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/8.1/sys/net/if.c 209524 2010-06-25 21:26:34Z qingli $
   31  */
   32 
   33 #include "opt_compat.h"
   34 #include "opt_inet6.h"
   35 #include "opt_inet.h"
   36 #include "opt_carp.h"
   37 #include "opt_ddb.h"
   38 
   39 #include <sys/param.h>
   40 #include <sys/types.h>
   41 #include <sys/conf.h>
   42 #include <sys/malloc.h>
   43 #include <sys/sbuf.h>
   44 #include <sys/bus.h>
   45 #include <sys/mbuf.h>
   46 #include <sys/systm.h>
   47 #include <sys/priv.h>
   48 #include <sys/proc.h>
   49 #include <sys/socket.h>
   50 #include <sys/socketvar.h>
   51 #include <sys/protosw.h>
   52 #include <sys/kernel.h>
   53 #include <sys/lock.h>
   54 #include <sys/refcount.h>
   55 #include <sys/module.h>
   56 #include <sys/rwlock.h>
   57 #include <sys/sockio.h>
   58 #include <sys/syslog.h>
   59 #include <sys/sysctl.h>
   60 #include <sys/taskqueue.h>
   61 #include <sys/domain.h>
   62 #include <sys/jail.h>
   63 #include <machine/stdarg.h>
   64 #include <vm/uma.h>
   65 
   66 #ifdef DDB
   67 #include <ddb/ddb.h>
   68 #endif
   69 
   70 #include <net/if.h>
   71 #include <net/if_arp.h>
   72 #include <net/if_clone.h>
   73 #include <net/if_dl.h>
   74 #include <net/if_types.h>
   75 #include <net/if_var.h>
   76 #include <net/radix.h>
   77 #include <net/route.h>
   78 #include <net/vnet.h>
   79 
   80 #if defined(INET) || defined(INET6)
   81 /*XXX*/
   82 #include <netinet/in.h>
   83 #include <netinet/in_var.h>
   84 #ifdef INET6
   85 #include <netinet6/in6_var.h>
   86 #include <netinet6/in6_ifattach.h>
   87 #endif
   88 #endif
   89 #ifdef INET
   90 #include <netinet/if_ether.h>
   91 #endif
   92 #if defined(INET) || defined(INET6)
   93 #ifdef DEV_CARP
   94 #include <netinet/ip_carp.h>
   95 #endif
   96 #endif
   97 
   98 #include <security/mac/mac_framework.h>
   99 
  100 struct ifindex_entry {
  101         struct  ifnet *ife_ifnet;
  102 };
  103 
  104 static int slowtimo_started;
  105 
  106 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
  107 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
  108 
  109 /* Log link state change events */
  110 static int log_link_state_change = 1;
  111 
  112 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
  113         &log_link_state_change, 0,
  114         "log interface link state change events");
  115 
  116 /* Interface description */
  117 static unsigned int ifdescr_maxlen = 1024;
  118 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
  119         &ifdescr_maxlen, 0,
  120         "administrative maximum length for interface description");
  121 
  122 MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
  123 
  124 /* global sx for non-critical path ifdescr */
  125 static struct sx ifdescr_sx;
  126 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
  127 
  128 void    (*bstp_linkstate_p)(struct ifnet *ifp, int state);
  129 void    (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
  130 void    (*lagg_linkstate_p)(struct ifnet *ifp, int state);
  131 
  132 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
  133 
  134 /*
  135  * XXX: Style; these should be sorted alphabetically, and unprototyped
  136  * static functions should be prototyped. Currently they are sorted by
  137  * declaration order.
  138  */
  139 static void     if_attachdomain(void *);
  140 static void     if_attachdomain1(struct ifnet *);
  141 static int      ifconf(u_long, caddr_t);
  142 static void     if_freemulti(struct ifmultiaddr *);
  143 static void     if_init(void *);
  144 static void     if_grow(void);
  145 static void     if_check(void *);
  146 static void     if_route(struct ifnet *, int flag, int fam);
  147 static int      if_setflag(struct ifnet *, int, int, int *, int);
  148 static void     if_slowtimo(void *);
  149 static int      if_transmit(struct ifnet *ifp, struct mbuf *m);
  150 static void     if_unroute(struct ifnet *, int flag, int fam);
  151 static void     link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  152 static int      if_rtdel(struct radix_node *, void *);
  153 static int      ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
  154 static int      if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
  155 static void     do_link_state_change(void *, int);
  156 static int      if_getgroup(struct ifgroupreq *, struct ifnet *);
  157 static int      if_getgroupmembers(struct ifgroupreq *);
  158 static void     if_delgroups(struct ifnet *);
  159 static void     if_attach_internal(struct ifnet *, int);
  160 static void     if_detach_internal(struct ifnet *, int);
  161 
  162 #ifdef INET6
  163 /*
  164  * XXX: declare here to avoid to include many inet6 related files..
  165  * should be more generalized?
  166  */
  167 extern void     nd6_setmtu(struct ifnet *);
  168 #endif
  169 
  170 VNET_DEFINE(int, if_index);
  171 int     ifqmaxlen = IFQ_MAXLEN;
  172 VNET_DEFINE(struct ifnethead, ifnet);   /* depend on static init XXX */
  173 VNET_DEFINE(struct ifgrouphead, ifg_head);
  174 
  175 static VNET_DEFINE(int, if_indexlim) = 8;
  176 
  177 /* Table of ifnet by index. */
  178 static VNET_DEFINE(struct ifindex_entry *, ifindex_table);
  179 
  180 #define V_if_indexlim           VNET(if_indexlim)
  181 #define V_ifindex_table         VNET(ifindex_table)
  182 
  183 /*
  184  * The global network interface list (V_ifnet) and related state (such as
  185  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
  186  * an rwlock.  Either may be acquired shared to stablize the list, but both
  187  * must be acquired writable to modify the list.  This model allows us to
  188  * both stablize the interface list during interrupt thread processing, but
  189  * also to stablize it over long-running ioctls, without introducing priority
  190  * inversions and deadlocks.
  191  */
  192 struct rwlock ifnet_rwlock;
  193 struct sx ifnet_sxlock;
  194 
  195 /*
  196  * The allocation of network interfaces is a rather non-atomic affair; we
  197  * need to select an index before we are ready to expose the interface for
  198  * use, so will use this pointer value to indicate reservation.
  199  */
  200 #define IFNET_HOLD      (void *)(uintptr_t)(-1)
  201 
  202 static  if_com_alloc_t *if_com_alloc[256];
  203 static  if_com_free_t *if_com_free[256];
  204 
  205 /*
  206  * System initialization
  207  */
  208 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL);
  209 
  210 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
  211 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
  212 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
  213 
  214 struct ifnet *
  215 ifnet_byindex_locked(u_short idx)
  216 {
  217 
  218         if (idx > V_if_index)
  219                 return (NULL);
  220         if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
  221                 return (NULL);
  222         return (V_ifindex_table[idx].ife_ifnet);
  223 }
  224 
  225 struct ifnet *
  226 ifnet_byindex(u_short idx)
  227 {
  228         struct ifnet *ifp;
  229 
  230         IFNET_RLOCK_NOSLEEP();
  231         ifp = ifnet_byindex_locked(idx);
  232         IFNET_RUNLOCK_NOSLEEP();
  233         return (ifp);
  234 }
  235 
  236 struct ifnet *
  237 ifnet_byindex_ref(u_short idx)
  238 {
  239         struct ifnet *ifp;
  240 
  241         IFNET_RLOCK_NOSLEEP();
  242         ifp = ifnet_byindex_locked(idx);
  243         if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
  244                 IFNET_RUNLOCK_NOSLEEP();
  245                 return (NULL);
  246         }
  247         if_ref(ifp);
  248         IFNET_RUNLOCK_NOSLEEP();
  249         return (ifp);
  250 }
  251 
  252 /*
  253  * Allocate an ifindex array entry; return 0 on success or an error on
  254  * failure.
  255  */
  256 static int
  257 ifindex_alloc_locked(u_short *idxp)
  258 {
  259         u_short idx;
  260 
  261         IFNET_WLOCK_ASSERT();
  262 
  263         /*
  264          * Try to find an empty slot below V_if_index.  If we fail, take the
  265          * next slot.
  266          */
  267         for (idx = 1; idx <= V_if_index; idx++) {
  268                 if (V_ifindex_table[idx].ife_ifnet == NULL)
  269                         break;
  270         }
  271 
  272         /* Catch if_index overflow. */
  273         if (idx < 1)
  274                 return (ENOSPC);
  275         if (idx > V_if_index)
  276                 V_if_index = idx;
  277         if (V_if_index >= V_if_indexlim)
  278                 if_grow();
  279         *idxp = idx;
  280         return (0);
  281 }
  282 
  283 static void
  284 ifindex_free_locked(u_short idx)
  285 {
  286 
  287         IFNET_WLOCK_ASSERT();
  288 
  289         V_ifindex_table[idx].ife_ifnet = NULL;
  290         while (V_if_index > 0 &&
  291             V_ifindex_table[V_if_index].ife_ifnet == NULL)
  292                 V_if_index--;
  293 }
  294 
  295 static void
  296 ifindex_free(u_short idx)
  297 {
  298 
  299         IFNET_WLOCK();
  300         ifindex_free_locked(idx);
  301         IFNET_WUNLOCK();
  302 }
  303 
  304 static void
  305 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
  306 {
  307 
  308         IFNET_WLOCK_ASSERT();
  309 
  310         V_ifindex_table[idx].ife_ifnet = ifp;
  311 }
  312 
  313 static void
  314 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
  315 {
  316 
  317         IFNET_WLOCK();
  318         ifnet_setbyindex_locked(idx, ifp);
  319         IFNET_WUNLOCK();
  320 }
  321 
  322 struct ifaddr *
  323 ifaddr_byindex(u_short idx)
  324 {
  325         struct ifaddr *ifa;
  326 
  327         IFNET_RLOCK_NOSLEEP();
  328         ifa = ifnet_byindex_locked(idx)->if_addr;
  329         if (ifa != NULL)
  330                 ifa_ref(ifa);
  331         IFNET_RUNLOCK_NOSLEEP();
  332         return (ifa);
  333 }
  334 
  335 /*
  336  * Network interface utility routines.
  337  *
  338  * Routines with ifa_ifwith* names take sockaddr *'s as
  339  * parameters.
  340  */
  341 
  342 static void
  343 vnet_if_init(const void *unused __unused)
  344 {
  345 
  346         TAILQ_INIT(&V_ifnet);
  347         TAILQ_INIT(&V_ifg_head);
  348         if_grow();                              /* create initial table */
  349         vnet_if_clone_init();
  350 }
  351 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_FIRST, vnet_if_init,
  352     NULL);
  353 
  354 /* ARGSUSED*/
  355 static void
  356 if_init(void *dummy __unused)
  357 {
  358 
  359         IFNET_LOCK_INIT();
  360         if_clone_init();
  361 }
  362 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_SECOND, if_init, NULL);
  363 
  364 
  365 #ifdef VIMAGE
  366 static void
  367 vnet_if_uninit(const void *unused __unused)
  368 {
  369 
  370         VNET_ASSERT(TAILQ_EMPTY(&V_ifnet));
  371         VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head));
  372 
  373         free((caddr_t)V_ifindex_table, M_IFNET);
  374 }
  375 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
  376     vnet_if_uninit, NULL);
  377 #endif
  378 
  379 static void
  380 if_grow(void)
  381 {
  382         u_int n;
  383         struct ifindex_entry *e;
  384 
  385         V_if_indexlim <<= 1;
  386         n = V_if_indexlim * sizeof(*e);
  387         e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
  388         if (V_ifindex_table != NULL) {
  389                 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
  390                 free((caddr_t)V_ifindex_table, M_IFNET);
  391         }
  392         V_ifindex_table = e;
  393 }
  394 
  395 static void
  396 if_check(void *dummy __unused)
  397 {
  398 
  399         /*
  400          * If at least one interface added during boot uses
  401          * if_watchdog then start the timer.
  402          */
  403         if (slowtimo_started)
  404                 if_slowtimo(0);
  405 }
  406 
  407 /*
  408  * Allocate a struct ifnet and an index for an interface.  A layer 2
  409  * common structure will also be allocated if an allocation routine is
  410  * registered for the passed type.
  411  */
  412 struct ifnet *
  413 if_alloc(u_char type)
  414 {
  415         struct ifnet *ifp;
  416         u_short idx;
  417 
  418         ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
  419         IFNET_WLOCK();
  420         if (ifindex_alloc_locked(&idx) != 0) {
  421                 IFNET_WUNLOCK();
  422                 free(ifp, M_IFNET);
  423                 return (NULL);
  424         }
  425         ifnet_setbyindex_locked(idx, IFNET_HOLD);
  426         IFNET_WUNLOCK();
  427         ifp->if_index = idx;
  428         ifp->if_type = type;
  429         ifp->if_alloctype = type;
  430         if (if_com_alloc[type] != NULL) {
  431                 ifp->if_l2com = if_com_alloc[type](type, ifp);
  432                 if (ifp->if_l2com == NULL) {
  433                         free(ifp, M_IFNET);
  434                         ifindex_free(idx);
  435                         return (NULL);
  436                 }
  437         }
  438 
  439         IF_ADDR_LOCK_INIT(ifp);
  440         TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
  441         ifp->if_afdata_initialized = 0;
  442         IF_AFDATA_LOCK_INIT(ifp);
  443         TAILQ_INIT(&ifp->if_addrhead);
  444         TAILQ_INIT(&ifp->if_prefixhead);
  445         TAILQ_INIT(&ifp->if_multiaddrs);
  446         TAILQ_INIT(&ifp->if_groups);
  447 #ifdef MAC
  448         mac_ifnet_init(ifp);
  449 #endif
  450         ifq_init(&ifp->if_snd, ifp);
  451 
  452         refcount_init(&ifp->if_refcount, 1);    /* Index reference. */
  453         ifnet_setbyindex(ifp->if_index, ifp);
  454         return (ifp);
  455 }
  456 
  457 /*
  458  * Do the actual work of freeing a struct ifnet, associated index, and layer
  459  * 2 common structure.  This call is made when the last reference to an
  460  * interface is released.
  461  */
  462 static void
  463 if_free_internal(struct ifnet *ifp)
  464 {
  465 
  466         KASSERT((ifp->if_flags & IFF_DYING),
  467             ("if_free_internal: interface not dying"));
  468 
  469         IFNET_WLOCK();
  470         KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
  471             ("%s: freeing unallocated ifnet", ifp->if_xname));
  472 
  473         ifindex_free_locked(ifp->if_index);
  474         IFNET_WUNLOCK();
  475 
  476         if (if_com_free[ifp->if_alloctype] != NULL)
  477                 if_com_free[ifp->if_alloctype](ifp->if_l2com,
  478                     ifp->if_alloctype);
  479 
  480 #ifdef MAC
  481         mac_ifnet_destroy(ifp);
  482 #endif /* MAC */
  483         if (ifp->if_description != NULL)
  484                 free(ifp->if_description, M_IFDESCR);
  485         IF_AFDATA_DESTROY(ifp);
  486         IF_ADDR_LOCK_DESTROY(ifp);
  487         ifq_delete(&ifp->if_snd);
  488         free(ifp, M_IFNET);
  489 }
  490 
  491 /*
  492  * This version should only be called by intefaces that switch their type
  493  * after calling if_alloc().  if_free_type() will go away again now that we
  494  * have if_alloctype to cache the original allocation type.  For now, assert
  495  * that they match, since we require that in practice.
