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  * $FreeBSD$
   31  */
   32 
   33 #include "opt_compat.h"
   34 #include "opt_inet6.h"
   35 #include "opt_inet.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/types.h>
   39 #include <sys/conf.h>
   40 #include <sys/malloc.h>
   41 #include <sys/sbuf.h>
   42 #include <sys/bus.h>
   43 #include <sys/mbuf.h>
   44 #include <sys/systm.h>
   45 #include <sys/priv.h>
   46 #include <sys/proc.h>
   47 #include <sys/socket.h>
   48 #include <sys/socketvar.h>
   49 #include <sys/protosw.h>
   50 #include <sys/kernel.h>
   51 #include <sys/lock.h>
   52 #include <sys/refcount.h>
   53 #include <sys/module.h>
   54 #include <sys/rwlock.h>
   55 #include <sys/sockio.h>
   56 #include <sys/syslog.h>
   57 #include <sys/sysctl.h>
   58 #include <sys/taskqueue.h>
   59 #include <sys/domain.h>
   60 #include <sys/jail.h>
   61 #include <sys/priv.h>
   62 
   63 #include <machine/stdarg.h>
   64 #include <vm/uma.h>
   65 
   66 #include <net/if.h>
   67 #include <net/if_arp.h>
   68 #include <net/if_clone.h>
   69 #include <net/if_dl.h>
   70 #include <net/if_types.h>
   71 #include <net/if_var.h>
   72 #include <net/radix.h>
   73 #include <net/route.h>
   74 #include <net/vnet.h>
   75 
   76 #if defined(INET) || defined(INET6)
   77 #include <net/ethernet.h>
   78 #include <netinet/in.h>
   79 #include <netinet/in_var.h>
   80 #include <netinet/ip.h>
   81 #include <netinet/ip_carp.h>
   82 #ifdef INET
   83 #include <netinet/if_ether.h>
   84 #endif /* INET */
   85 #ifdef INET6
   86 #include <netinet6/in6_var.h>
   87 #include <netinet6/in6_ifattach.h>
   88 #endif /* INET6 */
   89 #endif /* INET || INET6 */
   90 
   91 #include <security/mac/mac_framework.h>
   92 
   93 #ifdef COMPAT_FREEBSD32
   94 #include <sys/mount.h>
   95 #include <compat/freebsd32/freebsd32.h>
   96 #endif
   97 
   98 struct ifindex_entry {
   99         struct  ifnet *ife_ifnet;
  100 };
  101 
  102 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
  103 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
  104 
  105 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
  106 SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
  107     &ifqmaxlen, 0, "max send queue size");
  108 
  109 /* Log link state change events */
  110 static int log_link_state_change = 1;
  111 
  112 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
  113         &log_link_state_change, 0,
  114         "log interface link state change events");
  115 
  116 /* Interface description */
  117 static unsigned int ifdescr_maxlen = 1024;
  118 SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
  119         &ifdescr_maxlen, 0,
  120         "administrative maximum length for interface description");
  121 
  122 static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
  123 
  124 /* global sx for non-critical path ifdescr */
  125 static struct sx ifdescr_sx;
  126 SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
  127 
  128 void    (*bridge_linkstate_p)(struct ifnet *ifp);
  129 void    (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
  130 void    (*lagg_linkstate_p)(struct ifnet *ifp, int state);
  131 /* These are external hooks for CARP. */
  132 void    (*carp_linkstate_p)(struct ifnet *ifp);
  133 #if defined(INET) || defined(INET6)
  134 struct ifnet *(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
  135 int     (*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
  136     struct sockaddr *sa, struct rtentry *rt);
  137 #endif
  138 #ifdef INET
  139 int (*carp_iamatch_p)(struct ifnet *, struct in_ifaddr *, struct in_addr *,
  140     u_int8_t **);
  141 #endif
  142 #ifdef INET6
  143 struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
  144 caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
  145     const struct in6_addr *taddr);
  146 #endif
  147 
  148 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
  149 
  150 /*
  151  * XXX: Style; these should be sorted alphabetically, and unprototyped
  152  * static functions should be prototyped. Currently they are sorted by
  153  * declaration order.
  154  */
  155 static void     if_attachdomain(void *);
  156 static void     if_attachdomain1(struct ifnet *);
  157 static int      ifconf(u_long, caddr_t);
  158 static void     if_freemulti(struct ifmultiaddr *);
  159 static void     if_init(void *);
  160 static void     if_grow(void);
  161 static void     if_input_default(struct ifnet *, struct mbuf *);
  162 static void     if_route(struct ifnet *, int flag, int fam);
  163 static int      if_setflag(struct ifnet *, int, int, int *, int);
  164 static int      if_transmit(struct ifnet *ifp, struct mbuf *m);
  165 static void     if_unroute(struct ifnet *, int flag, int fam);
  166 static void     link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  167 static int      if_rtdel(struct radix_node *, void *);
  168 static int      ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
  169 static int      if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
  170 static void     do_link_state_change(void *, int);
  171 static int      if_getgroup(struct ifgroupreq *, struct ifnet *);
  172 static int      if_getgroupmembers(struct ifgroupreq *);
  173 static void     if_delgroups(struct ifnet *);
  174 static void     if_attach_internal(struct ifnet *, int);
  175 static void     if_detach_internal(struct ifnet *, int);
  176 
  177 #ifdef INET6
  178 /*
  179  * XXX: declare here to avoid to include many inet6 related files..
  180  * should be more generalized?
  181  */
  182 extern void     nd6_setmtu(struct ifnet *);
  183 #endif
  184 
  185 VNET_DEFINE(int, if_index);
  186 int     ifqmaxlen = IFQ_MAXLEN;
  187 VNET_DEFINE(struct ifnethead, ifnet);   /* depend on static init XXX */
  188 VNET_DEFINE(struct ifgrouphead, ifg_head);
  189 
  190 static VNET_DEFINE(int, if_indexlim) = 8;
  191 
  192 /* Table of ifnet by index. */
  193 VNET_DEFINE(struct ifindex_entry *, ifindex_table);
  194 
  195 #define V_if_indexlim           VNET(if_indexlim)
  196 #define V_ifindex_table         VNET(ifindex_table)
  197 
  198 /*
  199  * The global network interface list (V_ifnet) and related state (such as
  200  * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
  201  * an rwlock.  Either may be acquired shared to stablize the list, but both
  202  * must be acquired writable to modify the list.  This model allows us to
  203  * both stablize the interface list during interrupt thread processing, but
  204  * also to stablize it over long-running ioctls, without introducing priority
  205  * inversions and deadlocks.
  206  */
  207 struct rwlock ifnet_rwlock;
  208 struct sx ifnet_sxlock;
  209 
  210 /*
  211  * The allocation of network interfaces is a rather non-atomic affair; we
  212  * need to select an index before we are ready to expose the interface for
  213  * use, so will use this pointer value to indicate reservation.
  214  */
  215 #define IFNET_HOLD      (void *)(uintptr_t)(-1)
  216 
  217 static  if_com_alloc_t *if_com_alloc[256];
  218 static  if_com_free_t *if_com_free[256];
  219 
  220 static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
  221 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
  222 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
  223 
  224 struct ifnet *
  225 ifnet_byindex_locked(u_short idx)
  226 {
  227 
  228         if (idx > V_if_index)
  229                 return (NULL);
  230         if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
  231                 return (NULL);
  232         return (V_ifindex_table[idx].ife_ifnet);
  233 }
  234 
  235 struct ifnet *
  236 ifnet_byindex(u_short idx)
  237 {
  238         struct ifnet *ifp;
  239 
  240         IFNET_RLOCK_NOSLEEP();
  241         ifp = ifnet_byindex_locked(idx);
  242         IFNET_RUNLOCK_NOSLEEP();
  243         return (ifp);
  244 }
  245 
  246 struct ifnet *
  247 ifnet_byindex_ref(u_short idx)
  248 {
  249         struct ifnet *ifp;
  250 
  251         IFNET_RLOCK_NOSLEEP();
  252         ifp = ifnet_byindex_locked(idx);
  253         if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
  254                 IFNET_RUNLOCK_NOSLEEP();
  255                 return (NULL);
  256         }
  257         if_ref(ifp);
  258         IFNET_RUNLOCK_NOSLEEP();
  259         return (ifp);
  260 }
  261 
  262 /*
  263  * Allocate an ifindex array entry; return 0 on success or an error on
  264  * failure.
  265  */
  266 static int
  267 ifindex_alloc_locked(u_short *idxp)
  268 {
  269         u_short idx;
  270 
  271         IFNET_WLOCK_ASSERT();
  272 
  273 retry:
  274         /*
  275          * Try to find an empty slot below V_if_index.  If we fail, take the
  276          * next slot.
  277          */
  278         for (idx = 1; idx <= V_if_index; idx++) {
  279                 if (V_ifindex_table[idx].ife_ifnet == NULL)
  280                         break;
  281         }
  282 
  283         /* Catch if_index overflow. */
  284         if (idx < 1)
  285                 return (ENOSPC);
  286         if (idx >= V_if_indexlim) {
  287                 if_grow();
  288                 goto retry;
  289         }
  290         if (idx > V_if_index)
  291                 V_if_index = idx;
  292         *idxp = idx;
  293         return (0);
  294 }
  295 
  296 static void
  297 ifindex_free_locked(u_short idx)
  298 {
  299 
  300         IFNET_WLOCK_ASSERT();
  301 
  302         V_ifindex_table[idx].ife_ifnet = NULL;
  303         while (V_if_index > 0 &&
  304             V_ifindex_table[V_if_index].ife_ifnet == NULL)
  305                 V_if_index--;
  306 }
  307 
  308 static void
  309 ifindex_free(u_short idx)
  310 {
  311 
  312         IFNET_WLOCK();
  313         ifindex_free_locked(idx);
  314         IFNET_WUNLOCK();
  315 }
  316 
  317 static void
  318 ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
  319 {
  320 
  321         IFNET_WLOCK_ASSERT();
  322 
  323         V_ifindex_table[idx].ife_ifnet = ifp;
  324 }
  325 
  326 static void
  327 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
  328 {
  329 
  330         IFNET_WLOCK();
  331         ifnet_setbyindex_locked(idx, ifp);
  332         IFNET_WUNLOCK();
  333 }
  334 
  335 struct ifaddr *
  336 ifaddr_byindex(u_short idx)
  337 {
  338         struct ifaddr *ifa;
  339 
  340         IFNET_RLOCK_NOSLEEP();
  341         ifa = ifnet_byindex_locked(idx)->if_addr;
  342         if (ifa != NULL)
  343                 ifa_ref(ifa);
  344         IFNET_RUNLOCK_NOSLEEP();
  345         return (ifa);
  346 }
  347 
  348 /*
  349  * Network interface utility routines.
  350  *
  351  * Routines with ifa_ifwith* names take sockaddr *'s as
  352  * parameters.
  353  */
  354 
  355 static void
  356 vnet_if_init(const void *unused __unused)
  357 {
  358 
  359         TAILQ_INIT(&V_ifnet);
  360         TAILQ_INIT(&V_ifg_head);
  361         IFNET_WLOCK();
  362         if_grow();                              /* create initial table */
  363         IFNET_WUNLOCK();
  364         vnet_if_clone_init();
  365 }
  366 VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
  367     NULL);
  368 
  369 /* ARGSUSED*/
  370 static void
  371 if_init(void *dummy __unused)
  372 {
  373 
  374         IFNET_LOCK_INIT();
  375         if_clone_init();
  376 }
  377 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
  378 
  379 
  380 #ifdef VIMAGE
  381 static void
  382 vnet_if_uninit(const void *unused __unused)
  383 {
  384 
  385         VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
  386             "not empty", __func__, __LINE__, &V_ifnet));
  387         VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
  388             "not empty", __func__, __LINE__, &V_ifg_head));
  389 
  390         free((caddr_t)V_ifindex_table, M_IFNET);
  391 }
  392 VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
  393     vnet_if_uninit, NULL);
  394 #endif
  395 
  396 static void
  397 if_grow(void)
  398 {
  399         int oldlim;
  400         u_int n;
  401         struct ifindex_entry *e;
  402 
  403         IFNET_WLOCK_ASSERT();
  404         oldlim = V_if_indexlim;
  405         IFNET_WUNLOCK();
  406         n = (oldlim << 1) * sizeof(*e);
  407         e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
  408         IFNET_WLOCK();
  409         if (V_if_indexlim != oldlim) {
  410                 free(e, M_IFNET);
  411                 return;
  412         }
  413         if (V_ifindex_table != NULL) {
  414                 memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
  415                 free((caddr_t)V_ifindex_table, M_IFNET);
  416         }
  417         V_if_indexlim <<= 1;
  418         V_ifindex_table = e;
  419 }
  420 
  421 /*
  422  * Allocate a struct ifnet and an index for an interface.  A layer 2
  423  * common structure will also be allocated if an allocation routine is
  424  * registered for the passed type.
  425  */
  426 struct ifnet *
  427 if_alloc(u_char type)
  428 {
  429         struct ifnet *ifp;
  430         u_short idx;
  431 
  432         ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
  433         IFNET_WLOCK();
  434         if (ifindex_alloc_locked(&idx) != 0) {
  435                 IFNET_WUNLOCK();
  436                 free(ifp, M_IFNET);
  437                 return (NULL);
  438         }
  439         ifnet_setbyindex_locked(idx, IFNET_HOLD);
  440         IFNET_WUNLOCK();
  441         ifp->if_index = idx;
  442         ifp->if_type = type;
  443         ifp->if_alloctype = type;
  444         if (if_com_alloc[type] != NULL) {
  445                 ifp->if_l2com = if_com_alloc[type](type, ifp);
  446                 if (ifp->if_l2com == NULL) {
  447                         free(ifp, M_IFNET);
  448                         ifindex_free(idx);
  449                         return (NULL);
  450                 }
  451         }
  452 
  453         IF_ADDR_LOCK_INIT(ifp);
  454         TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
  455         ifp->if_afdata_initialized = 0;
  456         IF_AFDATA_LOCK_INIT(ifp);
  457         TAILQ_INIT(&ifp->if_addrhead);
  458         TAILQ_INIT(&ifp->if_prefixhead);
  459         TAILQ_INIT(&ifp->if_multiaddrs);
  460         TAILQ_INIT(&ifp->if_groups);
  461 #ifdef MAC
  462         mac_ifnet_init(ifp);
  463 #endif
  464         ifq_init(&ifp->if_snd, ifp);
  465 
  466         refcount_init(&ifp->if_refcount, 1);    /* Index reference. */
  467         ifnet_setbyindex(ifp->if_index, ifp);
  468         return (ifp);
  469 }
  470 
  471 /*
  472  * Do the actual work of freeing a struct ifnet, and layer 2 common
  473  * structure.  This call is made when the last reference to an
  474  * interface is released.
  475  */
  476 static void
  477 if_free_internal(struct ifnet *ifp)
  478 {
  479 
  480         KASSERT((ifp->if_flags & IFF_DYING),
  481             ("if_free_internal: interface not dying"));
  482 
  483         if (if_com_free[ifp->if_alloctype] != NULL)
  484                 if_com_free[ifp->if_alloctype](ifp->if_l2com,
  485                     ifp->if_alloctype);
  486 
  487 #ifdef MAC
  488         mac_ifnet_destroy(ifp);
  489 #endif /* MAC */
  490         if (ifp->if_description != NULL)
  491                 free(ifp->if_description, M_IFDESCR);
  492         IF_AFDATA_DESTROY(ifp);
  493         IF_ADDR_LOCK_DESTROY(ifp);
  494         ifq_delete(&ifp->if_snd);
  495         free(ifp, M_IFNET);
  496 }
  497 
  498 /*
  499  * This version should only be called by intefaces that switch their type
  500  * after calling if_alloc().  if_free_type() will go away again now that we
  501  * have if_alloctype to cache the original allocation type.  For now, assert
  502  * that they match, since we require that in practice.
