2033                 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i));
 2034                 if (error != 0 && error != ENOENT)
 2035                         goto done_deref;
 2036                 i = !i;
 2037                 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i));
 2038                 if (error != 0 && error != ENOENT)
 2039                         goto done_deref;
 2040         }
 2041         i = (pr->pr_uref == 0);
 2042         error = vfs_setopt(opts, "dying", &i, sizeof(i));
 2043         if (error != 0 && error != ENOENT)
 2044                 goto done_deref;
 2045         i = !i;
 2046         error = vfs_setopt(opts, "nodying", &i, sizeof(i));
 2047         if (error != 0 && error != ENOENT)
 2048                 goto done_deref;
 2049 
 2050         /* Get the module parameters. */
 2051         mtx_unlock(&pr->pr_mtx);
 2052         locked = 0;
 2053         error = osd_jail_call(pr, PR_METHOD_GET, opts);
 2054         if (error)
 2055                 goto done_deref;
 2056         prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED);
 2057 
 2058         /* By now, all parameters should have been noted. */
 2059         TAILQ_FOREACH(opt, opts, link) {
 2060                 if (!opt->seen && strcmp(opt->name, "errmsg")) {
 2061                         error = EINVAL;
 2062                         vfs_opterror(opts, "unknown parameter: %s", opt->name);
 2063                         goto done_errmsg;
 2064                 }
 2065         }
 2066 
 2067         /* Write the fetched parameters back to userspace. */
 2068         error = 0;
 2069         TAILQ_FOREACH(opt, opts, link) {
 2070                 if (opt->pos >= 0 && opt->pos != errmsg_pos) {
 2071                         pos = 2 * opt->pos + 1;
 2072                         optuio->uio_iov[pos].iov_len = opt->len;
 2073                         if (opt->value != NULL) {
 2074                                 if (optuio->uio_segflg == UIO_SYSSPACE) {
 2075                                         bcopy(opt->value,
 2076                                             optuio->uio_iov[pos].iov_base,
 2077                                             opt->len);
 2078                                 } else {
 2079                                         error = copyout(opt->value,
 2080                                             optuio->uio_iov[pos].iov_base,
 2081                                             opt->len);
 2082                                         if (error)
 2083                                                 break;
 2084                                 }
 2085                         }
 2086                 }
 2087         }
 2088         goto done_errmsg;
 2089 
 2090  done_deref:
 2091         prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED);
 2092         goto done_errmsg;
 2093 
 2094  done_unlock_list:
 2095         sx_sunlock(&allprison_lock);
 2096  done_errmsg:
 2097         if (error && errmsg_pos >= 0) {
 2098                 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len);
 2099                 errmsg_pos = 2 * errmsg_pos + 1;
 2100                 if (errmsg_len > 0) {
 2101                         if (optuio->uio_segflg == UIO_SYSSPACE)
 2102                                 bcopy(errmsg,
 2103                                     optuio->uio_iov[errmsg_pos].iov_base,
 2104                                     errmsg_len);
 2105                         else
 2106                                 copyout(errmsg,
 2107                                     optuio->uio_iov[errmsg_pos].iov_base,
 2108                                     errmsg_len);
 2109                 }
 2110         }
 2111         vfs_freeopts(opts);
 2112         return (error);
 2113 }
 2114 
 2115 
 2116 /*
 2117  * struct jail_remove_args {
 2118  *      int jid;
 2119  * };
 2120  */
 2121 int
 2122 sys_jail_remove(struct thread *td, struct jail_remove_args *uap)
 2123 {
 2124         struct prison *pr, *cpr, *lpr, *tpr;
 2125         int descend, error;
 2126 
 2127         error = priv_check(td, PRIV_JAIL_REMOVE);
 2128         if (error)
 2129                 return (error);
 2130 
 2131         sx_xlock(&allprison_lock);
 2132         pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
 2133         if (pr == NULL) {
 2134                 sx_xunlock(&allprison_lock);
 2135                 return (EINVAL);
 2136         }
 2137 
 2138         /* Remove all descendants of this prison, then remove this prison. */
 2139         pr->pr_ref++;
 2140         pr->pr_flags |= PR_REMOVE;
 2141         if (!LIST_EMPTY(&pr->pr_children)) {
 2142                 mtx_unlock(&pr->pr_mtx);
 2143                 lpr = NULL;
 2144                 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
 2145                         mtx_lock(&cpr->pr_mtx);
 2146                         if (cpr->pr_ref > 0) {
 2147                                 tpr = cpr;
 2148                                 cpr->pr_ref++;
 2149                                 cpr->pr_flags |= PR_REMOVE;
 2150                         } else {
 2151                                 /* Already removed - do not do it again. */
 2152                                 tpr = NULL;
 2153                         }
 2154                         mtx_unlock(&cpr->pr_mtx);
 2155                         if (lpr != NULL) {
 2156                                 mtx_lock(&lpr->pr_mtx);
 2157                                 prison_remove_one(lpr);
 2158                                 sx_xlock(&allprison_lock);
 2159                         }
 2160                         lpr = tpr;
 2161                 }
 2162                 if (lpr != NULL) {
 2163                         mtx_lock(&lpr->pr_mtx);
 2164                         prison_remove_one(lpr);
 2165                         sx_xlock(&allprison_lock);
 2166                 }
 2167                 mtx_lock(&pr->pr_mtx);
 2168         }
 2169         prison_remove_one(pr);
 2170         return (0);
 2171 }
 2172 
 2173 static void
 2174 prison_remove_one(struct prison *pr)
 2175 {
 2176         struct proc *p;
 2177         int deuref;
 2178 
 2179         /* If the prison was persistent, it is not anymore. */
 2180         deuref = 0;
 2181         if (pr->pr_flags & PR_PERSIST) {
 2182                 pr->pr_ref--;
 2183                 deuref = PD_DEUREF;
 2184                 pr->pr_flags &= ~PR_PERSIST;
 2185         }
 2186 
 2187         /*
 2188          * jail_remove added a reference.  If that's the only one, remove
 2189          * the prison now.
 2190          */
 2191         KASSERT(pr->pr_ref > 0,
 2192             ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id));
 2193         if (pr->pr_ref == 1) {
 2194                 prison_deref(pr,
 2195                     deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED);
 2196                 return;
 2197         }
 2198 
 2199         mtx_unlock(&pr->pr_mtx);
 2200         sx_xunlock(&allprison_lock);
 2201         /*
 2202          * Kill all processes unfortunate enough to be attached to this prison.
 2203          */
 2204         sx_slock(&allproc_lock);
 2205         LIST_FOREACH(p, &allproc, p_list) {
 2206                 PROC_LOCK(p);
 2207                 if (p->p_state != PRS_NEW && p->p_ucred &&
 2208                     p->p_ucred->cr_prison == pr)
 2209                         kern_psignal(p, SIGKILL);
 2210                 PROC_UNLOCK(p);
 2211         }
 2212         sx_sunlock(&allproc_lock);
 2213         /* Remove the temporary reference added by jail_remove. */
 2214         prison_deref(pr, deuref | PD_DEREF);
 2215 }
 2216 
 2217 
 2218 /*
 2219  * struct jail_attach_args {
 2220  *      int jid;
 2221  * };
 2222  */
 2223 int
 2224 sys_jail_attach(struct thread *td, struct jail_attach_args *uap)
 2225 {
 2226         struct prison *pr;
 2227         int error;
 2228 
 2229         error = priv_check(td, PRIV_JAIL_ATTACH);
 2230         if (error)
 2231                 return (error);
 2232 
 2233         sx_slock(&allprison_lock);
 2234         pr = prison_find_child(td->td_ucred->cr_prison, uap->jid);
 2235         if (pr == NULL) {
 2236                 sx_sunlock(&allprison_lock);
 2237                 return (EINVAL);
 2238         }
 2239 
 2240         /*
 2241          * Do not allow a process to attach to a prison that is not
 2242          * considered to be "alive".
 2243          */
 2244         if (pr->pr_uref == 0) {
 2245                 mtx_unlock(&pr->pr_mtx);
 2246                 sx_sunlock(&allprison_lock);
 2247                 return (EINVAL);
 2248         }
 2249 
 2250         return (do_jail_attach(td, pr));
 2251 }
 2252 
 2253 static int
 2254 do_jail_attach(struct thread *td, struct prison *pr)
 2255 {
 2256         struct prison *ppr;
 2257         struct proc *p;
 2258         struct ucred *newcred, *oldcred;
 2259         int vfslocked, error;
 2260 
 2261         /*
 2262          * XXX: Note that there is a slight race here if two threads
 2263          * in the same privileged process attempt to attach to two
 2264          * different jails at the same time.  It is important for
 2265          * user processes not to do this, or they might end up with
 2266          * a process root from one prison, but attached to the jail
 2267          * of another.
 2268          */
 2269         pr->pr_ref++;
 2270         pr->pr_uref++;
 2271         mtx_unlock(&pr->pr_mtx);
 2272 
 2273         /* Let modules do whatever they need to prepare for attaching. */
 2274         error = osd_jail_call(pr, PR_METHOD_ATTACH, td);
 2275         if (error) {
 2276                 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED);
 2277                 return (error);
 2278         }
 2279         sx_sunlock(&allprison_lock);
 2280 
 2281         /*
 2282          * Reparent the newly attached process to this jail.
 2283          */
 2284         ppr = td->td_ucred->cr_prison;
 2285         p = td->td_proc;
 2286         error = cpuset_setproc_update_set(p, pr->pr_cpuset);
 2287         if (error)
 2288                 goto e_revert_osd;
 2289 
 2290         vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
 2291         vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY);
 2292         if ((error = change_dir(pr->pr_root, td)) != 0)
 2293                 goto e_unlock;
 2294 #ifdef MAC
 2295         if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root)))
 2296                 goto e_unlock;
 2297 #endif
 2298         VOP_UNLOCK(pr->pr_root, 0);
 2299         if ((error = change_root(pr->pr_root, td)))
 2300                 goto e_unlock_giant;
 2301         VFS_UNLOCK_GIANT(vfslocked);
 2302 
 2303         newcred = crget();
 2304         PROC_LOCK(p);
 2305         oldcred = p->p_ucred;
 2306         setsugid(p);
 2307         crcopy(newcred, oldcred);
 2308         newcred->cr_prison = pr;
 2309         p->p_ucred = newcred;
 2310         PROC_UNLOCK(p);
 2311 #ifdef RACCT
 2312         racct_proc_ucred_changed(p, oldcred, newcred);
 2313 #endif
 2314         crfree(oldcred);
 2315         prison_deref(ppr, PD_DEREF | PD_DEUREF);
 2316         return (0);
 2317  e_unlock:
 2318         VOP_UNLOCK(pr->pr_root, 0);
 2319  e_unlock_giant:
 2320         VFS_UNLOCK_GIANT(vfslocked);
 2321  e_revert_osd:
 2322         /* Tell modules this thread is still in its old jail after all. */
 2323         (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td);
 2324         prison_deref(pr, PD_DEREF | PD_DEUREF);
 2325         return (error);
 2326 }
 2327 
 2328 
 2329 /*
 2330  * Returns a locked prison instance, or NULL on failure.
