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/6.1/sys/net/if.c 155707 2006-02-15 03:37:15Z ps $
   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/mac.h>
   43 #include <sys/malloc.h>
   44 #include <sys/sbuf.h>
   45 #include <sys/bus.h>
   46 #include <sys/mbuf.h>
   47 #include <sys/systm.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_arp.h>
   63 #include <net/if_clone.h>
   64 #include <net/if_dl.h>
   65 #include <net/if_types.h>
   66 #include <net/if_var.h>
   67 #include <net/radix.h>
   68 #include <net/route.h>
   69 
   70 #if defined(INET) || defined(INET6)
   71 /*XXX*/
   72 #include <netinet/in.h>
   73 #include <netinet/in_var.h>
   74 #ifdef INET6
   75 #include <netinet6/in6_var.h>
   76 #include <netinet6/in6_ifattach.h>
   77 #endif
   78 #endif
   79 #ifdef INET
   80 #include <netinet/if_ether.h>
   81 #endif
   82 #ifdef DEV_CARP
   83 #include <netinet/ip_carp.h>
   84 #endif
   85 
   86 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
   87 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
   88 
   89 /* Log link state change events */
   90 static int log_link_state_change = 1;
   91 
   92 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
   93         &log_link_state_change, 0,
   94         "log interface link state change events");
   95 
   96 void    (*bstp_linkstate_p)(struct ifnet *ifp, int state);
   97 void    (*ng_ether_link_state_p)(struct ifnet *ifp, int state);
   98 
   99 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
  100 
  101 static void     if_attachdomain(void *);
  102 static void     if_attachdomain1(struct ifnet *);
  103 static int      ifconf(u_long, caddr_t);
  104 static void     if_grow(void);
  105 static void     if_init(void *);
  106 static void     if_check(void *);
  107 static void     if_qflush(struct ifaltq *);
  108 static void     if_route(struct ifnet *, int flag, int fam);
  109 static int      if_setflag(struct ifnet *, int, int, int *, int);
  110 static void     if_slowtimo(void *);
  111 static void     if_unroute(struct ifnet *, int flag, int fam);
  112 static void     link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  113 static int      if_rtdel(struct radix_node *, void *);
  114 static int      ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
  115 static void     if_start_deferred(void *context, int pending);
  116 static void     do_link_state_change(void *, int);
  117 #ifdef INET6
  118 /*
  119  * XXX: declare here to avoid to include many inet6 related files..
  120  * should be more generalized?
  121  */
  122 extern void     nd6_setmtu(struct ifnet *);
  123 #endif
  124 
  125 int     if_index = 0;
  126 struct  ifindex_entry *ifindex_table = NULL;
  127 int     ifqmaxlen = IFQ_MAXLEN;
  128 struct  ifnethead ifnet;        /* depend on static init XXX */
  129 struct  mtx ifnet_lock;
  130 static  if_com_alloc_t *if_com_alloc[256];
  131 static  if_com_free_t *if_com_free[256];
  132 
  133 static int      if_indexlim = 8;
  134 static struct   knlist ifklist;
  135 
  136 static void     filt_netdetach(struct knote *kn);
  137 static int      filt_netdev(struct knote *kn, long hint);
  138 
  139 static struct filterops netdev_filtops =
  140     { 1, NULL, filt_netdetach, filt_netdev };
  141 
  142 /*
  143  * System initialization
  144  */
  145 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
  146 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
  147 
  148 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
  149 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
  150 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
  151 
  152 static d_open_t         netopen;
  153 static d_close_t        netclose;
  154 static d_ioctl_t        netioctl;
  155 static d_kqfilter_t     netkqfilter;
  156 
  157 static struct cdevsw net_cdevsw = {
  158         .d_version =    D_VERSION,
  159         .d_flags =      D_NEEDGIANT,
  160         .d_open =       netopen,
  161         .d_close =      netclose,
  162         .d_ioctl =      netioctl,
  163         .d_name =       "net",
  164         .d_kqfilter =   netkqfilter,
  165 };
  166 
  167 static int
  168 netopen(struct cdev *dev, int flag, int mode, struct thread *td)
  169 {
  170         return (0);
  171 }
  172 
  173 static int
  174 netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
  175 {
  176         return (0);
  177 }
  178 
  179 static int
  180 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
  181 {
  182         struct ifnet *ifp;
  183         int error, idx;
  184 
  185         /* only support interface specific ioctls */
  186         if (IOCGROUP(cmd) != 'i')
  187                 return (EOPNOTSUPP);
  188         idx = minor(dev);
  189         if (idx == 0) {
  190                 /*
  191                  * special network device, not interface.
  192                  */
  193                 if (cmd == SIOCGIFCONF)
  194                         return (ifconf(cmd, data));     /* XXX remove cmd */
  195 #ifdef __amd64__
  196                 if (cmd == SIOCGIFCONF32)
  197                         return (ifconf(cmd, data));     /* XXX remove cmd */
  198 #endif
  199                 return (EOPNOTSUPP);
  200         }
  201 
  202         ifp = ifnet_byindex(idx);
  203         if (ifp == NULL)
  204                 return (ENXIO);
  205 
  206         error = ifhwioctl(cmd, ifp, data, td);
  207         if (error == ENOIOCTL)
  208                 error = EOPNOTSUPP;
  209         return (error);
  210 }
  211 
  212 static int
  213 netkqfilter(struct cdev *dev, struct knote *kn)
  214 {
  215         struct knlist *klist;
  216         struct ifnet *ifp;
  217         int idx;
  218 
  219         switch (kn->kn_filter) {
  220         case EVFILT_NETDEV:
  221                 kn->kn_fop = &netdev_filtops;
  222                 break;
  223         default:
  224                 return (EINVAL);
  225         }
  226 
  227         idx = minor(dev);
  228         if (idx == 0) {
  229                 klist = &ifklist;
  230         } else {
  231                 ifp = ifnet_byindex(idx);
  232                 if (ifp == NULL)
  233                         return (1);
  234                 klist = &ifp->if_klist;
  235         }
  236 
  237         kn->kn_hook = (caddr_t)klist;
  238 
  239         knlist_add(klist, kn, 0);
  240 
  241         return (0);
  242 }
  243 
  244 static void
  245 filt_netdetach(struct knote *kn)
  246 {
  247         struct knlist *klist = (struct knlist *)kn->kn_hook;
  248 
  249         knlist_remove(klist, kn, 0);
  250 }
  251 
  252 static int
  253 filt_netdev(struct knote *kn, long hint)
  254 {
  255         struct knlist *klist = (struct knlist *)kn->kn_hook;
  256 
  257         /*
  258          * Currently NOTE_EXIT is abused to indicate device detach.
  259          */
  260         if (hint == NOTE_EXIT) {
  261                 kn->kn_data = NOTE_LINKINV;
  262                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
  263                 knlist_remove_inevent(klist, kn);
  264                 return (1);
  265         }
  266         if (hint != 0)
  267                 kn->kn_data = hint;                     /* current status */
  268         if (kn->kn_sfflags & hint)
  269                 kn->kn_fflags |= hint;
  270         return (kn->kn_fflags != 0);
  271 }
  272 
  273 /*
  274  * Network interface utility routines.
  275  *
  276  * Routines with ifa_ifwith* names take sockaddr *'s as
  277  * parameters.
  278  */
  279 /* ARGSUSED*/
  280 static void
  281 if_init(void *dummy __unused)
  282 {
  283 
  284         IFNET_LOCK_INIT();
  285         TAILQ_INIT(&ifnet);
  286         knlist_init(&ifklist, NULL, NULL, NULL, NULL);
  287         if_grow();                              /* create initial table */
  288         ifdev_byindex(0) = make_dev(&net_cdevsw, 0,
  289             UID_ROOT, GID_WHEEL, 0600, "network");
  290         if_clone_init();
  291 }
  292 
  293 static void
  294 if_grow(void)
  295 {
  296         u_int n;
  297         struct ifindex_entry *e;
  298 
  299         if_indexlim <<= 1;
  300         n = if_indexlim * sizeof(*e);
  301         e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
  302         if (ifindex_table != NULL) {
  303                 memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
  304                 free((caddr_t)ifindex_table, M_IFNET);
  305         }
  306         ifindex_table = e;
  307 }
  308 
  309 /* ARGSUSED*/
  310 static void
  311 if_check(void *dummy __unused)
  312 {
  313         struct ifnet *ifp;
  314         int s;
  315 
  316         s = splimp();
  317         IFNET_RLOCK();  /* could sleep on rare error; mostly okay XXX */
  318         TAILQ_FOREACH(ifp, &ifnet, if_link) {
  319                 if (ifp->if_snd.ifq_maxlen == 0) {
  320                         if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
  321                         ifp->if_snd.ifq_maxlen = ifqmaxlen;
  322                 }
  323                 if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
  324                         if_printf(ifp,
  325                             "XXX: driver didn't initialize queue mtx\n");
  326                         mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
  327                             MTX_NETWORK_LOCK, MTX_DEF);
  328                 }
  329         }
  330         IFNET_RUNLOCK();
  331         splx(s);
  332         if_slowtimo(0);
  333 }
  334 
  335 /*
  336  * Allocate a struct ifnet and in index for an interface.
  337  */
  338 struct ifnet*
  339 if_alloc(u_char type)
  340 {
  341         struct ifnet *ifp;
  342 
  343         ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
  344 
  345         /*
  346          * Try to find an empty slot below if_index.  If we fail, take
  347          * the next slot.
  348          *
  349          * XXX: should be locked!