  503  */
  504 void
  505 if_free_type(struct ifnet *ifp, u_char type)
  506 {
  507 
  508         KASSERT(ifp->if_alloctype == type,
  509             ("if_free_type: type (%d) != alloctype (%d)", type,
  510             ifp->if_alloctype));
  511 
  512         ifp->if_flags |= IFF_DYING;                     /* XXX: Locking */
  513 
  514         CURVNET_SET_QUIET(ifp->if_vnet);
  515         IFNET_WLOCK();
  516         KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
  517             ("%s: freeing unallocated ifnet", ifp->if_xname));
  518 
  519         ifindex_free_locked(ifp->if_index);
  520         IFNET_WUNLOCK();
  521 
  522         if (refcount_release(&ifp->if_refcount))
  523                 if_free_internal(ifp);
  524         CURVNET_RESTORE();
  525 }
  526 
  527 /*
  528  * This is the normal version of if_free(), used by device drivers to free a
  529  * detached network interface.  The contents of if_free_type() will move into
  530  * here when if_free_type() goes away.
  531  */
  532 void
  533 if_free(struct ifnet *ifp)
  534 {
  535 
  536         if_free_type(ifp, ifp->if_alloctype);
  537 }
  538 
  539 /*
  540  * Interfaces to keep an ifnet type-stable despite the possibility of the
  541  * driver calling if_free().  If there are additional references, we defer
  542  * freeing the underlying data structure.
  543  */
  544 void
  545 if_ref(struct ifnet *ifp)
  546 {
  547 
  548         /* We don't assert the ifnet list lock here, but arguably should. */
  549         refcount_acquire(&ifp->if_refcount);
  550 }
  551 
  552 void
  553 if_rele(struct ifnet *ifp)
  554 {
  555 
  556         if (!refcount_release(&ifp->if_refcount))
  557                 return;
  558         if_free_internal(ifp);
  559 }
  560 
  561 void
  562 ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
  563 {
  564         
  565         mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
  566 
  567         if (ifq->ifq_maxlen == 0) 
  568                 ifq->ifq_maxlen = ifqmaxlen;
  569 
  570         ifq->altq_type = 0;
  571         ifq->altq_disc = NULL;
  572         ifq->altq_flags &= ALTQF_CANTCHANGE;
  573         ifq->altq_tbr  = NULL;
  574         ifq->altq_ifp  = ifp;
  575 }
  576 
  577 void
  578 ifq_delete(struct ifaltq *ifq)
  579 {
  580         mtx_destroy(&ifq->ifq_mtx);
  581 }
  582 
  583 /*
  584  * Perform generic interface initalization tasks and attach the interface
  585  * to the list of "active" interfaces.  If vmove flag is set on entry
  586  * to if_attach_internal(), perform only a limited subset of initialization
  587  * tasks, given that we are moving from one vnet to another an ifnet which
  588  * has already been fully initialized.
  589  *
  590  * XXX:
  591  *  - The decision to return void and thus require this function to
  592  *    succeed is questionable.
  593  *  - We should probably do more sanity checking.  For instance we don't
  594  *    do anything to insure if_xname is unique or non-empty.
  595  */
  596 void
  597 if_attach(struct ifnet *ifp)
  598 {
  599 
  600         if_attach_internal(ifp, 0);
  601 }
  602 
  603 /*
  604  * Compute the least common TSO limit.
  605  */
  606 void
  607 if_hw_tsomax_common(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
  608 {
  609         /*
  610          * 1) If there is no limit currently, take the limit from
  611          * the network adapter.
  612          *
  613          * 2) If the network adapter has a limit below the current
  614          * limit, apply it.
  615          */
  616         if (pmax->tsomaxbytes == 0 || (ifp->if_hw_tsomax != 0 &&
  617             ifp->if_hw_tsomax < pmax->tsomaxbytes)) {
  618                 pmax->tsomaxbytes = ifp->if_hw_tsomax;
  619         }
  620         if (pmax->tsomaxsegcount == 0 || (ifp->if_hw_tsomaxsegcount != 0 &&
  621             ifp->if_hw_tsomaxsegcount < pmax->tsomaxsegcount)) {
  622                 pmax->tsomaxsegcount = ifp->if_hw_tsomaxsegcount;
  623         }
  624         if (pmax->tsomaxsegsize == 0 || (ifp->if_hw_tsomaxsegsize != 0 &&
  625             ifp->if_hw_tsomaxsegsize < pmax->tsomaxsegsize)) {
  626                 pmax->tsomaxsegsize = ifp->if_hw_tsomaxsegsize;
  627         }
  628 }
  629 
  630 /*
  631  * Update TSO limit of a network adapter.
  632  *
  633  * Returns zero if no change. Else non-zero.
  634  */
  635 int
  636 if_hw_tsomax_update(struct ifnet *ifp, struct ifnet_hw_tsomax *pmax)
  637 {
  638         int retval = 0;
  639         if (ifp->if_hw_tsomax != pmax->tsomaxbytes) {
  640                 ifp->if_hw_tsomax = pmax->tsomaxbytes;
  641                 retval++;
  642         }
  643         if (ifp->if_hw_tsomaxsegsize != pmax->tsomaxsegsize) {
  644                 ifp->if_hw_tsomaxsegsize = pmax->tsomaxsegsize;
  645                 retval++;
  646         }
  647         if (ifp->if_hw_tsomaxsegcount != pmax->tsomaxsegcount) {
  648                 ifp->if_hw_tsomaxsegcount = pmax->tsomaxsegcount;
  649                 retval++;
  650         }
  651         return (retval);
  652 }
  653 
  654 static void
  655 if_attach_internal(struct ifnet *ifp, int vmove)
  656 {
  657         unsigned socksize, ifasize;
  658         int namelen, masklen;
  659         struct sockaddr_dl *sdl;
  660         struct ifaddr *ifa;
  661 
  662         if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
  663                 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
  664                     ifp->if_xname);
  665 
  666 #ifdef VIMAGE
  667         ifp->if_vnet = curvnet;
  668         if (ifp->if_home_vnet == NULL)
  669                 ifp->if_home_vnet = curvnet;
  670 #endif
  671 
  672         if_addgroup(ifp, IFG_ALL);
  673 
  674         getmicrotime(&ifp->if_lastchange);
  675         ifp->if_data.ifi_epoch = time_uptime;
  676         ifp->if_data.ifi_datalen = sizeof(struct if_data);
  677 
  678         KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
  679             (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
  680             ("transmit and qflush must both either be set or both be NULL"));
  681         if (ifp->if_transmit == NULL) {
  682                 ifp->if_transmit = if_transmit;
  683                 ifp->if_qflush = if_qflush;
  684         }
  685         if (ifp->if_input == NULL)
  686                 ifp->if_input = if_input_default;
  687 
  688         if (!vmove) {
  689 #ifdef MAC
  690                 mac_ifnet_create(ifp);
  691 #endif
  692 
  693                 /*
  694                  * Create a Link Level name for this device.
  695                  */
  696                 namelen = strlen(ifp->if_xname);
  697                 /*
  698                  * Always save enough space for any possiable name so we
  699                  * can do a rename in place later.
  700                  */
  701                 masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
  702                 socksize = masklen + ifp->if_addrlen;
  703                 if (socksize < sizeof(*sdl))
  704                         socksize = sizeof(*sdl);
  705                 socksize = roundup2(socksize, sizeof(long));
  706                 ifasize = sizeof(*ifa) + 2 * socksize;
  707                 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
  708                 ifa_init(ifa);
  709                 sdl = (struct sockaddr_dl *)(ifa + 1);
  710                 sdl->sdl_len = socksize;
  711                 sdl->sdl_family = AF_LINK;
  712                 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
  713                 sdl->sdl_nlen = namelen;
  714                 sdl->sdl_index = ifp->if_index;
  715                 sdl->sdl_type = ifp->if_type;
  716                 ifp->if_addr = ifa;
  717                 ifa->ifa_ifp = ifp;
  718                 ifa->ifa_rtrequest = link_rtrequest;
  719                 ifa->ifa_addr = (struct sockaddr *)sdl;
  720                 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
  721                 ifa->ifa_netmask = (struct sockaddr *)sdl;
  722                 sdl->sdl_len = masklen;
  723                 while (namelen != 0)
  724                         sdl->sdl_data[--namelen] = 0xff;
  725                 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
  726                 /* Reliably crash if used uninitialized. */
  727                 ifp->if_broadcastaddr = NULL;
  728 
  729 #if defined(INET) || defined(INET6)
  730                 /* Use defaults for TSO, if nothing is set */
  731                 if (ifp->if_hw_tsomax == 0 &&
  732                     ifp->if_hw_tsomaxsegcount == 0 &&
  733                     ifp->if_hw_tsomaxsegsize == 0) {
  734                         /*
  735                          * The TSO defaults needs to be such that an
  736                          * NFS mbuf list of 35 mbufs totalling just
  737                          * below 64K works and that a chain of mbufs
  738                          * can be defragged into at most 32 segments:
  739                          */
  740                         ifp->if_hw_tsomax = min(IP_MAXPACKET, (32 * MCLBYTES) -
  741                             (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN));
  742                         ifp->if_hw_tsomaxsegcount = 35;
  743                         ifp->if_hw_tsomaxsegsize = 2048;        /* 2K */
  744 
  745                         /* XXX some drivers set IFCAP_TSO after ethernet attach */
  746                         if (ifp->if_capabilities & IFCAP_TSO) {
  747                                 if_printf(ifp, "Using defaults for TSO: %u/%u/%u\n",
  748                                     ifp->if_hw_tsomax,
  749                                     ifp->if_hw_tsomaxsegcount,
  750                                     ifp->if_hw_tsomaxsegsize);
  751                         }
  752                 }
  753 #endif
  754         }
  755 #ifdef VIMAGE
  756         else {
  757                 /*
  758                  * Update the interface index in the link layer address
  759                  * of the interface.
  760                  */
  761                 for (ifa = ifp->if_addr; ifa != NULL;
  762                     ifa = TAILQ_NEXT(ifa, ifa_link)) {
  763                         if (ifa->ifa_addr->sa_family == AF_LINK) {
  764                                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
  765                                 sdl->sdl_index = ifp->if_index;
  766                         }
  767                 }
  768         }
  769 #endif
  770 
  771         IFNET_WLOCK();
  772         TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
  773 #ifdef VIMAGE
  774         curvnet->vnet_ifcnt++;
  775 #endif
  776         IFNET_WUNLOCK();
  777 
  778         if (domain_init_status >= 2)
  779                 if_attachdomain1(ifp);
  780 
  781         EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
  782         if (IS_DEFAULT_VNET(curvnet))
  783                 devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
  784 
  785         /* Announce the interface. */
  786         rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
  787 }
  788 
  789 static void
  790 if_attachdomain(void *dummy)
  791 {
  792         struct ifnet *ifp;
  793         int s;
  794 
  795         s = splnet();
  796         TAILQ_FOREACH(ifp, &V_ifnet, if_link)
  797                 if_attachdomain1(ifp);
  798         splx(s);
  799 }
  800 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
  801     if_attachdomain, NULL);
  802 
  803 static void
  804 if_attachdomain1(struct ifnet *ifp)
  805 {
  806         struct domain *dp;
  807         int s;
  808 
  809         s = splnet();
  810 
  811         /*
  812          * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
  813          * cannot lock ifp->if_afdata initialization, entirely.
  814          */
  815         if (IF_AFDATA_TRYLOCK(ifp) == 0) {
  816                 splx(s);
  817                 return;
  818         }
  819         if (ifp->if_afdata_initialized >= domain_init_status) {
  820                 IF_AFDATA_UNLOCK(ifp);
  821                 splx(s);
  822                 printf("if_attachdomain called more than once on %s\n",
  823                     ifp->if_xname);
  824                 return;
  825         }
  826         ifp->if_afdata_initialized = domain_init_status;
  827         IF_AFDATA_UNLOCK(ifp);
  828 
  829         /* address family dependent data region */
  830         bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
  831         for (dp = domains; dp; dp = dp->dom_next) {
  832                 if (dp->dom_ifattach)
  833                         ifp->if_afdata[dp->dom_family] =
  834                             (*dp->dom_ifattach)(ifp);
  835         }
  836 
  837         splx(s);
  838 }
  839 
  840 /*
  841  * Remove any unicast or broadcast network addresses from an interface.
  842  */
  843 void
  844 if_purgeaddrs(struct ifnet *ifp)
  845 {
  846         struct ifaddr *ifa, *next;
  847 
  848         TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  849                 if (ifa->ifa_addr->sa_family == AF_LINK)
  850                         continue;
  851 #ifdef INET
  852                 /* XXX: Ugly!! ad hoc just for INET */
  853                 if (ifa->ifa_addr->sa_family == AF_INET) {
  854                         struct ifaliasreq ifr;
  855 
  856                         bzero(&ifr, sizeof(ifr));
  857                         ifr.ifra_addr = *ifa->ifa_addr;
  858                         if (ifa->ifa_dstaddr)
  859                                 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
  860                         if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
  861                             NULL) == 0)
  862                                 continue;
  863                 }
  864 #endif /* INET */
  865 #ifdef INET6
  866                 if (ifa->ifa_addr->sa_family == AF_INET6) {
  867                         in6_purgeaddr(ifa);
  868                         /* ifp_addrhead is already updated */
  869                         continue;
  870                 }
  871 #endif /* INET6 */
  872                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  873                 ifa_free(ifa);
  874         }
  875 }
  876 
  877 /*
  878  * Remove any multicast network addresses from an interface when an ifnet
  879  * is going away.
  880  */
  881 static void
  882 if_purgemaddrs(struct ifnet *ifp)
  883 {
  884         struct ifmultiaddr *ifma;
  885         struct ifmultiaddr *next;
  886 
  887         IF_ADDR_WLOCK(ifp);
  888         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
  889                 if_delmulti_locked(ifp, ifma, 1);
  890         IF_ADDR_WUNLOCK(ifp);
  891 }
  892 
  893 /*
  894  * Detach an interface, removing it from the list of "active" interfaces.
  895  * If vmove flag is set on entry to if_detach_internal(), perform only a
  896  * limited subset of cleanup tasks, given that we are moving an ifnet from
  897  * one vnet to another, where it must be fully operational.
  898  *
  899  * XXXRW: There are some significant questions about event ordering, and
  900  * how to prevent things from starting to use the interface during detach.
