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/8.2/sys/net/if.c 215342 2010-11-15 17:48:13Z sobomax $
   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 <machine/stdarg.h>
   62 #include <vm/uma.h>
   63 
   64 #include <net/if.h>
   65 #include <net/if_arp.h>
   66 #include <net/if_clone.h>
   67 #include <net/if_dl.h>
   68 #include <net/if_types.h>
   69 #include <net/if_var.h>
   70 #include <net/radix.h>
   71 #include <net/route.h>
   72 #include <net/vnet.h>
   73 
   74 #if defined(INET) || defined(INET6)
   75 /*XXX*/
   76 #include <netinet/in.h>
   77 #include <netinet/in_var.h>
   78 #include <netinet/ip_carp.h>
   79 #ifdef INET6
   80 #include <netinet6/in6_var.h>
   81 #include <netinet6/in6_ifattach.h>
   82 #endif
   83 #endif
   84 #ifdef INET
   85 #include <netinet/if_ether.h>
   86 #endif
   87 
   88 #include <security/mac/mac_framework.h>
   89 
   90 #ifdef COMPAT_FREEBSD32
   91 #include <sys/mount.h>
   92 #include <compat/freebsd32/freebsd32.h>
   93 #endif
   94 
   95 struct ifindex_entry {
   96         struct  ifnet *ife_ifnet;
   97 };
   98 
   99 static int slowtimo_started;
  100 
  101 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
  102 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
  103 
  104 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
  105 SYSCTL_UINT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
  106     &ifqmaxlen, 0, "max send queue size");
  107 
  108 /* Log link state change events */
  109 static int log_link_state_change = 1;
  110 
  111 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
  112         &log_link_state_change, 0,
  113         "log interface link state change events");
  114 
  115 /* Interface description */
  116 static unsigned int ifdescr_maxlen = 1024;
  117 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
  118         &ifdescr_maxlen, 0,
  119         "administrative maximum length for interface description");
  120 
  121 MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
  122 
  123 /* global sx for non-critical path ifdescr */
  124 static struct sx ifdescr_sx;
  125 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
  126 
  127 void    (*bstp_linkstate_p)(struct ifnet *ifp, int state);
  128 void    (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
  129 void    (*lagg_linkstate_p)(struct ifnet *ifp, int state);
  130 /* These are external hooks for CARP. */
  131 void    (*carp_linkstate_p)(struct ifnet *ifp);
  132 #if defined(INET) || defined(INET6)
  133 struct ifnet *(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
  134 int     (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
  135     struct sockaddr *sa, struct rtentry *rt);
  136 #endif
  137 #ifdef INET
  138 int (*carp_iamatch_p)(struct ifnet *, struct in_ifaddr *, struct in_addr *,
  139     u_int8_t **);
  140 #endif
  141 #ifdef INET6
  142 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
  143 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
  144     const struct in6_addr *taddr);
  145 #endif
  146 
  147 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
  148 
  149 /*
  150  * XXX: Style; these should be sorted alphabetically, and unprototyped
  151  * static functions should be prototyped. Currently they are sorted by
  152  * declaration order.
  153  */
  154 static void     if_attachdomain(void *);
  155 static void     if_attachdomain1(struct ifnet *);
  156 static int      ifconf(u_long, caddr_t);
  157 static void     if_freemulti(struct ifmultiaddr *);
  158 static void     if_init(void *);
  159 static void     if_grow(void);
  160 static void     if_check(void *);
  161 static void     if_route(struct ifnet *, int flag, int fam);
  162 static int      if_setflag(struct ifnet *, int, int, int *, int);
  163 static void     if_slowtimo(void *);
  164 static int      if_transmit(struct ifnet *ifp, struct mbuf *m);
  165 static void     if_unroute(struct ifnet *, int flag, int fam);
  166 static void     link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  167 static int      if_rtdel(struct radix_node *, void *);
  168 static int      ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
  169 static int      if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
  170 static void     do_link_state_change(void *, int);
  171 static int      if_getgroup(struct ifgroupreq *, struct ifnet *);
  172 static int      if_getgroupmembers(struct ifgroupreq *);
  173 static void     if_delgroups(struct ifnet *);
  174 static void     if_attach_internal(struct ifnet *, int);
  175 static void     if_detach_internal(struct ifnet *, int);
  176 
  177 #ifdef INET6
  178 /*
  179  * XXX: declare here to avoid to include many inet6 related files..
  180  * should be more generalized?
  181  */
  182 extern void     nd6_setmtu(struct ifnet *);
  183 #endif
  184 
  185 VNET_DEFINE(int, if_index);
  186 int     ifqmaxlen = IFQ_MAXLEN;
  187 VNET_DEFINE(struct ifnethead, ifnet);   /* depend on static init XXX */
  188 VNET_DEFINE(struct ifgrouphead, ifg_head);
  189 
  190 static VNET_DEFINE(int, if_indexlim) = 8;
  191 
  192 /* Table of ifnet by index. */
  193 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
  194 
  195 #define V_if_indexlim           VNET(if_indexlim)
  196 #define V_ifindex_table         VNET(ifindex_table)
  197 
  198 /*
  199  * The global network interface list (V_ifnet) and related state (such as
  200  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
  201  * an rwlock.  Either may be acquired shared to stablize the list, but both
  202  * must be acquired writable to modify the list.  This model allows us to
  203  * both stablize the interface list during interrupt thread processing, but
  204  * also to stablize it over long-running ioctls, without introducing priority
  205  * inversions and deadlocks.
  206  */
  207 struct rwlock ifnet_rwlock;
  208 struct sx ifnet_sxlock;
  209 
  210 /*
  211  * The allocation of network interfaces is a rather non-atomic affair; we
  212  * need to select an index before we are ready to expose the interface for
  213  * use, so will use this pointer value to indicate reservation.
  214  */
  215 #define IFNET_HOLD      (void *)(uintptr_t)(-1)
  216 
  217 static  if_com_alloc_t *if_com_alloc[256];
  218 static  if_com_free_t *if_com_free[256];
  219 
  220 /*
  221  * System initialization
  222  */
  223 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL);
  224 
  225 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
  226 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
  227 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
  228 
  229 struct ifnet *
  230 ifnet_byindex_locked(u_short idx)
  231 {
  232 
  233         if (idx > V_if_index)
  234                 return (NULL);
  235         if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
  236                 return (NULL);
  237         return (V_ifindex_table[idx].ife_ifnet);
  238 }
  239 
  240 struct ifnet *
  241 ifnet_byindex(u_short idx)
  242 {
  243         struct ifnet *ifp;
  244 
  245         IFNET_RLOCK_NOSLEEP();
  246         ifp = ifnet_byindex_locked(idx);
  247         IFNET_RUNLOCK_NOSLEEP();
  248         return (ifp);
  249 }
  250 
  251 struct ifnet *
  252 ifnet_byindex_ref(u_short idx)
  253 {
  254         struct ifnet *ifp;
  255 
  256         IFNET_RLOCK_NOSLEEP();
  257         ifp = ifnet_byindex_locked(idx);
  258         if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
  259                 IFNET_RUNLOCK_NOSLEEP();
  260                 return (NULL);
  261         }
  262         if_ref(ifp);
  263         IFNET_RUNLOCK_NOSLEEP();
  264         return (ifp);
  265 }
  266 
  267 /*
  268  * Allocate an ifindex array entry; return 0 on success or an error on
  269  * failure.
  270  */
  271 static int
  272 ifindex_alloc_locked(u_short *idxp)
  273 {
  274         u_short idx;
  275 
  276         IFNET_WLOCK_ASSERT();
  277 
  278         /*
  279          * Try to find an empty slot below V_if_index.  If we fail, take the
  280          * next slot.
  281          */
  282         for (idx = 1; idx <= V_if_index; idx++) {
  283                 if (V_ifindex_table[idx].ife_ifnet == NULL)
  284                         break;
  285         }
  286 
  287         /* Catch if_index overflow. */
  288         if (idx < 1)
  289                 return (ENOSPC);
  290         if (idx > V_if_index)
  291                 V_if_index = idx;
  292         if (V_if_index >= V_if_indexlim)
  293                 if_grow();
  294         *idxp = idx;
  295         return (0);
  296 }
  297 
  298 static void
  299 ifindex_free_locked(u_short idx)
  300 {
  301 
  302         IFNET_WLOCK_ASSERT();
  303 
  304         V_ifindex_table[idx].ife_ifnet = NULL;
  305         while (V_if_index > 0 &&
  306             V_ifindex_table[V_if_index].ife_ifnet == NULL)
  307                 V_if_index--;
  308 }
  309 
  310 static void
  311 ifindex_free(u_short idx)
  312 {
  313 
  314         IFNET_WLOCK();
  315         ifindex_free_locked(idx);
  316         IFNET_WUNLOCK();
  317 }
  318 
  319 static void
  320 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
  321 {
  322 
  323         IFNET_WLOCK_ASSERT();
  324 
  325         V_ifindex_table[idx].ife_ifnet = ifp;
  326 }
  327 
  328 static void
  329 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
  330 {
  331 
  332         IFNET_WLOCK();
  333         ifnet_setbyindex_locked(idx, ifp);
  334         IFNET_WUNLOCK();
  335 }
  336 
  337 struct ifaddr *
  338 ifaddr_byindex(u_short idx)
  339 {
  340         struct ifaddr *ifa;
  341 
  342         IFNET_RLOCK_NOSLEEP();
  343         ifa = ifnet_byindex_locked(idx)->if_addr;
  344         if (ifa != NULL)
  345                 ifa_ref(ifa);
  346         IFNET_RUNLOCK_NOSLEEP();
  347         return (ifa);
  348 }
  349 
  350 /*
  351  * Network interface utility routines.
  352  *
  353  * Routines with ifa_ifwith* names take sockaddr *'s as
  354  * parameters.
  355  */
  356 
  357 static void
  358 vnet_if_init(const void *unused __unused)
  359 {
  360 
  361         TAILQ_INIT(&V_ifnet);
  362         TAILQ_INIT(&V_ifg_head);
  363         if_grow();                              /* create initial table */
  364         vnet_if_clone_init();
  365 }
  366 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_init,
  367     NULL);
  368 
  369 /* ARGSUSED*/
  370 static void
  371 if_init(void *dummy __unused)
  372 {
  373 
  374         IFNET_LOCK_INIT();
  375         if_clone_init();
  376 }
  377 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
  378 
  379 
  380 #ifdef VIMAGE
  381 static void
  382 vnet_if_uninit(const void *unused __unused)
  383 {
  384 
  385         VNET_ASSERT(TAILQ_EMPTY(&V_ifnet));
  386         VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head));
  387 
  388         free((caddr_t)V_ifindex_table, M_IFNET);
  389 }
  390 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
  391     vnet_if_uninit, NULL);
  392 #endif
  393 
  394 static void
  395 if_grow(void)
  396 {
  397         u_int n;
  398         struct ifindex_entry *e;
  399 
  400         V_if_indexlim <<= 1;
  401         n = V_if_indexlim * sizeof(*e);
  402         e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
  403         if (V_ifindex_table != NULL) {
  404                 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
  405                 free((caddr_t)V_ifindex_table, M_IFNET);
  406         }
  407         V_ifindex_table = e;
  408 }
  409 
  410 static void
  411 if_check(void *dummy __unused)
  412 {
  413 
  414         /*
  415          * If at least one interface added during boot uses
  416          * if_watchdog then start the timer.
  417          */
  418         if (slowtimo_started)
  419                 if_slowtimo(0);
  420 }
  421 
  422 /*
  423  * Allocate a struct ifnet and an index for an interface.  A layer 2
  424  * common structure will also be allocated if an allocation routine is
  425  * registered for the passed type.
  426  */
  427 struct ifnet *
  428 if_alloc(u_char type)
  429 {
  430         struct ifnet *ifp;
  431         u_short idx;
  432 
  433         ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
  434         IFNET_WLOCK();
  435         if (ifindex_alloc_locked(&idx) != 0) {
  436                 IFNET_WUNLOCK();
  437                 free(ifp, M_IFNET);
  438                 return (NULL);
  439         }
  440         ifnet_setbyindex_locked(idx, IFNET_HOLD);
  441         IFNET_WUNLOCK();
  442         ifp->if_index = idx;
  443         ifp->if_type = type;
  444         ifp->if_alloctype = type;
  445         if (if_com_alloc[type] != NULL) {
  446                 ifp->if_l2com = if_com_alloc[type](type, ifp);
  447                 if (ifp->if_l2com == NULL) {
  448                         free(ifp, M_IFNET);
  449                         ifindex_free(idx);
  450                         return (NULL);
  451                 }
  452         }
  453 
  454         IF_ADDR_LOCK_INIT(ifp);
  455         TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
  456         ifp->if_afdata_initialized = 0;
  457         IF_AFDATA_LOCK_INIT(ifp);
  458         TAILQ_INIT(&ifp->if_addrhead);
  459         TAILQ_INIT(&ifp->if_prefixhead);
  460         TAILQ_INIT(&ifp->if_multiaddrs);
  461         TAILQ_INIT(&ifp->if_groups);
  462 #ifdef MAC
  463         mac_ifnet_init(ifp);
  464 #endif
  465         ifq_init(&ifp->if_snd, ifp);
  466 
  467         refcount_init(&ifp->if_refcount, 1);    /* Index reference. */
  468         ifnet_setbyindex(ifp->if_index, ifp);
  469         return (ifp);
  470 }
  471 
  472 /*
  473  * Do the actual work of freeing a struct ifnet, associated index, and layer
  474  * 2 common structure.  This call is made when the last reference to an
  475  * interface is released.