  496  */
  497 void
  498 if_free_type(struct ifnet *ifp, u_char type)
  499 {
  500 
  501         KASSERT(ifp->if_alloctype == type,
  502             ("if_free_type: type (%d) != alloctype (%d)", type,
  503             ifp->if_alloctype));
  504 
  505         ifp->if_flags |= IFF_DYING;                     /* XXX: Locking */
  506         if (!refcount_release(&ifp->if_refcount))
  507                 return;
  508         if_free_internal(ifp);
  509 }
  510 
  511 /*
  512  * This is the normal version of if_free(), used by device drivers to free a
  513  * detached network interface.  The contents of if_free_type() will move into
  514  * here when if_free_type() goes away.
  515  */
  516 void
  517 if_free(struct ifnet *ifp)
  518 {
  519 
  520         if_free_type(ifp, ifp->if_alloctype);
  521 }
  522 
  523 /*
  524  * Interfaces to keep an ifnet type-stable despite the possibility of the
  525  * driver calling if_free().  If there are additional references, we defer
  526  * freeing the underlying data structure.
  527  */
  528 void
  529 if_ref(struct ifnet *ifp)
  530 {
  531 
  532         /* We don't assert the ifnet list lock here, but arguably should. */
  533         refcount_acquire(&ifp->if_refcount);
  534 }
  535 
  536 void
  537 if_rele(struct ifnet *ifp)
  538 {
  539 
  540         if (!refcount_release(&ifp->if_refcount))
  541                 return;
  542         if_free_internal(ifp);
  543 }
  544 
  545 void
  546 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
  547 {
  548         
  549         mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
  550 
  551         if (ifq->ifq_maxlen == 0) 
  552                 ifq->ifq_maxlen = ifqmaxlen;
  553 
  554         ifq->altq_type = 0;
  555         ifq->altq_disc = NULL;
  556         ifq->altq_flags &= ALTQF_CANTCHANGE;
  557         ifq->altq_tbr  = NULL;
  558         ifq->altq_ifp  = ifp;
  559 }
  560 
  561 void
  562 ifq_delete(struct ifaltq *ifq)
  563 {
  564         mtx_destroy(&ifq->ifq_mtx);
  565 }
  566 
  567 /*
  568  * Perform generic interface initalization tasks and attach the interface
  569  * to the list of "active" interfaces.  If vmove flag is set on entry
  570  * to if_attach_internal(), perform only a limited subset of initialization
  571  * tasks, given that we are moving from one vnet to another an ifnet which
  572  * has already been fully initialized.
  573  *
  574  * XXX:
  575  *  - The decision to return void and thus require this function to
  576  *    succeed is questionable.
  577  *  - We should probably do more sanity checking.  For instance we don't
  578  *    do anything to insure if_xname is unique or non-empty.
  579  */
  580 void
  581 if_attach(struct ifnet *ifp)
  582 {
  583 
  584         if_attach_internal(ifp, 0);
  585 }
  586 
  587 static void
  588 if_attach_internal(struct ifnet *ifp, int vmove)
  589 {
  590         unsigned socksize, ifasize;
  591         int namelen, masklen;
  592         struct sockaddr_dl *sdl;
  593         struct ifaddr *ifa;
  594 
  595         if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
  596                 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
  597                     ifp->if_xname);
  598 
  599 #ifdef VIMAGE
  600         ifp->if_vnet = curvnet;
  601         if (ifp->if_home_vnet == NULL)
  602                 ifp->if_home_vnet = curvnet;
  603 #endif
  604 
  605         if_addgroup(ifp, IFG_ALL);
  606 
  607         getmicrotime(&ifp->if_lastchange);
  608         ifp->if_data.ifi_epoch = time_uptime;
  609         ifp->if_data.ifi_datalen = sizeof(struct if_data);
  610 
  611         KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
  612             (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
  613             ("transmit and qflush must both either be set or both be NULL"));
  614         if (ifp->if_transmit == NULL) {
  615                 ifp->if_transmit = if_transmit;
  616                 ifp->if_qflush = if_qflush;
  617         }
  618         
  619         if (!vmove) {
  620 #ifdef MAC
  621                 mac_ifnet_create(ifp);
  622 #endif
  623 
  624                 /*
  625                  * Create a Link Level name for this device.
  626                  */
  627                 namelen = strlen(ifp->if_xname);
  628                 /*
  629                  * Always save enough space for any possiable name so we
  630                  * can do a rename in place later.
  631                  */
  632                 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
  633                 socksize = masklen + ifp->if_addrlen;
  634                 if (socksize < sizeof(*sdl))
  635                         socksize = sizeof(*sdl);
  636                 socksize = roundup2(socksize, sizeof(long));
  637                 ifasize = sizeof(*ifa) + 2 * socksize;
  638                 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
  639                 ifa_init(ifa);
  640                 sdl = (struct sockaddr_dl *)(ifa + 1);
  641                 sdl->sdl_len = socksize;
  642                 sdl->sdl_family = AF_LINK;
  643                 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
  644                 sdl->sdl_nlen = namelen;
  645                 sdl->sdl_index = ifp->if_index;
  646                 sdl->sdl_type = ifp->if_type;
  647                 ifp->if_addr = ifa;
  648                 ifa->ifa_ifp = ifp;
  649                 ifa->ifa_rtrequest = link_rtrequest;
  650                 ifa->ifa_addr = (struct sockaddr *)sdl;
  651                 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
  652                 ifa->ifa_netmask = (struct sockaddr *)sdl;
  653                 sdl->sdl_len = masklen;
  654                 while (namelen != 0)
  655                         sdl->sdl_data[--namelen] = 0xff;
  656                 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
  657                 /* Reliably crash if used uninitialized. */
  658                 ifp->if_broadcastaddr = NULL;
  659         }
  660 #ifdef VIMAGE
  661         else {
  662                 /*
  663                  * Update the interface index in the link layer address
  664                  * of the interface.
  665                  */
  666                 for (ifa = ifp->if_addr; ifa != NULL;
  667                     ifa = TAILQ_NEXT(ifa, ifa_link)) {
  668                         if (ifa->ifa_addr->sa_family == AF_LINK) {
  669                                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
  670                                 sdl->sdl_index = ifp->if_index;
  671                         }
  672                 }
  673         }
  674 #endif
  675 
  676         IFNET_WLOCK();
  677         TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
  678 #ifdef VIMAGE
  679         curvnet->vnet_ifcnt++;
  680 #endif
  681         IFNET_WUNLOCK();
  682 
  683         if (domain_init_status >= 2)
  684                 if_attachdomain1(ifp);
  685 
  686         EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
  687         if (IS_DEFAULT_VNET(curvnet))
  688                 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
  689 
  690         /* Announce the interface. */
  691         rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
  692 
  693         if (!vmove && ifp->if_watchdog != NULL) {
  694                 if_printf(ifp,
  695                     "WARNING: using obsoleted if_watchdog interface\n");
  696 
  697                 /*
  698                  * Note that we need if_slowtimo().  If this happens after
  699                  * boot, then call if_slowtimo() directly.
  700                  */
  701                 if (atomic_cmpset_int(&slowtimo_started, 0, 1) && !cold)
  702                         if_slowtimo(0);
  703         }
  704 }
  705 
  706 static void
  707 if_attachdomain(void *dummy)
  708 {
  709         struct ifnet *ifp;
  710         int s;
  711 
  712         s = splnet();
  713         TAILQ_FOREACH(ifp, &V_ifnet, if_link)
  714                 if_attachdomain1(ifp);
  715         splx(s);
  716 }
  717 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
  718     if_attachdomain, NULL);
  719 
  720 static void
  721 if_attachdomain1(struct ifnet *ifp)
  722 {
  723         struct domain *dp;
  724         int s;
  725 
  726         s = splnet();
  727 
  728         /*
  729          * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
  730          * cannot lock ifp->if_afdata initialization, entirely.
  731          */
  732         if (IF_AFDATA_TRYLOCK(ifp) == 0) {
  733                 splx(s);
  734                 return;
  735         }
  736         if (ifp->if_afdata_initialized >= domain_init_status) {
  737                 IF_AFDATA_UNLOCK(ifp);
  738                 splx(s);
  739                 printf("if_attachdomain called more than once on %s\n",
  740                     ifp->if_xname);
  741                 return;
  742         }
  743         ifp->if_afdata_initialized = domain_init_status;
  744         IF_AFDATA_UNLOCK(ifp);
  745 
  746         /* address family dependent data region */
  747         bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
  748         for (dp = domains; dp; dp = dp->dom_next) {
  749                 if (dp->dom_ifattach)
  750                         ifp->if_afdata[dp->dom_family] =
  751                             (*dp->dom_ifattach)(ifp);
  752         }
  753 
  754         splx(s);
  755 }
  756 
  757 /*
  758  * Remove any unicast or broadcast network addresses from an interface.
  759  */
  760 void
  761 if_purgeaddrs(struct ifnet *ifp)
  762 {
  763         struct ifaddr *ifa, *next;
  764 
  765         TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  766                 if (ifa->ifa_addr->sa_family == AF_LINK)
  767                         continue;
  768 #ifdef INET
  769                 /* XXX: Ugly!! ad hoc just for INET */
  770                 if (ifa->ifa_addr->sa_family == AF_INET) {
  771                         struct ifaliasreq ifr;
  772 
  773                         bzero(&ifr, sizeof(ifr));
  774                         ifr.ifra_addr = *ifa->ifa_addr;
  775                         if (ifa->ifa_dstaddr)
  776                                 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
  777                         if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
  778                             NULL) == 0)
  779                                 continue;
  780                 }
  781 #endif /* INET */
  782 #ifdef INET6
  783                 if (ifa->ifa_addr->sa_family == AF_INET6) {
  784                         in6_purgeaddr(ifa);
  785                         /* ifp_addrhead is already updated */
  786                         continue;
  787                 }
  788 #endif /* INET6 */
  789                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  790                 ifa_free(ifa);
  791         }
  792 }
  793 
  794 /*
  795  * Remove any multicast network addresses from an interface when an ifnet
  796  * is going away.
  797  */
  798 static void
  799 if_purgemaddrs(struct ifnet *ifp)
  800 {
  801         struct ifmultiaddr *ifma;
  802         struct ifmultiaddr *next;
  803 
  804         IF_ADDR_LOCK(ifp);
  805         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
  806                 if_delmulti_locked(ifp, ifma, 1);
  807         IF_ADDR_UNLOCK(ifp);
  808 }
  809 
  810 /*
  811  * Detach an interface, removing it from the list of "active" interfaces.
  812  * If vmove flag is set on entry to if_detach_internal(), perform only a
  813  * limited subset of cleanup tasks, given that we are moving an ifnet from
  814  * one vnet to another, where it must be fully operational.
  815  *
  816  * XXXRW: There are some significant questions about event ordering, and
  817  * how to prevent things from starting to use the interface during detach.
  818  */
  819 void
  820 if_detach(struct ifnet *ifp)
  821 {
  822 
  823         if_detach_internal(ifp, 0);
  824 }
  825 
  826 static void
  827 if_detach_internal(struct ifnet *ifp, int vmove)
  828 {
  829         struct ifaddr *ifa;
  830         struct radix_node_head  *rnh;
  831         int i, j;
  832         struct domain *dp;
  833         struct ifnet *iter;
  834         int found = 0;
  835 
  836         IFNET_WLOCK();
  837         TAILQ_FOREACH(iter, &V_ifnet, if_link)
  838                 if (iter == ifp) {
  839                         TAILQ_REMOVE(&V_ifnet, ifp, if_link);
  840                         found = 1;
  841                         break;
  842                 }
  843 #ifdef VIMAGE
  844         if (found)
  845                 curvnet->vnet_ifcnt--;
  846 #endif
  847         IFNET_WUNLOCK();
  848         if (!found) {
  849                 if (vmove)
  850                         panic("%s: ifp=%p not on the ifnet tailq %p",
  851                             __func__, ifp, &V_ifnet);
  852                 else
  853                         return; /* XXX this should panic as well? */
  854         }
  855 
  856         /*
  857          * Remove/wait for pending events.
  858          */
  859         taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
  860 
  861         /*
  862          * Remove routes and flush queues.
  863          */
  864         if_down(ifp);
  865 #ifdef ALTQ
  866         if (ALTQ_IS_ENABLED(&ifp->if_snd))
  867                 altq_disable(&ifp->if_snd);
  868         if (ALTQ_IS_ATTACHED(&ifp->if_snd))
  869                 altq_detach(&ifp->if_snd);
  870 #endif
  871 
  872         if_purgeaddrs(ifp);
  873 
  874 #ifdef INET
  875         in_ifdetach(ifp);
  876 #endif
  877 
  878 #ifdef INET6
  879         /*
  880          * Remove all IPv6 kernel structs related to ifp.  This should be done
  881          * before removing routing entries below, since IPv6 interface direct
  882          * routes are expected to be removed by the IPv6-specific kernel API.
  883          * Otherwise, the kernel will detect some inconsistency and bark it.
  884          */
  885         in6_ifdetach(ifp);
  886 #endif
  887         if_purgemaddrs(ifp);
  888 
  889         if (!vmove) {
  890                 /*
  891                  * Prevent further calls into the device driver via ifnet.
  892                  */
  893                 if_dead(ifp);
  894 
  895                 /*
  896                  * Remove link ifaddr pointer and maybe decrement if_index.
  897                  * Clean up all addresses.
  898                  */
  899                 ifp->if_addr = NULL;
  900 
  901                 /* We can now free link ifaddr. */
  902                 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
  903                         ifa = TAILQ_FIRST(&ifp->if_addrhead);
  904                         TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  905                         ifa_free(ifa);
  906                 }
  907         }
  908 
  909         /*
  910          * Delete all remaining routes using this interface
  911          * Unfortuneatly the only way to do this is to slog through
  912          * the entire routing table looking for routes which point
  913          * to this interface...oh well...
  914          */
  915         for (i = 1; i <= AF_MAX; i++) {
  916                 for (j = 0; j < rt_numfibs; j++) {
  917                         rnh = rt_tables_get_rnh(j, i);
  918                         if (rnh == NULL)
  919                                 continue;
  920                         RADIX_NODE_HEAD_LOCK(rnh);
  921                         (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
  922                         RADIX_NODE_HEAD_UNLOCK(rnh);
  923                 }
  924         }
  925 
  926         /* Announce that the interface is gone. */
  927         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
  928         EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
  929         if (IS_DEFAULT_VNET(curvnet))
  930                 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
  931         if_delgroups(ifp);
  932 
  933         /*
  934          * We cannot hold the lock over dom_ifdetach calls as they might
  935          * sleep, for example trying to drain a callout, thus open up the
  936          * theoretical race with re-attaching.