  901  */
  902 void
  903 if_detach(struct ifnet *ifp)
  904 {
  905 
  906         CURVNET_SET_QUIET(ifp->if_vnet);
  907         if_detach_internal(ifp, 0);
  908         CURVNET_RESTORE();
  909 }
  910 
  911 static void
  912 if_detach_internal(struct ifnet *ifp, int vmove)
  913 {
  914         struct ifaddr *ifa;
  915         struct radix_node_head  *rnh;
  916         int i, j;
  917         struct domain *dp;
  918         struct ifnet *iter;
  919         int found = 0;
  920 
  921         IFNET_WLOCK();
  922         TAILQ_FOREACH(iter, &V_ifnet, if_link)
  923                 if (iter == ifp) {
  924                         TAILQ_REMOVE(&V_ifnet, ifp, if_link);
  925                         found = 1;
  926                         break;
  927                 }
  928 #ifdef VIMAGE
  929         if (found)
  930                 curvnet->vnet_ifcnt--;
  931 #endif
  932         IFNET_WUNLOCK();
  933         if (!found) {
  934                 if (vmove)
  935                         panic("%s: ifp=%p not on the ifnet tailq %p",
  936                             __func__, ifp, &V_ifnet);
  937                 else
  938                         return; /* XXX this should panic as well? */
  939         }
  940 
  941         /*
  942          * Remove/wait for pending events.
  943          */
  944         taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
  945 
  946         /*
  947          * Remove routes and flush queues.
  948          */
  949         if_down(ifp);
  950 #ifdef ALTQ
  951         if (ALTQ_IS_ENABLED(&ifp->if_snd))
  952                 altq_disable(&ifp->if_snd);
  953         if (ALTQ_IS_ATTACHED(&ifp->if_snd))
  954                 altq_detach(&ifp->if_snd);
  955 #endif
  956 
  957         if_purgeaddrs(ifp);
  958 
  959 #ifdef INET
  960         in_ifdetach(ifp);
  961 #endif
  962 
  963 #ifdef INET6
  964         /*
  965          * Remove all IPv6 kernel structs related to ifp.  This should be done
  966          * before removing routing entries below, since IPv6 interface direct
  967          * routes are expected to be removed by the IPv6-specific kernel API.
  968          * Otherwise, the kernel will detect some inconsistency and bark it.
  969          */
  970         in6_ifdetach(ifp);
  971 #endif
  972         if_purgemaddrs(ifp);
  973 
  974         if (!vmove) {
  975                 /*
  976                  * Prevent further calls into the device driver via ifnet.
  977                  */
  978                 if_dead(ifp);
  979 
  980                 /*
  981                  * Remove link ifaddr pointer and maybe decrement if_index.
  982                  * Clean up all addresses.
  983                  */
  984                 ifp->if_addr = NULL;
  985 
  986                 /* We can now free link ifaddr. */
  987                 if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
  988                         ifa = TAILQ_FIRST(&ifp->if_addrhead);
  989                         TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  990                         ifa_free(ifa);
  991                 }
  992         }
  993 
  994         /*
  995          * Delete all remaining routes using this interface
  996          * Unfortuneatly the only way to do this is to slog through
  997          * the entire routing table looking for routes which point
  998          * to this interface...oh well...
  999          */
 1000         for (i = 1; i <= AF_MAX; i++) {
 1001                 for (j = 0; j < rt_numfibs; j++) {
 1002                         rnh = rt_tables_get_rnh(j, i);
 1003                         if (rnh == NULL)
 1004                                 continue;
 1005                         RADIX_NODE_HEAD_LOCK(rnh);
 1006                         (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
 1007                         RADIX_NODE_HEAD_UNLOCK(rnh);
 1008                 }
 1009         }
 1010 
 1011         /* Announce that the interface is gone. */
 1012         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 1013         EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 1014         if (IS_DEFAULT_VNET(curvnet))
 1015                 devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
 1016         if_delgroups(ifp);
 1017 
 1018         /*
 1019          * We cannot hold the lock over dom_ifdetach calls as they might
 1020          * sleep, for example trying to drain a callout, thus open up the
 1021          * theoretical race with re-attaching.
 1022          */
 1023         IF_AFDATA_LOCK(ifp);
 1024         i = ifp->if_afdata_initialized;
 1025         ifp->if_afdata_initialized = 0;
 1026         IF_AFDATA_UNLOCK(ifp);
 1027         for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
 1028                 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
 1029                         (*dp->dom_ifdetach)(ifp,
 1030                             ifp->if_afdata[dp->dom_family]);
 1031         }
 1032 }
 1033 
 1034 #ifdef VIMAGE
 1035 /*
 1036  * if_vmove() performs a limited version of if_detach() in current
 1037  * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
 1038  * An attempt is made to shrink if_index in current vnet, find an
 1039  * unused if_index in target vnet and calls if_grow() if necessary,
 1040  * and finally find an unused if_xname for the target vnet.
 1041  */
 1042 void
 1043 if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
 1044 {
 1045         u_short idx;
 1046 
 1047         /*
 1048          * Detach from current vnet, but preserve LLADDR info, do not
 1049          * mark as dead etc. so that the ifnet can be reattached later.
 1050          */
 1051         if_detach_internal(ifp, 1);
 1052 
 1053         /*
 1054          * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
 1055          * the if_index for that vnet if possible.
 1056          *
 1057          * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
 1058          * or we'd lock on one vnet and unlock on another.
 1059          */
 1060         IFNET_WLOCK();
 1061         ifindex_free_locked(ifp->if_index);
 1062         IFNET_WUNLOCK();
 1063 
 1064         /*
 1065          * Perform interface-specific reassignment tasks, if provided by
 1066          * the driver.
 1067          */
 1068         if (ifp->if_reassign != NULL)
 1069                 ifp->if_reassign(ifp, new_vnet, NULL);
 1070 
 1071         /*
 1072          * Switch to the context of the target vnet.
 1073          */
 1074         CURVNET_SET_QUIET(new_vnet);
 1075 
 1076         IFNET_WLOCK();
 1077         if (ifindex_alloc_locked(&idx) != 0) {
 1078                 IFNET_WUNLOCK();
 1079                 panic("if_index overflow");
 1080         }
 1081         ifp->if_index = idx;
 1082         ifnet_setbyindex_locked(ifp->if_index, ifp);
 1083         IFNET_WUNLOCK();
 1084 
 1085         if_attach_internal(ifp, 1);
 1086 
 1087         CURVNET_RESTORE();
 1088 }
 1089 
 1090 /*
 1091  * Move an ifnet to or from another child prison/vnet, specified by the jail id.
 1092  */
 1093 static int
 1094 if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
 1095 {
 1096         struct prison *pr;
 1097         struct ifnet *difp;
 1098 
 1099         /* Try to find the prison within our visibility. */
 1100         sx_slock(&allprison_lock);
 1101         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1102         sx_sunlock(&allprison_lock);
 1103         if (pr == NULL)
 1104                 return (ENXIO);
 1105         prison_hold_locked(pr);
 1106         mtx_unlock(&pr->pr_mtx);
 1107 
 1108         /* Do not try to move the iface from and to the same prison. */
 1109         if (pr->pr_vnet == ifp->if_vnet) {
 1110                 prison_free(pr);
 1111                 return (EEXIST);
 1112         }
 1113 
 1114         /* Make sure the named iface does not exists in the dst. prison/vnet. */
 1115         /* XXX Lock interfaces to avoid races. */
 1116         CURVNET_SET_QUIET(pr->pr_vnet);
 1117         difp = ifunit(ifname);
 1118         CURVNET_RESTORE();
 1119         if (difp != NULL) {
 1120                 prison_free(pr);
 1121                 return (EEXIST);
 1122         }
 1123 
 1124         /* Move the interface into the child jail/vnet. */
 1125         if_vmove(ifp, pr->pr_vnet);
 1126 
 1127         /* Report the new if_xname back to the userland. */
 1128         sprintf(ifname, "%s", ifp->if_xname);
 1129 
 1130         prison_free(pr);
 1131         return (0);
 1132 }
 1133 
 1134 static int
 1135 if_vmove_reclaim(struct thread *td, char *ifname, int jid)
 1136 {
 1137         struct prison *pr;
 1138         struct vnet *vnet_dst;
 1139         struct ifnet *ifp;
 1140 
 1141         /* Try to find the prison within our visibility. */
 1142         sx_slock(&allprison_lock);
 1143         pr = prison_find_child(td->td_ucred->cr_prison, jid);
 1144         sx_sunlock(&allprison_lock);
 1145         if (pr == NULL)
 1146                 return (ENXIO);
 1147         prison_hold_locked(pr);
 1148         mtx_unlock(&pr->pr_mtx);
 1149 
 1150         /* Make sure the named iface exists in the source prison/vnet. */
 1151         CURVNET_SET(pr->pr_vnet);
 1152         ifp = ifunit(ifname);           /* XXX Lock to avoid races. */
 1153         if (ifp == NULL) {
 1154                 CURVNET_RESTORE();
 1155                 prison_free(pr);
 1156                 return (ENXIO);
 1157         }
 1158 
 1159         /* Do not try to move the iface from and to the same prison. */
 1160         vnet_dst = TD_TO_VNET(td);
 1161         if (vnet_dst == ifp->if_vnet) {
 1162                 CURVNET_RESTORE();
 1163                 prison_free(pr);
 1164                 return (EEXIST);
 1165         }
 1166 
 1167         /* Get interface back from child jail/vnet. */
 1168         if_vmove(ifp, vnet_dst);
 1169         CURVNET_RESTORE();
 1170 
 1171         /* Report the new if_xname back to the userland. */
 1172         sprintf(ifname, "%s", ifp->if_xname);
 1173 
 1174         prison_free(pr);
 1175         return (0);
 1176 }
 1177 #endif /* VIMAGE */
 1178 
 1179 /*
 1180  * Add a group to an interface
 1181  */
 1182 int
 1183 if_addgroup(struct ifnet *ifp, const char *groupname)
 1184 {
 1185         struct ifg_list         *ifgl;
 1186         struct ifg_group        *ifg = NULL;
 1187         struct ifg_member       *ifgm;
 1188 
 1189         if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
 1190             groupname[strlen(groupname) - 1] <= '9')
 1191                 return (EINVAL);
 1192 
 1193         IFNET_WLOCK();
 1194         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1195                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
 1196                         IFNET_WUNLOCK();
 1197                         return (EEXIST);
 1198                 }
 1199 
 1200         if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
 1201             M_NOWAIT)) == NULL) {
 1202                 IFNET_WUNLOCK();
 1203                 return (ENOMEM);
 1204         }
 1205 
 1206         if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
 1207             M_TEMP, M_NOWAIT)) == NULL) {
 1208                 free(ifgl, M_TEMP);
 1209                 IFNET_WUNLOCK();
 1210                 return (ENOMEM);
 1211         }
 1212 
 1213         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1214                 if (!strcmp(ifg->ifg_group, groupname))
 1215                         break;
 1216 
 1217         if (ifg == NULL) {
 1218                 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
 1219                     M_TEMP, M_NOWAIT)) == NULL) {
 1220                         free(ifgl, M_TEMP);
 1221                         free(ifgm, M_TEMP);
 1222                         IFNET_WUNLOCK();
 1223                         return (ENOMEM);
 1224                 }
 1225                 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
 1226                 ifg->ifg_refcnt = 0;
 1227                 TAILQ_INIT(&ifg->ifg_members);
 1228                 EVENTHANDLER_INVOKE(group_attach_event, ifg);
 1229                 TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
 1230         }
 1231 
 1232         ifg->ifg_refcnt++;
 1233         ifgl->ifgl_group = ifg;
 1234         ifgm->ifgm_ifp = ifp;
 1235 
 1236         IF_ADDR_WLOCK(ifp);
 1237         TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
 1238         TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
 1239         IF_ADDR_WUNLOCK(ifp);
 1240 
 1241         IFNET_WUNLOCK();
 1242 
 1243         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1244 
 1245         return (0);
 1246 }
 1247 
 1248 /*
 1249  * Remove a group from an interface
 1250  */
 1251 int
 1252 if_delgroup(struct ifnet *ifp, const char *groupname)
 1253 {
 1254         struct ifg_list         *ifgl;
 1255         struct ifg_member       *ifgm;
 1256 
 1257         IFNET_WLOCK();
 1258         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1259                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
 1260                         break;
 1261         if (ifgl == NULL) {
 1262                 IFNET_WUNLOCK();
 1263                 return (ENOENT);
 1264         }
 1265 
 1266         IF_ADDR_WLOCK(ifp);
 1267         TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1268         IF_ADDR_WUNLOCK(ifp);
 1269 
 1270         TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1271                 if (ifgm->ifgm_ifp == ifp)
 1272                         break;
 1273 
 1274         if (ifgm != NULL) {
 1275                 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
 1276                 free(ifgm, M_TEMP);
 1277         }
 1278 
 1279         if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1280                 TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1281                 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
 1282                 free(ifgl->ifgl_group, M_TEMP);
 1283         }
 1284         IFNET_WUNLOCK();
 1285 
 1286         free(ifgl, M_TEMP);
 1287 
 1288         EVENTHANDLER_INVOKE(group_change_event, groupname);
 1289 
 1290         return (0);
 1291 }
 1292 
 1293 /*
 1294  * Remove an interface from all groups
 1295  */
 1296 static void
 1297 if_delgroups(struct ifnet *ifp)
 1298 {
 1299         struct ifg_list         *ifgl;
 1300         struct ifg_member       *ifgm;
 1301         char groupname[IFNAMSIZ];
 1302 
 1303         IFNET_WLOCK();
 1304         while (!TAILQ_EMPTY(&ifp->if_groups)) {
 1305                 ifgl = TAILQ_FIRST(&ifp->if_groups);
 1306 
 1307                 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
 1308 
 1309                 IF_ADDR_WLOCK(ifp);
 1310                 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
 1311                 IF_ADDR_WUNLOCK(ifp);
 1312 
 1313                 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
 1314                         if (ifgm->ifgm_ifp == ifp)
 1315                                 break;
 1316 
 1317                 if (ifgm != NULL) {
 1318                         TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
 1319                             ifgm_next);
 1320                         free(ifgm, M_TEMP);
 1321                 }
 1322 
 1323                 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1324                         TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
 1325                         EVENTHANDLER_INVOKE(group_detach_event,
 1326                             ifgl->ifgl_group);
 1327                         free(ifgl->ifgl_group, M_TEMP);
 1328                 }
 1329                 IFNET_WUNLOCK();
 1330 
 1331                 free(ifgl, M_TEMP);
 1332 
 1333                 EVENTHANDLER_INVOKE(group_change_event, groupname);
 1334 
 1335                 IFNET_WLOCK();
 1336         }
 1337         IFNET_WUNLOCK();
 1338 }
 1339 
 1340 /*
 1341  * Stores all groups from an interface in memory pointed
 1342  * to by data
 1343  */
 1344 static int
 1345 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
 1346 {
 1347         int                      len, error;
 1348         struct ifg_list         *ifgl;
 1349         struct ifg_req           ifgrq, *ifgp;
 1350         struct ifgroupreq       *ifgr = data;
 1351 
 1352         if (ifgr->ifgr_len == 0) {
 1353                 IF_ADDR_RLOCK(ifp);
 1354                 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1355                         ifgr->ifgr_len += sizeof(struct ifg_req);
 1356                 IF_ADDR_RUNLOCK(ifp);
 1357                 return (0);
 1358         }
 1359 
 1360         len = ifgr->ifgr_len;
 1361         ifgp = ifgr->ifgr_groups;
 1362         /* XXX: wire */
 1363         IF_ADDR_RLOCK(ifp);
 1364         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
 1365                 if (len < sizeof(ifgrq)) {
 1366                         IF_ADDR_RUNLOCK(ifp);
 1367                         return (EINVAL);
 1368                 }
 1369                 bzero(&ifgrq, sizeof ifgrq);
 1370                 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
 1371                     sizeof(ifgrq.ifgrq_group));
 1372                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1373                         IF_ADDR_RUNLOCK(ifp);
 1374                         return (error);
 1375                 }
 1376                 len -= sizeof(ifgrq);
 1377                 ifgp++;
 1378         }
 1379         IF_ADDR_RUNLOCK(ifp);
 1380 
 1381         return (0);
 1382 }
 1383 
 1384 /*
 1385  * Stores all members of a group in memory pointed to by data
 1386  */
 1387 static int
 1388 if_getgroupmembers(struct ifgroupreq *data)
 1389 {
 1390         struct ifgroupreq       *ifgr = data;
 1391         struct ifg_group        *ifg;
 1392         struct ifg_member       *ifgm;
 1393         struct ifg_req           ifgrq, *ifgp;
 1394         int                      len, error;
 1395 
 1396         IFNET_RLOCK();
 1397         TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
 1398                 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
 1399                         break;
 1400         if (ifg == NULL) {
 1401                 IFNET_RUNLOCK();
 1402                 return (ENOENT);
 1403         }
 1404 
 1405         if (ifgr->ifgr_len == 0) {
 1406                 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
 1407                         ifgr->ifgr_len += sizeof(ifgrq);
 1408                 IFNET_RUNLOCK();
 1409                 return (0);
 1410         }
 1411 
 1412         len = ifgr->ifgr_len;
 1413         ifgp = ifgr->ifgr_groups;
 1414         TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
 1415                 if (len < sizeof(ifgrq)) {
 1416                         IFNET_RUNLOCK();
 1417                         return (EINVAL);
 1418                 }
 1419                 bzero(&ifgrq, sizeof ifgrq);
 1420                 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
 1421                     sizeof(ifgrq.ifgrq_member));
 1422                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1423                         IFNET_RUNLOCK();
 1424                         return (error);
 1425                 }
 1426                 len -= sizeof(ifgrq);
 1427                 ifgp++;
 1428         }
 1429         IFNET_RUNLOCK();
 1430 
 1431         return (0);
 1432 }
 1433 
 1434 /*
 1435  * Delete Routes for a Network Interface
 1436  *
 1437  * Called for each routing entry via the rnh->rnh_walktree() call above
 1438  * to delete all route entries referencing a detaching network interface.