  476  */
  477 static void
  478 if_free_internal(struct ifnet *ifp)
  479 {
  480 
  481         KASSERT((ifp->if_flags & IFF_DYING),
  482             ("if_free_internal: interface not dying"));
  483 
  484         IFNET_WLOCK();
  485         KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
  486             ("%s: freeing unallocated ifnet", ifp->if_xname));
  487 
  488         ifindex_free_locked(ifp->if_index);
  489         IFNET_WUNLOCK();
  490 
  491         if (if_com_free[ifp->if_alloctype] != NULL)
  492                 if_com_free[ifp->if_alloctype](ifp->if_l2com,
  493                     ifp->if_alloctype);
  494 
  495 #ifdef MAC
  496         mac_ifnet_destroy(ifp);
  497 #endif /* MAC */
  498         if (ifp->if_description != NULL)
  499                 free(ifp->if_description, M_IFDESCR);
  500         IF_AFDATA_DESTROY(ifp);
  501         IF_ADDR_LOCK_DESTROY(ifp);
  502         ifq_delete(&ifp->if_snd);
  503         free(ifp, M_IFNET);
  504 }
  505 
  506 /*
  507  * This version should only be called by intefaces that switch their type
  508  * after calling if_alloc().  if_free_type() will go away again now that we
  509  * have if_alloctype to cache the original allocation type.  For now, assert
  510  * that they match, since we require that in practice.
  511  */
  512 void
  513 if_free_type(struct ifnet *ifp, u_char type)
  514 {
  515 
  516         KASSERT(ifp->if_alloctype == type,
  517             ("if_free_type: type (%d) != alloctype (%d)", type,
  518             ifp->if_alloctype));
  519 
  520         ifp->if_flags |= IFF_DYING;                     /* XXX: Locking */
  521         if (!refcount_release(&ifp->if_refcount))
  522                 return;
  523         if_free_internal(ifp);
  524 }
  525 
  526 /*
  527  * This is the normal version of if_free(), used by device drivers to free a
  528  * detached network interface.  The contents of if_free_type() will move into
  529  * here when if_free_type() goes away.
  530  */
  531 void
  532 if_free(struct ifnet *ifp)
  533 {
  534 
  535         if_free_type(ifp, ifp->if_alloctype);
  536 }
  537 
  538 /*
  539  * Interfaces to keep an ifnet type-stable despite the possibility of the
  540  * driver calling if_free().  If there are additional references, we defer
  541  * freeing the underlying data structure.
  542  */
  543 void
  544 if_ref(struct ifnet *ifp)
  545 {
  546 
  547         /* We don't assert the ifnet list lock here, but arguably should. */
  548         refcount_acquire(&ifp->if_refcount);
  549 }
  550 
  551 void
  552 if_rele(struct ifnet *ifp)
  553 {
  554 
  555         if (!refcount_release(&ifp->if_refcount))
  556                 return;
  557         if_free_internal(ifp);
  558 }
  559 
  560 void
  561 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
  562 {
  563         
  564         mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
  565 
  566         if (ifq->ifq_maxlen == 0) 
  567                 ifq->ifq_maxlen = ifqmaxlen;
  568 
  569         ifq->altq_type = 0;
  570         ifq->altq_disc = NULL;
  571         ifq->altq_flags &= ALTQF_CANTCHANGE;
  572         ifq->altq_tbr  = NULL;
  573         ifq->altq_ifp  = ifp;
  574 }
  575 
  576 void
  577 ifq_delete(struct ifaltq *ifq)
  578 {
  579         mtx_destroy(&ifq->ifq_mtx);
  580 }
  581 
  582 /*
  583  * Perform generic interface initalization tasks and attach the interface
  584  * to the list of "active" interfaces.  If vmove flag is set on entry
  585  * to if_attach_internal(), perform only a limited subset of initialization
  586  * tasks, given that we are moving from one vnet to another an ifnet which
  587  * has already been fully initialized.
  588  *
  589  * XXX:
  590  *  - The decision to return void and thus require this function to
  591  *    succeed is questionable.
  592  *  - We should probably do more sanity checking.  For instance we don't
  593  *    do anything to insure if_xname is unique or non-empty.
  594  */
  595 void
  596 if_attach(struct ifnet *ifp)
  597 {
  598 
  599         if_attach_internal(ifp, 0);
  600 }
  601 
  602 static void
  603 if_attach_internal(struct ifnet *ifp, int vmove)
  604 {
  605         unsigned socksize, ifasize;
  606         int namelen, masklen;
  607         struct sockaddr_dl *sdl;
  608         struct ifaddr *ifa;
  609 
  610         if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
  611                 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
  612                     ifp->if_xname);
  613 
  614 #ifdef VIMAGE
  615         ifp->if_vnet = curvnet;
  616         if (ifp->if_home_vnet == NULL)
  617                 ifp->if_home_vnet = curvnet;
  618 #endif
  619 
  620         if_addgroup(ifp, IFG_ALL);
  621 
  622         getmicrotime(&ifp->if_lastchange);
  623         ifp->if_data.ifi_epoch = time_uptime;
  624         ifp->if_data.ifi_datalen = sizeof(struct if_data);
  625 
  626         KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
  627             (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
  628             ("transmit and qflush must both either be set or both be NULL"));
  629         if (ifp->if_transmit == NULL) {
  630                 ifp->if_transmit = if_transmit;
  631                 ifp->if_qflush = if_qflush;
  632         }
  633         
  634         if (!vmove) {
  635 #ifdef MAC
  636                 mac_ifnet_create(ifp);
  637 #endif
  638 
  639                 /*
  640                  * Create a Link Level name for this device.
  641                  */
  642                 namelen = strlen(ifp->if_xname);
  643                 /*
  644                  * Always save enough space for any possiable name so we
  645                  * can do a rename in place later.
  646                  */
  647                 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
  648                 socksize = masklen + ifp->if_addrlen;
  649                 if (socksize < sizeof(*sdl))
  650                         socksize = sizeof(*sdl);
  651                 socksize = roundup2(socksize, sizeof(long));
  652                 ifasize = sizeof(*ifa) + 2 * socksize;
  653                 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
  654                 ifa_init(ifa);
  655                 sdl = (struct sockaddr_dl *)(ifa + 1);
  656                 sdl->sdl_len = socksize;
  657                 sdl->sdl_family = AF_LINK;
  658                 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
  659                 sdl->sdl_nlen = namelen;
  660                 sdl->sdl_index = ifp->if_index;
  661                 sdl->sdl_type = ifp->if_type;
  662                 ifp->if_addr = ifa;
  663                 ifa->ifa_ifp = ifp;
  664                 ifa->ifa_rtrequest = link_rtrequest;
  665                 ifa->ifa_addr = (struct sockaddr *)sdl;
  666                 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
  667                 ifa->ifa_netmask = (struct sockaddr *)sdl;
  668                 sdl->sdl_len = masklen;
  669                 while (namelen != 0)
  670                         sdl->sdl_data[--namelen] = 0xff;
  671                 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
  672                 /* Reliably crash if used uninitialized. */
  673                 ifp->if_broadcastaddr = NULL;
  674         }
  675 #ifdef VIMAGE
  676         else {
  677                 /*
  678                  * Update the interface index in the link layer address
  679                  * of the interface.
  680                  */
  681                 for (ifa = ifp->if_addr; ifa != NULL;
  682                     ifa = TAILQ_NEXT(ifa, ifa_link)) {
  683                         if (ifa->ifa_addr->sa_family == AF_LINK) {
  684                                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
  685                                 sdl->sdl_index = ifp->if_index;
  686                         }
  687                 }
  688         }
  689 #endif
  690 
  691         IFNET_WLOCK();
  692         TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
  693 #ifdef VIMAGE
  694         curvnet->vnet_ifcnt++;
  695 #endif
  696         IFNET_WUNLOCK();
  697 
  698         if (domain_init_status >= 2)
  699                 if_attachdomain1(ifp);
  700 
  701         EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
  702         if (IS_DEFAULT_VNET(curvnet))
  703                 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
  704 
  705         /* Announce the interface. */
  706         rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
  707 
  708         if (!vmove && ifp->if_watchdog != NULL) {
  709                 if_printf(ifp,
  710                     "WARNING: using obsoleted if_watchdog interface\n");
  711 
  712                 /*
  713                  * Note that we need if_slowtimo().  If this happens after
  714                  * boot, then call if_slowtimo() directly.
  715                  */
  716                 if (atomic_cmpset_int(&slowtimo_started, 0, 1) && !cold)
  717                         if_slowtimo(0);
  718         }
  719 }
  720 
  721 static void
  722 if_attachdomain(void *dummy)
  723 {
  724         struct ifnet *ifp;
  725         int s;
  726 
  727         s = splnet();
  728         TAILQ_FOREACH(ifp, &V_ifnet, if_link)
  729                 if_attachdomain1(ifp);
  730         splx(s);
  731 }
  732 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
  733     if_attachdomain, NULL);
  734 
  735 static void
  736 if_attachdomain1(struct ifnet *ifp)
  737 {
  738         struct domain *dp;
  739         int s;
  740 
  741         s = splnet();
  742 
  743         /*
  744          * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
  745          * cannot lock ifp->if_afdata initialization, entirely.
  746          */
  747         if (IF_AFDATA_TRYLOCK(ifp) == 0) {
  748                 splx(s);
  749                 return;
  750         }
  751         if (ifp->if_afdata_initialized >= domain_init_status) {
  752                 IF_AFDATA_UNLOCK(ifp);
  753                 splx(s);
  754                 printf("if_attachdomain called more than once on %s\n",
  755                     ifp->if_xname);
  756                 return;
  757         }
  758         ifp->if_afdata_initialized = domain_init_status;
  759         IF_AFDATA_UNLOCK(ifp);
  760 
  761         /* address family dependent data region */
  762         bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
  763         for (dp = domains; dp; dp = dp->dom_next) {
  764                 if (dp->dom_ifattach)
  765                         ifp->if_afdata[dp->dom_family] =
  766                             (*dp->dom_ifattach)(ifp);
  767         }
  768 
  769         splx(s);
  770 }
  771 
  772 /*
  773  * Remove any unicast or broadcast network addresses from an interface.
  774  */
  775 void
  776 if_purgeaddrs(struct ifnet *ifp)
  777 {
  778         struct ifaddr *ifa, *next;
  779 
  780         TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  781                 if (ifa->ifa_addr->sa_family == AF_LINK)
  782                         continue;
  783 #ifdef INET
  784                 /* XXX: Ugly!! ad hoc just for INET */
  785                 if (ifa->ifa_addr->sa_family == AF_INET) {
  786                         struct ifaliasreq ifr;
  787 
  788                         bzero(&ifr, sizeof(ifr));
  789                         ifr.ifra_addr = *ifa->ifa_addr;
  790                         if (ifa->ifa_dstaddr)
  791                                 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
  792                         if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
  793                             NULL) == 0)
  794                                 continue;
  795                 }
  796 #endif /* INET */
  797 #ifdef INET6
  798                 if (ifa->ifa_addr->sa_family == AF_INET6) {
  799                         in6_purgeaddr(ifa);
  800                         /* ifp_addrhead is already updated */
  801                         continue;
  802                 }
  803 #endif /* INET6 */
  804                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  805                 ifa_free(ifa);
  806         }
  807 }
  808 
  809 /*
  810  * Remove any multicast network addresses from an interface when an ifnet
  811  * is going away.
  812  */
  813 static void
  814 if_purgemaddrs(struct ifnet *ifp)
  815 {
  816         struct ifmultiaddr *ifma;
  817         struct ifmultiaddr *next;
  818 
  819         IF_ADDR_LOCK(ifp);
  820         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
  821                 if_delmulti_locked(ifp, ifma, 1);
  822         IF_ADDR_UNLOCK(ifp);
  823 }
  824 
  825 /*
  826  * Detach an interface, removing it from the list of "active" interfaces.
  827  * If vmove flag is set on entry to if_detach_internal(), perform only a
  828  * limited subset of cleanup tasks, given that we are moving an ifnet from
  829  * one vnet to another, where it must be fully operational.
  830  *
  831  * XXXRW: There are some significant questions about event ordering, and
  832  * how to prevent things from starting to use the interface during detach.
  833  */
  834 void
  835 if_detach(struct ifnet *ifp)
  836 {
  837 
  838         if_detach_internal(ifp, 0);
  839 }
  840 
  841 static void
  842 if_detach_internal(struct ifnet *ifp, int vmove)
  843 {
  844         struct ifaddr *ifa;
  845         struct radix_node_head  *rnh;
  846         int i, j;
  847         struct domain *dp;
  848         struct ifnet *iter;
  849         int found = 0;
  850 
  851         IFNET_WLOCK();
  852         TAILQ_FOREACH(iter, &V_ifnet, if_link)
  853                 if (iter == ifp) {
  854                         TAILQ_REMOVE(&V_ifnet, ifp, if_link);
  855                         found = 1;
  856                         break;
  857                 }
  858 #ifdef VIMAGE
  859         if (found)
  860                 curvnet->vnet_ifcnt--;
  861 #endif
  862         IFNET_WUNLOCK();
  863         if (!found) {
  864                 if (vmove)
  865                         panic("%s: ifp=%p not on the ifnet tailq %p",
  866                             __func__, ifp, &V_ifnet);
  867                 else
  868                         return; /* XXX this should panic as well? */
  869         }
  870 
  871         /*
  872          * Remove/wait for pending events.
  873          */
  874         taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
  875 
  876         /*
  877          * Remove routes and flush queues.
  878          */
  879         if_down(ifp);
  880 #ifdef ALTQ
  881         if (ALTQ_IS_ENABLED(&ifp->if_snd))
  882                 altq_disable(&ifp->if_snd);
  883         if (ALTQ_IS_ATTACHED(&ifp->if_snd))
  884                 altq_detach(&ifp->if_snd);
  885 #endif
  886 
  887         if_purgeaddrs(ifp);
  888 
  889 #ifdef INET
  890         in_ifdetach(ifp);
  891 #endif
  892 
  893 #ifdef INET6
  894         /*
  895          * Remove all IPv6 kernel structs related to ifp.  This should be done
  896          * before removing routing entries below, since IPv6 interface direct
  897          * routes are expected to be removed by the IPv6-specific kernel API.
  898          * Otherwise, the kernel will detect some inconsistency and bark it.
  899          */
  900         in6_ifdetach(ifp);
  901 #endif
  902         if_purgemaddrs(ifp);
  903 
  904         if (!vmove) {
  905                 /*
  906                  * Prevent further calls into the device driver via ifnet.
  907                  */
  908                 if_dead(ifp);
  909 
  910                 /*
  911                  * Remove link ifaddr pointer and maybe decrement if_index.
  912                  * Clean up all addresses.