  937          */
  938         IF_AFDATA_LOCK(ifp);
  939         i = ifp->if_afdata_initialized;
  940         ifp->if_afdata_initialized = 0;
  941         IF_AFDATA_UNLOCK(ifp);
  942         for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
  943                 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
  944                         (*dp->dom_ifdetach)(ifp,
  945                             ifp->if_afdata[dp->dom_family]);
  946         }
  947 }
  948 
  949 #ifdef VIMAGE
  950 /*
  951  * if_vmove() performs a limited version of if_detach() in current
  952  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
  953  * An attempt is made to shrink if_index in current vnet, find an
  954  * unused if_index in target vnet and calls if_grow() if necessary,
  955  * and finally find an unused if_xname for the target vnet.
  956  */
  957 void
  958 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
  959 {
  960         u_short idx;
  961 
  962         /*
  963          * Detach from current vnet, but preserve LLADDR info, do not
  964          * mark as dead etc. so that the ifnet can be reattached later.
  965          */
  966         if_detach_internal(ifp, 1);
  967 
  968         /*
  969          * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
  970          * the if_index for that vnet if possible.
  971          *
  972          * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
  973          * or we'd lock on one vnet and unlock on another.
  974          */
  975         IFNET_WLOCK();
  976         ifindex_free_locked(ifp->if_index);
  977 
  978         /*
  979          * Switch to the context of the target vnet.
  980          */
  981         CURVNET_SET_QUIET(new_vnet);
  982 
  983         if (ifindex_alloc_locked(&idx) != 0) {
  984                 IFNET_WUNLOCK();
  985                 panic("if_index overflow");
  986         }
  987         ifp->if_index = idx;
  988         ifnet_setbyindex_locked(ifp->if_index, ifp);
  989         IFNET_WUNLOCK();
  990 
  991         if_attach_internal(ifp, 1);
  992 
  993         CURVNET_RESTORE();
  994 }
  995 
  996 /*
  997  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
  998  */
  999 static int
 1000 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
 1001 {
 1002         struct prison *pr;
 1003         struct ifnet *difp;
 1004 
 1005         /* Try to find the prison within our visibility. */
 1006         sx_slock(&allprison_lock);
 1007         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1008         sx_sunlock(&allprison_lock);
 1009         if (pr == NULL)
 1010                 return (ENXIO);
 1011         prison_hold_locked(pr);
 1012         mtx_unlock(&pr->pr_mtx);
 1013 
 1014         /* Do not try to move the iface from and to the same prison. */
 1015         if (pr->pr_vnet == ifp->if_vnet) {
 1016                 prison_free(pr);
 1017                 return (EEXIST);
 1018         }
 1019 
 1020         /* Make sure the named iface does not exists in the dst. prison/vnet. */
 1021         /* XXX Lock interfaces to avoid races. */
 1022         CURVNET_SET_QUIET(pr->pr_vnet);
 1023         difp = ifunit(ifname);
 1024         CURVNET_RESTORE();
 1025         if (difp != NULL) {
 1026                 prison_free(pr);
 1027                 return (EEXIST);
 1028         }
 1029 
 1030         /* Move the interface into the child jail/vnet. */
 1031         if_vmove(ifp, pr->pr_vnet);
 1032 
 1033         /* Report the new if_xname back to the userland. */
 1034         sprintf(ifname, "%s", ifp->if_xname);
 1035 
 1036         prison_free(pr);
 1037         return (0);
 1038 }
 1039 
 1040 static int
 1041 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
 1042 {
 1043         struct prison *pr;
 1044         struct vnet *vnet_dst;
 1045         struct ifnet *ifp;
 1046 
 1047         /* Try to find the prison within our visibility. */
 1048         sx_slock(&allprison_lock);
 1049         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1050         sx_sunlock(&allprison_lock);
 1051         if (pr == NULL)
 1052                 return (ENXIO);
 1053         prison_hold_locked(pr);
 1054         mtx_unlock(&pr->pr_mtx);
 1055 
 1056         /* Make sure the named iface exists in the source prison/vnet. */
 1057         CURVNET_SET(pr->pr_vnet);
 1058         ifp = ifunit(ifname);           /* XXX Lock to avoid races. */
 1059         if (ifp == NULL) {
 1060                 CURVNET_RESTORE();
 1061                 prison_free(pr);
 1062                 return (ENXIO);
 1063         }
 1064 
 1065         /* Do not try to move the iface from and to the same prison. */
 1066         vnet_dst = TD_TO_VNET(td);
 1067         if (vnet_dst == ifp->if_vnet) {
 1068                 CURVNET_RESTORE();
 1069                 prison_free(pr);
 1070                 return (EEXIST);
 1071         }
 1072 
 1073         /* Get interface back from child jail/vnet. */
 1074         if_vmove(ifp, vnet_dst);
 1075         CURVNET_RESTORE();
 1076 
 1077         /* Report the new if_xname back to the userland. */
 1078         sprintf(ifname, "%s", ifp->if_xname);
 1079 
 1080         prison_free(pr);
 1081         return (0);
 1082 }
 1083 #endif /* VIMAGE */
 1084 
 1085 /*
 1086  * Add a group to an interface
 1087  */
 1088 int
 1089 if_addgroup(struct ifnet *ifp, const char *groupname)
 1090 {
 1091         struct ifg_list         *ifgl;
 1092         struct ifg_group        *ifg = NULL;
 1093         struct ifg_member       *ifgm;
 1094 
 1095         if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
 1096             groupname[strlen(groupname) - 1] <= '9')
 1097                 return (EINVAL);
 1098 
 1099         IFNET_WLOCK();
 1100         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1101                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
 1102                         IFNET_WUNLOCK();
 1103                         return (EEXIST);
 1104                 }
 1105 
 1106         if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
 1107             M_NOWAIT)) == NULL) {
 1108                 IFNET_WUNLOCK();
 1109                 return (ENOMEM);
 1110         }
 1111 
 1112         if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
 1113             M_TEMP, M_NOWAIT)) == NULL) {
 1114                 free(ifgl, M_TEMP);
 1115                 IFNET_WUNLOCK();
 1116                 return (ENOMEM);
 1117         }
 1118 
 1119         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1120                 if (!strcmp(ifg->ifg_group, groupname))
 1121                         break;
 1122 
 1123         if (ifg == NULL) {
 1124                 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
 1125                     M_TEMP, M_NOWAIT)) == NULL) {
 1126                         free(ifgl, M_TEMP);
 1127                         free(ifgm, M_TEMP);
 1128                         IFNET_WUNLOCK();
 1129                         return (ENOMEM);
 1130                 }
 1131                 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
 1132                 ifg->ifg_refcnt = 0;
 1133                 TAILQ_INIT(&ifg->ifg_members);
 1134                 EVENTHANDLER_INVOKE(group_attach_event, ifg);
 1135                 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
 1136         }
 1137 
 1138         ifg->ifg_refcnt++;
 1139         ifgl->ifgl_group = ifg;
 1140         ifgm->ifgm_ifp = ifp;
 1141 
 1142         IF_ADDR_LOCK(ifp);
 1143         TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
 1144         TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
 1145         IF_ADDR_UNLOCK(ifp);
 1146 
 1147         IFNET_WUNLOCK();
 1148 
 1149         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1150 
 1151         return (0);
 1152 }
 1153 
 1154 /*
 1155  * Remove a group from an interface
 1156  */
 1157 int
 1158 if_delgroup(struct ifnet *ifp, const char *groupname)
 1159 {
 1160         struct ifg_list         *ifgl;
 1161         struct ifg_member       *ifgm;
 1162 
 1163         IFNET_WLOCK();
 1164         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1165                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
 1166                         break;
 1167         if (ifgl == NULL) {
 1168                 IFNET_WUNLOCK();
 1169                 return (ENOENT);
 1170         }
 1171 
 1172         IF_ADDR_LOCK(ifp);
 1173         TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1174         IF_ADDR_UNLOCK(ifp);
 1175 
 1176         TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1177                 if (ifgm->ifgm_ifp == ifp)
 1178                         break;
 1179 
 1180         if (ifgm != NULL) {
 1181                 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
 1182                 free(ifgm, M_TEMP);
 1183         }
 1184 
 1185         if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1186                 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1187                 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
 1188                 free(ifgl->ifgl_group, M_TEMP);
 1189         }
 1190         IFNET_WUNLOCK();
 1191 
 1192         free(ifgl, M_TEMP);
 1193 
 1194         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1195 
 1196         return (0);
 1197 }
 1198 
 1199 /*
 1200  * Remove an interface from all groups
 1201  */
 1202 static void
 1203 if_delgroups(struct ifnet *ifp)
 1204 {
 1205         struct ifg_list         *ifgl;
 1206         struct ifg_member       *ifgm;
 1207         char groupname[IFNAMSIZ];
 1208 
 1209         IFNET_WLOCK();
 1210         while (!TAILQ_EMPTY(&ifp->if_groups)) {
 1211                 ifgl = TAILQ_FIRST(&ifp->if_groups);
 1212 
 1213                 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
 1214 
 1215                 IF_ADDR_LOCK(ifp);
 1216                 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1217                 IF_ADDR_UNLOCK(ifp);
 1218 
 1219                 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1220                         if (ifgm->ifgm_ifp == ifp)
 1221                                 break;
 1222 
 1223                 if (ifgm != NULL) {
 1224                         TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
 1225                             ifgm_next);
 1226                         free(ifgm, M_TEMP);
 1227                 }
 1228 
 1229                 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1230                         TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1231                         EVENTHANDLER_INVOKE(group_detach_event,
 1232                             ifgl->ifgl_group);
 1233                         free(ifgl->ifgl_group, M_TEMP);
 1234                 }
 1235                 IFNET_WUNLOCK();
 1236 
 1237                 free(ifgl, M_TEMP);
 1238 
 1239                 EVENTHANDLER_INVOKE(group_change_event, groupname);
 1240 
 1241                 IFNET_WLOCK();
 1242         }
 1243         IFNET_WUNLOCK();
 1244 }
 1245 
 1246 /*
 1247  * Stores all groups from an interface in memory pointed
 1248  * to by data
 1249  */
 1250 static int
 1251 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
 1252 {
 1253         int                      len, error;
 1254         struct ifg_list         *ifgl;
 1255         struct ifg_req           ifgrq, *ifgp;
 1256         struct ifgroupreq       *ifgr = data;
 1257 
 1258         if (ifgr->ifgr_len == 0) {
 1259                 IF_ADDR_LOCK(ifp);
 1260                 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1261                         ifgr->ifgr_len += sizeof(struct ifg_req);
 1262                 IF_ADDR_UNLOCK(ifp);
 1263                 return (0);
 1264         }
 1265 
 1266         len = ifgr->ifgr_len;
 1267         ifgp = ifgr->ifgr_groups;
 1268         /* XXX: wire */
 1269         IF_ADDR_LOCK(ifp);
 1270         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
 1271                 if (len < sizeof(ifgrq)) {
 1272                         IF_ADDR_UNLOCK(ifp);
 1273                         return (EINVAL);
 1274                 }
 1275                 bzero(&ifgrq, sizeof ifgrq);
 1276                 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
 1277                     sizeof(ifgrq.ifgrq_group));
 1278                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1279                         IF_ADDR_UNLOCK(ifp);
 1280                         return (error);
 1281                 }
 1282                 len -= sizeof(ifgrq);
 1283                 ifgp++;
 1284         }
 1285         IF_ADDR_UNLOCK(ifp);
 1286 
 1287         return (0);
 1288 }
 1289 
 1290 /*
 1291  * Stores all members of a group in memory pointed to by data
 1292  */
 1293 static int
 1294 if_getgroupmembers(struct ifgroupreq *data)
 1295 {
 1296         struct ifgroupreq       *ifgr = data;
 1297         struct ifg_group        *ifg;
 1298         struct ifg_member       *ifgm;
 1299         struct ifg_req           ifgrq, *ifgp;
 1300         int                      len, error;
 1301 
 1302         IFNET_RLOCK();
 1303         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1304                 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
 1305                         break;
 1306         if (ifg == NULL) {
 1307                 IFNET_RUNLOCK();
 1308                 return (ENOENT);
 1309         }
 1310 
 1311         if (ifgr->ifgr_len == 0) {
 1312                 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
 1313                         ifgr->ifgr_len += sizeof(ifgrq);
 1314                 IFNET_RUNLOCK();
 1315                 return (0);
 1316         }
 1317 
 1318         len = ifgr->ifgr_len;
 1319         ifgp = ifgr->ifgr_groups;
 1320         TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
 1321                 if (len < sizeof(ifgrq)) {
 1322                         IFNET_RUNLOCK();
 1323                         return (EINVAL);
 1324                 }
 1325                 bzero(&ifgrq, sizeof ifgrq);
 1326                 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
 1327                     sizeof(ifgrq.ifgrq_member));
 1328                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1329                         IFNET_RUNLOCK();
 1330                         return (error);
 1331                 }
 1332                 len -= sizeof(ifgrq);
 1333                 ifgp++;
 1334         }
 1335         IFNET_RUNLOCK();
 1336 
 1337         return (0);
 1338 }
 1339 
 1340 /*
 1341  * Delete Routes for a Network Interface
 1342  *
 1343  * Called for each routing entry via the rnh->rnh_walktree() call above
 1344  * to delete all route entries referencing a detaching network interface.
 1345  *
 1346  * Arguments:
 1347  *      rn      pointer to node in the routing table
 1348  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
 1349  *
 1350  * Returns:
 1351  *      0       successful
 1352  *      errno   failed - reason indicated
 1353  *
 1354  */
 1355 static int
 1356 if_rtdel(struct radix_node *rn, void *arg)
 1357 {
 1358         struct rtentry  *rt = (struct rtentry *)rn;
 1359         struct ifnet    *ifp = arg;
 1360         int             err;
 1361 
 1362         if (rt->rt_ifp == ifp) {
 1363 
 1364                 /*
 1365                  * Protect (sorta) against walktree recursion problems
 1366                  * with cloned routes
 1367                  */
 1368                 if ((rt->rt_flags & RTF_UP) == 0)
 1369                         return (0);
 1370 
 1371                 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
 1372                                 rt_mask(rt), rt->rt_flags|RTF_RNH_LOCKED,
 1373                                 (struct rtentry **) NULL, rt->rt_fibnum);
 1374                 if (err) {
 1375                         log(LOG_WARNING, "if_rtdel: error %d\n", err);
 1376                 }
 1377         }
 1378 
 1379         return (0);
 1380 }
 1381 
 1382 /*
 1383  * Wrapper functions for struct ifnet address list locking macros.  These are
 1384  * used by kernel modules to avoid encoding programming interface or binary
 1385  * interface assumptions that may be violated when kernel-internal locking
 1386  * approaches change.
