The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/net/if.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1980, 1986, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 4. Neither the name of the University nor the names of its contributors
   14  *    may be used to endorse or promote products derived from this software
   15  *    without specific prior written permission.
   16  *
   17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  *
   29  *      @(#)if.c        8.5 (Berkeley) 1/9/95
   30  * $FreeBSD: releng/7.4/sys/net/if.c 215368 2010-11-16 04:40:03Z sobomax $
   31  */
   32 
   33 #include "opt_compat.h"
   34 #include "opt_inet6.h"
   35 #include "opt_inet.h"
   36 #include "opt_mac.h"
   37 #include "opt_carp.h"
   38 
   39 #include <sys/param.h>
   40 #include <sys/types.h>
   41 #include <sys/conf.h>
   42 #include <sys/malloc.h>
   43 #include <sys/sbuf.h>
   44 #include <sys/bus.h>
   45 #include <sys/mbuf.h>
   46 #include <sys/systm.h>
   47 #include <sys/priv.h>
   48 #include <sys/proc.h>
   49 #include <sys/socket.h>
   50 #include <sys/socketvar.h>
   51 #include <sys/protosw.h>
   52 #include <sys/kernel.h>
   53 #include <sys/sockio.h>
   54 #include <sys/syslog.h>
   55 #include <sys/sysctl.h>
   56 #include <sys/taskqueue.h>
   57 #include <sys/domain.h>
   58 #include <sys/jail.h>
   59 #include <machine/stdarg.h>
   60 
   61 #include <net/if.h>
   62 #include <net/if_clone.h>
   63 #include <net/if_dl.h>
   64 #include <net/if_types.h>
   65 #include <net/if_var.h>
   66 #include <net/radix.h>
   67 #include <net/route.h>
   68 
   69 #if defined(INET) || defined(INET6)
   70 /*XXX*/
   71 #include <netinet/in.h>
   72 #include <netinet/in_var.h>
   73 #ifdef INET6
   74 #include <netinet6/in6_var.h>
   75 #include <netinet6/in6_ifattach.h>
   76 #endif
   77 #endif
   78 #ifdef INET
   79 #include <netinet/if_ether.h>
   80 #endif
   81 #ifdef DEV_CARP
   82 #include <netinet/ip_carp.h>
   83 #endif
   84 
   85 #include <security/mac/mac_framework.h>
   86 
   87 static int slowtimo_started;
   88 
   89 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
   90 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
   91 
   92 TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
   93 SYSCTL_UINT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
   94     &ifqmaxlen, 0, "max send queue size");
   95 
   96 /* Log link state change events */
   97 static int log_link_state_change = 1;
   98 
   99 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
  100         &log_link_state_change, 0,
  101         "log interface link state change events");
  102 
  103 void    (*bstp_linkstate_p)(struct ifnet *ifp, int state);
  104 void    (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
  105 void    (*lagg_linkstate_p)(struct ifnet *ifp, int state);
  106 
  107 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
  108 
  109 /*
  110  * XXX: Style; these should be sorted alphabetically, and unprototyped
  111  * static functions should be prototyped. Currently they are sorted by
  112  * declaration order.
  113  */
  114 static void     if_attachdomain(void *);
  115 static void     if_attachdomain1(struct ifnet *);
  116 static void     if_purgemaddrs(struct ifnet *);
  117 static int      ifconf(u_long, caddr_t);
  118 static void     if_freemulti(struct ifmultiaddr *);
  119 static void     if_grow(void);
  120 static void     if_init(void *);
  121 static void     if_check(void *);
  122 static void     if_qflush(struct ifaltq *);
  123 static void     if_route(struct ifnet *, int flag, int fam);
  124 static int      if_setflag(struct ifnet *, int, int, int *, int);
  125 static void     if_slowtimo(void *);
  126 static void     if_unroute(struct ifnet *, int flag, int fam);
  127 static void     link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  128 static int      if_rtdel(struct radix_node *, void *);
  129 static int      ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
  130 static int      if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
  131 static void     if_start_deferred(void *context, int pending);
  132 static void     do_link_state_change(void *, int);
  133 static int      if_getgroup(struct ifgroupreq *, struct ifnet *);
  134 static int      if_getgroupmembers(struct ifgroupreq *);
  135 static void     if_delgroups(struct ifnet *);
  136 #ifdef INET6
  137 /*
  138  * XXX: declare here to avoid to include many inet6 related files..
  139  * should be more generalized?
  140  */
  141 extern void     nd6_setmtu(struct ifnet *);
  142 #endif
  143 
  144 int     if_index = 0;
  145 int     ifqmaxlen = IFQ_MAXLEN;
  146 struct  ifnethead ifnet;        /* depend on static init XXX */
  147 struct  ifgrouphead ifg_head;
  148 struct  mtx ifnet_lock;
  149 static  if_com_alloc_t *if_com_alloc[256];
  150 static  if_com_free_t *if_com_free[256];
  151 
  152 static int      if_indexlim = 8;
  153 static struct   knlist ifklist;
  154 
  155 /*
  156  * Table of ifnet/cdev by index.  Locked with ifnet_lock.
  157  */
  158 static struct ifindex_entry *ifindex_table = NULL;
  159 
  160 static void     filt_netdetach(struct knote *kn);
  161 static int      filt_netdev(struct knote *kn, long hint);
  162 
  163 static struct filterops netdev_filtops =
  164     { 1, NULL, filt_netdetach, filt_netdev };
  165 
  166 /*
  167  * System initialization
  168  */
  169 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
  170 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL);
  171 
  172 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
  173 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
  174 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
  175 
  176 struct ifnet *
  177 ifnet_byindex(u_short idx)
  178 {
  179         struct ifnet *ifp;
  180 
  181         IFNET_RLOCK();
  182         ifp = ifindex_table[idx].ife_ifnet;
  183         IFNET_RUNLOCK();
  184         return (ifp);
  185 }
  186 
  187 static void
  188 ifnet_setbyindex(u_short idx, struct ifnet *ifp)
  189 {
  190 
  191         IFNET_WLOCK_ASSERT();
  192 
  193         ifindex_table[idx].ife_ifnet = ifp;
  194 }
  195 
  196 struct ifaddr *
  197 ifaddr_byindex(u_short idx)
  198 {
  199         struct ifaddr *ifa;
  200 
  201         IFNET_RLOCK();
  202         ifa = ifnet_byindex(idx)->if_addr;
  203         IFNET_RUNLOCK();
  204         return (ifa);
  205 }
  206 
  207 struct cdev *
  208 ifdev_byindex(u_short idx)
  209 {
  210         struct cdev *cdev;
  211 
  212         IFNET_RLOCK();
  213         cdev = ifindex_table[idx].ife_dev;
  214         IFNET_RUNLOCK();
  215         return (cdev);
  216 }
  217 
  218 static void
  219 ifdev_setbyindex(u_short idx, struct cdev *cdev)
  220 {
  221 
  222         IFNET_WLOCK();
  223         ifindex_table[idx].ife_dev = cdev;
  224         IFNET_WUNLOCK();
  225 }
  226 
  227 static d_open_t         netopen;
  228 static d_close_t        netclose;
  229 static d_ioctl_t        netioctl;
  230 static d_kqfilter_t     netkqfilter;
  231 
  232 static struct cdevsw net_cdevsw = {
  233         .d_version =    D_VERSION,
  234         .d_flags =      D_NEEDGIANT,
  235         .d_open =       netopen,
  236         .d_close =      netclose,
  237         .d_ioctl =      netioctl,
  238         .d_name =       "net",
  239         .d_kqfilter =   netkqfilter,
  240 };
  241 
  242 static int
  243 netopen(struct cdev *dev, int flag, int mode, struct thread *td)
  244 {
  245         return (0);
  246 }
  247 
  248 static int
  249 netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
  250 {
  251         return (0);
  252 }
  253 
  254 static int
  255 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
  256 {
  257         struct ifnet *ifp;
  258         int error, idx;
  259 
  260         /* only support interface specific ioctls */
  261         if (IOCGROUP(cmd) != 'i')
  262                 return (EOPNOTSUPP);
  263         idx = minor(dev);
  264         if (idx == 0) {
  265                 /*
  266                  * special network device, not interface.
  267                  */
  268                 if (cmd == SIOCGIFCONF)
  269                         return (ifconf(cmd, data));     /* XXX remove cmd */
  270 #ifdef __amd64__
  271                 if (cmd == SIOCGIFCONF32)
  272                         return (ifconf(cmd, data));     /* XXX remove cmd */
  273 #endif
  274                 return (EOPNOTSUPP);
  275         }
  276 
  277         ifp = ifnet_byindex(idx);
  278         if (ifp == NULL)
  279                 return (ENXIO);
  280 
  281         error = ifhwioctl(cmd, ifp, data, td);
  282         if (error == ENOIOCTL)
  283                 error = EOPNOTSUPP;
  284         return (error);
  285 }
  286 
  287 static int
  288 netkqfilter(struct cdev *dev, struct knote *kn)
  289 {
  290         struct knlist *klist;
  291         struct ifnet *ifp;
  292         int idx;
  293 
  294         switch (kn->kn_filter) {
  295         case EVFILT_NETDEV:
  296                 kn->kn_fop = &netdev_filtops;
  297                 break;
  298         default:
  299                 return (EINVAL);
  300         }
  301 
  302         idx = minor(dev);
  303         if (idx == 0) {
  304                 klist = &ifklist;
  305         } else {
  306                 ifp = ifnet_byindex(idx);
  307                 if (ifp == NULL)
  308                         return (1);
  309                 klist = &ifp->if_klist;
  310         }
  311 
  312         kn->kn_hook = (caddr_t)klist;
  313 
  314         knlist_add(klist, kn, 0);
  315 
  316         return (0);
  317 }
  318 
  319 static void
  320 filt_netdetach(struct knote *kn)
  321 {
  322         struct knlist *klist = (struct knlist *)kn->kn_hook;
  323 
  324         knlist_remove(klist, kn, 0);
  325 }
  326 
  327 static int
  328 filt_netdev(struct knote *kn, long hint)
  329 {
  330         struct knlist *klist = (struct knlist *)kn->kn_hook;
  331 
  332         /*
  333          * Currently NOTE_EXIT is abused to indicate device detach.
  334          */
  335         if (hint == NOTE_EXIT) {
  336                 kn->kn_data = NOTE_LINKINV;
  337                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
  338                 knlist_remove_inevent(klist, kn);
  339                 return (1);
  340         }
  341         if (hint != 0)
  342                 kn->kn_data = hint;                     /* current status */
  343         if (kn->kn_sfflags & hint)
  344                 kn->kn_fflags |= hint;
  345         return (kn->kn_fflags != 0);
  346 }
  347 
  348 /*
  349  * Network interface utility routines.
  350  *
  351  * Routines with ifa_ifwith* names take sockaddr *'s as
  352  * parameters.
  353  */
  354 
  355 /* ARGSUSED*/
  356 static void
  357 if_init(void *dummy __unused)
  358 {
  359 
  360         IFNET_LOCK_INIT();
  361         TAILQ_INIT(&ifnet);
  362         TAILQ_INIT(&ifg_head);
  363         knlist_init(&ifklist, NULL, NULL, NULL, NULL);
  364         if_grow();                              /* create initial table */
  365         ifdev_setbyindex(0, make_dev(&net_cdevsw, 0, UID_ROOT, GID_WHEEL,
  366             0600, "network"));
  367         if_clone_init();
  368 }
  369 
  370 static void
  371 if_grow(void)
  372 {
  373         u_int n;
  374         struct ifindex_entry *e;
  375 
  376         if_indexlim <<= 1;
  377         n = if_indexlim * sizeof(*e);
  378         e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
  379         if (ifindex_table != NULL) {
  380                 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
  381                 free((caddr_t)ifindex_table, M_IFNET);
  382         }
  383         ifindex_table = e;
  384 }
  385 
  386 /* ARGSUSED*/
  387 static void
  388 if_check(void *dummy __unused)
  389 {
  390         struct ifnet *ifp;
  391         int s;
  392 
  393         s = splimp();
  394         IFNET_RLOCK();  /* could sleep on rare error; mostly okay XXX */
  395         TAILQ_FOREACH(ifp, &ifnet, if_link) {
  396                 if (ifp->if_snd.ifq_maxlen == 0) {
  397                         if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
  398                         ifp->if_snd.ifq_maxlen = ifqmaxlen;
  399                 }
  400                 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
  401                         if_printf(ifp,
  402                             "XXX: driver didn't initialize queue mtx\n");
  403                         mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
  404                             MTX_NETWORK_LOCK, MTX_DEF);
  405                 }
  406         }
  407         IFNET_RUNLOCK();
  408         splx(s);
  409 
  410         /*
  411          * If at least one interface added during boot uses
  412          * if_watchdog then start the timer.
