sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*-
   4  * SPDX-License-Identifier: BSD-4-Clause
   5  *
   6  * Copyright 2001 Wasabi Systems, Inc.
   7  * All rights reserved.
   8  *
   9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed for the NetBSD Project by
  22  *      Wasabi Systems, Inc.
  23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  24  *    or promote products derived from this software without specific prior
  25  *    written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 /*
  41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  42  * All rights reserved.
  43  *
  44  * Redistribution and use in source and binary forms, with or without
  45  * modification, are permitted provided that the following conditions
  46  * are met:
  47  * 1. Redistributions of source code must retain the above copyright
  48  *    notice, this list of conditions and the following disclaimer.
  49  * 2. Redistributions in binary form must reproduce the above copyright
  50  *    notice, this list of conditions and the following disclaimer in the
  51  *    documentation and/or other materials provided with the distribution.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  63  * POSSIBILITY OF SUCH DAMAGE.
  64  *
  65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  66  */
  67
  68 /*
  69  * Network interface bridge support.
  70  *
  71  * TODO:
  72  *
  73  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  74  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  75  *        to bridge other types of interfaces (maybe consider
  76  *        heterogeneous bridges).
  77  */
  78
  79 #include <sys/cdefs.h>
  80 __FBSDID("$FreeBSD$");
  81
  82 #include "opt_inet.h"
  83 #include "opt_inet6.h"
  84
  85 #include <sys/param.h>
  86 #include <sys/eventhandler.h>
  87 #include <sys/mbuf.h>
  88 #include <sys/malloc.h>
  89 #include <sys/protosw.h>
  90 #include <sys/systm.h>
  91 #include <sys/jail.h>
  92 #include <sys/time.h>
  93 #include <sys/socket.h> /* for net/if.h */
  94 #include <sys/sockio.h>
  95 #include <sys/ctype.h>  /* string functions */
  96 #include <sys/kernel.h>
  97 #include <sys/random.h>
  98 #include <sys/syslog.h>
  99 #include <sys/sysctl.h>
 100 #include <vm/uma.h>
 101 #include <sys/module.h>
 102 #include <sys/priv.h>
 103 #include <sys/proc.h>
 104 #include <sys/lock.h>
 105 #include <sys/mutex.h>
 106
 107 #include <net/bpf.h>
 108 #include <net/if.h>
 109 #include <net/if_clone.h>
 110 #include <net/if_dl.h>
 111 #include <net/if_types.h>
 112 #include <net/if_var.h>
 113 #include <net/pfil.h>
 114 #include <net/vnet.h>
 115
 116 #include <netinet/in.h>
 117 #include <netinet/in_systm.h>
 118 #include <netinet/in_var.h>
 119 #include <netinet/ip.h>
 120 #include <netinet/ip_var.h>
 121 #ifdef INET6
 122 #include <netinet/ip6.h>
 123 #include <netinet6/ip6_var.h>
 124 #include <netinet6/in6_ifattach.h>
 125 #endif
 126 #if defined(INET) || defined(INET6)
 127 #include <netinet/ip_carp.h>
 128 #endif
 129 #include <machine/in_cksum.h>
 130 #include <netinet/if_ether.h>
 131 #include <net/bridgestp.h>
 132 #include <net/if_bridgevar.h>
 133 #include <net/if_llc.h>
 134 #include <net/if_vlan_var.h>
 135
 136 #include <net/route.h>
 137
 138 #ifdef INET6
 139 /*
 140  * XXX: declare here to avoid to include many inet6 related files..
 141  * should be more generalized?
 142  */
 143 extern void     nd6_setmtu(struct ifnet *);
 144 #endif
 145
 146 /*
 147  * Size of the route hash table.  Must be a power of two.
 148  */
 149 #ifndef BRIDGE_RTHASH_SIZE
 150 #define BRIDGE_RTHASH_SIZE              1024
 151 #endif
 152
 153 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 154
 155 /*
 156  * Default maximum number of addresses to cache.
 157  */
 158 #ifndef BRIDGE_RTABLE_MAX
 159 #define BRIDGE_RTABLE_MAX               2000
 160 #endif
 161
 162 /*
 163  * Timeout (in seconds) for entries learned dynamically.
 164  */
 165 #ifndef BRIDGE_RTABLE_TIMEOUT
 166 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 167 #endif
 168
 169 /*
 170  * Number of seconds between walks of the route list.
 171  */
 172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 173 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 174 #endif
 175
 176 /*
 177  * List of capabilities to possibly mask on the member interface.
 178  */
 179 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 180                                          IFCAP_TXCSUM_IPV6)
 181
 182 /*
 183  * List of capabilities to strip
 184  */
 185 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 186
 187 /*
 188  * Bridge locking
 189  *
 190  * The bridge relies heavily on the epoch(9) system to protect its data
 191  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 192  * must ensure there is only one writer at a time.
 193  *
 194  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 195  * accesses we must hold:
 196  *
 197  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 198  *  - BRIDGE_LOCK, for any other change
 199  *
 200  * The BRIDGE_LOCK is a sleepable lock, because it is held accross ioctl()
 201  * calls to bridge member interfaces and these ioctl()s can sleep.
 202  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 203  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 204  */
 205 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 206         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 207         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 208 } while (0)
 209 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 210         sx_destroy(&(_sc)->sc_sx);              \
 211         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 212 } while (0)
 213 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 214 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 215 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 216 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 217             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 218 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 219 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 220 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 221 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 222 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 223             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 224
 225 /*
 226  * Bridge interface list entry.
 227  */
 228 struct bridge_iflist {
 229         CK_LIST_ENTRY(bridge_iflist) bif_next;
 230         struct ifnet            *bif_ifp;       /* member if */
 231         struct bstp_port        bif_stp;        /* STP state */
 232         uint32_t                bif_flags;      /* member if flags */
 233         int                     bif_savedcaps;  /* saved capabilities */
 234         uint32_t                bif_addrmax;    /* max # of addresses */
 235         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 236         uint32_t                bif_addrexceeded;/* # of address violations */
 237         struct epoch_context    bif_epoch_ctx;
 238 };
 239
 240 /*
 241  * Bridge route node.
 242  */
 243 struct bridge_rtnode {
 244         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 245         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 246         struct bridge_iflist    *brt_dst;       /* destination if */
 247         unsigned long           brt_expire;     /* expiration time */
 248         uint8_t                 brt_flags;      /* address flags */
 249         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 250         uint16_t                brt_vlan;       /* vlan id */
 251         struct  vnet            *brt_vnet;
 252         struct  epoch_context   brt_epoch_ctx;
 253 };
 254 #define brt_ifp                 brt_dst->bif_ifp
 255
 256 /*
 257  * Software state for each bridge.
 258  */
 259 struct bridge_softc {
 260         struct ifnet            *sc_ifp;        /* make this an interface */
 261         LIST_ENTRY(bridge_softc) sc_list;
 262         struct sx               sc_sx;
 263         struct mtx              sc_rt_mtx;
 264         uint32_t                sc_brtmax;      /* max # of addresses */
 265         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 266         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 267         struct callout          sc_brcallout;   /* bridge callout */
 268         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 269         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 270         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 271         uint32_t                sc_rthash_key;  /* key for hash */
 272         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 273         struct bstp_state       sc_stp;         /* STP state */
 274         uint32_t                sc_brtexceeded; /* # of cache drops */
 275         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 276         struct ether_addr       sc_defaddr;     /* Default MAC address */
 277         struct epoch_context    sc_epoch_ctx;
 278 };
 279
 280 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 281 #define V_bridge_list_sx        VNET(bridge_list_sx)
 282 static eventhandler_tag bridge_detach_cookie;
 283
 284 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 285
 286 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 287 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 288
 289 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 290 static void     bridge_clone_destroy(struct ifnet *);
 291
 292 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 293 static void     bridge_mutecaps(struct bridge_softc *);
 294 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 295                     int);
 296 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 297 static void     bridge_init(void *);
 298 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 299 static void     bridge_stop(struct ifnet *, int);
 300 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 301 static void     bridge_qflush(struct ifnet *);
 302 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 303 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 304                     struct rtentry *);
 305 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 306                     struct mbuf *);
 307 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 308
 309 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 310                     struct mbuf *m);
 311
 312 static void     bridge_timer(void *);
 313
 314 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 315                     struct mbuf *, int);
 316 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 317
 318 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 319                     uint16_t, struct bridge_iflist *, int, uint8_t);
 320 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 321                     uint16_t);
 322 static void     bridge_rttrim(struct bridge_softc *);
 323 static void     bridge_rtage(struct bridge_softc *);
 324 static void     bridge_rtflush(struct bridge_softc *, int);
 325 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 326                     uint16_t);
 327
 328 static void     bridge_rtable_init(struct bridge_softc *);
 329 static void     bridge_rtable_fini(struct bridge_softc *);
 330
 331 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 332 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 333                     const uint8_t *, uint16_t);
 334 static int      bridge_rtnode_insert(struct bridge_softc *,
 335                     struct bridge_rtnode *);
 336 static void     bridge_rtnode_destroy(struct bridge_softc *,
 337                     struct bridge_rtnode *);
 338 static void     bridge_rtable_expire(struct ifnet *, int);
 339 static void     bridge_state_change(struct ifnet *, int);
 340
 341 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 342                     const char *name);
 343 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 344                     struct ifnet *ifp);
 345 static void     bridge_delete_member(struct bridge_softc *,
 346                     struct bridge_iflist *, int);
 347 static void     bridge_delete_span(struct bridge_softc *,
 348                     struct bridge_iflist *);
 349
 350 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 351 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 352 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 353 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 354 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 381 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 382                     int);
 383 static int      bridge_ip_checkbasic(struct mbuf **mp);
 384 #ifdef INET6
 385 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 386 #endif /* INET6 */
 387 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 388                     struct ether_header *, int, struct llc *);
 389 static void     bridge_linkstate(struct ifnet *ifp);
 390 static void     bridge_linkcheck(struct bridge_softc *sc);
 391
 392 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 393 #define VLANTAGOF(_m)   \
 394     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 395
 396 static struct bstp_cb_ops bridge_ops = {
 397         .bcb_state = bridge_state_change,
 398         .bcb_rtage = bridge_rtable_expire
 399 };
 400
 401 SYSCTL_DECL(_net_link);
 402 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 403     "Bridge");
 404
 405 /* only pass IP[46] packets when pfil is enabled */
 406 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 407 #define V_pfil_onlyip   VNET(pfil_onlyip)
 408 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 409     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 410     "Only pass IP packets when pfil is enabled");
 411
 412 /* run pfil hooks on the bridge interface */
 413 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 414 #define V_pfil_bridge   VNET(pfil_bridge)
 415 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 416     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 417     "Packet filter on the bridge interface");
 418
 419 /* layer2 filter with ipfw */
 420 VNET_DEFINE_STATIC(int, pfil_ipfw);
 421 #define V_pfil_ipfw     VNET(pfil_ipfw)
 422
 423 /* layer2 ARP filter with ipfw */
 424 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 425 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 426 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 427     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 428     "Filter ARP packets through IPFW layer2");
 429
 430 /* run pfil hooks on the member interface */
 431 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 432 #define V_pfil_member   VNET(pfil_member)
 433 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 434     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 435     "Packet filter on the member interface");
 436
 437 /* run pfil hooks on the physical interface for locally destined packets */
 438 VNET_DEFINE_STATIC(int, pfil_local_phys);
 439 #define V_pfil_local_phys       VNET(pfil_local_phys)
 440 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 441     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 442     "Packet filter on the physical interface for locally destined packets");
 443
 444 /* log STP state changes */
 445 VNET_DEFINE_STATIC(int, log_stp);
 446 #define V_log_stp       VNET(log_stp)
 447 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 448     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 449     "Log STP state changes");
 450
 451 /* share MAC with first bridge member */
 452 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 453 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 454 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 455     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 456     "Inherit MAC address from the first bridge member");
 457
 458 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 459 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 460 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 461     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 462     "Allow overlap of link-local scope "
 463     "zones of a bridge interface and the member interfaces");
 464
 465 struct bridge_control {
