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/net/if.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) 1980, 1986, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 4. Neither the name of the University nor the names of its contributors
   14  *    may be used to endorse or promote products derived from this software
   15  *    without specific prior written permission.
   16  *
   17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  *
   29  *      @(#)if.c        8.5 (Berkeley) 1/9/95
   30  * $FreeBSD: releng/10.1/sys/net/if.c 269046 2014-07-24 06:02:03Z kevlo $
   31  */
   32 
   33 #include "opt_compat.h"
   34 #include "opt_inet6.h"
   35 #include "opt_inet.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/types.h>
   39 #include <sys/conf.h>
   40 #include <sys/malloc.h>
   41 #include <sys/sbuf.h>
   42 #include <sys/bus.h>
   43 #include <sys/mbuf.h>
   44 #include <sys/systm.h>
   45 #include <sys/priv.h>
   46 #include <sys/proc.h>
   47 #include <sys/socket.h>
   48 #include <sys/socketvar.h>
   49 #include <sys/protosw.h>
   50 #include <sys/kernel.h>
   51 #include <sys/lock.h>
   52 #include <sys/refcount.h>
   53 #include <sys/module.h>
   54 #include <sys/rwlock.h>
   55 #include <sys/sockio.h>
   56 #include <sys/syslog.h>
   57 #include <sys/sysctl.h>
   58 #include <sys/taskqueue.h>
   59 #include <sys/domain.h>
   60 #include <sys/jail.h>
   61 #include <sys/priv.h>
   62 
   63 #include <machine/stdarg.h>
   64 #include <vm/uma.h>
   65 
   66 #include <net/if.h>
   67 #include <net/if_arp.h>
   68 #include <net/if_clone.h>
   69 #include <net/if_dl.h>
   70 #include <net/if_types.h>
   71 #include <net/if_var.h>
   72 #include <net/radix.h>
   73 #include <net/route.h>
   74 #include <net/vnet.h>
   75 
   76 #if defined(INET) || defined(INET6)
   77 #include <net/ethernet.h>
   78 #include <netinet/in.h>
   79 #include <netinet/in_var.h>
   80 #include <netinet/ip.h>
   81 #include <netinet/ip_carp.h>
   82 #ifdef INET
   83 #include <netinet/if_ether.h>
   84 #endif /* INET */
   85 #ifdef INET6
   86 #include <netinet6/in6_var.h>
   87 #include <netinet6/in6_ifattach.h>
   88 #endif /* INET6 */
   89 #endif /* INET || INET6 */
   90 
   91 #include <security/mac/mac_framework.h>
   92 
   93 #ifdef COMPAT_FREEBSD32
   94 #include <sys/mount.h>
   95 #include <compat/freebsd32/freebsd32.h>
   96 #endif
   97 
   98 struct ifindex_entry {
   99         struct  ifnet *ife_ifnet;
  100 };
  101 
  102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
  103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
  104 
  105 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
  106 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
  107     &ifqmaxlen, 0, "max send queue size");
  108 
  109 /* Log link state change events */
  110 static int log_link_state_change = 1;
  111 
  112 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
  113         &log_link_state_change, 0,
  114         "log interface link state change events");
  115 
  116 /* Interface description */
  117 static unsigned int ifdescr_maxlen = 1024;
  118 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
  119         &ifdescr_maxlen, 0,
  120         "administrative maximum length for interface description");
  121 
  122 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
  123 
  124 /* global sx for non-critical path ifdescr */
  125 static struct sx ifdescr_sx;
  126 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
  127 
  128 void    (*bridge_linkstate_p)(struct ifnet *ifp);
  129 void    (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
  130 void    (*lagg_linkstate_p)(struct ifnet *ifp, int state);
  131 /* These are external hooks for CARP. */
  132 void    (*carp_linkstate_p)(struct ifnet *ifp);
  133 void    (*carp_demote_adj_p)(int, char *);
  134 int     (*carp_master_p)(struct ifaddr *);
  135 #if defined(INET) || defined(INET6)
  136 int     (*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
  137 int     (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
  138     const struct sockaddr *sa);
  139 int     (*carp_ioctl_p)(struct ifreq *, u_long, struct thread *);   
  140 int     (*carp_attach_p)(struct ifaddr *, int);
  141 void    (*carp_detach_p)(struct ifaddr *);
  142 #endif
  143 #ifdef INET
  144 int     (*carp_iamatch_p)(struct ifaddr *, uint8_t **);
  145 #endif
  146 #ifdef INET6
  147 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
  148 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
  149     const struct in6_addr *taddr);
  150 #endif
  151 
  152 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
  153 
  154 /*
  155  * XXX: Style; these should be sorted alphabetically, and unprototyped
  156  * static functions should be prototyped. Currently they are sorted by
  157  * declaration order.
  158  */
  159 static void     if_attachdomain(void *);
  160 static void     if_attachdomain1(struct ifnet *);
  161 static int      ifconf(u_long, caddr_t);
  162 static void     if_freemulti(struct ifmultiaddr *);
  163 static void     if_init(void *);
  164 static void     if_grow(void);
  165 static void     if_route(struct ifnet *, int flag, int fam);
  166 static int      if_setflag(struct ifnet *, int, int, int *, int);
  167 static int      if_transmit(struct ifnet *ifp, struct mbuf *m);
  168 static void     if_unroute(struct ifnet *, int flag, int fam);
  169 static void     link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  170 static int      if_rtdel(struct radix_node *, void *);
  171 static int      ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
  172 static int      if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
  173 static void     do_link_state_change(void *, int);
  174 static int      if_getgroup(struct ifgroupreq *, struct ifnet *);
  175 static int      if_getgroupmembers(struct ifgroupreq *);
  176 static void     if_delgroups(struct ifnet *);
  177 static void     if_attach_internal(struct ifnet *, int);
  178 static void     if_detach_internal(struct ifnet *, int);
  179 
  180 #ifdef INET6
  181 /*
  182  * XXX: declare here to avoid to include many inet6 related files..
  183  * should be more generalized?
  184  */
  185 extern void     nd6_setmtu(struct ifnet *);
  186 #endif
  187 
  188 VNET_DEFINE(int, if_index);
  189 int     ifqmaxlen = IFQ_MAXLEN;
  190 VNET_DEFINE(struct ifnethead, ifnet);   /* depend on static init XXX */
  191 VNET_DEFINE(struct ifgrouphead, ifg_head);
  192 
  193 static VNET_DEFINE(int, if_indexlim) = 8;
  194 
  195 /* Table of ifnet by index. */
  196 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
  197 
  198 #define V_if_indexlim           VNET(if_indexlim)
  199 #define V_ifindex_table         VNET(ifindex_table)
  200 
  201 /*
  202  * The global network interface list (V_ifnet) and related state (such as
  203  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
  204  * an rwlock.  Either may be acquired shared to stablize the list, but both
  205  * must be acquired writable to modify the list.  This model allows us to
  206  * both stablize the interface list during interrupt thread processing, but
  207  * also to stablize it over long-running ioctls, without introducing priority
  208  * inversions and deadlocks.
  209  */
  210 struct rwlock ifnet_rwlock;
  211 struct sx ifnet_sxlock;
  212 
  213 /*
  214  * The allocation of network interfaces is a rather non-atomic affair; we
  215  * need to select an index before we are ready to expose the interface for
  216  * use, so will use this pointer value to indicate reservation.
  217  */
  218 #define IFNET_HOLD      (void *)(uintptr_t)(-1)
  219 
  220 static  if_com_alloc_t *if_com_alloc[256];
  221 static  if_com_free_t *if_com_free[256];
  222 
  223 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
  224 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
  225 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
  226 
  227 struct ifnet *
  228 ifnet_byindex_locked(u_short idx)
  229 {
  230 
  231         if (idx > V_if_index)
  232                 return (NULL);
  233         if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
  234                 return (NULL);
  235         return (V_ifindex_table[idx].ife_ifnet);
  236 }
  237 
  238 struct ifnet *
  239 ifnet_byindex(u_short idx)
  240 {
  241         struct ifnet *ifp;
  242 
  243         IFNET_RLOCK_NOSLEEP();
  244         ifp = ifnet_byindex_locked(idx);
  245         IFNET_RUNLOCK_NOSLEEP();
  246         return (ifp);
  247 }
  248 
  249 struct ifnet *
  250 ifnet_byindex_ref(u_short idx)
  251 {
  252         struct ifnet *ifp;
  253 
  254         IFNET_RLOCK_NOSLEEP();
  255         ifp = ifnet_byindex_locked(idx);
  256         if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
  257                 IFNET_RUNLOCK_NOSLEEP();
  258                 return (NULL);
  259         }
  260         if_ref(ifp);
  261         IFNET_RUNLOCK_NOSLEEP();
  262         return (ifp);
  263 }
  264 
  265 /*
  266  * Allocate an ifindex array entry; return 0 on success or an error on
  267  * failure.
  268  */
  269 static int
  270 ifindex_alloc_locked(u_short *idxp)
  271 {
  272         u_short idx;
  273 
  274         IFNET_WLOCK_ASSERT();
  275 
  276 retry:
  277         /*
  278          * Try to find an empty slot below V_if_index.  If we fail, take the
  279          * next slot.
  280          */
  281         for (idx = 1; idx <= V_if_index; idx++) {
  282                 if (V_ifindex_table[idx].ife_ifnet == NULL)
  283                         break;
  284         }
  285 
  286         /* Catch if_index overflow. */
  287         if (idx < 1)
  288                 return (ENOSPC);
  289         if (idx >= V_if_indexlim) {
  290                 if_grow();
  291                 goto retry;
  292         }
  293         if (idx > V_if_index)
  294                 V_if_index = idx;
  295         *idxp = idx;
  296         return (0);
  297 }
  298 
  299 static void
  300 ifindex_free_locked(u_short idx)
  301 {
  302 
  303         IFNET_WLOCK_ASSERT();
  304 
  305         V_ifindex_table[idx].ife_ifnet = NULL;
  306         while (V_if_index > 0 &&
  307             V_ifindex_table[V_if_index].ife_ifnet == NULL)
  308                 V_if_index--;
  309 }
  310 
  311 static void
  312 ifindex_free(u_short idx)
  313 {
  314 
  315         IFNET_WLOCK();
  316         ifindex_free_locked(idx);
  317         IFNET_WUNLOCK();
  318 }
  319 
  320 static void
  321 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
  322 {
  323 
  324         IFNET_WLOCK_ASSERT();
  325 
  326         V_ifindex_table[idx].ife_ifnet = ifp;
  327 }
  328 
  329 static void
  330 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
  331 {
  332 
  333         IFNET_WLOCK();
  334         ifnet_setbyindex_locked(idx, ifp);
  335         IFNET_WUNLOCK();
  336 }
  337 
  338 struct ifaddr *
  339 ifaddr_byindex(u_short idx)
  340 {
  341         struct ifaddr *ifa;
  342 
  343         IFNET_RLOCK_NOSLEEP();
  344         ifa = ifnet_byindex_locked(idx)->if_addr;
  345         if (ifa != NULL)
  346                 ifa_ref(ifa);
  347         IFNET_RUNLOCK_NOSLEEP();
  348         return (ifa);
  349 }
  350 
  351 /*
  352  * Network interface utility routines.
  353  *
  354  * Routines with ifa_ifwith* names take sockaddr *'s as
  355  * parameters.
  356  */
  357 
  358 static void
  359 vnet_if_init(const void *unused __unused)
  360 {
  361 
  362         TAILQ_INIT(&V_ifnet);
  363         TAILQ_INIT(&V_ifg_head);
  364         IFNET_WLOCK();
  365         if_grow();                              /* create initial table */
  366         IFNET_WUNLOCK();
  367         vnet_if_clone_init();
  368 }
  369 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
  370     NULL);
  371 
  372 /* ARGSUSED*/
  373 static void
  374 if_init(void *dummy __unused)
  375 {
  376 
  377         IFNET_LOCK_INIT();
  378         if_clone_init();
  379 }
  380 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
  381 
  382 
  383 #ifdef VIMAGE
  384 static void
  385 vnet_if_uninit(const void *unused __unused)
  386 {
  387 
  388         VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
  389             "not empty", __func__, __LINE__, &V_ifnet));
  390         VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
  391             "not empty", __func__, __LINE__, &V_ifg_head));
  392 
  393         free((caddr_t)V_ifindex_table, M_IFNET);
  394 }
  395 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
  396     vnet_if_uninit, NULL);
  397 #endif
  398 
  399 static void
  400 if_grow(void)
  401 {
  402         int oldlim;
  403         u_int n;
  404         struct ifindex_entry *e;
  405 
  406         IFNET_WLOCK_ASSERT();
  407         oldlim = V_if_indexlim;
  408         IFNET_WUNLOCK();
  409         n = (oldlim << 1) * sizeof(*e);
  410         e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
  411         IFNET_WLOCK();
  412         if (V_if_indexlim != oldlim) {
  413                 free(e, M_IFNET);
  414                 return;
  415         }
  416         if (V_ifindex_table != NULL) {
  417                 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
  418                 free((caddr_t)V_ifindex_table, M_IFNET);
  419         }
  420         V_if_indexlim <<= 1;
  421         V_ifindex_table = e;
  422 }
  423 
  424 /*
  425  * Allocate a struct ifnet and an index for an interface.  A layer 2
  426  * common structure will also be allocated if an allocation routine is
  427  * registered for the passed type.
  428  */
  429 struct ifnet *
  430 if_alloc(u_char type)
  431 {
  432         struct ifnet *ifp;
  433         u_short idx;
  434 
  435         ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
  436         IFNET_WLOCK();
  437         if (ifindex_alloc_locked(&idx) != 0) {
  438                 IFNET_WUNLOCK();
  439                 free(ifp, M_IFNET);
  440                 return (NULL);
  441         }
  442         ifnet_setbyindex_locked(idx, IFNET_HOLD);
  443         IFNET_WUNLOCK();
  444         ifp->if_index = idx;
  445         ifp->if_type = type;
  446         ifp->if_alloctype = type;
  447         if (if_com_alloc[type] != NULL) {
  448                 ifp->if_l2com = if_com_alloc[type](type, ifp);
  449                 if (ifp->if_l2com == NULL) {
  450                         free(ifp, M_IFNET);
  451                         ifindex_free(idx);
  452                         return (NULL);
  453                 }
  454         }
  455 
  456         IF_ADDR_LOCK_INIT(ifp);
  457         TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
  458         ifp->if_afdata_initialized = 0;
  459         IF_AFDATA_LOCK_INIT(ifp);
  460         TAILQ_INIT(&ifp->if_addrhead);
  461         TAILQ_INIT(&ifp->if_multiaddrs);
  462         TAILQ_INIT(&ifp->if_groups);
  463 #ifdef MAC
  464         mac_ifnet_init(ifp);
  465 #endif
  466         ifq_init(&ifp->if_snd, ifp);
  467 
  468         refcount_init(&ifp->if_refcount, 1);    /* Index reference. */
  469         ifnet_setbyindex(ifp->if_index, ifp);
  470         return (ifp);
  471 }
  472 
  473 /*
  474  * Do the actual work of freeing a struct ifnet, and layer 2 common
  475  * structure.  This call is made when the last reference to an
  476  * interface is released.