  413          */
  414         if (slowtimo_started)
  415                 if_slowtimo(0);
  416 }
  417 
  418 /*
  419  * Allocate a struct ifnet and an index for an interface.  A layer 2
  420  * common structure will also be allocated if an allocation routine is
  421  * registered for the passed type.
  422  */
  423 struct ifnet*
  424 if_alloc(u_char type)
  425 {
  426         struct ifnet *ifp;
  427 
  428         ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
  429 
  430         /*
  431          * Try to find an empty slot below if_index.  If we fail, take
  432          * the next slot.
  433          *
  434          * XXX: should be locked!
  435          */
  436         for (ifp->if_index = 1; ifp->if_index <= if_index; ifp->if_index++) {
  437                 if (ifnet_byindex(ifp->if_index) == NULL)
  438                         break;
  439         }
  440         /* Catch if_index overflow. */
  441         if (ifp->if_index < 1) {
  442                 free(ifp, M_IFNET);
  443                 return (NULL);
  444         }
  445         if (ifp->if_index > if_index)
  446                 if_index = ifp->if_index;
  447         if (if_index >= if_indexlim)
  448                 if_grow();
  449 
  450         ifp->if_type = type;
  451 
  452         if (if_com_alloc[type] != NULL) {
  453                 ifp->if_l2com = if_com_alloc[type](type, ifp);
  454                 if (ifp->if_l2com == NULL) {
  455                         free(ifp, M_IFNET);
  456                         return (NULL);
  457                 }
  458         }
  459         IFNET_WLOCK();
  460         ifnet_setbyindex(ifp->if_index, ifp);
  461         IFNET_WUNLOCK();
  462         IF_ADDR_LOCK_INIT(ifp);
  463 
  464         return (ifp);
  465 }
  466 
  467 /*
  468  * Free the struct ifnet, the associated index, and the layer 2 common
  469  * structure if needed.  All the work is done in if_free_type().
  470  *
  471  * Do not add code to this function!  Add it to if_free_type().
  472  */
  473 void
  474 if_free(struct ifnet *ifp)
  475 {
  476 
  477         if_free_type(ifp, ifp->if_type);
  478 }
  479 
  480 /*
  481  * Do the actual work of freeing a struct ifnet, associated index, and
  482  * layer 2 common structure.  This version should only be called by
  483  * intefaces that switch their type after calling if_alloc().
  484  */
  485 void
  486 if_free_type(struct ifnet *ifp, u_char type)
  487 {
  488 
  489         if (ifp != ifnet_byindex(ifp->if_index)) {
  490                 if_printf(ifp, "%s: value was not if_alloced, skipping\n",
  491                     __func__);
  492                 return;
  493         }
  494 
  495         IFNET_WLOCK();
  496         ifnet_setbyindex(ifp->if_index, NULL);
  497 
  498         /* XXX: should be locked with if_findindex() */
  499         while (if_index > 0 && ifnet_byindex(if_index) == NULL)
  500                 if_index--;
  501         IFNET_WUNLOCK();
  502 
  503         if (if_com_free[type] != NULL)
  504                 if_com_free[type](ifp->if_l2com, type);
  505 
  506         IF_ADDR_LOCK_DESTROY(ifp);
  507         free(ifp, M_IFNET);
  508 };
  509 
  510 /*
  511  * Perform generic interface initalization tasks and attach the interface
  512  * to the list of "active" interfaces.
  513  *
  514  * XXX:
  515  *  - The decision to return void and thus require this function to
  516  *    succeed is questionable.
  517  *  - We do more initialization here then is probably a good idea.
  518  *    Some of this should probably move to if_alloc().
  519  *  - We should probably do more sanity checking.  For instance we don't
  520  *    do anything to insure if_xname is unique or non-empty.
  521  */
  522 void
  523 if_attach(struct ifnet *ifp)
  524 {
  525         unsigned socksize, ifasize;
  526         int namelen, masklen;
  527         struct sockaddr_dl *sdl;
  528         struct ifaddr *ifa;
  529 
  530         if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
  531                 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
  532                     ifp->if_xname);
  533 
  534         TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
  535         TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
  536         IF_AFDATA_LOCK_INIT(ifp);
  537         ifp->if_afdata_initialized = 0;
  538 
  539         TAILQ_INIT(&ifp->if_addrhead);
  540         TAILQ_INIT(&ifp->if_prefixhead);
  541         TAILQ_INIT(&ifp->if_multiaddrs);
  542         TAILQ_INIT(&ifp->if_groups);
  543 
  544         if_addgroup(ifp, IFG_ALL);
  545 
  546         knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
  547         getmicrotime(&ifp->if_lastchange);
  548         ifp->if_data.ifi_epoch = time_uptime;
  549         ifp->if_data.ifi_datalen = sizeof(struct if_data);
  550 
  551 #ifdef MAC
  552         mac_init_ifnet(ifp);
  553         mac_create_ifnet(ifp);
  554 #endif
  555 
  556         ifdev_setbyindex(ifp->if_index, make_dev(&net_cdevsw,
  557             unit2minor(ifp->if_index), UID_ROOT, GID_WHEEL, 0600, "%s/%s",
  558             net_cdevsw.d_name, ifp->if_xname));
  559         make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
  560             net_cdevsw.d_name, ifp->if_index);
  561 
  562         mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
  563 
  564         /*
  565          * create a Link Level name for this device
  566          */
  567         namelen = strlen(ifp->if_xname);
  568         /*
  569          * Always save enough space for any possiable name so we can do
  570          * a rename in place later.
  571          */
  572         masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
  573         socksize = masklen + ifp->if_addrlen;
  574         if (socksize < sizeof(*sdl))
  575                 socksize = sizeof(*sdl);
  576         socksize = roundup2(socksize, sizeof(long));
  577         ifasize = sizeof(*ifa) + 2 * socksize;
  578         ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
  579         IFA_LOCK_INIT(ifa);
  580         sdl = (struct sockaddr_dl *)(ifa + 1);
  581         sdl->sdl_len = socksize;
  582         sdl->sdl_family = AF_LINK;
  583         bcopy(ifp->if_xname, sdl->sdl_data, namelen);
  584         sdl->sdl_nlen = namelen;
  585         sdl->sdl_index = ifp->if_index;
  586         sdl->sdl_type = ifp->if_type;
  587         ifp->if_addr = ifa;
  588         ifa->ifa_ifp = ifp;
  589         ifa->ifa_rtrequest = link_rtrequest;
  590         ifa->ifa_addr = (struct sockaddr *)sdl;
  591         sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
  592         ifa->ifa_netmask = (struct sockaddr *)sdl;
  593         sdl->sdl_len = masklen;
  594         while (namelen != 0)
  595                 sdl->sdl_data[--namelen] = 0xff;
  596         ifa->ifa_refcnt = 1;
  597         TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
  598         ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
  599         ifp->if_snd.altq_type = 0;
  600         ifp->if_snd.altq_disc = NULL;
  601         ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
  602         ifp->if_snd.altq_tbr  = NULL;
  603         ifp->if_snd.altq_ifp  = ifp;
  604 
  605         IFNET_WLOCK();
  606         TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
  607         IFNET_WUNLOCK();
  608 
  609         if (domain_init_status >= 2)
  610                 if_attachdomain1(ifp);
  611 
  612         EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
  613         devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
  614 
  615         /* Announce the interface. */
  616         rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
  617 
  618         if (ifp->if_watchdog != NULL) {
  619                 if_printf(ifp,
  620                     "WARNING: using obsoleted if_watchdog interface\n");
  621               
  622                 /*
  623                  * Note that we need if_slowtimo().  If this happens after
  624                  * boot, then call if_slowtimo() directly.
  625                  */
  626                 if (atomic_cmpset_int(&slowtimo_started, 0, 1) && !cold)
  627                         if_slowtimo(0);
  628         }
  629         if (ifp->if_flags & IFF_NEEDSGIANT)
  630                 if_printf(ifp,
  631                     "WARNING: using obsoleted IFF_NEEDSGIANT flag\n");
  632 }
  633 
  634 static void
  635 if_attachdomain(void *dummy)
  636 {
  637         struct ifnet *ifp;
  638         int s;
  639 
  640         s = splnet();
  641         TAILQ_FOREACH(ifp, &ifnet, if_link)
  642                 if_attachdomain1(ifp);
  643         splx(s);
  644 }
  645 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
  646     if_attachdomain, NULL);
  647 
  648 static void
  649 if_attachdomain1(struct ifnet *ifp)
  650 {
  651         struct domain *dp;
  652         int s;
  653 
  654         s = splnet();
  655 
  656         /*
  657          * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
  658          * cannot lock ifp->if_afdata initialization, entirely.
  659          */
  660         if (IF_AFDATA_TRYLOCK(ifp) == 0) {
  661                 splx(s);
  662                 return;
  663         }
  664         if (ifp->if_afdata_initialized >= domain_init_status) {
  665                 IF_AFDATA_UNLOCK(ifp);
  666                 splx(s);
  667                 printf("if_attachdomain called more than once on %s\n",
  668                     ifp->if_xname);
  669                 return;
  670         }
  671         ifp->if_afdata_initialized = domain_init_status;
  672         IF_AFDATA_UNLOCK(ifp);
  673 
  674         /* address family dependent data region */
  675         bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
  676         for (dp = domains; dp; dp = dp->dom_next) {
  677                 if (dp->dom_ifattach)
  678                         ifp->if_afdata[dp->dom_family] =
  679                             (*dp->dom_ifattach)(ifp);
  680         }
  681 
  682         splx(s);
  683 }
  684 
  685 /*
  686  * Remove any unicast or broadcast network addresses from an interface.
  687  */
  688 void
  689 if_purgeaddrs(struct ifnet *ifp)
  690 {
  691         struct ifaddr *ifa, *next;
  692 
  693         TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  694                 if (ifa->ifa_addr->sa_family == AF_LINK)
  695                         continue;
  696 #ifdef INET
  697                 /* XXX: Ugly!! ad hoc just for INET */
  698                 if (ifa->ifa_addr->sa_family == AF_INET) {
  699                         struct ifaliasreq ifr;
  700 
  701                         bzero(&ifr, sizeof(ifr));
  702                         ifr.ifra_addr = *ifa->ifa_addr;
  703                         if (ifa->ifa_dstaddr)
  704                                 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
  705                         if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
  706                             NULL) == 0)
  707                                 continue;
  708                 }
  709 #endif /* INET */
  710 #ifdef INET6
  711                 if (ifa->ifa_addr->sa_family == AF_INET6) {
  712                         in6_purgeaddr(ifa);
  713                         /* ifp_addrhead is already updated */
  714                         continue;
  715                 }
  716 #endif /* INET6 */
  717                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  718                 IFAFREE(ifa);
  719         }
  720 }
  721 
  722 /*
  723  * Remove any multicast network addresses from an interface.
  724  */
  725 static void
  726 if_purgemaddrs(struct ifnet *ifp)
  727 {
  728         struct ifmultiaddr *ifma;
  729         struct ifmultiaddr *next;
  730 
  731         IF_ADDR_LOCK(ifp);
  732         TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
  733                 if_delmulti_locked(ifp, ifma, 1);
  734         IF_ADDR_UNLOCK(ifp);
  735 }
  736 
  737 /*
  738  * Detach an interface, removing it from the
  739  * list of "active" interfaces.
  740  *
  741  * XXXRW: There are some significant questions about event ordering, and
  742  * how to prevent things from starting to use the interface during detach.
  743  */
  744 void
  745 if_detach(struct ifnet *ifp)
  746 {
  747         struct ifaddr *ifa;
  748         struct radix_node_head  *rnh;
  749         int s;
  750         int i;
  751         struct domain *dp;
  752         struct ifnet *iter;
  753         int found = 0;
  754 
  755         IFNET_WLOCK();
  756         TAILQ_FOREACH(iter, &ifnet, if_link)
  757                 if (iter == ifp) {
  758                         TAILQ_REMOVE(&ifnet, ifp, if_link);
  759                         found = 1;
  760                         break;
  761                 }
  762         IFNET_WUNLOCK();
  763         if (!found)
  764                 return;
  765 
  766         /*
  767          * Remove/wait for pending events.
  768          */
  769         taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
  770 
  771         /*
  772          * Remove routes and flush queues.