 466         int     (*bc_func)(struct bridge_softc *, void *);
 467         int     bc_argsize;
 468         int     bc_flags;
 469 };
 470
 471 #define BC_F_COPYIN             0x01    /* copy arguments in */
 472 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 473 #define BC_F_SUSER              0x04    /* do super-user check */
 474
 475 const struct bridge_control bridge_control_table[] = {
 476         { bridge_ioctl_add,             sizeof(struct ifbreq),
 477           BC_F_COPYIN|BC_F_SUSER },
 478         { bridge_ioctl_del,             sizeof(struct ifbreq),
 479           BC_F_COPYIN|BC_F_SUSER },
 480
 481         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 482           BC_F_COPYIN|BC_F_COPYOUT },
 483         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 484           BC_F_COPYIN|BC_F_SUSER },
 485
 486         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 487           BC_F_COPYIN|BC_F_SUSER },
 488         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 489           BC_F_COPYOUT },
 490
 491         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 492           BC_F_COPYIN|BC_F_COPYOUT },
 493         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 494           BC_F_COPYIN|BC_F_COPYOUT },
 495
 496         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 497           BC_F_COPYIN|BC_F_SUSER },
 498
 499         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 500           BC_F_COPYIN|BC_F_SUSER },
 501         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 502           BC_F_COPYOUT },
 503
 504         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 505           BC_F_COPYIN|BC_F_SUSER },
 506
 507         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 508           BC_F_COPYIN|BC_F_SUSER },
 509
 510         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 511           BC_F_COPYOUT },
 512         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 513           BC_F_COPYIN|BC_F_SUSER },
 514
 515         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 516           BC_F_COPYOUT },
 517         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 518           BC_F_COPYIN|BC_F_SUSER },
 519
 520         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 521           BC_F_COPYOUT },
 522         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 523           BC_F_COPYIN|BC_F_SUSER },
 524
 525         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 526           BC_F_COPYOUT },
 527         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 528           BC_F_COPYIN|BC_F_SUSER },
 529
 530         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 531           BC_F_COPYIN|BC_F_SUSER },
 532
 533         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 534           BC_F_COPYIN|BC_F_SUSER },
 535
 536         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 537           BC_F_COPYIN|BC_F_SUSER },
 538         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 539           BC_F_COPYIN|BC_F_SUSER },
 540
 541         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 542           BC_F_COPYOUT },
 543
 544         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 545           BC_F_COPYOUT },
 546
 547         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 548           BC_F_COPYIN|BC_F_COPYOUT },
 549
 550         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 551           BC_F_COPYIN|BC_F_SUSER },
 552
 553         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 554           BC_F_COPYIN|BC_F_SUSER },
 555
 556         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 557           BC_F_COPYIN|BC_F_SUSER },
 558
 559 };
 560 const int bridge_control_table_size = nitems(bridge_control_table);
 561
 562 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 563 #define V_bridge_list   VNET(bridge_list)
 564 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 565                                             "if_bridge list")
 566 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 567 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 568 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 569
 570 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 571 #define V_bridge_cloner VNET(bridge_cloner)
 572
 573 static const char bridge_name[] = "bridge";
 574
 575 static void
 576 vnet_bridge_init(const void *unused __unused)
 577 {
 578
 579         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 580             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 581             UMA_ALIGN_PTR, 0);
 582         BRIDGE_LIST_LOCK_INIT();
 583         LIST_INIT(&V_bridge_list);
 584         V_bridge_cloner = if_clone_simple(bridge_name,
 585             bridge_clone_create, bridge_clone_destroy, 0);
 586 }
 587 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 588     vnet_bridge_init, NULL);
 589
 590 static void
 591 vnet_bridge_uninit(const void *unused __unused)
 592 {
 593
 594         if_clone_detach(V_bridge_cloner);
 595         V_bridge_cloner = NULL;
 596         BRIDGE_LIST_LOCK_DESTROY();
 597
 598         /* Callbacks may use the UMA zone. */
 599         epoch_drain_callbacks(net_epoch_preempt);
 600
 601         uma_zdestroy(V_bridge_rtnode_zone);
 602 }
 603 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 604     vnet_bridge_uninit, NULL);
 605
 606 static int
 607 bridge_modevent(module_t mod, int type, void *data)
 608 {
 609
 610         switch (type) {
 611         case MOD_LOAD:
 612                 bridge_dn_p = bridge_dummynet;
 613                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 614                     ifnet_departure_event, bridge_ifdetach, NULL,
 615                     EVENTHANDLER_PRI_ANY);
 616                 break;
 617         case MOD_UNLOAD:
 618                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 619                     bridge_detach_cookie);
 620                 bridge_dn_p = NULL;
 621                 break;
 622         default:
 623                 return (EOPNOTSUPP);
 624         }
 625         return (0);
 626 }
 627
 628 static moduledata_t bridge_mod = {
 629         "if_bridge",
 630         bridge_modevent,
 631         0
 632 };
 633
 634 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 635 MODULE_VERSION(if_bridge, 1);
 636 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 637
 638 /*
 639  * handler for net.link.bridge.ipfw
 640  */
 641 static int
 642 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 643 {
 644         int enable = V_pfil_ipfw;
 645         int error;
 646
 647         error = sysctl_handle_int(oidp, &enable, 0, req);
 648         enable &= 1;
 649
 650         if (enable != V_pfil_ipfw) {
 651                 V_pfil_ipfw = enable;
 652
 653                 /*
 654                  * Disable pfil so that ipfw doesnt run twice, if the user
 655                  * really wants both then they can re-enable pfil_bridge and/or
 656                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 657                  * layer2 type.
 658                  */
 659                 if (V_pfil_ipfw) {
 660                         V_pfil_onlyip = 0;
 661                         V_pfil_bridge = 0;
 662                         V_pfil_member = 0;
 663                 }
 664         }
 665
 666         return (error);
 667 }
 668 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 669     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
 670     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 671     "Layer2 filter with IPFW");
 672
 673 /*
 674  * bridge_clone_create:
 675  *
 676  *      Create a new bridge instance.
 677  */
 678 static int
 679 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 680 {
 681         struct bridge_softc *sc;
 682         struct ifnet *ifp;
 683
 684         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 685         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 686         if (ifp == NULL) {
 687                 free(sc, M_DEVBUF);
 688                 return (ENOSPC);
 689         }
 690
 691         BRIDGE_LOCK_INIT(sc);
 692         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 693         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 694
 695         /* Initialize our routing table. */
 696         bridge_rtable_init(sc);
 697
 698         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 699
 700         CK_LIST_INIT(&sc->sc_iflist);
 701         CK_LIST_INIT(&sc->sc_spanlist);
 702
 703         ifp->if_softc = sc;
 704         if_initname(ifp, bridge_name, unit);
 705         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 706         ifp->if_ioctl = bridge_ioctl;
 707         ifp->if_transmit = bridge_transmit;
 708         ifp->if_qflush = bridge_qflush;
 709         ifp->if_init = bridge_init;
 710         ifp->if_type = IFT_BRIDGE;
 711
 712         ether_gen_addr(ifp, &sc->sc_defaddr);
 713
 714         bstp_attach(&sc->sc_stp, &bridge_ops);
 715         ether_ifattach(ifp, sc->sc_defaddr.octet);
 716         /* Now undo some of the damage... */
 717         ifp->if_baudrate = 0;
 718         ifp->if_type = IFT_BRIDGE;
 719
 720         BRIDGE_LIST_LOCK();
 721         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 722         BRIDGE_LIST_UNLOCK();
 723
 724         return (0);
 725 }
 726
 727 static void
 728 bridge_clone_destroy_cb(struct epoch_context *ctx)
 729 {
 730         struct bridge_softc *sc;
 731
 732         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 733
 734         BRIDGE_LOCK_DESTROY(sc);
 735         free(sc, M_DEVBUF);
 736 }
 737
 738 /*
 739  * bridge_clone_destroy:
 740  *
 741  *      Destroy a bridge instance.
 742  */
 743 static void
 744 bridge_clone_destroy(struct ifnet *ifp)
 745 {
 746         struct bridge_softc *sc = ifp->if_softc;
 747         struct bridge_iflist *bif;
 748         struct epoch_tracker et;
 749
 750         BRIDGE_LOCK(sc);
 751
 752         bridge_stop(ifp, 1);
 753         ifp->if_flags &= ~IFF_UP;
 754
 755         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 756                 bridge_delete_member(sc, bif, 0);
 757
 758         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 759                 bridge_delete_span(sc, bif);
 760         }
 761
 762         /* Tear down the routing table. */
 763         bridge_rtable_fini(sc);
 764
 765         BRIDGE_UNLOCK(sc);
 766
 767         NET_EPOCH_ENTER(et);
 768
 769         callout_drain(&sc->sc_brcallout);
 770
 771         BRIDGE_LIST_LOCK();
 772         LIST_REMOVE(sc, sc_list);
 773         BRIDGE_LIST_UNLOCK();
 774
 775         bstp_detach(&sc->sc_stp);
 776         NET_EPOCH_EXIT(et);
 777
 778         ether_ifdetach(ifp);
 779         if_free(ifp);
 780
 781         NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
 782 }
 783
 784 /*
 785  * bridge_ioctl:
 786  *
 787  *      Handle a control request from the operator.
 788  */
 789 static int
 790 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 791 {
 792         struct bridge_softc *sc = ifp->if_softc;
 793         struct ifreq *ifr = (struct ifreq *)data;
 794         struct bridge_iflist *bif;
 795         struct thread *td = curthread;
 796         union {
 797                 struct ifbreq ifbreq;
 798                 struct ifbifconf ifbifconf;
 799                 struct ifbareq ifbareq;
 800                 struct ifbaconf ifbaconf;
 801                 struct ifbrparam ifbrparam;
 802                 struct ifbropreq ifbropreq;
 803         } args;
 804         struct ifdrv *ifd = (struct ifdrv *) data;
 805         const struct bridge_control *bc;
 806         int error = 0, oldmtu;
 807
 808         BRIDGE_LOCK(sc);
 809
 810         switch (cmd) {
 811         case SIOCADDMULTI:
 812         case SIOCDELMULTI:
 813                 break;
 814
 815         case SIOCGDRVSPEC:
 816         case SIOCSDRVSPEC:
 817                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 818                         error = EINVAL;
 819                         break;
 820                 }
 821                 bc = &bridge_control_table[ifd->ifd_cmd];
 822
 823                 if (cmd == SIOCGDRVSPEC &&
 824                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 825                         error = EINVAL;
 826                         break;
 827                 }
 828                 else if (cmd == SIOCSDRVSPEC &&
 829                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 830                         error = EINVAL;
 831                         break;
 832                 }
 833
 834                 if (bc->bc_flags & BC_F_SUSER) {
 835                         error = priv_check(td, PRIV_NET_BRIDGE);
 836                         if (error)
 837                                 break;
 838                 }
 839
 840                 if (ifd->ifd_len != bc->bc_argsize ||
 841                     ifd->ifd_len > sizeof(args)) {
 842                         error = EINVAL;
 843                         break;
 844                 }
 845
 846                 bzero(&args, sizeof(args));
 847                 if (bc->bc_flags & BC_F_COPYIN) {
 848                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 849                         if (error)
 850                                 break;
 851                 }
 852
 853                 oldmtu = ifp->if_mtu;
 854                 error = (*bc->bc_func)(sc, &args);
 855                 if (error)
 856                         break;
 857
 858                 /*
 859                  * Bridge MTU may change during addition of the first port.
 860                  * If it did, do network layer specific procedure.
 861                  */
 862                 if (ifp->if_mtu != oldmtu) {
 863 #ifdef INET6
 864                         nd6_setmtu(ifp);
 865 #endif
 866                         rt_updatemtu(ifp);
 867                 }
 868
 869                 if (bc->bc_flags & BC_F_COPYOUT)
 870                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 871
 872                 break;
 873
 874         case SIOCSIFFLAGS:
 875                 if (!(ifp->if_flags & IFF_UP) &&
 876                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 877                         /*
 878                          * If interface is marked down and it is running,
 879                          * then stop and disable it.
 880                          */
 881                         bridge_stop(ifp, 1);
 882                 } else if ((ifp->if_flags & IFF_UP) &&
 883                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 884                         /*
 885                          * If interface is marked up and it is stopped, then
 886                          * start it.
 887                          */
 888                         BRIDGE_UNLOCK(sc);
 889                         (*ifp->if_init)(sc);
 890                         BRIDGE_LOCK(sc);
 891                 }
 892                 break;
 893
 894         case SIOCSIFMTU:
 895                 if (ifr->ifr_mtu < 576) {
 896                         error = EINVAL;
 897                         break;
 898                 }
 899                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 900                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 901                         break;
 902                 }
 903                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 904                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 905                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 906                                     " != %d\n", sc->sc_ifp->if_xname,
 907                                     bif->bif_ifp->if_mtu,
 908                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 909                                 error = EINVAL;
 910                                 break;
 911                         }
 912                 }
 913                 if (!error)
 914                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 915                 break;
 916         default:
 917                 /*
 918                  * drop the lock as ether_ioctl() will call bridge_start() and
 919                  * cause the lock to be recursed.