 2331  */
 2332 struct prison *
 2333 prison_find(int prid)
 2334 {
 2335         struct prison *pr;
 2336 
 2337         sx_assert(&allprison_lock, SX_LOCKED);
 2338         TAILQ_FOREACH(pr, &allprison, pr_list) {
 2339                 if (pr->pr_id == prid) {
 2340                         mtx_lock(&pr->pr_mtx);
 2341                         if (pr->pr_ref > 0)
 2342                                 return (pr);
 2343                         mtx_unlock(&pr->pr_mtx);
 2344                 }
 2345         }
 2346         return (NULL);
 2347 }
 2348 
 2349 /*
 2350  * Find a prison that is a descendant of mypr.  Returns a locked prison or NULL.
 2351  */
 2352 struct prison *
 2353 prison_find_child(struct prison *mypr, int prid)
 2354 {
 2355         struct prison *pr;
 2356         int descend;
 2357 
 2358         sx_assert(&allprison_lock, SX_LOCKED);
 2359         FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
 2360                 if (pr->pr_id == prid) {
 2361                         mtx_lock(&pr->pr_mtx);
 2362                         if (pr->pr_ref > 0)
 2363                                 return (pr);
 2364                         mtx_unlock(&pr->pr_mtx);
 2365                 }
 2366         }
 2367         return (NULL);
 2368 }
 2369 
 2370 /*
 2371  * Look for the name relative to mypr.  Returns a locked prison or NULL.
 2372  */
 2373 struct prison *
 2374 prison_find_name(struct prison *mypr, const char *name)
 2375 {
 2376         struct prison *pr, *deadpr;
 2377         size_t mylen;
 2378         int descend;
 2379 
 2380         sx_assert(&allprison_lock, SX_LOCKED);
 2381         mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1;
 2382  again:
 2383         deadpr = NULL;
 2384         FOREACH_PRISON_DESCENDANT(mypr, pr, descend) {
 2385                 if (!strcmp(pr->pr_name + mylen, name)) {
 2386                         mtx_lock(&pr->pr_mtx);
 2387                         if (pr->pr_ref > 0) {
 2388                                 if (pr->pr_uref > 0)
 2389                                         return (pr);
 2390                                 deadpr = pr;
 2391                         }
 2392                         mtx_unlock(&pr->pr_mtx);
 2393                 }
 2394         }
 2395         /* There was no valid prison - perhaps there was a dying one. */
 2396         if (deadpr != NULL) {
 2397                 mtx_lock(&deadpr->pr_mtx);
 2398                 if (deadpr->pr_ref == 0) {
 2399                         mtx_unlock(&deadpr->pr_mtx);
 2400                         goto again;
 2401                 }
 2402         }
 2403         return (deadpr);
 2404 }
 2405 
 2406 /*
 2407  * See if a prison has the specific flag set.
 2408  */
 2409 int
 2410 prison_flag(struct ucred *cred, unsigned flag)
 2411 {
 2412 
 2413         /* This is an atomic read, so no locking is necessary. */
 2414         return (cred->cr_prison->pr_flags & flag);
 2415 }
 2416 
 2417 int
 2418 prison_allow(struct ucred *cred, unsigned flag)
 2419 {
 2420 
 2421         /* This is an atomic read, so no locking is necessary. */
 2422         return (cred->cr_prison->pr_allow & flag);
 2423 }
 2424 
 2425 /*
 2426  * Remove a prison reference.  If that was the last reference, remove the
 2427  * prison itself - but not in this context in case there are locks held.
 2428  */
 2429 void
 2430 prison_free_locked(struct prison *pr)
 2431 {
 2432 
 2433         mtx_assert(&pr->pr_mtx, MA_OWNED);
 2434         pr->pr_ref--;
 2435         if (pr->pr_ref == 0) {
 2436                 mtx_unlock(&pr->pr_mtx);
 2437                 TASK_INIT(&pr->pr_task, 0, prison_complete, pr);
 2438                 taskqueue_enqueue(taskqueue_thread, &pr->pr_task);
 2439                 return;
 2440         }
 2441         mtx_unlock(&pr->pr_mtx);
 2442 }
 2443 
 2444 void
 2445 prison_free(struct prison *pr)
 2446 {
 2447 
 2448         mtx_lock(&pr->pr_mtx);
 2449         prison_free_locked(pr);
 2450 }
 2451 
 2452 static void
 2453 prison_complete(void *context, int pending)
 2454 {
 2455 
 2456         prison_deref((struct prison *)context, 0);
 2457 }
 2458 
 2459 /*
 2460  * Remove a prison reference (usually).  This internal version assumes no
 2461  * mutexes are held, except perhaps the prison itself.  If there are no more
 2462  * references, release and delist the prison.  On completion, the prison lock
 2463  * and the allprison lock are both unlocked.
 2464  */
 2465 static void
 2466 prison_deref(struct prison *pr, int flags)
 2467 {
 2468         struct prison *ppr, *tpr;
 2469         int vfslocked;
 2470 
 2471         if (!(flags & PD_LOCKED))
 2472                 mtx_lock(&pr->pr_mtx);
 2473         for (;;) {
 2474                 if (flags & PD_DEUREF) {
 2475                         pr->pr_uref--;
 2476                         KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0"));
 2477                 }
 2478                 if (flags & PD_DEREF)
 2479                         pr->pr_ref--;
 2480                 /* If the prison still has references, nothing else to do. */
 2481                 if (pr->pr_ref > 0) {
 2482                         mtx_unlock(&pr->pr_mtx);
 2483                         if (flags & PD_LIST_SLOCKED)
 2484                                 sx_sunlock(&allprison_lock);
 2485                         else if (flags & PD_LIST_XLOCKED)
 2486                                 sx_xunlock(&allprison_lock);
 2487                         return;
 2488                 }
 2489 
 2490                 mtx_unlock(&pr->pr_mtx);
 2491                 if (flags & PD_LIST_SLOCKED) {
 2492                         if (!sx_try_upgrade(&allprison_lock)) {
 2493                                 sx_sunlock(&allprison_lock);
 2494                                 sx_xlock(&allprison_lock);
 2495                         }
 2496                 } else if (!(flags & PD_LIST_XLOCKED))
 2497                         sx_xlock(&allprison_lock);
 2498 
 2499                 TAILQ_REMOVE(&allprison, pr, pr_list);
 2500                 LIST_REMOVE(pr, pr_sibling);
 2501                 ppr = pr->pr_parent;
 2502                 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent)
 2503                         tpr->pr_childcount--;
 2504                 sx_xunlock(&allprison_lock);
 2505 
 2506 #ifdef VIMAGE
 2507                 if (pr->pr_vnet != ppr->pr_vnet)
 2508                         vnet_destroy(pr->pr_vnet);
 2509 #endif
 2510                 if (pr->pr_root != NULL) {
 2511                         vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount);
 2512                         vrele(pr->pr_root);
 2513                         VFS_UNLOCK_GIANT(vfslocked);
 2514                 }
 2515                 mtx_destroy(&pr->pr_mtx);
 2516 #ifdef INET
 2517                 free(pr->pr_ip4, M_PRISON);
 2518 #endif
 2519 #ifdef INET6
 2520                 free(pr->pr_ip6, M_PRISON);
 2521 #endif
 2522                 if (pr->pr_cpuset != NULL)
 2523                         cpuset_rel(pr->pr_cpuset);
 2524                 osd_jail_exit(pr);
 2525 #ifdef RACCT
 2526                 prison_racct_detach(pr);
 2527 #endif
 2528                 free(pr, M_PRISON);
 2529 
 2530                 /* Removing a prison frees a reference on its parent. */
 2531                 pr = ppr;
 2532                 mtx_lock(&pr->pr_mtx);
 2533                 flags = PD_DEREF | PD_DEUREF;
 2534         }
 2535 }
 2536 
 2537 void
 2538 prison_hold_locked(struct prison *pr)
 2539 {
 2540 
 2541         mtx_assert(&pr->pr_mtx, MA_OWNED);
 2542         KASSERT(pr->pr_ref > 0,
 2543             ("Trying to hold dead prison (jid=%d).", pr->pr_id));
 2544         pr->pr_ref++;
 2545 }
 2546 
 2547 void
 2548 prison_hold(struct prison *pr)
 2549 {
 2550 
 2551         mtx_lock(&pr->pr_mtx);
 2552         prison_hold_locked(pr);
 2553         mtx_unlock(&pr->pr_mtx);
 2554 }
 2555 
 2556 void
 2557 prison_proc_hold(struct prison *pr)
 2558 {
 2559 
 2560         mtx_lock(&pr->pr_mtx);
 2561         KASSERT(pr->pr_uref > 0,
 2562             ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id));
 2563         pr->pr_uref++;
 2564         mtx_unlock(&pr->pr_mtx);
 2565 }
 2566 
 2567 void
 2568 prison_proc_free(struct prison *pr)
 2569 {
 2570 
 2571         mtx_lock(&pr->pr_mtx);
 2572         KASSERT(pr->pr_uref > 0,
 2573             ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id));
 2574         prison_deref(pr, PD_DEUREF | PD_LOCKED);
 2575 }
 2576 
 2577 
 2578 #ifdef INET
 2579 /*
 2580  * Restrict a prison's IP address list with its parent's, possibly replacing
 2581  * it.  Return true if the replacement buffer was used (or would have been).
 2582  */
 2583 static int
 2584 prison_restrict_ip4(struct prison *pr, struct in_addr *newip4)
 2585 {
 2586         int ii, ij, used;
 2587         struct prison *ppr;
 2588 
 2589         ppr = pr->pr_parent;
 2590         if (!(pr->pr_flags & PR_IP4_USER)) {
 2591                 /* This has no user settings, so just copy the parent's list. */
 2592                 if (pr->pr_ip4s < ppr->pr_ip4s) {
 2593                         /*
 2594                          * There's no room for the parent's list.  Use the
 2595                          * new list buffer, which is assumed to be big enough
 2596                          * (if it was passed).  If there's no buffer, try to
 2597                          * allocate one.