  350          */
  351         for (ifp->if_index = 1; ifp->if_index <= if_index; ifp->if_index++) {
  352                 if (ifnet_byindex(ifp->if_index) == NULL)
  353                         break;
  354         }
  355         /* Catch if_index overflow. */
  356         if (ifp->if_index < 1) {
  357                 free(ifp, M_IFNET);
  358                 return (NULL);
  359         }
  360         if (ifp->if_index > if_index)
  361                 if_index = ifp->if_index;
  362         if (if_index >= if_indexlim)
  363                 if_grow();
  364         ifnet_byindex(ifp->if_index) = ifp;
  365 
  366         ifp->if_type = type;
  367 
  368         if (if_com_alloc[type] != NULL) {
  369                 ifp->if_l2com = if_com_alloc[type](type, ifp);
  370                 if (ifp->if_l2com == NULL) {
  371                         free(ifp, M_IFNET);
  372                         return (NULL);
  373                 }
  374         }
  375         IF_ADDR_LOCK_INIT(ifp);
  376 
  377         return (ifp);
  378 }
  379 
  380 void
  381 if_free(struct ifnet *ifp)
  382 {
  383 
  384         /* Do not add code to this function!  Add it to if_free_type(). */
  385         if_free_type(ifp, ifp->if_type);
  386 }
  387 
  388 void
  389 if_free_type(struct ifnet *ifp, u_char type)
  390 {
  391 
  392         if (ifp != ifnet_byindex(ifp->if_index)) {
  393                 if_printf(ifp, "%s: value was not if_alloced, skipping\n",
  394                     __func__);
  395                 return;
  396         }
  397 
  398         IF_ADDR_LOCK_DESTROY(ifp);
  399 
  400         ifnet_byindex(ifp->if_index) = NULL;
  401 
  402         /* XXX: should be locked with if_findindex() */
  403         while (if_index > 0 && ifnet_byindex(if_index) == NULL)
  404                 if_index--;
  405 
  406         if (if_com_free[type] != NULL)
  407                 if_com_free[type](ifp->if_l2com, type);
  408 
  409         free(ifp, M_IFNET);
  410 };
  411 
  412 /*
  413  * Attach an interface to the
  414  * list of "active" interfaces.
  415  */
  416 void
  417 if_attach(struct ifnet *ifp)
  418 {
  419         unsigned socksize, ifasize;
  420         int namelen, masklen;
  421         struct sockaddr_dl *sdl;
  422         struct ifaddr *ifa;
  423 
  424         if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
  425                 panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
  426                     ifp->if_xname);
  427 
  428         TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
  429         TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
  430         IF_AFDATA_LOCK_INIT(ifp);
  431         ifp->if_afdata_initialized = 0;
  432         IFNET_WLOCK();
  433         TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
  434         IFNET_WUNLOCK();
  435         /*
  436          * XXX -
  437          * The old code would work if the interface passed a pre-existing
  438          * chain of ifaddrs to this code.  We don't trust our callers to
  439          * properly initialize the tailq, however, so we no longer allow
  440          * this unlikely case.
  441          */
  442         TAILQ_INIT(&ifp->if_addrhead);
  443         TAILQ_INIT(&ifp->if_prefixhead);
  444         TAILQ_INIT(&ifp->if_multiaddrs);
  445         knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
  446         getmicrotime(&ifp->if_lastchange);
  447         ifp->if_data.ifi_epoch = time_uptime;
  448         ifp->if_data.ifi_datalen = sizeof(struct if_data);
  449 
  450 #ifdef MAC
  451         mac_init_ifnet(ifp);
  452         mac_create_ifnet(ifp);
  453 #endif
  454 
  455         ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw,
  456             unit2minor(ifp->if_index),
  457             UID_ROOT, GID_WHEEL, 0600, "%s/%s",
  458             net_cdevsw.d_name, ifp->if_xname);
  459         make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
  460             net_cdevsw.d_name, ifp->if_index);
  461 
  462         mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
  463 
  464         /*
  465          * create a Link Level name for this device
  466          */
  467         namelen = strlen(ifp->if_xname);
  468         /*
  469          * Always save enough space for any possiable name so we can do
  470          * a rename in place later.
  471          */
  472         masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
  473         socksize = masklen + ifp->if_addrlen;
  474         if (socksize < sizeof(*sdl))
  475                 socksize = sizeof(*sdl);
  476         socksize = roundup2(socksize, sizeof(long));
  477         ifasize = sizeof(*ifa) + 2 * socksize;
  478         ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
  479         IFA_LOCK_INIT(ifa);
  480         sdl = (struct sockaddr_dl *)(ifa + 1);
  481         sdl->sdl_len = socksize;
  482         sdl->sdl_family = AF_LINK;
  483         bcopy(ifp->if_xname, sdl->sdl_data, namelen);
  484         sdl->sdl_nlen = namelen;
  485         sdl->sdl_index = ifp->if_index;
  486         sdl->sdl_type = ifp->if_type;
  487         ifaddr_byindex(ifp->if_index) = ifa;
  488         ifa->ifa_ifp = ifp;
  489         ifa->ifa_rtrequest = link_rtrequest;
  490         ifa->ifa_addr = (struct sockaddr *)sdl;
  491         sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
  492         ifa->ifa_netmask = (struct sockaddr *)sdl;
  493         sdl->sdl_len = masklen;
  494         while (namelen != 0)
  495                 sdl->sdl_data[--namelen] = 0xff;
  496         ifa->ifa_refcnt = 1;
  497         TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
  498         ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
  499         ifp->if_snd.altq_type = 0;
  500         ifp->if_snd.altq_disc = NULL;
  501         ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
  502         ifp->if_snd.altq_tbr  = NULL;
  503         ifp->if_snd.altq_ifp  = ifp;
  504 
  505         if (domain_init_status >= 2)
  506                 if_attachdomain1(ifp);
  507 
  508         EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
  509 
  510         /* Announce the interface. */
  511         rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
  512 }
  513 
  514 static void
  515 if_attachdomain(void *dummy)
  516 {
  517         struct ifnet *ifp;
  518         int s;
  519 
  520         s = splnet();
  521         TAILQ_FOREACH(ifp, &ifnet, if_link)
  522                 if_attachdomain1(ifp);
  523         splx(s);
  524 }
  525 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
  526     if_attachdomain, NULL);
  527 
  528 static void
  529 if_attachdomain1(struct ifnet *ifp)
  530 {
  531         struct domain *dp;
  532         int s;
  533 
  534         s = splnet();
  535 
  536         /*
  537          * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
  538          * cannot lock ifp->if_afdata initialization, entirely.
  539          */
  540         if (IF_AFDATA_TRYLOCK(ifp) == 0) {
  541                 splx(s);
  542                 return;
  543         }
  544         if (ifp->if_afdata_initialized >= domain_init_status) {
  545                 IF_AFDATA_UNLOCK(ifp);
  546                 splx(s);
  547                 printf("if_attachdomain called more than once on %s\n",
  548                     ifp->if_xname);
  549                 return;
  550         }
  551         ifp->if_afdata_initialized = domain_init_status;
  552         IF_AFDATA_UNLOCK(ifp);
  553 
  554         /* address family dependent data region */
  555         bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
  556         for (dp = domains; dp; dp = dp->dom_next) {
  557                 if (dp->dom_ifattach)
  558                         ifp->if_afdata[dp->dom_family] =
  559                             (*dp->dom_ifattach)(ifp);
  560         }
  561 
  562         splx(s);
  563 }
  564 
  565 /*
  566  * Remove any network addresses from an interface.
  567  */
  568 
  569 void
  570 if_purgeaddrs(struct ifnet *ifp)
  571 {
  572         struct ifaddr *ifa, *next;
  573 
  574         TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  575 
  576                 if (ifa->ifa_addr->sa_family == AF_LINK)
  577                         continue;
  578 #ifdef INET
  579                 /* XXX: Ugly!! ad hoc just for INET */
  580                 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
  581                         struct ifaliasreq ifr;
  582 
  583                         bzero(&ifr, sizeof(ifr));
  584                         ifr.ifra_addr = *ifa->ifa_addr;
  585                         if (ifa->ifa_dstaddr)
  586                                 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
  587                         if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
  588                             NULL) == 0)
  589                                 continue;
  590                 }
  591 #endif /* INET */
  592 #ifdef INET6
  593                 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
  594                         in6_purgeaddr(ifa);
  595                         /* ifp_addrhead is already updated */
  596                         continue;
  597                 }
  598 #endif /* INET6 */
  599                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  600                 IFAFREE(ifa);
  601         }
  602 }
  603 
  604 /*
  605  * Detach an interface, removing it from the
  606  * list of "active" interfaces and freeing the struct ifnet.
  607  *
  608  * XXXRW: There are some significant questions about event ordering, and
  609  * how to prevent things from starting to use the interface during detach.
  610  */
  611 void
  612 if_detach(struct ifnet *ifp)
  613 {
  614         struct ifaddr *ifa;
  615         struct radix_node_head  *rnh;
  616         int s;
  617         int i;
  618         struct domain *dp;
  619         struct ifnet *iter;
  620         int found;
  621 
  622         /*
  623          * Remove/wait for pending events.
  624          */
  625         taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
  626 
  627 #ifdef DEV_CARP
  628         /* Maybe hook to the generalized departure handler above?!? */
  629         if (ifp->if_carp)
  630                 carp_ifdetach(ifp);
  631 #endif
  632 
  633         /*
  634          * Remove routes and flush queues.
  635          */
  636         s = splnet();
  637         if_down(ifp);
  638 #ifdef ALTQ
  639         if (ALTQ_IS_ENABLED(&ifp->if_snd))
  640                 altq_disable(&ifp->if_snd);
  641         if (ALTQ_IS_ATTACHED(&ifp->if_snd))
  642                 altq_detach(&ifp->if_snd);
  643 #endif
  644 
  645         if_purgeaddrs(ifp);
  646 
  647 #ifdef INET
  648         in_ifdetach(ifp);
  649 #endif
  650 
  651 #ifdef INET6
  652         /*
  653          * Remove all IPv6 kernel structs related to ifp.  This should be done
  654          * before removing routing entries below, since IPv6 interface direct
  655          * routes are expected to be removed by the IPv6-specific kernel API.