 1439  *
 1440  * Arguments:
 1441  *      rn      pointer to node in the routing table
 1442  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
 1443  *
 1444  * Returns:
 1445  *      0       successful
 1446  *      errno   failed - reason indicated
 1447  *
 1448  */
 1449 static int
 1450 if_rtdel(struct radix_node *rn, void *arg)
 1451 {
 1452         struct rtentry  *rt = (struct rtentry *)rn;
 1453         struct ifnet    *ifp = arg;
 1454         int             err;
 1455 
 1456         if (rt->rt_ifp == ifp) {
 1457 
 1458                 /*
 1459                  * Protect (sorta) against walktree recursion problems
 1460                  * with cloned routes
 1461                  */
 1462                 if ((rt->rt_flags & RTF_UP) == 0)
 1463                         return (0);
 1464 
 1465                 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
 1466                                 rt_mask(rt),
 1467                                 rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
 1468                                 (struct rtentry **) NULL, rt->rt_fibnum);
 1469                 if (err) {
 1470                         log(LOG_WARNING, "if_rtdel: error %d\n", err);
 1471                 }
 1472         }
 1473 
 1474         return (0);
 1475 }
 1476 
 1477 /*
 1478  * Wrapper functions for struct ifnet address list locking macros.  These are
 1479  * used by kernel modules to avoid encoding programming interface or binary
 1480  * interface assumptions that may be violated when kernel-internal locking
 1481  * approaches change.
 1482  */
 1483 void
 1484 if_addr_rlock(struct ifnet *ifp)
 1485 {
 1486 
 1487         IF_ADDR_RLOCK(ifp);
 1488 }
 1489 
 1490 void
 1491 if_addr_runlock(struct ifnet *ifp)
 1492 {
 1493 
 1494         IF_ADDR_RUNLOCK(ifp);
 1495 }
 1496 
 1497 void
 1498 if_maddr_rlock(struct ifnet *ifp)
 1499 {
 1500 
 1501         IF_ADDR_RLOCK(ifp);
 1502 }
 1503 
 1504 void
 1505 if_maddr_runlock(struct ifnet *ifp)
 1506 {
 1507 
 1508         IF_ADDR_RUNLOCK(ifp);
 1509 }
 1510 
 1511 /*
 1512  * Reference count functions for ifaddrs.
 1513  */
 1514 void
 1515 ifa_init(struct ifaddr *ifa)
 1516 {
 1517 
 1518         mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
 1519         refcount_init(&ifa->ifa_refcnt, 1);
 1520 }
 1521 
 1522 void
 1523 ifa_ref(struct ifaddr *ifa)
 1524 {
 1525 
 1526         refcount_acquire(&ifa->ifa_refcnt);
 1527 }
 1528 
 1529 void
 1530 ifa_free(struct ifaddr *ifa)
 1531 {
 1532 
 1533         if (refcount_release(&ifa->ifa_refcnt)) {
 1534                 mtx_destroy(&ifa->ifa_mtx);
 1535                 free(ifa, M_IFADDR);
 1536         }
 1537 }
 1538 
 1539 int
 1540 ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1541 {
 1542         int error = 0;
 1543         struct rtentry *rt = NULL;
 1544         struct rt_addrinfo info;
 1545         static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1546 
 1547         bzero(&info, sizeof(info));
 1548         info.rti_ifp = V_loif;
 1549         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1550         info.rti_info[RTAX_DST] = ia;
 1551         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1552         error = rtrequest1_fib(RTM_ADD, &info, &rt, ifa->ifa_ifp->if_fib);
 1553 
 1554         if (error == 0 && rt != NULL) {
 1555                 RT_LOCK(rt);
 1556                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
 1557                         ifa->ifa_ifp->if_type;
 1558                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
 1559                         ifa->ifa_ifp->if_index;
 1560                 RT_REMREF(rt);
 1561                 RT_UNLOCK(rt);
 1562         } else if (error != 0)
 1563                 log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
 1564 
 1565         return (error);
 1566 }
 1567 
 1568 int
 1569 ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
 1570 {
 1571         int error = 0;
 1572         struct rt_addrinfo info;
 1573         struct sockaddr_dl null_sdl;
 1574 
 1575         bzero(&null_sdl, sizeof(null_sdl));
 1576         null_sdl.sdl_len = sizeof(null_sdl);
 1577         null_sdl.sdl_family = AF_LINK;
 1578         null_sdl.sdl_type = ifa->ifa_ifp->if_type;
 1579         null_sdl.sdl_index = ifa->ifa_ifp->if_index;
 1580         bzero(&info, sizeof(info));
 1581         info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
 1582         info.rti_info[RTAX_DST] = ia;
 1583         info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
 1584         error = rtrequest1_fib(RTM_DELETE, &info, NULL, ifa->ifa_ifp->if_fib);
 1585 
 1586         if (error != 0)
 1587                 log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
 1588 
 1589         return (error);
 1590 }
 1591 
 1592 /*
 1593  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
 1594  * structs used to represent other address families, it is necessary
 1595  * to perform a different comparison.
 1596  */
 1597 
 1598 #define sa_equal(a1, a2)        \
 1599         (bcmp((a1), (a2), ((a1))->sa_len) == 0)
 1600 
 1601 #define sa_dl_equal(a1, a2)     \
 1602         ((((struct sockaddr_dl *)(a1))->sdl_len ==                      \
 1603          ((struct sockaddr_dl *)(a2))->sdl_len) &&                      \
 1604          (bcmp(LLADDR((struct sockaddr_dl *)(a1)),                      \
 1605                LLADDR((struct sockaddr_dl *)(a2)),                      \
 1606                ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
 1607 
 1608 /*
 1609  * Locate an interface based on a complete address.
 1610  */
 1611 /*ARGSUSED*/
 1612 static struct ifaddr *
 1613 ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
 1614 {
 1615         struct ifnet *ifp;
 1616         struct ifaddr *ifa;
 1617 
 1618         IFNET_RLOCK_NOSLEEP();
 1619         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1620                 IF_ADDR_RLOCK(ifp);
 1621                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1622                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1623                                 continue;
 1624                         if (sa_equal(addr, ifa->ifa_addr)) {
 1625                                 if (getref)
 1626                                         ifa_ref(ifa);
 1627                                 IF_ADDR_RUNLOCK(ifp);
 1628                                 goto done;
 1629                         }
 1630                         /* IP6 doesn't have broadcast */
 1631                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1632                             ifa->ifa_broadaddr &&
 1633                             ifa->ifa_broadaddr->sa_len != 0 &&
 1634                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1635                                 if (getref)
 1636                                         ifa_ref(ifa);
 1637                                 IF_ADDR_RUNLOCK(ifp);
 1638                                 goto done;
 1639                         }
 1640                 }
 1641                 IF_ADDR_RUNLOCK(ifp);
 1642         }
 1643         ifa = NULL;
 1644 done:
 1645         IFNET_RUNLOCK_NOSLEEP();
 1646         return (ifa);
 1647 }
 1648 
 1649 struct ifaddr *
 1650 ifa_ifwithaddr(struct sockaddr *addr)
 1651 {
 1652 
 1653         return (ifa_ifwithaddr_internal(addr, 1));
 1654 }
 1655 
 1656 int
 1657 ifa_ifwithaddr_check(struct sockaddr *addr)
 1658 {
 1659 
 1660         return (ifa_ifwithaddr_internal(addr, 0) != NULL);
 1661 }
 1662 
 1663 /*
 1664  * Locate an interface based on the broadcast address.
 1665  */
 1666 /* ARGSUSED */
 1667 struct ifaddr *
 1668 ifa_ifwithbroadaddr(struct sockaddr *addr)
 1669 {
 1670         struct ifnet *ifp;
 1671         struct ifaddr *ifa;
 1672 
 1673         IFNET_RLOCK_NOSLEEP();
 1674         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1675                 IF_ADDR_RLOCK(ifp);
 1676                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1677                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1678                                 continue;
 1679                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1680                             ifa->ifa_broadaddr &&
 1681                             ifa->ifa_broadaddr->sa_len != 0 &&
 1682                             sa_equal(ifa->ifa_broadaddr, addr)) {
 1683                                 ifa_ref(ifa);
 1684                                 IF_ADDR_RUNLOCK(ifp);
 1685                                 goto done;
 1686                         }
 1687                 }
 1688                 IF_ADDR_RUNLOCK(ifp);
 1689         }
 1690         ifa = NULL;
 1691 done:
 1692         IFNET_RUNLOCK_NOSLEEP();
 1693         return (ifa);
 1694 }
 1695 
 1696 /*
 1697  * Locate the point to point interface with a given destination address.
 1698  */
 1699 /*ARGSUSED*/
 1700 struct ifaddr *
 1701 ifa_ifwithdstaddr_fib(struct sockaddr *addr, int fibnum)
 1702 {
 1703         struct ifnet *ifp;
 1704         struct ifaddr *ifa;
 1705 
 1706         IFNET_RLOCK_NOSLEEP();
 1707         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1708                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
 1709                         continue;
 1710                 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
 1711                         continue;
 1712                 IF_ADDR_RLOCK(ifp);
 1713                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1714                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1715                                 continue;
 1716                         if (ifa->ifa_dstaddr != NULL &&
 1717                             sa_equal(addr, ifa->ifa_dstaddr)) {
 1718                                 ifa_ref(ifa);
 1719                                 IF_ADDR_RUNLOCK(ifp);
 1720                                 goto done;
 1721                         }
 1722                 }
 1723                 IF_ADDR_RUNLOCK(ifp);
 1724         }
 1725         ifa = NULL;
 1726 done:
 1727         IFNET_RUNLOCK_NOSLEEP();
 1728         return (ifa);
 1729 }
 1730 
 1731 struct ifaddr *
 1732 ifa_ifwithdstaddr(struct sockaddr *addr)
 1733 {
 1734 
 1735         return (ifa_ifwithdstaddr_fib(addr, RT_ALL_FIBS));
 1736 }
 1737 
 1738 /*
 1739  * Find an interface on a specific network.  If many, choice
 1740  * is most specific found.
 1741  */
 1742 struct ifaddr *
 1743 ifa_ifwithnet_fib(struct sockaddr *addr, int ignore_ptp, int fibnum)
 1744 {
 1745         struct ifnet *ifp;
 1746         struct ifaddr *ifa;
 1747         struct ifaddr *ifa_maybe = NULL;
 1748         u_int af = addr->sa_family;
 1749         char *addr_data = addr->sa_data, *cplim;
 1750 
 1751         /*
 1752          * AF_LINK addresses can be looked up directly by their index number,
 1753          * so do that if we can.
 1754          */
 1755         if (af == AF_LINK) {
 1756             struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
 1757             if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
 1758                 return (ifaddr_byindex(sdl->sdl_index));
 1759         }
 1760 
 1761         /*
 1762          * Scan though each interface, looking for ones that have addresses
 1763          * in this address family and the requested fib.  Maintain a reference
 1764          * on ifa_maybe once we find one, as we release the IF_ADDR_RLOCK() that
 1765          * kept it stable when we move onto the next interface.
 1766          */
 1767         IFNET_RLOCK_NOSLEEP();
 1768         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 1769                 if ((fibnum != RT_ALL_FIBS) && (ifp->if_fib != fibnum))
 1770                         continue;
 1771                 IF_ADDR_RLOCK(ifp);
 1772                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1773                         char *cp, *cp2, *cp3;
 1774 
 1775                         if (ifa->ifa_addr->sa_family != af)
 1776 next:                           continue;
 1777                         if (af == AF_INET && 
 1778                             ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
 1779                                 /*
 1780                                  * This is a bit broken as it doesn't
 1781                                  * take into account that the remote end may
 1782                                  * be a single node in the network we are
 1783                                  * looking for.
 1784                                  * The trouble is that we don't know the
 1785                                  * netmask for the remote end.
 1786                                  */
 1787                                 if (ifa->ifa_dstaddr != NULL &&
 1788                                     sa_equal(addr, ifa->ifa_dstaddr)) {
 1789                                         ifa_ref(ifa);
 1790                                         IF_ADDR_RUNLOCK(ifp);
 1791                                         goto done;
 1792                                 }
 1793                         } else {
 1794                                 /*
 1795                                  * if we have a special address handler,
 1796                                  * then use it instead of the generic one.
 1797                                  */
 1798                                 if (ifa->ifa_claim_addr) {
 1799                                         if ((*ifa->ifa_claim_addr)(ifa, addr)) {
 1800                                                 ifa_ref(ifa);
 1801                                                 IF_ADDR_RUNLOCK(ifp);
 1802                                                 goto done;
 1803                                         }
 1804                                         continue;
 1805                                 }
 1806 
 1807                                 /*
 1808                                  * Scan all the bits in the ifa's address.
 1809                                  * If a bit dissagrees with what we are
 1810                                  * looking for, mask it with the netmask
 1811                                  * to see if it really matters.