  477  */
  478 static void
  479 if_free_internal(struct ifnet *ifp)
  480 {
  481 
  482         KASSERT((ifp->if_flags & IFF_DYING),
  483             ("if_free_internal: interface not dying"));
  484 
  485         if (if_com_free[ifp->if_alloctype] != NULL)
  486                 if_com_free[ifp->if_alloctype](ifp->if_l2com,
  487                     ifp->if_alloctype);
  488 
  489 #ifdef MAC
  490         mac_ifnet_destroy(ifp);
  491 #endif /* MAC */
  492         if (ifp->if_description != NULL)
  493                 free(ifp->if_description, M_IFDESCR);
  494         IF_AFDATA_DESTROY(ifp);
  495         IF_ADDR_LOCK_DESTROY(ifp);
  496         ifq_delete(&ifp->if_snd);
  497         free(ifp, M_IFNET);
  498 }
  499 
  500 /*
  501  * Deregister an interface and free the associated storage.
  502  */
  503 void
  504 if_free(struct ifnet *ifp)
  505 {
  506 
  507         ifp->if_flags |= IFF_DYING;                     /* XXX: Locking */
  508 
  509         CURVNET_SET_QUIET(ifp->if_vnet);
  510         IFNET_WLOCK();
  511         KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
  512             ("%s: freeing unallocated ifnet", ifp->if_xname));
  513 
  514         ifindex_free_locked(ifp->if_index);
  515         IFNET_WUNLOCK();
  516 
  517         if (refcount_release(&ifp->if_refcount))
  518                 if_free_internal(ifp);
  519         CURVNET_RESTORE();
  520 }
  521 
  522 /*
  523  * Interfaces to keep an ifnet type-stable despite the possibility of the
  524  * driver calling if_free().  If there are additional references, we defer
  525  * freeing the underlying data structure.
  526  */
  527 void
  528 if_ref(struct ifnet *ifp)
  529 {
  530 
  531         /* We don't assert the ifnet list lock here, but arguably should. */
  532         refcount_acquire(&ifp->if_refcount);
  533 }
  534 
  535 void
  536 if_rele(struct ifnet *ifp)
  537 {
  538 
  539         if (!refcount_release(&ifp->if_refcount))
  540                 return;
  541         if_free_internal(ifp);
  542 }
  543 
  544 void
  545 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
  546 {
  547         
  548         mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
  549 
  550         if (ifq->ifq_maxlen == 0) 
  551                 ifq->ifq_maxlen = ifqmaxlen;
  552 
  553         ifq->altq_type = 0;
  554         ifq->altq_disc = NULL;
  555         ifq->altq_flags &= ALTQF_CANTCHANGE;
  556         ifq->altq_tbr  = NULL;
  557         ifq->altq_ifp  = ifp;
  558 }
  559 
  560 void
  561 ifq_delete(struct ifaltq *ifq)
  562 {
  563         mtx_destroy(&ifq->ifq_mtx);
  564 }
  565 
  566 /*
  567  * Perform generic interface initalization tasks and attach the interface
  568  * to the list of "active" interfaces.  If vmove flag is set on entry
  569  * to if_attach_internal(), perform only a limited subset of initialization
  570  * tasks, given that we are moving from one vnet to another an ifnet which
  571  * has already been fully initialized.
  572  *
  573  * XXX:
  574  *  - The decision to return void and thus require this function to
  575  *    succeed is questionable.
  576  *  - We should probably do more sanity checking.  For instance we don't
  577  *    do anything to insure if_xname is unique or non-empty.
  578  */
  579 void
  580 if_attach(struct ifnet *ifp)
  581 {
  582 
  583         if_attach_internal(ifp, 0);
  584 }
  585 
  586 static void
  587 if_attach_internal(struct ifnet *ifp, int vmove)
  588 {
  589         unsigned socksize, ifasize;
  590         int namelen, masklen;
  591         struct sockaddr_dl *sdl;
  592         struct ifaddr *ifa;
  593 
  594         if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
  595                 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
  596                     ifp->if_xname);
  597 
  598 #ifdef VIMAGE
  599         ifp->if_vnet = curvnet;
  600         if (ifp->if_home_vnet == NULL)
  601                 ifp->if_home_vnet = curvnet;
  602 #endif
  603 
  604         if_addgroup(ifp, IFG_ALL);
  605 
  606         getmicrotime(&ifp->if_lastchange);
  607         ifp->if_data.ifi_epoch = time_uptime;
  608         ifp->if_data.ifi_datalen = sizeof(struct if_data);
  609 
  610         KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
  611             (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
  612             ("transmit and qflush must both either be set or both be NULL"));
  613         if (ifp->if_transmit == NULL) {
  614                 ifp->if_transmit = if_transmit;
  615                 ifp->if_qflush = if_qflush;
  616         }
  617         
  618         if (!vmove) {
  619 #ifdef MAC
  620                 mac_ifnet_create(ifp);
  621 #endif
  622 
  623                 /*
  624                  * Create a Link Level name for this device.
  625                  */
  626                 namelen = strlen(ifp->if_xname);
  627                 /*
  628                  * Always save enough space for any possiable name so we
  629                  * can do a rename in place later.
  630                  */
  631                 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
  632                 socksize = masklen + ifp->if_addrlen;
  633                 if (socksize < sizeof(*sdl))
  634                         socksize = sizeof(*sdl);
  635                 socksize = roundup2(socksize, sizeof(long));
  636                 ifasize = sizeof(*ifa) + 2 * socksize;
  637                 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
  638                 ifa_init(ifa);
  639                 sdl = (struct sockaddr_dl *)(ifa + 1);
  640                 sdl->sdl_len = socksize;
  641                 sdl->sdl_family = AF_LINK;
  642                 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
  643                 sdl->sdl_nlen = namelen;
  644                 sdl->sdl_index = ifp->if_index;
  645                 sdl->sdl_type = ifp->if_type;
  646                 ifp->if_addr = ifa;
  647                 ifa->ifa_ifp = ifp;
  648                 ifa->ifa_rtrequest = link_rtrequest;
  649                 ifa->ifa_addr = (struct sockaddr *)sdl;
  650                 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
  651                 ifa->ifa_netmask = (struct sockaddr *)sdl;
  652                 sdl->sdl_len = masklen;
  653                 while (namelen != 0)
  654                         sdl->sdl_data[--namelen] = 0xff;
  655                 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
  656                 /* Reliably crash if used uninitialized. */
  657                 ifp->if_broadcastaddr = NULL;
  658 
  659 #if defined(INET) || defined(INET6)
  660                 /* Initialize to max value. */
  661                 if (ifp->if_hw_tsomax == 0)
  662                         ifp->if_hw_tsomax = min(IP_MAXPACKET, 32 * MCLBYTES -
  663                             (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
  664                 KASSERT(ifp->if_hw_tsomax <= IP_MAXPACKET &&
  665                     ifp->if_hw_tsomax >= IP_MAXPACKET / 8,
  666                     ("%s: tsomax outside of range", __func__));
  667 #endif
  668         }
  669 #ifdef VIMAGE
  670         else {
  671                 /*
  672                  * Update the interface index in the link layer address
  673                  * of the interface.
  674                  */
  675                 for (ifa = ifp->if_addr; ifa != NULL;
  676                     ifa = TAILQ_NEXT(ifa, ifa_link)) {
  677                         if (ifa->ifa_addr->sa_family == AF_LINK) {
  678                                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
  679                                 sdl->sdl_index = ifp->if_index;
  680                         }
  681                 }
  682         }
  683 #endif
  684 
  685         IFNET_WLOCK();
  686         TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
  687 #ifdef VIMAGE
  688         curvnet->vnet_ifcnt++;
  689 #endif
  690         IFNET_WUNLOCK();
  691 
  692         if (domain_init_status >= 2)
  693                 if_attachdomain1(ifp);
  694 
  695         EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
  696         if (IS_DEFAULT_VNET(curvnet))
  697                 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
  698 
  699         /* Announce the interface. */
  700         rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
  701 }
  702 
  703 static void
  704 if_attachdomain(void *dummy)
  705 {
  706         struct ifnet *ifp;
  707 
  708         TAILQ_FOREACH(ifp, &V_ifnet, if_link)
  709                 if_attachdomain1(ifp);
  710 }
  711 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
  712     if_attachdomain, NULL);
  713 
  714 static void
  715 if_attachdomain1(struct ifnet *ifp)
  716 {
  717         struct domain *dp;
  718 
  719         /*
  720          * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
  721          * cannot lock ifp->if_afdata initialization, entirely.
  722          */
  723         if (IF_AFDATA_TRYLOCK(ifp) == 0)
  724                 return;
  725         if (ifp->if_afdata_initialized >= domain_init_status) {
  726                 IF_AFDATA_UNLOCK(ifp);
  727                 log(LOG_WARNING, "%s called more than once on %s\n",
  728                     __func__, ifp->if_xname);
  729                 return;
  730         }
  731         ifp->if_afdata_initialized = domain_init_status;
  732         IF_AFDATA_UNLOCK(ifp);
  733 
  734         /* address family dependent data region */
  735         bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
  736         for (dp = domains; dp; dp = dp->dom_next) {
  737                 if (dp->dom_ifattach)
  738                         ifp->if_afdata[dp->dom_family] =
  739                             (*dp->dom_ifattach)(ifp);
  740         }
  741 }
  742 
  743 /*
  744  * Remove any unicast or broadcast network addresses from an interface.
  745  */
  746 void
  747 if_purgeaddrs(struct ifnet *ifp)
  748 {
  749         struct ifaddr *ifa, *next;
  750 
  751         TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  752                 if (ifa->ifa_addr->sa_family == AF_LINK)
  753                         continue;
  754 #ifdef INET
  755                 /* XXX: Ugly!! ad hoc just for INET */
  756                 if (ifa->ifa_addr->sa_family == AF_INET) {
  757                         struct ifaliasreq ifr;
  758 
  759                         bzero(&ifr, sizeof(ifr));
  760                         ifr.ifra_addr = *ifa->ifa_addr;
  761                         if (ifa->ifa_dstaddr)
  762                                 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
  763                         if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
  764                             NULL) == 0)
  765                                 continue;
  766                 }
  767 #endif /* INET */
  768 #ifdef INET6
  769                 if (ifa->ifa_addr->sa_family == AF_INET6) {
  770                         in6_purgeaddr(ifa);
  771                         /* ifp_addrhead is already updated */
  772                         continue;
  773                 }
  774 #endif /* INET6 */
  775                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  776                 ifa_free(ifa);
  777         }
  778 }
  779 
  780 /*
  781  * Remove any multicast network addresses from an interface when an ifnet
  782  * is going away.
  783  */
  784 static void
  785 if_purgemaddrs(struct ifnet *ifp)
  786 {
  787         struct ifmultiaddr *ifma;
  788         struct ifmultiaddr *next;
  789 
  790         IF_ADDR_WLOCK(ifp);
  791         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
  792                 if_delmulti_locked(ifp, ifma, 1);
  793         IF_ADDR_WUNLOCK(ifp);
  794 }
  795 
  796 /*
  797  * Detach an interface, removing it from the list of "active" interfaces.
  798  * If vmove flag is set on entry to if_detach_internal(), perform only a
  799  * limited subset of cleanup tasks, given that we are moving an ifnet from
  800  * one vnet to another, where it must be fully operational.
  801  *
  802  * XXXRW: There are some significant questions about event ordering, and
  803  * how to prevent things from starting to use the interface during detach.
  804  */
  805 void
  806 if_detach(struct ifnet *ifp)
  807 {
  808 
  809         CURVNET_SET_QUIET(ifp->if_vnet);
  810         if_detach_internal(ifp, 0);
  811         CURVNET_RESTORE();
  812 }
  813 
  814 static void
  815 if_detach_internal(struct ifnet *ifp, int vmove)
  816 {
  817         struct ifaddr *ifa;
  818         struct radix_node_head  *rnh;
  819         int i, j;
  820         struct domain *dp;
  821         struct ifnet *iter;
  822         int found = 0;
  823 
  824         IFNET_WLOCK();
  825         TAILQ_FOREACH(iter, &V_ifnet, if_link)
  826                 if (iter == ifp) {
  827                         TAILQ_REMOVE(&V_ifnet, ifp, if_link);
  828                         found = 1;
  829                         break;
  830                 }
  831 #ifdef VIMAGE
  832         if (found)
  833                 curvnet->vnet_ifcnt--;
  834 #endif
  835         IFNET_WUNLOCK();
  836         if (!found) {
  837                 if (vmove)
  838                         panic("%s: ifp=%p not on the ifnet tailq %p",
  839                             __func__, ifp, &V_ifnet);
  840                 else
  841                         return; /* XXX this should panic as well? */
  842         }
  843 
  844         /*
  845          * Remove/wait for pending events.
  846          */
  847         taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
  848 
  849         /*
  850          * Remove routes and flush queues.
  851          */
  852         if_down(ifp);
  853 #ifdef ALTQ
  854         if (ALTQ_IS_ENABLED(&ifp->if_snd))
  855                 altq_disable(&ifp->if_snd);
  856         if (ALTQ_IS_ATTACHED(&ifp->if_snd))
  857                 altq_detach(&ifp->if_snd);
  858 #endif
  859 
  860         if_purgeaddrs(ifp);
  861 
  862 #ifdef INET
  863         in_ifdetach(ifp);
  864 #endif
  865 
  866 #ifdef INET6
  867         /*
  868          * Remove all IPv6 kernel structs related to ifp.  This should be done
  869          * before removing routing entries below, since IPv6 interface direct
  870          * routes are expected to be removed by the IPv6-specific kernel API.
  871          * Otherwise, the kernel will detect some inconsistency and bark it.
  872          */
  873         in6_ifdetach(ifp);
  874 #endif
  875         if_purgemaddrs(ifp);
  876 
  877         if (!vmove) {
  878                 /*
  879                  * Prevent further calls into the device driver via ifnet.
  880                  */
  881                 if_dead(ifp);
  882 
  883                 /*
  884                  * Remove link ifaddr pointer and maybe decrement if_index.
  885                  * Clean up all addresses.
  886                  */
  887                 ifp->if_addr = NULL;
  888 
  889                 /* We can now free link ifaddr. */
  890                 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
  891                         ifa = TAILQ_FIRST(&ifp->if_addrhead);
  892                         TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  893                         ifa_free(ifa);
  894                 }
  895         }
  896 
  897         /*
  898          * Delete all remaining routes using this interface
  899          * Unfortuneatly the only way to do this is to slog through
  900          * the entire routing table looking for routes which point
  901          * to this interface...oh well...