  656          * Otherwise, the kernel will detect some inconsistency and bark it.
  657          */
  658         in6_ifdetach(ifp);
  659 #endif
  660         /*
  661          * Remove address from ifindex_table[] and maybe decrement if_index.
  662          * Clean up all addresses.
  663          */
  664         ifaddr_byindex(ifp->if_index) = NULL;
  665         destroy_dev(ifdev_byindex(ifp->if_index));
  666         ifdev_byindex(ifp->if_index) = NULL;
  667 
  668         /* We can now free link ifaddr. */
  669         if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
  670                 ifa = TAILQ_FIRST(&ifp->if_addrhead);
  671                 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
  672                 IFAFREE(ifa);
  673         }
  674 
  675         /*
  676          * Delete all remaining routes using this interface
  677          * Unfortuneatly the only way to do this is to slog through
  678          * the entire routing table looking for routes which point
  679          * to this interface...oh well...
  680          */
  681         for (i = 1; i <= AF_MAX; i++) {
  682                 if ((rnh = rt_tables[i]) == NULL)
  683                         continue;
  684                 RADIX_NODE_HEAD_LOCK(rnh);
  685                 (void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
  686                 RADIX_NODE_HEAD_UNLOCK(rnh);
  687         }
  688 
  689         /* Announce that the interface is gone. */
  690         rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
  691         EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
  692 
  693         IF_AFDATA_LOCK(ifp);
  694         for (dp = domains; dp; dp = dp->dom_next) {
  695                 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
  696                         (*dp->dom_ifdetach)(ifp,
  697                             ifp->if_afdata[dp->dom_family]);
  698         }
  699         IF_AFDATA_UNLOCK(ifp);
  700 
  701 #ifdef MAC
  702         mac_destroy_ifnet(ifp);
  703 #endif /* MAC */
  704         KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
  705         knlist_clear(&ifp->if_klist, 0);
  706         knlist_destroy(&ifp->if_klist);
  707         IFNET_WLOCK();
  708         found = 0;
  709         TAILQ_FOREACH(iter, &ifnet, if_link)
  710                 if (iter == ifp) {
  711                         found = 1;
  712                         break;
  713                 }
  714         if (found)
  715                 TAILQ_REMOVE(&ifnet, ifp, if_link);
  716         IFNET_WUNLOCK();
  717         mtx_destroy(&ifp->if_snd.ifq_mtx);
  718         IF_AFDATA_DESTROY(ifp);
  719         splx(s);
  720 }
  721 
  722 /*
  723  * Delete Routes for a Network Interface
  724  *
  725  * Called for each routing entry via the rnh->rnh_walktree() call above
  726  * to delete all route entries referencing a detaching network interface.
  727  *
  728  * Arguments:
  729  *      rn      pointer to node in the routing table
  730  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
  731  *
  732  * Returns:
  733  *      0       successful
  734  *      errno   failed - reason indicated
  735  *
  736  */
  737 static int
  738 if_rtdel(struct radix_node *rn, void *arg)
  739 {
  740         struct rtentry  *rt = (struct rtentry *)rn;
  741         struct ifnet    *ifp = arg;
  742         int             err;
  743 
  744         if (rt->rt_ifp == ifp) {
  745 
  746                 /*
  747                  * Protect (sorta) against walktree recursion problems
  748                  * with cloned routes
  749                  */
  750                 if ((rt->rt_flags & RTF_UP) == 0)
  751                         return (0);
  752 
  753                 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
  754                                 rt_mask(rt), rt->rt_flags,
  755                                 (struct rtentry **) NULL);
  756                 if (err) {
  757                         log(LOG_WARNING, "if_rtdel: error %d\n", err);
  758                 }
  759         }
  760 
  761         return (0);
  762 }
  763 
  764 #define sa_equal(a1, a2)        (bcmp((a1), (a2), ((a1))->sa_len) == 0)
  765 
  766 /*
  767  * Locate an interface based on a complete address.
  768  */
  769 /*ARGSUSED*/
  770 struct ifaddr *
  771 ifa_ifwithaddr(struct sockaddr *addr)
  772 {
  773         struct ifnet *ifp;
  774         struct ifaddr *ifa;
  775 
  776         IFNET_RLOCK();
  777         TAILQ_FOREACH(ifp, &ifnet, if_link)
  778                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  779                         if (ifa->ifa_addr->sa_family != addr->sa_family)
  780                                 continue;
  781                         if (sa_equal(addr, ifa->ifa_addr))
  782                                 goto done;
  783                         /* IP6 doesn't have broadcast */
  784                         if ((ifp->if_flags & IFF_BROADCAST) &&
  785                             ifa->ifa_broadaddr &&
  786                             ifa->ifa_broadaddr->sa_len != 0 &&
  787                             sa_equal(ifa->ifa_broadaddr, addr))
  788                                 goto done;
  789                 }
  790         ifa = NULL;
  791 done:
  792         IFNET_RUNLOCK();
  793         return (ifa);
  794 }
  795 
  796 /*
  797  * Locate the point to point interface with a given destination address.
  798  */
  799 /*ARGSUSED*/
  800 struct ifaddr *
  801 ifa_ifwithdstaddr(struct sockaddr *addr)
  802 {
  803         struct ifnet *ifp;
  804         struct ifaddr *ifa;
  805 
  806         IFNET_RLOCK();
  807         TAILQ_FOREACH(ifp, &ifnet, if_link) {
  808                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
  809                         continue;
  810                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  811                         if (ifa->ifa_addr->sa_family != addr->sa_family)
  812                                 continue;
  813                         if (ifa->ifa_dstaddr &&
  814                             sa_equal(addr, ifa->ifa_dstaddr))
  815                                 goto done;
  816                 }
  817         }
  818         ifa = NULL;
  819 done:
  820         IFNET_RUNLOCK();
  821         return (ifa);
  822 }
  823 
  824 /*
  825  * Find an interface on a specific network.  If many, choice
  826  * is most specific found.
  827  */
  828 struct ifaddr *
  829 ifa_ifwithnet(struct sockaddr *addr)
  830 {
  831         struct ifnet *ifp;
  832         struct ifaddr *ifa;
  833         struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
  834         u_int af = addr->sa_family;
  835         char *addr_data = addr->sa_data, *cplim;
  836 
  837         /*
  838          * AF_LINK addresses can be looked up directly by their index number,
  839          * so do that if we can.
  840          */
  841         if (af == AF_LINK) {
  842             struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
  843             if (sdl->sdl_index && sdl->sdl_index <= if_index)
  844                 return (ifaddr_byindex(sdl->sdl_index));
  845         }
  846 
  847         /*
  848          * Scan though each interface, looking for ones that have
  849          * addresses in this address family.
  850          */
  851         IFNET_RLOCK();
  852         TAILQ_FOREACH(ifp, &ifnet, if_link) {
  853                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  854                         char *cp, *cp2, *cp3;
  855 
  856                         if (ifa->ifa_addr->sa_family != af)
  857 next:                           continue;
  858                         if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
  859                                 /*
  860                                  * This is a bit broken as it doesn't
  861                                  * take into account that the remote end may
  862                                  * be a single node in the network we are
  863                                  * looking for.
  864                                  * The trouble is that we don't know the
  865                                  * netmask for the remote end.
  866                                  */
  867                                 if (ifa->ifa_dstaddr != 0 &&
  868                                     sa_equal(addr, ifa->ifa_dstaddr))
  869                                         goto done;
  870                         } else {
  871                                 /*
  872                                  * if we have a special address handler,
  873                                  * then use it instead of the generic one.
  874                                  */
  875                                 if (ifa->ifa_claim_addr) {
  876                                         if ((*ifa->ifa_claim_addr)(ifa, addr))
  877                                                 goto done;
  878                                         continue;
  879                                 }
  880 
  881                                 /*
  882                                  * Scan all the bits in the ifa's address.
  883                                  * If a bit dissagrees with what we are
  884                                  * looking for, mask it with the netmask
  885                                  * to see if it really matters.
  886                                  * (A byte at a time)
  887                                  */
  888                                 if (ifa->ifa_netmask == 0)
  889                                         continue;
  890                                 cp = addr_data;
  891                                 cp2 = ifa->ifa_addr->sa_data;
  892                                 cp3 = ifa->ifa_netmask->sa_data;
  893                                 cplim = ifa->ifa_netmask->sa_len
  894                                         + (char *)ifa->ifa_netmask;
  895                                 while (cp3 < cplim)
  896                                         if ((*cp++ ^ *cp2++) & *cp3++)
  897                                                 goto next; /* next address! */
  898                                 /*
  899                                  * If the netmask of what we just found
  900                                  * is more specific than what we had before
  901                                  * (if we had one) then remember the new one
  902                                  * before continuing to search
  903                                  * for an even better one.
  904                                  */
  905                                 if (ifa_maybe == 0 ||
  906                                     rn_refines((caddr_t)ifa->ifa_netmask,
  907                                     (caddr_t)ifa_maybe->ifa_netmask))
  908                                         ifa_maybe = ifa;
  909                         }
  910                 }
  911         }
  912         ifa = ifa_maybe;
  913 done:
  914         IFNET_RUNLOCK();
  915         return (ifa);
  916 }
  917 
  918 /*
  919  * Find an interface address specific to an interface best matching
  920  * a given address.