 920                  */
 921                 BRIDGE_UNLOCK(sc);
 922                 error = ether_ioctl(ifp, cmd, data);
 923                 BRIDGE_LOCK(sc);
 924                 break;
 925         }
 926
 927         BRIDGE_UNLOCK(sc);
 928
 929         return (error);
 930 }
 931
 932 /*
 933  * bridge_mutecaps:
 934  *
 935  *      Clear or restore unwanted capabilities on the member interface
 936  */
 937 static void
 938 bridge_mutecaps(struct bridge_softc *sc)
 939 {
 940         struct bridge_iflist *bif;
 941         int enabled, mask;
 942
 943         BRIDGE_LOCK_ASSERT(sc);
 944
 945         /* Initial bitmask of capabilities to test */
 946         mask = BRIDGE_IFCAPS_MASK;
 947
 948         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 949                 /* Every member must support it or its disabled */
 950                 mask &= bif->bif_savedcaps;
 951         }
 952
 953         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 954                 enabled = bif->bif_ifp->if_capenable;
 955                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 956                 /* strip off mask bits and enable them again if allowed */
 957                 enabled &= ~BRIDGE_IFCAPS_MASK;
 958                 enabled |= mask;
 959                 bridge_set_ifcap(sc, bif, enabled);
 960         }
 961 }
 962
 963 static void
 964 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 965 {
 966         struct ifnet *ifp = bif->bif_ifp;
 967         struct ifreq ifr;
 968         int error, mask, stuck;
 969
 970         bzero(&ifr, sizeof(ifr));
 971         ifr.ifr_reqcap = set;
 972
 973         if (ifp->if_capenable != set) {
 974                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 975                 if (error)
 976                         if_printf(sc->sc_ifp,
 977                             "error setting capabilities on %s: %d\n",
 978                             ifp->if_xname, error);
 979                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
 980                 stuck = ifp->if_capenable & mask & ~set;
 981                 if (stuck != 0)
 982                         if_printf(sc->sc_ifp,
 983                             "can't disable some capabilities on %s: 0x%x\n",
 984                             ifp->if_xname, stuck);
 985         }
 986 }
 987
 988 /*
 989  * bridge_lookup_member:
 990  *
 991  *      Lookup a bridge member interface.
 992  */
 993 static struct bridge_iflist *
 994 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 995 {
 996         struct bridge_iflist *bif;
 997         struct ifnet *ifp;
 998
 999         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1000
1001         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1002                 ifp = bif->bif_ifp;
1003                 if (strcmp(ifp->if_xname, name) == 0)
1004                         return (bif);
1005         }
1006
1007         return (NULL);
1008 }
1009
1010 /*
1011  * bridge_lookup_member_if:
1012  *
1013  *      Lookup a bridge member interface by ifnet*.
1014  */
1015 static struct bridge_iflist *
1016 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1017 {
1018         struct bridge_iflist *bif;
1019
1020         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1021
1022         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1023                 if (bif->bif_ifp == member_ifp)
1024                         return (bif);
1025         }
1026
1027         return (NULL);
1028 }
1029
1030 static void
1031 bridge_delete_member_cb(struct epoch_context *ctx)
1032 {
1033         struct bridge_iflist *bif;
1034
1035         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1036
1037         free(bif, M_DEVBUF);
1038 }
1039
1040 /*
1041  * bridge_delete_member:
1042  *
1043  *      Delete the specified member interface.
1044  */
1045 static void
1046 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1047     int gone)
1048 {
1049         struct ifnet *ifs = bif->bif_ifp;
1050         struct ifnet *fif = NULL;
1051         struct bridge_iflist *bifl;
1052
1053         BRIDGE_LOCK_ASSERT(sc);
1054
1055         if (bif->bif_flags & IFBIF_STP)
1056                 bstp_disable(&bif->bif_stp);
1057
1058         ifs->if_bridge = NULL;
1059         CK_LIST_REMOVE(bif, bif_next);
1060
1061         /*
1062          * If removing the interface that gave the bridge its mac address, set
1063          * the mac address of the bridge to the address of the next member, or
1064          * to its default address if no members are left.
1065          */
1066         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1067                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1068                         bcopy(&sc->sc_defaddr,
1069                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1070                         sc->sc_ifaddr = NULL;
1071                 } else {
1072                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1073                         fif = bifl->bif_ifp;
1074                         bcopy(IF_LLADDR(fif),
1075                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1076                         sc->sc_ifaddr = fif;
1077                 }
1078                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1079         }
1080
1081         bridge_linkcheck(sc);
1082         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1083         BRIDGE_RT_LOCK(sc);
1084         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1085         BRIDGE_RT_UNLOCK(sc);
1086         KASSERT(bif->bif_addrcnt == 0,
1087             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1088
1089         ifs->if_bridge_output = NULL;
1090         ifs->if_bridge_input = NULL;
1091         ifs->if_bridge_linkstate = NULL;
1092         if (!gone) {
1093                 switch (ifs->if_type) {
1094                 case IFT_ETHER:
1095                 case IFT_L2VLAN:
1096                         /*
1097                          * Take the interface out of promiscuous mode, but only
1098                          * if it was promiscuous in the first place. It might
1099                          * not be if we're in the bridge_ioctl_add() error path.
1100                          */
1101                         if (ifs->if_flags & IFF_PROMISC)
1102                                 (void) ifpromisc(ifs, 0);
1103                         break;
1104
1105                 case IFT_GIF:
1106                         break;
1107
1108                 default:
1109 #ifdef DIAGNOSTIC
1110                         panic("bridge_delete_member: impossible");
1111 #endif
1112                         break;
1113                 }
1114                 /* reneable any interface capabilities */
1115                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1116         }
1117         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1118
1119         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1120 }
1121
1122 /*
1123  * bridge_delete_span:
1124  *
1125  *      Delete the specified span interface.
1126  */
1127 static void
1128 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1129 {
1130         BRIDGE_LOCK_ASSERT(sc);
1131
1132         KASSERT(bif->bif_ifp->if_bridge == NULL,
1133             ("%s: not a span interface", __func__));
1134
1135         CK_LIST_REMOVE(bif, bif_next);
1136
1137         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1138 }
1139
1140 static int
1141 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1142 {
1143         struct ifbreq *req = arg;
1144         struct bridge_iflist *bif = NULL;
1145         struct ifnet *ifs;
1146         int error = 0;
1147
1148         ifs = ifunit(req->ifbr_ifsname);
1149         if (ifs == NULL)
1150                 return (ENOENT);
1151         if (ifs->if_ioctl == NULL)      /* must be supported */
1152                 return (EINVAL);
1153
1154         /* If it's in the span list, it can't be a member. */
1155         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1156                 if (ifs == bif->bif_ifp)
1157                         return (EBUSY);
1158
1159         if (ifs->if_bridge == sc)
1160                 return (EEXIST);
1161
1162         if (ifs->if_bridge != NULL)
1163                 return (EBUSY);
1164
1165         switch (ifs->if_type) {
1166         case IFT_ETHER:
1167         case IFT_L2VLAN:
1168         case IFT_GIF:
1169                 /* permitted interface types */
1170                 break;
1171         default:
1172                 return (EINVAL);
1173         }
1174
1175 #ifdef INET6
1176         /*
1177          * Two valid inet6 addresses with link-local scope must not be
1178          * on the parent interface and the member interfaces at the
1179          * same time.  This restriction is needed to prevent violation
1180          * of link-local scope zone.  Attempts to add a member
1181          * interface which has inet6 addresses when the parent has
1182          * inet6 triggers removal of all inet6 addresses on the member
1183          * interface.
1184          */
1185
1186         /* Check if the parent interface has a link-local scope addr. */
1187         if (V_allow_llz_overlap == 0 &&
1188             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1189                 /*
1190                  * If any, remove all inet6 addresses from the member
1191                  * interfaces.
1192                  */
1193                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1194                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1195                                 in6_ifdetach(bif->bif_ifp);
1196                                 if_printf(sc->sc_ifp,
1197                                     "IPv6 addresses on %s have been removed "
1198                                     "before adding it as a member to prevent "
1199                                     "IPv6 address scope violation.\n",
1200                                     bif->bif_ifp->if_xname);
1201                         }
1202                 }
1203                 if (in6ifa_llaonifp(ifs)) {
1204                         in6_ifdetach(ifs);
1205                         if_printf(sc->sc_ifp,
1206                             "IPv6 addresses on %s have been removed "
1207                             "before adding it as a member to prevent "
1208                             "IPv6 address scope violation.\n",
1209                             ifs->if_xname);
1210                 }
1211         }
1212 #endif
1213         /* Allow the first Ethernet member to define the MTU */
1214         if (CK_LIST_EMPTY(&sc->sc_iflist))
1215                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1216         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1217                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1218                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1219                 return (EINVAL);
1220         }
1221
1222         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1223         if (bif == NULL)
1224                 return (ENOMEM);
1225
1226         bif->bif_ifp = ifs;
1227         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1228         bif->bif_savedcaps = ifs->if_capenable;
1229
1230         /*
1231          * Assign the interface's MAC address to the bridge if it's the first
1232          * member and the MAC address of the bridge has not been changed from
1233          * the default randomly generated one.
1234          */
1235         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1236             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1237                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1238                 sc->sc_ifaddr = ifs;
1239                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1240         }
1241
1242         ifs->if_bridge = sc;
1243         ifs->if_bridge_output = bridge_output;
1244         ifs->if_bridge_input = bridge_input;
1245         ifs->if_bridge_linkstate = bridge_linkstate;
1246         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1247         /*
1248          * XXX: XLOCK HERE!?!
1249          *
1250          * NOTE: insert_***HEAD*** should be safe for the traversals.