  902          */
  903         for (i = 1; i <= AF_MAX; i++) {
  904                 for (j = 0; j < rt_numfibs; j++) {
  905                         rnh = rt_tables_get_rnh(j, i);
  906                         if (rnh == NULL)
  907                                 continue;
  908                         RADIX_NODE_HEAD_LOCK(rnh);
  909                         (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
  910                         RADIX_NODE_HEAD_UNLOCK(rnh);
  911                 }
  912         }
  913 
  914         /* Announce that the interface is gone. */
  915         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
  916         EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
  917         if (IS_DEFAULT_VNET(curvnet))
  918                 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
  919         if_delgroups(ifp);
  920 
  921         /*
  922          * We cannot hold the lock over dom_ifdetach calls as they might
  923          * sleep, for example trying to drain a callout, thus open up the
  924          * theoretical race with re-attaching.
  925          */
  926         IF_AFDATA_LOCK(ifp);
  927         i = ifp->if_afdata_initialized;
  928         ifp->if_afdata_initialized = 0;
  929         IF_AFDATA_UNLOCK(ifp);
  930         for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
  931                 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
  932                         (*dp->dom_ifdetach)(ifp,
  933                             ifp->if_afdata[dp->dom_family]);
  934         }
  935 }
  936 
  937 #ifdef VIMAGE
  938 /*
  939  * if_vmove() performs a limited version of if_detach() in current
  940  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
  941  * An attempt is made to shrink if_index in current vnet, find an
  942  * unused if_index in target vnet and calls if_grow() if necessary,
  943  * and finally find an unused if_xname for the target vnet.
  944  */
  945 void
  946 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
  947 {
  948         u_short idx;
  949 
  950         /*
  951          * Detach from current vnet, but preserve LLADDR info, do not
  952          * mark as dead etc. so that the ifnet can be reattached later.
  953          */
  954         if_detach_internal(ifp, 1);
  955 
  956         /*
  957          * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
  958          * the if_index for that vnet if possible.
  959          *
  960          * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
  961          * or we'd lock on one vnet and unlock on another.
  962          */
  963         IFNET_WLOCK();
  964         ifindex_free_locked(ifp->if_index);
  965         IFNET_WUNLOCK();
  966 
  967         /*
  968          * Perform interface-specific reassignment tasks, if provided by
  969          * the driver.
  970          */
  971         if (ifp->if_reassign != NULL)
  972                 ifp->if_reassign(ifp, new_vnet, NULL);
  973 
  974         /*
  975          * Switch to the context of the target vnet.
  976          */
  977         CURVNET_SET_QUIET(new_vnet);
  978 
  979         IFNET_WLOCK();
  980         if (ifindex_alloc_locked(&idx) != 0) {
  981                 IFNET_WUNLOCK();
  982                 panic("if_index overflow");
  983         }
  984         ifp->if_index = idx;
  985         ifnet_setbyindex_locked(ifp->if_index, ifp);
  986         IFNET_WUNLOCK();
  987 
  988         if_attach_internal(ifp, 1);
  989 
  990         CURVNET_RESTORE();
  991 }
  992 
  993 /*
  994  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
  995  */
  996 static int
  997 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
  998 {
  999         struct prison *pr;
 1000         struct ifnet *difp;
 1001 
 1002         /* Try to find the prison within our visibility. */
 1003         sx_slock(&allprison_lock);
 1004         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1005         sx_sunlock(&allprison_lock);
 1006         if (pr == NULL)
 1007                 return (ENXIO);
 1008         prison_hold_locked(pr);
 1009         mtx_unlock(&pr->pr_mtx);
 1010 
 1011         /* Do not try to move the iface from and to the same prison. */
 1012         if (pr->pr_vnet == ifp->if_vnet) {
 1013                 prison_free(pr);
 1014                 return (EEXIST);
 1015         }
 1016 
 1017         /* Make sure the named iface does not exists in the dst. prison/vnet. */
 1018         /* XXX Lock interfaces to avoid races. */
 1019         CURVNET_SET_QUIET(pr->pr_vnet);
 1020         difp = ifunit(ifname);
 1021         CURVNET_RESTORE();
 1022         if (difp != NULL) {
 1023                 prison_free(pr);
 1024                 return (EEXIST);
 1025         }
 1026 
 1027         /* Move the interface into the child jail/vnet. */
 1028         if_vmove(ifp, pr->pr_vnet);
 1029 
 1030         /* Report the new if_xname back to the userland. */
 1031         sprintf(ifname, "%s", ifp->if_xname);
 1032 
 1033         prison_free(pr);
 1034         return (0);
 1035 }
 1036 
 1037 static int
 1038 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
 1039 {
 1040         struct prison *pr;
 1041         struct vnet *vnet_dst;
 1042         struct ifnet *ifp;
 1043 
 1044         /* Try to find the prison within our visibility. */
 1045         sx_slock(&allprison_lock);
 1046         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1047         sx_sunlock(&allprison_lock);
 1048         if (pr == NULL)
 1049                 return (ENXIO);
 1050         prison_hold_locked(pr);
 1051         mtx_unlock(&pr->pr_mtx);
 1052 
 1053         /* Make sure the named iface exists in the source prison/vnet. */
 1054         CURVNET_SET(pr->pr_vnet);
 1055         ifp = ifunit(ifname);           /* XXX Lock to avoid races. */
 1056         if (ifp == NULL) {
 1057                 CURVNET_RESTORE();
 1058                 prison_free(pr);
 1059                 return (ENXIO);
 1060         }
 1061 
 1062         /* Do not try to move the iface from and to the same prison. */
 1063         vnet_dst = TD_TO_VNET(td);
 1064         if (vnet_dst == ifp->if_vnet) {
 1065                 CURVNET_RESTORE();
 1066                 prison_free(pr);
 1067                 return (EEXIST);
 1068         }
 1069 
 1070         /* Get interface back from child jail/vnet. */
 1071         if_vmove(ifp, vnet_dst);
 1072         CURVNET_RESTORE();
 1073 
 1074         /* Report the new if_xname back to the userland. */
 1075         sprintf(ifname, "%s", ifp->if_xname);
 1076 
 1077         prison_free(pr);
 1078         return (0);
 1079 }
 1080 #endif /* VIMAGE */
 1081 
 1082 /*
 1083  * Add a group to an interface
 1084  */
 1085 int
 1086 if_addgroup(struct ifnet *ifp, const char *groupname)
 1087 {
 1088         struct ifg_list         *ifgl;
 1089         struct ifg_group        *ifg = NULL;
 1090         struct ifg_member       *ifgm;
 1091         int                      new = 0;
 1092 
 1093         if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
 1094             groupname[strlen(groupname) - 1] <= '9')
 1095                 return (EINVAL);
 1096 
 1097         IFNET_WLOCK();
 1098         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1099                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
 1100                         IFNET_WUNLOCK();
 1101                         return (EEXIST);
 1102                 }
 1103 
 1104         if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
 1105             M_NOWAIT)) == NULL) {
 1106                 IFNET_WUNLOCK();
 1107                 return (ENOMEM);
 1108         }
 1109 
 1110         if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
 1111             M_TEMP, M_NOWAIT)) == NULL) {
 1112                 free(ifgl, M_TEMP);
 1113                 IFNET_WUNLOCK();
 1114                 return (ENOMEM);
 1115         }
 1116 
 1117         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1118                 if (!strcmp(ifg->ifg_group, groupname))
 1119                         break;
 1120 
 1121         if (ifg == NULL) {
 1122                 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
 1123                     M_TEMP, M_NOWAIT)) == NULL) {
 1124                         free(ifgl, M_TEMP);
 1125                         free(ifgm, M_TEMP);
 1126                         IFNET_WUNLOCK();
 1127                         return (ENOMEM);
 1128                 }
 1129                 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
 1130                 ifg->ifg_refcnt = 0;
 1131                 TAILQ_INIT(&ifg->ifg_members);
 1132                 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
 1133                 new = 1;
 1134         }
 1135 
 1136         ifg->ifg_refcnt++;
 1137         ifgl->ifgl_group = ifg;
 1138         ifgm->ifgm_ifp = ifp;
 1139 
 1140         IF_ADDR_WLOCK(ifp);
 1141         TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
 1142         TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
 1143         IF_ADDR_WUNLOCK(ifp);
 1144 
 1145         IFNET_WUNLOCK();
 1146 
 1147         if (new)
 1148                 EVENTHANDLER_INVOKE(group_attach_event, ifg);
 1149         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1150 
 1151         return (0);
 1152 }
 1153 
 1154 /*
 1155  * Remove a group from an interface
 1156  */
 1157 int
 1158 if_delgroup(struct ifnet *ifp, const char *groupname)
 1159 {
 1160         struct ifg_list         *ifgl;
 1161         struct ifg_member       *ifgm;
 1162 
 1163         IFNET_WLOCK();
 1164         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1165                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
 1166                         break;
 1167         if (ifgl == NULL) {
 1168                 IFNET_WUNLOCK();
 1169                 return (ENOENT);
 1170         }
 1171 
 1172         IF_ADDR_WLOCK(ifp);
 1173         TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1174         IF_ADDR_WUNLOCK(ifp);
 1175 
 1176         TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1177                 if (ifgm->ifgm_ifp == ifp)
 1178                         break;
 1179 
 1180         if (ifgm != NULL) {
 1181                 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
 1182                 free(ifgm, M_TEMP);
 1183         }
 1184 
 1185         if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1186                 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1187                 IFNET_WUNLOCK();
 1188                 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
 1189                 free(ifgl->ifgl_group, M_TEMP);
 1190         } else
 1191                 IFNET_WUNLOCK();
 1192 
 1193         free(ifgl, M_TEMP);
 1194 
 1195         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1196 
 1197         return (0);
 1198 }
 1199 
 1200 /*
 1201  * Remove an interface from all groups
 1202  */
 1203 static void
 1204 if_delgroups(struct ifnet *ifp)
 1205 {
 1206         struct ifg_list         *ifgl;
 1207         struct ifg_member       *ifgm;
 1208         char groupname[IFNAMSIZ];
 1209 
 1210         IFNET_WLOCK();
 1211         while (!TAILQ_EMPTY(&ifp->if_groups)) {
 1212                 ifgl = TAILQ_FIRST(&ifp->if_groups);
 1213 
 1214                 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
 1215 
 1216                 IF_ADDR_WLOCK(ifp);
 1217                 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1218                 IF_ADDR_WUNLOCK(ifp);
 1219 
 1220                 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1221                         if (ifgm->ifgm_ifp == ifp)
 1222                                 break;
 1223 
 1224                 if (ifgm != NULL) {
 1225                         TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
 1226                             ifgm_next);
 1227                         free(ifgm, M_TEMP);
 1228                 }
 1229 
 1230                 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1231                         TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1232                         IFNET_WUNLOCK();
 1233                         EVENTHANDLER_INVOKE(group_detach_event,
 1234                             ifgl->ifgl_group);
 1235                         free(ifgl->ifgl_group, M_TEMP);
 1236                 } else
 1237                         IFNET_WUNLOCK();
 1238 
 1239                 free(ifgl, M_TEMP);
 1240 
 1241                 EVENTHANDLER_INVOKE(group_change_event, groupname);
 1242 
 1243                 IFNET_WLOCK();
 1244         }
 1245         IFNET_WUNLOCK();
 1246 }
 1247 
 1248 /*
 1249  * Stores all groups from an interface in memory pointed
 1250  * to by data
 1251  */
 1252 static int
 1253 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
 1254 {
 1255         int                      len, error;
 1256         struct ifg_list         *ifgl;
 1257         struct ifg_req           ifgrq, *ifgp;
 1258         struct ifgroupreq       *ifgr = data;
 1259 
 1260         if (ifgr->ifgr_len == 0) {
 1261                 IF_ADDR_RLOCK(ifp);
 1262                 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1263                         ifgr->ifgr_len += sizeof(struct ifg_req);
 1264                 IF_ADDR_RUNLOCK(ifp);
 1265                 return (0);
 1266         }
 1267 
 1268         len = ifgr->ifgr_len;
 1269         ifgp = ifgr->ifgr_groups;
 1270         /* XXX: wire */
 1271         IF_ADDR_RLOCK(ifp);
 1272         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
 1273                 if (len < sizeof(ifgrq)) {
 1274                         IF_ADDR_RUNLOCK(ifp);
 1275                         return (EINVAL);
 1276                 }
 1277                 bzero(&ifgrq, sizeof ifgrq);
 1278                 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
 1279                     sizeof(ifgrq.ifgrq_group));
 1280                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1281                         IF_ADDR_RUNLOCK(ifp);
 1282                         return (error);
 1283                 }
 1284                 len -= sizeof(ifgrq);
 1285                 ifgp++;
 1286         }
 1287         IF_ADDR_RUNLOCK(ifp);
 1288 
 1289         return (0);
 1290 }
 1291 
 1292 /*
 1293  * Stores all members of a group in memory pointed to by data
 1294  */
 1295 static int
 1296 if_getgroupmembers(struct ifgroupreq *data)
 1297 {
 1298         struct ifgroupreq       *ifgr = data;
 1299         struct ifg_group        *ifg;
 1300         struct ifg_member       *ifgm;
 1301         struct ifg_req           ifgrq, *ifgp;
 1302         int                      len, error;
 1303 
 1304         IFNET_RLOCK();
 1305         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1306                 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
 1307                         break;
 1308         if (ifg == NULL) {
 1309                 IFNET_RUNLOCK();
 1310                 return (ENOENT);
 1311         }
 1312 
 1313         if (ifgr->ifgr_len == 0) {
 1314                 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
 1315                         ifgr->ifgr_len += sizeof(ifgrq);
 1316                 IFNET_RUNLOCK();
 1317                 return (0);
 1318         }
 1319 
 1320         len = ifgr->ifgr_len;
 1321         ifgp = ifgr->ifgr_groups;
 1322         TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
 1323                 if (len < sizeof(ifgrq)) {
 1324                         IFNET_RUNLOCK();
 1325                         return (EINVAL);
 1326                 }
 1327                 bzero(&ifgrq, sizeof ifgrq);
 1328                 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
 1329                     sizeof(ifgrq.ifgrq_member));
 1330                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1331                         IFNET_RUNLOCK();
 1332                         return (error);
 1333                 }
 1334                 len -= sizeof(ifgrq);
 1335                 ifgp++;
 1336         }
 1337         IFNET_RUNLOCK();
 1338 
 1339         return (0);
 1340 }
 1341 
 1342 /*
 1343  * Delete Routes for a Network Interface
 1344  *
 1345  * Called for each routing entry via the rnh->rnh_walktree() call above
 1346  * to delete all route entries referencing a detaching network interface.