  921  */
  922 struct ifaddr *
  923 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
  924 {
  925         struct ifaddr *ifa;
  926         char *cp, *cp2, *cp3;
  927         char *cplim;
  928         struct ifaddr *ifa_maybe = 0;
  929         u_int af = addr->sa_family;
  930 
  931         if (af >= AF_MAX)
  932                 return (0);
  933         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  934                 if (ifa->ifa_addr->sa_family != af)
  935                         continue;
  936                 if (ifa_maybe == 0)
  937                         ifa_maybe = ifa;
  938                 if (ifa->ifa_netmask == 0) {
  939                         if (sa_equal(addr, ifa->ifa_addr) ||
  940                             (ifa->ifa_dstaddr &&
  941                             sa_equal(addr, ifa->ifa_dstaddr)))
  942                                 goto done;
  943                         continue;
  944                 }
  945                 if (ifp->if_flags & IFF_POINTOPOINT) {
  946                         if (sa_equal(addr, ifa->ifa_dstaddr))
  947                                 goto done;
  948                 } else {
  949                         cp = addr->sa_data;
  950                         cp2 = ifa->ifa_addr->sa_data;
  951                         cp3 = ifa->ifa_netmask->sa_data;
  952                         cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
  953                         for (; cp3 < cplim; cp3++)
  954                                 if ((*cp++ ^ *cp2++) & *cp3)
  955                                         break;
  956                         if (cp3 == cplim)
  957                                 goto done;
  958                 }
  959         }
  960         ifa = ifa_maybe;
  961 done:
  962         return (ifa);
  963 }
  964 
  965 #include <net/route.h>
  966 
  967 /*
  968  * Default action when installing a route with a Link Level gateway.
  969  * Lookup an appropriate real ifa to point to.
  970  * This should be moved to /sys/net/link.c eventually.
  971  */
  972 static void
  973 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
  974 {
  975         struct ifaddr *ifa, *oifa;
  976         struct sockaddr *dst;
  977         struct ifnet *ifp;
  978 
  979         RT_LOCK_ASSERT(rt);
  980 
  981         if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
  982             ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
  983                 return;
  984         ifa = ifaof_ifpforaddr(dst, ifp);
  985         if (ifa) {
  986                 IFAREF(ifa);            /* XXX */
  987                 oifa = rt->rt_ifa;
  988                 rt->rt_ifa = ifa;
  989                 IFAFREE(oifa);
  990                 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
  991                         ifa->ifa_rtrequest(cmd, rt, info);
  992         }
  993 }
  994 
  995 /*
  996  * Mark an interface down and notify protocols of
  997  * the transition.
  998  * NOTE: must be called at splnet or eqivalent.
  999  */
 1000 static void
 1001 if_unroute(struct ifnet *ifp, int flag, int fam)
 1002 {
 1003         struct ifaddr *ifa;
 1004 
 1005         KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
 1006 
 1007         ifp->if_flags &= ~flag;
 1008         getmicrotime(&ifp->if_lastchange);
 1009         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1010                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1011                         pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
 1012         if_qflush(&ifp->if_snd);
 1013 #ifdef DEV_CARP
 1014         if (ifp->if_carp)
 1015                 carp_carpdev_state(ifp->if_carp);
 1016 #endif
 1017         rt_ifmsg(ifp);
 1018 }
 1019 
 1020 /*
 1021  * Mark an interface up and notify protocols of
 1022  * the transition.
 1023  * NOTE: must be called at splnet or eqivalent.
 1024  */
 1025 static void
 1026 if_route(struct ifnet *ifp, int flag, int fam)
 1027 {
 1028         struct ifaddr *ifa;
 1029 
 1030         KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
 1031 
 1032         ifp->if_flags |= flag;
 1033         getmicrotime(&ifp->if_lastchange);
 1034         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
 1035                 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
 1036                         pfctlinput(PRC_IFUP, ifa->ifa_addr);
 1037 #ifdef DEV_CARP
 1038         if (ifp->if_carp)
 1039                 carp_carpdev_state(ifp->if_carp);
 1040 #endif
 1041         rt_ifmsg(ifp);
 1042 #ifdef INET6
 1043         in6_if_up(ifp);
 1044 #endif
 1045 }
 1046 
 1047 void    (*vlan_link_state_p)(struct ifnet *, int);      /* XXX: private from if_vlan */
 1048 
 1049 /*
 1050  * Handle a change in the interface link state. To avoid LORs
 1051  * between driver lock and upper layer locks, as well as possible
 1052  * recursions, we post event to taskqueue, and all job
 1053  * is done in static do_link_state_change().
 1054  */
 1055 void
 1056 if_link_state_change(struct ifnet *ifp, int link_state)
 1057 {
 1058         /* Return if state hasn't changed. */
 1059         if (ifp->if_link_state == link_state)
 1060                 return;
 1061 
 1062         ifp->if_link_state = link_state;
 1063 
 1064         taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
 1065 }
 1066 
 1067 static void
 1068 do_link_state_change(void *arg, int pending)
 1069 {
 1070         struct ifnet *ifp = (struct ifnet *)arg;
 1071         int link_state = ifp->if_link_state;
 1072         int link;
 1073 
 1074         /* Notify that the link state has changed. */
 1075         rt_ifmsg(ifp);
 1076         if (link_state == LINK_STATE_UP)
 1077                 link = NOTE_LINKUP;
 1078         else if (link_state == LINK_STATE_DOWN)
 1079                 link = NOTE_LINKDOWN;
 1080         else
 1081                 link = NOTE_LINKINV;
 1082         KNOTE_UNLOCKED(&ifp->if_klist, link);
 1083         if (ifp->if_nvlans != 0)
 1084                 (*vlan_link_state_p)(ifp, link);
 1085 
 1086         if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
 1087             IFP2AC(ifp)->ac_netgraph != NULL)
 1088                 (*ng_ether_link_state_p)(ifp, link_state);
 1089 #ifdef DEV_CARP
 1090         if (ifp->if_carp)
 1091                 carp_carpdev_state(ifp->if_carp);
 1092 #endif
 1093         if (ifp->if_bridge) {
 1094                 KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
 1095                 (*bstp_linkstate_p)(ifp, link_state);
 1096         }
 1097 
 1098         devctl_notify("IFNET", ifp->if_xname,
 1099             (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
 1100         if (pending > 1)
 1101                 if_printf(ifp, "%d link states coalesced\n", pending);
 1102         if (log_link_state_change)
 1103                 log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
 1104                     (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
 1105 }
 1106 
 1107 /*
 1108  * Mark an interface down and notify protocols of
 1109  * the transition.
 1110  * NOTE: must be called at splnet or eqivalent.
 1111  */
 1112 void
 1113 if_down(struct ifnet *ifp)
 1114 {
 1115 
 1116         if_unroute(ifp, IFF_UP, AF_UNSPEC);
 1117 }
 1118 
 1119 /*
 1120  * Mark an interface up and notify protocols of
 1121  * the transition.
 1122  * NOTE: must be called at splnet or eqivalent.
 1123  */
 1124 void
 1125 if_up(struct ifnet *ifp)
 1126 {
 1127 
 1128         if_route(ifp, IFF_UP, AF_UNSPEC);
 1129 }
 1130 
 1131 /*
 1132  * Flush an interface queue.
 1133  */
 1134 static void
 1135 if_qflush(struct ifaltq *ifq)
 1136 {
 1137         struct mbuf *m, *n;
 1138 
 1139         IFQ_LOCK(ifq);
 1140 #ifdef ALTQ
 1141         if (ALTQ_IS_ENABLED(ifq))
 1142                 ALTQ_PURGE(ifq);
 1143 #endif
 1144         n = ifq->ifq_head;
 1145         while ((m = n) != 0) {
 1146                 n = m->m_act;
 1147                 m_freem(m);
 1148         }
 1149         ifq->ifq_head = 0;
 1150         ifq->ifq_tail = 0;
 1151         ifq->ifq_len = 0;
 1152         IFQ_UNLOCK(ifq);
 1153 }
 1154 
 1155 /*
 1156  * Handle interface watchdog timer routines.  Called
 1157  * from softclock, we decrement timers (if set) and
 1158  * call the appropriate interface routine on expiration.
 1159  *
 1160  * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
 1161  * holding Giant.  If we switch to an MPSAFE callout, we likely need to grab
 1162  * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
 1163  */
 1164 static void
 1165 if_slowtimo(void *arg)
 1166 {
 1167         struct ifnet *ifp;
 1168         int s = splimp();
 1169 
 1170         IFNET_RLOCK();
 1171         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1172                 if (ifp->if_timer == 0 || --ifp->if_timer)
 1173                         continue;
 1174                 if (ifp->if_watchdog)
 1175                         (*ifp->if_watchdog)(ifp);
 1176         }
 1177         IFNET_RUNLOCK();
 1178         splx(s);
 1179         timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
 1180 }
 1181 
 1182 /*
 1183  * Map interface name to
 1184  * interface structure pointer.
 1185  */
 1186 struct ifnet *
 1187 ifunit(const char *name)
 1188 {
 1189         struct ifnet *ifp;
 1190 
 1191         IFNET_RLOCK();
 1192         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1193                 if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
 1194                         break;
 1195         }
 1196         IFNET_RUNLOCK();
 1197         return (ifp);
 1198 }
 1199 
 1200 /*
 1201  * Hardware specific interface ioctls.