 1387  */
 1388 void
 1389 if_addr_rlock(struct ifnet *ifp)
 1390 {
 1391 
 1392         IF_ADDR_LOCK(ifp);
 1393 }
 1394 
 1395 void
 1396 if_addr_runlock(struct ifnet *ifp)
 1397 {
 1398 
 1399         IF_ADDR_UNLOCK(ifp);
 1400 }
 1401 
 1402 void
 1403 if_maddr_rlock(struct ifnet *ifp)
 1404 {
 1405 
 1406         IF_ADDR_LOCK(ifp);
 1407 }
 1408 
 1409 void
 1410 if_maddr_runlock(struct ifnet *ifp)
 1411 {
 1412 
 1413         IF_ADDR_UNLOCK(ifp);
 1414 }
 1415 
 1416 /*
 1417  * Reference count functions for ifaddrs.
 1418  */
 1419 void
 1420 ifa_init(struct ifaddr *ifa)
 1421 {
 1422 
 1423         mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
 1424         refcount_init(&ifa->ifa_refcnt, 1);
 1425 }
 1426 
 1427 void
 1428 ifa_ref(struct ifaddr *ifa)
 1429 {
 1430 
 1431         refcount_acquire(&ifa->ifa_refcnt);
 1432 }
 1433 
 1434 void
 1435 ifa_free(struct ifaddr *ifa)
 1436 {
 1437 
 1438         if (refcount_release(&ifa->ifa_refcnt)) {
 1439                 mtx_destroy(&ifa->ifa_mtx);
 1440                 free(ifa, M_IFADDR);
 1441         }
 1442 }
 1443 
 1444 int
 1445 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1446 {
 1447         int error = 0;
 1448         struct rtentry *rt = NULL;
 1449         struct rt_addrinfo info;
 1450         static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1451 
 1452         bzero(&info, sizeof(info));
 1453         info.rti_ifp = V_loif;
 1454         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1455         info.rti_info[RTAX_DST] = ia;
 1456         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1457         error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
 1458 
 1459         if (error == 0 && rt != NULL) {
 1460                 RT_LOCK(rt);
 1461                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
 1462                         ifa->ifa_ifp->if_type;
 1463                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
 1464                         ifa->ifa_ifp->if_index;
 1465                 RT_REMREF(rt);
 1466                 RT_UNLOCK(rt);
 1467         } else if (error != 0)
 1468                 log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
 1469 
 1470         return (error);
 1471 }
 1472 
 1473 int
 1474 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1475 {
 1476         int error = 0;
 1477         struct rt_addrinfo info;
 1478         struct sockaddr_dl null_sdl;
 1479 
 1480         bzero(&null_sdl, sizeof(null_sdl));
 1481         null_sdl.sdl_len = sizeof(null_sdl);
 1482         null_sdl.sdl_family = AF_LINK;
 1483         null_sdl.sdl_type = ifa->ifa_ifp->if_type;
 1484         null_sdl.sdl_index = ifa->ifa_ifp->if_index;
 1485         bzero(&info, sizeof(info));
 1486         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1487         info.rti_info[RTAX_DST] = ia;
 1488         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1489         error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
 1490 
 1491         if (error != 0)
 1492                 log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
 1493 
 1494         return (error);
 1495 }
 1496 
 1497 /*
 1498  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
 1499  * structs used to represent other address families, it is necessary
 1500  * to perform a different comparison.
 1501  */
 1502 
 1503 #define sa_equal(a1, a2)        \
 1504         (bcmp((a1), (a2), ((a1))->sa_len) == 0)
 1505 
 1506 #define sa_dl_equal(a1, a2)     \
 1507         ((((struct sockaddr_dl *)(a1))->sdl_len ==                      \
 1508          ((struct sockaddr_dl *)(a2))->sdl_len) &&                      \
 1509          (bcmp(LLADDR((struct sockaddr_dl *)(a1)),                      \
 1510                LLADDR((struct sockaddr_dl *)(a2)),                      \
 1511                ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
 1512 
 1513 /*
 1514  * Locate an interface based on a complete address.
 1515  */
 1516 /*ARGSUSED*/
 1517 static struct ifaddr *
 1518 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
 1519 {
 1520         struct ifnet *ifp;
 1521         struct ifaddr *ifa;
 1522 
 1523         IFNET_RLOCK_NOSLEEP();
 1524         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1525                 IF_ADDR_LOCK(ifp);
 1526                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1527                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1528                                 continue;
 1529                         if (sa_equal(addr, ifa->ifa_addr)) {
 1530                                 if (getref)
 1531                                         ifa_ref(ifa);
 1532                                 IF_ADDR_UNLOCK(ifp);
 1533                                 goto done;
 1534                         }
 1535                         /* IP6 doesn't have broadcast */
 1536                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1537                             ifa->ifa_broadaddr &&
 1538                             ifa->ifa_broadaddr->sa_len != 0 &&
 1539                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1540                                 if (getref)
 1541                                         ifa_ref(ifa);
 1542                                 IF_ADDR_UNLOCK(ifp);
 1543                                 goto done;
 1544                         }
 1545                 }
 1546                 IF_ADDR_UNLOCK(ifp);
 1547         }
 1548         ifa = NULL;
 1549 done:
 1550         IFNET_RUNLOCK_NOSLEEP();
 1551         return (ifa);
 1552 }
 1553 
 1554 struct ifaddr *
 1555 ifa_ifwithaddr(struct sockaddr *addr)
 1556 {
 1557 
 1558         return (ifa_ifwithaddr_internal(addr, 1));
 1559 }
 1560 
 1561 int
 1562 ifa_ifwithaddr_check(struct sockaddr *addr)
 1563 {
 1564 
 1565         return (ifa_ifwithaddr_internal(addr, 0) != NULL);
 1566 }
 1567 
 1568 /*
 1569  * Locate an interface based on the broadcast address.
 1570  */
 1571 /* ARGSUSED */
 1572 struct ifaddr *
 1573 ifa_ifwithbroadaddr(struct sockaddr *addr)
 1574 {
 1575         struct ifnet *ifp;
 1576         struct ifaddr *ifa;
 1577 
 1578         IFNET_RLOCK_NOSLEEP();
 1579         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1580                 IF_ADDR_LOCK(ifp);
 1581                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1582                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1583                                 continue;
 1584                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1585                             ifa->ifa_broadaddr &&
 1586                             ifa->ifa_broadaddr->sa_len != 0 &&
 1587                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1588                                 ifa_ref(ifa);
 1589                                 IF_ADDR_UNLOCK(ifp);
 1590                                 goto done;
 1591                         }
 1592                 }
 1593                 IF_ADDR_UNLOCK(ifp);
 1594         }
 1595         ifa = NULL;
 1596 done:
 1597         IFNET_RUNLOCK_NOSLEEP();
 1598         return (ifa);
 1599 }
 1600 
 1601 /*
 1602  * Locate the point to point interface with a given destination address.
 1603  */
 1604 /*ARGSUSED*/
 1605 struct ifaddr *
 1606 ifa_ifwithdstaddr(struct sockaddr *addr)
 1607 {
 1608         struct ifnet *ifp;
 1609         struct ifaddr *ifa;
 1610 
 1611         IFNET_RLOCK_NOSLEEP();
 1612         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1613                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
 1614                         continue;
 1615                 IF_ADDR_LOCK(ifp);
 1616                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1617                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1618                                 continue;
 1619                         if (ifa->ifa_dstaddr != NULL &&
 1620                             sa_equal(addr, ifa->ifa_dstaddr)) {
 1621                                 ifa_ref(ifa);
 1622                                 IF_ADDR_UNLOCK(ifp);
 1623                                 goto done;
 1624                         }
 1625                 }
 1626                 IF_ADDR_UNLOCK(ifp);
 1627         }
 1628         ifa = NULL;
 1629 done:
 1630         IFNET_RUNLOCK_NOSLEEP();
 1631         return (ifa);
 1632 }
 1633 
 1634 /*
 1635  * Find an interface on a specific network.  If many, choice
 1636  * is most specific found.
 1637  */
 1638 struct ifaddr *
 1639 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
 1640 {
 1641         struct ifnet *ifp;
 1642         struct ifaddr *ifa;
 1643         struct ifaddr *ifa_maybe = NULL;
 1644         u_int af = addr->sa_family;
 1645         char *addr_data = addr->sa_data, *cplim;
 1646 
 1647         /*
 1648          * AF_LINK addresses can be looked up directly by their index number,
 1649          * so do that if we can.
 1650          */
 1651         if (af == AF_LINK) {
 1652             struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
 1653             if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
 1654                 return (ifaddr_byindex(sdl->sdl_index));
 1655         }
 1656 
 1657         /*
 1658          * Scan though each interface, looking for ones that have addresses
 1659          * in this address family.  Maintain a reference on ifa_maybe once
 1660          * we find one, as we release the IF_ADDR_LOCK() that kept it stable
 1661          * when we move onto the next interface.
 1662          */
 1663         IFNET_RLOCK_NOSLEEP();
 1664         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1665                 IF_ADDR_LOCK(ifp);
 1666                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1667                         char *cp, *cp2, *cp3;
 1668 
 1669                         if (ifa->ifa_addr->sa_family != af)
 1670 next:                           continue;
 1671                         if (af == AF_INET && 
 1672                             ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
 1673                                 /*
 1674                                  * This is a bit broken as it doesn't
 1675                                  * take into account that the remote end may
 1676                                  * be a single node in the network we are
 1677                                  * looking for.
 1678                                  * The trouble is that we don't know the
 1679                                  * netmask for the remote end.
 1680                                  */
 1681                                 if (ifa->ifa_dstaddr != NULL &&
 1682                                     sa_equal(addr, ifa->ifa_dstaddr)) {
 1683                                         ifa_ref(ifa);
 1684                                         IF_ADDR_UNLOCK(ifp);
 1685                                         goto done;
 1686                                 }
 1687                         } else {
 1688                                 /*
 1689                                  * if we have a special address handler,
 1690                                  * then use it instead of the generic one.
 1691                                  */
 1692                                 if (ifa->ifa_claim_addr) {
 1693                                         if ((*ifa->ifa_claim_addr)(ifa, addr)) {
 1694                                                 ifa_ref(ifa);
 1695                                                 IF_ADDR_UNLOCK(ifp);
 1696                                                 goto done;
 1697                                         }
 1698                                         continue;
 1699                                 }
 1700 
 1701                                 /*
 1702                                  * Scan all the bits in the ifa's address.
 1703                                  * If a bit dissagrees with what we are
 1704                                  * looking for, mask it with the netmask
 1705                                  * to see if it really matters.
 1706                                  * (A byte at a time)
 1707                                  */
 1708                                 if (ifa->ifa_netmask == 0)
 1709                                         continue;
 1710                                 cp = addr_data;
 1711                                 cp2 = ifa->ifa_addr->sa_data;
 1712                                 cp3 = ifa->ifa_netmask->sa_data;
 1713                                 cplim = ifa->ifa_netmask->sa_len
 1714                                         + (char *)ifa->ifa_netmask;
 1715                                 while (cp3 < cplim)
 1716                                         if ((*cp++ ^ *cp2++) & *cp3++)
 1717                                                 goto next; /* next address! */
 1718                                 /*
 1719                                  * If the netmask of what we just found
 1720                                  * is more specific than what we had before
 1721                                  * (if we had one) then remember the new one
 1722                                  * before continuing to search
 1723                                  * for an even better one.
 1724                                  */
 1725                                 if (ifa_maybe == NULL ||
 1726                                     rn_refines((caddr_t)ifa->ifa_netmask,
 1727                                     (caddr_t)ifa_maybe->ifa_netmask)) {
 1728                                         if (ifa_maybe != NULL)
 1729                                                 ifa_free(ifa_maybe);
 1730                                         ifa_maybe = ifa;
 1731                                         ifa_ref(ifa_maybe);
 1732                                 }
 1733                         }
 1734                 }
 1735                 IF_ADDR_UNLOCK(ifp);
 1736         }
 1737         ifa = ifa_maybe;
 1738         ifa_maybe = NULL;
 1739 done:
 1740         IFNET_RUNLOCK_NOSLEEP();
 1741         if (ifa_maybe != NULL)
 1742                 ifa_free(ifa_maybe);
 1743         return (ifa);
 1744 }
 1745 
 1746 /*
 1747  * Find an interface address specific to an interface best matching
 1748  * a given address.
 1749  */
 1750 struct ifaddr *
 1751 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
 1752 {
 1753         struct ifaddr *ifa;
 1754         char *cp, *cp2, *cp3;
 1755         char *cplim;
 1756         struct ifaddr *ifa_maybe = NULL;
 1757         u_int af = addr->sa_family;
 1758 
 1759         if (af >= AF_MAX)
 1760                 return (0);
 1761         IF_ADDR_LOCK(ifp);
 1762         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1763                 if (ifa->ifa_addr->sa_family != af)
 1764                         continue;
 1765                 if (ifa_maybe == NULL)
 1766                         ifa_maybe = ifa;
 1767                 if (ifa->ifa_netmask == 0) {
 1768                         if (sa_equal(addr, ifa->ifa_addr) ||
 1769                             (ifa->ifa_dstaddr &&
 1770                             sa_equal(addr, ifa->ifa_dstaddr)))
 1771                                 goto done;
 1772                         continue;
 1773                 }
 1774                 if (ifp->if_flags & IFF_POINTOPOINT) {
 1775                         if (sa_equal(addr, ifa->ifa_dstaddr))
 1776                                 goto done;
 1777                 } else {
 1778                         cp = addr->sa_data;
 1779                         cp2 = ifa->ifa_addr->sa_data;
 1780                         cp3 = ifa->ifa_netmask->sa_data;
 1781                         cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
 1782                         for (; cp3 < cplim; cp3++)
 1783                                 if ((*cp++ ^ *cp2++) & *cp3)
 1784                                         break;
 1785                         if (cp3 == cplim)
 1786                                 goto done;
 1787                 }
 1788         }
 1789         ifa = ifa_maybe;
 1790 done:
 1791         if (ifa != NULL)
 1792                 ifa_ref(ifa);
 1793         IF_ADDR_UNLOCK(ifp);
 1794         return (ifa);
 1795 }
 1796 
 1797 #include <net/if_llatbl.h>
 1798 
 1799 /*
 1800  * Default action when installing a route with a Link Level gateway.
 1801  * Lookup an appropriate real ifa to point to.
 1802  * This should be moved to /sys/net/link.c eventually.