 1347  *
 1348  * Arguments:
 1349  *      rn      pointer to node in the routing table
 1350  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
 1351  *
 1352  * Returns:
 1353  *      0       successful
 1354  *      errno   failed - reason indicated
 1355  *
 1356  */
 1357 static int
 1358 if_rtdel(struct radix_node *rn, void *arg)
 1359 {
 1360         struct rtentry  *rt = (struct rtentry *)rn;
 1361         struct ifnet    *ifp = arg;
 1362         int             err;
 1363 
 1364         if (rt->rt_ifp == ifp) {
 1365 
 1366                 /*
 1367                  * Protect (sorta) against walktree recursion problems
 1368                  * with cloned routes
 1369                  */
 1370                 if ((rt->rt_flags & RTF_UP) == 0)
 1371                         return (0);
 1372 
 1373                 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
 1374                                 rt_mask(rt),
 1375                                 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
 1376                                 (struct rtentry **) NULL, rt->rt_fibnum);
 1377                 if (err) {
 1378                         log(LOG_WARNING, "if_rtdel: error %d\n", err);
 1379                 }
 1380         }
 1381 
 1382         return (0);
 1383 }
 1384 
 1385 /*
 1386  * Wrapper functions for struct ifnet address list locking macros.  These are
 1387  * used by kernel modules to avoid encoding programming interface or binary
 1388  * interface assumptions that may be violated when kernel-internal locking
 1389  * approaches change.
 1390  */
 1391 void
 1392 if_addr_rlock(struct ifnet *ifp)
 1393 {
 1394 
 1395         IF_ADDR_RLOCK(ifp);
 1396 }
 1397 
 1398 void
 1399 if_addr_runlock(struct ifnet *ifp)
 1400 {
 1401 
 1402         IF_ADDR_RUNLOCK(ifp);
 1403 }
 1404 
 1405 void
 1406 if_maddr_rlock(struct ifnet *ifp)
 1407 {
 1408 
 1409         IF_ADDR_RLOCK(ifp);
 1410 }
 1411 
 1412 void
 1413 if_maddr_runlock(struct ifnet *ifp)
 1414 {
 1415 
 1416         IF_ADDR_RUNLOCK(ifp);
 1417 }
 1418 
 1419 /*
 1420  * Initialization, destruction and refcounting functions for ifaddrs.
 1421  */
 1422 void
 1423 ifa_init(struct ifaddr *ifa)
 1424 {
 1425 
 1426         mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
 1427         refcount_init(&ifa->ifa_refcnt, 1);
 1428         ifa->if_data.ifi_datalen = sizeof(ifa->if_data);
 1429 }
 1430 
 1431 void
 1432 ifa_ref(struct ifaddr *ifa)
 1433 {
 1434 
 1435         refcount_acquire(&ifa->ifa_refcnt);
 1436 }
 1437 
 1438 void
 1439 ifa_free(struct ifaddr *ifa)
 1440 {
 1441 
 1442         if (refcount_release(&ifa->ifa_refcnt)) {
 1443                 mtx_destroy(&ifa->ifa_mtx);
 1444                 free(ifa, M_IFADDR);
 1445         }
 1446 }
 1447 
 1448 int
 1449 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1450 {
 1451         int error = 0;
 1452         struct rtentry *rt = NULL;
 1453         struct rt_addrinfo info;
 1454         static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1455 
 1456         bzero(&info, sizeof(info));
 1457         info.rti_ifp = V_loif;
 1458         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1459         info.rti_info[RTAX_DST] = ia;
 1460         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1461         error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
 1462 
 1463         if (error == 0 && rt != NULL) {
 1464                 RT_LOCK(rt);
 1465                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
 1466                         ifa->ifa_ifp->if_type;
 1467                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
 1468                         ifa->ifa_ifp->if_index;
 1469                 RT_REMREF(rt);
 1470                 RT_UNLOCK(rt);
 1471         } else if (error != 0)
 1472                 log(LOG_DEBUG, "%s: insertion failed: %u\n", __func__, error);
 1473 
 1474         return (error);
 1475 }
 1476 
 1477 int
 1478 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1479 {
 1480         int error = 0;
 1481         struct rt_addrinfo info;
 1482         struct sockaddr_dl null_sdl;
 1483 
 1484         bzero(&null_sdl, sizeof(null_sdl));
 1485         null_sdl.sdl_len = sizeof(null_sdl);
 1486         null_sdl.sdl_family = AF_LINK;
 1487         null_sdl.sdl_type = ifa->ifa_ifp->if_type;
 1488         null_sdl.sdl_index = ifa->ifa_ifp->if_index;
 1489         bzero(&info, sizeof(info));
 1490         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1491         info.rti_info[RTAX_DST] = ia;
 1492         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1493         error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
 1494 
 1495         if (error != 0)
 1496                 log(LOG_DEBUG, "%s: deletion failed: %u\n", __func__, error);
 1497 
 1498         return (error);
 1499 }
 1500 
 1501 /*
 1502  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
 1503  * structs used to represent other address families, it is necessary
 1504  * to perform a different comparison.
 1505  */
 1506 
 1507 #define sa_equal(a1, a2)        \
 1508         (bcmp((a1), (a2), ((a1))->sa_len) == 0)
 1509 
 1510 #define sa_dl_equal(a1, a2)     \
 1511         ((((struct sockaddr_dl *)(a1))->sdl_len ==                      \
 1512          ((struct sockaddr_dl *)(a2))->sdl_len) &&                      \
 1513          (bcmp(LLADDR((struct sockaddr_dl *)(a1)),                      \
 1514                LLADDR((struct sockaddr_dl *)(a2)),                      \
 1515                ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
 1516 
 1517 /*
 1518  * Locate an interface based on a complete address.
 1519  */
 1520 /*ARGSUSED*/
 1521 static struct ifaddr *
 1522 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
 1523 {
 1524         struct ifnet *ifp;
 1525         struct ifaddr *ifa;
 1526 
 1527         IFNET_RLOCK_NOSLEEP();
 1528         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1529                 IF_ADDR_RLOCK(ifp);
 1530                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1531                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1532                                 continue;
 1533                         if (sa_equal(addr, ifa->ifa_addr)) {
 1534                                 if (getref)
 1535                                         ifa_ref(ifa);
 1536                                 IF_ADDR_RUNLOCK(ifp);
 1537                                 goto done;
 1538                         }
 1539                         /* IP6 doesn't have broadcast */
 1540                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1541                             ifa->ifa_broadaddr &&
 1542                             ifa->ifa_broadaddr->sa_len != 0 &&
 1543                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1544                                 if (getref)
 1545                                         ifa_ref(ifa);
 1546                                 IF_ADDR_RUNLOCK(ifp);
 1547                                 goto done;
 1548                         }
 1549                 }
 1550                 IF_ADDR_RUNLOCK(ifp);
 1551         }
 1552         ifa = NULL;
 1553 done:
 1554         IFNET_RUNLOCK_NOSLEEP();
 1555         return (ifa);
 1556 }
 1557 
 1558 struct ifaddr *
 1559 ifa_ifwithaddr(struct sockaddr *addr)
 1560 {
 1561 
 1562         return (ifa_ifwithaddr_internal(addr, 1));
 1563 }
 1564 
 1565 int
 1566 ifa_ifwithaddr_check(struct sockaddr *addr)
 1567 {
 1568 
 1569         return (ifa_ifwithaddr_internal(addr, 0) != NULL);
 1570 }
 1571 
 1572 /*
 1573  * Locate an interface based on the broadcast address.
 1574  */
 1575 /* ARGSUSED */
 1576 struct ifaddr *
 1577 ifa_ifwithbroadaddr(struct sockaddr *addr)
 1578 {
 1579         struct ifnet *ifp;
 1580         struct ifaddr *ifa;
 1581 
 1582         IFNET_RLOCK_NOSLEEP();
 1583         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1584                 IF_ADDR_RLOCK(ifp);
 1585                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1586                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1587                                 continue;
 1588                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1589                             ifa->ifa_broadaddr &&
 1590                             ifa->ifa_broadaddr->sa_len != 0 &&
 1591                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1592                                 ifa_ref(ifa);
 1593                                 IF_ADDR_RUNLOCK(ifp);
 1594                                 goto done;
 1595                         }
 1596                 }
 1597                 IF_ADDR_RUNLOCK(ifp);
 1598         }
 1599         ifa = NULL;
 1600 done:
 1601         IFNET_RUNLOCK_NOSLEEP();
 1602         return (ifa);
 1603 }
 1604 
 1605 /*
 1606  * Locate the point to point interface with a given destination address.
 1607  */
 1608 /*ARGSUSED*/
 1609 struct ifaddr *
 1610 ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum)
 1611 {
 1612         struct ifnet *ifp;
 1613         struct ifaddr *ifa;
 1614 
 1615         IFNET_RLOCK_NOSLEEP();
 1616         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1617                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
 1618                         continue;
 1619                 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
 1620                         continue;
 1621                 IF_ADDR_RLOCK(ifp);
 1622                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1623                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1624                                 continue;
 1625                         if (ifa->ifa_dstaddr != NULL &&
 1626                             sa_equal(addr, ifa->ifa_dstaddr)) {
 1627                                 ifa_ref(ifa);
 1628                                 IF_ADDR_RUNLOCK(ifp);
 1629                                 goto done;
 1630                         }
 1631                 }
 1632                 IF_ADDR_RUNLOCK(ifp);
 1633         }
 1634         ifa = NULL;
 1635 done:
 1636         IFNET_RUNLOCK_NOSLEEP();
 1637         return (ifa);
 1638 }
 1639 
 1640 struct ifaddr *
 1641 ifa_ifwithdstaddr(struct sockaddr *addr)
 1642 {
 1643 
 1644         return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS));
 1645 }
 1646 
 1647 /*
 1648  * Find an interface on a specific network.  If many, choice
 1649  * is most specific found.
 1650  */
 1651 struct ifaddr *
 1652 ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum)
 1653 {
 1654         struct ifnet *ifp;
 1655         struct ifaddr *ifa;
 1656         struct ifaddr *ifa_maybe = NULL;
 1657         u_int af = addr->sa_family;
 1658         char *addr_data = addr->sa_data, *cplim;
 1659 
 1660         /*
 1661          * AF_LINK addresses can be looked up directly by their index number,
 1662          * so do that if we can.
 1663          */
 1664         if (af == AF_LINK) {
 1665             struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
 1666             if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
 1667                 return (ifaddr_byindex(sdl->sdl_index));
 1668         }
 1669 
 1670         /*
 1671          * Scan though each interface, looking for ones that have addresses
 1672          * in this address family and the requested fib.  Maintain a reference
 1673          * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
 1674          * kept it stable when we move onto the next interface.
 1675          */
 1676         IFNET_RLOCK_NOSLEEP();
 1677         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1678                 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
 1679                         continue;
 1680                 IF_ADDR_RLOCK(ifp);
 1681                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1682                         char *cp, *cp2, *cp3;
 1683 
 1684                         if (ifa->ifa_addr->sa_family != af)
 1685 next:                           continue;
 1686                         if (af == AF_INET && 
 1687                             ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
 1688                                 /*
 1689                                  * This is a bit broken as it doesn't
 1690                                  * take into account that the remote end may
 1691                                  * be a single node in the network we are
 1692                                  * looking for.
 1693                                  * The trouble is that we don't know the
 1694                                  * netmask for the remote end.
 1695                                  */
 1696                                 if (ifa->ifa_dstaddr != NULL &&
 1697                                     sa_equal(addr, ifa->ifa_dstaddr)) {
 1698                                         ifa_ref(ifa);
 1699                                         IF_ADDR_RUNLOCK(ifp);
 1700                                         goto done;
 1701                                 }
 1702                         } else {
 1703                                 /*
 1704                                  * if we have a special address handler,
 1705                                  * then use it instead of the generic one.
 1706                                  */
 1707                                 if (ifa->ifa_claim_addr) {
 1708                                         if ((*ifa->ifa_claim_addr)(ifa, addr)) {
 1709                                                 ifa_ref(ifa);
 1710                                                 IF_ADDR_RUNLOCK(ifp);
 1711                                                 goto done;
 1712                                         }
 1713                                         continue;
 1714                                 }
 1715 
 1716                                 /*
 1717                                  * Scan all the bits in the ifa's address.
 1718                                  * If a bit dissagrees with what we are
 1719                                  * looking for, mask it with the netmask
 1720                                  * to see if it really matters.
 1721                                  * (A byte at a time)
 1722                                  */
 1723                                 if (ifa->ifa_netmask == 0)
 1724                                         continue;
 1725                                 cp = addr_data;
 1726                                 cp2 = ifa->ifa_addr->sa_data;
 1727                                 cp3 = ifa->ifa_netmask->sa_data;
 1728                                 cplim = ifa->ifa_netmask->sa_len
 1729                                         + (char *)ifa->ifa_netmask;
 1730                                 while (cp3 < cplim)
 1731                                         if ((*cp++ ^ *cp2++) & *cp3++)
 1732                                                 goto next; /* next address! */
 1733                                 /*
 1734                                  * If the netmask of what we just found
 1735                                  * is more specific than what we had before
 1736                                  * (if we had one), or if the virtual status
 1737                                  * of new prefix is better than of the old one,
 1738                                  * then remember the new one before continuing
 1739                                  * to search for an even better one.
 1740                                  */
 1741                                 if (ifa_maybe == NULL ||
 1742                                     ifa_preferred(ifa_maybe, ifa) ||
 1743                                     rn_refines((caddr_t)ifa->ifa_netmask,
 1744                                     (caddr_t)ifa_maybe->ifa_netmask)) {
 1745                                         if (ifa_maybe != NULL)
 1746                                                 ifa_free(ifa_maybe);
 1747                                         ifa_maybe = ifa;
 1748                                         ifa_ref(ifa_maybe);
 1749                                 }
 1750                         }
 1751                 }
 1752                 IF_ADDR_RUNLOCK(ifp);
 1753         }
 1754         ifa = ifa_maybe;
 1755         ifa_maybe = NULL;
 1756 done:
 1757         IFNET_RUNLOCK_NOSLEEP();
 1758         if (ifa_maybe != NULL)
 1759                 ifa_free(ifa_maybe);
 1760         return (ifa);
 1761 }
 1762 
 1763 struct ifaddr *
 1764 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
 1765 {
 1766 
 1767         return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS));
 1768 }
 1769 
 1770 /*
 1771  * Find an interface address specific to an interface best matching
 1772  * a given address.