  773          */
  774         s = splnet();
  775         if_down(ifp);
  776 #ifdef ALTQ
  777         if (ALTQ_IS_ENABLED(&ifp->if_snd))
  778                 altq_disable(&ifp->if_snd);
  779         if (ALTQ_IS_ATTACHED(&ifp->if_snd))
  780                 altq_detach(&ifp->if_snd);
  781 #endif
  782 
  783         if_purgeaddrs(ifp);
  784 
  785 #ifdef INET
  786         in_ifdetach(ifp);
  787 #endif
  788 
  789 #ifdef INET6
  790         /*
  791          * Remove all IPv6 kernel structs related to ifp.  This should be done
  792          * before removing routing entries below, since IPv6 interface direct
  793          * routes are expected to be removed by the IPv6-specific kernel API.
  794          * Otherwise, the kernel will detect some inconsistency and bark it.
  795          */
  796         in6_ifdetach(ifp);
  797 #endif
  798         if_purgemaddrs(ifp);
  799 
  800         /*
  801          * Remove link ifaddr pointer and maybe decrement if_index.
  802          * Clean up all addresses.
  803          */
  804         ifp->if_addr = NULL;
  805         destroy_dev(ifdev_byindex(ifp->if_index));
  806         ifdev_setbyindex(ifp->if_index, NULL);  
  807 
  808         /* We can now free link ifaddr. */
  809         if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
  810                 ifa = TAILQ_FIRST(&ifp->if_addrhead);
  811                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  812                 IFAFREE(ifa);
  813         }
  814 
  815         /*
  816          * Delete all remaining routes using this interface
  817          * Unfortuneatly the only way to do this is to slog through
  818          * the entire routing table looking for routes which point
  819          * to this interface...oh well...
  820          */
  821         for (i = 1; i <= AF_MAX; i++) {
  822             int j;
  823             for (j = 0; j < rt_numfibs; j++) {
  824                 if ((rnh = rt_tables[j][i]) == NULL)
  825                         continue;
  826                 RADIX_NODE_HEAD_LOCK(rnh);
  827                 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
  828                 RADIX_NODE_HEAD_UNLOCK(rnh);
  829             }
  830         }
  831 
  832         /* Announce that the interface is gone. */
  833         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
  834         EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
  835         devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
  836         if_delgroups(ifp);
  837 
  838         IF_AFDATA_LOCK(ifp);
  839         for (dp = domains; dp; dp = dp->dom_next) {
  840                 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
  841                         (*dp->dom_ifdetach)(ifp,
  842                             ifp->if_afdata[dp->dom_family]);
  843         }
  844         IF_AFDATA_UNLOCK(ifp);
  845 
  846 #ifdef MAC
  847         mac_destroy_ifnet(ifp);
  848 #endif /* MAC */
  849         KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
  850         knlist_clear(&ifp->if_klist, 0);
  851         knlist_destroy(&ifp->if_klist);
  852         mtx_destroy(&ifp->if_snd.ifq_mtx);
  853         IF_AFDATA_DESTROY(ifp);
  854         splx(s);
  855 }
  856 
  857 /*
  858  * Add a group to an interface
  859  */
  860 int
  861 if_addgroup(struct ifnet *ifp, const char *groupname)
  862 {
  863         struct ifg_list         *ifgl;
  864         struct ifg_group        *ifg = NULL;
  865         struct ifg_member       *ifgm;
  866 
  867         if (groupname[0] && groupname[strlen(groupname) - 1] >= '' &&
  868             groupname[strlen(groupname) - 1] <= '9')
  869                 return (EINVAL);
  870 
  871         IFNET_WLOCK();
  872         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
  873                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
  874                         IFNET_WUNLOCK();
  875                         return (EEXIST);
  876                 }
  877 
  878         if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
  879             M_NOWAIT)) == NULL) {
  880                 IFNET_WUNLOCK();
  881                 return (ENOMEM);
  882         }
  883 
  884         if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
  885             M_TEMP, M_NOWAIT)) == NULL) {
  886                 free(ifgl, M_TEMP);
  887                 IFNET_WUNLOCK();
  888                 return (ENOMEM);
  889         }
  890 
  891         TAILQ_FOREACH(ifg, &ifg_head, ifg_next)
  892                 if (!strcmp(ifg->ifg_group, groupname))
  893                         break;
  894 
  895         if (ifg == NULL) {
  896                 if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
  897                     M_TEMP, M_NOWAIT)) == NULL) {
  898                         free(ifgl, M_TEMP);
  899                         free(ifgm, M_TEMP);
  900                         IFNET_WUNLOCK();
  901                         return (ENOMEM);
  902                 }
  903                 strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
  904                 ifg->ifg_refcnt = 0;
  905                 TAILQ_INIT(&ifg->ifg_members);
  906                 EVENTHANDLER_INVOKE(group_attach_event, ifg);
  907                 TAILQ_INSERT_TAIL(&ifg_head, ifg, ifg_next);
  908         }
  909 
  910         ifg->ifg_refcnt++;
  911         ifgl->ifgl_group = ifg;
  912         ifgm->ifgm_ifp = ifp;
  913 
  914         IF_ADDR_LOCK(ifp);
  915         TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
  916         TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
  917         IF_ADDR_UNLOCK(ifp);
  918 
  919         IFNET_WUNLOCK();
  920 
  921         EVENTHANDLER_INVOKE(group_change_event, groupname);
  922 
  923         return (0);
  924 }
  925 
  926 /*
  927  * Remove a group from an interface
  928  */
  929 int
  930 if_delgroup(struct ifnet *ifp, const char *groupname)
  931 {
  932         struct ifg_list         *ifgl;
  933         struct ifg_member       *ifgm;
  934 
  935         IFNET_WLOCK();
  936         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
  937                 if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
  938                         break;
  939         if (ifgl == NULL) {
  940                 IFNET_WUNLOCK();
  941                 return (ENOENT);
  942         }
  943 
  944         IF_ADDR_LOCK(ifp);
  945         TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
  946         IF_ADDR_UNLOCK(ifp);
  947 
  948         TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
  949                 if (ifgm->ifgm_ifp == ifp)
  950                         break;
  951 
  952         if (ifgm != NULL) {
  953                 TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
  954                 free(ifgm, M_TEMP);
  955         }
  956 
  957         if (--ifgl->ifgl_group->ifg_refcnt == 0) {
  958                 TAILQ_REMOVE(&ifg_head, ifgl->ifgl_group, ifg_next);
  959                 EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
  960                 free(ifgl->ifgl_group, M_TEMP);
  961         }
  962         IFNET_WUNLOCK();
  963 
  964         free(ifgl, M_TEMP);
  965 
  966         EVENTHANDLER_INVOKE(group_change_event, groupname);
  967 
  968         return (0);
  969 }
  970 
  971 /*
  972  * Remove an interface from all groups
  973  */
  974 static void
  975 if_delgroups(struct ifnet *ifp)
  976 {
  977         struct ifg_list         *ifgl;
  978         struct ifg_member       *ifgm;
  979         char groupname[IFNAMSIZ];
  980 
  981         IFNET_WLOCK();
  982         while (!TAILQ_EMPTY(&ifp->if_groups)) {
  983                 ifgl = TAILQ_FIRST(&ifp->if_groups);
  984 
  985                 strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
  986 
  987                 IF_ADDR_LOCK(ifp);
  988                 TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
  989                 IF_ADDR_UNLOCK(ifp);
  990 
  991                 TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
  992                         if (ifgm->ifgm_ifp == ifp)
  993                                 break;
  994 
  995                 if (ifgm != NULL) {
  996                         TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
  997                             ifgm_next);
  998                         free(ifgm, M_TEMP);
  999                 }
 1000 
 1001                 if (--ifgl->ifgl_group->ifg_refcnt == 0) {
 1002                         TAILQ_REMOVE(&ifg_head, ifgl->ifgl_group, ifg_next);
 1003                         EVENTHANDLER_INVOKE(group_detach_event,
 1004                             ifgl->ifgl_group);
 1005                         free(ifgl->ifgl_group, M_TEMP);
 1006                 }
 1007                 IFNET_WUNLOCK();
 1008 
 1009                 free(ifgl, M_TEMP);
 1010 
 1011                 EVENTHANDLER_INVOKE(group_change_event, groupname);
 1012 
 1013                 IFNET_WLOCK();
 1014         }
 1015         IFNET_WUNLOCK();
 1016 }
 1017 
 1018 /*
 1019  * Stores all groups from an interface in memory pointed
 1020  * to by data
 1021  */
 1022 static int
 1023 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
 1024 {
 1025         int                      len, error;
 1026         struct ifg_list         *ifgl;
 1027         struct ifg_req           ifgrq, *ifgp;
 1028         struct ifgroupreq       *ifgr = data;
 1029 
 1030         if (ifgr->ifgr_len == 0) {
 1031                 IF_ADDR_LOCK(ifp);
 1032                 TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
 1033                         ifgr->ifgr_len += sizeof(struct ifg_req);
 1034                 IF_ADDR_UNLOCK(ifp);
 1035                 return (0);
 1036         }
 1037 
 1038         len = ifgr->ifgr_len;
 1039         ifgp = ifgr->ifgr_groups;
 1040         /* XXX: wire */
 1041         IF_ADDR_LOCK(ifp);
 1042         TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
 1043                 if (len < sizeof(ifgrq)) {
 1044                         IF_ADDR_UNLOCK(ifp);
 1045                         return (EINVAL);
 1046                 }
 1047                 bzero(&ifgrq, sizeof ifgrq);
 1048                 strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
 1049                     sizeof(ifgrq.ifgrq_group));
 1050                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1051                         IF_ADDR_UNLOCK(ifp);
 1052                         return (error);
 1053                 }
 1054                 len -= sizeof(ifgrq);
 1055                 ifgp++;
 1056         }
 1057         IF_ADDR_UNLOCK(ifp);
 1058 
 1059         return (0);
 1060 }
 1061 
 1062 /*
 1063  * Stores all members of a group in memory pointed to by data
 1064  */
 1065 static int
 1066 if_getgroupmembers(struct ifgroupreq *data)
 1067 {
 1068         struct ifgroupreq       *ifgr = data;
 1069         struct ifg_group        *ifg;
 1070         struct ifg_member       *ifgm;
 1071         struct ifg_req           ifgrq, *ifgp;
 1072         int                      len, error;
 1073 
 1074         IFNET_RLOCK();
 1075         TAILQ_FOREACH(ifg, &ifg_head, ifg_next)
 1076                 if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
 1077                         break;
 1078         if (ifg == NULL) {
 1079                 IFNET_RUNLOCK();
 1080                 return (ENOENT);
 1081         }
 1082 
 1083         if (ifgr->ifgr_len == 0) {
 1084                 TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
 1085                         ifgr->ifgr_len += sizeof(ifgrq);
 1086                 IFNET_RUNLOCK();
 1087                 return (0);
 1088         }
 1089 
 1090         len = ifgr->ifgr_len;
 1091         ifgp = ifgr->ifgr_groups;
 1092         TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
 1093                 if (len < sizeof(ifgrq)) {
 1094                         IFNET_RUNLOCK();
 1095                         return (EINVAL);
 1096                 }
 1097                 bzero(&ifgrq, sizeof ifgrq);
 1098                 strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
 1099                     sizeof(ifgrq.ifgrq_member));
 1100                 if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
 1101                         IFNET_RUNLOCK();
 1102                         return (error);
 1103                 }
 1104                 len -= sizeof(ifgrq);
 1105                 ifgp++;
 1106         }
 1107         IFNET_RUNLOCK();
 1108 
 1109         return (0);
 1110 }
 1111 
 1112 /*
 1113  * Delete Routes for a Network Interface
 1114  *
 1115  * Called for each routing entry via the rnh->rnh_walktree() call above
 1116  * to delete all route entries referencing a detaching network interface.
 1117  *
 1118  * Arguments:
 1119  *      rn      pointer to node in the routing table
 1120  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
 1121  *
 1122  * Returns:
 1123  *      0       successful
 1124  *      errno   failed - reason indicated
 1125  *
 1126  */
 1127 static int
 1128 if_rtdel(struct radix_node *rn, void *arg)
 1129 {
 1130         struct rtentry  *rt = (struct rtentry *)rn;
 1131         struct ifnet    *ifp = arg;
 1132         int             err;
 1133 
 1134         if (rt->rt_ifp == ifp) {
 1135 
 1136                 /*
 1137                  * Protect (sorta) against walktree recursion problems
 1138                  * with cloned routes
 1139                  */
 1140                 if ((rt->rt_flags & RTF_UP) == 0)
 1141                         return (0);
 1142 
 1143                 err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
 1144                                 rt_mask(rt), rt->rt_flags,
 1145                                 (struct rtentry **) NULL, rt->rt_fibnum);
 1146                 if (err) {
 1147                         log(LOG_WARNING, "if_rtdel: error %d\n", err);
 1148                 }
 1149         }
 1150 
 1151         return (0);
 1152 }
 1153 
 1154 /*
 1155  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
 1156  * structs used to represent other address families, it is necessary
 1157  * to perform a different comparison.