1251          */
1252         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1253
1254         /* Set interface capabilities to the intersection set of all members */
1255         bridge_mutecaps(sc);
1256         bridge_linkcheck(sc);
1257
1258         /* Place the interface into promiscuous mode */
1259         switch (ifs->if_type) {
1260                 case IFT_ETHER:
1261                 case IFT_L2VLAN:
1262                         error = ifpromisc(ifs, 1);
1263                         break;
1264         }
1265
1266         if (error)
1267                 bridge_delete_member(sc, bif, 0);
1268         return (error);
1269 }
1270
1271 static int
1272 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1273 {
1274         struct ifbreq *req = arg;
1275         struct bridge_iflist *bif;
1276
1277         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1278         if (bif == NULL)
1279                 return (ENOENT);
1280
1281         bridge_delete_member(sc, bif, 0);
1282
1283         return (0);
1284 }
1285
1286 static int
1287 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1288 {
1289         struct ifbreq *req = arg;
1290         struct bridge_iflist *bif;
1291         struct bstp_port *bp;
1292
1293         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1294         if (bif == NULL)
1295                 return (ENOENT);
1296
1297         bp = &bif->bif_stp;
1298         req->ifbr_ifsflags = bif->bif_flags;
1299         req->ifbr_state = bp->bp_state;
1300         req->ifbr_priority = bp->bp_priority;
1301         req->ifbr_path_cost = bp->bp_path_cost;
1302         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1303         req->ifbr_proto = bp->bp_protover;
1304         req->ifbr_role = bp->bp_role;
1305         req->ifbr_stpflags = bp->bp_flags;
1306         req->ifbr_addrcnt = bif->bif_addrcnt;
1307         req->ifbr_addrmax = bif->bif_addrmax;
1308         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1309
1310         /* Copy STP state options as flags */
1311         if (bp->bp_operedge)
1312                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1313         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1314                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1315         if (bp->bp_ptp_link)
1316                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1317         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1318                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1319         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1320                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1321         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1322                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1323         return (0);
1324 }
1325
1326 static int
1327 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1328 {
1329         struct ifbreq *req = arg;
1330         struct bridge_iflist *bif;
1331         struct bstp_port *bp;
1332         int error;
1333
1334         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1335         if (bif == NULL)
1336                 return (ENOENT);
1337         bp = &bif->bif_stp;
1338
1339         if (req->ifbr_ifsflags & IFBIF_SPAN)
1340                 /* SPAN is readonly */
1341                 return (EINVAL);
1342
1343         if (req->ifbr_ifsflags & IFBIF_STP) {
1344                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1345                         error = bstp_enable(&bif->bif_stp);
1346                         if (error)
1347                                 return (error);
1348                 }
1349         } else {
1350                 if ((bif->bif_flags & IFBIF_STP) != 0)
1351                         bstp_disable(&bif->bif_stp);
1352         }
1353
1354         /* Pass on STP flags */
1355         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1356         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1357         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1358         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1359
1360         /* Save the bits relating to the bridge */
1361         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1362
1363         return (0);
1364 }
1365
1366 static int
1367 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1368 {
1369         struct ifbrparam *param = arg;
1370
1371         sc->sc_brtmax = param->ifbrp_csize;
1372         bridge_rttrim(sc);
1373
1374         return (0);
1375 }
1376
1377 static int
1378 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1379 {
1380         struct ifbrparam *param = arg;
1381
1382         param->ifbrp_csize = sc->sc_brtmax;
1383
1384         return (0);
1385 }
1386
1387 static int
1388 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1389 {
1390         struct ifbifconf *bifc = arg;
1391         struct bridge_iflist *bif;
1392         struct ifbreq breq;
1393         char *buf, *outbuf;
1394         int count, buflen, len, error = 0;
1395
1396         count = 0;
1397         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1398                 count++;
1399         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1400                 count++;
1401
1402         buflen = sizeof(breq) * count;
1403         if (bifc->ifbic_len == 0) {
1404                 bifc->ifbic_len = buflen;
1405                 return (0);
1406         }
1407         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1408         if (outbuf == NULL)
1409                 return (ENOMEM);
1410
1411         count = 0;
1412         buf = outbuf;
1413         len = min(bifc->ifbic_len, buflen);
1414         bzero(&breq, sizeof(breq));
1415         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1416                 if (len < sizeof(breq))
1417                         break;
1418
1419                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1420                     sizeof(breq.ifbr_ifsname));
1421                 /* Fill in the ifbreq structure */
1422                 error = bridge_ioctl_gifflags(sc, &breq);
1423                 if (error)
1424                         break;
1425                 memcpy(buf, &breq, sizeof(breq));
1426                 count++;
1427                 buf += sizeof(breq);
1428                 len -= sizeof(breq);
1429         }
1430         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1431                 if (len < sizeof(breq))
1432                         break;
1433
1434                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1435                     sizeof(breq.ifbr_ifsname));
1436                 breq.ifbr_ifsflags = bif->bif_flags;
1437                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1438                 memcpy(buf, &breq, sizeof(breq));
1439                 count++;
1440                 buf += sizeof(breq);
1441                 len -= sizeof(breq);
1442         }
1443
1444         bifc->ifbic_len = sizeof(breq) * count;
1445         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1446         free(outbuf, M_TEMP);
1447         return (error);
1448 }
1449
1450 static int
1451 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1452 {
1453         struct ifbaconf *bac = arg;
1454         struct bridge_rtnode *brt;
1455         struct ifbareq bareq;
1456         char *buf, *outbuf;
1457         int count, buflen, len, error = 0;
1458
1459         if (bac->ifbac_len == 0)
1460                 return (0);
1461
1462         count = 0;
1463         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1464                 count++;
1465         buflen = sizeof(bareq) * count;
1466
1467         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1468         if (outbuf == NULL)
1469                 return (ENOMEM);
1470
1471         count = 0;
1472         buf = outbuf;
1473         len = min(bac->ifbac_len, buflen);
1474         bzero(&bareq, sizeof(bareq));
1475         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1476                 if (len < sizeof(bareq))
1477                         goto out;
1478                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1479                     sizeof(bareq.ifba_ifsname));
1480                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1481                 bareq.ifba_vlan = brt->brt_vlan;
1482                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1483                                 time_uptime < brt->brt_expire)
1484                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1485                 else
1486                         bareq.ifba_expire = 0;
1487                 bareq.ifba_flags = brt->brt_flags;
1488
1489                 memcpy(buf, &bareq, sizeof(bareq));
1490                 count++;
1491                 buf += sizeof(bareq);
1492                 len -= sizeof(bareq);
1493         }
1494 out:
1495         bac->ifbac_len = sizeof(bareq) * count;
1496         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1497         free(outbuf, M_TEMP);
1498         return (error);
1499 }
1500
1501 static int
1502 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1503 {
1504         struct ifbareq *req = arg;
1505         struct bridge_iflist *bif;
1506         struct epoch_tracker et;
1507         int error;
1508
1509         NET_EPOCH_ENTER(et);
1510         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1511         if (bif == NULL) {
1512                 NET_EPOCH_EXIT(et);
1513                 return (ENOENT);
1514         }
1515
1516         /* bridge_rtupdate() may acquire the lock. */
1517         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1518             req->ifba_flags);
1519         NET_EPOCH_EXIT(et);
1520
1521         return (error);
1522 }
1523
1524 static int
1525 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1526 {
1527         struct ifbrparam *param = arg;
1528
1529         sc->sc_brttimeout = param->ifbrp_ctime;
1530         return (0);
1531 }
1532
1533 static int
1534 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1535 {
1536         struct ifbrparam *param = arg;
1537
1538         param->ifbrp_ctime = sc->sc_brttimeout;
1539         return (0);
1540 }
1541
1542 static int
1543 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1544 {
1545         struct ifbareq *req = arg;
1546
1547         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1548 }
1549
1550 static int
1551 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1552 {
1553         struct ifbreq *req = arg;
1554
1555         BRIDGE_RT_LOCK(sc);
1556         bridge_rtflush(sc, req->ifbr_ifsflags);
1557         BRIDGE_RT_UNLOCK(sc);
1558
1559         return (0);
1560 }
1561
1562 static int
1563 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1564 {
1565         struct ifbrparam *param = arg;
1566         struct bstp_state *bs = &sc->sc_stp;
1567
1568         param->ifbrp_prio = bs->bs_bridge_priority;
1569         return (0);
1570 }
1571
1572 static int
1573 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1574 {
1575         struct ifbrparam *param = arg;
1576
1577         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1578 }
1579
1580 static int
1581 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1582 {
1583         struct ifbrparam *param = arg;
1584         struct bstp_state *bs = &sc->sc_stp;
1585
1586         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1587         return (0);
1588 }
1589
1590 static int
1591 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1592 {
1593         struct ifbrparam *param = arg;
1594
1595         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1596 }
1597
1598 static int
1599 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1600 {
1601         struct ifbrparam *param = arg;
1602         struct bstp_state *bs = &sc->sc_stp;
1603
1604         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1605         return (0);
1606 }
1607
1608 static int
1609 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1610 {
1611         struct ifbrparam *param = arg;
1612
1613         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1614 }
1615
1616 static int
1617 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1618 {
1619         struct ifbrparam *param = arg;
1620         struct bstp_state *bs = &sc->sc_stp;
1621
1622         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1623         return (0);
1624 }
1625
1626 static int
1627 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1628 {
1629         struct ifbrparam *param = arg;
1630
1631         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1632 }
1633
1634 static int
1635 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1636 {
1637         struct ifbreq *req = arg;
1638         struct bridge_iflist *bif;
1639
1640         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1641         if (bif == NULL)
1642                 return (ENOENT);
1643
1644         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1645 }
1646
1647 static int
1648 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1649 {
1650         struct ifbreq *req = arg;
1651         struct bridge_iflist *bif;
1652
1653         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1654         if (bif == NULL)
1655                 return (ENOENT);
1656
1657         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1658 }
1659
1660 static int
1661 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1662 {
1663         struct ifbreq *req = arg;
1664         struct bridge_iflist *bif;
1665
1666         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1667         if (bif == NULL)
1668                 return (ENOENT);
1669
1670         bif->bif_addrmax = req->ifbr_addrmax;
1671         return (0);
1672 }
1673
1674 static int
1675 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1676 {
1677         struct ifbreq *req = arg;
1678         struct bridge_iflist *bif = NULL;
1679         struct ifnet *ifs;
1680
1681         ifs = ifunit(req->ifbr_ifsname);
1682         if (ifs == NULL)
1683                 return (ENOENT);
1684
1685         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1686                 if (ifs == bif->bif_ifp)
1687                         return (EBUSY);
1688
1689         if (ifs->if_bridge != NULL)
1690                 return (EBUSY);
1691
1692         switch (ifs->if_type) {
1693                 case IFT_ETHER:
1694                 case IFT_GIF:
1695                 case IFT_L2VLAN:
1696                         break;
1697                 default:
1698                         return (EINVAL);
1699         }
1700
1701         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1702         if (bif == NULL)
1703                 return (ENOMEM);
1704
1705         bif->bif_ifp = ifs;
1706         bif->bif_flags = IFBIF_SPAN;
1707
1708         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1709
1710         return (0);
1711 }
1712
1713 static int
1714 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1715 {
1716         struct ifbreq *req = arg;
1717         struct bridge_iflist *bif;
1718         struct ifnet *ifs;
1719
1720         ifs = ifunit(req->ifbr_ifsname);
1721         if (ifs == NULL)
1722                 return (ENOENT);
1723
1724         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1725                 if (ifs == bif->bif_ifp)
1726                         break;
1727
1728         if (bif == NULL)
1729                 return (ENOENT);
1730
1731         bridge_delete_span(sc, bif);
1732
1733         return (0);
1734 }
1735
1736 static int
1737 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1738 {
1739         struct ifbropreq *req = arg;
1740         struct bstp_state *bs = &sc->sc_stp;
1741         struct bstp_port *root_port;
1742
1743         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1744         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1745         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1746
1747         root_port = bs->bs_root_port;
1748         if (root_port == NULL)
1749                 req->ifbop_root_port = 0;
1750         else
1751                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1752
1753         req->ifbop_holdcount = bs->bs_txholdcount;
1754         req->ifbop_priority = bs->bs_bridge_priority;
1755         req->ifbop_protocol = bs->bs_protover;
1756         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1757         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1758         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1759         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1760         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1761         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1762
1763         return (0);
1764 }
1765
1766 static int
1767 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1768 {
1769         struct ifbrparam *param = arg;
1770
1771         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1772         return (0);
1773 }
1774
1775 static int
1776 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1777 {
1778         struct ifbpstpconf *bifstp = arg;
1779         struct bridge_iflist *bif;
1780         struct bstp_port *bp;
1781         struct ifbpstpreq bpreq;
1782         char *buf, *outbuf;
1783         int count, buflen, len, error = 0;
1784
1785         count = 0;
1786         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1787                 if ((bif->bif_flags & IFBIF_STP) != 0)
1788                         count++;
1789         }
1790
1791         buflen = sizeof(bpreq) * count;
1792         if (bifstp->ifbpstp_len == 0) {
1793                 bifstp->ifbpstp_len = buflen;
1794                 return (0);
1795         }
1796
1797         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1798         if (outbuf == NULL)
1799                 return (ENOMEM);
1800
1801         count = 0;
1802         buf = outbuf;
1803         len = min(bifstp->ifbpstp_len, buflen);
1804         bzero(&bpreq, sizeof(bpreq));
1805         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1806                 if (len < sizeof(bpreq))
1807                         break;
1808
1809                 if ((bif->bif_flags & IFBIF_STP) == 0)
1810                         continue;
1811
1812                 bp = &bif->bif_stp;
1813                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1814                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1815                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1816                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1817                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1818                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1819
1820                 memcpy(buf, &bpreq, sizeof(bpreq));
1821                 count++;
1822                 buf += sizeof(bpreq);
1823                 len -= sizeof(bpreq);
1824         }
1825
1826         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1827         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1828         free(outbuf, M_TEMP);
1829         return (error);
1830 }
1831
1832 static int
1833 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1834 {
1835         struct ifbrparam *param = arg;
1836
1837         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1838 }
1839
1840 static int
1841 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1842 {
1843         struct ifbrparam *param = arg;
1844
1845         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1846 }
1847
1848 /*
1849  * bridge_ifdetach:
1850  *
1851  *      Detach an interface from a bridge.  Called when a member
1852  *      interface is detaching.
1853  */
1854 static void
1855 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1856 {
1857         struct bridge_softc *sc = ifp->if_bridge;
1858         struct bridge_iflist *bif;
1859
1860         if (ifp->if_flags & IFF_RENAMING)
1861                 return;
1862         if (V_bridge_cloner == NULL) {
1863                 /*
1864                  * This detach handler can be called after
1865                  * vnet_bridge_uninit().  Just return in that case.