 1773  */
 1774 struct ifaddr *
 1775 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
 1776 {
 1777         struct ifaddr *ifa;
 1778         char *cp, *cp2, *cp3;
 1779         char *cplim;
 1780         struct ifaddr *ifa_maybe = NULL;
 1781         u_int af = addr->sa_family;
 1782 
 1783         if (af >= AF_MAX)
 1784                 return (NULL);
 1785         IF_ADDR_RLOCK(ifp);
 1786         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1787                 if (ifa->ifa_addr->sa_family != af)
 1788                         continue;
 1789                 if (ifa_maybe == NULL)
 1790                         ifa_maybe = ifa;
 1791                 if (ifa->ifa_netmask == 0) {
 1792                         if (sa_equal(addr, ifa->ifa_addr) ||
 1793                             (ifa->ifa_dstaddr &&
 1794                             sa_equal(addr, ifa->ifa_dstaddr)))
 1795                                 goto done;
 1796                         continue;
 1797                 }
 1798                 if (ifp->if_flags & IFF_POINTOPOINT) {
 1799                         if (sa_equal(addr, ifa->ifa_dstaddr))
 1800                                 goto done;
 1801                 } else {
 1802                         cp = addr->sa_data;
 1803                         cp2 = ifa->ifa_addr->sa_data;
 1804                         cp3 = ifa->ifa_netmask->sa_data;
 1805                         cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
 1806                         for (; cp3 < cplim; cp3++)
 1807                                 if ((*cp++ ^ *cp2++) & *cp3)
 1808                                         break;
 1809                         if (cp3 == cplim)
 1810                                 goto done;
 1811                 }
 1812         }
 1813         ifa = ifa_maybe;
 1814 done:
 1815         if (ifa != NULL)
 1816                 ifa_ref(ifa);
 1817         IF_ADDR_RUNLOCK(ifp);
 1818         return (ifa);
 1819 }
 1820 
 1821 /*
 1822  * See whether new ifa is better than current one:
 1823  * 1) A non-virtual one is preferred over virtual.
 1824  * 2) A virtual in master state preferred over any other state.
 1825  *
 1826  * Used in several address selecting functions.
 1827  */
 1828 int
 1829 ifa_preferred(struct ifaddr *cur, struct ifaddr *next)
 1830 {
 1831 
 1832         return (cur->ifa_carp && (!next->ifa_carp ||
 1833             ((*carp_master_p)(next) && !(*carp_master_p)(cur))));
 1834 }
 1835 
 1836 #include <net/if_llatbl.h>
 1837 
 1838 /*
 1839  * Default action when installing a route with a Link Level gateway.
 1840  * Lookup an appropriate real ifa to point to.
 1841  * This should be moved to /sys/net/link.c eventually.
 1842  */
 1843 static void
 1844 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
 1845 {
 1846         struct ifaddr *ifa, *oifa;
 1847         struct sockaddr *dst;
 1848         struct ifnet *ifp;
 1849 
 1850         RT_LOCK_ASSERT(rt);
 1851 
 1852         if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
 1853             ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
 1854                 return;
 1855         ifa = ifaof_ifpforaddr(dst, ifp);
 1856         if (ifa) {
 1857                 oifa = rt->rt_ifa;
 1858                 rt->rt_ifa = ifa;
 1859                 ifa_free(oifa);
 1860                 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
 1861                         ifa->ifa_rtrequest(cmd, rt, info);
 1862         }
 1863 }
 1864 
 1865 /*
 1866  * Mark an interface down and notify protocols of
 1867  * the transition.
 1868  */
 1869 static void
 1870 if_unroute(struct ifnet *ifp, int flag, int fam)
 1871 {
 1872         struct ifaddr *ifa;
 1873 
 1874         KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
 1875 
 1876         ifp->if_flags &= ~flag;
 1877         getmicrotime(&ifp->if_lastchange);
 1878         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1879                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1880                         pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
 1881         ifp->if_qflush(ifp);
 1882 
 1883         if (ifp->if_carp)
 1884                 (*carp_linkstate_p)(ifp);
 1885         rt_ifmsg(ifp);
 1886 }
 1887 
 1888 /*
 1889  * Mark an interface up and notify protocols of
 1890  * the transition.
 1891  */
 1892 static void
 1893 if_route(struct ifnet *ifp, int flag, int fam)
 1894 {
 1895         struct ifaddr *ifa;
 1896 
 1897         KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
 1898 
 1899         ifp->if_flags |= flag;
 1900         getmicrotime(&ifp->if_lastchange);
 1901         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1902                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1903                         pfctlinput(PRC_IFUP, ifa->ifa_addr);
 1904         if (ifp->if_carp)
 1905                 (*carp_linkstate_p)(ifp);
 1906         rt_ifmsg(ifp);
 1907 #ifdef INET6
 1908         in6_if_up(ifp);
 1909 #endif
 1910 }
 1911 
 1912 void    (*vlan_link_state_p)(struct ifnet *);   /* XXX: private from if_vlan */
 1913 void    (*vlan_trunk_cap_p)(struct ifnet *);            /* XXX: private from if_vlan */
 1914 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
 1915 struct  ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
 1916 int     (*vlan_tag_p)(struct ifnet *, uint16_t *);
 1917 int     (*vlan_setcookie_p)(struct ifnet *, void *);
 1918 void    *(*vlan_cookie_p)(struct ifnet *);
 1919 
 1920 /*
 1921  * Handle a change in the interface link state. To avoid LORs
 1922  * between driver lock and upper layer locks, as well as possible
 1923  * recursions, we post event to taskqueue, and all job
 1924  * is done in static do_link_state_change().
 1925  */
 1926 void
 1927 if_link_state_change(struct ifnet *ifp, int link_state)
 1928 {
 1929         /* Return if state hasn't changed. */
 1930         if (ifp->if_link_state == link_state)
 1931                 return;
 1932 
 1933         ifp->if_link_state = link_state;
 1934 
 1935         taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
 1936 }
 1937 
 1938 static void
 1939 do_link_state_change(void *arg, int pending)
 1940 {
 1941         struct ifnet *ifp = (struct ifnet *)arg;
 1942         int link_state = ifp->if_link_state;
 1943         CURVNET_SET(ifp->if_vnet);
 1944 
 1945         /* Notify that the link state has changed. */
 1946         rt_ifmsg(ifp);
 1947         if (ifp->if_vlantrunk != NULL)
 1948                 (*vlan_link_state_p)(ifp);
 1949 
 1950         if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
 1951             IFP2AC(ifp)->ac_netgraph != NULL)
 1952                 (*ng_ether_link_state_p)(ifp, link_state);
 1953         if (ifp->if_carp)
 1954                 (*carp_linkstate_p)(ifp);
 1955         if (ifp->if_bridge)
 1956                 (*bridge_linkstate_p)(ifp);
 1957         if (ifp->if_lagg)
 1958                 (*lagg_linkstate_p)(ifp, link_state);
 1959 
 1960         if (IS_DEFAULT_VNET(curvnet))
 1961                 devctl_notify("IFNET", ifp->if_xname,
 1962                     (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
 1963                     NULL);
 1964         if (pending > 1)
 1965                 if_printf(ifp, "%d link states coalesced\n", pending);
 1966         if (log_link_state_change)
 1967                 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
 1968                     (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
 1969         EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
 1970         CURVNET_RESTORE();
 1971 }
 1972 
 1973 /*
 1974  * Mark an interface down and notify protocols of
 1975  * the transition.
 1976  */
 1977 void
 1978 if_down(struct ifnet *ifp)
 1979 {
 1980 
 1981         if_unroute(ifp, IFF_UP, AF_UNSPEC);
 1982 }
 1983 
 1984 /*
 1985  * Mark an interface up and notify protocols of
 1986  * the transition.
 1987  */
 1988 void
 1989 if_up(struct ifnet *ifp)
 1990 {
 1991 
 1992         if_route(ifp, IFF_UP, AF_UNSPEC);
 1993 }
 1994 
 1995 /*
 1996  * Flush an interface queue.
 1997  */
 1998 void
 1999 if_qflush(struct ifnet *ifp)
 2000 {
 2001         struct mbuf *m, *n;
 2002         struct ifaltq *ifq;
 2003         
 2004         ifq = &ifp->if_snd;
 2005         IFQ_LOCK(ifq);
 2006 #ifdef ALTQ
 2007         if (ALTQ_IS_ENABLED(ifq))
 2008                 ALTQ_PURGE(ifq);
 2009 #endif
 2010         n = ifq->ifq_head;
 2011         while ((m = n) != 0) {
 2012                 n = m->m_nextpkt;
 2013                 m_freem(m);
 2014         }
 2015         ifq->ifq_head = 0;
 2016         ifq->ifq_tail = 0;
 2017         ifq->ifq_len = 0;
 2018         IFQ_UNLOCK(ifq);
 2019 }
 2020 
 2021 /*
 2022  * Map interface name to interface structure pointer, with or without
 2023  * returning a reference.
 2024  */
 2025 struct ifnet *
 2026 ifunit_ref(const char *name)
 2027 {
 2028         struct ifnet *ifp;
 2029 
 2030         IFNET_RLOCK_NOSLEEP();
 2031         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2032                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
 2033                     !(ifp->if_flags & IFF_DYING))
 2034                         break;
 2035         }
 2036         if (ifp != NULL)
 2037                 if_ref(ifp);
 2038         IFNET_RUNLOCK_NOSLEEP();
 2039         return (ifp);
 2040 }
 2041 
 2042 struct ifnet *
 2043 ifunit(const char *name)
 2044 {
 2045         struct ifnet *ifp;
 2046 
 2047         IFNET_RLOCK_NOSLEEP();
 2048         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2049                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
 2050                         break;
 2051         }
 2052         IFNET_RUNLOCK_NOSLEEP();
 2053         return (ifp);
 2054 }
 2055 
 2056 /*
 2057  * Hardware specific interface ioctls.
 2058  */
 2059 static int
 2060 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
 2061 {
 2062         struct ifreq *ifr;
 2063         struct ifstat *ifs;
 2064         int error = 0;
 2065         int new_flags, temp_flags;
 2066         size_t namelen, onamelen;
 2067         size_t descrlen;
 2068         char *descrbuf, *odescrbuf;
 2069         char new_name[IFNAMSIZ];
 2070         struct ifaddr *ifa;
 2071         struct sockaddr_dl *sdl;
 2072 
 2073         ifr = (struct ifreq *)data;
 2074         switch (cmd) {
 2075         case SIOCGIFINDEX:
 2076                 ifr->ifr_index = ifp->if_index;
 2077                 break;
 2078 
 2079         case SIOCGIFFLAGS:
 2080                 temp_flags = ifp->if_flags | ifp->if_drv_flags;
 2081                 ifr->ifr_flags = temp_flags & 0xffff;
 2082                 ifr->ifr_flagshigh = temp_flags >> 16;
 2083                 break;
 2084 
 2085         case SIOCGIFCAP:
 2086                 ifr->ifr_reqcap = ifp->if_capabilities;
 2087                 ifr->ifr_curcap = ifp->if_capenable;
 2088                 break;
 2089 
 2090 #ifdef MAC
 2091         case SIOCGIFMAC:
 2092                 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
 2093                 break;
 2094 #endif
 2095 
 2096         case SIOCGIFMETRIC:
 2097                 ifr->ifr_metric = ifp->if_metric;
 2098                 break;
 2099 
 2100         case SIOCGIFMTU:
 2101                 ifr->ifr_mtu = ifp->if_mtu;
 2102                 break;
 2103 
 2104         case SIOCGIFPHYS:
 2105                 ifr->ifr_phys = ifp->if_physical;
 2106                 break;
 2107 
 2108         case SIOCGIFDESCR:
 2109                 error = 0;
 2110                 sx_slock(&ifdescr_sx);
 2111                 if (ifp->if_description == NULL)
 2112                         error = ENOMSG;
 2113                 else {
 2114                         /* space for terminating nul */
 2115                         descrlen = strlen(ifp->if_description) + 1;
 2116                         if (ifr->ifr_buffer.length < descrlen)
 2117                                 ifr->ifr_buffer.buffer = NULL;
 2118                         else
 2119                                 error = copyout(ifp->if_description,
 2120                                     ifr->ifr_buffer.buffer, descrlen);
 2121                         ifr->ifr_buffer.length = descrlen;
 2122                 }
 2123                 sx_sunlock(&ifdescr_sx);
 2124                 break;
 2125 
 2126         case SIOCSIFDESCR:
 2127                 error = priv_check(td, PRIV_NET_SETIFDESCR);
 2128                 if (error)
 2129                         return (error);
 2130 
 2131                 /*
 2132                  * Copy only (length-1) bytes to make sure that
 2133                  * if_description is always nul terminated.  The
 2134                  * length parameter is supposed to count the
 2135                  * terminating nul in.
 2136                  */
 2137                 if (ifr->ifr_buffer.length > ifdescr_maxlen)
 2138                         return (ENAMETOOLONG);
 2139                 else if (ifr->ifr_buffer.length == 0)
 2140                         descrbuf = NULL;
 2141                 else {
 2142                         descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
 2143                             M_WAITOK | M_ZERO);
 2144                         error = copyin(ifr->ifr_buffer.buffer, descrbuf,
 2145                             ifr->ifr_buffer.length - 1);
 2146                         if (error) {
 2147                                 free(descrbuf, M_IFDESCR);
 2148                                 break;
 2149                         }
 2150                 }
 2151 
 2152                 sx_xlock(&ifdescr_sx);
 2153                 odescrbuf = ifp->if_description;
 2154                 ifp->if_description = descrbuf;
 2155                 sx_xunlock(&ifdescr_sx);
 2156 
 2157                 getmicrotime(&ifp->if_lastchange);
 2158                 free(odescrbuf, M_IFDESCR);
 2159                 break;
 2160 
 2161         case SIOCGIFFIB:
 2162                 ifr->ifr_fib = ifp->if_fib;
 2163                 break;
 2164 
 2165         case SIOCSIFFIB:
 2166                 error = priv_check(td, PRIV_NET_SETIFFIB);
 2167                 if (error)
 2168                         return (error);
 2169                 if (ifr->ifr_fib >= rt_numfibs)
 2170                         return (EINVAL);
 2171 
 2172                 ifp->if_fib = ifr->ifr_fib;
 2173                 break;
 2174 
 2175         case SIOCSIFFLAGS:
 2176                 error = priv_check(td, PRIV_NET_SETIFFLAGS);
 2177                 if (error)
 2178                         return (error);
 2179                 /*
 2180                  * Currently, no driver owned flags pass the IFF_CANTCHANGE
 2181                  * check, so we don't need special handling here yet.