 1812                                  * (A byte at a time)
 1813                                  */
 1814                                 if (ifa->ifa_netmask == 0)
 1815                                         continue;
 1816                                 cp = addr_data;
 1817                                 cp2 = ifa->ifa_addr->sa_data;
 1818                                 cp3 = ifa->ifa_netmask->sa_data;
 1819                                 cplim = ifa->ifa_netmask->sa_len
 1820                                         + (char *)ifa->ifa_netmask;
 1821                                 while (cp3 < cplim)
 1822                                         if ((*cp++ ^ *cp2++) & *cp3++)
 1823                                                 goto next; /* next address! */
 1824                                 /*
 1825                                  * If the netmask of what we just found
 1826                                  * is more specific than what we had before
 1827                                  * (if we had one) then remember the new one
 1828                                  * before continuing to search
 1829                                  * for an even better one.
 1830                                  */
 1831                                 if (ifa_maybe == NULL ||
 1832                                     rn_refines((caddr_t)ifa->ifa_netmask,
 1833                                     (caddr_t)ifa_maybe->ifa_netmask)) {
 1834                                         if (ifa_maybe != NULL)
 1835                                                 ifa_free(ifa_maybe);
 1836                                         ifa_maybe = ifa;
 1837                                         ifa_ref(ifa_maybe);
 1838                                 }
 1839                         }
 1840                 }
 1841                 IF_ADDR_RUNLOCK(ifp);
 1842         }
 1843         ifa = ifa_maybe;
 1844         ifa_maybe = NULL;
 1845 done:
 1846         IFNET_RUNLOCK_NOSLEEP();
 1847         if (ifa_maybe != NULL)
 1848                 ifa_free(ifa_maybe);
 1849         return (ifa);
 1850 }
 1851 
 1852 struct ifaddr *
 1853 ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
 1854 {
 1855 
 1856         return (ifa_ifwithnet_fib(addr, ignore_ptp, RT_ALL_FIBS));
 1857 }
 1858 
 1859 /*
 1860  * Find an interface address specific to an interface best matching
 1861  * a given address.
 1862  */
 1863 struct ifaddr *
 1864 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
 1865 {
 1866         struct ifaddr *ifa;
 1867         char *cp, *cp2, *cp3;
 1868         char *cplim;
 1869         struct ifaddr *ifa_maybe = NULL;
 1870         u_int af = addr->sa_family;
 1871 
 1872         if (af >= AF_MAX)
 1873                 return (NULL);
 1874         IF_ADDR_RLOCK(ifp);
 1875         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1876                 if (ifa->ifa_addr->sa_family != af)
 1877                         continue;
 1878                 if (ifa_maybe == NULL)
 1879                         ifa_maybe = ifa;
 1880                 if (ifa->ifa_netmask == 0) {
 1881                         if (sa_equal(addr, ifa->ifa_addr) ||
 1882                             (ifa->ifa_dstaddr &&
 1883                             sa_equal(addr, ifa->ifa_dstaddr)))
 1884                                 goto done;
 1885                         continue;
 1886                 }
 1887                 if (ifp->if_flags & IFF_POINTOPOINT) {
 1888                         if (sa_equal(addr, ifa->ifa_dstaddr))
 1889                                 goto done;
 1890                 } else {
 1891                         cp = addr->sa_data;
 1892                         cp2 = ifa->ifa_addr->sa_data;
 1893                         cp3 = ifa->ifa_netmask->sa_data;
 1894                         cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
 1895                         for (; cp3 < cplim; cp3++)
 1896                                 if ((*cp++ ^ *cp2++) & *cp3)
 1897                                         break;
 1898                         if (cp3 == cplim)
 1899                                 goto done;
 1900                 }
 1901         }
 1902         ifa = ifa_maybe;
 1903 done:
 1904         if (ifa != NULL)
 1905                 ifa_ref(ifa);
 1906         IF_ADDR_RUNLOCK(ifp);
 1907         return (ifa);
 1908 }
 1909 
 1910 #include <net/if_llatbl.h>
 1911 
 1912 /*
 1913  * Default action when installing a route with a Link Level gateway.
 1914  * Lookup an appropriate real ifa to point to.
 1915  * This should be moved to /sys/net/link.c eventually.
 1916  */
 1917 static void
 1918 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
 1919 {
 1920         struct ifaddr *ifa, *oifa;
 1921         struct sockaddr *dst;
 1922         struct ifnet *ifp;
 1923 
 1924         RT_LOCK_ASSERT(rt);
 1925 
 1926         if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
 1927             ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
 1928                 return;
 1929         ifa = ifaof_ifpforaddr(dst, ifp);
 1930         if (ifa) {
 1931                 oifa = rt->rt_ifa;
 1932                 rt->rt_ifa = ifa;
 1933                 ifa_free(oifa);
 1934                 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
 1935                         ifa->ifa_rtrequest(cmd, rt, info);
 1936         }
 1937 }
 1938 
 1939 /*
 1940  * Mark an interface down and notify protocols of
 1941  * the transition.
 1942  * NOTE: must be called at splnet or eqivalent.
 1943  */
 1944 static void
 1945 if_unroute(struct ifnet *ifp, int flag, int fam)
 1946 {
 1947         struct ifaddr *ifa;
 1948 
 1949         KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
 1950 
 1951         ifp->if_flags &= ~flag;
 1952         getmicrotime(&ifp->if_lastchange);
 1953         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1954                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1955                         pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
 1956         ifp->if_qflush(ifp);
 1957 
 1958         if (ifp->if_carp)
 1959                 (*carp_linkstate_p)(ifp);
 1960         rt_ifmsg(ifp);
 1961 }
 1962 
 1963 /*
 1964  * Mark an interface up and notify protocols of
 1965  * the transition.
 1966  * NOTE: must be called at splnet or eqivalent.
 1967  */
 1968 static void
 1969 if_route(struct ifnet *ifp, int flag, int fam)
 1970 {
 1971         struct ifaddr *ifa;
 1972 
 1973         KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
 1974 
 1975         ifp->if_flags |= flag;
 1976         getmicrotime(&ifp->if_lastchange);
 1977         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1978                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1979                         pfctlinput(PRC_IFUP, ifa->ifa_addr);
 1980         if (ifp->if_carp)
 1981                 (*carp_linkstate_p)(ifp);
 1982         rt_ifmsg(ifp);
 1983 #ifdef INET6
 1984         in6_if_up(ifp);
 1985 #endif
 1986 }
 1987 
 1988 void    (*vlan_link_state_p)(struct ifnet *);   /* XXX: private from if_vlan */
 1989 void    (*vlan_trunk_cap_p)(struct ifnet *);            /* XXX: private from if_vlan */
 1990 struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
 1991 struct  ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
 1992 int     (*vlan_tag_p)(struct ifnet *, uint16_t *);
 1993 int     (*vlan_setcookie_p)(struct ifnet *, void *);
 1994 void    *(*vlan_cookie_p)(struct ifnet *);
 1995 
 1996 /*
 1997  * Handle a change in the interface link state. To avoid LORs
 1998  * between driver lock and upper layer locks, as well as possible
 1999  * recursions, we post event to taskqueue, and all job
 2000  * is done in static do_link_state_change().
 2001  */
 2002 void
 2003 if_link_state_change(struct ifnet *ifp, int link_state)
 2004 {
 2005         /* Return if state hasn't changed. */
 2006         if (ifp->if_link_state == link_state)
 2007                 return;
 2008 
 2009         ifp->if_link_state = link_state;
 2010 
 2011         taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
 2012 }
 2013 
 2014 static void
 2015 do_link_state_change(void *arg, int pending)
 2016 {
 2017         struct ifnet *ifp = (struct ifnet *)arg;
 2018         int link_state = ifp->if_link_state;
 2019         CURVNET_SET(ifp->if_vnet);
 2020 
 2021         /* Notify that the link state has changed. */
 2022         rt_ifmsg(ifp);
 2023         if (ifp->if_vlantrunk != NULL)
 2024                 (*vlan_link_state_p)(ifp);
 2025 
 2026         if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
 2027             IFP2AC(ifp)->ac_netgraph != NULL)
 2028                 (*ng_ether_link_state_p)(ifp, link_state);
 2029         if (ifp->if_carp)
 2030                 (*carp_linkstate_p)(ifp);
 2031         if (ifp->if_bridge)
 2032                 (*bridge_linkstate_p)(ifp);
 2033         if (ifp->if_lagg)
 2034                 (*lagg_linkstate_p)(ifp, link_state);
 2035 
 2036         if (IS_DEFAULT_VNET(curvnet))
 2037                 devctl_notify("IFNET", ifp->if_xname,
 2038                     (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
 2039                     NULL);
 2040         if (pending > 1)
 2041                 if_printf(ifp, "%d link states coalesced\n", pending);
 2042         if (log_link_state_change)
 2043                 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
 2044                     (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
 2045         EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
 2046         CURVNET_RESTORE();
 2047 }
 2048 
 2049 /*
 2050  * Mark an interface down and notify protocols of
 2051  * the transition.
 2052  * NOTE: must be called at splnet or eqivalent.
 2053  */
 2054 void
 2055 if_down(struct ifnet *ifp)
 2056 {
 2057 
 2058         if_unroute(ifp, IFF_UP, AF_UNSPEC);
 2059 }
 2060 
 2061 /*
 2062  * Mark an interface up and notify protocols of
 2063  * the transition.
 2064  * NOTE: must be called at splnet or eqivalent.
 2065  */
 2066 void
 2067 if_up(struct ifnet *ifp)
 2068 {
 2069 
 2070         if_route(ifp, IFF_UP, AF_UNSPEC);
 2071 }
 2072 
 2073 /*
 2074  * Flush an interface queue.
 2075  */
 2076 void
 2077 if_qflush(struct ifnet *ifp)
 2078 {
 2079         struct mbuf *m, *n;
 2080         struct ifaltq *ifq;
 2081         
 2082         ifq = &ifp->if_snd;
 2083         IFQ_LOCK(ifq);
 2084 #ifdef ALTQ
 2085         if (ALTQ_IS_ENABLED(ifq))
 2086                 ALTQ_PURGE(ifq);
 2087 #endif
 2088         n = ifq->ifq_head;
 2089         while ((m = n) != 0) {
 2090                 n = m->m_act;
 2091                 m_freem(m);
 2092         }
 2093         ifq->ifq_head = 0;
 2094         ifq->ifq_tail = 0;
 2095         ifq->ifq_len = 0;
 2096         IFQ_UNLOCK(ifq);
 2097 }
 2098 
 2099 /*
 2100  * Map interface name to interface structure pointer, with or without
 2101  * returning a reference.
 2102  */
 2103 struct ifnet *
 2104 ifunit_ref(const char *name)
 2105 {
 2106         struct ifnet *ifp;
 2107 
 2108         IFNET_RLOCK_NOSLEEP();
 2109         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2110                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
 2111                     !(ifp->if_flags & IFF_DYING))
 2112                         break;
 2113         }
 2114         if (ifp != NULL)
 2115                 if_ref(ifp);
 2116         IFNET_RUNLOCK_NOSLEEP();
 2117         return (ifp);
 2118 }
 2119 
 2120 struct ifnet *
 2121 ifunit(const char *name)
 2122 {
 2123         struct ifnet *ifp;
 2124 
 2125         IFNET_RLOCK_NOSLEEP();
 2126         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2127                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
 2128                         break;
 2129         }
 2130         IFNET_RUNLOCK_NOSLEEP();
 2131         return (ifp);
 2132 }
 2133 
 2134 /*
 2135  * Hardware specific interface ioctls.
 2136  */
 2137 static int
 2138 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
 2139 {
 2140         struct ifreq *ifr;
 2141         struct ifstat *ifs;
 2142         int error = 0;
 2143         int new_flags, temp_flags;
 2144         size_t namelen, onamelen;
 2145         size_t descrlen;
 2146         char *descrbuf, *odescrbuf;
 2147         char new_name[IFNAMSIZ];
 2148         struct ifaddr *ifa;
 2149         struct sockaddr_dl *sdl;
 2150 
 2151         ifr = (struct ifreq *)data;
 2152         switch (cmd) {
 2153         case SIOCGIFINDEX:
 2154                 ifr->ifr_index = ifp->if_index;
 2155                 break;
 2156 
 2157         case SIOCGIFFLAGS:
 2158                 temp_flags = ifp->if_flags | ifp->if_drv_flags;
 2159                 ifr->ifr_flags = temp_flags & 0xffff;
 2160                 ifr->ifr_flagshigh = temp_flags >> 16;
 2161                 break;
 2162 
 2163         case SIOCGIFCAP:
 2164                 ifr->ifr_reqcap = ifp->if_capabilities;
 2165                 ifr->ifr_curcap = ifp->if_capenable;
 2166                 break;
 2167 
 2168 #ifdef MAC
 2169         case SIOCGIFMAC:
 2170                 error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
 2171                 break;
 2172 #endif
 2173 
 2174         case SIOCGIFMETRIC:
 2175                 ifr->ifr_metric = ifp->if_metric;
 2176                 break;
 2177 
 2178         case SIOCGIFMTU:
 2179                 ifr->ifr_mtu = ifp->if_mtu;
 2180                 break;
 2181 
 2182         case SIOCGIFPHYS:
 2183                 ifr->ifr_phys = ifp->if_physical;
 2184                 break;
 2185 
 2186         case SIOCGIFDESCR:
 2187                 error = 0;
 2188                 sx_slock(&ifdescr_sx);
 2189                 if (ifp->if_description == NULL)
 2190                         error = ENOMSG;
 2191                 else {
 2192                         /* space for terminating nul */
 2193                         descrlen = strlen(ifp->if_description) + 1;
 2194                         if (ifr->ifr_buffer.length < descrlen)
 2195                                 ifr->ifr_buffer.buffer = NULL;
 2196                         else
 2197                                 error = copyout(ifp->if_description,
 2198                                     ifr->ifr_buffer.buffer, descrlen);
 2199                         ifr->ifr_buffer.length = descrlen;
 2200                 }
 2201                 sx_sunlock(&ifdescr_sx);
 2202                 break;
 2203 
 2204         case SIOCSIFDESCR:
 2205                 error = priv_check(td, PRIV_NET_SETIFDESCR);
 2206                 if (error)
 2207                         return (error);
 2208 
 2209                 /*
 2210                  * Copy only (length-1) bytes to make sure that
 2211                  * if_description is always nul terminated.  The
 2212                  * length parameter is supposed to count the
 2213                  * terminating nul in.
 2214                  */
 2215                 if (ifr->ifr_buffer.length > ifdescr_maxlen)
 2216                         return (ENAMETOOLONG);
 2217                 else if (ifr->ifr_buffer.length == 0)
 2218                         descrbuf = NULL;
 2219                 else {
 2220                         descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
 2221                             M_WAITOK | M_ZERO);
 2222                         error = copyin(ifr->ifr_buffer.buffer, descrbuf,
 2223                             ifr->ifr_buffer.length - 1);
 2224                         if (error) {
 2225                                 free(descrbuf, M_IFDESCR);
 2226                                 break;
 2227                         }
 2228                 }
 2229 
 2230                 sx_xlock(&ifdescr_sx);
 2231                 odescrbuf = ifp->if_description;
 2232                 ifp->if_description = descrbuf;
 2233                 sx_xunlock(&ifdescr_sx);
 2234 
 2235                 getmicrotime(&ifp->if_lastchange);
 2236                 free(odescrbuf, M_IFDESCR);
 2237                 break;
 2238 
 2239         case SIOCGIFFIB:
 2240                 ifr->ifr_fib = ifp->if_fib;
 2241                 break;
 2242 
 2243         case SIOCSIFFIB:
 2244                 error = priv_check(td, PRIV_NET_SETIFFIB);
 2245                 if (error)
 2246                         return (error);
 2247                 if (ifr->ifr_fib >= rt_numfibs)
 2248                         return (EINVAL);
 2249 
 2250                 ifp->if_fib = ifr->ifr_fib;
 2251                 break;
 2252 
 2253         case SIOCSIFFLAGS:
 2254                 error = priv_check(td, PRIV_NET_SETIFFLAGS);
 2255                 if (error)
 2256                         return (error);
 2257                 /*
 2258                  * Currently, no driver owned flags pass the IFF_CANTCHANGE
 2259                  * check, so we don't need special handling here yet.