  913                  */
  914                 ifp->if_addr = NULL;
  915 
  916                 /* We can now free link ifaddr. */
  917                 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
  918                         ifa = TAILQ_FIRST(&ifp->if_addrhead);
  919                         TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  920                         ifa_free(ifa);
  921                 }
  922         }
  923 
  924         /*
  925          * Delete all remaining routes using this interface
  926          * Unfortuneatly the only way to do this is to slog through
  927          * the entire routing table looking for routes which point
  928          * to this interface...oh well...
  929          */
  930         for (i = 1; i <= AF_MAX; i++) {
  931                 for (j = 0; j < rt_numfibs; j++) {
  932                         rnh = rt_tables_get_rnh(j, i);
  933                         if (rnh == NULL)
  934                                 continue;
  935                         RADIX_NODE_HEAD_LOCK(rnh);
  936                         (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
  937                         RADIX_NODE_HEAD_UNLOCK(rnh);
  938                 }
  939         }
  940 
  941         /* Announce that the interface is gone. */
  942         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
  943         EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
  944         if (IS_DEFAULT_VNET(curvnet))
  945                 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
  946         if_delgroups(ifp);
  947 
  948         /*
  949          * We cannot hold the lock over dom_ifdetach calls as they might
  950          * sleep, for example trying to drain a callout, thus open up the
  951          * theoretical race with re-attaching.
  952          */
  953         IF_AFDATA_LOCK(ifp);
  954         i = ifp->if_afdata_initialized;
  955         ifp->if_afdata_initialized = 0;
  956         IF_AFDATA_UNLOCK(ifp);
  957         for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
  958                 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
  959                         (*dp->dom_ifdetach)(ifp,
  960                             ifp->if_afdata[dp->dom_family]);
  961         }
  962 }
  963 
  964 #ifdef VIMAGE
  965 /*
  966  * if_vmove() performs a limited version of if_detach() in current
  967  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
  968  * An attempt is made to shrink if_index in current vnet, find an
  969  * unused if_index in target vnet and calls if_grow() if necessary,
  970  * and finally find an unused if_xname for the target vnet.
  971  */
  972 void
  973 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
  974 {
  975         u_short idx;
  976 
  977         /*
  978          * Detach from current vnet, but preserve LLADDR info, do not
  979          * mark as dead etc. so that the ifnet can be reattached later.
  980          */
  981         if_detach_internal(ifp, 1);
  982 
  983         /*
  984          * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
  985          * the if_index for that vnet if possible.
  986          *
  987          * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
  988          * or we'd lock on one vnet and unlock on another.
  989          */
  990         IFNET_WLOCK();
  991         ifindex_free_locked(ifp->if_index);
  992         IFNET_WUNLOCK();
  993 
  994         /*
  995          * Perform interface-specific reassignment tasks, if provided by
  996          * the driver.
  997          */
  998         if (ifp->if_reassign != NULL)
  999                 ifp->if_reassign(ifp, new_vnet, NULL);
 1000 
 1001         /*
 1002          * Switch to the context of the target vnet.
 1003          */
 1004         CURVNET_SET_QUIET(new_vnet);
 1005 
 1006         IFNET_WLOCK();
 1007         if (ifindex_alloc_locked(&idx) != 0) {
 1008                 IFNET_WUNLOCK();
 1009                 panic("if_index overflow");
 1010         }
 1011         ifp->if_index = idx;
 1012         ifnet_setbyindex_locked(ifp->if_index, ifp);
 1013         IFNET_WUNLOCK();
 1014 
 1015         if_attach_internal(ifp, 1);
 1016 
 1017         CURVNET_RESTORE();
 1018 }
 1019 
 1020 /*
 1021  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
 1022  */
 1023 static int
 1024 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
 1025 {
 1026         struct prison *pr;
 1027         struct ifnet *difp;
 1028 
 1029         /* Try to find the prison within our visibility. */
 1030         sx_slock(&allprison_lock);
 1031         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1032         sx_sunlock(&allprison_lock);
 1033         if (pr == NULL)
 1034                 return (ENXIO);
 1035         prison_hold_locked(pr);
 1036         mtx_unlock(&pr->pr_mtx);
 1037 
 1038         /* Do not try to move the iface from and to the same prison. */
 1039         if (pr->pr_vnet == ifp->if_vnet) {
 1040                 prison_free(pr);
 1041                 return (EEXIST);
 1042         }
 1043 
 1044         /* Make sure the named iface does not exists in the dst. prison/vnet. */
 1045         /* XXX Lock interfaces to avoid races. */
 1046         CURVNET_SET_QUIET(pr->pr_vnet);
 1047         difp = ifunit(ifname);
 1048         CURVNET_RESTORE();
 1049         if (difp != NULL) {
 1050                 prison_free(pr);
 1051                 return (EEXIST);
 1052         }
 1053 
 1054         /* Move the interface into the child jail/vnet. */
 1055         if_vmove(ifp, pr->pr_vnet);
 1056 
 1057         /* Report the new if_xname back to the userland. */
 1058         sprintf(ifname, "%s", ifp->if_xname);
 1059 
 1060         prison_free(pr);
 1061         return (0);
 1062 }
 1063 
 1064 static int
 1065 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
 1066 {
 1067         struct prison *pr;
 1068         struct vnet *vnet_dst;
 1069         struct ifnet *ifp;
 1070 
 1071         /* Try to find the prison within our visibility. */
 1072         sx_slock(&allprison_lock);
 1073         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1074         sx_sunlock(&allprison_lock);
 1075         if (pr == NULL)
 1076                 return (ENXIO);
 1077         prison_hold_locked(pr);
 1078         mtx_unlock(&pr->pr_mtx);
 1079 
 1080         /* Make sure the named iface exists in the source prison/vnet. */
 1081         CURVNET_SET(pr->pr_vnet);
 1082         ifp = ifunit(ifname);           /* XXX Lock to avoid races. */
 1083         if (ifp == NULL) {
 1084                 CURVNET_RESTORE();
 1085                 prison_free(pr);
 1086                 return (ENXIO);
 1087         }
 1088 
 1089         /* Do not try to move the iface from and to the same prison. */
 1090         vnet_dst = TD_TO_VNET(td);
 1091         if (vnet_dst == ifp->if_vnet) {
 1092                 CURVNET_RESTORE();
 1093                 prison_free(pr);
 1094                 return (EEXIST);
 1095         }
 1096 
 1097         /* Get interface back from child jail/vnet. */
 1098         if_vmove(ifp, vnet_dst);
 1099         CURVNET_RESTORE();
 1100 
 1101         /* Report the new if_xname back to the userland. */
 1102         sprintf(ifname, "%s", ifp->if_xname);
 1103 
 1104         prison_free(pr);
 1105         return (0);
 1106 }
 1107 #endif /* VIMAGE */
 1108 
 1109 /*
 1110  * Add a group to an interface
 1111  */
 1112 int
 1113 if_addgroup(struct ifnet *ifp, const char *groupname)
 1114 {
 1115         struct ifg_list         *ifgl;
 1116         struct ifg_group        *ifg = NULL;
 1117         struct ifg_member       *ifgm;
 1118 
 1119         if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
 1120             groupname[strlen(groupname) - 1] <= '9')
 1121                 return (EINVAL);
 1122 
 1123         IFNET_WLOCK();
 1124         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1125                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
 1126                         IFNET_WUNLOCK();
 1127                         return (EEXIST);
 1128                 }
 1129 
 1130         if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
 1131             M_NOWAIT)) == NULL) {
 1132                 IFNET_WUNLOCK();
 1133                 return (ENOMEM);
 1134         }
 1135 
 1136         if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
 1137             M_TEMP, M_NOWAIT)) == NULL) {
 1138                 free(ifgl, M_TEMP);
 1139                 IFNET_WUNLOCK();
 1140                 return (ENOMEM);
 1141         }
 1142 
 1143         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1144                 if (!strcmp(ifg->ifg_group, groupname))
 1145                         break;
 1146 
 1147         if (ifg == NULL) {
 1148                 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
 1149                     M_TEMP, M_NOWAIT)) == NULL) {
 1150                         free(ifgl, M_TEMP);
 1151                         free(ifgm, M_TEMP);
 1152                         IFNET_WUNLOCK();
 1153                         return (ENOMEM);
 1154                 }
 1155                 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
 1156                 ifg->ifg_refcnt = 0;
 1157                 TAILQ_INIT(&ifg->ifg_members);
 1158                 EVENTHANDLER_INVOKE(group_attach_event, ifg);
 1159                 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
 1160         }
 1161 
 1162         ifg->ifg_refcnt++;
 1163         ifgl->ifgl_group = ifg;
 1164         ifgm->ifgm_ifp = ifp;
 1165 
 1166         IF_ADDR_LOCK(ifp);
 1167         TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
 1168         TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
 1169         IF_ADDR_UNLOCK(ifp);
 1170 
 1171         IFNET_WUNLOCK();
 1172 
 1173         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1174 
 1175         return (0);
 1176 }
 1177 
 1178 /*
 1179  * Remove a group from an interface
 1180  */
 1181 int
 1182 if_delgroup(struct ifnet *ifp, const char *groupname)
 1183 {
 1184         struct ifg_list         *ifgl;
 1185         struct ifg_member       *ifgm;
 1186 
 1187         IFNET_WLOCK();
 1188         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1189                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
 1190                         break;
 1191         if (ifgl == NULL) {
 1192                 IFNET_WUNLOCK();
 1193                 return (ENOENT);
 1194         }
 1195 
 1196         IF_ADDR_LOCK(ifp);
 1197         TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1198         IF_ADDR_UNLOCK(ifp);
 1199 
 1200         TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1201                 if (ifgm->ifgm_ifp == ifp)
 1202                         break;
 1203 
 1204         if (ifgm != NULL) {
 1205                 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
 1206                 free(ifgm, M_TEMP);
 1207         }
 1208 
 1209         if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1210                 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1211                 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
 1212                 free(ifgl->ifgl_group, M_TEMP);
 1213         }
 1214         IFNET_WUNLOCK();
 1215 
 1216         free(ifgl, M_TEMP);
 1217 
 1218         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1219 
 1220         return (0);
 1221 }
 1222 
 1223 /*
 1224  * Remove an interface from all groups
 1225  */
 1226 static void
 1227 if_delgroups(struct ifnet *ifp)
 1228 {
 1229         struct ifg_list         *ifgl;
 1230         struct ifg_member       *ifgm;
 1231         char groupname[IFNAMSIZ];
 1232 
 1233         IFNET_WLOCK();
 1234         while (!TAILQ_EMPTY(&ifp->if_groups)) {
 1235                 ifgl = TAILQ_FIRST(&ifp->if_groups);
 1236 
 1237                 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
 1238 
 1239                 IF_ADDR_LOCK(ifp);
 1240                 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1241                 IF_ADDR_UNLOCK(ifp);
 1242 
 1243                 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1244                         if (ifgm->ifgm_ifp == ifp)
 1245                                 break;
 1246 
 1247                 if (ifgm != NULL) {
 1248                         TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
 1249                             ifgm_next);
 1250                         free(ifgm, M_TEMP);
 1251                 }
 1252 
 1253                 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1254                         TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1255                         EVENTHANDLER_INVOKE(group_detach_event,
 1256                             ifgl->ifgl_group);
 1257                         free(ifgl->ifgl_group, M_TEMP);
 1258                 }
 1259                 IFNET_WUNLOCK();
 1260 
 1261                 free(ifgl, M_TEMP);
 1262 
 1263                 EVENTHANDLER_INVOKE(group_change_event, groupname);
 1264 
 1265                 IFNET_WLOCK();
 1266         }
 1267         IFNET_WUNLOCK();
 1268 }
 1269 
 1270 /*
 1271  * Stores all groups from an interface in memory pointed
 1272  * to by data
 1273  */
 1274 static int
 1275 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
 1276 {
 1277         int                      len, error;
 1278         struct ifg_list         *ifgl;
 1279         struct ifg_req           ifgrq, *ifgp;
 1280         struct ifgroupreq       *ifgr = data;
 1281 
 1282         if (ifgr->ifgr_len == 0) {
 1283                 IF_ADDR_LOCK(ifp);
 1284                 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1285                         ifgr->ifgr_len += sizeof(struct ifg_req);
 1286                 IF_ADDR_UNLOCK(ifp);
 1287                 return (0);
 1288         }
 1289 
 1290         len = ifgr->ifgr_len;
 1291         ifgp = ifgr->ifgr_groups;
 1292         /* XXX: wire */
 1293         IF_ADDR_LOCK(ifp);
 1294         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
 1295                 if (len < sizeof(ifgrq)) {
 1296                         IF_ADDR_UNLOCK(ifp);
 1297                         return (EINVAL);
 1298                 }
 1299                 bzero(&ifgrq, sizeof ifgrq);
 1300                 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
 1301                     sizeof(ifgrq.ifgrq_group));
 1302                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1303                         IF_ADDR_UNLOCK(ifp);
 1304                         return (error);
 1305                 }
 1306                 len -= sizeof(ifgrq);
 1307                 ifgp++;
 1308         }
 1309         IF_ADDR_UNLOCK(ifp);
 1310 
 1311         return (0);
 1312 }
 1313 
 1314 /*
 1315  * Stores all members of a group in memory pointed to by data
 1316  */
 1317 static int
 1318 if_getgroupmembers(struct ifgroupreq *data)
 1319 {
 1320         struct ifgroupreq       *ifgr = data;
 1321         struct ifg_group        *ifg;
 1322         struct ifg_member       *ifgm;
 1323         struct ifg_req           ifgrq, *ifgp;
 1324         int                      len, error;
 1325 
 1326         IFNET_RLOCK();
 1327         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1328                 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
 1329                         break;
 1330         if (ifg == NULL) {
 1331                 IFNET_RUNLOCK();
 1332                 return (ENOENT);
 1333         }
 1334 
 1335         if (ifgr->ifgr_len == 0) {
 1336                 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
 1337                         ifgr->ifgr_len += sizeof(ifgrq);
 1338                 IFNET_RUNLOCK();
 1339                 return (0);
 1340         }
 1341 
 1342         len = ifgr->ifgr_len;
 1343         ifgp = ifgr->ifgr_groups;
 1344         TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
 1345                 if (len < sizeof(ifgrq)) {
 1346                         IFNET_RUNLOCK();
 1347                         return (EINVAL);
 1348                 }
 1349                 bzero(&ifgrq, sizeof ifgrq);
 1350                 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
 1351                     sizeof(ifgrq.ifgrq_member));
 1352                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1353                         IFNET_RUNLOCK();
 1354                         return (error);
 1355                 }
 1356                 len -= sizeof(ifgrq);
 1357                 ifgp++;
 1358         }
 1359         IFNET_RUNLOCK();
 1360 
 1361         return (0);
 1362 }
 1363 
 1364 /*
 1365  * Delete Routes for a Network Interface
 1366  *
 1367  * Called for each routing entry via the rnh->rnh_walktree() call above
 1368  * to delete all route entries referencing a detaching network interface.