 2598                          */
 2599                         used = 1;
 2600                         if (newip4 == NULL) {
 2601                                 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4),
 2602                                     M_PRISON, M_NOWAIT);
 2603                                 if (newip4 != NULL)
 2604                                         used = 0;
 2605                         }
 2606                         if (newip4 != NULL) {
 2607                                 bcopy(ppr->pr_ip4, newip4,
 2608                                     ppr->pr_ip4s * sizeof(*newip4));
 2609                                 free(pr->pr_ip4, M_PRISON);
 2610                                 pr->pr_ip4 = newip4;
 2611                                 pr->pr_ip4s = ppr->pr_ip4s;
 2612                         }
 2613                         return (used);
 2614                 }
 2615                 pr->pr_ip4s = ppr->pr_ip4s;
 2616                 if (pr->pr_ip4s > 0)
 2617                         bcopy(ppr->pr_ip4, pr->pr_ip4,
 2618                             pr->pr_ip4s * sizeof(*newip4));
 2619                 else if (pr->pr_ip4 != NULL) {
 2620                         free(pr->pr_ip4, M_PRISON);
 2621                         pr->pr_ip4 = NULL;
 2622                 }
 2623         } else if (pr->pr_ip4s > 0) {
 2624                 /* Remove addresses that aren't in the parent. */
 2625                 for (ij = 0; ij < ppr->pr_ip4s; ij++)
 2626                         if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr)
 2627                                 break;
 2628                 if (ij < ppr->pr_ip4s)
 2629                         ii = 1;
 2630                 else {
 2631                         bcopy(pr->pr_ip4 + 1, pr->pr_ip4,
 2632                             --pr->pr_ip4s * sizeof(*pr->pr_ip4));
 2633                         ii = 0;
 2634                 }
 2635                 for (ij = 1; ii < pr->pr_ip4s; ) {
 2636                         if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) {
 2637                                 ii++;
 2638                                 continue;
 2639                         }
 2640                         switch (ij >= ppr->pr_ip4s ? -1 :
 2641                                 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) {
 2642                         case -1:
 2643                                 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii,
 2644                                     (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4));
 2645                                 break;
 2646                         case 0:
 2647                                 ii++;
 2648                                 ij++;
 2649                                 break;
 2650                         case 1:
 2651                                 ij++;
 2652                                 break;
 2653                         }
 2654                 }
 2655                 if (pr->pr_ip4s == 0) {
 2656                         pr->pr_flags |= PR_IP4_DISABLE;
 2657                         free(pr->pr_ip4, M_PRISON);
 2658                         pr->pr_ip4 = NULL;
 2659                 }
 2660         }
 2661         return (0);
 2662 }
 2663 
 2664 /*
 2665  * Pass back primary IPv4 address of this jail.
 2666  *
 2667  * If not restricted return success but do not alter the address.  Caller has
 2668  * to make sure to initialize it correctly (e.g. INADDR_ANY).
 2669  *
 2670  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
 2671  * Address returned in NBO.
 2672  */
 2673 int
 2674 prison_get_ip4(struct ucred *cred, struct in_addr *ia)
 2675 {
 2676         struct prison *pr;
 2677 
 2678         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2679         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2680 
 2681         pr = cred->cr_prison;
 2682         if (!(pr->pr_flags & PR_IP4))
 2683                 return (0);
 2684         mtx_lock(&pr->pr_mtx);
 2685         if (!(pr->pr_flags & PR_IP4)) {
 2686                 mtx_unlock(&pr->pr_mtx);
 2687                 return (0);
 2688         }
 2689         if (pr->pr_ip4 == NULL) {
 2690                 mtx_unlock(&pr->pr_mtx);
 2691                 return (EAFNOSUPPORT);
 2692         }
 2693 
 2694         ia->s_addr = pr->pr_ip4[0].s_addr;
 2695         mtx_unlock(&pr->pr_mtx);
 2696         return (0);
 2697 }
 2698 
 2699 /*
 2700  * Return 1 if we should do proper source address selection or are not jailed.
 2701  * We will return 0 if we should bypass source address selection in favour
 2702  * of the primary jail IPv4 address. Only in this case *ia will be updated and
 2703  * returned in NBO.
 2704  * Return EAFNOSUPPORT, in case this jail does not allow IPv4.
 2705  */
 2706 int
 2707 prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia)
 2708 {
 2709         struct prison *pr;
 2710         struct in_addr lia;
 2711         int error;
 2712 
 2713         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2714         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2715 
 2716         if (!jailed(cred))
 2717                 return (1);
 2718 
 2719         pr = cred->cr_prison;
 2720         if (pr->pr_flags & PR_IP4_SADDRSEL)
 2721                 return (1);
 2722 
 2723         lia.s_addr = INADDR_ANY;
 2724         error = prison_get_ip4(cred, &lia);
 2725         if (error)
 2726                 return (error);
 2727         if (lia.s_addr == INADDR_ANY)
 2728                 return (1);
 2729 
 2730         ia->s_addr = lia.s_addr;
 2731         return (0);
 2732 }
 2733 
 2734 /*
 2735  * Return true if pr1 and pr2 have the same IPv4 address restrictions.
 2736  */
 2737 int
 2738 prison_equal_ip4(struct prison *pr1, struct prison *pr2)
 2739 {
 2740 
 2741         if (pr1 == pr2)
 2742                 return (1);
 2743 
 2744         /*
 2745          * No need to lock since the PR_IP4_USER flag can't be altered for
 2746          * existing prisons.
 2747          */
 2748         while (pr1 != &prison0 &&
 2749 #ifdef VIMAGE
 2750                !(pr1->pr_flags & PR_VNET) &&
 2751 #endif
 2752                !(pr1->pr_flags & PR_IP4_USER))
 2753                 pr1 = pr1->pr_parent;
 2754         while (pr2 != &prison0 &&
 2755 #ifdef VIMAGE
 2756                !(pr2->pr_flags & PR_VNET) &&
 2757 #endif
 2758                !(pr2->pr_flags & PR_IP4_USER))
 2759                 pr2 = pr2->pr_parent;
 2760         return (pr1 == pr2);
 2761 }
 2762 
 2763 /*
 2764  * Make sure our (source) address is set to something meaningful to this
 2765  * jail.
 2766  *
 2767  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
 2768  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 2769  * doesn't allow IPv4.  Address passed in in NBO and returned in NBO.
 2770  */
 2771 int
 2772 prison_local_ip4(struct ucred *cred, struct in_addr *ia)
 2773 {
 2774         struct prison *pr;
 2775         struct in_addr ia0;
 2776         int error;
 2777 
 2778         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2779         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2780 
 2781         pr = cred->cr_prison;
 2782         if (!(pr->pr_flags & PR_IP4))
 2783                 return (0);
 2784         mtx_lock(&pr->pr_mtx);
 2785         if (!(pr->pr_flags & PR_IP4)) {
 2786                 mtx_unlock(&pr->pr_mtx);
 2787                 return (0);
 2788         }
 2789         if (pr->pr_ip4 == NULL) {
 2790                 mtx_unlock(&pr->pr_mtx);
 2791                 return (EAFNOSUPPORT);
 2792         }
 2793 
 2794         ia0.s_addr = ntohl(ia->s_addr);
 2795         if (ia0.s_addr == INADDR_LOOPBACK) {
 2796                 ia->s_addr = pr->pr_ip4[0].s_addr;
 2797                 mtx_unlock(&pr->pr_mtx);
 2798                 return (0);
 2799         }
 2800 
 2801         if (ia0.s_addr == INADDR_ANY) {
 2802                 /*
 2803                  * In case there is only 1 IPv4 address, bind directly.
 2804                  */
 2805                 if (pr->pr_ip4s == 1)
 2806                         ia->s_addr = pr->pr_ip4[0].s_addr;
 2807                 mtx_unlock(&pr->pr_mtx);
 2808                 return (0);
 2809         }
 2810 
 2811         error = _prison_check_ip4(pr, ia);
 2812         mtx_unlock(&pr->pr_mtx);
 2813         return (error);
 2814 }
 2815 
 2816 /*
 2817  * Rewrite destination address in case we will connect to loopback address.
 2818  *
 2819  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4.
 2820  * Address passed in in NBO and returned in NBO.
 2821  */
 2822 int
 2823 prison_remote_ip4(struct ucred *cred, struct in_addr *ia)
 2824 {
 2825         struct prison *pr;
 2826 
 2827         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2828         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2829 
 2830         pr = cred->cr_prison;
 2831         if (!(pr->pr_flags & PR_IP4))
 2832                 return (0);
 2833         mtx_lock(&pr->pr_mtx);
 2834         if (!(pr->pr_flags & PR_IP4)) {
 2835                 mtx_unlock(&pr->pr_mtx);
 2836                 return (0);
 2837         }
 2838         if (pr->pr_ip4 == NULL) {
 2839                 mtx_unlock(&pr->pr_mtx);
 2840                 return (EAFNOSUPPORT);
 2841         }
 2842 
 2843         if (ntohl(ia->s_addr) == INADDR_LOOPBACK) {
 2844                 ia->s_addr = pr->pr_ip4[0].s_addr;
 2845                 mtx_unlock(&pr->pr_mtx);
 2846                 return (0);
 2847         }
 2848 
 2849         /*
 2850          * Return success because nothing had to be changed.
 2851          */
 2852         mtx_unlock(&pr->pr_mtx);
 2853         return (0);
 2854 }
 2855 
 2856 /*
 2857  * Check if given address belongs to the jail referenced by cred/prison.
 2858  *
 2859  * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail,
 2860  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 2861  * doesn't allow IPv4.  Address passed in in NBO.
 2862  */
 2863 static int
 2864 _prison_check_ip4(struct prison *pr, struct in_addr *ia)
 2865 {
 2866         int i, a, z, d;
 2867 
 2868         /*
 2869          * Check the primary IP.
 2870          */
 2871         if (pr->pr_ip4[0].s_addr == ia->s_addr)
 2872                 return (0);
 2873 
 2874         /*
 2875          * All the other IPs are sorted so we can do a binary search.
 2876          */
 2877         a = 0;
 2878         z = pr->pr_ip4s - 2;
 2879         while (a <= z) {
 2880                 i = (a + z) / 2;
 2881                 d = qcmp_v4(&pr->pr_ip4[i+1], ia);
 2882                 if (d > 0)
 2883                         z = i - 1;
 2884                 else if (d < 0)
 2885                         a = i + 1;
 2886                 else
 2887                         return (0);
 2888         }
 2889 
 2890         return (EADDRNOTAVAIL);
 2891 }
 2892 
 2893 int
 2894 prison_check_ip4(struct ucred *cred, struct in_addr *ia)
 2895 {
 2896         struct prison *pr;
 2897         int error;
 2898 
 2899         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 2900         KASSERT(ia != NULL, ("%s: ia is NULL", __func__));
 2901 
 2902         pr = cred->cr_prison;
 2903         if (!(pr->pr_flags & PR_IP4))
 2904                 return (0);
 2905         mtx_lock(&pr->pr_mtx);
 2906         if (!(pr->pr_flags & PR_IP4)) {
 2907                 mtx_unlock(&pr->pr_mtx);
 2908                 return (0);
 2909         }
 2910         if (pr->pr_ip4 == NULL) {
 2911                 mtx_unlock(&pr->pr_mtx);
 2912                 return (EAFNOSUPPORT);
 2913         }
 2914 
 2915         error = _prison_check_ip4(pr, ia);
 2916         mtx_unlock(&pr->pr_mtx);
 2917         return (error);
 2918 }
 2919 #endif
 2920 
 2921 #ifdef INET6
 2922 static int
 2923 prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6)
 2924 {
 2925         int ii, ij, used;
 2926         struct prison *ppr;
 2927 
 2928         ppr = pr->pr_parent;
 2929         if (!(pr->pr_flags & PR_IP6_USER)) {
 2930                 /* This has no user settings, so just copy the parent's list. */
 2931                 if (pr->pr_ip6s < ppr->pr_ip6s) {
 2932                         /*
 2933                          * There's no room for the parent's list.  Use the
 2934                          * new list buffer, which is assumed to be big enough
 2935                          * (if it was passed).  If there's no buffer, try to
 2936                          * allocate one.