 1803  */
 1804 static void
 1805 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
 1806 {
 1807         struct ifaddr *ifa, *oifa;
 1808         struct sockaddr *dst;
 1809         struct ifnet *ifp;
 1810 
 1811         RT_LOCK_ASSERT(rt);
 1812 
 1813         if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
 1814             ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
 1815                 return;
 1816         ifa = ifaof_ifpforaddr(dst, ifp);
 1817         if (ifa) {
 1818                 oifa = rt->rt_ifa;
 1819                 rt->rt_ifa = ifa;
 1820                 ifa_free(oifa);
 1821                 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
 1822                         ifa->ifa_rtrequest(cmd, rt, info);
 1823         }
 1824 }
 1825 
 1826 /*
 1827  * Mark an interface down and notify protocols of
 1828  * the transition.
 1829  * NOTE: must be called at splnet or eqivalent.
 1830  */
 1831 static void
 1832 if_unroute(struct ifnet *ifp, int flag, int fam)
 1833 {
 1834         struct ifaddr *ifa;
 1835 
 1836         KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
 1837 
 1838         ifp->if_flags &= ~flag;
 1839         getmicrotime(&ifp->if_lastchange);
 1840         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1841                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1842                         pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
 1843         ifp->if_qflush(ifp);
 1844 
 1845 #if defined(INET) || defined(INET6)
 1846 #ifdef DEV_CARP
 1847         if (ifp->if_carp)
 1848                 carp_carpdev_state(ifp->if_carp);
 1849 #endif
 1850 #endif
 1851         rt_ifmsg(ifp);
 1852 }
 1853 
 1854 /*
 1855  * Mark an interface up and notify protocols of
 1856  * the transition.
 1857  * NOTE: must be called at splnet or eqivalent.
 1858  */
 1859 static void
 1860 if_route(struct ifnet *ifp, int flag, int fam)
 1861 {
 1862         struct ifaddr *ifa;
 1863 
 1864         KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
 1865 
 1866         ifp->if_flags |= flag;
 1867         getmicrotime(&ifp->if_lastchange);
 1868         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1869                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1870                         pfctlinput(PRC_IFUP, ifa->ifa_addr);
 1871 #if defined(INET) || defined(INET6)
 1872 #ifdef DEV_CARP
 1873         if (ifp->if_carp)
 1874                 carp_carpdev_state(ifp->if_carp);
 1875 #endif
 1876 #endif
 1877         rt_ifmsg(ifp);
 1878 #ifdef INET6
 1879         in6_if_up(ifp);
 1880 #endif
 1881 }
 1882 
 1883 void    (*vlan_link_state_p)(struct ifnet *, int);      /* XXX: private from if_vlan */
 1884 void    (*vlan_trunk_cap_p)(struct ifnet *);            /* XXX: private from if_vlan */
 1885 
 1886 /*
 1887  * Handle a change in the interface link state. To avoid LORs
 1888  * between driver lock and upper layer locks, as well as possible
 1889  * recursions, we post event to taskqueue, and all job
 1890  * is done in static do_link_state_change().
 1891  */
 1892 void
 1893 if_link_state_change(struct ifnet *ifp, int link_state)
 1894 {
 1895         /* Return if state hasn't changed. */
 1896         if (ifp->if_link_state == link_state)
 1897                 return;
 1898 
 1899         ifp->if_link_state = link_state;
 1900 
 1901         taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
 1902 }
 1903 
 1904 static void
 1905 do_link_state_change(void *arg, int pending)
 1906 {
 1907         struct ifnet *ifp = (struct ifnet *)arg;
 1908         int link_state = ifp->if_link_state;
 1909         CURVNET_SET(ifp->if_vnet);
 1910 
 1911         /* Notify that the link state has changed. */
 1912         rt_ifmsg(ifp);
 1913         if (ifp->if_vlantrunk != NULL)
 1914                 (*vlan_link_state_p)(ifp, 0);
 1915 
 1916         if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
 1917             IFP2AC(ifp)->ac_netgraph != NULL)
 1918                 (*ng_ether_link_state_p)(ifp, link_state);
 1919 #if defined(INET) || defined(INET6)
 1920 #ifdef DEV_CARP
 1921         if (ifp->if_carp)
 1922                 carp_carpdev_state(ifp->if_carp);
 1923 #endif
 1924 #endif
 1925         if (ifp->if_bridge) {
 1926                 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
 1927                 (*bstp_linkstate_p)(ifp, link_state);
 1928         }
 1929         if (ifp->if_lagg) {
 1930                 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
 1931                 (*lagg_linkstate_p)(ifp, link_state);
 1932         }
 1933 
 1934         if (IS_DEFAULT_VNET(curvnet))
 1935                 devctl_notify("IFNET", ifp->if_xname,
 1936                     (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
 1937                     NULL);
 1938         if (pending > 1)
 1939                 if_printf(ifp, "%d link states coalesced\n", pending);
 1940         if (log_link_state_change)
 1941                 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
 1942                     (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
 1943         CURVNET_RESTORE();
 1944 }
 1945 
 1946 /*
 1947  * Mark an interface down and notify protocols of
 1948  * the transition.
 1949  * NOTE: must be called at splnet or eqivalent.
 1950  */
 1951 void
 1952 if_down(struct ifnet *ifp)
 1953 {
 1954 
 1955         if_unroute(ifp, IFF_UP, AF_UNSPEC);
 1956 }
 1957 
 1958 /*
 1959  * Mark an interface up and notify protocols of
 1960  * the transition.
 1961  * NOTE: must be called at splnet or eqivalent.
 1962  */
 1963 void
 1964 if_up(struct ifnet *ifp)
 1965 {
 1966 
 1967         if_route(ifp, IFF_UP, AF_UNSPEC);
 1968 }
 1969 
 1970 /*
 1971  * Flush an interface queue.
 1972  */
 1973 void
 1974 if_qflush(struct ifnet *ifp)
 1975 {
 1976         struct mbuf *m, *n;
 1977         struct ifaltq *ifq;
 1978         
 1979         ifq = &ifp->if_snd;
 1980         IFQ_LOCK(ifq);
 1981 #ifdef ALTQ
 1982         if (ALTQ_IS_ENABLED(ifq))
 1983                 ALTQ_PURGE(ifq);
 1984 #endif
 1985         n = ifq->ifq_head;
 1986         while ((m = n) != 0) {
 1987                 n = m->m_act;
 1988                 m_freem(m);
 1989         }
 1990         ifq->ifq_head = 0;
 1991         ifq->ifq_tail = 0;
 1992         ifq->ifq_len = 0;
 1993         IFQ_UNLOCK(ifq);
 1994 }
 1995 
 1996 /*
 1997  * Handle interface watchdog timer routines.  Called
 1998  * from softclock, we decrement timers (if set) and
 1999  * call the appropriate interface routine on expiration.
 2000  *
 2001  * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
 2002  * holding Giant.
 2003  */
 2004 static void
 2005 if_slowtimo(void *arg)
 2006 {
 2007         VNET_ITERATOR_DECL(vnet_iter);
 2008         struct ifnet *ifp;
 2009         int s = splimp();
 2010 
 2011         VNET_LIST_RLOCK_NOSLEEP();
 2012         IFNET_RLOCK_NOSLEEP();
 2013         VNET_FOREACH(vnet_iter) {
 2014                 CURVNET_SET(vnet_iter);
 2015                 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2016                         if (ifp->if_timer == 0 || --ifp->if_timer)
 2017                                 continue;
 2018                         if (ifp->if_watchdog)
 2019                                 (*ifp->if_watchdog)(ifp);
 2020                 }
 2021                 CURVNET_RESTORE();
 2022         }
 2023         IFNET_RUNLOCK_NOSLEEP();
 2024         VNET_LIST_RUNLOCK_NOSLEEP();
 2025         splx(s);
 2026         timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
 2027 }
 2028 
 2029 /*
 2030  * Map interface name to interface structure pointer, with or without
 2031  * returning a reference.
 2032  */
 2033 struct ifnet *
 2034 ifunit_ref(const char *name)
 2035 {
 2036         struct ifnet *ifp;
 2037 
 2038         IFNET_RLOCK_NOSLEEP();
 2039         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2040                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
 2041                     !(ifp->if_flags & IFF_DYING))
 2042                         break;
 2043         }
 2044         if (ifp != NULL)
 2045                 if_ref(ifp);
 2046         IFNET_RUNLOCK_NOSLEEP();
 2047         return (ifp);
 2048 }
 2049 
 2050 struct ifnet *
 2051 ifunit(const char *name)
 2052 {
 2053         struct ifnet *ifp;
 2054 
 2055         IFNET_RLOCK_NOSLEEP();
 2056         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2057                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
 2058                         break;
 2059         }
 2060         IFNET_RUNLOCK_NOSLEEP();
 2061         return (ifp);
 2062 }
 2063 
 2064 /*
 2065  * Hardware specific interface ioctls.
 2066  */
 2067 static int
 2068 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
 2069 {
 2070         struct ifreq *ifr;
 2071         struct ifstat *ifs;
 2072         int error = 0;
 2073         int new_flags, temp_flags;
 2074         size_t namelen, onamelen;
 2075         size_t descrlen;
 2076         char *descrbuf, *odescrbuf;
 2077         char new_name[IFNAMSIZ];
 2078         struct ifaddr *ifa;
 2079         struct sockaddr_dl *sdl;
 2080 
 2081         ifr = (struct ifreq *)data;
 2082         switch (cmd) {
 2083         case SIOCGIFINDEX:
 2084                 ifr->ifr_index = ifp->if_index;
 2085                 break;
 2086 
 2087         case SIOCGIFFLAGS:
 2088                 temp_flags = ifp->if_flags | ifp->if_drv_flags;
 2089                 ifr->ifr_flags = temp_flags & 0xffff;
 2090                 ifr->ifr_flagshigh = temp_flags >> 16;
 2091                 break;
 2092 
 2093         case SIOCGIFCAP:
 2094                 ifr->ifr_reqcap = ifp->if_capabilities;
 2095                 ifr->ifr_curcap = ifp->if_capenable;
 2096                 break;
 2097 
 2098 #ifdef MAC
 2099         case SIOCGIFMAC:
 2100                 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
 2101                 break;
 2102 #endif
 2103 
 2104         case SIOCGIFMETRIC:
 2105                 ifr->ifr_metric = ifp->if_metric;
 2106                 break;
 2107 
 2108         case SIOCGIFMTU:
 2109                 ifr->ifr_mtu = ifp->if_mtu;
 2110                 break;
 2111 
 2112         case SIOCGIFPHYS:
 2113                 ifr->ifr_phys = ifp->if_physical;
 2114                 break;
 2115 
 2116         case SIOCGIFDESCR:
 2117                 error = 0;
 2118                 sx_slock(&ifdescr_sx);
 2119                 if (ifp->if_description == NULL)
 2120                         error = ENOMSG;
 2121                 else {
 2122                         /* space for terminating nul */
 2123                         descrlen = strlen(ifp->if_description) + 1;
 2124                         if (ifr->ifr_buffer.length < descrlen)
 2125                                 ifr->ifr_buffer.buffer = NULL;
 2126                         else
 2127                                 error = copyout(ifp->if_description,
 2128                                     ifr->ifr_buffer.buffer, descrlen);
 2129                         ifr->ifr_buffer.length = descrlen;
 2130                 }
 2131                 sx_sunlock(&ifdescr_sx);
 2132                 break;
 2133 
 2134         case SIOCSIFDESCR:
 2135                 error = priv_check(td, PRIV_NET_SETIFDESCR);
 2136                 if (error)
 2137                         return (error);
 2138 
 2139                 /*
 2140                  * Copy only (length-1) bytes to make sure that
 2141                  * if_description is always nul terminated.  The
 2142                  * length parameter is supposed to count the
 2143                  * terminating nul in.
 2144                  */
 2145                 if (ifr->ifr_buffer.length > ifdescr_maxlen)
 2146                         return (ENAMETOOLONG);
 2147                 else if (ifr->ifr_buffer.length == 0)
 2148                         descrbuf = NULL;
 2149                 else {
 2150                         descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
 2151                             M_WAITOK | M_ZERO);
 2152                         error = copyin(ifr->ifr_buffer.buffer, descrbuf,
 2153                             ifr->ifr_buffer.length - 1);
 2154                         if (error) {
 2155                                 free(descrbuf, M_IFDESCR);
 2156                                 break;
 2157                         }
 2158                 }
 2159 
 2160                 sx_xlock(&ifdescr_sx);
 2161                 odescrbuf = ifp->if_description;
 2162                 ifp->if_description = descrbuf;
 2163                 sx_xunlock(&ifdescr_sx);
 2164 
 2165                 getmicrotime(&ifp->if_lastchange);
 2166                 free(odescrbuf, M_IFDESCR);
 2167                 break;
 2168 
 2169         case SIOCSIFFLAGS:
 2170                 error = priv_check(td, PRIV_NET_SETIFFLAGS);
 2171                 if (error)
 2172                         return (error);
 2173                 /*
 2174                  * Currently, no driver owned flags pass the IFF_CANTCHANGE
 2175                  * check, so we don't need special handling here yet.
 2176                  */
 2177                 new_flags = (ifr->ifr_flags & 0xffff) |
 2178                     (ifr->ifr_flagshigh << 16);
 2179                 if (ifp->if_flags & IFF_SMART) {
 2180                         /* Smart drivers twiddle their own routes */
 2181                 } else if (ifp->if_flags & IFF_UP &&
 2182                     (new_flags & IFF_UP) == 0) {
 2183                         int s = splimp();
 2184                         if_down(ifp);
 2185                         splx(s);
 2186                 } else if (new_flags & IFF_UP &&
 2187                     (ifp->if_flags & IFF_UP) == 0) {
 2188                         int s = splimp();
 2189                         if_up(ifp);
 2190                         splx(s);
 2191                 }
 2192                 /* See if permanently promiscuous mode bit is about to flip */
 2193                 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
 2194                         if (new_flags & IFF_PPROMISC)
 2195                                 ifp->if_flags |= IFF_PROMISC;
 2196                         else if (ifp->if_pcount == 0)
 2197                                 ifp->if_flags &= ~IFF_PROMISC;
 2198                         log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
 2199                             ifp->if_xname,
 2200                             (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
 2201                 }
 2202                 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
 2203                         (new_flags &~ IFF_CANTCHANGE);
 2204                 if (ifp->if_ioctl) {
 2205                         (void) (*ifp->if_ioctl)(ifp, cmd, data);
 2206                 }
 2207                 getmicrotime(&ifp->if_lastchange);
 2208                 break;
 2209 
 2210         case SIOCSIFCAP:
 2211                 error = priv_check(td, PRIV_NET_SETIFCAP);
 2212                 if (error)
 2213                         return (error);
 2214                 if (ifp->if_ioctl == NULL)
 2215                         return (EOPNOTSUPP);
 2216                 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
 2217                         return (EINVAL);
 2218                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2219                 if (error == 0)
 2220                         getmicrotime(&ifp->if_lastchange);
 2221                 break;
 2222 
 2223 #ifdef MAC
 2224         case SIOCSIFMAC:
 2225                 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
 2226                 break;
 2227 #endif
 2228 
 2229         case SIOCSIFNAME:
 2230                 error = priv_check(td, PRIV_NET_SETIFNAME);
 2231                 if (error)
 2232                         return (error);
 2233                 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
 2234                 if (error != 0)
 2235                         return (error);
 2236                 if (new_name[0] == '\0')
 2237                         return (EINVAL);
 2238                 if (ifunit(new_name) != NULL)
 2239                         return (EEXIST);
 2240 
 2241                 /*
 2242                  * XXX: Locking.  Nothing else seems to lock if_flags,
 2243                  * and there are numerous other races with the
 2244                  * ifunit() checks not being atomic with namespace
 2245                  * changes (renames, vmoves, if_attach, etc).