 1158  */
 1159 
 1160 #define sa_equal(a1, a2)        \
 1161         (bcmp((a1), (a2), ((a1))->sa_len) == 0)
 1162 
 1163 #define sa_dl_equal(a1, a2)     \
 1164         ((((struct sockaddr_dl *)(a1))->sdl_len ==                      \
 1165          ((struct sockaddr_dl *)(a2))->sdl_len) &&                      \
 1166          (bcmp(LLADDR((struct sockaddr_dl *)(a1)),                      \
 1167                LLADDR((struct sockaddr_dl *)(a2)),                      \
 1168                ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
 1169 
 1170 /*
 1171  * Locate an interface based on a complete address.
 1172  */
 1173 /*ARGSUSED*/
 1174 struct ifaddr *
 1175 ifa_ifwithaddr(struct sockaddr *addr)
 1176 {
 1177         struct ifnet *ifp;
 1178         struct ifaddr *ifa;
 1179 
 1180         IFNET_RLOCK();
 1181         TAILQ_FOREACH(ifp, &ifnet, if_link)
 1182                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1183                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1184                                 continue;
 1185                         if (sa_equal(addr, ifa->ifa_addr))
 1186                                 goto done;
 1187                         /* IP6 doesn't have broadcast */
 1188                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1189                             ifa->ifa_broadaddr &&
 1190                             ifa->ifa_broadaddr->sa_len != 0 &&
 1191                             sa_equal(ifa->ifa_broadaddr, addr))
 1192                                 goto done;
 1193                 }
 1194         ifa = NULL;
 1195 done:
 1196         IFNET_RUNLOCK();
 1197         return (ifa);
 1198 }
 1199 
 1200 /*
 1201  * Locate an interface based on the broadcast address.
 1202  */
 1203 /* ARGSUSED */
 1204 struct ifaddr *
 1205 ifa_ifwithbroadaddr(struct sockaddr *addr)
 1206 {
 1207         struct ifnet *ifp;
 1208         struct ifaddr *ifa;
 1209 
 1210         IFNET_RLOCK();
 1211         TAILQ_FOREACH(ifp, &ifnet, if_link)
 1212                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1213                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1214                                 continue;
 1215                         if ((ifp->if_flags & IFF_BROADCAST) &&
 1216                             ifa->ifa_broadaddr &&
 1217                             ifa->ifa_broadaddr->sa_len != 0 &&
 1218                             sa_equal(ifa->ifa_broadaddr, addr))
 1219                                 goto done;
 1220                 }
 1221         ifa = NULL;
 1222 done:
 1223         IFNET_RUNLOCK();
 1224         return (ifa);
 1225 }
 1226 
 1227 /*
 1228  * Locate the point to point interface with a given destination address.
 1229  */
 1230 /*ARGSUSED*/
 1231 struct ifaddr *
 1232 ifa_ifwithdstaddr(struct sockaddr *addr)
 1233 {
 1234         struct ifnet *ifp;
 1235         struct ifaddr *ifa;
 1236 
 1237         IFNET_RLOCK();
 1238         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1239                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
 1240                         continue;
 1241                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1242                         if (ifa->ifa_addr->sa_family != addr->sa_family)
 1243                                 continue;
 1244                         if (ifa->ifa_dstaddr != NULL &&
 1245                             sa_equal(addr, ifa->ifa_dstaddr))
 1246                                 goto done;
 1247                 }
 1248         }
 1249         ifa = NULL;
 1250 done:
 1251         IFNET_RUNLOCK();
 1252         return (ifa);
 1253 }
 1254 
 1255 /*
 1256  * Find an interface on a specific network.  If many, choice
 1257  * is most specific found.
 1258  */
 1259 struct ifaddr *
 1260 ifa_ifwithnet(struct sockaddr *addr)
 1261 {
 1262         struct ifnet *ifp;
 1263         struct ifaddr *ifa;
 1264         struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
 1265         u_int af = addr->sa_family;
 1266         char *addr_data = addr->sa_data, *cplim;
 1267 
 1268         /*
 1269          * AF_LINK addresses can be looked up directly by their index number,
 1270          * so do that if we can.
 1271          */
 1272         if (af == AF_LINK) {
 1273             struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
 1274             if (sdl->sdl_index && sdl->sdl_index <= if_index)
 1275                 return (ifaddr_byindex(sdl->sdl_index));
 1276         }
 1277 
 1278         /*
 1279          * Scan though each interface, looking for ones that have
 1280          * addresses in this address family.
 1281          */
 1282         IFNET_RLOCK();
 1283         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1284                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1285                         char *cp, *cp2, *cp3;
 1286 
 1287                         if (ifa->ifa_addr->sa_family != af)
 1288 next:                           continue;
 1289                         if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
 1290                                 /*
 1291                                  * This is a bit broken as it doesn't
 1292                                  * take into account that the remote end may
 1293                                  * be a single node in the network we are
 1294                                  * looking for.
 1295                                  * The trouble is that we don't know the
 1296                                  * netmask for the remote end.
 1297                                  */
 1298                                 if (ifa->ifa_dstaddr != NULL &&
 1299                                     sa_equal(addr, ifa->ifa_dstaddr))
 1300                                         goto done;
 1301                         } else {
 1302                                 /*
 1303                                  * if we have a special address handler,
 1304                                  * then use it instead of the generic one.
 1305                                  */
 1306                                 if (ifa->ifa_claim_addr) {
 1307                                         if ((*ifa->ifa_claim_addr)(ifa, addr))
 1308                                                 goto done;
 1309                                         continue;
 1310                                 }
 1311 
 1312                                 /*
 1313                                  * Scan all the bits in the ifa's address.
 1314                                  * If a bit dissagrees with what we are
 1315                                  * looking for, mask it with the netmask
 1316                                  * to see if it really matters.
 1317                                  * (A byte at a time)
 1318                                  */
 1319                                 if (ifa->ifa_netmask == 0)
 1320                                         continue;
 1321                                 cp = addr_data;
 1322                                 cp2 = ifa->ifa_addr->sa_data;
 1323                                 cp3 = ifa->ifa_netmask->sa_data;
 1324                                 cplim = ifa->ifa_netmask->sa_len
 1325                                         + (char *)ifa->ifa_netmask;
 1326                                 while (cp3 < cplim)
 1327                                         if ((*cp++ ^ *cp2++) & *cp3++)
 1328                                                 goto next; /* next address! */
 1329                                 /*
 1330                                  * If the netmask of what we just found
 1331                                  * is more specific than what we had before
 1332                                  * (if we had one) then remember the new one
 1333                                  * before continuing to search
 1334                                  * for an even better one.
 1335                                  */
 1336                                 if (ifa_maybe == 0 ||
 1337                                     rn_refines((caddr_t)ifa->ifa_netmask,
 1338                                     (caddr_t)ifa_maybe->ifa_netmask))
 1339                                         ifa_maybe = ifa;
 1340                         }
 1341                 }
 1342         }
 1343         ifa = ifa_maybe;
 1344 done:
 1345         IFNET_RUNLOCK();
 1346         return (ifa);
 1347 }
 1348 
 1349 /*
 1350  * Find an interface address specific to an interface best matching
 1351  * a given address.
 1352  */
 1353 struct ifaddr *
 1354 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
 1355 {
 1356         struct ifaddr *ifa;
 1357         char *cp, *cp2, *cp3;
 1358         char *cplim;
 1359         struct ifaddr *ifa_maybe = 0;
 1360         u_int af = addr->sa_family;
 1361 
 1362         if (af >= AF_MAX)
 1363                 return (NULL);
 1364         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1365                 if (ifa->ifa_addr->sa_family != af)
 1366                         continue;
 1367                 if (ifa_maybe == 0)
 1368                         ifa_maybe = ifa;
 1369                 if (ifa->ifa_netmask == 0) {
 1370                         if (sa_equal(addr, ifa->ifa_addr) ||
 1371                             (ifa->ifa_dstaddr &&
 1372                             sa_equal(addr, ifa->ifa_dstaddr)))
 1373                                 goto done;
 1374                         continue;
 1375                 }
 1376                 if (ifp->if_flags & IFF_POINTOPOINT) {
 1377                         if (sa_equal(addr, ifa->ifa_dstaddr))
 1378                                 goto done;
 1379                 } else {
 1380                         cp = addr->sa_data;
 1381                         cp2 = ifa->ifa_addr->sa_data;
 1382                         cp3 = ifa->ifa_netmask->sa_data;
 1383                         cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
 1384                         for (; cp3 < cplim; cp3++)
 1385                                 if ((*cp++ ^ *cp2++) & *cp3)
 1386                                         break;
 1387                         if (cp3 == cplim)
 1388                                 goto done;
 1389                 }
 1390         }
 1391         ifa = ifa_maybe;
 1392 done:
 1393         return (ifa);
 1394 }
 1395 
 1396 #include <net/route.h>
 1397 
 1398 /*
 1399  * Default action when installing a route with a Link Level gateway.
 1400  * Lookup an appropriate real ifa to point to.
 1401  * This should be moved to /sys/net/link.c eventually.
 1402  */
 1403 static void
 1404 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
 1405 {
 1406         struct ifaddr *ifa, *oifa;
 1407         struct sockaddr *dst;
 1408         struct ifnet *ifp;
 1409 
 1410         RT_LOCK_ASSERT(rt);
 1411 
 1412         if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
 1413             ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
 1414                 return;
 1415         ifa = ifaof_ifpforaddr(dst, ifp);
 1416         if (ifa) {
 1417                 IFAREF(ifa);            /* XXX */
 1418                 oifa = rt->rt_ifa;
 1419                 rt->rt_ifa = ifa;
 1420                 IFAFREE(oifa);
 1421                 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
 1422                         ifa->ifa_rtrequest(cmd, rt, info);
 1423         }
 1424 }
 1425 
 1426 /*
 1427  * Mark an interface down and notify protocols of
 1428  * the transition.
 1429  * NOTE: must be called at splnet or eqivalent.
 1430  */
 1431 static void
 1432 if_unroute(struct ifnet *ifp, int flag, int fam)
 1433 {
 1434         struct ifaddr *ifa;
 1435 
 1436         KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
 1437 
 1438         ifp->if_flags &= ~flag;
 1439         getmicrotime(&ifp->if_lastchange);
 1440         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1441                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1442                         pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
 1443         if_qflush(&ifp->if_snd);
 1444 #ifdef DEV_CARP
 1445         if (ifp->if_carp)
 1446                 carp_carpdev_state(ifp->if_carp);
 1447 #endif
 1448         rt_ifmsg(ifp);
 1449 }
 1450 
 1451 /*
 1452  * Mark an interface up and notify protocols of
 1453  * the transition.
 1454  * NOTE: must be called at splnet or eqivalent.
 1455  */
 1456 static void
 1457 if_route(struct ifnet *ifp, int flag, int fam)
 1458 {
 1459         struct ifaddr *ifa;
 1460 
 1461         KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
 1462 
 1463         ifp->if_flags |= flag;
 1464         getmicrotime(&ifp->if_lastchange);
 1465         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1466                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1467                         pfctlinput(PRC_IFUP, ifa->ifa_addr);
 1468 #ifdef DEV_CARP
 1469         if (ifp->if_carp)
 1470                 carp_carpdev_state(ifp->if_carp);
 1471 #endif
 1472         rt_ifmsg(ifp);
 1473 #ifdef INET6
 1474         in6_if_up(ifp);
 1475 #endif
 1476 }
 1477 
 1478 void    (*vlan_link_state_p)(struct ifnet *, int);      /* XXX: private from if_vlan */
 1479 void    (*vlan_trunk_cap_p)(struct ifnet *);            /* XXX: private from if_vlan */
 1480 
 1481 /*
 1482  * Handle a change in the interface link state. To avoid LORs
 1483  * between driver lock and upper layer locks, as well as possible
 1484  * recursions, we post event to taskqueue, and all job
 1485  * is done in static do_link_state_change().