 1369  *
 1370  * Arguments:
 1371  *      rn      pointer to node in the routing table
 1372  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
 1373  *
 1374  * Returns:
 1375  *      0       successful
 1376  *      errno   failed - reason indicated
 1377  *
 1378  */
 1379 static int
 1380 if_rtdel(struct radix_node *rn, void *arg)
 1381 {
 1382         struct rtentry  *rt = (struct rtentry *)rn;
 1383         struct ifnet    *ifp = arg;
 1384         int             err;
 1385 
 1386         if (rt->rt_ifp == ifp) {
 1387 
 1388                 /*
 1389                  * Protect (sorta) against walktree recursion problems
 1390                  * with cloned routes
 1391                  */
 1392                 if ((rt->rt_flags & RTF_UP) == 0)
 1393                         return (0);
 1394 
 1395                 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
 1396                                 rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
 1397                                 (struct rtentry **) NULL, rt->rt_fibnum);
 1398                 if (err) {
 1399                         log(LOG_WARNING, "if_rtdel: error %d\n", err);
 1400                 }
 1401         }
 1402 
 1403         return (0);
 1404 }
 1405 
 1406 /*
 1407  * Wrapper functions for struct ifnet address list locking macros.  These are
 1408  * used by kernel modules to avoid encoding programming interface or binary
 1409  * interface assumptions that may be violated when kernel-internal locking
 1410  * approaches change.
 1411  */
 1412 void
 1413 if_addr_rlock(struct ifnet *ifp)
 1414 {
 1415 
 1416         IF_ADDR_LOCK(ifp);
 1417 }
 1418 
 1419 void
 1420 if_addr_runlock(struct ifnet *ifp)
 1421 {
 1422 
 1423         IF_ADDR_UNLOCK(ifp);
 1424 }
 1425 
 1426 void
 1427 if_maddr_rlock(struct ifnet *ifp)
 1428 {
 1429 
 1430         IF_ADDR_LOCK(ifp);
 1431 }
 1432 
 1433 void
 1434 if_maddr_runlock(struct ifnet *ifp)
 1435 {
 1436 
 1437         IF_ADDR_UNLOCK(ifp);
 1438 }
 1439 
 1440 /*
 1441  * Reference count functions for ifaddrs.
 1442  */
 1443 void
 1444 ifa_init(struct ifaddr *ifa)
 1445 {
 1446 
 1447         mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
 1448         refcount_init(&ifa->ifa_refcnt, 1);
 1449 }
 1450 
 1451 void
 1452 ifa_ref(struct ifaddr *ifa)
 1453 {
 1454 
 1455         refcount_acquire(&ifa->ifa_refcnt);
 1456 }
 1457 
 1458 void
 1459 ifa_free(struct ifaddr *ifa)
 1460 {
 1461 
 1462         if (refcount_release(&ifa->ifa_refcnt)) {
 1463                 mtx_destroy(&ifa->ifa_mtx);
 1464                 free(ifa, M_IFADDR);
 1465         }
 1466 }
 1467 
 1468 int
 1469 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1470 {
 1471         int error = 0;
 1472         struct rtentry *rt = NULL;
 1473         struct rt_addrinfo info;
 1474         static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1475 
 1476         bzero(&info, sizeof(info));
 1477         info.rti_ifp = V_loif;
 1478         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1479         info.rti_info[RTAX_DST] = ia;
 1480         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1481         error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
 1482 
 1483         if (error == 0 && rt != NULL) {
 1484                 RT_LOCK(rt);
 1485                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
 1486                         ifa->ifa_ifp->if_type;
 1487                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
 1488                         ifa->ifa_ifp->if_index;
 1489                 RT_REMREF(rt);
 1490                 RT_UNLOCK(rt);
 1491         } else if (error != 0)
 1492                 log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
 1493 
 1494         return (error);
 1495 }
 1496 
 1497 int
 1498 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1499 {
 1500         int error = 0;
 1501         struct rt_addrinfo info;
 1502         struct sockaddr_dl null_sdl;
 1503 
 1504         bzero(&null_sdl, sizeof(null_sdl));
 1505         null_sdl.sdl_len = sizeof(null_sdl);
 1506         null_sdl.sdl_family = AF_LINK;
 1507         null_sdl.sdl_type = ifa->ifa_ifp->if_type;
 1508         null_sdl.sdl_index = ifa->ifa_ifp->if_index;
 1509         bzero(&info, sizeof(info));
 1510         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1511         info.rti_info[RTAX_DST] = ia;
 1512         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1513         error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
 1514 
 1515         if (error != 0)
 1516                 log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
 1517 
 1518         return (error);
 1519 }
 1520 
 1521 /*
 1522  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
 1523  * structs used to represent other address families, it is necessary
 1524  * to perform a different comparison.
 1525  */
 1526 
 1527 #define sa_equal(a1, a2)        \
 1528         (bcmp((a1), (a2), ((a1))->sa_len) == 0)
 1529 
 1530 #define sa_dl_equal(a1, a2)     \
 1531         ((((struct sockaddr_dl *)(a1))->sdl_len ==                      \
 1532          ((struct sockaddr_dl *)(a2))->sdl_len) &&                      \
 1533          (bcmp(LLADDR((struct sockaddr_dl *)(a1)),                      \
 1534                LLADDR((struct sockaddr_dl *)(a2)),                      \
 1535                ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
 1536 
 1537 /*
 1538  * Locate an interface based on a complete address.
 1539  */
 1540 /*ARGSUSED*/
 1541 static struct ifaddr *
 1542 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
 1543 {
 1544         struct ifnet *ifp;
 1545         struct ifaddr *ifa;
 1546 
 1547         IFNET_RLOCK_NOSLEEP();
 1548         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1549                 IF_ADDR_LOCK(ifp);
 1550                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1551                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1552                                 continue;
 1553                         if (sa_equal(addr, ifa->ifa_addr)) {
 1554                                 if (getref)
 1555                                         ifa_ref(ifa);
 1556                                 IF_ADDR_UNLOCK(ifp);
 1557                                 goto done;
 1558                         }
 1559                         /* IP6 doesn't have broadcast */
 1560                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1561                             ifa->ifa_broadaddr &&
 1562                             ifa->ifa_broadaddr->sa_len != 0 &&
 1563                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1564                                 if (getref)
 1565                                         ifa_ref(ifa);
 1566                                 IF_ADDR_UNLOCK(ifp);
 1567                                 goto done;
 1568                         }
 1569                 }
 1570                 IF_ADDR_UNLOCK(ifp);
 1571         }
 1572         ifa = NULL;
 1573 done:
 1574         IFNET_RUNLOCK_NOSLEEP();
 1575         return (ifa);
 1576 }
 1577 
 1578 struct ifaddr *
 1579 ifa_ifwithaddr(struct sockaddr *addr)
 1580 {
 1581 
 1582         return (ifa_ifwithaddr_internal(addr, 1));
 1583 }
 1584 
 1585 int
 1586 ifa_ifwithaddr_check(struct sockaddr *addr)
 1587 {
 1588 
 1589         return (ifa_ifwithaddr_internal(addr, 0) != NULL);
 1590 }
 1591 
 1592 /*
 1593  * Locate an interface based on the broadcast address.
 1594  */
 1595 /* ARGSUSED */
 1596 struct ifaddr *
 1597 ifa_ifwithbroadaddr(struct sockaddr *addr)
 1598 {
 1599         struct ifnet *ifp;
 1600         struct ifaddr *ifa;
 1601 
 1602         IFNET_RLOCK_NOSLEEP();
 1603         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1604                 IF_ADDR_LOCK(ifp);
 1605                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1606                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1607                                 continue;
 1608                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1609                             ifa->ifa_broadaddr &&
 1610                             ifa->ifa_broadaddr->sa_len != 0 &&
 1611                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1612                                 ifa_ref(ifa);
 1613                                 IF_ADDR_UNLOCK(ifp);
 1614                                 goto done;
 1615                         }
 1616                 }
 1617                 IF_ADDR_UNLOCK(ifp);
 1618         }
 1619         ifa = NULL;
 1620 done:
 1621         IFNET_RUNLOCK_NOSLEEP();
 1622         return (ifa);
 1623 }
 1624 
 1625 /*
 1626  * Locate the point to point interface with a given destination address.
 1627  */
 1628 /*ARGSUSED*/
 1629 struct ifaddr *
 1630 ifa_ifwithdstaddr(struct sockaddr *addr)
 1631 {
 1632         struct ifnet *ifp;
 1633         struct ifaddr *ifa;
 1634 
 1635         IFNET_RLOCK_NOSLEEP();
 1636         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1637                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
 1638                         continue;
 1639                 IF_ADDR_LOCK(ifp);
 1640                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1641                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1642                                 continue;
 1643                         if (ifa->ifa_dstaddr != NULL &&
 1644                             sa_equal(addr, ifa->ifa_dstaddr)) {
 1645                                 ifa_ref(ifa);
 1646                                 IF_ADDR_UNLOCK(ifp);
 1647                                 goto done;
 1648                         }
 1649                 }
 1650                 IF_ADDR_UNLOCK(ifp);
 1651         }
 1652         ifa = NULL;
 1653 done:
 1654         IFNET_RUNLOCK_NOSLEEP();
 1655         return (ifa);
 1656 }
 1657 
 1658 /*
 1659  * Find an interface on a specific network.  If many, choice
 1660  * is most specific found.
 1661  */
 1662 struct ifaddr *
 1663 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
 1664 {
 1665         struct ifnet *ifp;
 1666         struct ifaddr *ifa;
 1667         struct ifaddr *ifa_maybe = NULL;
 1668         u_int af = addr->sa_family;
 1669         char *addr_data = addr->sa_data, *cplim;
 1670 
 1671         /*
 1672          * AF_LINK addresses can be looked up directly by their index number,
 1673          * so do that if we can.
 1674          */
 1675         if (af == AF_LINK) {
 1676             struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
 1677             if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
 1678                 return (ifaddr_byindex(sdl->sdl_index));
 1679         }
 1680 
 1681         /*
 1682          * Scan though each interface, looking for ones that have addresses
 1683          * in this address family.  Maintain a reference on ifa_maybe once
 1684          * we find one, as we release the IF_ADDR_LOCK() that kept it stable
 1685          * when we move onto the next interface.
 1686          */
 1687         IFNET_RLOCK_NOSLEEP();
 1688         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1689                 IF_ADDR_LOCK(ifp);
 1690                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1691                         char *cp, *cp2, *cp3;
 1692 
 1693                         if (ifa->ifa_addr->sa_family != af)
 1694 next:                           continue;
 1695                         if (af == AF_INET && 
 1696                             ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
 1697                                 /*
 1698                                  * This is a bit broken as it doesn't
 1699                                  * take into account that the remote end may
 1700                                  * be a single node in the network we are
 1701                                  * looking for.
 1702                                  * The trouble is that we don't know the
 1703                                  * netmask for the remote end.
 1704                                  */
 1705                                 if (ifa->ifa_dstaddr != NULL &&
 1706                                     sa_equal(addr, ifa->ifa_dstaddr)) {
 1707                                         ifa_ref(ifa);
 1708                                         IF_ADDR_UNLOCK(ifp);
 1709                                         goto done;
 1710                                 }
 1711                         } else {
 1712                                 /*
 1713                                  * if we have a special address handler,
 1714                                  * then use it instead of the generic one.
 1715                                  */
 1716                                 if (ifa->ifa_claim_addr) {
 1717                                         if ((*ifa->ifa_claim_addr)(ifa, addr)) {
 1718                                                 ifa_ref(ifa);
 1719                                                 IF_ADDR_UNLOCK(ifp);
 1720                                                 goto done;
 1721                                         }
 1722                                         continue;
 1723                                 }
 1724 
 1725                                 /*
 1726                                  * Scan all the bits in the ifa's address.
 1727                                  * If a bit dissagrees with what we are
 1728                                  * looking for, mask it with the netmask
 1729                                  * to see if it really matters.
 1730                                  * (A byte at a time)
 1731                                  */
 1732                                 if (ifa->ifa_netmask == 0)
 1733                                         continue;
 1734                                 cp = addr_data;
 1735                                 cp2 = ifa->ifa_addr->sa_data;
 1736                                 cp3 = ifa->ifa_netmask->sa_data;
 1737                                 cplim = ifa->ifa_netmask->sa_len
 1738                                         + (char *)ifa->ifa_netmask;
 1739                                 while (cp3 < cplim)
 1740                                         if ((*cp++ ^ *cp2++) & *cp3++)
 1741                                                 goto next; /* next address! */
 1742                                 /*
 1743                                  * If the netmask of what we just found
 1744                                  * is more specific than what we had before
 1745                                  * (if we had one) then remember the new one
 1746                                  * before continuing to search
 1747                                  * for an even better one.
 1748                                  */
 1749                                 if (ifa_maybe == NULL ||
 1750                                     rn_refines((caddr_t)ifa->ifa_netmask,
 1751                                     (caddr_t)ifa_maybe->ifa_netmask)) {
 1752                                         if (ifa_maybe != NULL)
 1753                                                 ifa_free(ifa_maybe);
 1754                                         ifa_maybe = ifa;
 1755                                         ifa_ref(ifa_maybe);
 1756                                 }
 1757                         }
 1758                 }
 1759                 IF_ADDR_UNLOCK(ifp);
 1760         }
 1761         ifa = ifa_maybe;
 1762         ifa_maybe = NULL;
 1763 done:
 1764         IFNET_RUNLOCK_NOSLEEP();
 1765         if (ifa_maybe != NULL)
 1766                 ifa_free(ifa_maybe);
 1767         return (ifa);
 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_LOCK(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_UNLOCK(ifp);
 1818         return (ifa);
 1819 }
 1820 
 1821 #include <net/if_llatbl.h>
 1822 
 1823 /*
 1824  * Default action when installing a route with a Link Level gateway.