 2182                  */
 2183                 new_flags = (ifr->ifr_flags & 0xffff) |
 2184                     (ifr->ifr_flagshigh << 16);
 2185                 if (ifp->if_flags & IFF_SMART) {
 2186                         /* Smart drivers twiddle their own routes */
 2187                 } else if (ifp->if_flags & IFF_UP &&
 2188                     (new_flags & IFF_UP) == 0) {
 2189                         if_down(ifp);
 2190                 } else if (new_flags & IFF_UP &&
 2191                     (ifp->if_flags & IFF_UP) == 0) {
 2192                         if_up(ifp);
 2193                 }
 2194                 /* See if permanently promiscuous mode bit is about to flip */
 2195                 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
 2196                         if (new_flags & IFF_PPROMISC)
 2197                                 ifp->if_flags |= IFF_PROMISC;
 2198                         else if (ifp->if_pcount == 0)
 2199                                 ifp->if_flags &= ~IFF_PROMISC;
 2200                         log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
 2201                             ifp->if_xname,
 2202                             (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
 2203                 }
 2204                 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
 2205                         (new_flags &~ IFF_CANTCHANGE);
 2206                 if (ifp->if_ioctl) {
 2207                         (void) (*ifp->if_ioctl)(ifp, cmd, data);
 2208                 }
 2209                 getmicrotime(&ifp->if_lastchange);
 2210                 break;
 2211 
 2212         case SIOCSIFCAP:
 2213                 error = priv_check(td, PRIV_NET_SETIFCAP);
 2214                 if (error)
 2215                         return (error);
 2216                 if (ifp->if_ioctl == NULL)
 2217                         return (EOPNOTSUPP);
 2218                 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
 2219                         return (EINVAL);
 2220                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2221                 if (error == 0)
 2222                         getmicrotime(&ifp->if_lastchange);
 2223                 break;
 2224 
 2225 #ifdef MAC
 2226         case SIOCSIFMAC:
 2227                 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
 2228                 break;
 2229 #endif
 2230 
 2231         case SIOCSIFNAME:
 2232                 error = priv_check(td, PRIV_NET_SETIFNAME);
 2233                 if (error)
 2234                         return (error);
 2235                 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
 2236                 if (error != 0)
 2237                         return (error);
 2238                 if (new_name[0] == '\0')
 2239                         return (EINVAL);
 2240                 if (ifunit(new_name) != NULL)
 2241                         return (EEXIST);
 2242 
 2243                 /*
 2244                  * XXX: Locking.  Nothing else seems to lock if_flags,
 2245                  * and there are numerous other races with the
 2246                  * ifunit() checks not being atomic with namespace
 2247                  * changes (renames, vmoves, if_attach, etc).
 2248                  */
 2249                 ifp->if_flags |= IFF_RENAMING;
 2250                 
 2251                 /* Announce the departure of the interface. */
 2252                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 2253                 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 2254 
 2255                 log(LOG_INFO, "%s: changing name to '%s'\n",
 2256                     ifp->if_xname, new_name);
 2257 
 2258                 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
 2259                 ifa = ifp->if_addr;
 2260                 IFA_LOCK(ifa);
 2261                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 2262                 namelen = strlen(new_name);
 2263                 onamelen = sdl->sdl_nlen;
 2264                 /*
 2265                  * Move the address if needed.  This is safe because we
 2266                  * allocate space for a name of length IFNAMSIZ when we
 2267                  * create this in if_attach().
 2268                  */
 2269                 if (namelen != onamelen) {
 2270                         bcopy(sdl->sdl_data + onamelen,
 2271                             sdl->sdl_data + namelen, sdl->sdl_alen);
 2272                 }
 2273                 bcopy(new_name, sdl->sdl_data, namelen);
 2274                 sdl->sdl_nlen = namelen;
 2275                 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
 2276                 bzero(sdl->sdl_data, onamelen);
 2277                 while (namelen != 0)
 2278                         sdl->sdl_data[--namelen] = 0xff;
 2279                 IFA_UNLOCK(ifa);
 2280 
 2281                 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
 2282                 /* Announce the return of the interface. */
 2283                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 2284 
 2285                 ifp->if_flags &= ~IFF_RENAMING;
 2286                 break;
 2287 
 2288 #ifdef VIMAGE
 2289         case SIOCSIFVNET:
 2290                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2291                 if (error)
 2292                         return (error);
 2293                 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
 2294                 break;
 2295 #endif
 2296 
 2297         case SIOCSIFMETRIC:
 2298                 error = priv_check(td, PRIV_NET_SETIFMETRIC);
 2299                 if (error)
 2300                         return (error);
 2301                 ifp->if_metric = ifr->ifr_metric;
 2302                 getmicrotime(&ifp->if_lastchange);
 2303                 break;
 2304 
 2305         case SIOCSIFPHYS:
 2306                 error = priv_check(td, PRIV_NET_SETIFPHYS);
 2307                 if (error)
 2308                         return (error);
 2309                 if (ifp->if_ioctl == NULL)
 2310                         return (EOPNOTSUPP);
 2311                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2312                 if (error == 0)
 2313                         getmicrotime(&ifp->if_lastchange);
 2314                 break;
 2315 
 2316         case SIOCSIFMTU:
 2317         {
 2318                 u_long oldmtu = ifp->if_mtu;
 2319 
 2320                 error = priv_check(td, PRIV_NET_SETIFMTU);
 2321                 if (error)
 2322                         return (error);
 2323                 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
 2324                         return (EINVAL);
 2325                 if (ifp->if_ioctl == NULL)
 2326                         return (EOPNOTSUPP);
 2327                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2328                 if (error == 0) {
 2329                         getmicrotime(&ifp->if_lastchange);
 2330                         rt_ifmsg(ifp);
 2331                 }
 2332                 /*
 2333                  * If the link MTU changed, do network layer specific procedure.
 2334                  */
 2335                 if (ifp->if_mtu != oldmtu) {
 2336 #ifdef INET6
 2337                         nd6_setmtu(ifp);
 2338 #endif
 2339                 }
 2340                 break;
 2341         }
 2342 
 2343         case SIOCADDMULTI:
 2344         case SIOCDELMULTI:
 2345                 if (cmd == SIOCADDMULTI)
 2346                         error = priv_check(td, PRIV_NET_ADDMULTI);
 2347                 else
 2348                         error = priv_check(td, PRIV_NET_DELMULTI);
 2349                 if (error)
 2350                         return (error);
 2351 
 2352                 /* Don't allow group membership on non-multicast interfaces. */
 2353                 if ((ifp->if_flags & IFF_MULTICAST) == 0)
 2354                         return (EOPNOTSUPP);
 2355 
 2356                 /* Don't let users screw up protocols' entries. */
 2357                 if (ifr->ifr_addr.sa_family != AF_LINK)
 2358                         return (EINVAL);
 2359 
 2360                 if (cmd == SIOCADDMULTI) {
 2361                         struct ifmultiaddr *ifma;
 2362 
 2363                         /*
 2364                          * Userland is only permitted to join groups once
 2365                          * via the if_addmulti() KPI, because it cannot hold
 2366                          * struct ifmultiaddr * between calls. It may also
 2367                          * lose a race while we check if the membership
 2368                          * already exists.
 2369                          */
 2370                         IF_ADDR_RLOCK(ifp);
 2371                         ifma = if_findmulti(ifp, &ifr->ifr_addr);
 2372                         IF_ADDR_RUNLOCK(ifp);
 2373                         if (ifma != NULL)
 2374                                 error = EADDRINUSE;
 2375                         else
 2376                                 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
 2377                 } else {
 2378                         error = if_delmulti(ifp, &ifr->ifr_addr);
 2379                 }
 2380                 if (error == 0)
 2381                         getmicrotime(&ifp->if_lastchange);
 2382                 break;
 2383 
 2384         case SIOCSIFPHYADDR:
 2385         case SIOCDIFPHYADDR:
 2386 #ifdef INET6
 2387         case SIOCSIFPHYADDR_IN6:
 2388 #endif
 2389         case SIOCSLIFPHYADDR:
 2390         case SIOCSIFMEDIA:
 2391         case SIOCSIFGENERIC:
 2392                 error = priv_check(td, PRIV_NET_HWIOCTL);
 2393                 if (error)
 2394                         return (error);
 2395                 if (ifp->if_ioctl == NULL)
 2396                         return (EOPNOTSUPP);
 2397                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2398                 if (error == 0)
 2399                         getmicrotime(&ifp->if_lastchange);
 2400                 break;
 2401 
 2402         case SIOCGIFSTATUS:
 2403                 ifs = (struct ifstat *)data;
 2404                 ifs->ascii[0] = '\0';
 2405 
 2406         case SIOCGIFPSRCADDR:
 2407         case SIOCGIFPDSTADDR:
 2408         case SIOCGLIFPHYADDR:
 2409         case SIOCGIFMEDIA:
 2410         case SIOCGIFGENERIC:
 2411                 if (ifp->if_ioctl == NULL)
 2412                         return (EOPNOTSUPP);
 2413                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2414                 break;
 2415 
 2416         case SIOCSIFLLADDR:
 2417                 error = priv_check(td, PRIV_NET_SETLLADDR);
 2418                 if (error)
 2419                         return (error);
 2420                 error = if_setlladdr(ifp,
 2421                     ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
 2422                 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
 2423                 break;
 2424 
 2425         case SIOCAIFGROUP:
 2426         {
 2427                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2428 
 2429                 error = priv_check(td, PRIV_NET_ADDIFGROUP);
 2430                 if (error)
 2431                         return (error);
 2432                 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
 2433                         return (error);
 2434                 break;
 2435         }
 2436 
 2437         case SIOCGIFGROUP:
 2438                 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
 2439                         return (error);
 2440                 break;
 2441 
 2442         case SIOCDIFGROUP:
 2443         {
 2444                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2445 
 2446                 error = priv_check(td, PRIV_NET_DELIFGROUP);
 2447                 if (error)
 2448                         return (error);
 2449                 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
 2450                         return (error);
 2451                 break;
 2452         }
 2453 
 2454         default:
 2455                 error = ENOIOCTL;
 2456                 break;
 2457         }
 2458         return (error);
 2459 }
 2460 
 2461 #ifdef COMPAT_FREEBSD32
 2462 struct ifconf32 {
 2463         int32_t ifc_len;
 2464         union {
 2465                 uint32_t        ifcu_buf;
 2466                 uint32_t        ifcu_req;
 2467         } ifc_ifcu;
 2468 };
 2469 #define SIOCGIFCONF32   _IOWR('i', 36, struct ifconf32)
 2470 #endif
 2471 
 2472 /*
 2473  * Interface ioctls.
 2474  */
 2475 int
 2476 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
 2477 {
 2478         struct ifnet *ifp;
 2479         struct ifreq *ifr;
 2480         int error;
 2481         int oif_flags;
 2482 
 2483         CURVNET_SET(so->so_vnet);
 2484         switch (cmd) {
 2485         case SIOCGIFCONF:
 2486         case OSIOCGIFCONF:
 2487                 error = ifconf(cmd, data);
 2488                 CURVNET_RESTORE();
 2489                 return (error);
 2490 
 2491 #ifdef COMPAT_FREEBSD32
 2492         case SIOCGIFCONF32:
 2493                 {
 2494                         struct ifconf32 *ifc32;
 2495                         struct ifconf ifc;
 2496 
 2497                         ifc32 = (struct ifconf32 *)data;
 2498                         ifc.ifc_len = ifc32->ifc_len;
 2499                         ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
 2500 
 2501                         error = ifconf(SIOCGIFCONF, (void *)&ifc);
 2502                         CURVNET_RESTORE();
 2503                         if (error == 0)
 2504                                 ifc32->ifc_len = ifc.ifc_len;
 2505                         return (error);
 2506                 }
 2507 #endif
 2508         }
 2509         ifr = (struct ifreq *)data;
 2510 
 2511         switch (cmd) {
 2512 #ifdef VIMAGE
 2513         case SIOCSIFRVNET:
 2514                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2515                 if (error == 0)
 2516                         error = if_vmove_reclaim(td, ifr->ifr_name,
 2517                             ifr->ifr_jid);
 2518                 CURVNET_RESTORE();
 2519                 return (error);
 2520 #endif
 2521         case SIOCIFCREATE:
 2522         case SIOCIFCREATE2:
 2523                 error = priv_check(td, PRIV_NET_IFCREATE);
 2524                 if (error == 0)
 2525                         error = if_clone_create(ifr->ifr_name,
 2526                             sizeof(ifr->ifr_name),
 2527                             cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
 2528                 CURVNET_RESTORE();
 2529                 return (error);
 2530         case SIOCIFDESTROY:
 2531                 error = priv_check(td, PRIV_NET_IFDESTROY);
 2532                 if (error == 0)
 2533                         error = if_clone_destroy(ifr->ifr_name);
 2534                 CURVNET_RESTORE();
 2535                 return (error);
 2536 
 2537         case SIOCIFGCLONERS:
 2538                 error = if_clone_list((struct if_clonereq *)data);
 2539                 CURVNET_RESTORE();
 2540                 return (error);
 2541         case SIOCGIFGMEMB:
 2542                 error = if_getgroupmembers((struct ifgroupreq *)data);
 2543                 CURVNET_RESTORE();
 2544                 return (error);
 2545 #if defined(INET) || defined(INET6)
 2546         case SIOCSVH:
 2547         case SIOCGVH:
 2548                 if (carp_ioctl_p == NULL)
 2549                         error = EPROTONOSUPPORT;
 2550                 else
 2551                         error = (*carp_ioctl_p)(ifr, cmd, td);
 2552                 CURVNET_RESTORE();
 2553                 return (error);
 2554 #endif
 2555         }
 2556 
 2557         ifp = ifunit_ref(ifr->ifr_name);
 2558         if (ifp == NULL) {
 2559                 CURVNET_RESTORE();
 2560                 return (ENXIO);
 2561         }
 2562 
 2563         error = ifhwioctl(cmd, ifp, data, td);
 2564         if (error != ENOIOCTL) {
 2565                 if_rele(ifp);
 2566                 CURVNET_RESTORE();
 2567                 return (error);
 2568         }
 2569 
 2570         oif_flags = ifp->if_flags;
 2571         if (so->so_proto == NULL) {
 2572                 if_rele(ifp);
 2573                 CURVNET_RESTORE();
 2574                 return (EOPNOTSUPP);
 2575         }
 2576 
 2577         /*
 2578          * Pass the request on to the socket control method, and if the
 2579          * latter returns EOPNOTSUPP, directly to the interface.
 2580          *
 2581          * Make an exception for the legacy SIOCSIF* requests.  Drivers
 2582          * trust SIOCSIFADDR et al to come from an already privileged
 2583          * layer, and do not perform any credentials checks or input
 2584          * validation.