 1486  */
 1487 void
 1488 if_link_state_change(struct ifnet *ifp, int link_state)
 1489 {
 1490         /* Return if state hasn't changed. */
 1491         if (ifp->if_link_state == link_state)
 1492                 return;
 1493 
 1494         ifp->if_link_state = link_state;
 1495 
 1496         taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
 1497 }
 1498 
 1499 static void
 1500 do_link_state_change(void *arg, int pending)
 1501 {
 1502         struct ifnet *ifp = (struct ifnet *)arg;
 1503         int link_state = ifp->if_link_state;
 1504         int link;
 1505 
 1506         /* Notify that the link state has changed. */
 1507         rt_ifmsg(ifp);
 1508         if (link_state == LINK_STATE_UP)
 1509                 link = NOTE_LINKUP;
 1510         else if (link_state == LINK_STATE_DOWN)
 1511                 link = NOTE_LINKDOWN;
 1512         else
 1513                 link = NOTE_LINKINV;
 1514         KNOTE_UNLOCKED(&ifp->if_klist, link);
 1515         if (ifp->if_vlantrunk != NULL)
 1516                 (*vlan_link_state_p)(ifp, link);
 1517 
 1518         if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
 1519             IFP2AC(ifp)->ac_netgraph != NULL)
 1520                 (*ng_ether_link_state_p)(ifp, link_state);
 1521 #ifdef DEV_CARP
 1522         if (ifp->if_carp)
 1523                 carp_carpdev_state(ifp->if_carp);
 1524 #endif
 1525         if (ifp->if_bridge) {
 1526                 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
 1527                 (*bstp_linkstate_p)(ifp, link_state);
 1528         }
 1529         if (ifp->if_lagg) {
 1530                 KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
 1531                 (*lagg_linkstate_p)(ifp, link_state);
 1532         }
 1533 
 1534         devctl_notify("IFNET", ifp->if_xname,
 1535             (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
 1536         if (pending > 1)
 1537                 if_printf(ifp, "%d link states coalesced\n", pending);
 1538         if (log_link_state_change)
 1539                 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
 1540                     (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
 1541 }
 1542 
 1543 /*
 1544  * Mark an interface down and notify protocols of
 1545  * the transition.
 1546  * NOTE: must be called at splnet or eqivalent.
 1547  */
 1548 void
 1549 if_down(struct ifnet *ifp)
 1550 {
 1551 
 1552         if_unroute(ifp, IFF_UP, AF_UNSPEC);
 1553 }
 1554 
 1555 /*
 1556  * Mark an interface up and notify protocols of
 1557  * the transition.
 1558  * NOTE: must be called at splnet or eqivalent.
 1559  */
 1560 void
 1561 if_up(struct ifnet *ifp)
 1562 {
 1563 
 1564         if_route(ifp, IFF_UP, AF_UNSPEC);
 1565 }
 1566 
 1567 /*
 1568  * Flush an interface queue.
 1569  */
 1570 static void
 1571 if_qflush(struct ifaltq *ifq)
 1572 {
 1573         struct mbuf *m, *n;
 1574 
 1575         IFQ_LOCK(ifq);
 1576 #ifdef ALTQ
 1577         if (ALTQ_IS_ENABLED(ifq))
 1578                 ALTQ_PURGE(ifq);
 1579 #endif
 1580         n = ifq->ifq_head;
 1581         while ((m = n) != 0) {
 1582                 n = m->m_act;
 1583                 m_freem(m);
 1584         }
 1585         ifq->ifq_head = 0;
 1586         ifq->ifq_tail = 0;
 1587         ifq->ifq_len = 0;
 1588         IFQ_UNLOCK(ifq);
 1589 }
 1590 
 1591 /*
 1592  * Handle interface watchdog timer routines.  Called
 1593  * from softclock, we decrement timers (if set) and
 1594  * call the appropriate interface routine on expiration.
 1595  *
 1596  * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
 1597  * holding Giant.  If we switch to an MPSAFE callout, we likely need to grab
 1598  * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
 1599  */
 1600 static void
 1601 if_slowtimo(void *arg)
 1602 {
 1603         struct ifnet *ifp;
 1604         int s = splimp();
 1605 
 1606         IFNET_RLOCK();
 1607         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1608                 if (ifp->if_timer == 0 || --ifp->if_timer)
 1609                         continue;
 1610                 if (ifp->if_watchdog)
 1611                         (*ifp->if_watchdog)(ifp);
 1612         }
 1613         IFNET_RUNLOCK();
 1614         splx(s);
 1615         timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
 1616 }
 1617 
 1618 /*
 1619  * Map interface name to
 1620  * interface structure pointer.
 1621  */
 1622 struct ifnet *
 1623 ifunit(const char *name)
 1624 {
 1625         struct ifnet *ifp;
 1626 
 1627         IFNET_RLOCK();
 1628         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1629                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
 1630                         break;
 1631         }
 1632         IFNET_RUNLOCK();
 1633         return (ifp);
 1634 }
 1635 
 1636 /*
 1637  * Hardware specific interface ioctls.
 1638  */
 1639 static int
 1640 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
 1641 {
 1642         struct ifreq *ifr;
 1643         struct ifstat *ifs;
 1644         int error = 0;
 1645         int new_flags, temp_flags;
 1646         size_t namelen, onamelen;
 1647         char new_name[IFNAMSIZ];
 1648         struct ifaddr *ifa;
 1649         struct sockaddr_dl *sdl;
 1650 
 1651         ifr = (struct ifreq *)data;
 1652         switch (cmd) {
 1653         case SIOCGIFINDEX:
 1654                 ifr->ifr_index = ifp->if_index;
 1655                 break;
 1656 
 1657         case SIOCGIFFLAGS:
 1658                 temp_flags = ifp->if_flags | ifp->if_drv_flags;
 1659                 ifr->ifr_flags = temp_flags & 0xffff;
 1660                 ifr->ifr_flagshigh = temp_flags >> 16;
 1661                 break;
 1662 
 1663         case SIOCGIFCAP:
 1664                 ifr->ifr_reqcap = ifp->if_capabilities;
 1665                 ifr->ifr_curcap = ifp->if_capenable;
 1666                 break;
 1667 
 1668 #ifdef MAC
 1669         case SIOCGIFMAC:
 1670                 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp);
 1671                 break;
 1672 #endif
 1673 
 1674         case SIOCGIFMETRIC:
 1675                 ifr->ifr_metric = ifp->if_metric;
 1676                 break;
 1677 
 1678         case SIOCGIFMTU:
 1679                 ifr->ifr_mtu = ifp->if_mtu;
 1680                 break;
 1681 
 1682         case SIOCGIFPHYS:
 1683                 ifr->ifr_phys = ifp->if_physical;
 1684                 break;
 1685 
 1686         case SIOCSIFFLAGS:
 1687                 error = priv_check(td, PRIV_NET_SETIFFLAGS);
 1688                 if (error)
 1689                         return (error);
 1690                 /*
 1691                  * Currently, no driver owned flags pass the IFF_CANTCHANGE
 1692                  * check, so we don't need special handling here yet.
 1693                  */
 1694                 new_flags = (ifr->ifr_flags & 0xffff) |
 1695                     (ifr->ifr_flagshigh << 16);
 1696                 if (ifp->if_flags & IFF_SMART) {
 1697                         /* Smart drivers twiddle their own routes */
 1698                 } else if (ifp->if_flags & IFF_UP &&
 1699                     (new_flags & IFF_UP) == 0) {
 1700                         int s = splimp();
 1701                         if_down(ifp);
 1702                         splx(s);
 1703                 } else if (new_flags & IFF_UP &&
 1704                     (ifp->if_flags & IFF_UP) == 0) {
 1705                         int s = splimp();
 1706                         if_up(ifp);
 1707                         splx(s);
 1708                 }
 1709                 /* See if permanently promiscuous mode bit is about to flip */
 1710                 if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
 1711                         if (new_flags & IFF_PPROMISC)
 1712                                 ifp->if_flags |= IFF_PROMISC;
 1713                         else if (ifp->if_pcount == 0)
 1714                                 ifp->if_flags &= ~IFF_PROMISC;
 1715                         log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
 1716                             ifp->if_xname,
 1717                             (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
 1718                 }
 1719                 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
 1720                         (new_flags &~ IFF_CANTCHANGE);
 1721                 if (ifp->if_ioctl) {
 1722                         IFF_LOCKGIANT(ifp);
 1723                         (void) (*ifp->if_ioctl)(ifp, cmd, data);
 1724                         IFF_UNLOCKGIANT(ifp);
 1725                 }
 1726                 getmicrotime(&ifp->if_lastchange);
 1727                 break;
 1728 
 1729         case SIOCSIFCAP:
 1730                 error = priv_check(td, PRIV_NET_SETIFCAP);
 1731                 if (error)
 1732                         return (error);
 1733                 if (ifp->if_ioctl == NULL)
 1734                         return (EOPNOTSUPP);
 1735                 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
 1736                         return (EINVAL);
 1737                 IFF_LOCKGIANT(ifp);
 1738                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1739                 IFF_UNLOCKGIANT(ifp);
 1740                 if (error == 0)
 1741                         getmicrotime(&ifp->if_lastchange);
 1742                 break;
 1743 
 1744 #ifdef MAC
 1745         case SIOCSIFMAC:
 1746                 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp);
 1747                 break;
 1748 #endif
 1749 
 1750         case SIOCSIFNAME:
 1751                 error = priv_check(td, PRIV_NET_SETIFNAME);
 1752                 if (error)
 1753                         return (error);
 1754                 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
 1755                 if (error != 0)
 1756                         return (error);
 1757                 if (new_name[0] == '\0')
 1758                         return (EINVAL);
 1759                 if (ifunit(new_name) != NULL)
 1760                         return (EEXIST);
 1761 
 1762                 /*
 1763                  * XXX: Locking.  Nothing else seems to lock if_flags,
 1764                  * and there are numerous other races with the
 1765                  * ifunit() checks not being atomic with namespace
 1766                  * changes (renames, vmoves, if_attach, etc).
 1767                  */
 1768                 ifp->if_flags |= IFF_RENAMING;
 1769                 
 1770                 /* Announce the departure of the interface. */
 1771                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 1772                 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 1773 
 1774                 log(LOG_INFO, "%s: changing name to '%s'\n",
 1775                     ifp->if_xname, new_name);
 1776 
 1777                 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
 1778                 ifa = ifp->if_addr;
 1779                 IFA_LOCK(ifa);
 1780                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 1781                 namelen = strlen(new_name);
 1782                 onamelen = sdl->sdl_nlen;
 1783                 /*
 1784                  * Move the address if needed.  This is safe because we
 1785                  * allocate space for a name of length IFNAMSIZ when we
 1786                  * create this in if_attach().
 1787                  */
 1788                 if (namelen != onamelen) {
 1789                         bcopy(sdl->sdl_data + onamelen,
 1790                             sdl->sdl_data + namelen, sdl->sdl_alen);
 1791                 }
 1792                 bcopy(new_name, sdl->sdl_data, namelen);
 1793                 sdl->sdl_nlen = namelen;
 1794                 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
 1795                 bzero(sdl->sdl_data, onamelen);
 1796                 while (namelen != 0)
 1797                         sdl->sdl_data[--namelen] = 0xff;
 1798                 IFA_UNLOCK(ifa);
 1799 
 1800                 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
 1801                 /* Announce the return of the interface. */
 1802                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 1803 
 1804                 ifp->if_flags &= ~IFF_RENAMING;
 1805                 break;
 1806 
 1807         case SIOCSIFMETRIC:
 1808                 error = priv_check(td, PRIV_NET_SETIFMETRIC);
 1809                 if (error)
 1810                         return (error);
 1811                 ifp->if_metric = ifr->ifr_metric;
 1812                 getmicrotime(&ifp->if_lastchange);
 1813                 break;
 1814 
 1815         case SIOCSIFPHYS:
 1816                 error = priv_check(td, PRIV_NET_SETIFPHYS);
 1817                 if (error)
 1818                         return (error);
 1819                 if (ifp->if_ioctl == NULL)
 1820                         return (EOPNOTSUPP);
 1821                 IFF_LOCKGIANT(ifp);
 1822                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1823                 IFF_UNLOCKGIANT(ifp);
 1824                 if (error == 0)
 1825                         getmicrotime(&ifp->if_lastchange);
 1826                 break;
 1827 
 1828         case SIOCSIFMTU:
 1829         {
 1830                 u_long oldmtu = ifp->if_mtu;
 1831 
 1832                 error = priv_check(td, PRIV_NET_SETIFMTU);
 1833                 if (error)
 1834                         return (error);
 1835                 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
 1836                         return (EINVAL);
 1837                 if (ifp->if_ioctl == NULL)
 1838                         return (EOPNOTSUPP);
 1839                 IFF_LOCKGIANT(ifp);
 1840                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1841                 IFF_UNLOCKGIANT(ifp);
 1842                 if (error == 0) {
 1843                         getmicrotime(&ifp->if_lastchange);
 1844                         rt_ifmsg(ifp);
 1845                 }
 1846                 /*
 1847                  * If the link MTU changed, do network layer specific procedure.