 1825  * Lookup an appropriate real ifa to point to.
 1826  * This should be moved to /sys/net/link.c eventually.
 1827  */
 1828 static void
 1829 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
 1830 {
 1831         struct ifaddr *ifa, *oifa;
 1832         struct sockaddr *dst;
 1833         struct ifnet *ifp;
 1834 
 1835         RT_LOCK_ASSERT(rt);
 1836 
 1837         if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
 1838             ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
 1839                 return;
 1840         ifa = ifaof_ifpforaddr(dst, ifp);
 1841         if (ifa) {
 1842                 oifa = rt->rt_ifa;
 1843                 rt->rt_ifa = ifa;
 1844                 ifa_free(oifa);
 1845                 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
 1846                         ifa->ifa_rtrequest(cmd, rt, info);
 1847         }
 1848 }
 1849 
 1850 /*
 1851  * Mark an interface down and notify protocols of
 1852  * the transition.
 1853  * NOTE: must be called at splnet or eqivalent.
 1854  */
 1855 static void
 1856 if_unroute(struct ifnet *ifp, int flag, int fam)
 1857 {
 1858         struct ifaddr *ifa;
 1859 
 1860         KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
 1861 
 1862         ifp->if_flags &= ~flag;
 1863         getmicrotime(&ifp->if_lastchange);
 1864         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1865                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1866                         pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
 1867         ifp->if_qflush(ifp);
 1868 
 1869         if (ifp->if_carp)
 1870                 (*carp_linkstate_p)(ifp);
 1871         rt_ifmsg(ifp);
 1872 }
 1873 
 1874 /*
 1875  * Mark an interface up and notify protocols of
 1876  * the transition.
 1877  * NOTE: must be called at splnet or eqivalent.
 1878  */
 1879 static void
 1880 if_route(struct ifnet *ifp, int flag, int fam)
 1881 {
 1882         struct ifaddr *ifa;
 1883 
 1884         KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
 1885 
 1886         ifp->if_flags |= flag;
 1887         getmicrotime(&ifp->if_lastchange);
 1888         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1889                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1890                         pfctlinput(PRC_IFUP, ifa->ifa_addr);
 1891         if (ifp->if_carp)
 1892                 (*carp_linkstate_p)(ifp);
 1893         rt_ifmsg(ifp);
 1894 #ifdef INET6
 1895         in6_if_up(ifp);
 1896 #endif
 1897 }
 1898 
 1899 void    (*vlan_link_state_p)(struct ifnet *, int);      /* XXX: private from if_vlan */
 1900 void    (*vlan_trunk_cap_p)(struct ifnet *);            /* XXX: private from if_vlan */
 1901 
 1902 /*
 1903  * Handle a change in the interface link state. To avoid LORs
 1904  * between driver lock and upper layer locks, as well as possible
 1905  * recursions, we post event to taskqueue, and all job
 1906  * is done in static do_link_state_change().
 1907  */
 1908 void
 1909 if_link_state_change(struct ifnet *ifp, int link_state)
 1910 {
 1911         /* Return if state hasn't changed. */
 1912         if (ifp->if_link_state == link_state)
 1913                 return;
 1914 
 1915         ifp->if_link_state = link_state;
 1916 
 1917         taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
 1918 }
 1919 
 1920 static void
 1921 do_link_state_change(void *arg, int pending)
 1922 {
 1923         struct ifnet *ifp = (struct ifnet *)arg;
 1924         int link_state = ifp->if_link_state;
 1925         CURVNET_SET(ifp->if_vnet);
 1926 
 1927         /* Notify that the link state has changed. */
 1928         rt_ifmsg(ifp);
 1929         if (ifp->if_vlantrunk != NULL)
 1930                 (*vlan_link_state_p)(ifp, 0);
 1931 
 1932         if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
 1933             IFP2AC(ifp)->ac_netgraph != NULL)
 1934                 (*ng_ether_link_state_p)(ifp, link_state);
 1935         if (ifp->if_carp)
 1936                 (*carp_linkstate_p)(ifp);
 1937         if (ifp->if_bridge) {
 1938                 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
 1939                 (*bstp_linkstate_p)(ifp, link_state);
 1940         }
 1941         if (ifp->if_lagg) {
 1942                 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
 1943                 (*lagg_linkstate_p)(ifp, link_state);
 1944         }
 1945 
 1946         if (IS_DEFAULT_VNET(curvnet))
 1947                 devctl_notify("IFNET", ifp->if_xname,
 1948                     (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
 1949                     NULL);
 1950         if (pending > 1)
 1951                 if_printf(ifp, "%d link states coalesced\n", pending);
 1952         if (log_link_state_change)
 1953                 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
 1954                     (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
 1955         CURVNET_RESTORE();
 1956 }
 1957 
 1958 /*
 1959  * Mark an interface down and notify protocols of
 1960  * the transition.
 1961  * NOTE: must be called at splnet or eqivalent.
 1962  */
 1963 void
 1964 if_down(struct ifnet *ifp)
 1965 {
 1966 
 1967         if_unroute(ifp, IFF_UP, AF_UNSPEC);
 1968 }
 1969 
 1970 /*
 1971  * Mark an interface up and notify protocols of
 1972  * the transition.
 1973  * NOTE: must be called at splnet or eqivalent.
 1974  */
 1975 void
 1976 if_up(struct ifnet *ifp)
 1977 {
 1978 
 1979         if_route(ifp, IFF_UP, AF_UNSPEC);
 1980 }
 1981 
 1982 /*
 1983  * Flush an interface queue.
 1984  */
 1985 void
 1986 if_qflush(struct ifnet *ifp)
 1987 {
 1988         struct mbuf *m, *n;
 1989         struct ifaltq *ifq;
 1990         
 1991         ifq = &ifp->if_snd;
 1992         IFQ_LOCK(ifq);
 1993 #ifdef ALTQ
 1994         if (ALTQ_IS_ENABLED(ifq))
 1995                 ALTQ_PURGE(ifq);
 1996 #endif
 1997         n = ifq->ifq_head;
 1998         while ((m = n) != 0) {
 1999                 n = m->m_act;
 2000                 m_freem(m);
 2001         }
 2002         ifq->ifq_head = 0;
 2003         ifq->ifq_tail = 0;
 2004         ifq->ifq_len = 0;
 2005         IFQ_UNLOCK(ifq);
 2006 }
 2007 
 2008 /*
 2009  * Handle interface watchdog timer routines.  Called
 2010  * from softclock, we decrement timers (if set) and
 2011  * call the appropriate interface routine on expiration.
 2012  *
 2013  * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
 2014  * holding Giant.
 2015  */
 2016 static void
 2017 if_slowtimo(void *arg)
 2018 {
 2019         VNET_ITERATOR_DECL(vnet_iter);
 2020         struct ifnet *ifp;
 2021         int s = splimp();
 2022 
 2023         VNET_LIST_RLOCK_NOSLEEP();
 2024         IFNET_RLOCK_NOSLEEP();
 2025         VNET_FOREACH(vnet_iter) {
 2026                 CURVNET_SET(vnet_iter);
 2027                 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2028                         if (ifp->if_timer == 0 || --ifp->if_timer)
 2029                                 continue;
 2030                         if (ifp->if_watchdog)
 2031                                 (*ifp->if_watchdog)(ifp);
 2032                 }
 2033                 CURVNET_RESTORE();
 2034         }
 2035         IFNET_RUNLOCK_NOSLEEP();
 2036         VNET_LIST_RUNLOCK_NOSLEEP();
 2037         splx(s);
 2038         timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
 2039 }
 2040 
 2041 /*
 2042  * Map interface name to interface structure pointer, with or without
 2043  * returning a reference.
 2044  */
 2045 struct ifnet *
 2046 ifunit_ref(const char *name)
 2047 {
 2048         struct ifnet *ifp;
 2049 
 2050         IFNET_RLOCK_NOSLEEP();
 2051         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2052                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
 2053                     !(ifp->if_flags & IFF_DYING))
 2054                         break;
 2055         }
 2056         if (ifp != NULL)
 2057                 if_ref(ifp);
 2058         IFNET_RUNLOCK_NOSLEEP();
 2059         return (ifp);
 2060 }
 2061 
 2062 struct ifnet *
 2063 ifunit(const char *name)
 2064 {
 2065         struct ifnet *ifp;
 2066 
 2067         IFNET_RLOCK_NOSLEEP();
 2068         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2069                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
 2070                         break;
 2071         }
 2072         IFNET_RUNLOCK_NOSLEEP();
 2073         return (ifp);
 2074 }
 2075 
 2076 /*
 2077  * Hardware specific interface ioctls.
 2078  */
 2079 static int
 2080 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
 2081 {
 2082         struct ifreq *ifr;
 2083         struct ifstat *ifs;
 2084         int error = 0;
 2085         int new_flags, temp_flags;
 2086         size_t namelen, onamelen;
 2087         size_t descrlen;
 2088         char *descrbuf, *odescrbuf;
 2089         char new_name[IFNAMSIZ];
 2090         struct ifaddr *ifa;
 2091         struct sockaddr_dl *sdl;
 2092 
 2093         ifr = (struct ifreq *)data;
 2094         switch (cmd) {
 2095         case SIOCGIFINDEX:
 2096                 ifr->ifr_index = ifp->if_index;
 2097                 break;
 2098 
 2099         case SIOCGIFFLAGS:
 2100                 temp_flags = ifp->if_flags | ifp->if_drv_flags;
 2101                 ifr->ifr_flags = temp_flags & 0xffff;
 2102                 ifr->ifr_flagshigh = temp_flags >> 16;
 2103                 break;
 2104 
 2105         case SIOCGIFCAP:
 2106                 ifr->ifr_reqcap = ifp->if_capabilities;
 2107                 ifr->ifr_curcap = ifp->if_capenable;
 2108                 break;
 2109 
 2110 #ifdef MAC
 2111         case SIOCGIFMAC:
 2112                 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
 2113                 break;
 2114 #endif
 2115 
 2116         case SIOCGIFMETRIC:
 2117                 ifr->ifr_metric = ifp->if_metric;
 2118                 break;
 2119 
 2120         case SIOCGIFMTU:
 2121                 ifr->ifr_mtu = ifp->if_mtu;
 2122                 break;
 2123 
 2124         case SIOCGIFPHYS:
 2125                 ifr->ifr_phys = ifp->if_physical;
 2126                 break;
 2127 
 2128         case SIOCGIFDESCR:
 2129                 error = 0;
 2130                 sx_slock(&ifdescr_sx);
 2131                 if (ifp->if_description == NULL)
 2132                         error = ENOMSG;
 2133                 else {
 2134                         /* space for terminating nul */
 2135                         descrlen = strlen(ifp->if_description) + 1;
 2136                         if (ifr->ifr_buffer.length < descrlen)
 2137                                 ifr->ifr_buffer.buffer = NULL;
 2138                         else
 2139                                 error = copyout(ifp->if_description,
 2140                                     ifr->ifr_buffer.buffer, descrlen);
 2141                         ifr->ifr_buffer.length = descrlen;
 2142                 }
 2143                 sx_sunlock(&ifdescr_sx);
 2144                 break;
 2145 
 2146         case SIOCSIFDESCR:
 2147                 error = priv_check(td, PRIV_NET_SETIFDESCR);
 2148                 if (error)
 2149                         return (error);
 2150 
 2151                 /*
 2152                  * Copy only (length-1) bytes to make sure that
 2153                  * if_description is always nul terminated.  The
 2154                  * length parameter is supposed to count the
 2155                  * terminating nul in.
 2156                  */
 2157                 if (ifr->ifr_buffer.length > ifdescr_maxlen)
 2158                         return (ENAMETOOLONG);
 2159                 else if (ifr->ifr_buffer.length == 0)
 2160                         descrbuf = NULL;
 2161                 else {
 2162                         descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
 2163                             M_WAITOK | M_ZERO);
 2164                         error = copyin(ifr->ifr_buffer.buffer, descrbuf,
 2165                             ifr->ifr_buffer.length - 1);
 2166                         if (error) {
 2167                                 free(descrbuf, M_IFDESCR);
 2168                                 break;
 2169                         }
 2170                 }
 2171 
 2172                 sx_xlock(&ifdescr_sx);
 2173                 odescrbuf = ifp->if_description;
 2174                 ifp->if_description = descrbuf;
 2175                 sx_xunlock(&ifdescr_sx);
 2176 
 2177                 getmicrotime(&ifp->if_lastchange);
 2178                 free(odescrbuf, M_IFDESCR);
 2179                 break;
 2180 
 2181         case SIOCSIFFLAGS:
 2182                 error = priv_check(td, PRIV_NET_SETIFFLAGS);
 2183                 if (error)
 2184                         return (error);
 2185                 /*
 2186                  * Currently, no driver owned flags pass the IFF_CANTCHANGE
 2187                  * check, so we don't need special handling here yet.