 2585          */
 2586 #ifndef COMPAT_43
 2587         error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
 2588                                                                  data,
 2589                                                                  ifp, td));
 2590         if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
 2591             cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
 2592             cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
 2593                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2594 #else
 2595         {
 2596                 u_long ocmd = cmd;
 2597 
 2598                 switch (cmd) {
 2599 
 2600                 case SIOCSIFDSTADDR:
 2601                 case SIOCSIFADDR:
 2602                 case SIOCSIFBRDADDR:
 2603                 case SIOCSIFNETMASK:
 2604 #if BYTE_ORDER != BIG_ENDIAN
 2605                         if (ifr->ifr_addr.sa_family == 0 &&
 2606                             ifr->ifr_addr.sa_len < 16) {
 2607                                 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
 2608                                 ifr->ifr_addr.sa_len = 16;
 2609                         }
 2610 #else
 2611                         if (ifr->ifr_addr.sa_len == 0)
 2612                                 ifr->ifr_addr.sa_len = 16;
 2613 #endif
 2614                         break;
 2615 
 2616                 case OSIOCGIFADDR:
 2617                         cmd = SIOCGIFADDR;
 2618                         break;
 2619 
 2620                 case OSIOCGIFDSTADDR:
 2621                         cmd = SIOCGIFDSTADDR;
 2622                         break;
 2623 
 2624                 case OSIOCGIFBRDADDR:
 2625                         cmd = SIOCGIFBRDADDR;
 2626                         break;
 2627 
 2628                 case OSIOCGIFNETMASK:
 2629                         cmd = SIOCGIFNETMASK;
 2630                 }
 2631                 error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
 2632                                                                    cmd,
 2633                                                                    data,
 2634                                                                    ifp, td));
 2635                 if (error == EOPNOTSUPP && ifp != NULL &&
 2636                     ifp->if_ioctl != NULL &&
 2637                     cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
 2638                     cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
 2639                         error = (*ifp->if_ioctl)(ifp, cmd, data);
 2640                 switch (ocmd) {
 2641 
 2642                 case OSIOCGIFADDR:
 2643                 case OSIOCGIFDSTADDR:
 2644                 case OSIOCGIFBRDADDR:
 2645                 case OSIOCGIFNETMASK:
 2646                         *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
 2647 
 2648                 }
 2649         }
 2650 #endif /* COMPAT_43 */
 2651 
 2652         if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
 2653 #ifdef INET6
 2654                 if (ifp->if_flags & IFF_UP)
 2655                         in6_if_up(ifp);
 2656 #endif
 2657         }
 2658         if_rele(ifp);
 2659         CURVNET_RESTORE();
 2660         return (error);
 2661 }
 2662 
 2663 /*
 2664  * The code common to handling reference counted flags,
 2665  * e.g., in ifpromisc() and if_allmulti().
 2666  * The "pflag" argument can specify a permanent mode flag to check,
 2667  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
 2668  *
 2669  * Only to be used on stack-owned flags, not driver-owned flags.
 2670  */
 2671 static int
 2672 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
 2673 {
 2674         struct ifreq ifr;
 2675         int error;
 2676         int oldflags, oldcount;
 2677 
 2678         /* Sanity checks to catch programming errors */
 2679         KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
 2680             ("%s: setting driver-owned flag %d", __func__, flag));
 2681 
 2682         if (onswitch)
 2683                 KASSERT(*refcount >= 0,
 2684                     ("%s: increment negative refcount %d for flag %d",
 2685                     __func__, *refcount, flag));
 2686         else
 2687                 KASSERT(*refcount > 0,
 2688                     ("%s: decrement non-positive refcount %d for flag %d",
 2689                     __func__, *refcount, flag));
 2690 
 2691         /* In case this mode is permanent, just touch refcount */
 2692         if (ifp->if_flags & pflag) {
 2693                 *refcount += onswitch ? 1 : -1;
 2694                 return (0);
 2695         }
 2696 
 2697         /* Save ifnet parameters for if_ioctl() may fail */
 2698         oldcount = *refcount;
 2699         oldflags = ifp->if_flags;
 2700         
 2701         /*
 2702          * See if we aren't the only and touching refcount is enough.
 2703          * Actually toggle interface flag if we are the first or last.
 2704          */
 2705         if (onswitch) {
 2706                 if ((*refcount)++)
 2707                         return (0);
 2708                 ifp->if_flags |= flag;
 2709         } else {
 2710                 if (--(*refcount))
 2711                         return (0);
 2712                 ifp->if_flags &= ~flag;
 2713         }
 2714 
 2715         /* Call down the driver since we've changed interface flags */
 2716         if (ifp->if_ioctl == NULL) {
 2717                 error = EOPNOTSUPP;
 2718                 goto recover;
 2719         }
 2720         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2721         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2722         error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2723         if (error)
 2724                 goto recover;
 2725         /* Notify userland that interface flags have changed */
 2726         rt_ifmsg(ifp);
 2727         return (0);
 2728 
 2729 recover:
 2730         /* Recover after driver error */
 2731         *refcount = oldcount;
 2732         ifp->if_flags = oldflags;
 2733         return (error);
 2734 }
 2735 
 2736 /*
 2737  * Set/clear promiscuous mode on interface ifp based on the truth value
 2738  * of pswitch.  The calls are reference counted so that only the first
 2739  * "on" request actually has an effect, as does the final "off" request.
 2740  * Results are undefined if the "off" and "on" requests are not matched.
 2741  */
 2742 int
 2743 ifpromisc(struct ifnet *ifp, int pswitch)
 2744 {
 2745         int error;
 2746         int oldflags = ifp->if_flags;
 2747 
 2748         error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
 2749                            &ifp->if_pcount, pswitch);
 2750         /* If promiscuous mode status has changed, log a message */
 2751         if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
 2752                 log(LOG_INFO, "%s: promiscuous mode %s\n",
 2753                     ifp->if_xname,
 2754                     (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
 2755         return (error);
 2756 }
 2757 
 2758 /*
 2759  * Return interface configuration
 2760  * of system.  List may be used
 2761  * in later ioctl's (above) to get
 2762  * other information.
 2763  */
 2764 /*ARGSUSED*/
 2765 static int
 2766 ifconf(u_long cmd, caddr_t data)
 2767 {
 2768         struct ifconf *ifc = (struct ifconf *)data;
 2769         struct ifnet *ifp;
 2770         struct ifaddr *ifa;
 2771         struct ifreq ifr;
 2772         struct sbuf *sb;
 2773         int error, full = 0, valid_len, max_len;
 2774 
 2775         /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
 2776         max_len = MAXPHYS - 1;
 2777 
 2778         /* Prevent hostile input from being able to crash the system */
 2779         if (ifc->ifc_len <= 0)
 2780                 return (EINVAL);
 2781 
 2782 again:
 2783         if (ifc->ifc_len <= max_len) {
 2784                 max_len = ifc->ifc_len;
 2785                 full = 1;
 2786         }
 2787         sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
 2788         max_len = 0;
 2789         valid_len = 0;
 2790 
 2791         IFNET_RLOCK();
 2792         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2793                 int addrs;
 2794 
 2795                 /*
 2796                  * Zero the ifr_name buffer to make sure we don't
 2797                  * disclose the contents of the stack.
 2798                  */
 2799                 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
 2800 
 2801                 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
 2802                     >= sizeof(ifr.ifr_name)) {
 2803                         sbuf_delete(sb);
 2804                         IFNET_RUNLOCK();
 2805                         return (ENAMETOOLONG);
 2806                 }
 2807 
 2808                 addrs = 0;
 2809                 IF_ADDR_RLOCK(ifp);
 2810                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2811                         struct sockaddr *sa = ifa->ifa_addr;
 2812 
 2813                         if (prison_if(curthread->td_ucred, sa) != 0)
 2814                                 continue;
 2815                         addrs++;
 2816 #ifdef COMPAT_43
 2817                         if (cmd == OSIOCGIFCONF) {
 2818                                 struct osockaddr *osa =
 2819                                          (struct osockaddr *)&ifr.ifr_addr;
 2820                                 ifr.ifr_addr = *sa;
 2821                                 osa->sa_family = sa->sa_family;
 2822                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2823                                 max_len += sizeof(ifr);
 2824                         } else
 2825 #endif
 2826                         if (sa->sa_len <= sizeof(*sa)) {
 2827                                 ifr.ifr_addr = *sa;
 2828                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2829                                 max_len += sizeof(ifr);
 2830                         } else {
 2831                                 sbuf_bcat(sb, &ifr,
 2832                                     offsetof(struct ifreq, ifr_addr));
 2833                                 max_len += offsetof(struct ifreq, ifr_addr);
 2834                                 sbuf_bcat(sb, sa, sa->sa_len);
 2835                                 max_len += sa->sa_len;
 2836                         }
 2837 
 2838                         if (sbuf_error(sb) == 0)
 2839                                 valid_len = sbuf_len(sb);
 2840                 }
 2841                 IF_ADDR_RUNLOCK(ifp);
 2842                 if (addrs == 0) {
 2843                         bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
 2844                         sbuf_bcat(sb, &ifr, sizeof(ifr));
 2845                         max_len += sizeof(ifr);
 2846 
 2847                         if (sbuf_error(sb) == 0)
 2848                                 valid_len = sbuf_len(sb);
 2849                 }
 2850         }
 2851         IFNET_RUNLOCK();
 2852 
 2853         /*
 2854          * If we didn't allocate enough space (uncommon), try again.  If
 2855          * we have already allocated as much space as we are allowed,
 2856          * return what we've got.
 2857          */
 2858         if (valid_len != max_len && !full) {
 2859                 sbuf_delete(sb);
 2860                 goto again;
 2861         }
 2862 
 2863         ifc->ifc_len = valid_len;
 2864         sbuf_finish(sb);
 2865         error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
 2866         sbuf_delete(sb);
 2867         return (error);
 2868 }
 2869 
 2870 /*
 2871  * Just like ifpromisc(), but for all-multicast-reception mode.
 2872  */
 2873 int
 2874 if_allmulti(struct ifnet *ifp, int onswitch)
 2875 {
 2876 
 2877         return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
 2878 }
 2879 
 2880 struct ifmultiaddr *
 2881 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
 2882 {
 2883         struct ifmultiaddr *ifma;
 2884 
 2885         IF_ADDR_LOCK_ASSERT(ifp);
 2886 
 2887         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
 2888                 if (sa->sa_family == AF_LINK) {
 2889                         if (sa_dl_equal(ifma->ifma_addr, sa))
 2890                                 break;
 2891                 } else {
 2892                         if (sa_equal(ifma->ifma_addr, sa))
 2893                                 break;
 2894                 }
 2895         }
 2896 
 2897         return ifma;
 2898 }
 2899 
 2900 /*
 2901  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
 2902  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
 2903  * the ifnet multicast address list here, so the caller must do that and
 2904  * other setup work (such as notifying the device driver).  The reference
 2905  * count is initialized to 1.
 2906  */
 2907 static struct ifmultiaddr *
 2908 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
 2909     int mflags)
 2910 {
 2911         struct ifmultiaddr *ifma;
 2912         struct sockaddr *dupsa;
 2913 
 2914         ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
 2915             M_ZERO);
 2916         if (ifma == NULL)
 2917                 return (NULL);
 2918 
 2919         dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
 2920         if (dupsa == NULL) {
 2921                 free(ifma, M_IFMADDR);
 2922                 return (NULL);
 2923         }
 2924         bcopy(sa, dupsa, sa->sa_len);
 2925         ifma->ifma_addr = dupsa;
 2926 
 2927         ifma->ifma_ifp = ifp;
 2928         ifma->ifma_refcount = 1;
 2929         ifma->ifma_protospec = NULL;
 2930 
 2931         if (llsa == NULL) {
 2932                 ifma->ifma_lladdr = NULL;
 2933                 return (ifma);
 2934         }
 2935 
 2936         dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
 2937         if (dupsa == NULL) {
 2938                 free(ifma->ifma_addr, M_IFMADDR);
 2939                 free(ifma, M_IFMADDR);
 2940                 return (NULL);
 2941         }
 2942         bcopy(llsa, dupsa, llsa->sa_len);
 2943         ifma->ifma_lladdr = dupsa;
 2944 
 2945         return (ifma);
 2946 }
 2947 
 2948 /*
 2949  * if_freemulti: free ifmultiaddr structure and possibly attached related
 2950  * addresses.  The caller is responsible for implementing reference
 2951  * counting, notifying the driver, handling routing messages, and releasing
 2952  * any dependent link layer state.
 2953  */
 2954 static void
 2955 if_freemulti(struct ifmultiaddr *ifma)
 2956 {
 2957 
 2958         KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
 2959             ifma->ifma_refcount));
 2960         KASSERT(ifma->ifma_protospec == NULL,
 2961             ("if_freemulti: protospec not NULL"));
 2962 
 2963         if (ifma->ifma_lladdr != NULL)
 2964                 free(ifma->ifma_lladdr, M_IFMADDR);
 2965         free(ifma->ifma_addr, M_IFMADDR);
 2966         free(ifma, M_IFMADDR);
 2967 }
 2968 
 2969 /*
 2970  * Register an additional multicast address with a network interface.
 2971  *
 2972  * - If the address is already present, bump the reference count on the
 2973  *   address and return.
 2974  * - If the address is not link-layer, look up a link layer address.
 2975  * - Allocate address structures for one or both addresses, and attach to the
 2976  *   multicast address list on the interface.  If automatically adding a link
 2977  *   layer address, the protocol address will own a reference to the link
 2978  *   layer address, to be freed when it is freed.
 2979  * - Notify the network device driver of an addition to the multicast address
 2980  *   list.
 2981  *
 2982  * 'sa' points to caller-owned memory with the desired multicast address.
 2983  *
 2984  * 'retifma' will be used to return a pointer to the resulting multicast
 2985  * address reference, if desired.
 2986  */
 2987 int
 2988 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
 2989     struct ifmultiaddr **retifma)
 2990 {
 2991         struct ifmultiaddr *ifma, *ll_ifma;
 2992         struct sockaddr *llsa;
 2993         int error;
 2994 
 2995         /*
 2996          * If the address is already present, return a new reference to it;
 2997          * otherwise, allocate storage and set up a new address.
 2998          */
 2999         IF_ADDR_WLOCK(ifp);
 3000         ifma = if_findmulti(ifp, sa);
 3001         if (ifma != NULL) {
 3002                 ifma->ifma_refcount++;
 3003                 if (retifma != NULL)
 3004                         *retifma = ifma;
 3005                 IF_ADDR_WUNLOCK(ifp);
 3006                 return (0);
 3007         }
 3008 
 3009         /*
 3010          * The address isn't already present; resolve the protocol address
 3011          * into a link layer address, and then look that up, bump its
 3012          * refcount or allocate an ifma for that also.  If 'llsa' was
 3013          * returned, we will need to free it later.
 3014          */
 3015         llsa = NULL;
 3016         ll_ifma = NULL;
 3017         if (ifp->if_resolvemulti != NULL) {
 3018                 error = ifp->if_resolvemulti(ifp, &llsa, sa);
 3019                 if (error)
 3020                         goto unlock_out;
 3021         }
 3022 
 3023         /*
 3024          * Allocate the new address.  Don't hook it up yet, as we may also
 3025          * need to allocate a link layer multicast address.