 1202  */
 1203 static int
 1204 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
 1205 {
 1206         struct ifreq *ifr;
 1207         struct ifstat *ifs;
 1208         int error = 0;
 1209         int new_flags, temp_flags;
 1210         size_t namelen, onamelen;
 1211         char new_name[IFNAMSIZ];
 1212         struct ifaddr *ifa;
 1213         struct sockaddr_dl *sdl;
 1214 
 1215         ifr = (struct ifreq *)data;
 1216         switch (cmd) {
 1217         case SIOCGIFINDEX:
 1218                 ifr->ifr_index = ifp->if_index;
 1219                 break;
 1220 
 1221         case SIOCGIFFLAGS:
 1222                 temp_flags = ifp->if_flags | ifp->if_drv_flags;
 1223                 ifr->ifr_flags = temp_flags & 0xffff;
 1224                 ifr->ifr_flagshigh = temp_flags >> 16;
 1225                 break;
 1226 
 1227         case SIOCGIFCAP:
 1228                 ifr->ifr_reqcap = ifp->if_capabilities;
 1229                 ifr->ifr_curcap = ifp->if_capenable;
 1230                 break;
 1231 
 1232 #ifdef MAC
 1233         case SIOCGIFMAC:
 1234                 error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp);
 1235                 break;
 1236 #endif
 1237 
 1238         case SIOCGIFMETRIC:
 1239                 ifr->ifr_metric = ifp->if_metric;
 1240                 break;
 1241 
 1242         case SIOCGIFMTU:
 1243                 ifr->ifr_mtu = ifp->if_mtu;
 1244                 break;
 1245 
 1246         case SIOCGIFPHYS:
 1247                 ifr->ifr_phys = ifp->if_physical;
 1248                 break;
 1249 
 1250         case SIOCSIFFLAGS:
 1251                 error = suser(td);
 1252                 if (error)
 1253                         return (error);
 1254                 /*
 1255                  * Currently, no driver owned flags pass the IFF_CANTCHANGE
 1256                  * check, so we don't need special handling here yet.
 1257                  */
 1258                 new_flags = (ifr->ifr_flags & 0xffff) |
 1259                     (ifr->ifr_flagshigh << 16);
 1260                 if (ifp->if_flags & IFF_SMART) {
 1261                         /* Smart drivers twiddle their own routes */
 1262                 } else if (ifp->if_flags & IFF_UP &&
 1263                     (new_flags & IFF_UP) == 0) {
 1264                         int s = splimp();
 1265                         if_down(ifp);
 1266                         splx(s);
 1267                 } else if (new_flags & IFF_UP &&
 1268                     (ifp->if_flags & IFF_UP) == 0) {
 1269                         int s = splimp();
 1270                         if_up(ifp);
 1271                         splx(s);
 1272                 }
 1273                 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
 1274                         (new_flags &~ IFF_CANTCHANGE);
 1275                 if (new_flags & IFF_PPROMISC) {
 1276                         /* Permanently promiscuous mode requested */
 1277                         ifp->if_flags |= IFF_PROMISC;
 1278                 } else if (ifp->if_pcount == 0) {
 1279                         ifp->if_flags &= ~IFF_PROMISC;
 1280                 }
 1281                 if (ifp->if_ioctl != NULL) {
 1282                         IFF_LOCKGIANT(ifp);
 1283                         (void) (*ifp->if_ioctl)(ifp, cmd, data);
 1284                         IFF_UNLOCKGIANT(ifp);
 1285                 }
 1286                 getmicrotime(&ifp->if_lastchange);
 1287                 break;
 1288 
 1289         case SIOCSIFCAP:
 1290                 error = suser(td);
 1291                 if (error)
 1292                         return (error);
 1293                 if (ifp->if_ioctl == NULL)
 1294                         return (EOPNOTSUPP);
 1295                 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
 1296                         return (EINVAL);
 1297                 IFF_LOCKGIANT(ifp);
 1298                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1299                 IFF_UNLOCKGIANT(ifp);
 1300                 if (error == 0)
 1301                         getmicrotime(&ifp->if_lastchange);
 1302                 break;
 1303 
 1304 #ifdef MAC
 1305         case SIOCSIFMAC:
 1306                 error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp);
 1307                 break;
 1308 #endif
 1309 
 1310         case SIOCSIFNAME:
 1311                 error = suser(td);
 1312                 if (error != 0)
 1313                         return (error);
 1314                 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
 1315                 if (error != 0)
 1316                         return (error);
 1317                 if (new_name[0] == '\0')
 1318                         return (EINVAL);
 1319                 if (ifunit(new_name) != NULL)
 1320                         return (EEXIST);
 1321                 
 1322                 /* Announce the departure of the interface. */
 1323                 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
 1324                 EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
 1325 
 1326                 log(LOG_INFO, "%s: changing name to '%s'\n",
 1327                     ifp->if_xname, new_name);
 1328 
 1329                 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
 1330                 ifa = ifaddr_byindex(ifp->if_index);
 1331                 IFA_LOCK(ifa);
 1332                 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 1333                 namelen = strlen(new_name);
 1334                 onamelen = sdl->sdl_nlen;
 1335                 /*
 1336                  * Move the address if needed.  This is safe because we
 1337                  * allocate space for a name of length IFNAMSIZ when we
 1338                  * create this in if_attach().
 1339                  */
 1340                 if (namelen != onamelen) {
 1341                         bcopy(sdl->sdl_data + onamelen,
 1342                             sdl->sdl_data + namelen, sdl->sdl_alen);
 1343                 }
 1344                 bcopy(new_name, sdl->sdl_data, namelen);
 1345                 sdl->sdl_nlen = namelen;
 1346                 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
 1347                 bzero(sdl->sdl_data, onamelen);
 1348                 while (namelen != 0)
 1349                         sdl->sdl_data[--namelen] = 0xff;
 1350                 IFA_UNLOCK(ifa);
 1351 
 1352                 EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
 1353                 /* Announce the return of the interface. */
 1354                 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
 1355                 break;
 1356 
 1357         case SIOCSIFMETRIC:
 1358                 error = suser(td);
 1359                 if (error)
 1360                         return (error);
 1361                 ifp->if_metric = ifr->ifr_metric;
 1362                 getmicrotime(&ifp->if_lastchange);
 1363                 break;
 1364 
 1365         case SIOCSIFPHYS:
 1366                 error = suser(td);
 1367                 if (error)
 1368                         return (error);
 1369                 if (ifp->if_ioctl == NULL)
 1370                         return (EOPNOTSUPP);
 1371                 IFF_LOCKGIANT(ifp);
 1372                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1373                 IFF_UNLOCKGIANT(ifp);
 1374                 if (error == 0)
 1375                         getmicrotime(&ifp->if_lastchange);
 1376                 break;
 1377 
 1378         case SIOCSIFMTU:
 1379         {
 1380                 u_long oldmtu = ifp->if_mtu;
 1381 
 1382                 error = suser(td);
 1383                 if (error)
 1384                         return (error);
 1385                 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
 1386                         return (EINVAL);
 1387                 if (ifp->if_ioctl == NULL)
 1388                         return (EOPNOTSUPP);
 1389                 IFF_LOCKGIANT(ifp);
 1390                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1391                 IFF_UNLOCKGIANT(ifp);
 1392                 if (error == 0) {
 1393                         getmicrotime(&ifp->if_lastchange);
 1394                         rt_ifmsg(ifp);
 1395                 }
 1396                 /*
 1397                  * If the link MTU changed, do network layer specific procedure.
 1398                  */
 1399                 if (ifp->if_mtu != oldmtu) {
 1400 #ifdef INET6
 1401                         nd6_setmtu(ifp);
 1402 #endif
 1403                 }
 1404                 break;
 1405         }
 1406 
 1407         case SIOCADDMULTI:
 1408         case SIOCDELMULTI:
 1409                 error = suser(td);
 1410                 if (error)
 1411                         return (error);
 1412 
 1413                 /* Don't allow group membership on non-multicast interfaces. */
 1414                 if ((ifp->if_flags & IFF_MULTICAST) == 0)
 1415                         return (EOPNOTSUPP);
 1416 
 1417                 /* Don't let users screw up protocols' entries. */
 1418                 if (ifr->ifr_addr.sa_family != AF_LINK)
 1419                         return (EINVAL);
 1420 
 1421                 if (cmd == SIOCADDMULTI) {
 1422                         struct ifmultiaddr *ifma;
 1423                         error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
 1424                 } else {
 1425                         error = if_delmulti(ifp, &ifr->ifr_addr);
 1426                 }
 1427                 if (error == 0)
 1428                         getmicrotime(&ifp->if_lastchange);
 1429                 break;
 1430 
 1431         case SIOCSIFPHYADDR:
 1432         case SIOCDIFPHYADDR:
 1433 #ifdef INET6
 1434         case SIOCSIFPHYADDR_IN6:
 1435 #endif
 1436         case SIOCSLIFPHYADDR:
 1437         case SIOCSIFMEDIA:
 1438         case SIOCSIFGENERIC:
 1439                 error = suser(td);
 1440                 if (error)
 1441                         return (error);
 1442                 if (ifp->if_ioctl == NULL)
 1443                         return (EOPNOTSUPP);
 1444                 IFF_LOCKGIANT(ifp);
 1445                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1446                 IFF_UNLOCKGIANT(ifp);
 1447                 if (error == 0)
 1448                         getmicrotime(&ifp->if_lastchange);
 1449                 break;
 1450 
 1451         case SIOCGIFSTATUS:
 1452                 ifs = (struct ifstat *)data;
 1453                 ifs->ascii[0] = '\0';
 1454 
 1455         case SIOCGIFPSRCADDR:
 1456         case SIOCGIFPDSTADDR:
 1457         case SIOCGLIFPHYADDR:
 1458         case SIOCGIFMEDIA:
 1459         case SIOCGIFGENERIC:
 1460                 if (ifp->if_ioctl == NULL)
 1461                         return (EOPNOTSUPP);
 1462                 IFF_LOCKGIANT(ifp);
 1463                 error = (*ifp->if_ioctl)(ifp, cmd, data);
 1464                 IFF_UNLOCKGIANT(ifp);
 1465                 break;
 1466 
 1467         case SIOCSIFLLADDR:
 1468                 error = suser(td);
 1469                 if (error)
 1470                         return (error);
 1471                 error = if_setlladdr(ifp,
 1472                     ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
 1473                 break;
 1474 
 1475         default:
 1476                 error = ENOIOCTL;
 1477                 break;
 1478         }
 1479         return (error);
 1480 }
 1481 
 1482 /*
 1483  * Interface ioctls.
