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 #include "opt_inet.h"
  81 #include "opt_inet6.h"
  82
  83 #include <sys/param.h>
  84 #include <sys/eventhandler.h>
  85 #include <sys/mbuf.h>
  86 #include <sys/malloc.h>
  87 #include <sys/protosw.h>
  88 #include <sys/systm.h>
  89 #include <sys/jail.h>
  90 #include <sys/time.h>
  91 #include <sys/socket.h> /* for net/if.h */
  92 #include <sys/sockio.h>
  93 #include <sys/ctype.h>  /* string functions */
  94 #include <sys/kernel.h>
  95 #include <sys/random.h>
  96 #include <sys/syslog.h>
  97 #include <sys/sysctl.h>
  98 #include <vm/uma.h>
  99 #include <sys/module.h>
 100 #include <sys/priv.h>
 101 #include <sys/proc.h>
 102 #include <sys/lock.h>
 103 #include <sys/mutex.h>
 104
 105 #include <net/bpf.h>
 106 #include <net/if.h>
 107 #include <net/if_clone.h>
 108 #include <net/if_dl.h>
 109 #include <net/if_types.h>
 110 #include <net/if_var.h>
 111 #include <net/pfil.h>
 112 #include <net/vnet.h>
 113
 114 #include <netinet/in.h>
 115 #include <netinet/in_systm.h>
 116 #include <netinet/in_var.h>
 117 #include <netinet/ip.h>
 118 #include <netinet/ip_var.h>
 119 #ifdef INET6
 120 #include <netinet/ip6.h>
 121 #include <netinet6/ip6_var.h>
 122 #include <netinet6/in6_ifattach.h>
 123 #endif
 124 #if defined(INET) || defined(INET6)
 125 #include <netinet/ip_carp.h>
 126 #endif
 127 #include <machine/in_cksum.h>
 128 #include <netinet/if_ether.h>
 129 #include <net/bridgestp.h>
 130 #include <net/if_bridgevar.h>
 131 #include <net/if_llc.h>
 132 #include <net/if_vlan_var.h>
 133
 134 #include <net/route.h>
 135
 136 #ifdef INET6
 137 /*
 138  * XXX: declare here to avoid to include many inet6 related files..
 139  * should be more generalized?
 140  */
 141 extern void     nd6_setmtu(struct ifnet *);
 142 #endif
 143
 144 /*
 145  * Size of the route hash table.  Must be a power of two.
 146  */
 147 #ifndef BRIDGE_RTHASH_SIZE
 148 #define BRIDGE_RTHASH_SIZE              1024
 149 #endif
 150
 151 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 152
 153 /*
 154  * Default maximum number of addresses to cache.
 155  */
 156 #ifndef BRIDGE_RTABLE_MAX
 157 #define BRIDGE_RTABLE_MAX               2000
 158 #endif
 159
 160 /*
 161  * Timeout (in seconds) for entries learned dynamically.
 162  */
 163 #ifndef BRIDGE_RTABLE_TIMEOUT
 164 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 165 #endif
 166
 167 /*
 168  * Number of seconds between walks of the route list.
 169  */
 170 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 171 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 172 #endif
 173
 174 /*
 175  * List of capabilities to possibly mask on the member interface.
 176  */
 177 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 178                                          IFCAP_TXCSUM_IPV6)
 179
 180 /*
 181  * List of capabilities to strip
 182  */
 183 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 184
 185 /*
 186  * Bridge locking
 187  *
 188  * The bridge relies heavily on the epoch(9) system to protect its data
 189  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 190  * must ensure there is only one writer at a time.
 191  *
 192  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 193  * accesses we must hold:
 194  *
 195  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 196  *  - BRIDGE_LOCK, for any other change
 197  *
 198  * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
 199  * calls to bridge member interfaces and these ioctl()s can sleep.