 2246                  */
 2247                 ifp->if_flags |= IFF_RENAMING;
 2248                 
 2249                 /* Announce the departure of the interface. */
 2250                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 2251                 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 2252 
 2253                 log(LOG_INFO, "%s: changing name to '%s'\n",
 2254                     ifp->if_xname, new_name);
 2255 
 2256                 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
 2257                 ifa = ifp->if_addr;
 2258                 IFA_LOCK(ifa);
 2259                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 2260                 namelen = strlen(new_name);
 2261                 onamelen = sdl->sdl_nlen;
 2262                 /*
 2263                  * Move the address if needed.  This is safe because we
 2264                  * allocate space for a name of length IFNAMSIZ when we
 2265                  * create this in if_attach().
 2266                  */
 2267                 if (namelen != onamelen) {
 2268                         bcopy(sdl->sdl_data + onamelen,
 2269                             sdl->sdl_data + namelen, sdl->sdl_alen);
 2270                 }
 2271                 bcopy(new_name, sdl->sdl_data, namelen);
 2272                 sdl->sdl_nlen = namelen;
 2273                 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
 2274                 bzero(sdl->sdl_data, onamelen);
 2275                 while (namelen != 0)
 2276                         sdl->sdl_data[--namelen] = 0xff;
 2277                 IFA_UNLOCK(ifa);
 2278 
 2279                 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
 2280                 /* Announce the return of the interface. */
 2281                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 2282 
 2283                 ifp->if_flags &= ~IFF_RENAMING;
 2284                 break;
 2285 
 2286 #ifdef VIMAGE
 2287         case SIOCSIFVNET:
 2288                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2289                 if (error)
 2290                         return (error);
 2291                 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
 2292                 break;
 2293 #endif
 2294 
 2295         case SIOCSIFMETRIC:
 2296                 error = priv_check(td, PRIV_NET_SETIFMETRIC);
 2297                 if (error)
 2298                         return (error);
 2299                 ifp->if_metric = ifr->ifr_metric;
 2300                 getmicrotime(&ifp->if_lastchange);
 2301                 break;
 2302 
 2303         case SIOCSIFPHYS:
 2304                 error = priv_check(td, PRIV_NET_SETIFPHYS);
 2305                 if (error)
 2306                         return (error);
 2307                 if (ifp->if_ioctl == NULL)
 2308                         return (EOPNOTSUPP);
 2309                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2310                 if (error == 0)
 2311                         getmicrotime(&ifp->if_lastchange);
 2312                 break;
 2313 
 2314         case SIOCSIFMTU:
 2315         {
 2316                 u_long oldmtu = ifp->if_mtu;
 2317 
 2318                 error = priv_check(td, PRIV_NET_SETIFMTU);
 2319                 if (error)
 2320                         return (error);
 2321                 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
 2322                         return (EINVAL);
 2323                 if (ifp->if_ioctl == NULL)
 2324                         return (EOPNOTSUPP);
 2325                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2326                 if (error == 0) {
 2327                         getmicrotime(&ifp->if_lastchange);
 2328                         rt_ifmsg(ifp);
 2329                 }
 2330                 /*
 2331                  * If the link MTU changed, do network layer specific procedure.
 2332                  */
 2333                 if (ifp->if_mtu != oldmtu) {
 2334 #ifdef INET6
 2335                         nd6_setmtu(ifp);
 2336 #endif
 2337                 }
 2338                 break;
 2339         }
 2340 
 2341         case SIOCADDMULTI:
 2342         case SIOCDELMULTI:
 2343                 if (cmd == SIOCADDMULTI)
 2344                         error = priv_check(td, PRIV_NET_ADDMULTI);
 2345                 else
 2346                         error = priv_check(td, PRIV_NET_DELMULTI);
 2347                 if (error)
 2348                         return (error);
 2349 
 2350                 /* Don't allow group membership on non-multicast interfaces. */
 2351                 if ((ifp->if_flags & IFF_MULTICAST) == 0)
 2352                         return (EOPNOTSUPP);
 2353 
 2354                 /* Don't let users screw up protocols' entries. */
 2355                 if (ifr->ifr_addr.sa_family != AF_LINK)
 2356                         return (EINVAL);
 2357 
 2358                 if (cmd == SIOCADDMULTI) {
 2359                         struct ifmultiaddr *ifma;
 2360 
 2361                         /*
 2362                          * Userland is only permitted to join groups once
 2363                          * via the if_addmulti() KPI, because it cannot hold
 2364                          * struct ifmultiaddr * between calls. It may also
 2365                          * lose a race while we check if the membership
 2366                          * already exists.
 2367                          */
 2368                         IF_ADDR_LOCK(ifp);
 2369                         ifma = if_findmulti(ifp, &ifr->ifr_addr);
 2370                         IF_ADDR_UNLOCK(ifp);
 2371                         if (ifma != NULL)
 2372                                 error = EADDRINUSE;
 2373                         else
 2374                                 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
 2375                 } else {
 2376                         error = if_delmulti(ifp, &ifr->ifr_addr);
 2377                 }
 2378                 if (error == 0)
 2379                         getmicrotime(&ifp->if_lastchange);
 2380                 break;
 2381 
 2382         case SIOCSIFPHYADDR:
 2383         case SIOCDIFPHYADDR:
 2384 #ifdef INET6
 2385         case SIOCSIFPHYADDR_IN6:
 2386 #endif
 2387         case SIOCSLIFPHYADDR:
 2388         case SIOCSIFMEDIA:
 2389         case SIOCSIFGENERIC:
 2390                 error = priv_check(td, PRIV_NET_HWIOCTL);
 2391                 if (error)
 2392                         return (error);
 2393                 if (ifp->if_ioctl == NULL)
 2394                         return (EOPNOTSUPP);
 2395                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2396                 if (error == 0)
 2397                         getmicrotime(&ifp->if_lastchange);
 2398                 break;
 2399 
 2400         case SIOCGIFSTATUS:
 2401                 ifs = (struct ifstat *)data;
 2402                 ifs->ascii[0] = '\0';
 2403 
 2404         case SIOCGIFPSRCADDR:
 2405         case SIOCGIFPDSTADDR:
 2406         case SIOCGLIFPHYADDR:
 2407         case SIOCGIFMEDIA:
 2408         case SIOCGIFGENERIC:
 2409                 if (ifp->if_ioctl == NULL)
 2410                         return (EOPNOTSUPP);
 2411                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2412                 break;
 2413 
 2414         case SIOCSIFLLADDR:
 2415                 error = priv_check(td, PRIV_NET_SETLLADDR);
 2416                 if (error)
 2417                         return (error);
 2418                 error = if_setlladdr(ifp,
 2419                     ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
 2420                 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
 2421                 break;
 2422 
 2423         case SIOCAIFGROUP:
 2424         {
 2425                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2426 
 2427                 error = priv_check(td, PRIV_NET_ADDIFGROUP);
 2428                 if (error)
 2429                         return (error);
 2430                 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
 2431                         return (error);
 2432                 break;
 2433         }
 2434 
 2435         case SIOCGIFGROUP:
 2436                 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
 2437                         return (error);
 2438                 break;
 2439 
 2440         case SIOCDIFGROUP:
 2441         {
 2442                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2443 
 2444                 error = priv_check(td, PRIV_NET_DELIFGROUP);
 2445                 if (error)
 2446                         return (error);
 2447                 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
 2448                         return (error);
 2449                 break;
 2450         }
 2451 
 2452         default:
 2453                 error = ENOIOCTL;
 2454                 break;
 2455         }
 2456         return (error);
 2457 }
 2458 
 2459 /*
 2460  * Interface ioctls.
 2461  */
 2462 int
 2463 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
 2464 {
 2465         struct ifnet *ifp;
 2466         struct ifreq *ifr;
 2467         int error;
 2468         int oif_flags;
 2469 
 2470         switch (cmd) {
 2471         case SIOCGIFCONF:
 2472         case OSIOCGIFCONF:
 2473 #ifdef __amd64__
 2474         case SIOCGIFCONF32:
 2475 #endif
 2476                 return (ifconf(cmd, data));
 2477         }
 2478         ifr = (struct ifreq *)data;
 2479 
 2480         switch (cmd) {
 2481 #ifdef VIMAGE
 2482         case SIOCSIFRVNET:
 2483                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2484                 if (error)
 2485                         return (error);
 2486                 return (if_vmove_reclaim(td, ifr->ifr_name, ifr->ifr_jid));
 2487 #endif
 2488         case SIOCIFCREATE:
 2489         case SIOCIFCREATE2:
 2490                 error = priv_check(td, PRIV_NET_IFCREATE);
 2491                 if (error)
 2492                         return (error);
 2493                 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
 2494                         cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
 2495         case SIOCIFDESTROY:
 2496                 error = priv_check(td, PRIV_NET_IFDESTROY);
 2497                 if (error)
 2498                         return (error);
 2499                 return if_clone_destroy(ifr->ifr_name);
 2500 
 2501         case SIOCIFGCLONERS:
 2502                 return (if_clone_list((struct if_clonereq *)data));
 2503         case SIOCGIFGMEMB:
 2504                 return (if_getgroupmembers((struct ifgroupreq *)data));
 2505         }
 2506 
 2507         ifp = ifunit_ref(ifr->ifr_name);
 2508         if (ifp == NULL)
 2509                 return (ENXIO);
 2510 
 2511         error = ifhwioctl(cmd, ifp, data, td);
 2512         if (error != ENOIOCTL) {
 2513                 if_rele(ifp);
 2514                 return (error);
 2515         }
 2516 
 2517         oif_flags = ifp->if_flags;
 2518         if (so->so_proto == NULL) {
 2519                 if_rele(ifp);
 2520                 return (EOPNOTSUPP);
 2521         }
 2522 #ifndef COMPAT_43
 2523         error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
 2524                                                                  data,
 2525                                                                  ifp, td));
 2526         if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
 2527                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2528 #else
 2529         {
 2530                 u_long ocmd = cmd;
 2531 
 2532                 switch (cmd) {
 2533 
 2534                 case SIOCSIFDSTADDR:
 2535                 case SIOCSIFADDR:
 2536                 case SIOCSIFBRDADDR:
 2537                 case SIOCSIFNETMASK:
 2538 #if BYTE_ORDER != BIG_ENDIAN
 2539                         if (ifr->ifr_addr.sa_family == 0 &&
 2540                             ifr->ifr_addr.sa_len < 16) {
 2541                                 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
 2542                                 ifr->ifr_addr.sa_len = 16;
 2543                         }
 2544 #else
 2545                         if (ifr->ifr_addr.sa_len == 0)
 2546                                 ifr->ifr_addr.sa_len = 16;
 2547 #endif
 2548                         break;
 2549 
 2550                 case OSIOCGIFADDR:
 2551                         cmd = SIOCGIFADDR;
 2552                         break;
 2553 
 2554                 case OSIOCGIFDSTADDR:
 2555                         cmd = SIOCGIFDSTADDR;
 2556                         break;
 2557 
 2558                 case OSIOCGIFBRDADDR:
 2559                         cmd = SIOCGIFBRDADDR;
 2560                         break;
 2561 
 2562                 case OSIOCGIFNETMASK:
 2563                         cmd = SIOCGIFNETMASK;
 2564                 }
 2565                 error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
 2566                                                                    cmd,
 2567                                                                    data,
 2568                                                                    ifp, td));
 2569                 if (error == EOPNOTSUPP && ifp != NULL &&
 2570                     ifp->if_ioctl != NULL)
 2571                         error = (*ifp->if_ioctl)(ifp, cmd, data);
 2572                 switch (ocmd) {
 2573 
 2574                 case OSIOCGIFADDR:
 2575                 case OSIOCGIFDSTADDR:
 2576                 case OSIOCGIFBRDADDR:
 2577                 case OSIOCGIFNETMASK:
 2578                         *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
 2579 
 2580                 }
 2581         }
 2582 #endif /* COMPAT_43 */
 2583 
 2584         if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
 2585 #ifdef INET6
 2586                 if (ifp->if_flags & IFF_UP) {
 2587                         int s = splimp();
 2588                         in6_if_up(ifp);
 2589                         splx(s);
 2590                 }
 2591 #endif
 2592         }
 2593         if_rele(ifp);
 2594         return (error);
 2595 }
 2596 
 2597 /*
 2598  * The code common to handling reference counted flags,
 2599  * e.g., in ifpromisc() and if_allmulti().
 2600  * The "pflag" argument can specify a permanent mode flag to check,
 2601  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
 2602  *
 2603  * Only to be used on stack-owned flags, not driver-owned flags.
 2604  */
 2605 static int
 2606 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
 2607 {
 2608         struct ifreq ifr;
 2609         int error;
 2610         int oldflags, oldcount;
 2611 
 2612         /* Sanity checks to catch programming errors */
 2613         KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
 2614             ("%s: setting driver-owned flag %d", __func__, flag));
 2615 
 2616         if (onswitch)
 2617                 KASSERT(*refcount >= 0,
 2618                     ("%s: increment negative refcount %d for flag %d",
 2619                     __func__, *refcount, flag));
 2620         else
 2621                 KASSERT(*refcount > 0,
 2622                     ("%s: decrement non-positive refcount %d for flag %d",
 2623                     __func__, *refcount, flag));
 2624 
 2625         /* In case this mode is permanent, just touch refcount */
 2626         if (ifp->if_flags & pflag) {
 2627                 *refcount += onswitch ? 1 : -1;
 2628                 return (0);
 2629         }
 2630 
 2631         /* Save ifnet parameters for if_ioctl() may fail */
 2632         oldcount = *refcount;
 2633         oldflags = ifp->if_flags;
 2634         
 2635         /*
 2636          * See if we aren't the only and touching refcount is enough.
 2637          * Actually toggle interface flag if we are the first or last.
 2638          */
 2639         if (onswitch) {
 2640                 if ((*refcount)++)
 2641                         return (0);
 2642                 ifp->if_flags |= flag;
 2643         } else {
 2644                 if (--(*refcount))
 2645                         return (0);
 2646                 ifp->if_flags &= ~flag;
 2647         }
 2648 
 2649         /* Call down the driver since we've changed interface flags */
 2650         if (ifp->if_ioctl == NULL) {
 2651                 error = EOPNOTSUPP;
 2652                 goto recover;
 2653         }
 2654         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2655         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2656         error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2657         if (error)
 2658                 goto recover;
 2659         /* Notify userland that interface flags have changed */
 2660         rt_ifmsg(ifp);
 2661         return (0);
 2662 
 2663 recover:
 2664         /* Recover after driver error */
 2665         *refcount = oldcount;
 2666         ifp->if_flags = oldflags;
 2667         return (error);
 2668 }
 2669 
 2670 /*
 2671  * Set/clear promiscuous mode on interface ifp based on the truth value
 2672  * of pswitch.  The calls are reference counted so that only the first
 2673  * "on" request actually has an effect, as does the final "off" request.