 2937                          */
 2938                         used = 1;
 2939                         if (newip6 == NULL) {
 2940                                 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6),
 2941                                     M_PRISON, M_NOWAIT);
 2942                                 if (newip6 != NULL)
 2943                                         used = 0;
 2944                         }
 2945                         if (newip6 != NULL) {
 2946                                 bcopy(ppr->pr_ip6, newip6,
 2947                                     ppr->pr_ip6s * sizeof(*newip6));
 2948                                 free(pr->pr_ip6, M_PRISON);
 2949                                 pr->pr_ip6 = newip6;
 2950                                 pr->pr_ip6s = ppr->pr_ip6s;
 2951                         }
 2952                         return (used);
 2953                 }
 2954                 pr->pr_ip6s = ppr->pr_ip6s;
 2955                 if (pr->pr_ip6s > 0)
 2956                         bcopy(ppr->pr_ip6, pr->pr_ip6,
 2957                             pr->pr_ip6s * sizeof(*newip6));
 2958                 else if (pr->pr_ip6 != NULL) {
 2959                         free(pr->pr_ip6, M_PRISON);
 2960                         pr->pr_ip6 = NULL;
 2961                 }
 2962         } else if (pr->pr_ip6s > 0) {
 2963                 /* Remove addresses that aren't in the parent. */
 2964                 for (ij = 0; ij < ppr->pr_ip6s; ij++)
 2965                         if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0],
 2966                             &ppr->pr_ip6[ij]))
 2967                                 break;
 2968                 if (ij < ppr->pr_ip6s)
 2969                         ii = 1;
 2970                 else {
 2971                         bcopy(pr->pr_ip6 + 1, pr->pr_ip6,
 2972                             --pr->pr_ip6s * sizeof(*pr->pr_ip6));
 2973                         ii = 0;
 2974                 }
 2975                 for (ij = 1; ii < pr->pr_ip6s; ) {
 2976                         if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii],
 2977                             &ppr->pr_ip6[0])) {
 2978                                 ii++;
 2979                                 continue;
 2980                         }
 2981                         switch (ij >= ppr->pr_ip4s ? -1 :
 2982                                 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) {
 2983                         case -1:
 2984                                 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii,
 2985                                     (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6));
 2986                                 break;
 2987                         case 0:
 2988                                 ii++;
 2989                                 ij++;
 2990                                 break;
 2991                         case 1:
 2992                                 ij++;
 2993                                 break;
 2994                         }
 2995                 }
 2996                 if (pr->pr_ip6s == 0) {
 2997                         pr->pr_flags |= PR_IP6_DISABLE;
 2998                         free(pr->pr_ip6, M_PRISON);
 2999                         pr->pr_ip6 = NULL;
 3000                 }
 3001         }
 3002         return 0;
 3003 }
 3004 
 3005 /*
 3006  * Pass back primary IPv6 address for this jail.
 3007  *
 3008  * If not restricted return success but do not alter the address.  Caller has
 3009  * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT).
 3010  *
 3011  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
 3012  */
 3013 int
 3014 prison_get_ip6(struct ucred *cred, struct in6_addr *ia6)
 3015 {
 3016         struct prison *pr;
 3017 
 3018         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3019         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3020 
 3021         pr = cred->cr_prison;
 3022         if (!(pr->pr_flags & PR_IP6))
 3023                 return (0);
 3024         mtx_lock(&pr->pr_mtx);
 3025         if (!(pr->pr_flags & PR_IP6)) {
 3026                 mtx_unlock(&pr->pr_mtx);
 3027                 return (0);
 3028         }
 3029         if (pr->pr_ip6 == NULL) {
 3030                 mtx_unlock(&pr->pr_mtx);
 3031                 return (EAFNOSUPPORT);
 3032         }
 3033 
 3034         bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3035         mtx_unlock(&pr->pr_mtx);
 3036         return (0);
 3037 }
 3038 
 3039 /*
 3040  * Return 1 if we should do proper source address selection or are not jailed.
 3041  * We will return 0 if we should bypass source address selection in favour
 3042  * of the primary jail IPv6 address. Only in this case *ia will be updated and
 3043  * returned in NBO.
 3044  * Return EAFNOSUPPORT, in case this jail does not allow IPv6.
 3045  */
 3046 int
 3047 prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6)
 3048 {
 3049         struct prison *pr;
 3050         struct in6_addr lia6;
 3051         int error;
 3052 
 3053         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3054         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3055 
 3056         if (!jailed(cred))
 3057                 return (1);
 3058 
 3059         pr = cred->cr_prison;
 3060         if (pr->pr_flags & PR_IP6_SADDRSEL)
 3061                 return (1);
 3062 
 3063         lia6 = in6addr_any;
 3064         error = prison_get_ip6(cred, &lia6);
 3065         if (error)
 3066                 return (error);
 3067         if (IN6_IS_ADDR_UNSPECIFIED(&lia6))
 3068                 return (1);
 3069 
 3070         bcopy(&lia6, ia6, sizeof(struct in6_addr));
 3071         return (0);
 3072 }
 3073 
 3074 /*
 3075  * Return true if pr1 and pr2 have the same IPv6 address restrictions.
 3076  */
 3077 int
 3078 prison_equal_ip6(struct prison *pr1, struct prison *pr2)
 3079 {
 3080 
 3081         if (pr1 == pr2)
 3082                 return (1);
 3083 
 3084         while (pr1 != &prison0 &&
 3085 #ifdef VIMAGE
 3086                !(pr1->pr_flags & PR_VNET) &&
 3087 #endif
 3088                !(pr1->pr_flags & PR_IP6_USER))
 3089                 pr1 = pr1->pr_parent;
 3090         while (pr2 != &prison0 &&
 3091 #ifdef VIMAGE
 3092                !(pr2->pr_flags & PR_VNET) &&
 3093 #endif
 3094                !(pr2->pr_flags & PR_IP6_USER))
 3095                 pr2 = pr2->pr_parent;
 3096         return (pr1 == pr2);
 3097 }
 3098 
 3099 /*
 3100  * Make sure our (source) address is set to something meaningful to this jail.
 3101  *
 3102  * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0)
 3103  * when needed while binding.
 3104  *
 3105  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
 3106  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 3107  * doesn't allow IPv6.
 3108  */
 3109 int
 3110 prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only)
 3111 {
 3112         struct prison *pr;
 3113         int error;
 3114 
 3115         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3116         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3117 
 3118         pr = cred->cr_prison;
 3119         if (!(pr->pr_flags & PR_IP6))
 3120                 return (0);
 3121         mtx_lock(&pr->pr_mtx);
 3122         if (!(pr->pr_flags & PR_IP6)) {
 3123                 mtx_unlock(&pr->pr_mtx);
 3124                 return (0);
 3125         }
 3126         if (pr->pr_ip6 == NULL) {
 3127                 mtx_unlock(&pr->pr_mtx);
 3128                 return (EAFNOSUPPORT);
 3129         }
 3130 
 3131         if (IN6_IS_ADDR_LOOPBACK(ia6)) {
 3132                 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3133                 mtx_unlock(&pr->pr_mtx);
 3134                 return (0);
 3135         }
 3136 
 3137         if (IN6_IS_ADDR_UNSPECIFIED(ia6)) {
 3138                 /*
 3139                  * In case there is only 1 IPv6 address, and v6only is true,
 3140                  * then bind directly.
 3141                  */
 3142                 if (v6only != 0 && pr->pr_ip6s == 1)
 3143                         bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3144                 mtx_unlock(&pr->pr_mtx);
 3145                 return (0);
 3146         }
 3147 
 3148         error = _prison_check_ip6(pr, ia6);
 3149         mtx_unlock(&pr->pr_mtx);
 3150         return (error);
 3151 }
 3152 
 3153 /*
 3154  * Rewrite destination address in case we will connect to loopback address.
 3155  *
 3156  * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6.
 3157  */
 3158 int
 3159 prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6)
 3160 {
 3161         struct prison *pr;
 3162 
 3163         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3164         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3165 
 3166         pr = cred->cr_prison;
 3167         if (!(pr->pr_flags & PR_IP6))
 3168                 return (0);
 3169         mtx_lock(&pr->pr_mtx);
 3170         if (!(pr->pr_flags & PR_IP6)) {
 3171                 mtx_unlock(&pr->pr_mtx);
 3172                 return (0);
 3173         }
 3174         if (pr->pr_ip6 == NULL) {
 3175                 mtx_unlock(&pr->pr_mtx);
 3176                 return (EAFNOSUPPORT);
 3177         }
 3178 
 3179         if (IN6_IS_ADDR_LOOPBACK(ia6)) {
 3180                 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr));
 3181                 mtx_unlock(&pr->pr_mtx);
 3182                 return (0);
 3183         }
 3184 
 3185         /*
 3186          * Return success because nothing had to be changed.
 3187          */
 3188         mtx_unlock(&pr->pr_mtx);
 3189         return (0);
 3190 }
 3191 
 3192 /*
 3193  * Check if given address belongs to the jail referenced by cred/prison.
 3194  *
 3195  * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail,
 3196  * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail
 3197  * doesn't allow IPv6.
 3198  */
 3199 static int
 3200 _prison_check_ip6(struct prison *pr, struct in6_addr *ia6)
 3201 {
 3202         int i, a, z, d;
 3203 
 3204         /*
 3205          * Check the primary IP.
 3206          */
 3207         if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6))
 3208                 return (0);
 3209 
 3210         /*
 3211          * All the other IPs are sorted so we can do a binary search.
 3212          */
 3213         a = 0;
 3214         z = pr->pr_ip6s - 2;
 3215         while (a <= z) {
 3216                 i = (a + z) / 2;
 3217                 d = qcmp_v6(&pr->pr_ip6[i+1], ia6);
 3218                 if (d > 0)
 3219                         z = i - 1;
 3220                 else if (d < 0)
 3221                         a = i + 1;
 3222                 else
 3223                         return (0);
 3224         }
 3225 
 3226         return (EADDRNOTAVAIL);
 3227 }
 3228 
 3229 int
 3230 prison_check_ip6(struct ucred *cred, struct in6_addr *ia6)
 3231 {
 3232         struct prison *pr;
 3233         int error;
 3234 
 3235         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3236         KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__));
 3237 
 3238         pr = cred->cr_prison;
 3239         if (!(pr->pr_flags & PR_IP6))
 3240                 return (0);
 3241         mtx_lock(&pr->pr_mtx);
 3242         if (!(pr->pr_flags & PR_IP6)) {
 3243                 mtx_unlock(&pr->pr_mtx);
 3244                 return (0);
 3245         }
 3246         if (pr->pr_ip6 == NULL) {
 3247                 mtx_unlock(&pr->pr_mtx);
 3248                 return (EAFNOSUPPORT);
 3249         }
 3250 
 3251         error = _prison_check_ip6(pr, ia6);
 3252         mtx_unlock(&pr->pr_mtx);
 3253         return (error);
 3254 }
 3255 #endif
 3256 
 3257 /*
 3258  * Check if a jail supports the given address family.
 3259  *
 3260  * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT
 3261  * if not.