1866                  */
1867                 return;
1868         }
1869         /* Check if the interface is a bridge member */
1870         if (sc != NULL) {
1871                 BRIDGE_LOCK(sc);
1872
1873                 bif = bridge_lookup_member_if(sc, ifp);
1874                 if (bif != NULL)
1875                         bridge_delete_member(sc, bif, 1);
1876
1877                 BRIDGE_UNLOCK(sc);
1878                 return;
1879         }
1880
1881         /* Check if the interface is a span port */
1882         BRIDGE_LIST_LOCK();
1883         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1884                 BRIDGE_LOCK(sc);
1885                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1886                         if (ifp == bif->bif_ifp) {
1887                                 bridge_delete_span(sc, bif);
1888                                 break;
1889                         }
1890
1891                 BRIDGE_UNLOCK(sc);
1892         }
1893         BRIDGE_LIST_UNLOCK();
1894 }
1895
1896 /*
1897  * bridge_init:
1898  *
1899  *      Initialize a bridge interface.
1900  */
1901 static void
1902 bridge_init(void *xsc)
1903 {
1904         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1905         struct ifnet *ifp = sc->sc_ifp;
1906
1907         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1908                 return;
1909
1910         BRIDGE_LOCK(sc);
1911         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1912             bridge_timer, sc);
1913
1914         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1915         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1916
1917         BRIDGE_UNLOCK(sc);
1918 }
1919
1920 /*
1921  * bridge_stop:
1922  *
1923  *      Stop the bridge interface.
1924  */
1925 static void
1926 bridge_stop(struct ifnet *ifp, int disable)
1927 {
1928         struct bridge_softc *sc = ifp->if_softc;
1929
1930         BRIDGE_LOCK_ASSERT(sc);
1931
1932         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1933                 return;
1934
1935         BRIDGE_RT_LOCK(sc);
1936         callout_stop(&sc->sc_brcallout);
1937
1938         bstp_stop(&sc->sc_stp);
1939
1940         bridge_rtflush(sc, IFBF_FLUSHDYN);
1941         BRIDGE_RT_UNLOCK(sc);
1942
1943         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1944 }
1945
1946 /*
1947  * bridge_enqueue:
1948  *
1949  *      Enqueue a packet on a bridge member interface.
1950  *
1951  */
1952 static int
1953 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1954 {
1955         int len, err = 0;
1956         short mflags;
1957         struct mbuf *m0;
1958
1959         /* We may be sending a fragment so traverse the mbuf */
1960         for (; m; m = m0) {
1961                 m0 = m->m_nextpkt;
1962                 m->m_nextpkt = NULL;
1963                 len = m->m_pkthdr.len;
1964                 mflags = m->m_flags;
1965
1966                 /*
1967                  * If underlying interface can not do VLAN tag insertion itself
1968                  * then attach a packet tag that holds it.
1969                  */
1970                 if ((m->m_flags & M_VLANTAG) &&
1971                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1972                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1973                         if (m == NULL) {
1974                                 if_printf(dst_ifp,
1975                                     "unable to prepend VLAN header\n");
1976                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
1977                                 continue;
1978                         }
1979                         m->m_flags &= ~M_VLANTAG;
1980                 }
1981
1982                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
1983                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1984                         m_freem(m0);
1985                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
1986                         break;
1987                 }
1988
1989                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
1990                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
1991                 if (mflags & M_MCAST)
1992                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
1993         }
1994
1995         return (err);
1996 }
1997
1998 /*
1999  * bridge_dummynet:
2000  *
2001  *      Receive a queued packet from dummynet and pass it on to the output
2002  *      interface.
2003  *
2004  *      The mbuf has the Ethernet header already attached.
2005  */
2006 static void
2007 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2008 {
2009         struct bridge_softc *sc;
2010
2011         sc = ifp->if_bridge;
2012
2013         /*
2014          * The packet didnt originate from a member interface. This should only
2015          * ever happen if a member interface is removed while packets are
2016          * queued for it.
2017          */
2018         if (sc == NULL) {
2019                 m_freem(m);
2020                 return;
2021         }
2022
2023         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2024 #ifdef INET6
2025             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2026 #endif
2027             ) {
2028                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2029                         return;
2030                 if (m == NULL)
2031                         return;
2032         }
2033
2034         bridge_enqueue(sc, ifp, m);
2035 }
2036
2037 /*
2038  * bridge_output:
2039  *
2040  *      Send output from a bridge member interface.  This
2041  *      performs the bridging function for locally originated
2042  *      packets.
2043  *
2044  *      The mbuf has the Ethernet header already attached.  We must
2045  *      enqueue or free the mbuf before returning.
2046  */
2047 static int
2048 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2049     struct rtentry *rt)
2050 {
2051         struct ether_header *eh;
2052         struct ifnet *bifp, *dst_if;
2053         struct bridge_softc *sc;
2054         uint16_t vlan;
2055
2056         NET_EPOCH_ASSERT();
2057
2058         if (m->m_len < ETHER_HDR_LEN) {
2059                 m = m_pullup(m, ETHER_HDR_LEN);
2060                 if (m == NULL)
2061                         return (0);
2062         }
2063
2064         eh = mtod(m, struct ether_header *);
2065         sc = ifp->if_bridge;
2066         vlan = VLANTAGOF(m);
2067
2068         bifp = sc->sc_ifp;
2069
2070         /*
2071          * If bridge is down, but the original output interface is up,
2072          * go ahead and send out that interface.  Otherwise, the packet
2073          * is dropped below.
2074          */
2075         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2076                 dst_if = ifp;
2077                 goto sendunicast;
2078         }
2079
2080         /*
2081          * If the packet is a multicast, or we don't know a better way to
2082          * get there, send to all interfaces.
2083          */
2084         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2085                 dst_if = NULL;
2086         else
2087                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2088         /* Tap any traffic not passing back out the originating interface */
2089         if (dst_if != ifp)
2090                 ETHER_BPF_MTAP(bifp, m);
2091         if (dst_if == NULL) {
2092                 struct bridge_iflist *bif;
2093                 struct mbuf *mc;
2094                 int used = 0;
2095
2096                 bridge_span(sc, m);
2097
2098                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2099                         dst_if = bif->bif_ifp;
2100
2101                         if (dst_if->if_type == IFT_GIF)
2102                                 continue;
2103                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2104                                 continue;
2105
2106                         /*
2107                          * If this is not the original output interface,
2108                          * and the interface is participating in spanning
2109                          * tree, make sure the port is in a state that
2110                          * allows forwarding.
2111                          */
2112                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2113                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2114                                 continue;
2115
2116                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2117                                 used = 1;
2118                                 mc = m;
2119                         } else {
2120                                 mc = m_copypacket(m, M_NOWAIT);
2121                                 if (mc == NULL) {
2122                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2123                                         continue;
2124                                 }
2125                         }
2126
2127                         bridge_enqueue(sc, dst_if, mc);
2128                 }
2129                 if (used == 0)
2130                         m_freem(m);
2131                 return (0);
2132         }
2133
2134 sendunicast:
2135         /*
2136          * XXX Spanning tree consideration here?
2137          */
2138
2139         bridge_span(sc, m);
2140         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2141                 m_freem(m);
2142                 return (0);
2143         }
2144
2145         bridge_enqueue(sc, dst_if, m);
2146         return (0);
2147 }
2148
2149 /*
2150  * bridge_transmit:
2151  *
2152  *      Do output on a bridge.
2153  *
2154  */
2155 static int
2156 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2157 {
2158         struct bridge_softc *sc;
2159         struct ether_header *eh;
2160         struct ifnet *dst_if;
2161         int error = 0;
2162
2163         sc = ifp->if_softc;
2164
2165         ETHER_BPF_MTAP(ifp, m);
2166
2167         eh = mtod(m, struct ether_header *);
2168
2169         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2170             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2171                 error = bridge_enqueue(sc, dst_if, m);
2172         } else
2173                 bridge_broadcast(sc, ifp, m, 0);
2174
2175         return (error);
2176 }
2177
2178 /*
2179  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2180  */
2181 static void
2182 bridge_qflush(struct ifnet *ifp __unused)
2183 {
2184 }
2185
2186 /*
2187  * bridge_forward:
2188  *
2189  *      The forwarding function of the bridge.
2190  *
2191  *      NOTE: Releases the lock on return.
2192  */
2193 static void
2194 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2195     struct mbuf *m)
2196 {
2197         struct bridge_iflist *dbif;
2198         struct ifnet *src_if, *dst_if, *ifp;
2199         struct ether_header *eh;
2200         uint16_t vlan;
2201         uint8_t *dst;
2202         int error;
2203
2204         NET_EPOCH_ASSERT();
2205
2206         src_if = m->m_pkthdr.rcvif;
2207         ifp = sc->sc_ifp;
2208
2209         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2210         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2211         vlan = VLANTAGOF(m);
2212
2213         if ((sbif->bif_flags & IFBIF_STP) &&
2214             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2215                 goto drop;
2216
2217         eh = mtod(m, struct ether_header *);
2218         dst = eh->ether_dhost;
2219
2220         /* If the interface is learning, record the address. */
2221         if (sbif->bif_flags & IFBIF_LEARNING) {
2222                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2223                     sbif, 0, IFBAF_DYNAMIC);
2224                 /*
2225                  * If the interface has addresses limits then deny any source
2226                  * that is not in the cache.
2227                  */
2228                 if (error && sbif->bif_addrmax)
2229                         goto drop;
2230         }
2231
2232         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2233             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2234                 goto drop;
2235
2236         /*
2237          * At this point, the port either doesn't participate
2238          * in spanning tree or it is in the forwarding state.
2239          */
2240
2241         /*
2242          * If the packet is unicast, destined for someone on
2243          * "this" side of the bridge, drop it.
2244          */
2245         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2246                 dst_if = bridge_rtlookup(sc, dst, vlan);
2247                 if (src_if == dst_if)
2248                         goto drop;
2249         } else {
2250                 /*
2251                  * Check if its a reserved multicast address, any address
2252                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2253                  * bridge.
2254                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2255                  */
2256                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2257                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2258                     dst[4] == 0x00 && dst[5] <= 0x0f)
2259                         goto drop;
2260
2261                 /* ...forward it to all interfaces. */
2262                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2263                 dst_if = NULL;
2264         }
2265
2266         /*
2267          * If we have a destination interface which is a member of our bridge,
2268          * OR this is a unicast packet, push it through the bpf(4) machinery.
2269          * For broadcast or multicast packets, don't bother because it will
2270          * be reinjected into ether_input. We do this before we pass the packets
2271          * through the pfil(9) framework, as it is possible that pfil(9) will
2272          * drop the packet, or possibly modify it, making it difficult to debug
2273          * firewall issues on the bridge.
2274          */
2275         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2276                 ETHER_BPF_MTAP(ifp, m);
2277
2278         /* run the packet filter */
2279         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2280 #ifdef INET6
2281             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2282 #endif
2283             ) {
2284                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2285                         return;
2286                 if (m == NULL)
2287                         return;
2288         }
2289
2290         if (dst_if == NULL) {
2291                 bridge_broadcast(sc, src_if, m, 1);
2292                 return;
2293         }
2294
2295         /*
2296          * At this point, we're dealing with a unicast frame
2297          * going to a different interface.
2298          */
2299         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2300                 goto drop;
2301
2302         dbif = bridge_lookup_member_if(sc, dst_if);
2303         if (dbif == NULL)
2304                 /* Not a member of the bridge (anymore?) */
2305                 goto drop;
2306
2307         /* Private segments can not talk to each other */
2308         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2309                 goto drop;
2310
2311         if ((dbif->bif_flags & IFBIF_STP) &&
2312             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2313                 goto drop;
2314
2315         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2316 #ifdef INET6
2317             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2318 #endif
2319             ) {
2320                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2321                         return;
2322                 if (m == NULL)
2323                         return;
2324         }
2325
2326         bridge_enqueue(sc, dst_if, m);
2327         return;
2328
2329 drop:
2330         m_freem(m);
2331 }
2332
2333 /*
2334  * bridge_input:
2335  *
2336  *      Receive input from a member interface.  Queue the packet for
2337  *      bridging if it is not for us.
2338  */
2339 static struct mbuf *
2340 bridge_input(struct ifnet *ifp, struct mbuf *m)
2341 {
2342         struct bridge_softc *sc = ifp->if_bridge;
2343         struct bridge_iflist *bif, *bif2;
2344         struct ifnet *bifp;
2345         struct ether_header *eh;
2346         struct mbuf *mc, *mc2;
2347         uint16_t vlan;
2348         int error;
2349
2350         NET_EPOCH_ASSERT();
2351
2352         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2353                 return (m);
2354
2355         bifp = sc->sc_ifp;
2356         vlan = VLANTAGOF(m);
2357
2358         /*
2359          * Implement support for bridge monitoring. If this flag has been
2360          * set on this interface, discard the packet once we push it through
2361          * the bpf(4) machinery, but before we do, increment the byte and
2362          * packet counters associated with this interface.