 1848                  */
 1849                 if (ifp->if_mtu != oldmtu) {
 1850 #ifdef INET6
 1851                         nd6_setmtu(ifp);
 1852 #endif
 1853                 }
 1854                 break;
 1855         }
 1856 
 1857         case SIOCADDMULTI:
 1858         case SIOCDELMULTI:
 1859                 if (cmd == SIOCADDMULTI)
 1860                         error = priv_check(td, PRIV_NET_ADDMULTI);
 1861                 else
 1862                         error = priv_check(td, PRIV_NET_DELMULTI);
 1863                 if (error)
 1864                         return (error);
 1865 
 1866                 /* Don't allow group membership on non-multicast interfaces. */
 1867                 if ((ifp->if_flags & IFF_MULTICAST) == 0)
 1868                         return (EOPNOTSUPP);
 1869 
 1870                 /* Don't let users screw up protocols' entries. */
 1871                 if (ifr->ifr_addr.sa_family != AF_LINK)
 1872                         return (EINVAL);
 1873 
 1874                 if (cmd == SIOCADDMULTI) {
 1875                         struct ifmultiaddr *ifma;
 1876 
 1877                         /*
 1878                          * Userland is only permitted to join groups once
 1879                          * via the if_addmulti() KPI, because it cannot hold
 1880                          * struct ifmultiaddr * between calls. It may also
 1881                          * lose a race while we check if the membership
 1882                          * already exists.
 1883                          */
 1884                         IF_ADDR_LOCK(ifp);
 1885                         ifma = if_findmulti(ifp, &ifr->ifr_addr);
 1886                         IF_ADDR_UNLOCK(ifp);
 1887                         if (ifma != NULL)
 1888                                 error = EADDRINUSE;
 1889                         else
 1890                                 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
 1891                 } else {
 1892                         error = if_delmulti(ifp, &ifr->ifr_addr);
 1893                 }
 1894                 if (error == 0)
 1895                         getmicrotime(&ifp->if_lastchange);
 1896                 break;
 1897 
 1898         case SIOCSIFPHYADDR:
 1899         case SIOCDIFPHYADDR:
 1900 #ifdef INET6
 1901         case SIOCSIFPHYADDR_IN6:
 1902 #endif
 1903         case SIOCSLIFPHYADDR:
 1904         case SIOCSIFMEDIA:
 1905         case SIOCSIFGENERIC:
 1906                 error = priv_check(td, PRIV_NET_HWIOCTL);
 1907                 if (error)
 1908                         return (error);
 1909                 if (ifp->if_ioctl == NULL)
 1910                         return (EOPNOTSUPP);
 1911                 IFF_LOCKGIANT(ifp);
 1912                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1913                 IFF_UNLOCKGIANT(ifp);
 1914                 if (error == 0)
 1915                         getmicrotime(&ifp->if_lastchange);
 1916                 break;
 1917 
 1918         case SIOCGIFSTATUS:
 1919                 ifs = (struct ifstat *)data;
 1920                 ifs->ascii[0] = '\0';
 1921 
 1922         case SIOCGIFPSRCADDR:
 1923         case SIOCGIFPDSTADDR:
 1924         case SIOCGLIFPHYADDR:
 1925         case SIOCGIFMEDIA:
 1926         case SIOCGIFGENERIC:
 1927                 if (ifp->if_ioctl == NULL)
 1928                         return (EOPNOTSUPP);
 1929                 IFF_LOCKGIANT(ifp);
 1930                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1931                 IFF_UNLOCKGIANT(ifp);
 1932                 break;
 1933 
 1934         case SIOCSIFLLADDR:
 1935                 error = priv_check(td, PRIV_NET_SETLLADDR);
 1936                 if (error)
 1937                         return (error);
 1938                 error = if_setlladdr(ifp,
 1939                     ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
 1940                 break;
 1941 
 1942         case SIOCAIFGROUP:
 1943         {
 1944                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 1945 
 1946                 error = priv_check(td, PRIV_NET_ADDIFGROUP);
 1947                 if (error)
 1948                         return (error);
 1949                 if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
 1950                         return (error);
 1951                 break;
 1952         }
 1953 
 1954         case SIOCGIFGROUP:
 1955                 if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
 1956                         return (error);
 1957                 break;
 1958 
 1959         case SIOCDIFGROUP:
 1960         {
 1961                 struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
 1962 
 1963                 error = priv_check(td, PRIV_NET_DELIFGROUP);
 1964                 if (error)
 1965                         return (error);
 1966                 if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
 1967                         return (error);
 1968                 break;
 1969         }
 1970 
 1971         default:
 1972                 error = ENOIOCTL;
 1973                 break;
 1974         }
 1975         return (error);
 1976 }
 1977 
 1978 /*
 1979  * Interface ioctls.
 1980  */
 1981 int
 1982 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
 1983 {
 1984         struct ifnet *ifp;
 1985         struct ifreq *ifr;
 1986         int error;
 1987         int oif_flags;
 1988 
 1989         switch (cmd) {
 1990         case SIOCGIFCONF:
 1991         case OSIOCGIFCONF:
 1992 #ifdef __amd64__
 1993         case SIOCGIFCONF32:
 1994 #endif
 1995                 return (ifconf(cmd, data));
 1996         }
 1997         ifr = (struct ifreq *)data;
 1998 
 1999         switch (cmd) {
 2000         case SIOCIFCREATE:
 2001         case SIOCIFCREATE2:
 2002                 error = priv_check(td, PRIV_NET_IFCREATE);
 2003                 if (error)
 2004                         return (error);
 2005                 return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
 2006                         cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
 2007         case SIOCIFDESTROY:
 2008                 error = priv_check(td, PRIV_NET_IFDESTROY);
 2009                 if (error)
 2010                         return (error);
 2011                 return if_clone_destroy(ifr->ifr_name);
 2012 
 2013         case SIOCIFGCLONERS:
 2014                 return (if_clone_list((struct if_clonereq *)data));
 2015         case SIOCGIFGMEMB:
 2016                 return (if_getgroupmembers((struct ifgroupreq *)data));
 2017         }
 2018 
 2019         ifp = ifunit(ifr->ifr_name);
 2020         if (ifp == 0)
 2021                 return (ENXIO);
 2022 
 2023         error = ifhwioctl(cmd, ifp, data, td);
 2024         if (error != ENOIOCTL)
 2025                 return (error);
 2026 
 2027         oif_flags = ifp->if_flags;
 2028         if (so->so_proto == 0)
 2029                 return (EOPNOTSUPP);
 2030 #ifndef COMPAT_43
 2031         error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
 2032                                                                  data,
 2033                                                                  ifp, td));
 2034         if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
 2035                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 2036 #else
 2037         {
 2038                 int ocmd = cmd;
 2039 
 2040                 switch (cmd) {
 2041 
 2042                 case SIOCSIFDSTADDR:
 2043                 case SIOCSIFADDR:
 2044                 case SIOCSIFBRDADDR:
 2045                 case SIOCSIFNETMASK:
 2046 #if BYTE_ORDER != BIG_ENDIAN
 2047                         if (ifr->ifr_addr.sa_family == 0 &&
 2048                             ifr->ifr_addr.sa_len < 16) {
 2049                                 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
 2050                                 ifr->ifr_addr.sa_len = 16;
 2051                         }
 2052 #else
 2053                         if (ifr->ifr_addr.sa_len == 0)
 2054                                 ifr->ifr_addr.sa_len = 16;
 2055 #endif
 2056                         break;
 2057 
 2058                 case OSIOCGIFADDR:
 2059                         cmd = SIOCGIFADDR;
 2060                         break;
 2061 
 2062                 case OSIOCGIFDSTADDR:
 2063                         cmd = SIOCGIFDSTADDR;
 2064                         break;
 2065 
 2066                 case OSIOCGIFBRDADDR:
 2067                         cmd = SIOCGIFBRDADDR;
 2068                         break;
 2069 
 2070                 case OSIOCGIFNETMASK:
 2071                         cmd = SIOCGIFNETMASK;
 2072                 }
 2073                 error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
 2074                                                                    cmd,
 2075                                                                    data,
 2076                                                                    ifp, td));
 2077                 if (error == EOPNOTSUPP && ifp != NULL &&
 2078                     ifp->if_ioctl != NULL)
 2079                         error = (*ifp->if_ioctl)(ifp, cmd, data);
 2080                 switch (ocmd) {
 2081 
 2082                 case OSIOCGIFADDR:
 2083                 case OSIOCGIFDSTADDR:
 2084                 case OSIOCGIFBRDADDR:
 2085                 case OSIOCGIFNETMASK:
 2086                         *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
 2087 
 2088                 }
 2089         }
 2090 #endif /* COMPAT_43 */
 2091 
 2092         if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
 2093 #ifdef INET6
 2094                 DELAY(100);/* XXX: temporary workaround for fxp issue*/
 2095                 if (ifp->if_flags & IFF_UP) {
 2096                         int s = splimp();
 2097                         in6_if_up(ifp);
 2098                         splx(s);
 2099                 }
 2100 #endif
 2101         }
 2102         return (error);
 2103 }
 2104 
 2105 /*
 2106  * The code common to handling reference counted flags,
 2107  * e.g., in ifpromisc() and if_allmulti().
 2108  * The "pflag" argument can specify a permanent mode flag to check,
 2109  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
 2110  *
 2111  * Only to be used on stack-owned flags, not driver-owned flags.
 2112  */
 2113 static int
 2114 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
 2115 {
 2116         struct ifreq ifr;
 2117         int error;
 2118         int oldflags, oldcount;
 2119 
 2120         /* Sanity checks to catch programming errors */
 2121         KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
 2122             ("%s: setting driver-owned flag %d", __func__, flag));
 2123 
 2124         if (onswitch)
 2125                 KASSERT(*refcount >= 0,
 2126                     ("%s: increment negative refcount %d for flag %d",
 2127                     __func__, *refcount, flag));
 2128         else
 2129                 KASSERT(*refcount > 0,
 2130                     ("%s: decrement non-positive refcount %d for flag %d",
 2131                     __func__, *refcount, flag));
 2132 
 2133         /* In case this mode is permanent, just touch refcount */
 2134         if (ifp->if_flags & pflag) {
 2135                 *refcount += onswitch ? 1 : -1;
 2136                 return (0);
 2137         }
 2138 
 2139         /* Save ifnet parameters for if_ioctl() may fail */
 2140         oldcount = *refcount;
 2141         oldflags = ifp->if_flags;
 2142         
 2143         /*
 2144          * See if we aren't the only and touching refcount is enough.
 2145          * Actually toggle interface flag if we are the first or last.
 2146          */
 2147         if (onswitch) {
 2148                 if ((*refcount)++)
 2149                         return (0);
 2150                 ifp->if_flags |= flag;
 2151         } else {
 2152                 if (--(*refcount))
 2153                         return (0);
 2154                 ifp->if_flags &= ~flag;
 2155         }
 2156 
 2157         /* Call down the driver since we've changed interface flags */
 2158         if (ifp->if_ioctl == NULL) {
 2159                 error = EOPNOTSUPP;
 2160                 goto recover;
 2161         }
 2162         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2163         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2164         IFF_LOCKGIANT(ifp);
 2165         error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2166         IFF_UNLOCKGIANT(ifp);
 2167         if (error)
 2168                 goto recover;
 2169         /* Notify userland that interface flags have changed */
 2170         rt_ifmsg(ifp);
 2171         return (0);
 2172 
 2173 recover:
 2174         /* Recover after driver error */
 2175         *refcount = oldcount;
 2176         ifp->if_flags = oldflags;
 2177         return (error);
 2178 }
 2179 
 2180 /*
 2181  * Set/clear promiscuous mode on interface ifp based on the truth value
 2182  * of pswitch.  The calls are reference counted so that only the first
 2183  * "on" request actually has an effect, as does the final "off" request.
 2184  * Results are undefined if the "off" and "on" requests are not matched.
 2185  */
 2186 int
 2187 ifpromisc(struct ifnet *ifp, int pswitch)
 2188 {
 2189         int error;
 2190         int oldflags = ifp->if_flags;
 2191 
 2192         error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
 2193                            &ifp->if_pcount, pswitch);
 2194         /* If promiscuous mode status has changed, log a message */
 2195         if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
 2196                 log(LOG_INFO, "%s: promiscuous mode %s\n",
 2197                     ifp->if_xname,
 2198                     (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
 2199         return (error);
 2200 }
 2201 
 2202 /*
 2203  * Return interface configuration
 2204  * of system.  List may be used
 2205  * in later ioctl's (above) to get
 2206  * other information.