 3262  */
 3263 int
 3264 prison_check_af(struct ucred *cred, int af)
 3265 {
 3266         struct prison *pr;
 3267         int error;
 3268 
 3269         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3270 
 3271         pr = cred->cr_prison;
 3272 #ifdef VIMAGE
 3273         /* Prisons with their own network stack are not limited. */
 3274         if (prison_owns_vnet(cred))
 3275                 return (0);
 3276 #endif
 3277 
 3278         error = 0;
 3279         switch (af)
 3280         {
 3281 #ifdef INET
 3282         case AF_INET:
 3283                 if (pr->pr_flags & PR_IP4)
 3284                 {
 3285                         mtx_lock(&pr->pr_mtx);
 3286                         if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL)
 3287                                 error = EAFNOSUPPORT;
 3288                         mtx_unlock(&pr->pr_mtx);
 3289                 }
 3290                 break;
 3291 #endif
 3292 #ifdef INET6
 3293         case AF_INET6:
 3294                 if (pr->pr_flags & PR_IP6)
 3295                 {
 3296                         mtx_lock(&pr->pr_mtx);
 3297                         if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL)
 3298                                 error = EAFNOSUPPORT;
 3299                         mtx_unlock(&pr->pr_mtx);
 3300                 }
 3301                 break;
 3302 #endif
 3303         case AF_LOCAL:
 3304         case AF_ROUTE:
 3305                 break;
 3306         default:
 3307                 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF))
 3308                         error = EAFNOSUPPORT;
 3309         }
 3310         return (error);
 3311 }
 3312 
 3313 /*
 3314  * Check if given address belongs to the jail referenced by cred (wrapper to
 3315  * prison_check_ip[46]).
 3316  *
 3317  * Returns 0 if jail doesn't restrict the address family or if address belongs
 3318  * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if
 3319  * the jail doesn't allow the address family.  IPv4 Address passed in in NBO.
 3320  */
 3321 int
 3322 prison_if(struct ucred *cred, struct sockaddr *sa)
 3323 {
 3324 #ifdef INET
 3325         struct sockaddr_in *sai;
 3326 #endif
 3327 #ifdef INET6
 3328         struct sockaddr_in6 *sai6;
 3329 #endif
 3330         int error;
 3331 
 3332         KASSERT(cred != NULL, ("%s: cred is NULL", __func__));
 3333         KASSERT(sa != NULL, ("%s: sa is NULL", __func__));
 3334 
 3335 #ifdef VIMAGE
 3336         if (prison_owns_vnet(cred))
 3337                 return (0);
 3338 #endif
 3339 
 3340         error = 0;
 3341         switch (sa->sa_family)
 3342         {
 3343 #ifdef INET
 3344         case AF_INET:
 3345                 sai = (struct sockaddr_in *)sa;
 3346                 error = prison_check_ip4(cred, &sai->sin_addr);
 3347                 break;
 3348 #endif
 3349 #ifdef INET6
 3350         case AF_INET6:
 3351                 sai6 = (struct sockaddr_in6 *)sa;
 3352                 error = prison_check_ip6(cred, &sai6->sin6_addr);
 3353                 break;
 3354 #endif
 3355         default:
 3356                 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF))
 3357                         error = EAFNOSUPPORT;
 3358         }
 3359         return (error);
 3360 }
 3361 
 3362 /*
 3363  * Return 0 if jails permit p1 to frob p2, otherwise ESRCH.
 3364  */
 3365 int
 3366 prison_check(struct ucred *cred1, struct ucred *cred2)
 3367 {
 3368 
 3369         return ((cred1->cr_prison == cred2->cr_prison ||
 3370             prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH);
 3371 }
 3372 
 3373 /*
 3374  * Return 1 if p2 is a child of p1, otherwise 0.
 3375  */
 3376 int
 3377 prison_ischild(struct prison *pr1, struct prison *pr2)
 3378 {
 3379 
 3380         for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent)
 3381                 if (pr1 == pr2)
 3382                         return (1);
 3383         return (0);
 3384 }
 3385 
 3386 /*
 3387  * Return 1 if the passed credential is in a jail, otherwise 0.
 3388  */
 3389 int
 3390 jailed(struct ucred *cred)
 3391 {
 3392 
 3393         return (cred->cr_prison != &prison0);
 3394 }
 3395 
 3396 /*
 3397  * Return 1 if the passed credential is in a jail and that jail does not
 3398  * have its own virtual network stack, otherwise 0.
 3399  */
 3400 int
 3401 jailed_without_vnet(struct ucred *cred)
 3402 {
 3403 
 3404         if (!jailed(cred))
 3405                 return (0);
 3406 #ifdef VIMAGE
 3407         if (prison_owns_vnet(cred))
 3408                 return (0);
 3409 #endif
 3410 
 3411         return (1);
 3412 }
 3413 
 3414 /*
 3415  * Return the correct hostname (domainname, et al) for the passed credential.
 3416  */
 3417 void
 3418 getcredhostname(struct ucred *cred, char *buf, size_t size)
 3419 {
 3420         struct prison *pr;
 3421 
 3422         /*
 3423          * A NULL credential can be used to shortcut to the physical
 3424          * system's hostname.
 3425          */
 3426         pr = (cred != NULL) ? cred->cr_prison : &prison0;
 3427         mtx_lock(&pr->pr_mtx);
 3428         strlcpy(buf, pr->pr_hostname, size);
 3429         mtx_unlock(&pr->pr_mtx);
 3430 }
 3431 
 3432 void
 3433 getcreddomainname(struct ucred *cred, char *buf, size_t size)
 3434 {
 3435 
 3436         mtx_lock(&cred->cr_prison->pr_mtx);
 3437         strlcpy(buf, cred->cr_prison->pr_domainname, size);
 3438         mtx_unlock(&cred->cr_prison->pr_mtx);
 3439 }
 3440 
 3441 void
 3442 getcredhostuuid(struct ucred *cred, char *buf, size_t size)
 3443 {
 3444 
 3445         mtx_lock(&cred->cr_prison->pr_mtx);
 3446         strlcpy(buf, cred->cr_prison->pr_hostuuid, size);
 3447         mtx_unlock(&cred->cr_prison->pr_mtx);
 3448 }
 3449 
 3450 void
 3451 getcredhostid(struct ucred *cred, unsigned long *hostid)
 3452 {
 3453 
 3454         mtx_lock(&cred->cr_prison->pr_mtx);
 3455         *hostid = cred->cr_prison->pr_hostid;
 3456         mtx_unlock(&cred->cr_prison->pr_mtx);
 3457 }
 3458 
 3459 #ifdef VIMAGE
 3460 /*
 3461  * Determine whether the prison represented by cred owns
 3462  * its vnet rather than having it inherited.
 3463  *
 3464  * Returns 1 in case the prison owns the vnet, 0 otherwise.
 3465  */
 3466 int
 3467 prison_owns_vnet(struct ucred *cred)
 3468 {
 3469 
 3470         /*
 3471          * vnets cannot be added/removed after jail creation,
 3472          * so no need to lock here.
 3473          */
 3474         return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0);
 3475 }
 3476 #endif
 3477 
 3478 /*
 3479  * Determine whether the subject represented by cred can "see"
 3480  * status of a mount point.
 3481  * Returns: 0 for permitted, ENOENT otherwise.
 3482  * XXX: This function should be called cr_canseemount() and should be
 3483  *      placed in kern_prot.c.
 3484  */
 3485 int
 3486 prison_canseemount(struct ucred *cred, struct mount *mp)
 3487 {
 3488         struct prison *pr;
 3489         struct statfs *sp;
 3490         size_t len;
 3491 
 3492         pr = cred->cr_prison;
 3493         if (pr->pr_enforce_statfs == 0)
 3494                 return (0);
 3495         if (pr->pr_root->v_mount == mp)
 3496                 return (0);
 3497         if (pr->pr_enforce_statfs == 2)
 3498                 return (ENOENT);
 3499         /*
 3500          * If jail's chroot directory is set to "/" we should be able to see
 3501          * all mount-points from inside a jail.
 3502          * This is ugly check, but this is the only situation when jail's
 3503          * directory ends with '/'.
 3504          */
 3505         if (strcmp(pr->pr_path, "/") == 0)
 3506                 return (0);
 3507         len = strlen(pr->pr_path);
 3508         sp = &mp->mnt_stat;
 3509         if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0)
 3510                 return (ENOENT);
 3511         /*
 3512          * Be sure that we don't have situation where jail's root directory
 3513          * is "/some/path" and mount point is "/some/pathpath".
 3514          */
 3515         if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/')
 3516                 return (ENOENT);
 3517         return (0);
 3518 }
 3519 
 3520 void
 3521 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp)
 3522 {
 3523         char jpath[MAXPATHLEN];
 3524         struct prison *pr;
 3525         size_t len;
 3526 
 3527         pr = cred->cr_prison;
 3528         if (pr->pr_enforce_statfs == 0)
 3529                 return;
 3530         if (prison_canseemount(cred, mp) != 0) {
 3531                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
 3532                 strlcpy(sp->f_mntonname, "[restricted]",
 3533                     sizeof(sp->f_mntonname));
 3534                 return;
 3535         }
 3536         if (pr->pr_root->v_mount == mp) {
 3537                 /*
 3538                  * Clear current buffer data, so we are sure nothing from
 3539                  * the valid path left there.
 3540                  */
 3541                 bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
 3542                 *sp->f_mntonname = '/';
 3543                 return;
 3544         }
 3545         /*
 3546          * If jail's chroot directory is set to "/" we should be able to see
 3547          * all mount-points from inside a jail.
 3548          */
 3549         if (strcmp(pr->pr_path, "/") == 0)
 3550                 return;
 3551         len = strlen(pr->pr_path);
 3552         strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath));
 3553         /*
 3554          * Clear current buffer data, so we are sure nothing from
 3555          * the valid path left there.
 3556          */
 3557         bzero(sp->f_mntonname, sizeof(sp->f_mntonname));
 3558         if (*jpath == '\0') {
 3559                 /* Should never happen. */
 3560                 *sp->f_mntonname = '/';
 3561         } else {
 3562                 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname));
 3563         }
 3564 }
 3565 
 3566 /*
 3567  * Check with permission for a specific privilege is granted within jail.  We
 3568  * have a specific list of accepted privileges; the rest are denied.
 3569  */
 3570 int
 3571 prison_priv_check(struct ucred *cred, int priv)
 3572 {
 3573 
 3574         if (!jailed(cred))
 3575                 return (0);
 3576 
 3577 #ifdef VIMAGE
 3578         /*
 3579          * Privileges specific to prisons with a virtual network stack.
 3580          * There might be a duplicate entry here in case the privilege
 3581          * is only granted conditionally in the legacy jail case.
 3582          */
 3583         switch (priv) {
 3584 #ifdef notyet
 3585                 /*
 3586                  * NFS-specific privileges.
 3587                  */
 3588         case PRIV_NFS_DAEMON:
 3589         case PRIV_NFS_LOCKD:
 3590 #endif
 3591                 /*
 3592                  * Network stack privileges.