 2188                  */
 2189                 new_flags = (ifr->ifr_flags & 0xffff) |
 2190                     (ifr->ifr_flagshigh << 16);
 2191                 if (ifp->if_flags & IFF_SMART) {
 2192                         /* Smart drivers twiddle their own routes */
 2193                 } else if (ifp->if_flags & IFF_UP &&
 2194                     (new_flags & IFF_UP) == 0) {
 2195                         int s = splimp();
 2196                         if_down(ifp);
 2197                         splx(s);
 2198                 } else if (new_flags & IFF_UP &&
 2199                     (ifp->if_flags & IFF_UP) == 0) {
 2200                         int s = splimp();
 2201                         if_up(ifp);
 2202                         splx(s);
 2203                 }
 2204                 /* See if permanently promiscuous mode bit is about to flip */
 2205                 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
 2206                         if (new_flags & IFF_PPROMISC)
 2207                                 ifp->if_flags |= IFF_PROMISC;
 2208                         else if (ifp->if_pcount == 0)
 2209                                 ifp->if_flags &= ~IFF_PROMISC;
 2210                         log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
 2211                             ifp->if_xname,
 2212                             (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
 2213                 }
 2214                 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
 2215                         (new_flags &~ IFF_CANTCHANGE);
 2216                 if (ifp->if_ioctl) {
 2217                         (void) (*ifp->if_ioctl)(ifp, cmd, data);
 2218                 }
 2219                 getmicrotime(&ifp->if_lastchange);
 2220                 break;
 2221 
 2222         case SIOCSIFCAP:
 2223                 error = priv_check(td, PRIV_NET_SETIFCAP);
 2224                 if (error)
 2225                         return (error);
 2226                 if (ifp->if_ioctl == NULL)
 2227                         return (EOPNOTSUPP);
 2228                 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
 2229                         return (EINVAL);
 2230                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2231                 if (error == 0)
 2232                         getmicrotime(&ifp->if_lastchange);
 2233                 break;
 2234 
 2235 #ifdef MAC
 2236         case SIOCSIFMAC:
 2237                 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
 2238                 break;
 2239 #endif
 2240 
 2241         case SIOCSIFNAME:
 2242                 error = priv_check(td, PRIV_NET_SETIFNAME);
 2243                 if (error)
 2244                         return (error);
 2245                 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
 2246                 if (error != 0)
 2247                         return (error);
 2248                 if (new_name[0] == '\0')
 2249                         return (EINVAL);
 2250                 if (ifunit(new_name) != NULL)
 2251                         return (EEXIST);
 2252 
 2253                 /*
 2254                  * XXX: Locking.  Nothing else seems to lock if_flags,
 2255                  * and there are numerous other races with the
 2256                  * ifunit() checks not being atomic with namespace
 2257                  * changes (renames, vmoves, if_attach, etc).
 2258                  */
 2259                 ifp->if_flags |= IFF_RENAMING;
 2260                 
 2261                 /* Announce the departure of the interface. */
 2262                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 2263                 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 2264 
 2265                 log(LOG_INFO, "%s: changing name to '%s'\n",
 2266                     ifp->if_xname, new_name);
 2267 
 2268                 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
 2269                 ifa = ifp->if_addr;
 2270                 IFA_LOCK(ifa);
 2271                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 2272                 namelen = strlen(new_name);
 2273                 onamelen = sdl->sdl_nlen;
 2274                 /*
 2275                  * Move the address if needed.  This is safe because we
 2276                  * allocate space for a name of length IFNAMSIZ when we
 2277                  * create this in if_attach().
 2278                  */
 2279                 if (namelen != onamelen) {
 2280                         bcopy(sdl->sdl_data + onamelen,
 2281                             sdl->sdl_data + namelen, sdl->sdl_alen);
 2282                 }
 2283                 bcopy(new_name, sdl->sdl_data, namelen);
 2284                 sdl->sdl_nlen = namelen;
 2285                 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
 2286                 bzero(sdl->sdl_data, onamelen);
 2287                 while (namelen != 0)
 2288                         sdl->sdl_data[--namelen] = 0xff;
 2289                 IFA_UNLOCK(ifa);
 2290 
 2291                 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
 2292                 /* Announce the return of the interface. */
 2293                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 2294 
 2295                 ifp->if_flags &= ~IFF_RENAMING;
 2296                 break;
 2297 
 2298 #ifdef VIMAGE
 2299         case SIOCSIFVNET:
 2300                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2301                 if (error)
 2302                         return (error);
 2303                 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
 2304                 break;
 2305 #endif
 2306 
 2307         case SIOCSIFMETRIC:
 2308                 error = priv_check(td, PRIV_NET_SETIFMETRIC);
 2309                 if (error)
 2310                         return (error);
 2311                 ifp->if_metric = ifr->ifr_metric;
 2312                 getmicrotime(&ifp->if_lastchange);
 2313                 break;
 2314 
 2315         case SIOCSIFPHYS:
 2316                 error = priv_check(td, PRIV_NET_SETIFPHYS);
 2317                 if (error)
 2318                         return (error);
 2319                 if (ifp->if_ioctl == NULL)
 2320                         return (EOPNOTSUPP);
 2321                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2322                 if (error == 0)
 2323                         getmicrotime(&ifp->if_lastchange);
 2324                 break;
 2325 
 2326         case SIOCSIFMTU:
 2327         {
 2328                 u_long oldmtu = ifp->if_mtu;
 2329 
 2330                 error = priv_check(td, PRIV_NET_SETIFMTU);
 2331                 if (error)
 2332                         return (error);
 2333                 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
 2334                         return (EINVAL);
 2335                 if (ifp->if_ioctl == NULL)
 2336                         return (EOPNOTSUPP);
 2337                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2338                 if (error == 0) {
 2339                         getmicrotime(&ifp->if_lastchange);
 2340                         rt_ifmsg(ifp);
 2341                 }
 2342                 /*
 2343                  * If the link MTU changed, do network layer specific procedure.
 2344                  */
 2345                 if (ifp->if_mtu != oldmtu) {
 2346 #ifdef INET6
 2347                         nd6_setmtu(ifp);
 2348 #endif
 2349                 }
 2350                 break;
 2351         }
 2352 
 2353         case SIOCADDMULTI:
 2354         case SIOCDELMULTI:
 2355                 if (cmd == SIOCADDMULTI)
 2356                         error = priv_check(td, PRIV_NET_ADDMULTI);
 2357                 else
 2358                         error = priv_check(td, PRIV_NET_DELMULTI);
 2359                 if (error)
 2360                         return (error);
 2361 
 2362                 /* Don't allow group membership on non-multicast interfaces. */
 2363                 if ((ifp->if_flags & IFF_MULTICAST) == 0)
 2364                         return (EOPNOTSUPP);
 2365 
 2366                 /* Don't let users screw up protocols' entries. */
 2367                 if (ifr->ifr_addr.sa_family != AF_LINK)
 2368                         return (EINVAL);
 2369 
 2370                 if (cmd == SIOCADDMULTI) {
 2371                         struct ifmultiaddr *ifma;
 2372 
 2373                         /*
 2374                          * Userland is only permitted to join groups once
 2375                          * via the if_addmulti() KPI, because it cannot hold
 2376                          * struct ifmultiaddr * between calls. It may also
 2377                          * lose a race while we check if the membership
 2378                          * already exists.
 2379                          */
 2380                         IF_ADDR_LOCK(ifp);
 2381                         ifma = if_findmulti(ifp, &ifr->ifr_addr);
 2382                         IF_ADDR_UNLOCK(ifp);
 2383                         if (ifma != NULL)
 2384                                 error = EADDRINUSE;
 2385                         else
 2386                                 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
 2387                 } else {
 2388                         error = if_delmulti(ifp, &ifr->ifr_addr);
 2389                 }
 2390                 if (error == 0)
 2391                         getmicrotime(&ifp->if_lastchange);
 2392                 break;
 2393 
 2394         case SIOCSIFPHYADDR:
 2395         case SIOCDIFPHYADDR:
 2396 #ifdef INET6
 2397         case SIOCSIFPHYADDR_IN6:
 2398 #endif
 2399         case SIOCSLIFPHYADDR:
 2400         case SIOCSIFMEDIA:
 2401         case SIOCSIFGENERIC:
 2402                 error = priv_check(td, PRIV_NET_HWIOCTL);
 2403                 if (error)
 2404                         return (error);
 2405                 if (ifp->if_ioctl == NULL)
 2406                         return (EOPNOTSUPP);
 2407                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2408                 if (error == 0)
 2409                         getmicrotime(&ifp->if_lastchange);
 2410                 break;
 2411 
 2412         case SIOCGIFSTATUS:
 2413                 ifs = (struct ifstat *)data;
 2414                 ifs->ascii[0] = '\0';
 2415 
 2416         case SIOCGIFPSRCADDR:
 2417         case SIOCGIFPDSTADDR:
 2418         case SIOCGLIFPHYADDR:
 2419         case SIOCGIFMEDIA:
 2420         case SIOCGIFGENERIC:
 2421                 if (ifp->if_ioctl == NULL)
 2422                         return (EOPNOTSUPP);
 2423                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2424                 break;
 2425 
 2426         case SIOCSIFLLADDR:
 2427                 error = priv_check(td, PRIV_NET_SETLLADDR);
 2428                 if (error)
 2429                         return (error);
 2430                 error = if_setlladdr(ifp,
 2431                     ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
 2432                 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
 2433                 break;
 2434 
 2435         case SIOCAIFGROUP:
 2436         {
 2437                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2438 
 2439                 error = priv_check(td, PRIV_NET_ADDIFGROUP);
 2440                 if (error)
 2441                         return (error);
 2442                 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
 2443                         return (error);
 2444                 break;
 2445         }
 2446 
 2447         case SIOCGIFGROUP:
 2448                 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
 2449                         return (error);
 2450                 break;
 2451 
 2452         case SIOCDIFGROUP:
 2453         {
 2454                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2455 
 2456                 error = priv_check(td, PRIV_NET_DELIFGROUP);
 2457                 if (error)
 2458                         return (error);
 2459                 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
 2460                         return (error);
 2461                 break;
 2462         }
 2463 
 2464         default:
 2465                 error = ENOIOCTL;
 2466                 break;
 2467         }
 2468         return (error);
 2469 }
 2470 
 2471 #ifdef COMPAT_FREEBSD32
 2472 struct ifconf32 {
 2473         int32_t ifc_len;
 2474         union {
 2475                 uint32_t        ifcu_buf;
 2476                 uint32_t        ifcu_req;
 2477         } ifc_ifcu;
 2478 };
 2479 #define SIOCGIFCONF32   _IOWR('i', 36, struct ifconf32)
 2480 #endif
 2481 
 2482 /*
 2483  * Interface ioctls.
 2484  */
 2485 int
 2486 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
 2487 {
 2488         struct ifnet *ifp;
 2489         struct ifreq *ifr;
 2490         int error;
 2491         int oif_flags;
 2492 
 2493         switch (cmd) {
 2494         case SIOCGIFCONF:
 2495         case OSIOCGIFCONF:
 2496                 return (ifconf(cmd, data));
 2497 
 2498 #ifdef COMPAT_FREEBSD32
 2499         case SIOCGIFCONF32:
 2500                 {
 2501                         struct ifconf32 *ifc32;
 2502                         struct ifconf ifc;
 2503 
 2504                         ifc32 = (struct ifconf32 *)data;
 2505                         ifc.ifc_len = ifc32->ifc_len;
 2506                         ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
 2507 
 2508                         return (ifconf(SIOCGIFCONF, (void *)&ifc));
 2509                 }
 2510 #endif
 2511         }
 2512         ifr = (struct ifreq *)data;
 2513 
 2514         switch (cmd) {
 2515 #ifdef VIMAGE
 2516         case SIOCSIFRVNET:
 2517                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2518                 if (error)
 2519                         return (error);
 2520                 return (if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid));
 2521 #endif
 2522         case SIOCIFCREATE:
 2523         case SIOCIFCREATE2:
 2524                 error = priv_check(td, PRIV_NET_IFCREATE);
 2525                 if (error)
 2526                         return (error);
 2527                 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
 2528                         cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
 2529         case SIOCIFDESTROY:
 2530                 error = priv_check(td, PRIV_NET_IFDESTROY);
 2531                 if (error)
 2532                         return (error);
 2533                 return if_clone_destroy(ifr->ifr_name);
 2534 
 2535         case SIOCIFGCLONERS:
 2536                 return (if_clone_list((struct if_clonereq *)data));
 2537         case SIOCGIFGMEMB:
 2538                 return (if_getgroupmembers((struct ifgroupreq *)data));
 2539         }
 2540 
 2541         ifp = ifunit_ref(ifr->ifr_name);
 2542         if (ifp == NULL)
 2543                 return (ENXIO);
 2544 
 2545         error = ifhwioctl(cmd, ifp, data, td);
 2546         if (error != ENOIOCTL) {
 2547                 if_rele(ifp);
 2548                 return (error);
 2549         }
 2550 
 2551         oif_flags = ifp->if_flags;
 2552         if (so->so_proto == NULL) {
 2553                 if_rele(ifp);
 2554                 return (EOPNOTSUPP);
 2555         }
 2556 #ifndef COMPAT_43
 2557         error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
 2558                                                                  data,
 2559                                                                  ifp, td));
 2560         if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
 2561                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2562 #else
 2563         {
 2564                 u_long ocmd = cmd;
 2565 
 2566                 switch (cmd) {
 2567 
 2568                 case SIOCSIFDSTADDR:
 2569                 case SIOCSIFADDR:
 2570                 case SIOCSIFBRDADDR:
 2571                 case SIOCSIFNETMASK:
 2572 #if BYTE_ORDER != BIG_ENDIAN
 2573                         if (ifr->ifr_addr.sa_family == 0 &&
 2574                             ifr->ifr_addr.sa_len < 16) {
 2575                                 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
 2576                                 ifr->ifr_addr.sa_len = 16;
 2577                         }
 2578 #else
 2579                         if (ifr->ifr_addr.sa_len == 0)
 2580                                 ifr->ifr_addr.sa_len = 16;
 2581 #endif
 2582                         break;
 2583 
 2584                 case OSIOCGIFADDR:
 2585                         cmd = SIOCGIFADDR;
 2586                         break;
 2587 
 2588                 case OSIOCGIFDSTADDR:
 2589                         cmd = SIOCGIFDSTADDR;
 2590                         break;
 2591 
 2592                 case OSIOCGIFBRDADDR:
 2593                         cmd = SIOCGIFBRDADDR;
 2594                         break;
 2595 
 2596                 case OSIOCGIFNETMASK:
 2597                         cmd = SIOCGIFNETMASK;
 2598                 }
 2599                 error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
 2600                                                                    cmd,
 2601                                                                    data,
 2602                                                                    ifp, td));
 2603                 if (error == EOPNOTSUPP && ifp != NULL &&
 2604                     ifp->if_ioctl != NULL)
 2605                         error = (*ifp->if_ioctl)(ifp, cmd, data);
 2606                 switch (ocmd) {
 2607 
 2608                 case OSIOCGIFADDR:
 2609                 case OSIOCGIFDSTADDR:
 2610                 case OSIOCGIFBRDADDR:
 2611                 case OSIOCGIFNETMASK:
 2612                         *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
 2613 
 2614                 }
 2615         }
 2616 #endif /* COMPAT_43 */
 2617 
 2618         if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
 2619 #ifdef INET6
 2620                 if (ifp->if_flags & IFF_UP) {
 2621                         int s = splimp();
 2622                         in6_if_up(ifp);
 2623                         splx(s);
 2624                 }
 2625 #endif
 2626         }
 2627         if_rele(ifp);
 2628         return (error);
 2629 }
 2630 
 2631 /*
 2632  * The code common to handling reference counted flags,
 2633  * e.g., in ifpromisc() and if_allmulti().