 1484  */
 1485 int
 1486 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
 1487 {
 1488         struct ifnet *ifp;
 1489         struct ifreq *ifr;
 1490         int error;
 1491         int oif_flags;
 1492 
 1493         switch (cmd) {
 1494         case SIOCGIFCONF:
 1495         case OSIOCGIFCONF:
 1496 #ifdef __amd64__
 1497         case SIOCGIFCONF32:
 1498 #endif
 1499                 return (ifconf(cmd, data));
 1500         }
 1501         ifr = (struct ifreq *)data;
 1502 
 1503         switch (cmd) {
 1504         case SIOCIFCREATE:
 1505         case SIOCIFDESTROY:
 1506                 if ((error = suser(td)) != 0)
 1507                         return (error);
 1508                 return ((cmd == SIOCIFCREATE) ?
 1509                         if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
 1510                         if_clone_destroy(ifr->ifr_name));
 1511 
 1512         case SIOCIFGCLONERS:
 1513                 return (if_clone_list((struct if_clonereq *)data));
 1514         }
 1515 
 1516         ifp = ifunit(ifr->ifr_name);
 1517         if (ifp == 0)
 1518                 return (ENXIO);
 1519 
 1520         error = ifhwioctl(cmd, ifp, data, td);
 1521         if (error != ENOIOCTL)
 1522                 return (error);
 1523 
 1524         oif_flags = ifp->if_flags;
 1525         if (so->so_proto == 0)
 1526                 return (EOPNOTSUPP);
 1527 #ifndef COMPAT_43
 1528         error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
 1529                                                                  data,
 1530                                                                  ifp, td));
 1531 #else
 1532         {
 1533                 int ocmd = cmd;
 1534 
 1535                 switch (cmd) {
 1536 
 1537                 case SIOCSIFDSTADDR:
 1538                 case SIOCSIFADDR:
 1539                 case SIOCSIFBRDADDR:
 1540                 case SIOCSIFNETMASK:
 1541 #if BYTE_ORDER != BIG_ENDIAN
 1542                         if (ifr->ifr_addr.sa_family == 0 &&
 1543                             ifr->ifr_addr.sa_len < 16) {
 1544                                 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
 1545                                 ifr->ifr_addr.sa_len = 16;
 1546                         }
 1547 #else
 1548                         if (ifr->ifr_addr.sa_len == 0)
 1549                                 ifr->ifr_addr.sa_len = 16;
 1550 #endif
 1551                         break;
 1552 
 1553                 case OSIOCGIFADDR:
 1554                         cmd = SIOCGIFADDR;
 1555                         break;
 1556 
 1557                 case OSIOCGIFDSTADDR:
 1558                         cmd = SIOCGIFDSTADDR;
 1559                         break;
 1560 
 1561                 case OSIOCGIFBRDADDR:
 1562                         cmd = SIOCGIFBRDADDR;
 1563                         break;
 1564 
 1565                 case OSIOCGIFNETMASK:
 1566                         cmd = SIOCGIFNETMASK;
 1567                 }
 1568                 error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
 1569                                                                    cmd,
 1570                                                                    data,
 1571                                                                    ifp, td));
 1572                 switch (ocmd) {
 1573 
 1574                 case OSIOCGIFADDR:
 1575                 case OSIOCGIFDSTADDR:
 1576                 case OSIOCGIFBRDADDR:
 1577                 case OSIOCGIFNETMASK:
 1578                         *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
 1579 
 1580                 }
 1581         }
 1582 #endif /* COMPAT_43 */
 1583 
 1584         if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
 1585 #ifdef INET6
 1586                 DELAY(100);/* XXX: temporary workaround for fxp issue*/
 1587                 if (ifp->if_flags & IFF_UP) {
 1588                         int s = splimp();
 1589                         in6_if_up(ifp);
 1590                         splx(s);
 1591                 }
 1592 #endif
 1593         }
 1594         return (error);
 1595 }
 1596 
 1597 /*
 1598  * The code common to handling reference counted flags,
 1599  * e.g., in ifpromisc() and if_allmulti().
 1600  * The "pflag" argument can specify a permanent mode flag,
 1601  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
 1602  *
 1603  * Only to be used on stack-owned flags, not driver-owned flags.
 1604  */
 1605 static int
 1606 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
 1607 {
 1608         struct ifreq ifr;
 1609         int error;
 1610         int oldflags, oldcount;
 1611 
 1612         KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
 1613             ("if_setflag: setting driver-ownded flag %d", flag));
 1614 
 1615         /* Sanity checks to catch programming errors */
 1616         if (onswitch) {
 1617                 if (*refcount < 0) {
 1618                         if_printf(ifp,
 1619                             "refusing to increment negative refcount %d "
 1620                             "for interface flag %d\n", *refcount, flag);
 1621                         return (EINVAL);
 1622                 }
 1623         } else {
 1624                 if (*refcount <= 0) {
 1625                         if_printf(ifp,
 1626                             "refusing to decrement non-positive refcount %d"
 1627                             "for interface flag %d\n", *refcount, flag);
 1628                         return (EINVAL);
 1629                 }
 1630         }
 1631 
 1632         /* In case this mode is permanent, just touch refcount */
 1633         if (ifp->if_flags & pflag) {
 1634                 *refcount += onswitch ? 1 : -1;
 1635                 return (0);
 1636         }
 1637 
 1638         /* Save ifnet parameters for if_ioctl() may fail */
 1639         oldcount = *refcount;
 1640         oldflags = ifp->if_flags;
 1641         
 1642         /*
 1643          * See if we aren't the only and touching refcount is enough.
 1644          * Actually toggle interface flag if we are the first or last.
 1645          */
 1646         if (onswitch) {
 1647                 if ((*refcount)++)
 1648                         return (0);
 1649                 ifp->if_flags |= flag;
 1650         } else {
 1651                 if (--(*refcount))
 1652                         return (0);
 1653                 ifp->if_flags &= ~flag;
 1654         }
 1655 
 1656         /* Call down the driver since we've changed interface flags */
 1657         if (ifp->if_ioctl == NULL) {
 1658                 error = EOPNOTSUPP;
 1659                 goto recover;
 1660         }
 1661         ifr.ifr_flags = ifp->if_flags & 0xffff;
 1662         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 1663         IFF_LOCKGIANT(ifp);
 1664         error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 1665         IFF_UNLOCKGIANT(ifp);
 1666         if (error)
 1667                 goto recover;
 1668         /* Notify userland that interface flags have changed */
 1669         rt_ifmsg(ifp);
 1670         return (0);
 1671 
 1672 recover:
 1673         /* Recover after driver error */
 1674         *refcount = oldcount;
 1675         ifp->if_flags = oldflags;
 1676         return (error);
 1677 }
 1678 
 1679 /*
 1680  * Set/clear promiscuous mode on interface ifp based on the truth value
 1681  * of pswitch.  The calls are reference counted so that only the first
 1682  * "on" request actually has an effect, as does the final "off" request.
 1683  * Results are undefined if the "off" and "on" requests are not matched.
 1684  */
 1685 int
 1686 ifpromisc(struct ifnet *ifp, int pswitch)
 1687 {
 1688         int error;
 1689         int oldflags = ifp->if_flags;
 1690 
 1691         error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
 1692                            &ifp->if_pcount, pswitch);
 1693         /* If promiscuous mode status has changed, log a message */
 1694         if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
 1695                 log(LOG_INFO, "%s: promiscuous mode %s\n",
 1696                     ifp->if_xname,
 1697                     (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
 1698         return (error);
 1699 }
 1700 
 1701 /*
 1702  * Return interface configuration
 1703  * of system.  List may be used
 1704  * in later ioctl's (above) to get
 1705  * other information.
 1706  */
 1707 /*ARGSUSED*/
 1708 static int
 1709 ifconf(u_long cmd, caddr_t data)
 1710 {
 1711         struct ifconf *ifc = (struct ifconf *)data;
 1712 #ifdef __amd64__
 1713         struct ifconf32 *ifc32 = (struct ifconf32 *)data;
 1714         struct ifconf ifc_swab;
 1715 #endif
 1716         struct ifnet *ifp;
 1717         struct ifaddr *ifa;
 1718         struct ifreq ifr;
 1719         struct sbuf *sb;
 1720         int error, full = 0, valid_len, max_len;
 1721 
 1722 #ifdef __amd64__
 1723         if (cmd == SIOCGIFCONF32) {
 1724                 ifc_swab.ifc_len = ifc32->ifc_len;
 1725                 ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
 1726                 ifc = &ifc_swab;
 1727         }
 1728 #endif
 1729         /* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
 1730         max_len = MAXPHYS - 1;
 1731 
 1732         /* Prevent hostile input from being able to crash the system */
 1733         if (ifc->ifc_len <= 0)
 1734                 return (EINVAL);
 1735 
 1736 again:
 1737         if (ifc->ifc_len <= max_len) {
 1738                 max_len = ifc->ifc_len;
 1739                 full = 1;
 1740         }
 1741         sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
 1742         max_len = 0;
 1743         valid_len = 0;
 1744 
 1745         IFNET_RLOCK();          /* could sleep XXX */
 1746         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1747                 int addrs;
 1748 
 1749                 /*
 1750                  * Zero the ifr_name buffer to make sure we don't
 1751                  * disclose the contents of the stack.