 3593                  */
 3594         case PRIV_NET_BRIDGE:
 3595         case PRIV_NET_GRE:
 3596         case PRIV_NET_BPF:
 3597         case PRIV_NET_RAW:              /* Dup, cond. in legacy jail case. */
 3598         case PRIV_NET_ROUTE:
 3599         case PRIV_NET_TAP:
 3600         case PRIV_NET_SETIFMTU:
 3601         case PRIV_NET_SETIFFLAGS:
 3602         case PRIV_NET_SETIFCAP:
 3603         case PRIV_NET_SETIFDESCR:
 3604         case PRIV_NET_SETIFNAME :
 3605         case PRIV_NET_SETIFMETRIC:
 3606         case PRIV_NET_SETIFPHYS:
 3607         case PRIV_NET_SETIFMAC:
 3608         case PRIV_NET_ADDMULTI:
 3609         case PRIV_NET_DELMULTI:
 3610         case PRIV_NET_HWIOCTL:
 3611         case PRIV_NET_SETLLADDR:
 3612         case PRIV_NET_ADDIFGROUP:
 3613         case PRIV_NET_DELIFGROUP:
 3614         case PRIV_NET_IFCREATE:
 3615         case PRIV_NET_IFDESTROY:
 3616         case PRIV_NET_ADDIFADDR:
 3617         case PRIV_NET_DELIFADDR:
 3618         case PRIV_NET_LAGG:
 3619         case PRIV_NET_GIF:
 3620         case PRIV_NET_SETIFVNET:
 3621         case PRIV_NET_SETIFFIB:
 3622 
 3623                 /*
 3624                  * 802.11-related privileges.
 3625                  */
 3626         case PRIV_NET80211_GETKEY:
 3627 #ifdef notyet
 3628         case PRIV_NET80211_MANAGE:              /* XXX-BZ discuss with sam@ */
 3629 #endif
 3630 
 3631 #ifdef notyet
 3632                 /*
 3633                  * AppleTalk privileges.
 3634                  */
 3635         case PRIV_NETATALK_RESERVEDPORT:
 3636 
 3637                 /*
 3638                  * ATM privileges.
 3639                  */
 3640         case PRIV_NETATM_CFG:
 3641         case PRIV_NETATM_ADD:
 3642         case PRIV_NETATM_DEL:
 3643         case PRIV_NETATM_SET:
 3644 
 3645                 /*
 3646                  * Bluetooth privileges.
 3647                  */
 3648         case PRIV_NETBLUETOOTH_RAW:
 3649 #endif
 3650 
 3651                 /*
 3652                  * Netgraph and netgraph module privileges.
 3653                  */
 3654         case PRIV_NETGRAPH_CONTROL:
 3655 #ifdef notyet
 3656         case PRIV_NETGRAPH_TTY:
 3657 #endif
 3658 
 3659                 /*
 3660                  * IPv4 and IPv6 privileges.
 3661                  */
 3662         case PRIV_NETINET_IPFW:
 3663         case PRIV_NETINET_DIVERT:
 3664         case PRIV_NETINET_PF:
 3665         case PRIV_NETINET_DUMMYNET:
 3666         case PRIV_NETINET_CARP:
 3667         case PRIV_NETINET_MROUTE:
 3668         case PRIV_NETINET_RAW:
 3669         case PRIV_NETINET_ADDRCTRL6:
 3670         case PRIV_NETINET_ND6:
 3671         case PRIV_NETINET_SCOPE6:
 3672         case PRIV_NETINET_ALIFETIME6:
 3673         case PRIV_NETINET_IPSEC:
 3674         case PRIV_NETINET_BINDANY:
 3675 
 3676 #ifdef notyet
 3677                 /*
 3678                  * IPX/SPX privileges.
 3679                  */
 3680         case PRIV_NETIPX_RESERVEDPORT:
 3681         case PRIV_NETIPX_RAW:
 3682 
 3683                 /*
 3684                  * NCP privileges.
 3685                  */
 3686         case PRIV_NETNCP:
 3687 
 3688                 /*
 3689                  * SMB privileges.
 3690                  */
 3691         case PRIV_NETSMB:
 3692 #endif
 3693 
 3694         /*
 3695          * No default: or deny here.
 3696          * In case of no permit fall through to next switch().
 3697          */
 3698                 if (cred->cr_prison->pr_flags & PR_VNET)
 3699                         return (0);
 3700         }
 3701 #endif /* VIMAGE */
 3702 
 3703         switch (priv) {
 3704 
 3705                 /*
 3706                  * Allow ktrace privileges for root in jail.
 3707                  */
 3708         case PRIV_KTRACE:
 3709 
 3710 #if 0
 3711                 /*
 3712                  * Allow jailed processes to configure audit identity and
 3713                  * submit audit records (login, etc).  In the future we may
 3714                  * want to further refine the relationship between audit and
 3715                  * jail.
 3716                  */
 3717         case PRIV_AUDIT_GETAUDIT:
 3718         case PRIV_AUDIT_SETAUDIT:
 3719         case PRIV_AUDIT_SUBMIT:
 3720 #endif
 3721 
 3722                 /*
 3723                  * Allow jailed processes to manipulate process UNIX
 3724                  * credentials in any way they see fit.
 3725                  */
 3726         case PRIV_CRED_SETUID:
 3727         case PRIV_CRED_SETEUID:
 3728         case PRIV_CRED_SETGID:
 3729         case PRIV_CRED_SETEGID:
 3730         case PRIV_CRED_SETGROUPS:
 3731         case PRIV_CRED_SETREUID:
 3732         case PRIV_CRED_SETREGID:
 3733         case PRIV_CRED_SETRESUID:
 3734         case PRIV_CRED_SETRESGID:
 3735 
 3736                 /*
 3737                  * Jail implements visibility constraints already, so allow
 3738                  * jailed root to override uid/gid-based constraints.
 3739                  */
 3740         case PRIV_SEEOTHERGIDS:
 3741         case PRIV_SEEOTHERUIDS:
 3742 
 3743                 /*
 3744                  * Jail implements inter-process debugging limits already, so
 3745                  * allow jailed root various debugging privileges.
 3746                  */
 3747         case PRIV_DEBUG_DIFFCRED:
 3748         case PRIV_DEBUG_SUGID:
 3749         case PRIV_DEBUG_UNPRIV:
 3750 
 3751                 /*
 3752                  * Allow jail to set various resource limits and login
 3753                  * properties, and for now, exceed process resource limits.
 3754                  */
 3755         case PRIV_PROC_LIMIT:
 3756         case PRIV_PROC_SETLOGIN:
 3757         case PRIV_PROC_SETRLIMIT:
 3758 
 3759                 /*
 3760                  * System V and POSIX IPC privileges are granted in jail.
 3761                  */
 3762         case PRIV_IPC_READ:
 3763         case PRIV_IPC_WRITE:
 3764         case PRIV_IPC_ADMIN:
 3765         case PRIV_IPC_MSGSIZE:
 3766         case PRIV_MQ_ADMIN:
 3767 
 3768                 /*
 3769                  * Jail operations within a jail work on child jails.
 3770                  */
 3771         case PRIV_JAIL_ATTACH:
 3772         case PRIV_JAIL_SET:
 3773         case PRIV_JAIL_REMOVE:
 3774 
 3775                 /*
 3776                  * Jail implements its own inter-process limits, so allow
 3777                  * root processes in jail to change scheduling on other
 3778                  * processes in the same jail.  Likewise for signalling.
 3779                  */
 3780         case PRIV_SCHED_DIFFCRED:
 3781         case PRIV_SCHED_CPUSET:
 3782         case PRIV_SIGNAL_DIFFCRED:
 3783         case PRIV_SIGNAL_SUGID:
 3784 
 3785                 /*
 3786                  * Allow jailed processes to write to sysctls marked as jail
 3787                  * writable.
 3788                  */
 3789         case PRIV_SYSCTL_WRITEJAIL:
 3790 
 3791                 /*
 3792                  * Allow root in jail to manage a variety of quota
 3793                  * properties.  These should likely be conditional on a
 3794                  * configuration option.
 3795                  */
 3796         case PRIV_VFS_GETQUOTA:
 3797         case PRIV_VFS_SETQUOTA:
 3798 
 3799                 /*
 3800                  * Since Jail relies on chroot() to implement file system
 3801                  * protections, grant many VFS privileges to root in jail.
 3802                  * Be careful to exclude mount-related and NFS-related
 3803                  * privileges.
 3804                  */
 3805         case PRIV_VFS_READ:
 3806         case PRIV_VFS_WRITE:
 3807         case PRIV_VFS_ADMIN:
 3808         case PRIV_VFS_EXEC:
 3809         case PRIV_VFS_LOOKUP:
 3810         case PRIV_VFS_BLOCKRESERVE:     /* XXXRW: Slightly surprising. */
 3811         case PRIV_VFS_CHFLAGS_DEV:
 3812         case PRIV_VFS_CHOWN:
 3813         case PRIV_VFS_CHROOT:
 3814         case PRIV_VFS_RETAINSUGID:
 3815         case PRIV_VFS_FCHROOT:
 3816         case PRIV_VFS_LINK:
 3817         case PRIV_VFS_SETGID:
 3818         case PRIV_VFS_STAT:
 3819         case PRIV_VFS_STICKYFILE:
 3820                 return (0);
 3821 
 3822                 /*
 3823                  * Depending on the global setting, allow privilege of
 3824                  * setting system flags.
 3825                  */
 3826         case PRIV_VFS_SYSFLAGS:
 3827                 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS)
 3828                         return (0);
 3829                 else
 3830                         return (EPERM);
 3831 
 3832                 /*
 3833                  * Depending on the global setting, allow privilege of
 3834                  * mounting/unmounting file systems.
 3835                  */
 3836         case PRIV_VFS_MOUNT:
 3837         case PRIV_VFS_UNMOUNT:
 3838         case PRIV_VFS_MOUNT_NONUSER:
 3839         case PRIV_VFS_MOUNT_OWNER:
 3840                 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT &&
 3841                     cred->cr_prison->pr_enforce_statfs < 2)
 3842                         return (0);
 3843                 else
 3844                         return (EPERM);
 3845 
 3846                 /*
 3847                  * Allow jailed root to bind reserved ports and reuse in-use
 3848                  * ports.
 3849                  */
 3850         case PRIV_NETINET_RESERVEDPORT:
 3851         case PRIV_NETINET_REUSEPORT:
 3852                 return (0);
 3853 
 3854                 /*
 3855                  * Allow jailed root to set certian IPv4/6 (option) headers.
 3856                  */
 3857         case PRIV_NETINET_SETHDROPTS:
 3858                 return (0);
 3859 
 3860                 /*
 3861                  * Conditionally allow creating raw sockets in jail.
 3862                  */
 3863         case PRIV_NETINET_RAW:
 3864                 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS)
 3865                         return (0);
 3866                 else
 3867                         return (EPERM);
 3868 
 3869                 /*
 3870                  * Since jail implements its own visibility limits on netstat
 3871                  * sysctls, allow getcred.  This allows identd to work in
 3872                  * jail.