 2634  * The "pflag" argument can specify a permanent mode flag to check,
 2635  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
 2636  *
 2637  * Only to be used on stack-owned flags, not driver-owned flags.
 2638  */
 2639 static int
 2640 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
 2641 {
 2642         struct ifreq ifr;
 2643         int error;
 2644         int oldflags, oldcount;
 2645 
 2646         /* Sanity checks to catch programming errors */
 2647         KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
 2648             ("%s: setting driver-owned flag %d", __func__, flag));
 2649 
 2650         if (onswitch)
 2651                 KASSERT(*refcount >= 0,
 2652                     ("%s: increment negative refcount %d for flag %d",
 2653                     __func__, *refcount, flag));
 2654         else
 2655                 KASSERT(*refcount > 0,
 2656                     ("%s: decrement non-positive refcount %d for flag %d",
 2657                     __func__, *refcount, flag));
 2658 
 2659         /* In case this mode is permanent, just touch refcount */
 2660         if (ifp->if_flags & pflag) {
 2661                 *refcount += onswitch ? 1 : -1;
 2662                 return (0);
 2663         }
 2664 
 2665         /* Save ifnet parameters for if_ioctl() may fail */
 2666         oldcount = *refcount;
 2667         oldflags = ifp->if_flags;
 2668         
 2669         /*
 2670          * See if we aren't the only and touching refcount is enough.
 2671          * Actually toggle interface flag if we are the first or last.
 2672          */
 2673         if (onswitch) {
 2674                 if ((*refcount)++)
 2675                         return (0);
 2676                 ifp->if_flags |= flag;
 2677         } else {
 2678                 if (--(*refcount))
 2679                         return (0);
 2680                 ifp->if_flags &= ~flag;
 2681         }
 2682 
 2683         /* Call down the driver since we've changed interface flags */
 2684         if (ifp->if_ioctl == NULL) {
 2685                 error = EOPNOTSUPP;
 2686                 goto recover;
 2687         }
 2688         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2689         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2690         error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2691         if (error)
 2692                 goto recover;
 2693         /* Notify userland that interface flags have changed */
 2694         rt_ifmsg(ifp);
 2695         return (0);
 2696 
 2697 recover:
 2698         /* Recover after driver error */
 2699         *refcount = oldcount;
 2700         ifp->if_flags = oldflags;
 2701         return (error);
 2702 }
 2703 
 2704 /*
 2705  * Set/clear promiscuous mode on interface ifp based on the truth value
 2706  * of pswitch.  The calls are reference counted so that only the first
 2707  * "on" request actually has an effect, as does the final "off" request.
 2708  * Results are undefined if the "off" and "on" requests are not matched.
 2709  */
 2710 int
 2711 ifpromisc(struct ifnet *ifp, int pswitch)
 2712 {
 2713         int error;
 2714         int oldflags = ifp->if_flags;
 2715 
 2716         error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
 2717                            &ifp->if_pcount, pswitch);
 2718         /* If promiscuous mode status has changed, log a message */
 2719         if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
 2720                 log(LOG_INFO, "%s: promiscuous mode %s\n",
 2721                     ifp->if_xname,
 2722                     (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
 2723         return (error);
 2724 }
 2725 
 2726 /*
 2727  * Return interface configuration
 2728  * of system.  List may be used
 2729  * in later ioctl's (above) to get
 2730  * other information.
 2731  */
 2732 /*ARGSUSED*/
 2733 static int
 2734 ifconf(u_long cmd, caddr_t data)
 2735 {
 2736         struct ifconf *ifc = (struct ifconf *)data;
 2737         struct ifnet *ifp;
 2738         struct ifaddr *ifa;
 2739         struct ifreq ifr;
 2740         struct sbuf *sb;
 2741         int error, full = 0, valid_len, max_len;
 2742 
 2743         /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
 2744         max_len = MAXPHYS - 1;
 2745 
 2746         /* Prevent hostile input from being able to crash the system */
 2747         if (ifc->ifc_len <= 0)
 2748                 return (EINVAL);
 2749 
 2750 again:
 2751         if (ifc->ifc_len <= max_len) {
 2752                 max_len = ifc->ifc_len;
 2753                 full = 1;
 2754         }
 2755         sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
 2756         max_len = 0;
 2757         valid_len = 0;
 2758 
 2759         IFNET_RLOCK();
 2760         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2761                 int addrs;
 2762 
 2763                 /*
 2764                  * Zero the ifr_name buffer to make sure we don't
 2765                  * disclose the contents of the stack.
 2766                  */
 2767                 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
 2768 
 2769                 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
 2770                     >= sizeof(ifr.ifr_name)) {
 2771                         sbuf_delete(sb);
 2772                         IFNET_RUNLOCK();
 2773                         return (ENAMETOOLONG);
 2774                 }
 2775 
 2776                 addrs = 0;
 2777                 IF_ADDR_LOCK(ifp);
 2778                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2779                         struct sockaddr *sa = ifa->ifa_addr;
 2780 
 2781                         if (prison_if(curthread->td_ucred, sa) != 0)
 2782                                 continue;
 2783                         addrs++;
 2784 #ifdef COMPAT_43
 2785                         if (cmd == OSIOCGIFCONF) {
 2786                                 struct osockaddr *osa =
 2787                                          (struct osockaddr *)&ifr.ifr_addr;
 2788                                 ifr.ifr_addr = *sa;
 2789                                 osa->sa_family = sa->sa_family;
 2790                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2791                                 max_len += sizeof(ifr);
 2792                         } else
 2793 #endif
 2794                         if (sa->sa_len <= sizeof(*sa)) {
 2795                                 ifr.ifr_addr = *sa;
 2796                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2797                                 max_len += sizeof(ifr);
 2798                         } else {
 2799                                 sbuf_bcat(sb, &ifr,
 2800                                     offsetof(struct ifreq, ifr_addr));
 2801                                 max_len += offsetof(struct ifreq, ifr_addr);
 2802                                 sbuf_bcat(sb, sa, sa->sa_len);
 2803                                 max_len += sa->sa_len;
 2804                         }
 2805 
 2806                         if (!sbuf_overflowed(sb))
 2807                                 valid_len = sbuf_len(sb);
 2808                 }
 2809                 IF_ADDR_UNLOCK(ifp);
 2810                 if (addrs == 0) {
 2811                         bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
 2812                         sbuf_bcat(sb, &ifr, sizeof(ifr));
 2813                         max_len += sizeof(ifr);
 2814 
 2815                         if (!sbuf_overflowed(sb))
 2816                                 valid_len = sbuf_len(sb);
 2817                 }
 2818         }
 2819         IFNET_RUNLOCK();
 2820 
 2821         /*
 2822          * If we didn't allocate enough space (uncommon), try again.  If
 2823          * we have already allocated as much space as we are allowed,
 2824          * return what we've got.
 2825          */
 2826         if (valid_len != max_len && !full) {
 2827                 sbuf_delete(sb);
 2828                 goto again;
 2829         }
 2830 
 2831         ifc->ifc_len = valid_len;
 2832         sbuf_finish(sb);
 2833         error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
 2834         sbuf_delete(sb);
 2835         return (error);
 2836 }
 2837 
 2838 /*
 2839  * Just like ifpromisc(), but for all-multicast-reception mode.
 2840  */
 2841 int
 2842 if_allmulti(struct ifnet *ifp, int onswitch)
 2843 {
 2844 
 2845         return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
 2846 }
 2847 
 2848 struct ifmultiaddr *
 2849 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
 2850 {
 2851         struct ifmultiaddr *ifma;
 2852 
 2853         IF_ADDR_LOCK_ASSERT(ifp);
 2854 
 2855         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
 2856                 if (sa->sa_family == AF_LINK) {
 2857                         if (sa_dl_equal(ifma->ifma_addr, sa))
 2858                                 break;
 2859                 } else {
 2860                         if (sa_equal(ifma->ifma_addr, sa))
 2861                                 break;
 2862                 }
 2863         }
 2864 
 2865         return ifma;
 2866 }
 2867 
 2868 /*
 2869  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
 2870  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
 2871  * the ifnet multicast address list here, so the caller must do that and
 2872  * other setup work (such as notifying the device driver).  The reference
 2873  * count is initialized to 1.
 2874  */
 2875 static struct ifmultiaddr *
 2876 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
 2877     int mflags)
 2878 {
 2879         struct ifmultiaddr *ifma;
 2880         struct sockaddr *dupsa;
 2881 
 2882         ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
 2883             M_ZERO);
 2884         if (ifma == NULL)
 2885                 return (NULL);
 2886 
 2887         dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
 2888         if (dupsa == NULL) {
 2889                 free(ifma, M_IFMADDR);
 2890                 return (NULL);
 2891         }
 2892         bcopy(sa, dupsa, sa->sa_len);
 2893         ifma->ifma_addr = dupsa;
 2894 
 2895         ifma->ifma_ifp = ifp;
 2896         ifma->ifma_refcount = 1;
 2897         ifma->ifma_protospec = NULL;
 2898 
 2899         if (llsa == NULL) {
 2900                 ifma->ifma_lladdr = NULL;
 2901                 return (ifma);
 2902         }
 2903 
 2904         dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
 2905         if (dupsa == NULL) {
 2906                 free(ifma->ifma_addr, M_IFMADDR);
 2907                 free(ifma, M_IFMADDR);
 2908                 return (NULL);
 2909         }
 2910         bcopy(llsa, dupsa, llsa->sa_len);
 2911         ifma->ifma_lladdr = dupsa;
 2912 
 2913         return (ifma);
 2914 }
 2915 
 2916 /*
 2917  * if_freemulti: free ifmultiaddr structure and possibly attached related
 2918  * addresses.  The caller is responsible for implementing reference
 2919  * counting, notifying the driver, handling routing messages, and releasing
 2920  * any dependent link layer state.
 2921  */
 2922 static void
 2923 if_freemulti(struct ifmultiaddr *ifma)
 2924 {
 2925 
 2926         KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
 2927             ifma->ifma_refcount));
 2928         KASSERT(ifma->ifma_protospec == NULL,
 2929             ("if_freemulti: protospec not NULL"));
 2930 
 2931         if (ifma->ifma_lladdr != NULL)
 2932                 free(ifma->ifma_lladdr, M_IFMADDR);
 2933         free(ifma->ifma_addr, M_IFMADDR);
 2934         free(ifma, M_IFMADDR);
 2935 }
 2936 
 2937 /*
 2938  * Register an additional multicast address with a network interface.
 2939  *
 2940  * - If the address is already present, bump the reference count on the
 2941  *   address and return.
 2942  * - If the address is not link-layer, look up a link layer address.
 2943  * - Allocate address structures for one or both addresses, and attach to the
 2944  *   multicast address list on the interface.  If automatically adding a link
 2945  *   layer address, the protocol address will own a reference to the link
 2946  *   layer address, to be freed when it is freed.
 2947  * - Notify the network device driver of an addition to the multicast address
 2948  *   list.
 2949  *
 2950  * 'sa' points to caller-owned memory with the desired multicast address.
 2951  *
 2952  * 'retifma' will be used to return a pointer to the resulting multicast
 2953  * address reference, if desired.
 2954  */
 2955 int
 2956 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
 2957     struct ifmultiaddr **retifma)
 2958 {
 2959         struct ifmultiaddr *ifma, *ll_ifma;
 2960         struct sockaddr *llsa;
 2961         int error;
 2962 
 2963         /*
 2964          * If the address is already present, return a new reference to it;
 2965          * otherwise, allocate storage and set up a new address.
 2966          */
 2967         IF_ADDR_LOCK(ifp);
 2968         ifma = if_findmulti(ifp, sa);
 2969         if (ifma != NULL) {
 2970                 ifma->ifma_refcount++;
 2971                 if (retifma != NULL)
 2972                         *retifma = ifma;
 2973                 IF_ADDR_UNLOCK(ifp);
 2974                 return (0);
 2975         }
 2976 
 2977         /*
 2978          * The address isn't already present; resolve the protocol address
 2979          * into a link layer address, and then look that up, bump its
 2980          * refcount or allocate an ifma for that also.  If 'llsa' was
 2981          * returned, we will need to free it later.
 2982          */
 2983         llsa = NULL;
 2984         ll_ifma = NULL;
 2985         if (ifp->if_resolvemulti != NULL) {
 2986                 error = ifp->if_resolvemulti(ifp, &llsa, sa);
 2987                 if (error)
 2988                         goto unlock_out;
 2989         }
 2990 
 2991         /*
 2992          * Allocate the new address.  Don't hook it up yet, as we may also
 2993          * need to allocate a link layer multicast address.
 2994          */
 2995         ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
 2996         if (ifma == NULL) {
 2997                 error = ENOMEM;
 2998                 goto free_llsa_out;
 2999         }
 3000 
 3001         /*
 3002          * If a link layer address is found, we'll need to see if it's
 3003          * already present in the address list, or allocate is as well.
 3004          * When this block finishes, the link layer address will be on the
 3005          * list.