2363          */
2364         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2365                 m->m_pkthdr.rcvif  = bifp;
2366                 ETHER_BPF_MTAP(bifp, m);
2367                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2368                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2369                 m_freem(m);
2370                 return (NULL);
2371         }
2372         bif = bridge_lookup_member_if(sc, ifp);
2373         if (bif == NULL) {
2374                 return (m);
2375         }
2376
2377         eh = mtod(m, struct ether_header *);
2378
2379         bridge_span(sc, m);
2380
2381         if (m->m_flags & (M_BCAST|M_MCAST)) {
2382                 /* Tap off 802.1D packets; they do not get forwarded. */
2383                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2384                     ETHER_ADDR_LEN) == 0) {
2385                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2386                         return (NULL);
2387                 }
2388
2389                 if ((bif->bif_flags & IFBIF_STP) &&
2390                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2391                         return (m);
2392                 }
2393
2394                 /*
2395                  * Make a deep copy of the packet and enqueue the copy
2396                  * for bridge processing; return the original packet for
2397                  * local processing.
2398                  */
2399                 mc = m_dup(m, M_NOWAIT);
2400                 if (mc == NULL) {
2401                         return (m);
2402                 }
2403
2404                 /* Perform the bridge forwarding function with the copy. */
2405                 bridge_forward(sc, bif, mc);
2406
2407                 /*
2408                  * Reinject the mbuf as arriving on the bridge so we have a
2409                  * chance at claiming multicast packets. We can not loop back
2410                  * here from ether_input as a bridge is never a member of a
2411                  * bridge.
2412                  */
2413                 KASSERT(bifp->if_bridge == NULL,
2414                     ("loop created in bridge_input"));
2415                 mc2 = m_dup(m, M_NOWAIT);
2416                 if (mc2 != NULL) {
2417                         /* Keep the layer3 header aligned */
2418                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2419                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2420                 }
2421                 if (mc2 != NULL) {
2422                         mc2->m_pkthdr.rcvif = bifp;
2423                         (*bifp->if_input)(bifp, mc2);
2424                 }
2425
2426                 /* Return the original packet for local processing. */
2427                 return (m);
2428         }
2429
2430         if ((bif->bif_flags & IFBIF_STP) &&
2431             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2432                 return (m);
2433         }
2434
2435 #if (defined(INET) || defined(INET6))
2436 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2437         || ((iface)->if_carp \
2438             && (*carp_forus_p)((iface), eh->ether_dhost))
2439 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2440         || ((iface)->if_carp \
2441             && (*carp_forus_p)((iface), eh->ether_shost))
2442 #else
2443 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2444 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2445 #endif
2446
2447 #ifdef INET6
2448 #   define OR_PFIL_HOOKED_INET6 \
2449         || PFIL_HOOKED_IN(V_inet6_pfil_head)
2450 #else
2451 #   define OR_PFIL_HOOKED_INET6
2452 #endif
2453
2454 #define GRAB_OUR_PACKETS(iface) \
2455         if ((iface)->if_type == IFT_GIF) \
2456                 continue; \
2457         /* It is destined for us. */ \
2458         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2459             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2460             ) {                                                         \
2461                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2462                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2463                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2464                         if (error && bif->bif_addrmax) {                \
2465                                 m_freem(m);                             \
2466                                 return (NULL);                          \
2467                         }                                               \
2468                 }                                                       \
2469                 m->m_pkthdr.rcvif = iface;                              \
2470                 if ((iface) == ifp) {                                   \
2471                         /* Skip bridge processing... src == dest */     \
2472                         return (m);                                     \
2473                 }                                                       \
2474                 /* It's passing over or to the bridge, locally. */      \
2475                 ETHER_BPF_MTAP(bifp, m);                                \
2476                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2477                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); \
2478                 /* Filter on the physical interface. */                 \
2479                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) \
2480                      OR_PFIL_HOOKED_INET6)) {                           \
2481                         if (bridge_pfil(&m, NULL, ifp,                  \
2482                             PFIL_IN) != 0 || m == NULL) {               \
2483                                 return (NULL);                          \
2484                         }                                               \
2485                 }                                                       \
2486                 if ((iface) != bifp)                                    \
2487                         ETHER_BPF_MTAP(iface, m);                       \
2488                 return (m);                                             \
2489         }                                                               \
2490                                                                         \
2491         /* We just received a packet that we sent out. */               \
2492         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2493             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2494             ) {                                                         \
2495                 m_freem(m);                                             \
2496                 return (NULL);                                          \
2497         }
2498
2499         /*
2500          * Unicast.  Make sure it's not for the bridge.
2501          */
2502         do { GRAB_OUR_PACKETS(bifp) } while (0);
2503
2504         /*
2505          * Give a chance for ifp at first priority. This will help when the
2506          * packet comes through the interface like VLAN's with the same MACs
2507          * on several interfaces from the same bridge. This also will save
2508          * some CPU cycles in case the destination interface and the input
2509          * interface (eq ifp) are the same.
2510          */
2511         do { GRAB_OUR_PACKETS(ifp) } while (0);
2512
2513         /* Now check the all bridge members. */
2514         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2515                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2516         }
2517
2518 #undef OR_CARP_CHECK_WE_ARE_DST
2519 #undef OR_CARP_CHECK_WE_ARE_SRC
2520 #undef OR_PFIL_HOOKED_INET6
2521 #undef GRAB_OUR_PACKETS
2522
2523         /* Perform the bridge forwarding function. */
2524         bridge_forward(sc, bif, m);
2525
2526         return (NULL);
2527 }
2528
2529 /*
2530  * bridge_broadcast:
2531  *
2532  *      Send a frame to all interfaces that are members of
2533  *      the bridge, except for the one on which the packet
2534  *      arrived.
2535  *
2536  *      NOTE: Releases the lock on return.
2537  */
2538 static void
2539 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2540     struct mbuf *m, int runfilt)
2541 {
2542         struct bridge_iflist *dbif, *sbif;
2543         struct mbuf *mc;
2544         struct ifnet *dst_if;
2545         int used = 0, i;
2546
2547         NET_EPOCH_ASSERT();
2548
2549         sbif = bridge_lookup_member_if(sc, src_if);
2550
2551         /* Filter on the bridge interface before broadcasting */
2552         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2553 #ifdef INET6
2554             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2555 #endif
2556             )) {
2557                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2558                         return;
2559                 if (m == NULL)
2560                         return;
2561         }
2562
2563         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2564                 dst_if = dbif->bif_ifp;
2565                 if (dst_if == src_if)
2566                         continue;
2567
2568                 /* Private segments can not talk to each other */
2569                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2570                         continue;
2571
2572                 if ((dbif->bif_flags & IFBIF_STP) &&
2573                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2574                         continue;
2575
2576                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2577                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2578                         continue;
2579
2580                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2581                         continue;
2582
2583                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2584                         mc = m;
2585                         used = 1;
2586                 } else {
2587                         mc = m_dup(m, M_NOWAIT);
2588                         if (mc == NULL) {
2589                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2590                                 continue;
2591                         }
2592                 }
2593
2594                 /*
2595                  * Filter on the output interface. Pass a NULL bridge interface
2596                  * pointer so we do not redundantly filter on the bridge for
2597                  * each interface we broadcast on.
2598                  */
2599                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2600 #ifdef INET6
2601                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2602 #endif
2603                     )) {
2604                         if (used == 0) {
2605                                 /* Keep the layer3 header aligned */
2606                                 i = min(mc->m_pkthdr.len, max_protohdr);
2607                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2608                                 if (mc == NULL) {
2609                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2610                                         continue;
2611                                 }
2612                         }
2613                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2614                                 continue;
2615                         if (mc == NULL)
2616                                 continue;
2617                 }
2618
2619                 bridge_enqueue(sc, dst_if, mc);
2620         }
2621         if (used == 0)
2622                 m_freem(m);
2623 }
2624
2625 /*
2626  * bridge_span:
2627  *
2628  *      Duplicate a packet out one or more interfaces that are in span mode,
2629  *      the original mbuf is unmodified.
2630  */
2631 static void
2632 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2633 {
2634         struct bridge_iflist *bif;
2635         struct ifnet *dst_if;
2636         struct mbuf *mc;
2637
2638         NET_EPOCH_ASSERT();
2639
2640         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2641                 return;
2642
2643         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2644                 dst_if = bif->bif_ifp;
2645
2646                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2647                         continue;
2648
2649                 mc = m_copypacket(m, M_NOWAIT);
2650                 if (mc == NULL) {
2651                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2652                         continue;
2653                 }
2654
2655                 bridge_enqueue(sc, dst_if, mc);
2656         }
2657 }
2658
2659 /*
2660  * bridge_rtupdate:
2661  *
2662  *      Add a bridge routing entry.
2663  */
2664 static int
2665 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2666     struct bridge_iflist *bif, int setflags, uint8_t flags)
2667 {
2668         struct bridge_rtnode *brt;
2669         int error;
2670
2671         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2672
2673         /* Check the source address is valid and not multicast. */
2674         if (ETHER_IS_MULTICAST(dst) ||
2675             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2676              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2677                 return (EINVAL);
2678
2679         /* 802.1p frames map to vlan 1 */
2680         if (vlan == 0)
2681                 vlan = 1;
2682
2683         /*
2684          * A route for this destination might already exist.  If so,
2685          * update it, otherwise create a new one.
2686          */
2687         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2688                 BRIDGE_RT_LOCK(sc);
2689
2690                 /* Check again, now that we have the lock. There could have
2691                  * been a race and we only want to insert this once. */
2692                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2693                         BRIDGE_RT_UNLOCK(sc);
2694                         return (0);
2695                 }
2696
2697                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2698                         sc->sc_brtexceeded++;
2699                         BRIDGE_RT_UNLOCK(sc);
2700                         return (ENOSPC);
2701                 }
2702                 /* Check per interface address limits (if enabled) */
2703                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2704                         bif->bif_addrexceeded++;
2705                         BRIDGE_RT_UNLOCK(sc);
2706                         return (ENOSPC);
2707                 }
2708
2709                 /*
2710                  * Allocate a new bridge forwarding node, and
2711                  * initialize the expiration time and Ethernet
2712                  * address.
2713                  */
2714                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2715                 if (brt == NULL) {
2716                         BRIDGE_RT_UNLOCK(sc);
2717                         return (ENOMEM);
2718                 }
2719                 brt->brt_vnet = curvnet;
2720
2721                 if (bif->bif_flags & IFBIF_STICKY)
2722                         brt->brt_flags = IFBAF_STICKY;
2723                 else
2724                         brt->brt_flags = IFBAF_DYNAMIC;
2725
2726                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2727                 brt->brt_vlan = vlan;
2728
2729                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2730                         uma_zfree(V_bridge_rtnode_zone, brt);
2731                         BRIDGE_RT_UNLOCK(sc);
2732                         return (error);
2733                 }
2734                 brt->brt_dst = bif;
2735                 bif->bif_addrcnt++;
2736
2737                 BRIDGE_RT_UNLOCK(sc);
2738         }
2739
2740         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2741             brt->brt_dst != bif) {
2742                 BRIDGE_RT_LOCK(sc);
2743                 brt->brt_dst->bif_addrcnt--;
2744                 brt->brt_dst = bif;
2745                 brt->brt_dst->bif_addrcnt++;
2746                 BRIDGE_RT_UNLOCK(sc);
2747         }
2748
2749         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2750                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2751         if (setflags)
2752                 brt->brt_flags = flags;
2753
2754         return (0);
2755 }
2756
2757 /*
2758  * bridge_rtlookup:
2759  *
2760  *      Lookup the destination interface for an address.
2761  */
2762 static struct ifnet *
2763 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2764 {
2765         struct bridge_rtnode *brt;
2766
2767         NET_EPOCH_ASSERT();
2768
2769         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2770                 return (NULL);
2771
2772         return (brt->brt_ifp);
2773 }
2774
2775 /*
2776  * bridge_rttrim:
2777  *
2778  *      Trim the routine table so that we have a number
2779  *      of routing entries less than or equal to the
2780  *      maximum number.