 2260                  */
 2261                 new_flags = (ifr->ifr_flags & 0xffff) |
 2262                     (ifr->ifr_flagshigh << 16);
 2263                 if (ifp->if_flags & IFF_SMART) {
 2264                         /* Smart drivers twiddle their own routes */
 2265                 } else if (ifp->if_flags & IFF_UP &&
 2266                     (new_flags & IFF_UP) == 0) {
 2267                         int s = splimp();
 2268                         if_down(ifp);
 2269                         splx(s);
 2270                 } else if (new_flags & IFF_UP &&
 2271                     (ifp->if_flags & IFF_UP) == 0) {
 2272                         int s = splimp();
 2273                         if_up(ifp);
 2274                         splx(s);
 2275                 }
 2276                 /* See if permanently promiscuous mode bit is about to flip */
 2277                 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
 2278                         if (new_flags & IFF_PPROMISC)
 2279                                 ifp->if_flags |= IFF_PROMISC;
 2280                         else if (ifp->if_pcount == 0)
 2281                                 ifp->if_flags &= ~IFF_PROMISC;
 2282                         log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
 2283                             ifp->if_xname,
 2284                             (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
 2285                 }
 2286                 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
 2287                         (new_flags &~ IFF_CANTCHANGE);
 2288                 if (ifp->if_ioctl) {
 2289                         (void) (*ifp->if_ioctl)(ifp, cmd, data);
 2290                 }
 2291                 getmicrotime(&ifp->if_lastchange);
 2292                 break;
 2293 
 2294         case SIOCSIFCAP:
 2295                 error = priv_check(td, PRIV_NET_SETIFCAP);
 2296                 if (error)
 2297                         return (error);
 2298                 if (ifp->if_ioctl == NULL)
 2299                         return (EOPNOTSUPP);
 2300                 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
 2301                         return (EINVAL);
 2302                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2303                 if (error == 0)
 2304                         getmicrotime(&ifp->if_lastchange);
 2305                 break;
 2306 
 2307 #ifdef MAC
 2308         case SIOCSIFMAC:
 2309                 error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
 2310                 break;
 2311 #endif
 2312 
 2313         case SIOCSIFNAME:
 2314                 error = priv_check(td, PRIV_NET_SETIFNAME);
 2315                 if (error)
 2316                         return (error);
 2317                 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
 2318                 if (error != 0)
 2319                         return (error);
 2320                 if (new_name[0] == '\0')
 2321                         return (EINVAL);
 2322                 if (ifunit(new_name) != NULL)
 2323                         return (EEXIST);
 2324 
 2325                 /*
 2326                  * XXX: Locking.  Nothing else seems to lock if_flags,
 2327                  * and there are numerous other races with the
 2328                  * ifunit() checks not being atomic with namespace
 2329                  * changes (renames, vmoves, if_attach, etc).
 2330                  */
 2331                 ifp->if_flags |= IFF_RENAMING;
 2332                 
 2333                 /* Announce the departure of the interface. */
 2334                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 2335                 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 2336 
 2337                 log(LOG_INFO, "%s: changing name to '%s'\n",
 2338                     ifp->if_xname, new_name);
 2339 
 2340                 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
 2341                 ifa = ifp->if_addr;
 2342                 IFA_LOCK(ifa);
 2343                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 2344                 namelen = strlen(new_name);
 2345                 onamelen = sdl->sdl_nlen;
 2346                 /*
 2347                  * Move the address if needed.  This is safe because we
 2348                  * allocate space for a name of length IFNAMSIZ when we
 2349                  * create this in if_attach().
 2350                  */
 2351                 if (namelen != onamelen) {
 2352                         bcopy(sdl->sdl_data + onamelen,
 2353                             sdl->sdl_data + namelen, sdl->sdl_alen);
 2354                 }
 2355                 bcopy(new_name, sdl->sdl_data, namelen);
 2356                 sdl->sdl_nlen = namelen;
 2357                 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
 2358                 bzero(sdl->sdl_data, onamelen);
 2359                 while (namelen != 0)
 2360                         sdl->sdl_data[--namelen] = 0xff;
 2361                 IFA_UNLOCK(ifa);
 2362 
 2363                 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
 2364                 /* Announce the return of the interface. */
 2365                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 2366 
 2367                 ifp->if_flags &= ~IFF_RENAMING;
 2368                 break;
 2369 
 2370 #ifdef VIMAGE
 2371         case SIOCSIFVNET:
 2372                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2373                 if (error)
 2374                         return (error);
 2375                 error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
 2376                 break;
 2377 #endif
 2378 
 2379         case SIOCSIFMETRIC:
 2380                 error = priv_check(td, PRIV_NET_SETIFMETRIC);
 2381                 if (error)
 2382                         return (error);
 2383                 ifp->if_metric = ifr->ifr_metric;
 2384                 getmicrotime(&ifp->if_lastchange);
 2385                 break;
 2386 
 2387         case SIOCSIFPHYS:
 2388                 error = priv_check(td, PRIV_NET_SETIFPHYS);
 2389                 if (error)
 2390                         return (error);
 2391                 if (ifp->if_ioctl == NULL)
 2392                         return (EOPNOTSUPP);
 2393                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2394                 if (error == 0)
 2395                         getmicrotime(&ifp->if_lastchange);
 2396                 break;
 2397 
 2398         case SIOCSIFMTU:
 2399         {
 2400                 u_long oldmtu = ifp->if_mtu;
 2401 
 2402                 error = priv_check(td, PRIV_NET_SETIFMTU);
 2403                 if (error)
 2404                         return (error);
 2405                 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
 2406                         return (EINVAL);
 2407                 if (ifp->if_ioctl == NULL)
 2408                         return (EOPNOTSUPP);
 2409                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2410                 if (error == 0) {
 2411                         getmicrotime(&ifp->if_lastchange);
 2412                         rt_ifmsg(ifp);
 2413                 }
 2414                 /*
 2415                  * If the link MTU changed, do network layer specific procedure.
 2416                  */
 2417                 if (ifp->if_mtu != oldmtu) {
 2418 #ifdef INET6
 2419                         nd6_setmtu(ifp);
 2420 #endif
 2421                 }
 2422                 break;
 2423         }
 2424 
 2425         case SIOCADDMULTI:
 2426         case SIOCDELMULTI:
 2427                 if (cmd == SIOCADDMULTI)
 2428                         error = priv_check(td, PRIV_NET_ADDMULTI);
 2429                 else
 2430                         error = priv_check(td, PRIV_NET_DELMULTI);
 2431                 if (error)
 2432                         return (error);
 2433 
 2434                 /* Don't allow group membership on non-multicast interfaces. */
 2435                 if ((ifp->if_flags & IFF_MULTICAST) == 0)
 2436                         return (EOPNOTSUPP);
 2437 
 2438                 /* Don't let users screw up protocols' entries. */
 2439                 if (ifr->ifr_addr.sa_family != AF_LINK)
 2440                         return (EINVAL);
 2441 
 2442                 if (cmd == SIOCADDMULTI) {
 2443                         struct ifmultiaddr *ifma;
 2444 
 2445                         /*
 2446                          * Userland is only permitted to join groups once
 2447                          * via the if_addmulti() KPI, because it cannot hold
 2448                          * struct ifmultiaddr * between calls. It may also
 2449                          * lose a race while we check if the membership
 2450                          * already exists.
 2451                          */
 2452                         IF_ADDR_RLOCK(ifp);
 2453                         ifma = if_findmulti(ifp, &ifr->ifr_addr);
 2454                         IF_ADDR_RUNLOCK(ifp);
 2455                         if (ifma != NULL)
 2456                                 error = EADDRINUSE;
 2457                         else
 2458                                 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
 2459                 } else {
 2460                         error = if_delmulti(ifp, &ifr->ifr_addr);
 2461                 }
 2462                 if (error == 0)
 2463                         getmicrotime(&ifp->if_lastchange);
 2464                 break;
 2465 
 2466         case SIOCSIFPHYADDR:
 2467         case SIOCDIFPHYADDR:
 2468 #ifdef INET6
 2469         case SIOCSIFPHYADDR_IN6:
 2470 #endif
 2471         case SIOCSLIFPHYADDR:
 2472         case SIOCSIFMEDIA:
 2473         case SIOCSIFGENERIC:
 2474                 error = priv_check(td, PRIV_NET_HWIOCTL);
 2475                 if (error)
 2476                         return (error);
 2477                 if (ifp->if_ioctl == NULL)
 2478                         return (EOPNOTSUPP);
 2479                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2480                 if (error == 0)
 2481                         getmicrotime(&ifp->if_lastchange);
 2482                 break;
 2483 
 2484         case SIOCGIFSTATUS:
 2485                 ifs = (struct ifstat *)data;
 2486                 ifs->ascii[0] = '\0';
 2487 
 2488         case SIOCGIFPSRCADDR:
 2489         case SIOCGIFPDSTADDR:
 2490         case SIOCGLIFPHYADDR:
 2491         case SIOCGIFMEDIA:
 2492         case SIOCGIFGENERIC:
 2493                 if (ifp->if_ioctl == NULL)
 2494                         return (EOPNOTSUPP);
 2495                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2496                 break;
 2497 
 2498         case SIOCSIFLLADDR:
 2499                 error = priv_check(td, PRIV_NET_SETLLADDR);
 2500                 if (error)
 2501                         return (error);
 2502                 error = if_setlladdr(ifp,
 2503                     ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
 2504                 EVENTHANDLER_INVOKE(iflladdr_event, ifp);
 2505                 break;
 2506 
 2507         case SIOCAIFGROUP:
 2508         {
 2509                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2510 
 2511                 error = priv_check(td, PRIV_NET_ADDIFGROUP);
 2512                 if (error)
 2513                         return (error);
 2514                 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
 2515                         return (error);
 2516                 break;
 2517         }
 2518 
 2519         case SIOCGIFGROUP:
 2520                 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
 2521                         return (error);
 2522                 break;
 2523 
 2524         case SIOCDIFGROUP:
 2525         {
 2526                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 2527 
 2528                 error = priv_check(td, PRIV_NET_DELIFGROUP);
 2529                 if (error)
 2530                         return (error);
 2531                 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
 2532                         return (error);
 2533                 break;
 2534         }
 2535 
 2536         default:
 2537                 error = ENOIOCTL;
 2538                 break;
 2539         }
 2540         return (error);
 2541 }
 2542 
 2543 #ifdef COMPAT_FREEBSD32
 2544 struct ifconf32 {
 2545         int32_t ifc_len;
 2546         union {
 2547                 uint32_t        ifcu_buf;
 2548                 uint32_t        ifcu_req;
 2549         } ifc_ifcu;
 2550 };
 2551 #define SIOCGIFCONF32   _IOWR('i', 36, struct ifconf32)
 2552 #endif
 2553 
 2554 /*
 2555  * Interface ioctls.
 2556  */
 2557 int
 2558 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
 2559 {
 2560         struct ifnet *ifp;
 2561         struct ifreq *ifr;
 2562         int error;
 2563         int oif_flags;
 2564 
 2565         CURVNET_SET(so->so_vnet);
 2566         switch (cmd) {
 2567         case SIOCGIFCONF:
 2568         case OSIOCGIFCONF:
 2569                 error = ifconf(cmd, data);
 2570                 CURVNET_RESTORE();
 2571                 return (error);
 2572 
 2573 #ifdef COMPAT_FREEBSD32
 2574         case SIOCGIFCONF32:
 2575                 {
 2576                         struct ifconf32 *ifc32;
 2577                         struct ifconf ifc;
 2578 
 2579                         ifc32 = (struct ifconf32 *)data;
 2580                         ifc.ifc_len = ifc32->ifc_len;
 2581                         ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
 2582 
 2583                         error = ifconf(SIOCGIFCONF, (void *)&ifc);
 2584                         CURVNET_RESTORE();
 2585                         if (error == 0)
 2586                                 ifc32->ifc_len = ifc.ifc_len;
 2587                         return (error);
 2588                 }
 2589 #endif
 2590         }
 2591         ifr = (struct ifreq *)data;
 2592 
 2593         switch (cmd) {
 2594 #ifdef VIMAGE
 2595         case SIOCSIFRVNET:
 2596                 error = priv_check(td, PRIV_NET_SETIFVNET);
 2597                 if (error == 0)
 2598                         error = if_vmove_reclaim(td, ifr->ifr_name,
 2599                             ifr->ifr_jid);
 2600                 CURVNET_RESTORE();
 2601                 return (error);
 2602 #endif
 2603         case SIOCIFCREATE:
 2604         case SIOCIFCREATE2:
 2605                 error = priv_check(td, PRIV_NET_IFCREATE);
 2606                 if (error == 0)
 2607                         error = if_clone_create(ifr->ifr_name,
 2608                             sizeof(ifr->ifr_name),
 2609                             cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
 2610                 CURVNET_RESTORE();
 2611                 return (error);
 2612         case SIOCIFDESTROY:
 2613                 error = priv_check(td, PRIV_NET_IFDESTROY);
 2614                 if (error == 0)
 2615                         error = if_clone_destroy(ifr->ifr_name);
 2616                 CURVNET_RESTORE();
 2617                 return (error);
 2618 
 2619         case SIOCIFGCLONERS:
 2620                 error = if_clone_list((struct if_clonereq *)data);
 2621                 CURVNET_RESTORE();
 2622                 return (error);
 2623         case SIOCGIFGMEMB:
 2624                 error = if_getgroupmembers((struct ifgroupreq *)data);
 2625                 CURVNET_RESTORE();
 2626                 return (error);
 2627         }
 2628 
 2629         ifp = ifunit_ref(ifr->ifr_name);
 2630         if (ifp == NULL) {
 2631                 CURVNET_RESTORE();
 2632                 return (ENXIO);
 2633         }
 2634 
 2635         error = ifhwioctl(cmd, ifp, data, td);
 2636         if (error != ENOIOCTL) {
 2637                 if_rele(ifp);
 2638                 CURVNET_RESTORE();
 2639                 return (error);
 2640         }
 2641 
 2642         oif_flags = ifp->if_flags;
 2643         if (so->so_proto == NULL) {
 2644                 if_rele(ifp);
 2645                 CURVNET_RESTORE();
 2646                 return (EOPNOTSUPP);
 2647         }
 2648 
 2649         /*
 2650          * Pass the request on to the socket control method, and if the
 2651          * latter returns EOPNOTSUPP, directly to the interface.
 2652          *
 2653          * Make an exception for the legacy SIOCSIF* requests.  Drivers
 2654          * trust SIOCSIFADDR et al to come from an already privileged
 2655          * layer, and do not perform any credentials checks or input
 2656          * validation.