 1752                  */
 1753                 memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
 1754 
 1755                 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
 1756                     >= sizeof(ifr.ifr_name)) {
 1757                         sbuf_delete(sb);
 1758                         IFNET_RUNLOCK();
 1759                         return (ENAMETOOLONG);
 1760                 }
 1761 
 1762                 addrs = 0;
 1763                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1764                         struct sockaddr *sa = ifa->ifa_addr;
 1765 
 1766                         if (jailed(curthread->td_ucred) &&
 1767                             prison_if(curthread->td_ucred, sa))
 1768                                 continue;
 1769                         addrs++;
 1770 #ifdef COMPAT_43
 1771                         if (cmd == OSIOCGIFCONF) {
 1772                                 struct osockaddr *osa =
 1773                                          (struct osockaddr *)&ifr.ifr_addr;
 1774                                 ifr.ifr_addr = *sa;
 1775                                 osa->sa_family = sa->sa_family;
 1776                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 1777                                 max_len += sizeof(ifr);
 1778                         } else
 1779 #endif
 1780                         if (sa->sa_len <= sizeof(*sa)) {
 1781                                 ifr.ifr_addr = *sa;
 1782                                 sbuf_bcat(sb, &ifr, sizeof(ifr));
 1783                                 max_len += sizeof(ifr);
 1784                         } else {
 1785                                 sbuf_bcat(sb, &ifr,
 1786                                     offsetof(struct ifreq, ifr_addr));
 1787                                 max_len += offsetof(struct ifreq, ifr_addr);
 1788                                 sbuf_bcat(sb, sa, sa->sa_len);
 1789                                 max_len += sa->sa_len;
 1790                         }
 1791 
 1792                         if (!sbuf_overflowed(sb))
 1793                                 valid_len = sbuf_len(sb);
 1794                 }
 1795                 if (addrs == 0) {
 1796                         bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
 1797                         sbuf_bcat(sb, &ifr, sizeof(ifr));
 1798                         max_len += sizeof(ifr);
 1799 
 1800                         if (!sbuf_overflowed(sb))
 1801                                 valid_len = sbuf_len(sb);
 1802                 }
 1803         }
 1804         IFNET_RUNLOCK();
 1805 
 1806         /*
 1807          * If we didn't allocate enough space (uncommon), try again.  If
 1808          * we have already allocated as much space as we are allowed,
 1809          * return what we've got.
 1810          */
 1811         if (valid_len != max_len && !full) {
 1812                 sbuf_delete(sb);
 1813                 goto again;
 1814         }
 1815 
 1816         ifc->ifc_len = valid_len;
 1817 #ifdef __amd64__
 1818         if (cmd == SIOCGIFCONF32)
 1819                 ifc32->ifc_len = valid_len;
 1820 #endif
 1821         sbuf_finish(sb);
 1822         error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
 1823         sbuf_delete(sb);
 1824         return (error);
 1825 }
 1826 
 1827 /*
 1828  * Just like ifpromisc(), but for all-multicast-reception mode.
 1829  */
 1830 int
 1831 if_allmulti(struct ifnet *ifp, int onswitch)
 1832 {
 1833 
 1834         return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
 1835 }
 1836 
 1837 static struct ifmultiaddr *
 1838 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
 1839 {
 1840         struct ifmultiaddr *ifma;
 1841 
 1842         IF_ADDR_LOCK_ASSERT(ifp);
 1843 
 1844         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
 1845                 if (sa_equal(ifma->ifma_addr, sa))
 1846                         break;
 1847         }
 1848 
 1849         return ifma;
 1850 }
 1851 
 1852 /*
 1853  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
 1854  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
 1855  * the ifnet multicast address list here, so the caller must do that and
 1856  * other setup work (such as notifying the device driver).  The reference
 1857  * count is initialized to 1.
 1858  */
 1859 static struct ifmultiaddr *
 1860 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
 1861     int mflags)
 1862 {
 1863         struct ifmultiaddr *ifma;
 1864         struct sockaddr *dupsa;
 1865 
 1866         MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
 1867             M_ZERO);
 1868         if (ifma == NULL)
 1869                 return (NULL);
 1870 
 1871         MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
 1872         if (dupsa == NULL) {
 1873                 FREE(ifma, M_IFMADDR);
 1874                 return (NULL);
 1875         }
 1876         bcopy(sa, dupsa, sa->sa_len);
 1877         ifma->ifma_addr = dupsa;
 1878 
 1879         ifma->ifma_ifp = ifp;
 1880         ifma->ifma_refcount = 1;
 1881         ifma->ifma_protospec = NULL;
 1882 
 1883         if (llsa == NULL) {
 1884                 ifma->ifma_lladdr = NULL;
 1885                 return (ifma);
 1886         }
 1887 
 1888         MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
 1889         if (dupsa == NULL) {
 1890                 FREE(ifma->ifma_addr, M_IFMADDR);
 1891                 FREE(ifma, M_IFMADDR);
 1892                 return (NULL);
 1893         }
 1894         bcopy(llsa, dupsa, llsa->sa_len);
 1895         ifma->ifma_lladdr = dupsa;
 1896 
 1897         return (ifma);
 1898 }
 1899 
 1900 /*
 1901  * if_freemulti: free ifmultiaddr structure and possibly attached related
 1902  * addresses.  The caller is responsible for implementing reference
 1903  * counting, notifying the driver, handling routing messages, and releasing
 1904  * any dependent link layer state.
 1905  */
 1906 static void
 1907 if_freemulti(struct ifmultiaddr *ifma)
 1908 {
 1909 
 1910         KASSERT(ifma->ifma_refcount == 1, ("if_freemulti: refcount %d",
 1911             ifma->ifma_refcount));
 1912         KASSERT(ifma->ifma_protospec == NULL,
 1913             ("if_freemulti: protospec not NULL"));
 1914 
 1915         if (ifma->ifma_lladdr != NULL)
 1916                 FREE(ifma->ifma_lladdr, M_IFMADDR);
 1917         FREE(ifma->ifma_addr, M_IFMADDR);
 1918         FREE(ifma, M_IFMADDR);
 1919 }
 1920 
 1921 /*
 1922  * Register an additional multicast address with a network interface.
 1923  *
 1924  * - If the address is already present, bump the reference count on the
 1925  *   address and return.
 1926  * - If the address is not link-layer, look up a link layer address.
 1927  * - Allocate address structures for one or both addresses, and attach to the
 1928  *   multicast address list on the interface.  If automatically adding a link
 1929  *   layer address, the protocol address will own a reference to the link
 1930  *   layer address, to be freed when it is freed.
 1931  * - Notify the network device driver of an addition to the multicast address
 1932  *   list.
 1933  *
 1934  * 'sa' points to caller-owned memory with the desired multicast address.
 1935  *
 1936  * 'retifma' will be used to return a pointer to the resulting multicast
 1937  * address reference, if desired.
 1938  */
 1939 int
 1940 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
 1941     struct ifmultiaddr **retifma)
 1942 {
 1943         struct ifmultiaddr *ifma, *ll_ifma;
 1944         struct sockaddr *llsa;
 1945         int error;
 1946 
 1947         /*
 1948          * If the address is already present, return a new reference to it;
 1949          * otherwise, allocate storage and set up a new address.
 1950          */
 1951         IF_ADDR_LOCK(ifp);
 1952         ifma = if_findmulti(ifp, sa);
 1953         if (ifma != NULL) {
 1954                 ifma->ifma_refcount++;
 1955                 if (retifma != NULL)
 1956                         *retifma = ifma;
 1957                 IF_ADDR_UNLOCK(ifp);
 1958                 return (0);
 1959         }
 1960 
 1961         /*
 1962          * The address isn't already present; resolve the protocol address
 1963          * into a link layer address, and then look that up, bump its
 1964          * refcount or allocate an ifma for that also.  If 'llsa' was
 1965          * returned, we will need to free it later.
 1966          */
 1967         llsa = NULL;
 1968         ll_ifma = NULL;
 1969         if (ifp->if_resolvemulti != NULL) {
 1970                 error = ifp->if_resolvemulti(ifp, &llsa, sa);
 1971                 if (error)
 1972                         goto unlock_out;
 1973         }
 1974 
 1975         /*
 1976          * Allocate the new address.  Don't hook it up yet, as we may also
 1977          * need to allocate a link layer multicast address.
 1978          */
 1979         ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
 1980         if (ifma == NULL) {
 1981                 error = ENOMEM;
 1982                 goto free_llsa_out;
 1983         }
 1984 
 1985         /*
 1986          * If a link layer address is found, we'll need to see if it's
 1987          * already present in the address list, or allocate is as well.
 1988          * When this block finishes, the link layer address will be on the
 1989          * list.
 1990          */
 1991         if (llsa != NULL) {
 1992                 ll_ifma = if_findmulti(ifp, llsa);
 1993                 if (ll_ifma == NULL) {
 1994                         ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
 1995                         if (ll_ifma == NULL) {
 1996                                 if_freemulti(ifma);
 1997                                 error = ENOMEM;
 1998                                 goto free_llsa_out;
 1999                         }
 2000                         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
 2001                             ifma_link);
 2002                 } else
 2003                         ll_ifma->ifma_refcount++;
 2004         }
 2005 
 2006         /*
 2007          * We now have a new multicast address, ifma, and possibly a new or
 2008          * referenced link layer address.  Add the primary address to the
 2009          * ifnet address list.
 2010          */
 2011         TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
 2012 
 2013         if (retifma != NULL)
 2014                 *retifma = ifma;
 2015 
 2016         /*
 2017          * Must generate the message while holding the lock so that 'ifma'
 2018          * pointer is still valid.
 2019          *
 2020          * XXXRW: How come we don't announce ll_ifma?
 2021          */
 2022         rt_newmaddrmsg(RTM_NEWMADDR, ifma);
 2023         IF_ADDR_UNLOCK(ifp);
 2024 
 2025         /*
 2026          * We are certain we have added something, so call down to the
 2027          * interface to let them know about it.