2781  */
2782 static void
2783 bridge_rttrim(struct bridge_softc *sc)
2784 {
2785         struct bridge_rtnode *brt, *nbrt;
2786
2787         NET_EPOCH_ASSERT();
2788         BRIDGE_RT_LOCK_ASSERT(sc);
2789
2790         /* Make sure we actually need to do this. */
2791         if (sc->sc_brtcnt <= sc->sc_brtmax)
2792                 return;
2793
2794         /* Force an aging cycle; this might trim enough addresses. */
2795         bridge_rtage(sc);
2796         if (sc->sc_brtcnt <= sc->sc_brtmax)
2797                 return;
2798
2799         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2800                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2801                         bridge_rtnode_destroy(sc, brt);
2802                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2803                                 return;
2804                 }
2805         }
2806 }
2807
2808 /*
2809  * bridge_timer:
2810  *
2811  *      Aging timer for the bridge.
2812  */
2813 static void
2814 bridge_timer(void *arg)
2815 {
2816         struct bridge_softc *sc = arg;
2817
2818         BRIDGE_RT_LOCK_ASSERT(sc);
2819
2820         /* Destruction of rtnodes requires a proper vnet context */
2821         CURVNET_SET(sc->sc_ifp->if_vnet);
2822         bridge_rtage(sc);
2823
2824         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2825                 callout_reset(&sc->sc_brcallout,
2826                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2827         CURVNET_RESTORE();
2828 }
2829
2830 /*
2831  * bridge_rtage:
2832  *
2833  *      Perform an aging cycle.
2834  */
2835 static void
2836 bridge_rtage(struct bridge_softc *sc)
2837 {
2838         struct bridge_rtnode *brt, *nbrt;
2839
2840         BRIDGE_RT_LOCK_ASSERT(sc);
2841
2842         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2843                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2844                         if (time_uptime >= brt->brt_expire)
2845                                 bridge_rtnode_destroy(sc, brt);
2846                 }
2847         }
2848 }
2849
2850 /*
2851  * bridge_rtflush:
2852  *
2853  *      Remove all dynamic addresses from the bridge.
2854  */
2855 static void
2856 bridge_rtflush(struct bridge_softc *sc, int full)
2857 {
2858         struct bridge_rtnode *brt, *nbrt;
2859
2860         BRIDGE_RT_LOCK_ASSERT(sc);
2861
2862         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2863                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2864                         bridge_rtnode_destroy(sc, brt);
2865         }
2866 }
2867
2868 /*
2869  * bridge_rtdaddr:
2870  *
2871  *      Remove an address from the table.
2872  */
2873 static int
2874 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2875 {
2876         struct bridge_rtnode *brt;
2877         int found = 0;
2878
2879         BRIDGE_RT_LOCK(sc);
2880
2881         /*
2882          * If vlan is zero then we want to delete for all vlans so the lookup
2883          * may return more than one.
2884          */
2885         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2886                 bridge_rtnode_destroy(sc, brt);
2887                 found = 1;
2888         }
2889
2890         BRIDGE_RT_UNLOCK(sc);
2891
2892         return (found ? 0 : ENOENT);
2893 }
2894
2895 /*
2896  * bridge_rtdelete:
2897  *
2898  *      Delete routes to a speicifc member interface.
2899  */
2900 static void
2901 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2902 {
2903         struct bridge_rtnode *brt, *nbrt;
2904
2905         BRIDGE_RT_LOCK_ASSERT(sc);
2906
2907         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2908                 if (brt->brt_ifp == ifp && (full ||
2909                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2910                         bridge_rtnode_destroy(sc, brt);
2911         }
2912 }
2913
2914 /*
2915  * bridge_rtable_init:
2916  *
2917  *      Initialize the route table for this bridge.
2918  */
2919 static void
2920 bridge_rtable_init(struct bridge_softc *sc)
2921 {
2922         int i;
2923
2924         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2925             M_DEVBUF, M_WAITOK);
2926
2927         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2928                 CK_LIST_INIT(&sc->sc_rthash[i]);
2929
2930         sc->sc_rthash_key = arc4random();
2931         CK_LIST_INIT(&sc->sc_rtlist);
2932 }
2933
2934 /*
2935  * bridge_rtable_fini:
2936  *
2937  *      Deconstruct the route table for this bridge.
2938  */
2939 static void
2940 bridge_rtable_fini(struct bridge_softc *sc)
2941 {
2942
2943         KASSERT(sc->sc_brtcnt == 0,
2944             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2945         free(sc->sc_rthash, M_DEVBUF);
2946 }
2947
2948 /*
2949  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2950  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2951  */
2952 #define mix(a, b, c)                                                    \
2953 do {                                                                    \
2954         a -= b; a -= c; a ^= (c >> 13);                                 \
2955         b -= c; b -= a; b ^= (a << 8);                                  \
2956         c -= a; c -= b; c ^= (b >> 13);                                 \
2957         a -= b; a -= c; a ^= (c >> 12);                                 \
2958         b -= c; b -= a; b ^= (a << 16);                                 \
2959         c -= a; c -= b; c ^= (b >> 5);                                  \
2960         a -= b; a -= c; a ^= (c >> 3);                                  \
2961         b -= c; b -= a; b ^= (a << 10);                                 \
2962         c -= a; c -= b; c ^= (b >> 15);                                 \
2963 } while (/*CONSTCOND*/0)
2964
2965 static __inline uint32_t
2966 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2967 {
2968         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2969
2970         b += addr[5] << 8;
2971         b += addr[4];
2972         a += addr[3] << 24;
2973         a += addr[2] << 16;
2974         a += addr[1] << 8;
2975         a += addr[0];
2976
2977         mix(a, b, c);
2978
2979         return (c & BRIDGE_RTHASH_MASK);
2980 }
2981
2982 #undef mix
2983
2984 static int
2985 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2986 {
2987         int i, d;
2988
2989         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2990                 d = ((int)a[i]) - ((int)b[i]);
2991         }
2992
2993         return (d);
2994 }
2995
2996 /*
2997  * bridge_rtnode_lookup:
2998  *
2999  *      Look up a bridge route node for the specified destination. Compare the
3000  *      vlan id or if zero then just return the first match.
3001  */
3002 static struct bridge_rtnode *
3003 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3004 {
3005         struct bridge_rtnode *brt;
3006         uint32_t hash;
3007         int dir;
3008
3009         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3010
3011         hash = bridge_rthash(sc, addr);
3012         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3013                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3014                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3015                         return (brt);
3016                 if (dir > 0)
3017                         return (NULL);
3018         }
3019
3020         return (NULL);
3021 }
3022
3023 /*
3024  * bridge_rtnode_insert:
3025  *
3026  *      Insert the specified bridge node into the route table.  We
3027  *      assume the entry is not already in the table.
3028  */
3029 static int
3030 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3031 {
3032         struct bridge_rtnode *lbrt;
3033         uint32_t hash;
3034         int dir;
3035
3036         BRIDGE_RT_LOCK_ASSERT(sc);
3037
3038         hash = bridge_rthash(sc, brt->brt_addr);
3039
3040         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3041         if (lbrt == NULL) {
3042                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3043                 goto out;
3044         }
3045
3046         do {
3047                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3048                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3049                         return (EEXIST);
3050                 if (dir > 0) {
3051                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3052                         goto out;
3053                 }
3054                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3055                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3056                         goto out;
3057                 }
3058                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3059         } while (lbrt != NULL);
3060
3061 #ifdef DIAGNOSTIC
3062         panic("bridge_rtnode_insert: impossible");
3063 #endif
3064
3065 out:
3066         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3067         sc->sc_brtcnt++;
3068
3069         return (0);
3070 }
3071
3072 static void
3073 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3074 {
3075         struct bridge_rtnode *brt;
3076
3077         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3078
3079         CURVNET_SET(brt->brt_vnet);
3080         uma_zfree(V_bridge_rtnode_zone, brt);
3081         CURVNET_RESTORE();
3082 }
3083
3084 /*
3085  * bridge_rtnode_destroy:
3086  *
3087  *      Destroy a bridge rtnode.
3088  */
3089 static void
3090 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3091 {
3092         BRIDGE_RT_LOCK_ASSERT(sc);
3093
3094         CK_LIST_REMOVE(brt, brt_hash);
3095
3096         CK_LIST_REMOVE(brt, brt_list);
3097         sc->sc_brtcnt--;
3098         brt->brt_dst->bif_addrcnt--;
3099
3100         NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3101 }
3102
3103 /*
3104  * bridge_rtable_expire:
3105  *
3106  *      Set the expiry time for all routes on an interface.
3107  */
3108 static void
3109 bridge_rtable_expire(struct ifnet *ifp, int age)
3110 {
3111         struct bridge_softc *sc = ifp->if_bridge;
3112         struct bridge_rtnode *brt;
3113
3114         CURVNET_SET(ifp->if_vnet);
3115         BRIDGE_RT_LOCK(sc);
3116
3117         /*
3118          * If the age is zero then flush, otherwise set all the expiry times to
3119          * age for the interface
3120          */
3121         if (age == 0)
3122                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3123         else {
3124                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3125                         /* Cap the expiry time to 'age' */
3126                         if (brt->brt_ifp == ifp &&
3127                             brt->brt_expire > time_uptime + age &&
3128                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3129                                 brt->brt_expire = time_uptime + age;
3130                 }
3131         }
3132         BRIDGE_RT_UNLOCK(sc);
3133         CURVNET_RESTORE();
3134 }
3135
3136 /*
3137  * bridge_state_change:
3138  *
3139  *      Callback from the bridgestp code when a port changes states.
3140  */
3141 static void
3142 bridge_state_change(struct ifnet *ifp, int state)
3143 {
3144         struct bridge_softc *sc = ifp->if_bridge;
3145         static const char *stpstates[] = {
3146                 "disabled",
3147                 "listening",
3148                 "learning",
3149                 "forwarding",
3150                 "blocking",
3151                 "discarding"
3152         };
3153
3154         CURVNET_SET(ifp->if_vnet);
3155         if (V_log_stp)
3156                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3157                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3158         CURVNET_RESTORE();
3159 }
3160
3161 /*
3162  * Send bridge packets through pfil if they are one of the types pfil can deal
3163  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3164  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3165  * that interface.
3166  */
3167 static int
3168 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3169 {
3170         int snap, error, i, hlen;
3171         struct ether_header *eh1, eh2;
3172         struct ip *ip;
3173         struct llc llc1;
3174         u_int16_t ether_type;
3175         pfil_return_t rv;
3176
3177         snap = 0;
3178         error = -1;     /* Default error if not error == 0 */
3179
3180 #if 0
3181         /* we may return with the IP fields swapped, ensure its not shared */
3182         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3183 #endif
3184
3185         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3186                 return (0); /* filtering is disabled */
3187
3188         i = min((*mp)->m_pkthdr.len, max_protohdr);
3189         if ((*mp)->m_len < i) {
3190             *mp = m_pullup(*mp, i);
3191             if (*mp == NULL) {
3192                 printf("%s: m_pullup failed\n", __func__);
3193                 return (-1);
3194             }
3195         }
3196
3197         eh1 = mtod(*mp, struct ether_header *);
3198         ether_type = ntohs(eh1->ether_type);
3199
3200         /*
3201          * Check for SNAP/LLC.
3202          */
3203         if (ether_type < ETHERMTU) {
3204                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3205
3206                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3207                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3208                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3209                     llc2->llc_control == LLC_UI) {
3210                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3211                         snap = 1;
3212                 }
3213         }
3214
3215         /*
3216          * If we're trying to filter bridge traffic, don't look at anything
3217          * other than IP and ARP traffic.  If the filter doesn't understand
3218          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3219          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3220          * but of course we don't have an AppleTalk filter to begin with.
3221          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3222          * ARP traffic.)
3223          */
3224         switch (ether_type) {
3225                 case ETHERTYPE_ARP:
3226                 case ETHERTYPE_REVARP:
3227                         if (V_pfil_ipfw_arp == 0)
3228                                 return (0); /* Automatically pass */
3229                         break;
3230
3231                 case ETHERTYPE_IP:
3232 #ifdef INET6
3233                 case ETHERTYPE_IPV6:
3234 #endif /* INET6 */
3235                         break;
3236                 default:
3237                         /*
3238                          * Check to see if the user wants to pass non-ip
3239                          * packets, these will not be checked by pfil(9) and
3240                          * passed unconditionally so the default is to drop.