 2657          */
 2658 #ifndef COMPAT_43
 2659         error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
 2660                                                                  data,
 2661                                                                  ifp, td));
 2662         if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL &&
 2663             cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
 2664             cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
 2665                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2666 #else
 2667         {
 2668                 u_long ocmd = cmd;
 2669 
 2670                 switch (cmd) {
 2671 
 2672                 case SIOCSIFDSTADDR:
 2673                 case SIOCSIFADDR:
 2674                 case SIOCSIFBRDADDR:
 2675                 case SIOCSIFNETMASK:
 2676 #if BYTE_ORDER != BIG_ENDIAN
 2677                         if (ifr->ifr_addr.sa_family == 0 &&
 2678                             ifr->ifr_addr.sa_len < 16) {
 2679                                 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
 2680                                 ifr->ifr_addr.sa_len = 16;
 2681                         }
 2682 #else
 2683                         if (ifr->ifr_addr.sa_len == 0)
 2684                                 ifr->ifr_addr.sa_len = 16;
 2685 #endif
 2686                         break;
 2687 
 2688                 case OSIOCGIFADDR:
 2689                         cmd = SIOCGIFADDR;
 2690                         break;
 2691 
 2692                 case OSIOCGIFDSTADDR:
 2693                         cmd = SIOCGIFDSTADDR;
 2694                         break;
 2695 
 2696                 case OSIOCGIFBRDADDR:
 2697                         cmd = SIOCGIFBRDADDR;
 2698                         break;
 2699 
 2700                 case OSIOCGIFNETMASK:
 2701                         cmd = SIOCGIFNETMASK;
 2702                 }
 2703                 error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
 2704                                                                    cmd,
 2705                                                                    data,
 2706                                                                    ifp, td));
 2707                 if (error == EOPNOTSUPP && ifp != NULL &&
 2708                     ifp->if_ioctl != NULL &&
 2709                     cmd != SIOCSIFADDR && cmd != SIOCSIFBRDADDR &&
 2710                     cmd != SIOCSIFDSTADDR && cmd != SIOCSIFNETMASK)
 2711                         error = (*ifp->if_ioctl)(ifp, cmd, data);
 2712                 switch (ocmd) {
 2713 
 2714                 case OSIOCGIFADDR:
 2715                 case OSIOCGIFDSTADDR:
 2716                 case OSIOCGIFBRDADDR:
 2717                 case OSIOCGIFNETMASK:
 2718                         *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
 2719 
 2720                 }
 2721         }
 2722 #endif /* COMPAT_43 */
 2723 
 2724         if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
 2725 #ifdef INET6
 2726                 if (ifp->if_flags & IFF_UP) {
 2727                         int s = splimp();
 2728                         in6_if_up(ifp);
 2729                         splx(s);
 2730                 }
 2731 #endif
 2732         }
 2733         if_rele(ifp);
 2734         CURVNET_RESTORE();
 2735         return (error);
 2736 }
 2737 
 2738 /*
 2739  * The code common to handling reference counted flags,
 2740  * e.g., in ifpromisc() and if_allmulti().
 2741  * The "pflag" argument can specify a permanent mode flag to check,
 2742  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
 2743  *
 2744  * Only to be used on stack-owned flags, not driver-owned flags.
 2745  */
 2746 static int
 2747 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
 2748 {
 2749         struct ifreq ifr;
 2750         int error;
 2751         int oldflags, oldcount;
 2752 
 2753         /* Sanity checks to catch programming errors */
 2754         KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
 2755             ("%s: setting driver-owned flag %d", __func__, flag));
 2756 
 2757         if (onswitch)
 2758                 KASSERT(*refcount >= 0,
 2759                     ("%s: increment negative refcount %d for flag %d",
 2760                     __func__, *refcount, flag));
 2761         else
 2762                 KASSERT(*refcount > 0,
 2763                     ("%s: decrement non-positive refcount %d for flag %d",
 2764                     __func__, *refcount, flag));
 2765 
 2766         /* In case this mode is permanent, just touch refcount */
 2767         if (ifp->if_flags & pflag) {
 2768                 *refcount += onswitch ? 1 : -1;
 2769                 return (0);
 2770         }
 2771 
 2772         /* Save ifnet parameters for if_ioctl() may fail */
 2773         oldcount = *refcount;
 2774         oldflags = ifp->if_flags;
 2775         
 2776         /*
 2777          * See if we aren't the only and touching refcount is enough.
 2778          * Actually toggle interface flag if we are the first or last.
 2779          */
 2780         if (onswitch) {
 2781                 if ((*refcount)++)
 2782                         return (0);
 2783                 ifp->if_flags |= flag;
 2784         } else {
 2785                 if (--(*refcount))
 2786                         return (0);
 2787                 ifp->if_flags &= ~flag;
 2788         }
 2789 
 2790         /* Call down the driver since we've changed interface flags */
 2791         if (ifp->if_ioctl == NULL) {
 2792                 error = EOPNOTSUPP;
 2793                 goto recover;
 2794         }
 2795         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2796         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2797         error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2798         if (error)
 2799                 goto recover;
 2800         /* Notify userland that interface flags have changed */
 2801         rt_ifmsg(ifp);
 2802         return (0);
 2803 
 2804 recover:
 2805         /* Recover after driver error */
 2806         *refcount = oldcount;
 2807         ifp->if_flags = oldflags;
 2808         return (error);
 2809 }
 2810 
 2811 /*
 2812  * Set/clear promiscuous mode on interface ifp based on the truth value
 2813  * of pswitch.  The calls are reference counted so that only the first
 2814  * "on" request actually has an effect, as does the final "off" request.
 2815  * Results are undefined if the "off" and "on" requests are not matched.
 2816  */
 2817 int
 2818 ifpromisc(struct ifnet *ifp, int pswitch)
 2819 {
 2820         int error;
 2821         int oldflags = ifp->if_flags;
 2822 
 2823         error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
 2824                            &ifp->if_pcount, pswitch);
 2825         /* If promiscuous mode status has changed, log a message */
 2826         if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
 2827                 log(LOG_INFO, "%s: promiscuous mode %s\n",
 2828                     ifp->if_xname,
 2829                     (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
 2830         return (error);
 2831 }
 2832 
 2833 /*
 2834  * Return interface configuration
 2835  * of system.  List may be used
 2836  * in later ioctl's (above) to get
 2837  * other information.
 2838  */
 2839 /*ARGSUSED*/
 2840 static int
 2841 ifconf(u_long cmd, caddr_t data)
 2842 {
 2843         struct ifconf *ifc = (struct ifconf *)data;
 2844         struct ifnet *ifp;
 2845         struct ifaddr *ifa;
 2846         struct ifreq ifr;
 2847         struct sbuf *sb;
 2848         int error, full = 0, valid_len, max_len;
 2849 
 2850         /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
 2851         max_len = MAXPHYS - 1;
 2852 
 2853         /* Prevent hostile input from being able to crash the system */
 2854         if (ifc->ifc_len <= 0)
 2855                 return (EINVAL);
 2856 
 2857 again:
 2858         if (ifc->ifc_len <= max_len) {
 2859                 max_len = ifc->ifc_len;
 2860                 full = 1;
 2861         }
 2862         sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
 2863         max_len = 0;
 2864         valid_len = 0;
 2865 
 2866         IFNET_RLOCK();
 2867         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2868                 int addrs;
 2869 
 2870                 /*
 2871                  * Zero the ifr_name buffer to make sure we don't
 2872                  * disclose the contents of the stack.
 2873                  */
 2874                 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
 2875 
 2876                 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
 2877                     >= sizeof(ifr.ifr_name)) {
 2878                         sbuf_delete(sb);
 2879                         IFNET_RUNLOCK();
 2880                         return (ENAMETOOLONG);
 2881                 }
 2882 
 2883                 addrs = 0;
 2884                 IF_ADDR_RLOCK(ifp);
 2885                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2886                         struct sockaddr *sa = ifa->ifa_addr;
 2887 
 2888                         if (prison_if(curthread->td_ucred, sa) != 0)
 2889                                 continue;
 2890                         addrs++;
 2891 #ifdef COMPAT_43
 2892                         if (cmd == OSIOCGIFCONF) {
 2893                                 struct osockaddr *osa =
 2894                                          (struct osockaddr *)&ifr.ifr_addr;
 2895                                 ifr.ifr_addr = *sa;
 2896                                 osa->sa_family = sa->sa_family;
 2897                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2898                                 max_len += sizeof(ifr);
 2899                         } else
 2900 #endif
 2901                         if (sa->sa_len <= sizeof(*sa)) {
 2902                                 ifr.ifr_addr = *sa;
 2903                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2904                                 max_len += sizeof(ifr);
 2905                         } else {
 2906                                 sbuf_bcat(sb, &ifr,
 2907                                     offsetof(struct ifreq, ifr_addr));
 2908                                 max_len += offsetof(struct ifreq, ifr_addr);
 2909                                 sbuf_bcat(sb, sa, sa->sa_len);
 2910                                 max_len += sa->sa_len;
 2911                         }
 2912 
 2913                         if (sbuf_error(sb) == 0)
 2914                                 valid_len = sbuf_len(sb);
 2915                 }
 2916                 IF_ADDR_RUNLOCK(ifp);
 2917                 if (addrs == 0) {
 2918                         bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
 2919                         sbuf_bcat(sb, &ifr, sizeof(ifr));
 2920                         max_len += sizeof(ifr);
 2921 
 2922                         if (sbuf_error(sb) == 0)
 2923                                 valid_len = sbuf_len(sb);
 2924                 }
 2925         }
 2926         IFNET_RUNLOCK();
 2927 
 2928         /*
 2929          * If we didn't allocate enough space (uncommon), try again.  If
 2930          * we have already allocated as much space as we are allowed,
 2931          * return what we've got.
 2932          */
 2933         if (valid_len != max_len && !full) {
 2934                 sbuf_delete(sb);
 2935                 goto again;
 2936         }
 2937 
 2938         ifc->ifc_len = valid_len;
 2939         sbuf_finish(sb);
 2940         error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
 2941         sbuf_delete(sb);
 2942         return (error);
 2943 }
 2944 
 2945 /*
 2946  * Just like ifpromisc(), but for all-multicast-reception mode.
 2947  */
 2948 int
 2949 if_allmulti(struct ifnet *ifp, int onswitch)
 2950 {
 2951 
 2952         return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
 2953 }
 2954 
 2955 struct ifmultiaddr *
 2956 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
 2957 {
 2958         struct ifmultiaddr *ifma;
 2959 
 2960         IF_ADDR_LOCK_ASSERT(ifp);
 2961 
 2962         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
 2963                 if (sa->sa_family == AF_LINK) {
 2964                         if (sa_dl_equal(ifma->ifma_addr, sa))
 2965                                 break;
 2966                 } else {
 2967                         if (sa_equal(ifma->ifma_addr, sa))
 2968                                 break;
 2969                 }
 2970         }
 2971 
 2972         return ifma;
 2973 }
 2974 
 2975 /*
 2976  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
 2977  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
 2978  * the ifnet multicast address list here, so the caller must do that and
 2979  * other setup work (such as notifying the device driver).  The reference
 2980  * count is initialized to 1.
 2981  */
 2982 static struct ifmultiaddr *
 2983 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
 2984     int mflags)
 2985 {
 2986         struct ifmultiaddr *ifma;
 2987         struct sockaddr *dupsa;
 2988 
 2989         ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
 2990             M_ZERO);
 2991         if (ifma == NULL)
 2992                 return (NULL);
 2993 
 2994         dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
 2995         if (dupsa == NULL) {
 2996                 free(ifma, M_IFMADDR);
 2997                 return (NULL);
 2998         }
 2999         bcopy(sa, dupsa, sa->sa_len);
 3000         ifma->ifma_addr = dupsa;
 3001 
 3002         ifma->ifma_ifp = ifp;
 3003         ifma->ifma_refcount = 1;
 3004         ifma->ifma_protospec = NULL;
 3005 
 3006         if (llsa == NULL) {
 3007                 ifma->ifma_lladdr = NULL;
 3008                 return (ifma);
 3009         }
 3010 
 3011         dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
 3012         if (dupsa == NULL) {
 3013                 free(ifma->ifma_addr, M_IFMADDR);
 3014                 free(ifma, M_IFMADDR);
 3015                 return (NULL);
 3016         }
 3017         bcopy(llsa, dupsa, llsa->sa_len);
 3018         ifma->ifma_lladdr = dupsa;
 3019 
 3020         return (ifma);
 3021 }
 3022 
 3023 /*
 3024  * if_freemulti: free ifmultiaddr structure and possibly attached related
 3025  * addresses.  The caller is responsible for implementing reference
 3026  * counting, notifying the driver, handling routing messages, and releasing
 3027  * any dependent link layer state.
 3028  */
 3029 static void
 3030 if_freemulti(struct ifmultiaddr *ifma)
 3031 {
 3032 
 3033         KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
 3034             ifma->ifma_refcount));
 3035         KASSERT(ifma->ifma_protospec == NULL,
 3036             ("if_freemulti: protospec not NULL"));
 3037 
 3038         if (ifma->ifma_lladdr != NULL)
 3039                 free(ifma->ifma_lladdr, M_IFMADDR);
 3040         free(ifma->ifma_addr, M_IFMADDR);
 3041         free(ifma, M_IFMADDR);
 3042 }
 3043 
 3044 /*
 3045  * Register an additional multicast address with a network interface.
 3046  *
 3047  * - If the address is already present, bump the reference count on the
 3048  *   address and return.
 3049  * - If the address is not link-layer, look up a link layer address.
 3050  * - Allocate address structures for one or both addresses, and attach to the
 3051  *   multicast address list on the interface.  If automatically adding a link
 3052  *   layer address, the protocol address will own a reference to the link
 3053  *   layer address, to be freed when it is freed.
 3054  * - Notify the network device driver of an addition to the multicast address
 3055  *   list.
 3056  *
 3057  * 'sa' points to caller-owned memory with the desired multicast address.
 3058  *
 3059  * 'retifma' will be used to return a pointer to the resulting multicast
 3060  * address reference, if desired.
 3061  */
 3062 int
 3063 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
 3064     struct ifmultiaddr **retifma)
 3065 {
 3066         struct ifmultiaddr *ifma, *ll_ifma;
 3067         struct sockaddr *llsa;
 3068         int error;
 3069 
 3070         /*
 3071          * If the address is already present, return a new reference to it;
 3072          * otherwise, allocate storage and set up a new address.
 3073          */
 3074         IF_ADDR_WLOCK(ifp);
 3075         ifma = if_findmulti(ifp, sa);
 3076         if (ifma != NULL) {
 3077                 ifma->ifma_refcount++;
 3078                 if (retifma != NULL)
 3079                         *retifma = ifma;
 3080                 IF_ADDR_WUNLOCK(ifp);
 3081                 return (0);
 3082         }
 3083 
 3084         /*
 3085          * The address isn't already present; resolve the protocol address
 3086          * into a link layer address, and then look that up, bump its
 3087          * refcount or allocate an ifma for that also.  If 'llsa' was
 3088          * returned, we will need to free it later.
 3089          */
 3090         llsa = NULL;
 3091         ll_ifma = NULL;
 3092         if (ifp->if_resolvemulti != NULL) {
 3093                 error = ifp->if_resolvemulti(ifp, &llsa, sa);
 3094                 if (error)
 3095                         goto unlock_out;
 3096         }
 3097 
 3098         /*
 3099          * Allocate the new address.  Don't hook it up yet, as we may also
 3100          * need to allocate a link layer multicast address.