 2207  */
 2208 /*ARGSUSED*/
 2209 static int
 2210 ifconf(u_long cmd, caddr_t data)
 2211 {
 2212         struct ifconf *ifc = (struct ifconf *)data;
 2213 #ifdef __amd64__
 2214         struct ifconf32 *ifc32 = (struct ifconf32 *)data;
 2215         struct ifconf ifc_swab;
 2216 #endif
 2217         struct ifnet *ifp;
 2218         struct ifaddr *ifa;
 2219         struct ifreq ifr;
 2220         struct sbuf *sb;
 2221         int error, full = 0, valid_len, max_len;
 2222 
 2223 #ifdef __amd64__
 2224         if (cmd == SIOCGIFCONF32) {
 2225                 ifc_swab.ifc_len = ifc32->ifc_len;
 2226                 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
 2227                 ifc = &ifc_swab;
 2228         }
 2229 #endif
 2230         /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
 2231         max_len = MAXPHYS - 1;
 2232 
 2233         /* Prevent hostile input from being able to crash the system */
 2234         if (ifc->ifc_len <= 0)
 2235                 return (EINVAL);
 2236 
 2237 again:
 2238         if (ifc->ifc_len <= max_len) {
 2239                 max_len = ifc->ifc_len;
 2240                 full = 1;
 2241         }
 2242         sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
 2243         max_len = 0;
 2244         valid_len = 0;
 2245 
 2246         IFNET_RLOCK();          /* could sleep XXX */
 2247         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 2248                 int addrs;
 2249 
 2250                 /*
 2251                  * Zero the ifr_name buffer to make sure we don't
 2252                  * disclose the contents of the stack.
 2253                  */
 2254                 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
 2255 
 2256                 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
 2257                     >= sizeof(ifr.ifr_name)) {
 2258                         sbuf_delete(sb);
 2259                         IFNET_RUNLOCK();
 2260                         return (ENAMETOOLONG);
 2261                 }
 2262 
 2263                 addrs = 0;
 2264                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2265                         struct sockaddr *sa = ifa->ifa_addr;
 2266 
 2267                         if (prison_if(curthread->td_ucred, sa) != 0)
 2268                                 continue;
 2269                         addrs++;
 2270 #ifdef COMPAT_43
 2271                         if (cmd == OSIOCGIFCONF) {
 2272                                 struct osockaddr *osa =
 2273                                          (struct osockaddr *)&ifr.ifr_addr;
 2274                                 ifr.ifr_addr = *sa;
 2275                                 osa->sa_family = sa->sa_family;
 2276                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2277                                 max_len += sizeof(ifr);
 2278                         } else
 2279 #endif
 2280                         if (sa->sa_len <= sizeof(*sa)) {
 2281                                 ifr.ifr_addr = *sa;
 2282                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 2283                                 max_len += sizeof(ifr);
 2284                         } else {
 2285                                 sbuf_bcat(sb, &ifr,
 2286                                     offsetof(struct ifreq, ifr_addr));
 2287                                 max_len += offsetof(struct ifreq, ifr_addr);
 2288                                 sbuf_bcat(sb, sa, sa->sa_len);
 2289                                 max_len += sa->sa_len;
 2290                         }
 2291 
 2292                         if (!sbuf_overflowed(sb))
 2293                                 valid_len = sbuf_len(sb);
 2294                 }
 2295                 if (addrs == 0) {
 2296                         bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
 2297                         sbuf_bcat(sb, &ifr, sizeof(ifr));
 2298                         max_len += sizeof(ifr);
 2299 
 2300                         if (!sbuf_overflowed(sb))
 2301                                 valid_len = sbuf_len(sb);
 2302                 }
 2303         }
 2304         IFNET_RUNLOCK();
 2305 
 2306         /*
 2307          * If we didn't allocate enough space (uncommon), try again.  If
 2308          * we have already allocated as much space as we are allowed,
 2309          * return what we've got.
 2310          */
 2311         if (valid_len != max_len && !full) {
 2312                 sbuf_delete(sb);
 2313                 goto again;
 2314         }
 2315 
 2316         ifc->ifc_len = valid_len;
 2317 #ifdef __amd64__
 2318         if (cmd == SIOCGIFCONF32)
 2319                 ifc32->ifc_len = valid_len;
 2320 #endif
 2321         sbuf_finish(sb);
 2322         error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
 2323         sbuf_delete(sb);
 2324         return (error);
 2325 }
 2326 
 2327 /*
 2328  * Just like ifpromisc(), but for all-multicast-reception mode.
 2329  */
 2330 int
 2331 if_allmulti(struct ifnet *ifp, int onswitch)
 2332 {
 2333 
 2334         return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
 2335 }
 2336 
 2337 struct ifmultiaddr *
 2338 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
 2339 {
 2340         struct ifmultiaddr *ifma;
 2341 
 2342         IF_ADDR_LOCK_ASSERT(ifp);
 2343 
 2344         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
 2345                 if (sa->sa_family == AF_LINK) {
 2346                         if (sa_dl_equal(ifma->ifma_addr, sa))
 2347                                 break;
 2348                 } else {
 2349                         if (sa_equal(ifma->ifma_addr, sa))
 2350                                 break;
 2351                 }
 2352         }
 2353 
 2354         return ifma;
 2355 }
 2356 
 2357 /*
 2358  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
 2359  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
 2360  * the ifnet multicast address list here, so the caller must do that and
 2361  * other setup work (such as notifying the device driver).  The reference
 2362  * count is initialized to 1.
 2363  */
 2364 static struct ifmultiaddr *
 2365 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
 2366     int mflags)
 2367 {
 2368         struct ifmultiaddr *ifma;
 2369         struct sockaddr *dupsa;
 2370 
 2371         MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
 2372             M_ZERO);
 2373         if (ifma == NULL)
 2374                 return (NULL);
 2375 
 2376         MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
 2377         if (dupsa == NULL) {
 2378                 FREE(ifma, M_IFMADDR);
 2379                 return (NULL);
 2380         }
 2381         bcopy(sa, dupsa, sa->sa_len);
 2382         ifma->ifma_addr = dupsa;
 2383 
 2384         ifma->ifma_ifp = ifp;
 2385         ifma->ifma_refcount = 1;
 2386         ifma->ifma_protospec = NULL;
 2387 
 2388         if (llsa == NULL) {
 2389                 ifma->ifma_lladdr = NULL;
 2390                 return (ifma);
 2391         }
 2392 
 2393         MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
 2394         if (dupsa == NULL) {
 2395                 FREE(ifma->ifma_addr, M_IFMADDR);
 2396                 FREE(ifma, M_IFMADDR);
 2397                 return (NULL);
 2398         }
 2399         bcopy(llsa, dupsa, llsa->sa_len);
 2400         ifma->ifma_lladdr = dupsa;
 2401 
 2402         return (ifma);
 2403 }
 2404 
 2405 /*
 2406  * if_freemulti: free ifmultiaddr structure and possibly attached related
 2407  * addresses.  The caller is responsible for implementing reference
 2408  * counting, notifying the driver, handling routing messages, and releasing
 2409  * any dependent link layer state.
 2410  */
 2411 static void
 2412 if_freemulti(struct ifmultiaddr *ifma)
 2413 {
 2414 
 2415         KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
 2416             ifma->ifma_refcount));
 2417         KASSERT(ifma->ifma_protospec == NULL,
 2418             ("if_freemulti: protospec not NULL"));
 2419 
 2420         if (ifma->ifma_lladdr != NULL)
 2421                 FREE(ifma->ifma_lladdr, M_IFMADDR);
 2422         FREE(ifma->ifma_addr, M_IFMADDR);
 2423         FREE(ifma, M_IFMADDR);
 2424 }
 2425 
 2426 /*
 2427  * Register an additional multicast address with a network interface.
 2428  *
 2429  * - If the address is already present, bump the reference count on the
 2430  *   address and return.
 2431  * - If the address is not link-layer, look up a link layer address.
 2432  * - Allocate address structures for one or both addresses, and attach to the
 2433  *   multicast address list on the interface.  If automatically adding a link
 2434  *   layer address, the protocol address will own a reference to the link
 2435  *   layer address, to be freed when it is freed.
 2436  * - Notify the network device driver of an addition to the multicast address
 2437  *   list.
 2438  *
 2439  * 'sa' points to caller-owned memory with the desired multicast address.
 2440  *
 2441  * 'retifma' will be used to return a pointer to the resulting multicast
 2442  * address reference, if desired.
 2443  */
 2444 int
 2445 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
 2446     struct ifmultiaddr **retifma)
 2447 {
 2448         struct ifmultiaddr *ifma, *ll_ifma;
 2449         struct sockaddr *llsa;
 2450         int error;
 2451 
 2452         /*
 2453          * If the address is already present, return a new reference to it;
 2454          * otherwise, allocate storage and set up a new address.
 2455          */
 2456         IF_ADDR_LOCK(ifp);
 2457         ifma = if_findmulti(ifp, sa);
 2458         if (ifma != NULL) {
 2459                 ifma->ifma_refcount++;
 2460                 if (retifma != NULL)
 2461                         *retifma = ifma;
 2462                 IF_ADDR_UNLOCK(ifp);
 2463                 return (0);
 2464         }
 2465 
 2466         /*
 2467          * The address isn't already present; resolve the protocol address
 2468          * into a link layer address, and then look that up, bump its
 2469          * refcount or allocate an ifma for that also.  If 'llsa' was
 2470          * returned, we will need to free it later.
 2471          */
 2472         llsa = NULL;
 2473         ll_ifma = NULL;
 2474         if (ifp->if_resolvemulti != NULL) {
 2475                 error = ifp->if_resolvemulti(ifp, &llsa, sa);
 2476                 if (error)
 2477                         goto unlock_out;
 2478         }
 2479 
 2480         /*
 2481          * Allocate the new address.  Don't hook it up yet, as we may also
 2482          * need to allocate a link layer multicast address.
 2483          */
 2484         ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
 2485         if (ifma == NULL) {
 2486                 error = ENOMEM;
 2487                 goto free_llsa_out;
 2488         }
 2489 
 2490         /*
 2491          * If a link layer address is found, we'll need to see if it's
 2492          * already present in the address list, or allocate is as well.
 2493          * When this block finishes, the link layer address will be on the
 2494          * list.
 2495          */
 2496         if (llsa != NULL) {
 2497                 ll_ifma = if_findmulti(ifp, llsa);
 2498                 if (ll_ifma == NULL) {
 2499                         ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
 2500                         if (ll_ifma == NULL) {
 2501                                 --ifma->ifma_refcount;
 2502                                 if_freemulti(ifma);
 2503                                 error = ENOMEM;
 2504                                 goto free_llsa_out;
 2505                         }
 2506                         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
 2507                             ifma_link);
 2508                 } else
 2509                         ll_ifma->ifma_refcount++;
 2510                 ifma->ifma_llifma = ll_ifma;
 2511         }
 2512 
 2513         /*
 2514          * We now have a new multicast address, ifma, and possibly a new or
 2515          * referenced link layer address.  Add the primary address to the
 2516          * ifnet address list.
 2517          */
 2518         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
 2519 
 2520         if (retifma != NULL)
 2521                 *retifma = ifma;
 2522 
 2523         /*
 2524          * Must generate the message while holding the lock so that 'ifma'
 2525          * pointer is still valid.
 2526          */
 2527         rt_newmaddrmsg(RTM_NEWMADDR, ifma);
 2528         IF_ADDR_UNLOCK(ifp);
 2529 
 2530         /*
 2531          * We are certain we have added something, so call down to the
 2532          * interface to let them know about it.
 2533          */
 2534         if (ifp->if_ioctl != NULL) {
 2535                 IFF_LOCKGIANT(ifp);
 2536                 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
 2537                 IFF_UNLOCKGIANT(ifp);
 2538         }
 2539 
 2540         if (llsa != NULL)
 2541                 FREE(llsa, M_IFMADDR);
 2542 
 2543         return (0);
 2544 
 2545 free_llsa_out:
 2546         if (llsa != NULL)
 2547                 FREE(llsa, M_IFMADDR);
 2548 
 2549 unlock_out:
 2550         IF_ADDR_UNLOCK(ifp);
 2551         return (error);
 2552 }
 2553 
 2554 /*
 2555  * Delete a multicast group membership by network-layer group address.
 2556  *
 2557  * Returns ENOENT if the entry could not be found. If ifp no longer
 2558  * exists, results are undefined. This entry point should only be used
 2559  * from subsystems which do appropriate locking to hold ifp for the
 2560  * duration of the call.
 2561  * Network-layer protocol domains must use if_delmulti_ifma().