 2028          */
 2029         if (ifp->if_ioctl != NULL) {
 2030                 IFF_LOCKGIANT(ifp);
 2031                 (void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
 2032                 IFF_UNLOCKGIANT(ifp);
 2033         }
 2034 
 2035         if (llsa != NULL)
 2036                 FREE(llsa, M_IFMADDR);
 2037 
 2038         return (0);
 2039 
 2040 free_llsa_out:
 2041         if (llsa != NULL)
 2042                 FREE(llsa, M_IFMADDR);
 2043 
 2044 unlock_out:
 2045         IF_ADDR_UNLOCK(ifp);
 2046         return (error);
 2047 }
 2048 
 2049 /*
 2050  * Remove a reference to a multicast address on this interface.  Yell
 2051  * if the request does not match an existing membership.
 2052  */
 2053 int
 2054 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
 2055 {
 2056         struct ifmultiaddr *ifma, *ll_ifma;
 2057 
 2058         IF_ADDR_LOCK(ifp);
 2059         ifma = if_findmulti(ifp, sa);
 2060         if (ifma == NULL) {
 2061                 IF_ADDR_UNLOCK(ifp);
 2062                 return ENOENT;
 2063         }
 2064 
 2065         if (ifma->ifma_refcount > 1) {
 2066                 ifma->ifma_refcount--;
 2067                 IF_ADDR_UNLOCK(ifp);
 2068                 return 0;
 2069         }
 2070 
 2071         sa = ifma->ifma_lladdr;
 2072         if (sa != NULL)
 2073                 ll_ifma = if_findmulti(ifp, sa);
 2074         else
 2075                 ll_ifma = NULL;
 2076 
 2077         /*
 2078          * XXXRW: How come we don't announce ll_ifma?
 2079          */
 2080         rt_newmaddrmsg(RTM_DELMADDR, ifma);
 2081 
 2082         TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
 2083         if_freemulti(ifma);
 2084 
 2085         if (ll_ifma != NULL) {
 2086                 if (ll_ifma->ifma_refcount == 1) {
 2087                         TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma, ifma_link);
 2088                         if_freemulti(ll_ifma);
 2089                 } else
 2090                         ll_ifma->ifma_refcount--;
 2091         }
 2092         IF_ADDR_UNLOCK(ifp);
 2093 
 2094         /*
 2095          * Make sure the interface driver is notified
 2096          * in the case of a link layer mcast group being left.
 2097          */
 2098         if (ifp->if_ioctl) {
 2099                 IFF_LOCKGIANT(ifp);
 2100                 (void) (*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
 2101                 IFF_UNLOCKGIANT(ifp);
 2102         }
 2103 
 2104         return 0;
 2105 }
 2106 
 2107 /*
 2108  * Set the link layer address on an interface.
 2109  *
 2110  * At this time we only support certain types of interfaces,
 2111  * and we don't allow the length of the address to change.
 2112  */
 2113 int
 2114 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
 2115 {
 2116         struct sockaddr_dl *sdl;
 2117         struct ifaddr *ifa;
 2118         struct ifreq ifr;
 2119 
 2120         ifa = ifaddr_byindex(ifp->if_index);
 2121         if (ifa == NULL)
 2122                 return (EINVAL);
 2123         sdl = (struct sockaddr_dl *)ifa->ifa_addr;
 2124         if (sdl == NULL)
 2125                 return (EINVAL);
 2126         if (len != sdl->sdl_alen)       /* don't allow length to change */
 2127                 return (EINVAL);
 2128         switch (ifp->if_type) {
 2129         case IFT_ETHER:                 /* these types use struct arpcom */
 2130         case IFT_FDDI:
 2131         case IFT_XETHER:
 2132         case IFT_ISO88025:
 2133         case IFT_L2VLAN:
 2134         case IFT_BRIDGE:
 2135                 bcopy(lladdr, IFP2ENADDR(ifp), len);
 2136                 /*
 2137                  * XXX We also need to store the lladdr in LLADDR(sdl),
 2138                  * which is done below. This is a pain because we must
 2139                  * remember to keep the info in sync.
 2140                  */
 2141                 /* FALLTHROUGH */
 2142         case IFT_ARCNET:
 2143                 bcopy(lladdr, LLADDR(sdl), len);
 2144                 break;
 2145         default:
 2146                 return (ENODEV);
 2147         }
 2148         /*
 2149          * If the interface is already up, we need
 2150          * to re-init it in order to reprogram its
 2151          * address filter.
 2152          */
 2153         if ((ifp->if_flags & IFF_UP) != 0) {
 2154                 if (ifp->if_ioctl) {
 2155                         IFF_LOCKGIANT(ifp);
 2156                         ifp->if_flags &= ~IFF_UP;
 2157                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2158                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2159                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2160                         ifp->if_flags |= IFF_UP;
 2161                         ifr.ifr_flags = ifp->if_flags & 0xffff;
 2162                         ifr.ifr_flagshigh = ifp->if_flags >> 16;
 2163                         (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
 2164                         IFF_UNLOCKGIANT(ifp);
 2165                 }
 2166 #ifdef INET
 2167                 /*
 2168                  * Also send gratuitous ARPs to notify other nodes about
 2169                  * the address change.
 2170                  */
 2171                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 2172                         if (ifa->ifa_addr != NULL &&
 2173                             ifa->ifa_addr->sa_family == AF_INET)
 2174                                 arp_ifinit(ifp, ifa);
 2175                 }
 2176 #endif
 2177         }
 2178         return (0);
 2179 }
 2180 
 2181 /*
 2182  * The name argument must be a pointer to storage which will last as
 2183  * long as the interface does.  For physical devices, the result of
 2184  * device_get_name(dev) is a good choice and for pseudo-devices a
 2185  * static string works well.
 2186  */
 2187 void
 2188 if_initname(struct ifnet *ifp, const char *name, int unit)
 2189 {
 2190         ifp->if_dname = name;
 2191         ifp->if_dunit = unit;
 2192         if (unit != IF_DUNIT_NONE)
 2193                 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
 2194         else
 2195                 strlcpy(ifp->if_xname, name, IFNAMSIZ);
 2196 }
 2197 
 2198 int
 2199 if_printf(struct ifnet *ifp, const char * fmt, ...)
 2200 {
 2201         va_list ap;
 2202         int retval;
 2203 
 2204         retval = printf("%s: ", ifp->if_xname);
 2205         va_start(ap, fmt);
 2206         retval += vprintf(fmt, ap);
 2207         va_end(ap);
 2208         return (retval);
 2209 }
 2210 
 2211 /*
 2212  * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
 2213  * be called without Giant.  However, we often can't acquire the Giant lock
 2214  * at those points; instead, we run it via a task queue that holds Giant via
 2215  * if_start_deferred.
 2216  *
 2217  * XXXRW: We need to make sure that the ifnet isn't fully detached until any
 2218  * outstanding if_start_deferred() tasks that will run after the free.  This
 2219  * probably means waiting in if_detach().
 2220  */
 2221 void
 2222 if_start(struct ifnet *ifp)
 2223 {
 2224 
 2225         NET_ASSERT_GIANT();
 2226 
 2227         if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) {
 2228                 if (mtx_owned(&Giant))
 2229                         (*(ifp)->if_start)(ifp);
 2230                 else
 2231                         taskqueue_enqueue(taskqueue_swi_giant,
 2232                             &ifp->if_starttask);
 2233         } else
 2234                 (*(ifp)->if_start)(ifp);
 2235 }
 2236 
 2237 static void
 2238 if_start_deferred(void *context, int pending)
 2239 {
 2240         struct ifnet *ifp;
 2241 
 2242         /*
 2243          * This code must be entered with Giant, and should never run if
 2244          * we're not running with debug.mpsafenet.
 2245          */
 2246         KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet"));
 2247         GIANT_REQUIRED;
 2248 
 2249         ifp = context;
 2250         (ifp->if_start)(ifp);
 2251 }
 2252 
 2253 int
 2254 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
 2255 {
 2256         int active = 0;
 2257 
 2258         IF_LOCK(ifq);
 2259         if (_IF_QFULL(ifq)) {
 2260                 _IF_DROP(ifq);
 2261                 IF_UNLOCK(ifq);
 2262                 m_freem(m);
 2263                 return (0);
 2264         }
 2265         if (ifp != NULL) {
 2266                 ifp->if_obytes += m->m_pkthdr.len + adjust;
 2267                 if (m->m_flags & (M_BCAST|M_MCAST))
 2268                         ifp->if_omcasts++;
 2269                 active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
 2270         }
 2271         _IF_ENQUEUE(ifq, m);
 2272         IF_UNLOCK(ifq);
 2273         if (ifp != NULL && !active)
 2274                 if_start(ifp);
 2275         return (1);
 2276 }
 2277 
 2278 void
 2279 if_register_com_alloc(u_char type,
 2280     if_com_alloc_t *a, if_com_free_t *f)
 2281 {
 2282         
 2283         KASSERT(if_com_alloc[type] == NULL,
 2284             ("if_register_com_alloc: %d already registered", type));
 2285         KASSERT(if_com_free[type] == NULL,
 2286             ("if_register_com_alloc: %d free already registered", type));
 2287 
 2288         if_com_alloc[type] = a;
 2289         if_com_free[type] = f;
 2290 }
 2291 
 2292 void
 2293 if_deregister_com_alloc(u_char type)
 2294 {
 2295         
 2296         KASSERT(if_com_alloc[type] == NULL,
 2297             ("if_deregister_com_alloc: %d not registered", type));
 2298         KASSERT(if_com_free[type] == NULL,
 2299             ("if_deregister_com_alloc: %d free not registered", type));
 2300         if_com_alloc[type] = NULL;
 2301         if_com_free[type] = NULL;
 2302 }

Cache object: 94f51886a0b32aecab1830c497cbfe86


[ 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.