 3101          */
 3102         ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
 3103         if (ifma == NULL) {
 3104                 error = ENOMEM;
 3105                 goto free_llsa_out;
 3106         }
 3107 
 3108         /*
 3109          * If a link layer address is found, we'll need to see if it's
 3110          * already present in the address list, or allocate is as well.
 3111          * When this block finishes, the link layer address will be on the
 3112          * list.
 3113          */
 3114         if (llsa != NULL) {
 3115                 ll_ifma = if_findmulti(ifp, llsa);
 3116                 if (ll_ifma == NULL) {
 3117                         ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
 3118                         if (ll_ifma == NULL) {
 3119                                 --ifma->ifma_refcount;
 3120                                 if_freemulti(ifma);
 3121                                 error = ENOMEM;
 3122                                 goto free_llsa_out;
 3123                         }
 3124                         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
 3125                             ifma_link);
 3126                 } else
 3127                         ll_ifma->ifma_refcount++;
 3128                 ifma->ifma_llifma = ll_ifma;
 3129         }
 3130 
 3131         /*
 3132          * We now have a new multicast address, ifma, and possibly a new or
 3133          * referenced link layer address.  Add the primary address to the
 3134          * ifnet address list.
 3135          */
 3136         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
 3137 
 3138         if (retifma != NULL)
 3139                 *retifma = ifma;
 3140 
 3141         /*
 3142          * Must generate the message while holding the lock so that 'ifma'
 3143          * pointer is still valid.
 3144          */
 3145         rt_newmaddrmsg(RTM_NEWMADDR, ifma);
 3146         IF_ADDR_WUNLOCK(ifp);
 3147 
 3148         /*
 3149          * We are certain we have added something, so call down to the
 3150          * interface to let them know about it.
 3151          */
 3152         if (ifp->if_ioctl != NULL) {
 3153                 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
 3154         }
 3155 
 3156         if (llsa != NULL)
 3157                 free(llsa, M_IFMADDR);
 3158 
 3159         return (0);
 3160 
 3161 free_llsa_out:
 3162         if (llsa != NULL)
 3163                 free(llsa, M_IFMADDR);
 3164 
 3165 unlock_out:
 3166         IF_ADDR_WUNLOCK(ifp);
 3167         return (error);
 3168 }
 3169 
 3170 /*
 3171  * Delete a multicast group membership by network-layer group address.
 3172  *
 3173  * Returns ENOENT if the entry could not be found. If ifp no longer
 3174  * exists, results are undefined. This entry point should only be used
 3175  * from subsystems which do appropriate locking to hold ifp for the
 3176  * duration of the call.
 3177  * Network-layer protocol domains must use if_delmulti_ifma().
 3178  */
 3179 int
 3180 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
 3181 {
 3182         struct ifmultiaddr *ifma;
 3183         int lastref;
 3184 #ifdef INVARIANTS
 3185         struct ifnet *oifp;
 3186 
 3187         IFNET_RLOCK_NOSLEEP();
 3188         TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3189                 if (ifp == oifp)
 3190                         break;
 3191         if (ifp != oifp)
 3192                 ifp = NULL;
 3193         IFNET_RUNLOCK_NOSLEEP();
 3194 
 3195         KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
 3196 #endif
 3197         if (ifp == NULL)
 3198                 return (ENOENT);
 3199 
 3200         IF_ADDR_WLOCK(ifp);
 3201         lastref = 0;
 3202         ifma = if_findmulti(ifp, sa);
 3203         if (ifma != NULL)
 3204                 lastref = if_delmulti_locked(ifp, ifma, 0);
 3205         IF_ADDR_WUNLOCK(ifp);
 3206 
 3207         if (ifma == NULL)
 3208                 return (ENOENT);
 3209 
 3210         if (lastref && ifp->if_ioctl != NULL) {
 3211                 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3212         }
 3213 
 3214         return (0);
 3215 }
 3216 
 3217 /*
 3218  * Delete all multicast group membership for an interface.
 3219  * Should be used to quickly flush all multicast filters.
 3220  */
 3221 void
 3222 if_delallmulti(struct ifnet *ifp)
 3223 {
 3224         struct ifmultiaddr *ifma;
 3225         struct ifmultiaddr *next;
 3226 
 3227         IF_ADDR_WLOCK(ifp);
 3228         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
 3229                 if_delmulti_locked(ifp, ifma, 0);
 3230         IF_ADDR_WUNLOCK(ifp);
 3231 }
 3232 
 3233 /*
 3234  * Delete a multicast group membership by group membership pointer.
 3235  * Network-layer protocol domains must use this routine.
 3236  *
 3237  * It is safe to call this routine if the ifp disappeared.
 3238  */
 3239 void
 3240 if_delmulti_ifma(struct ifmultiaddr *ifma)
 3241 {
 3242         struct ifnet *ifp;
 3243         int lastref;
 3244 
 3245         ifp = ifma->ifma_ifp;
 3246 #ifdef DIAGNOSTIC
 3247         if (ifp == NULL) {
 3248                 printf("%s: ifma_ifp seems to be detached\n", __func__);
 3249         } else {
 3250                 struct ifnet *oifp;
 3251 
 3252                 IFNET_RLOCK_NOSLEEP();
 3253                 TAILQ_FOREACH(oifp, &V_ifnet, if_link)
 3254                         if (ifp == oifp)
 3255                                 break;
 3256                 if (ifp != oifp) {
 3257                         printf("%s: ifnet %p disappeared\n", __func__, ifp);
 3258                         ifp = NULL;
 3259                 }
 3260                 IFNET_RUNLOCK_NOSLEEP();
 3261         }
 3262 #endif
 3263         /*
 3264          * If and only if the ifnet instance exists: Acquire the address lock.
 3265          */
 3266         if (ifp != NULL)
 3267                 IF_ADDR_WLOCK(ifp);
 3268 
 3269         lastref = if_delmulti_locked(ifp, ifma, 0);
 3270 
 3271         if (ifp != NULL) {
 3272                 /*
 3273                  * If and only if the ifnet instance exists:
 3274                  *  Release the address lock.
 3275                  *  If the group was left: update the hardware hash filter.
 3276                  */
 3277                 IF_ADDR_WUNLOCK(ifp);
 3278                 if (lastref && ifp->if_ioctl != NULL) {
 3279                         (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 3280                 }
 3281         }
 3282 }
 3283 
 3284 /*
 3285  * Perform deletion of network-layer and/or link-layer multicast address.
 3286  *
 3287  * Return 0 if the reference count was decremented.
 3288  * Return 1 if the final reference was released, indicating that the
 3289  * hardware hash filter should be reprogrammed.
 3290  */
 3291 static int
 3292 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
 3293 {
 3294         struct ifmultiaddr *ll_ifma;
 3295 
 3296         if (ifp != NULL && ifma->ifma_ifp != NULL) {
 3297                 KASSERT(ifma->ifma_ifp == ifp,
 3298                     ("%s: inconsistent ifp %p", __func__, ifp));
 3299                 IF_ADDR_WLOCK_ASSERT(ifp);
 3300         }
 3301 
 3302         ifp = ifma->ifma_ifp;
 3303 
 3304         /*
 3305          * If the ifnet is detaching, null out references to ifnet,
 3306          * so that upper protocol layers will notice, and not attempt
 3307          * to obtain locks for an ifnet which no longer exists. The
 3308          * routing socket announcement must happen before the ifnet
 3309          * instance is detached from the system.
 3310          */
 3311         if (detaching) {
 3312 #ifdef DIAGNOSTIC
 3313                 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
 3314 #endif
 3315                 /*
 3316                  * ifp may already be nulled out if we are being reentered
 3317                  * to delete the ll_ifma.
 3318                  */
 3319                 if (ifp != NULL) {
 3320                         rt_newmaddrmsg(RTM_DELMADDR, ifma);
 3321                         ifma->ifma_ifp = NULL;
 3322                 }
 3323         }
 3324 
 3325         if (--ifma->ifma_refcount > 0)
 3326                 return 0;
 3327 
 3328         /*
 3329          * If this ifma is a network-layer ifma, a link-layer ifma may
 3330          * have been associated with it. Release it first if so.
 3331          */
 3332         ll_ifma = ifma->ifma_llifma;
 3333         if (ll_ifma != NULL) {
 3334                 KASSERT(ifma->ifma_lladdr != NULL,
 3335                     ("%s: llifma w/o lladdr", __func__));
 3336                 if (detaching)
 3337                         ll_ifma->ifma_ifp = NULL;       /* XXX */
 3338                 if (--ll_ifma->ifma_refcount == 0) {
 3339                         if (ifp != NULL) {
 3340                                 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
 3341                                     ifma_link);
 3342                         }
 3343                         if_freemulti(ll_ifma);
 3344                 }
 3345         }
 3346 
 3347         if (ifp != NULL)
 3348                 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
 3349 
 3350         if_freemulti(ifma);
 3351 
 3352         /*
 3353          * The last reference to this instance of struct ifmultiaddr
 3354          * was released; the hardware should be notified of this change.
 3355          */
 3356         return 1;
 3357 }
 3358 
 3359 /*
 3360  * Set the link layer address on an interface.
 3361  *
 3362  * At this time we only support certain types of interfaces,
 3363  * and we don't allow the length of the address to change.
 3364  */
 3365 int
 3366 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
 3367 {
 3368         struct sockaddr_dl *sdl;
 3369         struct ifaddr *ifa;
 3370         struct ifreq ifr;
 3371 
 3372         IF_ADDR_RLOCK(ifp);
 3373         ifa = ifp->if_addr;
 3374         if (ifa == NULL) {
 3375                 IF_ADDR_RUNLOCK(ifp);
 3376                 return (EINVAL);
 3377         }
 3378         ifa_ref(ifa);
 3379         IF_ADDR_RUNLOCK(ifp);
 3380         sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 3381         if (sdl == NULL) {
 3382                 ifa_free(ifa);
 3383                 return (EINVAL);
 3384         }
 3385         if (len != sdl->sdl_alen) {     /* don't allow length to change */
 3386                 ifa_free(ifa);
 3387                 return (EINVAL);
 3388         }
 3389         switch (ifp->if_type) {
 3390         case IFT_ETHER:
 3391         case IFT_FDDI:
 3392         case IFT_XETHER:
 3393         case IFT_ISO88025:
 3394         case IFT_L2VLAN:
 3395         case IFT_BRIDGE:
 3396         case IFT_ARCNET:
 3397         case IFT_IEEE8023ADLAG:
 3398         case IFT_IEEE80211:
 3399                 bcopy(lladdr, LLADDR(sdl), len);
 3400                 ifa_free(ifa);
 3401                 break;
 3402         default:
 3403                 ifa_free(ifa);
 3404                 return (ENODEV);
 3405         }
 3406 
 3407         /*
 3408          * If the interface is already up, we need
 3409          * to re-init it in order to reprogram its
 3410          * address filter.
 3411          */
 3412         if ((ifp->if_flags & IFF_UP) != 0) {
 3413                 if (ifp->if_ioctl) {
 3414                         ifp->if_flags &= ~IFF_UP;
 3415                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3416                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3417                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3418                         ifp->if_flags |= IFF_UP;
 3419                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 3420                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 3421                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 3422                 }
 3423 #ifdef INET
 3424                 /*
 3425                  * Also send gratuitous ARPs to notify other nodes about
 3426                  * the address change.
 3427                  */
 3428                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 3429                         if (ifa->ifa_addr->sa_family == AF_INET)
 3430                                 arp_ifinit(ifp, ifa);
 3431                 }
 3432 #endif
 3433         }
 3434         return (0);
 3435 }
 3436 
 3437 /*
 3438  * The name argument must be a pointer to storage which will last as
 3439  * long as the interface does.  For physical devices, the result of
 3440  * device_get_name(dev) is a good choice and for pseudo-devices a
 3441  * static string works well.
 3442  */
 3443 void
 3444 if_initname(struct ifnet *ifp, const char *name, int unit)
 3445 {
 3446         ifp->if_dname = name;
 3447         ifp->if_dunit = unit;
 3448         if (unit != IF_DUNIT_NONE)
 3449                 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
 3450         else
 3451                 strlcpy(ifp->if_xname, name, IFNAMSIZ);
 3452 }
 3453 
 3454 int
 3455 if_printf(struct ifnet *ifp, const char * fmt, ...)
 3456 {
 3457         va_list ap;
 3458         int retval;
 3459 
 3460         retval = printf("%s: ", ifp->if_xname);
 3461         va_start(ap, fmt);
 3462         retval += vprintf(fmt, ap);
 3463         va_end(ap);
 3464         return (retval);
 3465 }
 3466 
 3467 void
 3468 if_start(struct ifnet *ifp)
 3469 {
 3470 
 3471         (*(ifp)->if_start)(ifp);
 3472 }
 3473 
 3474 /*
 3475  * Backwards compatibility interface for drivers 
 3476  * that have not implemented it
 3477  */
 3478 static int
 3479 if_transmit(struct ifnet *ifp, struct mbuf *m)
 3480 {
 3481         int error;
 3482 
 3483         IFQ_HANDOFF(ifp, m, error);
 3484         return (error);
 3485 }
 3486 
 3487 static void
 3488 if_input_default(struct ifnet *ifp __unused, struct mbuf *m)
 3489 {
 3490 
 3491         m_freem(m);
 3492 }
 3493 
 3494 int
 3495 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
 3496 {
 3497         int active = 0;
 3498 
 3499         IF_LOCK(ifq);
 3500         if (_IF_QFULL(ifq)) {
 3501                 _IF_DROP(ifq);
 3502                 IF_UNLOCK(ifq);
 3503                 m_freem(m);
 3504                 return (0);
 3505         }
 3506         if (ifp != NULL) {
 3507                 ifp->if_obytes += m->m_pkthdr.len + adjust;
 3508                 if (m->m_flags & (M_BCAST|M_MCAST))
 3509                         ifp->if_omcasts++;
 3510                 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
 3511         }
 3512         _IF_ENQUEUE(ifq, m);
 3513         IF_UNLOCK(ifq);
 3514         if (ifp != NULL && !active)
 3515                 (*(ifp)->if_start)(ifp);
 3516         return (1);
 3517 }
 3518 
 3519 void
 3520 if_register_com_alloc(u_char type,
 3521     if_com_alloc_t *a, if_com_free_t *f)
 3522 {
 3523         
 3524         KASSERT(if_com_alloc[type] == NULL,
 3525             ("if_register_com_alloc: %d already registered", type));
 3526         KASSERT(if_com_free[type] == NULL,
 3527             ("if_register_com_alloc: %d free already registered", type));
 3528 
 3529         if_com_alloc[type] = a;
 3530         if_com_free[type] = f;
 3531 }
 3532 
 3533 void
 3534 if_deregister_com_alloc(u_char type)
 3535 {
 3536         
 3537         KASSERT(if_com_alloc[type] != NULL,
 3538             ("if_deregister_com_alloc: %d not registered", type));
 3539         KASSERT(if_com_free[type] != NULL,
 3540             ("if_deregister_com_alloc: %d free not registered", type));
 3541         if_com_alloc[type] = NULL;
 3542         if_com_free[type] = NULL;
 3543 }

Cache object: 1e99f4cb4f5dd51d565605beedc10602


[ 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.