 3873                  */
 3874         case PRIV_NETINET_GETCRED:
 3875                 return (0);
 3876 
 3877                 /*
 3878                  * Allow jailed root to set loginclass.
 3879                  */
 3880         case PRIV_PROC_SETLOGINCLASS:
 3881                 return (0);
 3882 
 3883         default:
 3884                 /*
 3885                  * In all remaining cases, deny the privilege request.  This
 3886                  * includes almost all network privileges, many system
 3887                  * configuration privileges.
 3888                  */
 3889                 return (EPERM);
 3890         }
 3891 }
 3892 
 3893 /*
 3894  * Return the part of pr2's name that is relative to pr1, or the whole name
 3895  * if it does not directly follow.
 3896  */
 3897 
 3898 char *
 3899 prison_name(struct prison *pr1, struct prison *pr2)
 3900 {
 3901         char *name;
 3902 
 3903         /* Jails see themselves as "" (if they see themselves at all). */
 3904         if (pr1 == pr2)
 3905                 return "";
 3906         name = pr2->pr_name;
 3907         if (prison_ischild(pr1, pr2)) {
 3908                 /*
 3909                  * pr1 isn't locked (and allprison_lock may not be either)
 3910                  * so its length can't be counted on.  But the number of dots
 3911                  * can be counted on - and counted.
 3912                  */
 3913                 for (; pr1 != &prison0; pr1 = pr1->pr_parent)
 3914                         name = strchr(name, '.') + 1;
 3915         }
 3916         return (name);
 3917 }
 3918 
 3919 /*
 3920  * Return the part of pr2's path that is relative to pr1, or the whole path
 3921  * if it does not directly follow.
 3922  */
 3923 static char *
 3924 prison_path(struct prison *pr1, struct prison *pr2)
 3925 {
 3926         char *path1, *path2;
 3927         int len1;
 3928 
 3929         path1 = pr1->pr_path;
 3930         path2 = pr2->pr_path;
 3931         if (!strcmp(path1, "/"))
 3932                 return (path2);
 3933         len1 = strlen(path1);
 3934         if (strncmp(path1, path2, len1))
 3935                 return (path2);
 3936         if (path2[len1] == '\0')
 3937                 return "/";
 3938         if (path2[len1] == '/')
 3939                 return (path2 + len1);
 3940         return (path2);
 3941 }
 3942 
 3943 
 3944 /*
 3945  * Jail-related sysctls.
 3946  */
 3947 SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0,
 3948     "Jails");
 3949 
 3950 static int
 3951 sysctl_jail_list(SYSCTL_HANDLER_ARGS)
 3952 {
 3953         struct xprison *xp;
 3954         struct prison *pr, *cpr;
 3955 #ifdef INET
 3956         struct in_addr *ip4 = NULL;
 3957         int ip4s = 0;
 3958 #endif
 3959 #ifdef INET6
 3960         struct in6_addr *ip6 = NULL;
 3961         int ip6s = 0;
 3962 #endif
 3963         int descend, error;
 3964 
 3965         xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK);
 3966         pr = req->td->td_ucred->cr_prison;
 3967         error = 0;
 3968         sx_slock(&allprison_lock);
 3969         FOREACH_PRISON_DESCENDANT(pr, cpr, descend) {
 3970 #if defined(INET) || defined(INET6)
 3971  again:
 3972 #endif
 3973                 mtx_lock(&cpr->pr_mtx);
 3974 #ifdef INET
 3975                 if (cpr->pr_ip4s > 0) {
 3976                         if (ip4s < cpr->pr_ip4s) {
 3977                                 ip4s = cpr->pr_ip4s;
 3978                                 mtx_unlock(&cpr->pr_mtx);
 3979                                 ip4 = realloc(ip4, ip4s *
 3980                                     sizeof(struct in_addr), M_TEMP, M_WAITOK);
 3981                                 goto again;
 3982                         }
 3983                         bcopy(cpr->pr_ip4, ip4,
 3984                             cpr->pr_ip4s * sizeof(struct in_addr));
 3985                 }
 3986 #endif
 3987 #ifdef INET6
 3988                 if (cpr->pr_ip6s > 0) {
 3989                         if (ip6s < cpr->pr_ip6s) {
 3990                                 ip6s = cpr->pr_ip6s;
 3991                                 mtx_unlock(&cpr->pr_mtx);
 3992                                 ip6 = realloc(ip6, ip6s *
 3993                                     sizeof(struct in6_addr), M_TEMP, M_WAITOK);
 3994                                 goto again;
 3995                         }
 3996                         bcopy(cpr->pr_ip6, ip6,
 3997                             cpr->pr_ip6s * sizeof(struct in6_addr));
 3998                 }
 3999 #endif
 4000                 if (cpr->pr_ref == 0) {
 4001                         mtx_unlock(&cpr->pr_mtx);
 4002                         continue;
 4003                 }
 4004                 bzero(xp, sizeof(*xp));
 4005                 xp->pr_version = XPRISON_VERSION;
 4006                 xp->pr_id = cpr->pr_id;
 4007                 xp->pr_state = cpr->pr_uref > 0
 4008                     ? PRISON_STATE_ALIVE : PRISON_STATE_DYING;
 4009                 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path));
 4010                 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host));
 4011                 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name));
 4012 #ifdef INET
 4013                 xp->pr_ip4s = cpr->pr_ip4s;
 4014 #endif
 4015 #ifdef INET6
 4016                 xp->pr_ip6s = cpr->pr_ip6s;
 4017 #endif
 4018                 mtx_unlock(&cpr->pr_mtx);
 4019                 error = SYSCTL_OUT(req, xp, sizeof(*xp));
 4020                 if (error)
 4021                         break;
 4022 #ifdef INET
 4023                 if (xp->pr_ip4s > 0) {
 4024                         error = SYSCTL_OUT(req, ip4,
 4025                             xp->pr_ip4s * sizeof(struct in_addr));
 4026                         if (error)
 4027                                 break;
 4028                 }
 4029 #endif
 4030 #ifdef INET6
 4031                 if (xp->pr_ip6s > 0) {
 4032                         error = SYSCTL_OUT(req, ip6,
 4033                             xp->pr_ip6s * sizeof(struct in6_addr));
 4034                         if (error)
 4035                                 break;
 4036                 }
 4037 #endif
 4038         }
 4039         sx_sunlock(&allprison_lock);
 4040         free(xp, M_TEMP);
 4041 #ifdef INET
 4042         free(ip4, M_TEMP);
 4043 #endif
 4044 #ifdef INET6
 4045         free(ip6, M_TEMP);
 4046 #endif
 4047         return (error);
 4048 }
 4049 
 4050 SYSCTL_OID(_security_jail, OID_AUTO, list,
 4051     CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
 4052     sysctl_jail_list, "S", "List of active jails");
 4053 
 4054 static int
 4055 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS)
 4056 {
 4057         int error, injail;
 4058 
 4059         injail = jailed(req->td->td_ucred);
 4060         error = SYSCTL_OUT(req, &injail, sizeof(injail));
 4061 
 4062         return (error);
 4063 }
 4064 
 4065 SYSCTL_PROC(_security_jail, OID_AUTO, jailed,
 4066     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
 4067     sysctl_jail_jailed, "I", "Process in jail?");
 4068 
 4069 #if defined(INET) || defined(INET6)
 4070 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW,
 4071     &jail_max_af_ips, 0,
 4072     "Number of IP addresses a jail may have at most per address family");
 4073 #endif
 4074 
 4075 /*
 4076  * Default parameters for jail(2) compatability.  For historical reasons,
 4077  * the sysctl names have varying similarity to the parameter names.  Prisons
 4078  * just see their own parameters, and can't change them.
 4079  */
 4080 static int
 4081 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS)
 4082 {
 4083         struct prison *pr;
 4084         int allow, error, i;
 4085 
 4086         pr = req->td->td_ucred->cr_prison;
 4087         allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow;
 4088 
 4089         /* Get the current flag value, and convert it to a boolean. */
 4090         i = (allow & arg2) ? 1 : 0;
 4091         if (arg1 != NULL)
 4092                 i = !i;
 4093         error = sysctl_handle_int(oidp, &i, 0, req);
 4094         if (error || !req->newptr)
 4095                 return (error);
 4096         i = i ? arg2 : 0;
 4097         if (arg1 != NULL)
 4098                 i ^= arg2;
 4099         /*
 4100          * The sysctls don't have CTLFLAGS_PRISON, so assume prison0
 4101          * for writing.
 4102          */
 4103         mtx_lock(&prison0.pr_mtx);
 4104         jail_default_allow = (jail_default_allow & ~arg2) | i;
 4105         mtx_unlock(&prison0.pr_mtx);
 4106         return (0);
 4107 }
 4108 
 4109 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed,
 4110     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4111     NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I",
 4112     "Processes in jail can set their hostnames");
 4113 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only,
 4114     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4115     (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I",
 4116     "Processes in jail are limited to creating UNIX/IP/route sockets only");
 4117 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed,
 4118     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4119     NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I",
 4120     "Processes in jail can use System V IPC primitives");
 4121 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets,
 4122     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4123     NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I",
 4124     "Prison root can create raw sockets");
 4125 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed,
 4126     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4127     NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I",
 4128     "Processes in jail can alter system file flags");
 4129 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed,
 4130     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4131     NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I",
 4132     "Processes in jail can mount/unmount jail-friendly file systems");
 4133 
 4134 static int
 4135 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS)
 4136 {
 4137         struct prison *pr;
 4138         int level, error;
 4139 
 4140         pr = req->td->td_ucred->cr_prison;
 4141         level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2);
 4142         error = sysctl_handle_int(oidp, &level, 0, req);
 4143         if (error || !req->newptr)
 4144                 return (error);
 4145         *(int *)arg1 = level;
 4146         return (0);
 4147 }
 4148 
 4149 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs,
 4150     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
 4151     &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs),
 4152     sysctl_jail_default_level, "I",
 4153     "Processes in jail cannot see all mounted file systems");
 4154 
 4155 /*
 4156  * Nodes to describe jail parameters.  Maximum length of string parameters
 4157  * is returned in the string itself, and the other parameters exist merely
 4158  * to make themselves and their types known.