 3006          */
 3007         if (llsa != NULL) {
 3008                 ll_ifma = if_findmulti(ifp, llsa);
 3009                 if (ll_ifma == NULL) {
 3010                         ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
 3011                         if (ll_ifma == NULL) {
 3012                                 --ifma->ifma_refcount;
 3013                                 if_freemulti(ifma);
 3014                                 error = ENOMEM;
 3015                                 goto free_llsa_out;
 3016                         }
 3017                         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
 3018                             ifma_link);
 3019                 } else
 3020                         ll_ifma->ifma_refcount++;
 3021                 ifma->ifma_llifma = ll_ifma;
 3022         }
 3023 
 3024         /*
 3025          * We now have a new multicast address, ifma, and possibly a new or
 3026          * referenced link layer address.  Add the primary address to the
 3027          * ifnet address list.
 3028          */
 3029         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
 3030 
 3031         if (retifma != NULL)
 3032                 *retifma = ifma;
 3033 
 3034         /*
 3035          * Must generate the message while holding the lock so that 'ifma'
 3036          * pointer is still valid.
 3037          */
 3038         rt_newmaddrmsg(RTM_NEWMADDR, ifma);
 3039         IF_ADDR_UNLOCK(ifp);
 3040 
 3041         /*
 3042          * We are certain we have added something, so call down to the
 3043          * interface to let them know about it.
 3044          */
 3045         if (ifp->if_ioctl != NULL) {
 3046                 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
 3047         }
 3048 
 3049         if (llsa != NULL)
 3050                 free(llsa, M_IFMADDR);
 3051 
 3052         return (0);
 3053 
 3054 free_llsa_out:
 3055         if (llsa != NULL)
 3056                 free(llsa, M_IFMADDR);
 3057 
 3058 unlock_out:
 3059         IF_ADDR_UNLOCK(ifp);
 3060         return (error);
 3061 }
 3062 
 3063 /*
 3064  * Delete a multicast group membership by network-layer group address.
 3065  *
 3066  * Returns ENOENT if the entry could not be found. If ifp no longer
 3067  * exists, results are undefined. This entry point should only be used
 3068  * from subsystems which do appropriate locking to hold ifp for the
 3069  * duration of the call.
 3070  * Network-layer protocol domains must use if_delmulti_ifma().
 3071  */
 3072 int
 3073 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
 3074 {
 3075         struct ifmultiaddr *ifma;
 3076         int lastref;
 3077 #ifdef INVARIANTS
 3078         struct ifnet *oifp;
 3079 
 3080         IFNET_RLOCK_NOSLEEP();
 3081         TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3082                 if (ifp == oifp)
 3083                         break;
 3084         if (ifp != oifp)
 3085                 ifp = NULL;
 3086         IFNET_RUNLOCK_NOSLEEP();
 3087 
 3088         KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
 3089 #endif
 3090         if (ifp == NULL)
 3091                 return (ENOENT);
 3092 
 3093         IF_ADDR_LOCK(ifp);
 3094         lastref = 0;
 3095         ifma = if_findmulti(ifp, sa);
 3096         if (ifma != NULL)
 3097                 lastref = if_delmulti_locked(ifp, ifma, 0);
 3098         IF_ADDR_UNLOCK(ifp);
 3099 
 3100         if (ifma == NULL)
 3101                 return (ENOENT);
 3102 
 3103         if (lastref && ifp->if_ioctl != NULL) {
 3104                 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3105         }
 3106 
 3107         return (0);
 3108 }
 3109 
 3110 /*
 3111  * Delete all multicast group membership for an interface.
 3112  * Should be used to quickly flush all multicast filters.
 3113  */
 3114 void
 3115 if_delallmulti(struct ifnet *ifp)
 3116 {
 3117         struct ifmultiaddr *ifma;
 3118         struct ifmultiaddr *next;
 3119 
 3120         IF_ADDR_LOCK(ifp);
 3121         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
 3122                 if_delmulti_locked(ifp, ifma, 0);
 3123         IF_ADDR_UNLOCK(ifp);
 3124 }
 3125 
 3126 /*
 3127  * Delete a multicast group membership by group membership pointer.
 3128  * Network-layer protocol domains must use this routine.
 3129  *
 3130  * It is safe to call this routine if the ifp disappeared.
 3131  */
 3132 void
 3133 if_delmulti_ifma(struct ifmultiaddr *ifma)
 3134 {
 3135         struct ifnet *ifp;
 3136         int lastref;
 3137 
 3138         ifp = ifma->ifma_ifp;
 3139 #ifdef DIAGNOSTIC
 3140         if (ifp == NULL) {
 3141                 printf("%s: ifma_ifp seems to be detached\n", __func__);
 3142         } else {
 3143                 struct ifnet *oifp;
 3144 
 3145                 IFNET_RLOCK_NOSLEEP();
 3146                 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3147                         if (ifp == oifp)
 3148                                 break;
 3149                 if (ifp != oifp) {
 3150                         printf("%s: ifnet %p disappeared\n", __func__, ifp);
 3151                         ifp = NULL;
 3152                 }
 3153                 IFNET_RUNLOCK_NOSLEEP();
 3154         }
 3155 #endif
 3156         /*
 3157          * If and only if the ifnet instance exists: Acquire the address lock.
 3158          */
 3159         if (ifp != NULL)
 3160                 IF_ADDR_LOCK(ifp);
 3161 
 3162         lastref = if_delmulti_locked(ifp, ifma, 0);
 3163 
 3164         if (ifp != NULL) {
 3165                 /*
 3166                  * If and only if the ifnet instance exists:
 3167                  *  Release the address lock.
 3168                  *  If the group was left: update the hardware hash filter.
 3169                  */
 3170                 IF_ADDR_UNLOCK(ifp);
 3171                 if (lastref && ifp->if_ioctl != NULL) {
 3172                         (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3173                 }
 3174         }
 3175 }
 3176 
 3177 /*
 3178  * Perform deletion of network-layer and/or link-layer multicast address.
 3179  *
 3180  * Return 0 if the reference count was decremented.
 3181  * Return 1 if the final reference was released, indicating that the
 3182  * hardware hash filter should be reprogrammed.
 3183  */
 3184 static int
 3185 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
 3186 {
 3187         struct ifmultiaddr *ll_ifma;
 3188 
 3189         if (ifp != NULL && ifma->ifma_ifp != NULL) {
 3190                 KASSERT(ifma->ifma_ifp == ifp,
 3191                     ("%s: inconsistent ifp %p", __func__, ifp));
 3192                 IF_ADDR_LOCK_ASSERT(ifp);
 3193         }
 3194 
 3195         ifp = ifma->ifma_ifp;
 3196 
 3197         /*
 3198          * If the ifnet is detaching, null out references to ifnet,
 3199          * so that upper protocol layers will notice, and not attempt
 3200          * to obtain locks for an ifnet which no longer exists. The
 3201          * routing socket announcement must happen before the ifnet
 3202          * instance is detached from the system.
 3203          */
 3204         if (detaching) {
 3205 #ifdef DIAGNOSTIC
 3206                 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
 3207 #endif
 3208                 /*
 3209                  * ifp may already be nulled out if we are being reentered
 3210                  * to delete the ll_ifma.
 3211                  */
 3212                 if (ifp != NULL) {
 3213                         rt_newmaddrmsg(RTM_DELMADDR, ifma);
 3214                         ifma->ifma_ifp = NULL;
 3215                 }
 3216         }
 3217 
 3218         if (--ifma->ifma_refcount > 0)
 3219                 return 0;
 3220 
 3221         /*
 3222          * If this ifma is a network-layer ifma, a link-layer ifma may
 3223          * have been associated with it. Release it first if so.
 3224          */
 3225         ll_ifma = ifma->ifma_llifma;
 3226         if (ll_ifma != NULL) {
 3227                 KASSERT(ifma->ifma_lladdr != NULL,
 3228                     ("%s: llifma w/o lladdr", __func__));
 3229                 if (detaching)
 3230                         ll_ifma->ifma_ifp = NULL;       /* XXX */
 3231                 if (--ll_ifma->ifma_refcount == 0) {
 3232                         if (ifp != NULL) {
 3233                                 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
 3234                                     ifma_link);
 3235                         }
 3236                         if_freemulti(ll_ifma);
 3237                 }
 3238         }
 3239 
 3240         if (ifp != NULL)
 3241                 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
 3242 
 3243         if_freemulti(ifma);
 3244 
 3245         /*
 3246          * The last reference to this instance of struct ifmultiaddr
 3247          * was released; the hardware should be notified of this change.
 3248          */
 3249         return 1;
 3250 }
 3251 
 3252 /*
 3253  * Set the link layer address on an interface.
 3254  *
 3255  * At this time we only support certain types of interfaces,
 3256  * and we don't allow the length of the address to change.
 3257  */
 3258 int
 3259 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
 3260 {
 3261         struct sockaddr_dl *sdl;
 3262         struct ifaddr *ifa;
 3263         struct ifreq ifr;
 3264 
 3265         IF_ADDR_LOCK(ifp);
 3266         ifa = ifp->if_addr;
 3267         if (ifa == NULL) {
 3268                 IF_ADDR_UNLOCK(ifp);
 3269                 return (EINVAL);
 3270         }
 3271         ifa_ref(ifa);
 3272         IF_ADDR_UNLOCK(ifp);
 3273         sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 3274         if (sdl == NULL) {
 3275                 ifa_free(ifa);
 3276                 return (EINVAL);
 3277         }
 3278         if (len != sdl->sdl_alen) {     /* don't allow length to change */
 3279                 ifa_free(ifa);
 3280                 return (EINVAL);
 3281         }
 3282         switch (ifp->if_type) {
 3283         case IFT_ETHER:
 3284         case IFT_FDDI:
 3285         case IFT_XETHER:
 3286         case IFT_ISO88025:
 3287         case IFT_L2VLAN:
 3288         case IFT_BRIDGE:
 3289         case IFT_ARCNET:
 3290         case IFT_IEEE8023ADLAG:
 3291         case IFT_IEEE80211:
 3292                 bcopy(lladdr, LLADDR(sdl), len);
 3293                 ifa_free(ifa);
 3294                 break;
 3295         default:
 3296                 ifa_free(ifa);
 3297                 return (ENODEV);
 3298         }
 3299 
 3300         /*
 3301          * If the interface is already up, we need
 3302          * to re-init it in order to reprogram its
 3303          * address filter.
 3304          */
 3305         if ((ifp->if_flags & IFF_UP) != 0) {
 3306                 if (ifp->if_ioctl) {
 3307                         ifp->if_flags &= ~IFF_UP;
 3308                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3309                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3310                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3311                         ifp->if_flags |= IFF_UP;
 3312                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3313                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3314                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3315                 }
 3316 #ifdef INET
 3317                 /*
 3318                  * Also send gratuitous ARPs to notify other nodes about
 3319                  * the address change.
 3320                  */
 3321                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 3322                         if (ifa->ifa_addr->sa_family == AF_INET)
 3323                                 arp_ifinit(ifp, ifa);
 3324                 }
 3325 #endif
 3326         }
 3327         return (0);
 3328 }
 3329 
 3330 /*
 3331  * The name argument must be a pointer to storage which will last as
 3332  * long as the interface does.  For physical devices, the result of
 3333  * device_get_name(dev) is a good choice and for pseudo-devices a
 3334  * static string works well.
 3335  */
 3336 void
 3337 if_initname(struct ifnet *ifp, const char *name, int unit)
 3338 {
 3339         ifp->if_dname = name;
 3340         ifp->if_dunit = unit;
 3341         if (unit != IF_DUNIT_NONE)
 3342                 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
 3343         else
 3344                 strlcpy(ifp->if_xname, name, IFNAMSIZ);
 3345 }
 3346 
 3347 int
 3348 if_printf(struct ifnet *ifp, const char * fmt, ...)
 3349 {
 3350         va_list ap;
 3351         int retval;
 3352 
 3353         retval = printf("%s: ", ifp->if_xname);
 3354         va_start(ap, fmt);
 3355         retval += vprintf(fmt, ap);
 3356         va_end(ap);
 3357         return (retval);
 3358 }
 3359 
 3360 void
 3361 if_start(struct ifnet *ifp)
 3362 {
 3363 
 3364         (*(ifp)->if_start)(ifp);
 3365 }
 3366 
 3367 /*
 3368  * Backwards compatibility interface for drivers 
 3369  * that have not implemented it
 3370  */
 3371 static int
 3372 if_transmit(struct ifnet *ifp, struct mbuf *m)
 3373 {
 3374         int error;
 3375 
 3376         IFQ_HANDOFF(ifp, m, error);
 3377         return (error);
 3378 }
 3379 
 3380 int
 3381 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
 3382 {
 3383         int active = 0;
 3384 
 3385         IF_LOCK(ifq);
 3386         if (_IF_QFULL(ifq)) {
 3387                 _IF_DROP(ifq);
 3388                 IF_UNLOCK(ifq);
 3389                 m_freem(m);
 3390                 return (0);
 3391         }
 3392         if (ifp != NULL) {
 3393                 ifp->if_obytes += m->m_pkthdr.len + adjust;
 3394                 if (m->m_flags & (M_BCAST|M_MCAST))
 3395                         ifp->if_omcasts++;
 3396                 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
 3397         }
 3398         _IF_ENQUEUE(ifq, m);
 3399         IF_UNLOCK(ifq);
 3400         if (ifp != NULL && !active)
 3401                 (*(ifp)->if_start)(ifp);
 3402         return (1);
 3403 }
 3404 
 3405 void
 3406 if_register_com_alloc(u_char type,
 3407     if_com_alloc_t *a, if_com_free_t *f)
 3408 {
 3409         
 3410         KASSERT(if_com_alloc[type] == NULL,
 3411             ("if_register_com_alloc: %d already registered", type));
 3412         KASSERT(if_com_free[type] == NULL,
 3413             ("if_register_com_alloc: %d free already registered", type));
 3414 
 3415         if_com_alloc[type] = a;
 3416         if_com_free[type] = f;
 3417 }
 3418 
 3419 void
 3420 if_deregister_com_alloc(u_char type)
 3421 {
 3422         
 3423         KASSERT(if_com_alloc[type] != NULL,
 3424             ("if_deregister_com_alloc: %d not registered", type));
 3425         KASSERT(if_com_free[type] != NULL,
 3426             ("if_deregister_com_alloc: %d free not registered", type));
 3427         if_com_alloc[type] = NULL;
 3428         if_com_free[type] = NULL;
 3429 }

Cache object: ad1d20e9dad86827d0b0a7b441375a5b


[ 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.