 3026          */
 3027         ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
 3028         if (ifma == NULL) {
 3029                 error = ENOMEM;
 3030                 goto free_llsa_out;
 3031         }
 3032 
 3033         /*
 3034          * If a link layer address is found, we'll need to see if it's
 3035          * already present in the address list, or allocate is as well.
 3036          * When this block finishes, the link layer address will be on the
 3037          * list.
 3038          */
 3039         if (llsa != NULL) {
 3040                 ll_ifma = if_findmulti(ifp, llsa);
 3041                 if (ll_ifma == NULL) {
 3042                         ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
 3043                         if (ll_ifma == NULL) {
 3044                                 --ifma->ifma_refcount;
 3045                                 if_freemulti(ifma);
 3046                                 error = ENOMEM;
 3047                                 goto free_llsa_out;
 3048                         }
 3049                         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
 3050                             ifma_link);
 3051                 } else
 3052                         ll_ifma->ifma_refcount++;
 3053                 ifma->ifma_llifma = ll_ifma;
 3054         }
 3055 
 3056         /*
 3057          * We now have a new multicast address, ifma, and possibly a new or
 3058          * referenced link layer address.  Add the primary address to the
 3059          * ifnet address list.
 3060          */
 3061         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
 3062 
 3063         if (retifma != NULL)
 3064                 *retifma = ifma;
 3065 
 3066         /*
 3067          * Must generate the message while holding the lock so that 'ifma'
 3068          * pointer is still valid.
 3069          */
 3070         rt_newmaddrmsg(RTM_NEWMADDR, ifma);
 3071         IF_ADDR_WUNLOCK(ifp);
 3072 
 3073         /*
 3074          * We are certain we have added something, so call down to the
 3075          * interface to let them know about it.
 3076          */
 3077         if (ifp->if_ioctl != NULL) {
 3078                 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
 3079         }
 3080 
 3081         if (llsa != NULL)
 3082                 free(llsa, M_IFMADDR);
 3083 
 3084         return (0);
 3085 
 3086 free_llsa_out:
 3087         if (llsa != NULL)
 3088                 free(llsa, M_IFMADDR);
 3089 
 3090 unlock_out:
 3091         IF_ADDR_WUNLOCK(ifp);
 3092         return (error);
 3093 }
 3094 
 3095 /*
 3096  * Delete a multicast group membership by network-layer group address.
 3097  *
 3098  * Returns ENOENT if the entry could not be found. If ifp no longer
 3099  * exists, results are undefined. This entry point should only be used
 3100  * from subsystems which do appropriate locking to hold ifp for the
 3101  * duration of the call.
 3102  * Network-layer protocol domains must use if_delmulti_ifma().
 3103  */
 3104 int
 3105 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
 3106 {
 3107         struct ifmultiaddr *ifma;
 3108         int lastref;
 3109 #ifdef INVARIANTS
 3110         struct ifnet *oifp;
 3111 
 3112         IFNET_RLOCK_NOSLEEP();
 3113         TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3114                 if (ifp == oifp)
 3115                         break;
 3116         if (ifp != oifp)
 3117                 ifp = NULL;
 3118         IFNET_RUNLOCK_NOSLEEP();
 3119 
 3120         KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
 3121 #endif
 3122         if (ifp == NULL)
 3123                 return (ENOENT);
 3124 
 3125         IF_ADDR_WLOCK(ifp);
 3126         lastref = 0;
 3127         ifma = if_findmulti(ifp, sa);
 3128         if (ifma != NULL)
 3129                 lastref = if_delmulti_locked(ifp, ifma, 0);
 3130         IF_ADDR_WUNLOCK(ifp);
 3131 
 3132         if (ifma == NULL)
 3133                 return (ENOENT);
 3134 
 3135         if (lastref && ifp->if_ioctl != NULL) {
 3136                 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3137         }
 3138 
 3139         return (0);
 3140 }
 3141 
 3142 /*
 3143  * Delete all multicast group membership for an interface.
 3144  * Should be used to quickly flush all multicast filters.
 3145  */
 3146 void
 3147 if_delallmulti(struct ifnet *ifp)
 3148 {
 3149         struct ifmultiaddr *ifma;
 3150         struct ifmultiaddr *next;
 3151 
 3152         IF_ADDR_WLOCK(ifp);
 3153         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
 3154                 if_delmulti_locked(ifp, ifma, 0);
 3155         IF_ADDR_WUNLOCK(ifp);
 3156 }
 3157 
 3158 /*
 3159  * Delete a multicast group membership by group membership pointer.
 3160  * Network-layer protocol domains must use this routine.
 3161  *
 3162  * It is safe to call this routine if the ifp disappeared.
 3163  */
 3164 void
 3165 if_delmulti_ifma(struct ifmultiaddr *ifma)
 3166 {
 3167         struct ifnet *ifp;
 3168         int lastref;
 3169 
 3170         ifp = ifma->ifma_ifp;
 3171 #ifdef DIAGNOSTIC
 3172         if (ifp == NULL) {
 3173                 printf("%s: ifma_ifp seems to be detached\n", __func__);
 3174         } else {
 3175                 struct ifnet *oifp;
 3176 
 3177                 IFNET_RLOCK_NOSLEEP();
 3178                 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3179                         if (ifp == oifp)
 3180                                 break;
 3181                 if (ifp != oifp) {
 3182                         printf("%s: ifnet %p disappeared\n", __func__, ifp);
 3183                         ifp = NULL;
 3184                 }
 3185                 IFNET_RUNLOCK_NOSLEEP();
 3186         }
 3187 #endif
 3188         /*
 3189          * If and only if the ifnet instance exists: Acquire the address lock.
 3190          */
 3191         if (ifp != NULL)
 3192                 IF_ADDR_WLOCK(ifp);
 3193 
 3194         lastref = if_delmulti_locked(ifp, ifma, 0);
 3195 
 3196         if (ifp != NULL) {
 3197                 /*
 3198                  * If and only if the ifnet instance exists:
 3199                  *  Release the address lock.
 3200                  *  If the group was left: update the hardware hash filter.
 3201                  */
 3202                 IF_ADDR_WUNLOCK(ifp);
 3203                 if (lastref && ifp->if_ioctl != NULL) {
 3204                         (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3205                 }
 3206         }
 3207 }
 3208 
 3209 /*
 3210  * Perform deletion of network-layer and/or link-layer multicast address.
 3211  *
 3212  * Return 0 if the reference count was decremented.
 3213  * Return 1 if the final reference was released, indicating that the
 3214  * hardware hash filter should be reprogrammed.
 3215  */
 3216 static int
 3217 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
 3218 {
 3219         struct ifmultiaddr *ll_ifma;
 3220 
 3221         if (ifp != NULL && ifma->ifma_ifp != NULL) {
 3222                 KASSERT(ifma->ifma_ifp == ifp,
 3223                     ("%s: inconsistent ifp %p", __func__, ifp));
 3224                 IF_ADDR_WLOCK_ASSERT(ifp);
 3225         }
 3226 
 3227         ifp = ifma->ifma_ifp;
 3228 
 3229         /*
 3230          * If the ifnet is detaching, null out references to ifnet,
 3231          * so that upper protocol layers will notice, and not attempt
 3232          * to obtain locks for an ifnet which no longer exists. The
 3233          * routing socket announcement must happen before the ifnet
 3234          * instance is detached from the system.
 3235          */
 3236         if (detaching) {
 3237 #ifdef DIAGNOSTIC
 3238                 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
 3239 #endif
 3240                 /*
 3241                  * ifp may already be nulled out if we are being reentered
 3242                  * to delete the ll_ifma.
 3243                  */
 3244                 if (ifp != NULL) {
 3245                         rt_newmaddrmsg(RTM_DELMADDR, ifma);
 3246                         ifma->ifma_ifp = NULL;
 3247                 }
 3248         }
 3249 
 3250         if (--ifma->ifma_refcount > 0)
 3251                 return 0;
 3252 
 3253         /*
 3254          * If this ifma is a network-layer ifma, a link-layer ifma may
 3255          * have been associated with it. Release it first if so.
 3256          */
 3257         ll_ifma = ifma->ifma_llifma;
 3258         if (ll_ifma != NULL) {
 3259                 KASSERT(ifma->ifma_lladdr != NULL,
 3260                     ("%s: llifma w/o lladdr", __func__));
 3261                 if (detaching)
 3262                         ll_ifma->ifma_ifp = NULL;       /* XXX */
 3263                 if (--ll_ifma->ifma_refcount == 0) {
 3264                         if (ifp != NULL) {
 3265                                 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
 3266                                     ifma_link);
 3267                         }
 3268                         if_freemulti(ll_ifma);
 3269                 }
 3270         }
 3271 
 3272         if (ifp != NULL)
 3273                 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
 3274 
 3275         if_freemulti(ifma);
 3276 
 3277         /*
 3278          * The last reference to this instance of struct ifmultiaddr
 3279          * was released; the hardware should be notified of this change.
 3280          */
 3281         return 1;
 3282 }
 3283 
 3284 /*
 3285  * Set the link layer address on an interface.
 3286  *
 3287  * At this time we only support certain types of interfaces,
 3288  * and we don't allow the length of the address to change.
 3289  */
 3290 int
 3291 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
 3292 {
 3293         struct sockaddr_dl *sdl;
 3294         struct ifaddr *ifa;
 3295         struct ifreq ifr;
 3296 
 3297         IF_ADDR_RLOCK(ifp);
 3298         ifa = ifp->if_addr;
 3299         if (ifa == NULL) {
 3300                 IF_ADDR_RUNLOCK(ifp);
 3301                 return (EINVAL);
 3302         }
 3303         ifa_ref(ifa);
 3304         IF_ADDR_RUNLOCK(ifp);
 3305         sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 3306         if (sdl == NULL) {
 3307                 ifa_free(ifa);
 3308                 return (EINVAL);
 3309         }
 3310         if (len != sdl->sdl_alen) {     /* don't allow length to change */
 3311                 ifa_free(ifa);
 3312                 return (EINVAL);
 3313         }
 3314         switch (ifp->if_type) {
 3315         case IFT_ETHER:
 3316         case IFT_FDDI:
 3317         case IFT_XETHER:
 3318         case IFT_ISO88025:
 3319         case IFT_L2VLAN:
 3320         case IFT_BRIDGE:
 3321         case IFT_ARCNET:
 3322         case IFT_IEEE8023ADLAG:
 3323         case IFT_IEEE80211:
 3324                 bcopy(lladdr, LLADDR(sdl), len);
 3325                 ifa_free(ifa);
 3326                 break;
 3327         default:
 3328                 ifa_free(ifa);
 3329                 return (ENODEV);
 3330         }
 3331 
 3332         /*
 3333          * If the interface is already up, we need
 3334          * to re-init it in order to reprogram its
 3335          * address filter.
 3336          */
 3337         if ((ifp->if_flags & IFF_UP) != 0) {
 3338                 if (ifp->if_ioctl) {
 3339                         ifp->if_flags &= ~IFF_UP;
 3340                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3341                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3342                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3343                         ifp->if_flags |= IFF_UP;
 3344                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3345                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3346                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3347                 }
 3348 #ifdef INET
 3349                 /*
 3350                  * Also send gratuitous ARPs to notify other nodes about
 3351                  * the address change.
 3352                  */
 3353                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 3354                         if (ifa->ifa_addr->sa_family == AF_INET)
 3355                                 arp_ifinit(ifp, ifa);
 3356                 }
 3357 #endif
 3358         }
 3359         return (0);
 3360 }
 3361 
 3362 /*
 3363  * The name argument must be a pointer to storage which will last as
 3364  * long as the interface does.  For physical devices, the result of
 3365  * device_get_name(dev) is a good choice and for pseudo-devices a
 3366  * static string works well.
 3367  */
 3368 void
 3369 if_initname(struct ifnet *ifp, const char *name, int unit)
 3370 {
 3371         ifp->if_dname = name;
 3372         ifp->if_dunit = unit;
 3373         if (unit != IF_DUNIT_NONE)
 3374                 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
 3375         else
 3376                 strlcpy(ifp->if_xname, name, IFNAMSIZ);
 3377 }
 3378 
 3379 int
 3380 if_printf(struct ifnet *ifp, const char * fmt, ...)
 3381 {
 3382         va_list ap;
 3383         int retval;
 3384 
 3385         retval = printf("%s: ", ifp->if_xname);
 3386         va_start(ap, fmt);
 3387         retval += vprintf(fmt, ap);
 3388         va_end(ap);
 3389         return (retval);
 3390 }
 3391 
 3392 void
 3393 if_start(struct ifnet *ifp)
 3394 {
 3395 
 3396         (*(ifp)->if_start)(ifp);
 3397 }
 3398 
 3399 /*
 3400  * Backwards compatibility interface for drivers 
 3401  * that have not implemented it
 3402  */
 3403 static int
 3404 if_transmit(struct ifnet *ifp, struct mbuf *m)
 3405 {
 3406         int error;
 3407 
 3408         IFQ_HANDOFF(ifp, m, error);
 3409         return (error);
 3410 }
 3411 
 3412 int
 3413 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
 3414 {
 3415         int active = 0;
 3416 
 3417         IF_LOCK(ifq);
 3418         if (_IF_QFULL(ifq)) {
 3419                 _IF_DROP(ifq);
 3420                 IF_UNLOCK(ifq);
 3421                 m_freem(m);
 3422                 return (0);
 3423         }
 3424         if (ifp != NULL) {
 3425                 ifp->if_obytes += m->m_pkthdr.len + adjust;
 3426                 if (m->m_flags & (M_BCAST|M_MCAST))
 3427                         ifp->if_omcasts++;
 3428                 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
 3429         }
 3430         _IF_ENQUEUE(ifq, m);
 3431         IF_UNLOCK(ifq);
 3432         if (ifp != NULL && !active)
 3433                 (*(ifp)->if_start)(ifp);
 3434         return (1);
 3435 }
 3436 
 3437 void
 3438 if_register_com_alloc(u_char type,
 3439     if_com_alloc_t *a, if_com_free_t *f)
 3440 {
 3441         
 3442         KASSERT(if_com_alloc[type] == NULL,
 3443             ("if_register_com_alloc: %d already registered", type));
 3444         KASSERT(if_com_free[type] == NULL,
 3445             ("if_register_com_alloc: %d free already registered", type));
 3446 
 3447         if_com_alloc[type] = a;
 3448         if_com_free[type] = f;
 3449 }
 3450 
 3451 void
 3452 if_deregister_com_alloc(u_char type)
 3453 {
 3454         
 3455         KASSERT(if_com_alloc[type] != NULL,
 3456             ("if_deregister_com_alloc: %d not registered", type));
 3457         KASSERT(if_com_free[type] != NULL,
 3458             ("if_deregister_com_alloc: %d free not registered", type));
 3459         if_com_alloc[type] = NULL;
 3460         if_com_free[type] = NULL;
 3461 }

Cache object: acea6e55ab6054bb0907902c0f6de6da


[ 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.