3241                          */
3242                         if (V_pfil_onlyip)
3243                                 goto bad;
3244         }
3245
3246         /* Run the packet through pfil before stripping link headers */
3247         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3248             dir == PFIL_OUT && ifp != NULL) {
3249                 switch (pfil_run_hooks(V_link_pfil_head, mp, ifp, dir, NULL)) {
3250                 case PFIL_DROPPED:
3251                         return (EACCES);
3252                 case PFIL_CONSUMED:
3253                         return (0);
3254                 }
3255         }
3256
3257         /* Strip off the Ethernet header and keep a copy. */
3258         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3259         m_adj(*mp, ETHER_HDR_LEN);
3260
3261         /* Strip off snap header, if present */
3262         if (snap) {
3263                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3264                 m_adj(*mp, sizeof(struct llc));
3265         }
3266
3267         /*
3268          * Check the IP header for alignment and errors
3269          */
3270         if (dir == PFIL_IN) {
3271                 switch (ether_type) {
3272                         case ETHERTYPE_IP:
3273                                 error = bridge_ip_checkbasic(mp);
3274                                 break;
3275 #ifdef INET6
3276                         case ETHERTYPE_IPV6:
3277                                 error = bridge_ip6_checkbasic(mp);
3278                                 break;
3279 #endif /* INET6 */
3280                         default:
3281                                 error = 0;
3282                 }
3283                 if (error)
3284                         goto bad;
3285         }
3286
3287         error = 0;
3288
3289         /*
3290          * Run the packet through pfil
3291          */
3292         rv = PFIL_PASS;
3293         switch (ether_type) {
3294         case ETHERTYPE_IP:
3295                 /*
3296                  * Run pfil on the member interface and the bridge, both can
3297                  * be skipped by clearing pfil_member or pfil_bridge.
3298                  *
3299                  * Keep the order:
3300                  *   in_if -> bridge_if -> out_if
3301                  */
3302                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3303                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3304                     PFIL_PASS)
3305                         break;
3306
3307                 if (V_pfil_member && ifp != NULL && (rv =
3308                     pfil_run_hooks(V_inet_pfil_head, mp, ifp, dir, NULL)) !=
3309                     PFIL_PASS)
3310                         break;
3311
3312                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3313                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3314                     PFIL_PASS)
3315                         break;
3316
3317                 /* check if we need to fragment the packet */
3318                 /* bridge_fragment generates a mbuf chain of packets */
3319                 /* that already include eth headers */
3320                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3321                         i = (*mp)->m_pkthdr.len;
3322                         if (i > ifp->if_mtu) {
3323                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3324                                             &llc1);
3325                                 return (error);
3326                         }
3327                 }
3328
3329                 /* Recalculate the ip checksum. */
3330                 ip = mtod(*mp, struct ip *);
3331                 hlen = ip->ip_hl << 2;
3332                 if (hlen < sizeof(struct ip))
3333                         goto bad;
3334                 if (hlen > (*mp)->m_len) {
3335                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3336                                 goto bad;
3337                         ip = mtod(*mp, struct ip *);
3338                         if (ip == NULL)
3339                                 goto bad;
3340                 }
3341                 ip->ip_sum = 0;
3342                 if (hlen == sizeof(struct ip))
3343                         ip->ip_sum = in_cksum_hdr(ip);
3344                 else
3345                         ip->ip_sum = in_cksum(*mp, hlen);
3346
3347                 break;
3348 #ifdef INET6
3349         case ETHERTYPE_IPV6:
3350                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3351                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3352                     PFIL_PASS)
3353                         break;
3354
3355                 if (V_pfil_member && ifp != NULL && (rv =
3356                     pfil_run_hooks(V_inet6_pfil_head, mp, ifp, dir, NULL)) !=
3357                     PFIL_PASS)
3358                         break;
3359
3360                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3361                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3362                     PFIL_PASS)
3363                         break;
3364                 break;
3365 #endif
3366         }
3367
3368         switch (rv) {
3369         case PFIL_CONSUMED:
3370                 return (0);
3371         case PFIL_DROPPED:
3372                 return (EACCES);
3373         default:
3374                 break;
3375         }
3376
3377         error = -1;
3378
3379         /*
3380          * Finally, put everything back the way it was and return
3381          */
3382         if (snap) {
3383                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3384                 if (*mp == NULL)
3385                         return (error);
3386                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3387         }
3388
3389         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3390         if (*mp == NULL)
3391                 return (error);
3392         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3393
3394         return (0);
3395
3396 bad:
3397         m_freem(*mp);
3398         *mp = NULL;
3399         return (error);
3400 }
3401
3402 /*
3403  * Perform basic checks on header size since
3404  * pfil assumes ip_input has already processed
3405  * it for it.  Cut-and-pasted from ip_input.c.
3406  * Given how simple the IPv6 version is,
3407  * does the IPv4 version really need to be
3408  * this complicated?
3409  *
3410  * XXX Should we update ipstat here, or not?
3411  * XXX Right now we update ipstat but not
3412  * XXX csum_counter.
3413  */
3414 static int
3415 bridge_ip_checkbasic(struct mbuf **mp)
3416 {
3417         struct mbuf *m = *mp;
3418         struct ip *ip;
3419         int len, hlen;
3420         u_short sum;
3421
3422         if (*mp == NULL)
3423                 return (-1);
3424
3425         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3426                 if ((m = m_copyup(m, sizeof(struct ip),
3427                         (max_linkhdr + 3) & ~3)) == NULL) {
3428                         /* XXXJRT new stat, please */
3429                         KMOD_IPSTAT_INC(ips_toosmall);
3430                         goto bad;
3431                 }
3432         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3433                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3434                         KMOD_IPSTAT_INC(ips_toosmall);
3435                         goto bad;
3436                 }
3437         }
3438         ip = mtod(m, struct ip *);
3439         if (ip == NULL) goto bad;
3440
3441         if (ip->ip_v != IPVERSION) {
3442                 KMOD_IPSTAT_INC(ips_badvers);
3443                 goto bad;
3444         }
3445         hlen = ip->ip_hl << 2;
3446         if (hlen < sizeof(struct ip)) { /* minimum header length */
3447                 KMOD_IPSTAT_INC(ips_badhlen);
3448                 goto bad;
3449         }
3450         if (hlen > m->m_len) {
3451                 if ((m = m_pullup(m, hlen)) == NULL) {
3452                         KMOD_IPSTAT_INC(ips_badhlen);
3453                         goto bad;
3454                 }
3455                 ip = mtod(m, struct ip *);
3456                 if (ip == NULL) goto bad;
3457         }
3458
3459         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3460                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3461         } else {
3462                 if (hlen == sizeof(struct ip)) {
3463                         sum = in_cksum_hdr(ip);
3464                 } else {
3465                         sum = in_cksum(m, hlen);
3466                 }
3467         }
3468         if (sum) {
3469                 KMOD_IPSTAT_INC(ips_badsum);
3470                 goto bad;
3471         }
3472
3473         /* Retrieve the packet length. */
3474         len = ntohs(ip->ip_len);
3475
3476         /*
3477          * Check for additional length bogosity
3478          */
3479         if (len < hlen) {
3480                 KMOD_IPSTAT_INC(ips_badlen);
3481                 goto bad;
3482         }
3483
3484         /*
3485          * Check that the amount of data in the buffers
3486          * is as at least much as the IP header would have us expect.
3487          * Drop packet if shorter than we expect.
3488          */
3489         if (m->m_pkthdr.len < len) {
3490                 KMOD_IPSTAT_INC(ips_tooshort);
3491                 goto bad;
3492         }
3493
3494         /* Checks out, proceed */
3495         *mp = m;
3496         return (0);
3497
3498 bad:
3499         *mp = m;
3500         return (-1);
3501 }
3502
3503 #ifdef INET6
3504 /*
3505  * Same as above, but for IPv6.
3506  * Cut-and-pasted from ip6_input.c.
3507  * XXX Should we update ip6stat, or not?
3508  */
3509 static int
3510 bridge_ip6_checkbasic(struct mbuf **mp)
3511 {
3512         struct mbuf *m = *mp;
3513         struct ip6_hdr *ip6;
3514
3515         /*
3516          * If the IPv6 header is not aligned, slurp it up into a new
3517          * mbuf with space for link headers, in the event we forward
3518          * it.  Otherwise, if it is aligned, make sure the entire base
3519          * IPv6 header is in the first mbuf of the chain.
3520          */
3521         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3522                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3523                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3524                             (max_linkhdr + 3) & ~3)) == NULL) {
3525                         /* XXXJRT new stat, please */
3526                         IP6STAT_INC(ip6s_toosmall);
3527                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3528                         goto bad;
3529                 }
3530         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3531                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3532                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3533                         IP6STAT_INC(ip6s_toosmall);
3534                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3535                         goto bad;
3536                 }
3537         }
3538
3539         ip6 = mtod(m, struct ip6_hdr *);
3540
3541         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3542                 IP6STAT_INC(ip6s_badvers);
3543                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3544                 goto bad;
3545         }
3546
3547         /* Checks out, proceed */
3548         *mp = m;
3549         return (0);
3550
3551 bad:
3552         *mp = m;
3553         return (-1);
3554 }
3555 #endif /* INET6 */
3556
3557 /*
3558  * bridge_fragment:
3559  *
3560  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3561  */
3562 static int
3563 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3564     int snap, struct llc *llc)
3565 {
3566         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3567         struct ip *ip;
3568         int error = -1;
3569
3570         if (m->m_len < sizeof(struct ip) &&
3571             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3572                 goto dropit;
3573         ip = mtod(m, struct ip *);
3574
3575         m->m_pkthdr.csum_flags |= CSUM_IP;
3576         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3577         if (error)
3578                 goto dropit;
3579
3580         /*
3581          * Walk the chain and re-add the Ethernet header for
3582          * each mbuf packet.
3583          */
3584         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3585                 nextpkt = mcur->m_nextpkt;
3586                 mcur->m_nextpkt = NULL;
3587                 if (snap) {
3588                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3589                         if (mcur == NULL) {
3590                                 error = ENOBUFS;
3591                                 if (mprev != NULL)
3592                                         mprev->m_nextpkt = nextpkt;
3593                                 goto dropit;
3594                         }
3595                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3596                 }
3597
3598                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3599                 if (mcur == NULL) {
3600                         error = ENOBUFS;
3601                         if (mprev != NULL)
3602                                 mprev->m_nextpkt = nextpkt;
3603                         goto dropit;
3604                 }
3605                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3606
3607                 /*
3608                  * The previous two M_PREPEND could have inserted one or two
3609                  * mbufs in front so we have to update the previous packet's
3610                  * m_nextpkt.
3611                  */
3612                 mcur->m_nextpkt = nextpkt;
3613                 if (mprev != NULL)
3614                         mprev->m_nextpkt = mcur;
3615                 else {
3616                         /* The first mbuf in the original chain needs to be
3617                          * updated. */
3618                         *mp = mcur;
3619                 }
3620                 mprev = mcur;
3621         }
3622
3623         KMOD_IPSTAT_INC(ips_fragmented);
3624         return (error);
3625
3626 dropit:
3627         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3628                 m = mcur->m_nextpkt;
3629                 m_freem(mcur);
3630         }
3631         return (error);
3632 }
3633
3634 static void
3635 bridge_linkstate(struct ifnet *ifp)
3636 {
3637         struct bridge_softc *sc = ifp->if_bridge;
3638         struct bridge_iflist *bif;
3639         struct epoch_tracker et;
3640
3641         NET_EPOCH_ENTER(et);
3642
3643         bif = bridge_lookup_member_if(sc, ifp);
3644         if (bif == NULL) {
3645                 NET_EPOCH_EXIT(et);
3646                 return;
3647         }
3648         bridge_linkcheck(sc);
3649
3650         bstp_linkstate(&bif->bif_stp);
3651
3652         NET_EPOCH_EXIT(et);
3653 }
3654
3655 static void
3656 bridge_linkcheck(struct bridge_softc *sc)
3657 {
3658         struct bridge_iflist *bif;
3659         int new_link, hasls;
3660
3661         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3662
3663         new_link = LINK_STATE_DOWN;
3664         hasls = 0;
3665         /* Our link is considered up if at least one of our ports is active */
3666         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3667                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3668                         hasls++;
3669                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3670                         new_link = LINK_STATE_UP;
3671                         break;
3672                 }
3673         }
3674         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3675                 /* If no interfaces support link-state then we default to up */
3676                 new_link = LINK_STATE_UP;
3677         }
3678         if_link_state_change(sc->sc_ifp, new_link);
3679 }