 2674  * Results are undefined if the "off" and "on" requests are not matched.
 2675  */
 2676 int
 2677 ifpromisc(struct ifnet *ifp, int pswitch)
 2678 {
 2679         int error;
 2680         int oldflags = ifp->if_flags;
 2681 
 2682         error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
 2683                            &ifp->if_pcount, pswitch);
 2684         /* If promiscuous mode status has changed, log a message */
 2685         if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
 2686                 log(LOG_INFO, "%s: promiscuous mode %s\n",
 2687                     ifp->if_xname,
 2688                     (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
 2689         return (error);
 2690 }
 2691 
 2692 /*
 2693  * Return interface configuration
 2694  * of system.  List may be used
 2695  * in later ioctl's (above) to get
 2696  * other information.
 2697  */
 2698 /*ARGSUSED*/
 2699 static int
 2700 ifconf(u_long cmd, caddr_t data)
 2701 {
 2702         struct ifconf *ifc = (struct ifconf *)data;
 2703 #ifdef __amd64__
 2704         struct ifconf32 *ifc32 = (struct ifconf32 *)data;
 2705         struct ifconf ifc_swab;
 2706 #endif
 2707         struct ifnet *ifp;
 2708         struct ifaddr *ifa;
 2709         struct ifreq ifr;
 2710         struct sbuf *sb;
 2711         int error, full = 0, valid_len, max_len;
 2712 
 2713 #ifdef __amd64__
 2714         if (cmd == SIOCGIFCONF32) {
 2715                 ifc_swab.ifc_len = ifc32->ifc_len;
 2716                 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
 2717                 ifc = &ifc_swab;
 2718         }
 2719 #endif
 2720         /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
 2721         max_len = MAXPHYS - 1;
 2722 
 2723         /* Prevent hostile input from being able to crash the system */
 2724         if (ifc->ifc_len <= 0)
 2725                 return (EINVAL);
 2726 
 2727 again:
 2728         if (ifc->ifc_len <= max_len) {
 2729                 max_len = ifc->ifc_len;
 2730                 full = 1;
 2731         }
 2732         sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
 2733         max_len = 0;
 2734         valid_len = 0;
 2735 
 2736         IFNET_RLOCK();
 2737         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2738                 int addrs;
 2739 
 2740                 /*
 2741                  * Zero the ifr_name buffer to make sure we don't
 2742                  * disclose the contents of the stack.
 2743                  */
 2744                 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
 2745 
 2746                 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
 2747                     >= sizeof(ifr.ifr_name)) {
 2748                         sbuf_delete(sb);
 2749                         IFNET_RUNLOCK();
 2750                         return (ENAMETOOLONG);
 2751                 }
 2752 
 2753                 addrs = 0;
 2754                 IF_ADDR_LOCK(ifp);
 2755                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2756                         struct sockaddr *sa = ifa->ifa_addr;
 2757 
 2758                         if (prison_if(curthread->td_ucred, sa) != 0)
 2759                                 continue;
 2760                         addrs++;
 2761 #ifdef COMPAT_43
 2762                         if (cmd == OSIOCGIFCONF) {
 2763                                 struct osockaddr *osa =
 2764                                          (struct osockaddr *)&ifr.ifr_addr;
 2765                                 ifr.ifr_addr = *sa;
 2766                                 osa->sa_family = sa->sa_family;
 2767                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2768                                 max_len += sizeof(ifr);
 2769                         } else
 2770 #endif
 2771                         if (sa->sa_len <= sizeof(*sa)) {
 2772                                 ifr.ifr_addr = *sa;
 2773                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2774                                 max_len += sizeof(ifr);
 2775                         } else {
 2776                                 sbuf_bcat(sb, &ifr,
 2777                                     offsetof(struct ifreq, ifr_addr));
 2778                                 max_len += offsetof(struct ifreq, ifr_addr);
 2779                                 sbuf_bcat(sb, sa, sa->sa_len);
 2780                                 max_len += sa->sa_len;
 2781                         }
 2782 
 2783                         if (!sbuf_overflowed(sb))
 2784                                 valid_len = sbuf_len(sb);
 2785                 }
 2786                 IF_ADDR_UNLOCK(ifp);
 2787                 if (addrs == 0) {
 2788                         bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
 2789                         sbuf_bcat(sb, &ifr, sizeof(ifr));
 2790                         max_len += sizeof(ifr);
 2791 
 2792                         if (!sbuf_overflowed(sb))
 2793                                 valid_len = sbuf_len(sb);
 2794                 }
 2795         }
 2796         IFNET_RUNLOCK();
 2797 
 2798         /*
 2799          * If we didn't allocate enough space (uncommon), try again.  If
 2800          * we have already allocated as much space as we are allowed,
 2801          * return what we've got.
 2802          */
 2803         if (valid_len != max_len && !full) {
 2804                 sbuf_delete(sb);
 2805                 goto again;
 2806         }
 2807 
 2808         ifc->ifc_len = valid_len;
 2809 #ifdef __amd64__
 2810         if (cmd == SIOCGIFCONF32)
 2811                 ifc32->ifc_len = valid_len;
 2812 #endif
 2813         sbuf_finish(sb);
 2814         error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
 2815         sbuf_delete(sb);
 2816         return (error);
 2817 }
 2818 
 2819 /*
 2820  * Just like ifpromisc(), but for all-multicast-reception mode.
 2821  */
 2822 int
 2823 if_allmulti(struct ifnet *ifp, int onswitch)
 2824 {
 2825 
 2826         return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
 2827 }
 2828 
 2829 struct ifmultiaddr *
 2830 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
 2831 {
 2832         struct ifmultiaddr *ifma;
 2833 
 2834         IF_ADDR_LOCK_ASSERT(ifp);
 2835 
 2836         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
 2837                 if (sa->sa_family == AF_LINK) {
 2838                         if (sa_dl_equal(ifma->ifma_addr, sa))
 2839                                 break;
 2840                 } else {
 2841                         if (sa_equal(ifma->ifma_addr, sa))
 2842                                 break;
 2843                 }
 2844         }
 2845 
 2846         return ifma;
 2847 }
 2848 
 2849 /*
 2850  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
 2851  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
 2852  * the ifnet multicast address list here, so the caller must do that and
 2853  * other setup work (such as notifying the device driver).  The reference
 2854  * count is initialized to 1.
 2855  */
 2856 static struct ifmultiaddr *
 2857 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
 2858     int mflags)
 2859 {
 2860         struct ifmultiaddr *ifma;
 2861         struct sockaddr *dupsa;
 2862 
 2863         ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
 2864             M_ZERO);
 2865         if (ifma == NULL)
 2866                 return (NULL);
 2867 
 2868         dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
 2869         if (dupsa == NULL) {
 2870                 free(ifma, M_IFMADDR);
 2871                 return (NULL);
 2872         }
 2873         bcopy(sa, dupsa, sa->sa_len);
 2874         ifma->ifma_addr = dupsa;
 2875 
 2876         ifma->ifma_ifp = ifp;
 2877         ifma->ifma_refcount = 1;
 2878         ifma->ifma_protospec = NULL;
 2879 
 2880         if (llsa == NULL) {
 2881                 ifma->ifma_lladdr = NULL;
 2882                 return (ifma);
 2883         }
 2884 
 2885         dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
 2886         if (dupsa == NULL) {
 2887                 free(ifma->ifma_addr, M_IFMADDR);
 2888                 free(ifma, M_IFMADDR);
 2889                 return (NULL);
 2890         }
 2891         bcopy(llsa, dupsa, llsa->sa_len);
 2892         ifma->ifma_lladdr = dupsa;
 2893 
 2894         return (ifma);
 2895 }
 2896 
 2897 /*
 2898  * if_freemulti: free ifmultiaddr structure and possibly attached related
 2899  * addresses.  The caller is responsible for implementing reference
 2900  * counting, notifying the driver, handling routing messages, and releasing
 2901  * any dependent link layer state.
 2902  */
 2903 static void
 2904 if_freemulti(struct ifmultiaddr *ifma)
 2905 {
 2906 
 2907         KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
 2908             ifma->ifma_refcount));
 2909         KASSERT(ifma->ifma_protospec == NULL,
 2910             ("if_freemulti: protospec not NULL"));
 2911 
 2912         if (ifma->ifma_lladdr != NULL)
 2913                 free(ifma->ifma_lladdr, M_IFMADDR);
 2914         free(ifma->ifma_addr, M_IFMADDR);
 2915         free(ifma, M_IFMADDR);
 2916 }
 2917 
 2918 /*
 2919  * Register an additional multicast address with a network interface.
 2920  *
 2921  * - If the address is already present, bump the reference count on the
 2922  *   address and return.
 2923  * - If the address is not link-layer, look up a link layer address.
 2924  * - Allocate address structures for one or both addresses, and attach to the
 2925  *   multicast address list on the interface.  If automatically adding a link
 2926  *   layer address, the protocol address will own a reference to the link
 2927  *   layer address, to be freed when it is freed.
 2928  * - Notify the network device driver of an addition to the multicast address
 2929  *   list.
 2930  *
 2931  * 'sa' points to caller-owned memory with the desired multicast address.
 2932  *
 2933  * 'retifma' will be used to return a pointer to the resulting multicast
 2934  * address reference, if desired.
 2935  */
 2936 int
 2937 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
 2938     struct ifmultiaddr **retifma)
 2939 {
 2940         struct ifmultiaddr *ifma, *ll_ifma;
 2941         struct sockaddr *llsa;
 2942         int error;
 2943 
 2944         /*
 2945          * If the address is already present, return a new reference to it;
 2946          * otherwise, allocate storage and set up a new address.
 2947          */
 2948         IF_ADDR_LOCK(ifp);
 2949         ifma = if_findmulti(ifp, sa);
 2950         if (ifma != NULL) {
 2951                 ifma->ifma_refcount++;
 2952                 if (retifma != NULL)
 2953                         *retifma = ifma;
 2954                 IF_ADDR_UNLOCK(ifp);
 2955                 return (0);
 2956         }
 2957 
 2958         /*
 2959          * The address isn't already present; resolve the protocol address
 2960          * into a link layer address, and then look that up, bump its
 2961          * refcount or allocate an ifma for that also.  If 'llsa' was
 2962          * returned, we will need to free it later.
 2963          */
 2964         llsa = NULL;
 2965         ll_ifma = NULL;
 2966         if (ifp->if_resolvemulti != NULL) {
 2967                 error = ifp->if_resolvemulti(ifp, &llsa, sa);
 2968                 if (error)
 2969                         goto unlock_out;
 2970         }
 2971 
 2972         /*
 2973          * Allocate the new address.  Don't hook it up yet, as we may also
 2974          * need to allocate a link layer multicast address.
 2975          */
 2976         ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
 2977         if (ifma == NULL) {
 2978                 error = ENOMEM;
 2979                 goto free_llsa_out;
 2980         }
 2981 
 2982         /*
 2983          * If a link layer address is found, we'll need to see if it's
 2984          * already present in the address list, or allocate is as well.
 2985          * When this block finishes, the link layer address will be on the
 2986          * list.
 2987          */
 2988         if (llsa != NULL) {
 2989                 ll_ifma = if_findmulti(ifp, llsa);
 2990                 if (ll_ifma == NULL) {
 2991                         ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
 2992                         if (ll_ifma == NULL) {
 2993                                 --ifma->ifma_refcount;
 2994                                 if_freemulti(ifma);
 2995                                 error = ENOMEM;
 2996                                 goto free_llsa_out;
 2997                         }
 2998                         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
 2999                             ifma_link);
 3000                 } else
 3001                         ll_ifma->ifma_refcount++;
 3002                 ifma->ifma_llifma = ll_ifma;
 3003         }
 3004 
 3005         /*
 3006          * We now have a new multicast address, ifma, and possibly a new or
 3007          * referenced link layer address.  Add the primary address to the
 3008          * ifnet address list.
 3009          */
 3010         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
 3011 
 3012         if (retifma != NULL)
 3013                 *retifma = ifma;
 3014 
 3015         /*
 3016          * Must generate the message while holding the lock so that 'ifma'
 3017          * pointer is still valid.
 3018          */
 3019         rt_newmaddrmsg(RTM_NEWMADDR, ifma);
 3020         IF_ADDR_UNLOCK(ifp);
 3021 
 3022         /*
 3023          * We are certain we have added something, so call down to the
 3024          * interface to let them know about it.
 3025          */
 3026         if (ifp->if_ioctl != NULL) {
 3027                 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
 3028         }
 3029 
 3030         if (llsa != NULL)
 3031                 free(llsa, M_IFMADDR);
 3032 
 3033         return (0);
 3034 
 3035 free_llsa_out:
 3036         if (llsa != NULL)
 3037                 free(llsa, M_IFMADDR);
 3038 
 3039 unlock_out:
 3040         IF_ADDR_UNLOCK(ifp);
 3041         return (error);
 3042 }
 3043 
 3044 /*
 3045  * Delete a multicast group membership by network-layer group address.
 3046  *
 3047  * Returns ENOENT if the entry could not be found. If ifp no longer
 3048  * exists, results are undefined. This entry point should only be used
 3049  * from subsystems which do appropriate locking to hold ifp for the
 3050  * duration of the call.
 3051  * Network-layer protocol domains must use if_delmulti_ifma().
 3052  */
 3053 int
 3054 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
 3055 {
 3056         struct ifmultiaddr *ifma;
 3057         int lastref;
 3058 #ifdef INVARIANTS
 3059         struct ifnet *oifp;
 3060 
 3061         IFNET_RLOCK_NOSLEEP();
 3062         TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3063                 if (ifp == oifp)
 3064                         break;
 3065         if (ifp != oifp)
 3066                 ifp = NULL;
 3067         IFNET_RUNLOCK_NOSLEEP();
 3068 
 3069         KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
 3070 #endif
 3071         if (ifp == NULL)
 3072                 return (ENOENT);
 3073 
 3074         IF_ADDR_LOCK(ifp);
 3075         lastref = 0;
 3076         ifma = if_findmulti(ifp, sa);
 3077         if (ifma != NULL)
 3078                 lastref = if_delmulti_locked(ifp, ifma, 0);
 3079         IF_ADDR_UNLOCK(ifp);
 3080 
 3081         if (ifma == NULL)
 3082                 return (ENOENT);
 3083 
 3084         if (lastref && ifp->if_ioctl != NULL) {
 3085                 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3086         }
 3087 
 3088         return (0);
 3089 }
 3090 
 3091 /*
 3092  * Delete all multicast group membership for an interface.