 4159  */
 4160 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0,
 4161     "Jail parameters");
 4162 
 4163 int
 4164 sysctl_jail_param(SYSCTL_HANDLER_ARGS)
 4165 {
 4166         int i;
 4167         long l;
 4168         size_t s;
 4169         char numbuf[12];
 4170 
 4171         switch (oidp->oid_kind & CTLTYPE)
 4172         {
 4173         case CTLTYPE_LONG:
 4174         case CTLTYPE_ULONG:
 4175                 l = 0;
 4176 #ifdef SCTL_MASK32
 4177                 if (!(req->flags & SCTL_MASK32))
 4178 #endif
 4179                         return (SYSCTL_OUT(req, &l, sizeof(l)));
 4180         case CTLTYPE_INT:
 4181         case CTLTYPE_UINT:
 4182                 i = 0;
 4183                 return (SYSCTL_OUT(req, &i, sizeof(i)));
 4184         case CTLTYPE_STRING:
 4185                 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2);
 4186                 return
 4187                     (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req));
 4188         case CTLTYPE_STRUCT:
 4189                 s = (size_t)arg2;
 4190                 return (SYSCTL_OUT(req, &s, sizeof(s)));
 4191         }
 4192         return (0);
 4193 }
 4194 
 4195 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID");
 4196 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID");
 4197 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name");
 4198 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path");
 4199 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW,
 4200     "I", "Jail secure level");
 4201 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW,
 4202     "I", "Jail cannot see all mounted file systems");
 4203 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW,
 4204     "B", "Jail persistence");
 4205 #ifdef VIMAGE
 4206 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN,
 4207     "E,jailsys", "Virtual network stack");
 4208 #endif
 4209 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD,
 4210     "B", "Jail is in the process of shutting down");
 4211 
 4212 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails");
 4213 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD,
 4214     "I", "Current number of child jails");
 4215 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW,
 4216     "I", "Maximum number of child jails");
 4217 
 4218 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info");
 4219 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN,
 4220     "Jail hostname");
 4221 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN,
 4222     "Jail NIS domainname");
 4223 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN,
 4224     "Jail host UUID");
 4225 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW,
 4226     "LU", "Jail host ID");
 4227 
 4228 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset");
 4229 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID");
 4230 
 4231 #ifdef INET
 4232 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN,
 4233     "Jail IPv4 address virtualization");
 4234 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr),
 4235     "S,in_addr,a", "Jail IPv4 addresses");
 4236 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
 4237     "B", "Do (not) use IPv4 source address selection rather than the "
 4238     "primary jail IPv4 address.");
 4239 #endif
 4240 #ifdef INET6
 4241 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN,
 4242     "Jail IPv6 address virtualization");
 4243 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr),
 4244     "S,in6_addr,a", "Jail IPv6 addresses");
 4245 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW,
 4246     "B", "Do (not) use IPv6 source address selection rather than the "
 4247     "primary jail IPv6 address.");
 4248 #endif
 4249 
 4250 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags");
 4251 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW,
 4252     "B", "Jail may set hostname");
 4253 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW,
 4254     "B", "Jail may use SYSV IPC");
 4255 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW,
 4256     "B", "Jail may create raw sockets");
 4257 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW,
 4258     "B", "Jail may alter system file flags");
 4259 SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW,
 4260     "B", "Jail may mount/unmount jail-friendly file systems");
 4261 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW,
 4262     "B", "Jail may set file quotas");
 4263 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW,
 4264     "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route");
 4265 
 4266 void
 4267 prison_racct_foreach(void (*callback)(struct racct *racct,
 4268     void *arg2, void *arg3), void *arg2, void *arg3)
 4269 {
 4270         struct prison_racct *prr;
 4271 
 4272         sx_slock(&allprison_lock);
 4273         LIST_FOREACH(prr, &allprison_racct, prr_next)
 4274                 (callback)(prr->prr_racct, arg2, arg3);
 4275         sx_sunlock(&allprison_lock);
 4276 }
 4277 
 4278 static struct prison_racct *
 4279 prison_racct_find_locked(const char *name)
 4280 {
 4281         struct prison_racct *prr;
 4282 
 4283         sx_assert(&allprison_lock, SA_XLOCKED);
 4284 
 4285         if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN)
 4286                 return (NULL);
 4287 
 4288         LIST_FOREACH(prr, &allprison_racct, prr_next) {
 4289                 if (strcmp(name, prr->prr_name) != 0)
 4290                         continue;
 4291 
 4292                 /* Found prison_racct with a matching name? */
 4293                 prison_racct_hold(prr);
 4294                 return (prr);
 4295         }
 4296 
 4297         /* Add new prison_racct. */
 4298         prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK);
 4299         racct_create(&prr->prr_racct);
 4300 
 4301         strcpy(prr->prr_name, name);
 4302         refcount_init(&prr->prr_refcount, 1);
 4303         LIST_INSERT_HEAD(&allprison_racct, prr, prr_next);
 4304 
 4305         return (prr);
 4306 }
 4307 
 4308 struct prison_racct *
 4309 prison_racct_find(const char *name)
 4310 {
 4311         struct prison_racct *prr;
 4312 
 4313         sx_xlock(&allprison_lock);
 4314         prr = prison_racct_find_locked(name);
 4315         sx_xunlock(&allprison_lock);
 4316         return (prr);
 4317 }
 4318 
 4319 void
 4320 prison_racct_hold(struct prison_racct *prr)
 4321 {
 4322 
 4323         refcount_acquire(&prr->prr_refcount);
 4324 }
 4325 
 4326 void
 4327 prison_racct_free(struct prison_racct *prr)
 4328 {
 4329         int old;
 4330 
 4331         old = prr->prr_refcount;
 4332         if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1))
 4333                 return;
 4334 
 4335         sx_xlock(&allprison_lock);
 4336         if (refcount_release(&prr->prr_refcount)) {
 4337                 racct_destroy(&prr->prr_racct);
 4338                 LIST_REMOVE(prr, prr_next);
 4339                 sx_xunlock(&allprison_lock);
 4340                 free(prr, M_PRISON_RACCT);
 4341 
 4342                 return;
 4343         }
 4344         sx_xunlock(&allprison_lock);
 4345 }
 4346 
 4347 #ifdef RACCT
 4348 static void
 4349 prison_racct_attach(struct prison *pr)
 4350 {
 4351         struct prison_racct *prr;
 4352 
 4353         prr = prison_racct_find_locked(pr->pr_name);
 4354         KASSERT(prr != NULL, ("cannot find prison_racct"));
 4355 
 4356         pr->pr_prison_racct = prr;
 4357 }
 4358 
 4359 static void
 4360 prison_racct_detach(struct prison *pr)
 4361 {
 4362         prison_racct_free(pr->pr_prison_racct);
 4363         pr->pr_prison_racct = NULL;
 4364 }
 4365 #endif /* RACCT */
 4366 
 4367 #ifdef DDB
 4368 
 4369 static void
 4370 db_show_prison(struct prison *pr)
 4371 {
 4372         int fi;
 4373 #if defined(INET) || defined(INET6)
 4374         int ii;
 4375 #endif
 4376         unsigned jsf;
 4377 #ifdef INET6
 4378         char ip6buf[INET6_ADDRSTRLEN];
 4379 #endif
 4380 
 4381         db_printf("prison %p:\n", pr);
 4382         db_printf(" jid             = %d\n", pr->pr_id);
 4383         db_printf(" name            = %s\n", pr->pr_name);
 4384         db_printf(" parent          = %p\n", pr->pr_parent);
 4385         db_printf(" ref             = %d\n", pr->pr_ref);
 4386         db_printf(" uref            = %d\n", pr->pr_uref);
 4387         db_printf(" path            = %s\n", pr->pr_path);
 4388         db_printf(" cpuset          = %d\n", pr->pr_cpuset
 4389             ? pr->pr_cpuset->cs_id : -1);
 4390 #ifdef VIMAGE
 4391         db_printf(" vnet            = %p\n", pr->pr_vnet);
 4392 #endif
 4393         db_printf(" root            = %p\n", pr->pr_root);
 4394         db_printf(" securelevel     = %d\n", pr->pr_securelevel);
 4395         db_printf(" children.max    = %d\n", pr->pr_childmax);
 4396         db_printf(" children.cur    = %d\n", pr->pr_childcount);
 4397         db_printf(" child           = %p\n", LIST_FIRST(&pr->pr_children));
 4398         db_printf(" sibling         = %p\n", LIST_NEXT(pr, pr_sibling));
 4399         db_printf(" flags           = 0x%x", pr->pr_flags);
 4400         for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]);
 4401             fi++)
 4402                 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi)))
 4403                         db_printf(" %s", pr_flag_names[fi]);
 4404         for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]);
 4405             fi++) {
 4406                 jsf = pr->pr_flags &
 4407                     (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new);
 4408                 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name,
 4409                     pr_flag_jailsys[fi].disable && 
 4410                       (jsf == pr_flag_jailsys[fi].disable) ? "disable"
 4411                     : (jsf == pr_flag_jailsys[fi].new) ? "new"
 4412                     : "inherit");
 4413         }
 4414         db_printf(" allow           = 0x%x", pr->pr_allow);
 4415         for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]);
 4416             fi++)
 4417                 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi)))
 4418                         db_printf(" %s", pr_allow_names[fi]);
 4419         db_printf("\n");
 4420         db_printf(" enforce_statfs  = %d\n", pr->pr_enforce_statfs);
 4421         db_printf(" host.hostname   = %s\n", pr->pr_hostname);
 4422         db_printf(" host.domainname = %s\n", pr->pr_domainname);
 4423         db_printf(" host.hostuuid   = %s\n", pr->pr_hostuuid);
 4424         db_printf(" host.hostid     = %lu\n", pr->pr_hostid);
 4425 #ifdef INET
 4426         db_printf(" ip4s            = %d\n", pr->pr_ip4s);
 4427         for (ii = 0; ii < pr->pr_ip4s; ii++)
 4428                 db_printf(" %s %s\n",
 4429                     ii == 0 ? "ip4.addr        =" : "                 ",
 4430                     inet_ntoa(pr->pr_ip4[ii]));
 4431 #endif
 4432 #ifdef INET6
 4433         db_printf(" ip6s            = %d\n", pr->pr_ip6s);
 4434         for (ii = 0; ii < pr->pr_ip6s; ii++)
 4435                 db_printf(" %s %s\n",
 4436                     ii == 0 ? "ip6.addr        =" : "                 ",
 4437                     ip6_sprintf(ip6buf, &pr->pr_ip6[ii]));
 4438 #endif
 4439 }
 4440 
 4441 DB_SHOW_COMMAND(prison, db_show_prison_command)
 4442 {
 4443         struct prison *pr;
 4444 
 4445         if (!have_addr) {
 4446                 /*
 4447                  * Show all prisons in the list, and prison0 which is not
 4448                  * listed.
 4449                  */
 4450                 db_show_prison(&prison0);
 4451                 if (!db_pager_quit) {
 4452                         TAILQ_FOREACH(pr, &allprison, pr_list) {
 4453                                 db_show_prison(pr);
 4454                                 if (db_pager_quit)
 4455                                         break;
 4456                         }
 4457                 }
 4458                 return;
 4459         }
 4460 
 4461         if (addr == 0)
 4462                 pr = &prison0;
 4463         else {
 4464                 /* Look for a prison with the ID and with references. */
 4465                 TAILQ_FOREACH(pr, &allprison, pr_list)
 4466                         if (pr->pr_id == addr && pr->pr_ref > 0)
 4467                                 break;
 4468                 if (pr == NULL)
 4469                         /* Look again, without requiring a reference. */
 4470                         TAILQ_FOREACH(pr, &allprison, pr_list)
 4471                                 if (pr->pr_id == addr)
 4472                                         break;
 4473                 if (pr == NULL)
 4474                         /* Assume address points to a valid prison. */
 4475                         pr = (struct prison *)addr;
 4476         }
 4477         db_show_prison(pr);
 4478 }
 4479 
 4480 #endif /* DDB */

Cache object: 63a119f63d63b1d2077b890d2b67850c


[ 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.