 2562  */
 2563 int
 2564 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
 2565 {
 2566         struct ifmultiaddr *ifma;
 2567         int lastref;
 2568 #ifdef INVARIANTS
 2569         struct ifnet *oifp;
 2570 
 2571         IFNET_RLOCK();
 2572         TAILQ_FOREACH(oifp, &ifnet, if_link)
 2573                 if (ifp == oifp)
 2574                         break;
 2575         if (ifp != oifp)
 2576                 ifp = NULL;
 2577         IFNET_RUNLOCK();
 2578 
 2579         KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
 2580 #endif
 2581         if (ifp == NULL)
 2582                 return (ENOENT);
 2583 
 2584         IF_ADDR_LOCK(ifp);
 2585         lastref = 0;
 2586         ifma = if_findmulti(ifp, sa);
 2587         if (ifma != NULL)
 2588                 lastref = if_delmulti_locked(ifp, ifma, 0);
 2589         IF_ADDR_UNLOCK(ifp);
 2590 
 2591         if (ifma == NULL)
 2592                 return (ENOENT);
 2593 
 2594         if (lastref && ifp->if_ioctl != NULL) {
 2595                 IFF_LOCKGIANT(ifp);
 2596                 (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 2597                 IFF_UNLOCKGIANT(ifp);
 2598         }
 2599 
 2600         return (0);
 2601 }
 2602 
 2603 /*
 2604  * Delete a multicast group membership by group membership pointer.
 2605  * Network-layer protocol domains must use this routine.
 2606  *
 2607  * It is safe to call this routine if the ifp disappeared. Callers should
 2608  * hold IFF_LOCKGIANT() to avoid a LOR in case the hardware needs to be
 2609  * reconfigured.
 2610  */
 2611 void
 2612 if_delmulti_ifma(struct ifmultiaddr *ifma)
 2613 {
 2614         struct ifnet *ifp;
 2615         int lastref;
 2616 
 2617         ifp = ifma->ifma_ifp;
 2618 #ifdef DIAGNOSTIC
 2619         if (ifp == NULL) {
 2620                 printf("%s: ifma_ifp seems to be detached\n", __func__);
 2621         } else {
 2622                 struct ifnet *oifp;
 2623 
 2624                 IFNET_RLOCK();
 2625                 TAILQ_FOREACH(oifp, &ifnet, if_link)
 2626                         if (ifp == oifp)
 2627                                 break;
 2628                 if (ifp != oifp) {
 2629                         printf("%s: ifnet %p disappeared\n", __func__, ifp);
 2630                         ifp = NULL;
 2631                 }
 2632                 IFNET_RUNLOCK();
 2633         }
 2634 #endif
 2635         /*
 2636          * If and only if the ifnet instance exists: Acquire the address lock.
 2637          */
 2638         if (ifp != NULL)
 2639                 IF_ADDR_LOCK(ifp);
 2640 
 2641         lastref = if_delmulti_locked(ifp, ifma, 0);
 2642 
 2643         if (ifp != NULL) {
 2644                 /*
 2645                  * If and only if the ifnet instance exists:
 2646                  *  Release the address lock.
 2647                  *  If the group was left: update the hardware hash filter.
 2648                  */
 2649                 IF_ADDR_UNLOCK(ifp);
 2650                 if (lastref && ifp->if_ioctl != NULL) {
 2651                         IFF_LOCKGIANT(ifp);
 2652                         (void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 2653                         IFF_UNLOCKGIANT(ifp);
 2654                 }
 2655         }
 2656 }
 2657 
 2658 /*
 2659  * Perform deletion of network-layer and/or link-layer multicast address.
 2660  *
 2661  * Return 0 if the reference count was decremented.
 2662  * Return 1 if the final reference was released, indicating that the
 2663  * hardware hash filter should be reprogrammed.
 2664  */
 2665 static int
 2666 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
 2667 {
 2668         struct ifmultiaddr *ll_ifma;
 2669 
 2670         if (ifp != NULL && ifma->ifma_ifp != NULL) {
 2671                 KASSERT(ifma->ifma_ifp == ifp,
 2672                     ("%s: inconsistent ifp %p", __func__, ifp));
 2673                 IF_ADDR_LOCK_ASSERT(ifp);
 2674         }
 2675 
 2676         ifp = ifma->ifma_ifp;
 2677 
 2678         /*
 2679          * If the ifnet is detaching, null out references to ifnet,
 2680          * so that upper protocol layers will notice, and not attempt
 2681          * to obtain locks for an ifnet which no longer exists. The
 2682          * routing socket announcement must happen before the ifnet
 2683          * instance is detached from the system.
 2684          */
 2685         if (detaching) {
 2686 #ifdef DIAGNOSTIC
 2687                 printf("%s: detaching ifnet instance %p\n", __func__, ifp);
 2688 #endif
 2689                 /*
 2690                  * ifp may already be nulled out if we are being reentered
 2691                  * to delete the ll_ifma.
 2692                  */
 2693                 if (ifp != NULL) {
 2694                         rt_newmaddrmsg(RTM_DELMADDR, ifma);
 2695                         ifma->ifma_ifp = NULL;
 2696                 }
 2697         }
 2698 
 2699         if (--ifma->ifma_refcount > 0)
 2700                 return 0;
 2701 
 2702         /*
 2703          * If this ifma is a network-layer ifma, a link-layer ifma may
 2704          * have been associated with it. Release it first if so.
 2705          */
 2706         ll_ifma = ifma->ifma_llifma;
 2707         if (ll_ifma != NULL) {
 2708                 KASSERT(ifma->ifma_lladdr != NULL,
 2709                     ("%s: llifma w/o lladdr", __func__));
 2710                 if (detaching)
 2711                         ll_ifma->ifma_ifp = NULL;       /* XXX */
 2712                 if (--ll_ifma->ifma_refcount == 0) {
 2713                         if (ifp != NULL) {
 2714                                 TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
 2715                                     ifma_link);
 2716                         }
 2717                         if_freemulti(ll_ifma);
 2718                 }
 2719         }
 2720 
 2721         if (ifp != NULL)
 2722                 TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
 2723 
 2724         if_freemulti(ifma);
 2725 
 2726         /*
 2727          * The last reference to this instance of struct ifmultiaddr
 2728          * was released; the hardware should be notified of this change.
 2729          */
 2730         return 1;
 2731 }
 2732 
 2733 /*
 2734  * Set the link layer address on an interface.
 2735  *
 2736  * At this time we only support certain types of interfaces,
 2737  * and we don't allow the length of the address to change.
 2738  */
 2739 int
 2740 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
 2741 {
 2742         struct sockaddr_dl *sdl;
 2743         struct ifaddr *ifa;
 2744         struct ifreq ifr;
 2745 
 2746         ifa = ifp->if_addr;
 2747         if (ifa == NULL)
 2748                 return (EINVAL);
 2749         sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 2750         if (sdl == NULL)
 2751                 return (EINVAL);
 2752         if (len != sdl->sdl_alen)       /* don't allow length to change */
 2753                 return (EINVAL);
 2754         switch (ifp->if_type) {
 2755         case IFT_ETHER:
 2756         case IFT_FDDI:
 2757         case IFT_XETHER:
 2758         case IFT_ISO88025:
 2759         case IFT_L2VLAN:
 2760         case IFT_BRIDGE:
 2761         case IFT_ARCNET:
 2762         case IFT_IEEE8023ADLAG:
 2763                 bcopy(lladdr, LLADDR(sdl), len);
 2764                 break;
 2765         default:
 2766                 return (ENODEV);
 2767         }
 2768         /*
 2769          * If the interface is already up, we need
 2770          * to re-init it in order to reprogram its
 2771          * address filter.
 2772          */
 2773         if ((ifp->if_flags & IFF_UP) != 0) {
 2774                 if (ifp->if_ioctl) {
 2775                         IFF_LOCKGIANT(ifp);
 2776                         ifp->if_flags &= ~IFF_UP;
 2777                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2778                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2779                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2780                         ifp->if_flags |= IFF_UP;
 2781                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2782                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2783                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2784                         IFF_UNLOCKGIANT(ifp);
 2785                 }
 2786 #ifdef INET
 2787                 /*
 2788                  * Also send gratuitous ARPs to notify other nodes about
 2789                  * the address change.
 2790                  */
 2791                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2792                         if (ifa->ifa_addr->sa_family == AF_INET)
 2793                                 arp_ifinit(ifp, ifa);
 2794                 }
 2795 #endif
 2796         }
 2797         return (0);
 2798 }
 2799 
 2800 /*
 2801  * The name argument must be a pointer to storage which will last as
 2802  * long as the interface does.  For physical devices, the result of
 2803  * device_get_name(dev) is a good choice and for pseudo-devices a
 2804  * static string works well.
 2805  */
 2806 void
 2807 if_initname(struct ifnet *ifp, const char *name, int unit)
 2808 {
 2809         ifp->if_dname = name;
 2810         ifp->if_dunit = unit;
 2811         if (unit != IF_DUNIT_NONE)
 2812                 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
 2813         else
 2814                 strlcpy(ifp->if_xname, name, IFNAMSIZ);
 2815 }
 2816 
 2817 int
 2818 if_printf(struct ifnet *ifp, const char * fmt, ...)
 2819 {
 2820         va_list ap;
 2821         int retval;
 2822 
 2823         retval = printf("%s: ", ifp->if_xname);
 2824         va_start(ap, fmt);
 2825         retval += vprintf(fmt, ap);
 2826         va_end(ap);
 2827         return (retval);
 2828 }
 2829 
 2830 /*
 2831  * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
 2832  * be called without Giant.  However, we often can't acquire the Giant lock
 2833  * at those points; instead, we run it via a task queue that holds Giant via
 2834  * if_start_deferred.
 2835  *
 2836  * XXXRW: We need to make sure that the ifnet isn't fully detached until any
 2837  * outstanding if_start_deferred() tasks that will run after the free.  This
 2838  * probably means waiting in if_detach().
 2839  */
 2840 void
 2841 if_start(struct ifnet *ifp)
 2842 {
 2843 
 2844         if (ifp->if_flags & IFF_NEEDSGIANT) {
 2845                 if (mtx_owned(&Giant))
 2846                         (*(ifp)->if_start)(ifp);
 2847                 else
 2848                         taskqueue_enqueue(taskqueue_swi_giant,
 2849                             &ifp->if_starttask);
 2850         } else
 2851                 (*(ifp)->if_start)(ifp);
 2852 }
 2853 
 2854 static void
 2855 if_start_deferred(void *context, int pending)
 2856 {
 2857         struct ifnet *ifp;
 2858 
 2859         GIANT_REQUIRED;
 2860 
 2861         ifp = context;
 2862         (ifp->if_start)(ifp);
 2863 }
 2864 
 2865 int
 2866 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
 2867 {
 2868         int active = 0;
 2869 
 2870         IF_LOCK(ifq);
 2871         if (_IF_QFULL(ifq)) {
 2872                 _IF_DROP(ifq);
 2873                 IF_UNLOCK(ifq);
 2874                 m_freem(m);
 2875                 return (0);
 2876         }
 2877         if (ifp != NULL) {
 2878                 ifp->if_obytes += m->m_pkthdr.len + adjust;
 2879                 if (m->m_flags & (M_BCAST|M_MCAST))
 2880                         ifp->if_omcasts++;
 2881                 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
 2882         }
 2883         _IF_ENQUEUE(ifq, m);
 2884         IF_UNLOCK(ifq);
 2885         if (ifp != NULL && !active)
 2886                 if_start(ifp);
 2887         return (1);
 2888 }
 2889 
 2890 void
 2891 if_register_com_alloc(u_char type,
 2892     if_com_alloc_t *a, if_com_free_t *f)
 2893 {
 2894         
 2895         KASSERT(if_com_alloc[type] == NULL,
 2896             ("if_register_com_alloc: %d already registered", type));
 2897         KASSERT(if_com_free[type] == NULL,
 2898             ("if_register_com_alloc: %d free already registered", type));
 2899 
 2900         if_com_alloc[type] = a;
 2901         if_com_free[type] = f;
 2902 }
 2903 
 2904 void
 2905 if_deregister_com_alloc(u_char type)
 2906 {
 2907         
 2908         KASSERT(if_com_alloc[type] != NULL,
 2909             ("if_deregister_com_alloc: %d not registered", type));
 2910         KASSERT(if_com_free[type] != NULL,
 2911             ("if_deregister_com_alloc: %d free not registered", type));
 2912         if_com_alloc[type] = NULL;
 2913         if_com_free[type] = NULL;
 2914 }

Cache object: 939403a718a4665adfe1e8b66c33b3df


[ 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.