 3093  * Should be used to quickly flush all multicast filters.
 3094  */
 3095 void
 3096 if_delallmulti(struct ifnet *ifp)
 3097 {
 3098         struct ifmultiaddr *ifma;
 3099         struct ifmultiaddr *next;
 3100 
 3101         IF_ADDR_LOCK(ifp);
 3102         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
 3103                 if_delmulti_locked(ifp, ifma, 0);
 3104         IF_ADDR_UNLOCK(ifp);
 3105 }
 3106 
 3107 /*
 3108  * Delete a multicast group membership by group membership pointer.
 3109  * Network-layer protocol domains must use this routine.
 3110  *
 3111  * It is safe to call this routine if the ifp disappeared.
 3112  */
 3113 void
 3114 if_delmulti_ifma(struct ifmultiaddr *ifma)
 3115 {
 3116         struct ifnet *ifp;
 3117         int lastref;
 3118 
 3119         ifp = ifma->ifma_ifp;
 3120 #ifdef DIAGNOSTIC
 3121         if (ifp == NULL) {
 3122                 printf("%s: ifma_ifp seems to be detached\n", __func__);
 3123         } else {
 3124                 struct ifnet *oifp;
 3125 
 3126                 IFNET_RLOCK_NOSLEEP();
 3127                 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3128                         if (ifp == oifp)
 3129                                 break;
 3130                 if (ifp != oifp) {
 3131                         printf("%s: ifnet %p disappeared\n", __func__, ifp);
 3132                         ifp = NULL;
 3133                 }
 3134                 IFNET_RUNLOCK_NOSLEEP();
 3135         }
 3136 #endif
 3137         /*
 3138          * If and only if the ifnet instance exists: Acquire the address lock.
 3139          */
 3140         if (ifp != NULL)
 3141                 IF_ADDR_LOCK(ifp);
 3142 
 3143         lastref = if_delmulti_locked(ifp, ifma, 0);
 3144 
 3145         if (ifp != NULL) {
 3146                 /*
 3147                  * If and only if the ifnet instance exists:
 3148                  *  Release the address lock.
 3149                  *  If the group was left: update the hardware hash filter.
 3150                  */
 3151                 IF_ADDR_UNLOCK(ifp);
 3152                 if (lastref && ifp->if_ioctl != NULL) {
 3153                         (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3154                 }
 3155         }
 3156 }
 3157 
 3158 /*
 3159  * Perform deletion of network-layer and/or link-layer multicast address.
 3160  *
 3161  * Return 0 if the reference count was decremented.
 3162  * Return 1 if the final reference was released, indicating that the
 3163  * hardware hash filter should be reprogrammed.
 3164  */
 3165 static int
 3166 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
 3167 {
 3168         struct ifmultiaddr *ll_ifma;
 3169 
 3170         if (ifp != NULL && ifma->ifma_ifp != NULL) {
 3171                 KASSERT(ifma->ifma_ifp == ifp,
 3172                     ("%s: inconsistent ifp %p", __func__, ifp));
 3173                 IF_ADDR_LOCK_ASSERT(ifp);
 3174         }
 3175 
 3176         ifp = ifma->ifma_ifp;
 3177 
 3178         /*
 3179          * If the ifnet is detaching, null out references to ifnet,
 3180          * so that upper protocol layers will notice, and not attempt
 3181          * to obtain locks for an ifnet which no longer exists. The
 3182          * routing socket announcement must happen before the ifnet
 3183          * instance is detached from the system.
 3184          */
 3185         if (detaching) {
 3186 #ifdef DIAGNOSTIC
 3187                 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
 3188 #endif
 3189                 /*
 3190                  * ifp may already be nulled out if we are being reentered
 3191                  * to delete the ll_ifma.
 3192                  */
 3193                 if (ifp != NULL) {
 3194                         rt_newmaddrmsg(RTM_DELMADDR, ifma);
 3195                         ifma->ifma_ifp = NULL;
 3196                 }
 3197         }
 3198 
 3199         if (--ifma->ifma_refcount > 0)
 3200                 return 0;
 3201 
 3202         /*
 3203          * If this ifma is a network-layer ifma, a link-layer ifma may
 3204          * have been associated with it. Release it first if so.
 3205          */
 3206         ll_ifma = ifma->ifma_llifma;
 3207         if (ll_ifma != NULL) {
 3208                 KASSERT(ifma->ifma_lladdr != NULL,
 3209                     ("%s: llifma w/o lladdr", __func__));
 3210                 if (detaching)
 3211                         ll_ifma->ifma_ifp = NULL;       /* XXX */
 3212                 if (--ll_ifma->ifma_refcount == 0) {
 3213                         if (ifp != NULL) {
 3214                                 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
 3215                                     ifma_link);
 3216                         }
 3217                         if_freemulti(ll_ifma);
 3218                 }
 3219         }
 3220 
 3221         if (ifp != NULL)
 3222                 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
 3223 
 3224         if_freemulti(ifma);
 3225 
 3226         /*
 3227          * The last reference to this instance of struct ifmultiaddr
 3228          * was released; the hardware should be notified of this change.
 3229          */
 3230         return 1;
 3231 }
 3232 
 3233 /*
 3234  * Set the link layer address on an interface.
 3235  *
 3236  * At this time we only support certain types of interfaces,
 3237  * and we don't allow the length of the address to change.
 3238  */
 3239 int
 3240 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
 3241 {
 3242         struct sockaddr_dl *sdl;
 3243         struct ifaddr *ifa;
 3244         struct ifreq ifr;
 3245 
 3246         IF_ADDR_LOCK(ifp);
 3247         ifa = ifp->if_addr;
 3248         if (ifa == NULL) {
 3249                 IF_ADDR_UNLOCK(ifp);
 3250                 return (EINVAL);
 3251         }
 3252         ifa_ref(ifa);
 3253         IF_ADDR_UNLOCK(ifp);
 3254         sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 3255         if (sdl == NULL) {
 3256                 ifa_free(ifa);
 3257                 return (EINVAL);
 3258         }
 3259         if (len != sdl->sdl_alen) {     /* don't allow length to change */
 3260                 ifa_free(ifa);
 3261                 return (EINVAL);
 3262         }
 3263         switch (ifp->if_type) {
 3264         case IFT_ETHER:
 3265         case IFT_FDDI:
 3266         case IFT_XETHER:
 3267         case IFT_ISO88025:
 3268         case IFT_L2VLAN:
 3269         case IFT_BRIDGE:
 3270         case IFT_ARCNET:
 3271         case IFT_IEEE8023ADLAG:
 3272         case IFT_IEEE80211:
 3273                 bcopy(lladdr, LLADDR(sdl), len);
 3274                 ifa_free(ifa);
 3275                 break;
 3276         default:
 3277                 ifa_free(ifa);
 3278                 return (ENODEV);
 3279         }
 3280 
 3281         /*
 3282          * If the interface is already up, we need
 3283          * to re-init it in order to reprogram its
 3284          * address filter.
 3285          */
 3286         if ((ifp->if_flags & IFF_UP) != 0) {
 3287                 if (ifp->if_ioctl) {
 3288                         ifp->if_flags &= ~IFF_UP;
 3289                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3290                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3291                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3292                         ifp->if_flags |= IFF_UP;
 3293                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3294                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3295                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3296                 }
 3297 #ifdef INET
 3298                 /*
 3299                  * Also send gratuitous ARPs to notify other nodes about
 3300                  * the address change.
 3301                  */
 3302                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 3303                         if (ifa->ifa_addr->sa_family == AF_INET)
 3304                                 arp_ifinit(ifp, ifa);
 3305                 }
 3306 #endif
 3307         }
 3308         return (0);
 3309 }
 3310 
 3311 /*
 3312  * The name argument must be a pointer to storage which will last as
 3313  * long as the interface does.  For physical devices, the result of
 3314  * device_get_name(dev) is a good choice and for pseudo-devices a
 3315  * static string works well.
 3316  */
 3317 void
 3318 if_initname(struct ifnet *ifp, const char *name, int unit)
 3319 {
 3320         ifp->if_dname = name;
 3321         ifp->if_dunit = unit;
 3322         if (unit != IF_DUNIT_NONE)
 3323                 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
 3324         else
 3325                 strlcpy(ifp->if_xname, name, IFNAMSIZ);
 3326 }
 3327 
 3328 int
 3329 if_printf(struct ifnet *ifp, const char * fmt, ...)
 3330 {
 3331         va_list ap;
 3332         int retval;
 3333 
 3334         retval = printf("%s: ", ifp->if_xname);
 3335         va_start(ap, fmt);
 3336         retval += vprintf(fmt, ap);
 3337         va_end(ap);
 3338         return (retval);
 3339 }
 3340 
 3341 void
 3342 if_start(struct ifnet *ifp)
 3343 {
 3344 
 3345         (*(ifp)->if_start)(ifp);
 3346 }
 3347 
 3348 /*
 3349  * Backwards compatibility interface for drivers 
 3350  * that have not implemented it
 3351  */
 3352 static int
 3353 if_transmit(struct ifnet *ifp, struct mbuf *m)
 3354 {
 3355         int error;
 3356 
 3357         IFQ_HANDOFF(ifp, m, error);
 3358         return (error);
 3359 }
 3360 
 3361 int
 3362 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
 3363 {
 3364         int active = 0;
 3365 
 3366         IF_LOCK(ifq);
 3367         if (_IF_QFULL(ifq)) {
 3368                 _IF_DROP(ifq);
 3369                 IF_UNLOCK(ifq);
 3370                 m_freem(m);
 3371                 return (0);
 3372         }
 3373         if (ifp != NULL) {
 3374                 ifp->if_obytes += m->m_pkthdr.len + adjust;
 3375                 if (m->m_flags & (M_BCAST|M_MCAST))
 3376                         ifp->if_omcasts++;
 3377                 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
 3378         }
 3379         _IF_ENQUEUE(ifq, m);
 3380         IF_UNLOCK(ifq);
 3381         if (ifp != NULL && !active)
 3382                 (*(ifp)->if_start)(ifp);
 3383         return (1);
 3384 }
 3385 
 3386 void
 3387 if_register_com_alloc(u_char type,
 3388     if_com_alloc_t *a, if_com_free_t *f)
 3389 {
 3390         
 3391         KASSERT(if_com_alloc[type] == NULL,
 3392             ("if_register_com_alloc: %d already registered", type));
 3393         KASSERT(if_com_free[type] == NULL,
 3394             ("if_register_com_alloc: %d free already registered", type));
 3395 
 3396         if_com_alloc[type] = a;
 3397         if_com_free[type] = f;
 3398 }
 3399 
 3400 void
 3401 if_deregister_com_alloc(u_char type)
 3402 {
 3403         
 3404         KASSERT(if_com_alloc[type] != NULL,
 3405             ("if_deregister_com_alloc: %d not registered", type));
 3406         KASSERT(if_com_free[type] != NULL,
 3407             ("if_deregister_com_alloc: %d free not registered", type));
 3408         if_com_alloc[type] = NULL;
 3409         if_com_free[type] = NULL;
 3410 }
 3411 
 3412 #ifdef DDB
 3413 static void
 3414 if_show_ifnet(struct ifnet *ifp)
 3415 {
 3416 
 3417         if (ifp == NULL)
 3418                 return;
 3419         db_printf("%s:\n", ifp->if_xname);
 3420 #define IF_DB_PRINTF(f, e)      db_printf("   %s = " f "\n", #e, ifp->e);
 3421         IF_DB_PRINTF("%s", if_dname);
 3422         IF_DB_PRINTF("%d", if_dunit);
 3423         IF_DB_PRINTF("%s", if_description);
 3424         IF_DB_PRINTF("%u", if_index);
 3425         IF_DB_PRINTF("%u", if_refcount);
 3426         IF_DB_PRINTF("%p", if_softc);
 3427         IF_DB_PRINTF("%p", if_l2com);
 3428         IF_DB_PRINTF("%p", if_vnet);
 3429         IF_DB_PRINTF("%p", if_home_vnet);
 3430         IF_DB_PRINTF("%p", if_addr);
 3431         IF_DB_PRINTF("%p", if_llsoftc);
 3432         IF_DB_PRINTF("%p", if_label);
 3433         IF_DB_PRINTF("%u", if_pcount);
 3434         IF_DB_PRINTF("0x%08x", if_flags);
 3435         IF_DB_PRINTF("0x%08x", if_drv_flags);
 3436         IF_DB_PRINTF("0x%08x", if_capabilities);
 3437         IF_DB_PRINTF("0x%08x", if_capenable);
 3438         IF_DB_PRINTF("%p", if_snd.ifq_head);
 3439         IF_DB_PRINTF("%p", if_snd.ifq_tail);
 3440         IF_DB_PRINTF("%d", if_snd.ifq_len);
 3441         IF_DB_PRINTF("%d", if_snd.ifq_maxlen);
 3442         IF_DB_PRINTF("%d", if_snd.ifq_drops);
 3443         IF_DB_PRINTF("%p", if_snd.ifq_drv_head);
 3444         IF_DB_PRINTF("%p", if_snd.ifq_drv_tail);
 3445         IF_DB_PRINTF("%d", if_snd.ifq_drv_len);
 3446         IF_DB_PRINTF("%d", if_snd.ifq_drv_maxlen);
 3447         IF_DB_PRINTF("%d", if_snd.altq_type);
 3448         IF_DB_PRINTF("%x", if_snd.altq_flags);
 3449 #undef IF_DB_PRINTF
 3450 }
 3451 
 3452 DB_SHOW_COMMAND(ifnet, db_show_ifnet)
 3453 {
 3454 
 3455         if (!have_addr) {
 3456                 db_printf("usage: show ifnet <struct ifnet *>\n");
 3457                 return;
 3458         }
 3459 
 3460         if_show_ifnet((struct ifnet *)addr);
 3461 }
 3462 
 3463 DB_SHOW_ALL_COMMAND(ifnets, db_show_all_ifnets)
 3464 {
 3465         VNET_ITERATOR_DECL(vnet_iter);
 3466         struct ifnet *ifp;
 3467         u_short idx;
 3468 
 3469         VNET_FOREACH(vnet_iter) {
 3470                 CURVNET_SET_QUIET(vnet_iter);
 3471 #ifdef VIMAGE
 3472                 db_printf("vnet=%p\n", curvnet);
 3473 #endif
 3474                 for (idx = 1; idx <= V_if_index; idx++) {
 3475                         ifp = V_ifindex_table[idx].ife_ifnet;
 3476                         if (ifp == NULL)
 3477                                 continue;
 3478                         db_printf( "%20s ifp=%p\n", ifp->if_xname, ifp);
 3479                         if (db_pager_quit)
 3480                                 break;
 3481                 }
 3482                 CURVNET_RESTORE();
 3483         }
 3484 }
 3485 #endif

Cache object: 81c9a0fefdda8869de3c5ecf0c7ac927


[ 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.