 200  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 201  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 202  */
 203 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 204         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 205         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 206 } while (0)
 207 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 208         sx_destroy(&(_sc)->sc_sx);              \
 209         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 210 } while (0)
 211 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 212 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 213 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 214 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 215             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 216 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 217 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 218 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 219 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 220 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 221             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 222
 223 /*
 224  * Bridge interface list entry.
 225  */
 226 struct bridge_iflist {
 227         CK_LIST_ENTRY(bridge_iflist) bif_next;
 228         struct ifnet            *bif_ifp;       /* member if */
 229         struct bstp_port        bif_stp;        /* STP state */
 230         uint32_t                bif_flags;      /* member if flags */
 231         int                     bif_savedcaps;  /* saved capabilities */
 232         uint32_t                bif_addrmax;    /* max # of addresses */
 233         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 234         uint32_t                bif_addrexceeded;/* # of address violations */
 235         struct epoch_context    bif_epoch_ctx;
 236 };
 237
 238 /*
 239  * Bridge route node.
 240  */
 241 struct bridge_rtnode {
 242         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 243         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 244         struct bridge_iflist    *brt_dst;       /* destination if */
 245         unsigned long           brt_expire;     /* expiration time */
 246         uint8_t                 brt_flags;      /* address flags */
 247         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 248         uint16_t                brt_vlan;       /* vlan id */
 249         struct  vnet            *brt_vnet;
 250         struct  epoch_context   brt_epoch_ctx;
 251 };
 252 #define brt_ifp                 brt_dst->bif_ifp
 253
 254 /*
 255  * Software state for each bridge.
 256  */
 257 struct bridge_softc {
 258         struct ifnet            *sc_ifp;        /* make this an interface */
 259         LIST_ENTRY(bridge_softc) sc_list;
 260         struct sx               sc_sx;
 261         struct mtx              sc_rt_mtx;
 262         uint32_t                sc_brtmax;      /* max # of addresses */
 263         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 264         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 265         struct callout          sc_brcallout;   /* bridge callout */
 266         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 267         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 268         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 269         uint32_t                sc_rthash_key;  /* key for hash */
 270         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 271         struct bstp_state       sc_stp;         /* STP state */
 272         uint32_t                sc_brtexceeded; /* # of cache drops */
 273         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 274         struct ether_addr       sc_defaddr;     /* Default MAC address */
 275         struct epoch_context    sc_epoch_ctx;
 276 };
 277
 278 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 279 #define V_bridge_list_sx        VNET(bridge_list_sx)
 280 static eventhandler_tag bridge_detach_cookie;
 281
 282 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 283
 284 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 285 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 286
 287 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 288 static void     bridge_clone_destroy(struct ifnet *);
 289
 290 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 291 static void     bridge_mutecaps(struct bridge_softc *);
 292 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 293                     int);
 294 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 295 static void     bridge_init(void *);
 296 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 297 static void     bridge_stop(struct ifnet *, int);
 298 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 299 #ifdef ALTQ
 300 static void     bridge_altq_start(if_t);
 301 static int      bridge_altq_transmit(if_t, struct mbuf *);
 302 #endif
 303 static void     bridge_qflush(struct ifnet *);
 304 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 305 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 306                     struct rtentry *);
 307 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 308                     struct mbuf *);
 309 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 310
 311 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 312                     struct mbuf *m);
 313
 314 static void     bridge_timer(void *);
 315
 316 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 317                     struct mbuf *, int);
 318 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 319
 320 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 321                     uint16_t, struct bridge_iflist *, int, uint8_t);
 322 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 323                     uint16_t);
 324 static void     bridge_rttrim(struct bridge_softc *);
 325 static void     bridge_rtage(struct bridge_softc *);
 326 static void     bridge_rtflush(struct bridge_softc *, int);
 327 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 328                     uint16_t);
 329
 330 static void     bridge_rtable_init(struct bridge_softc *);
 331 static void     bridge_rtable_fini(struct bridge_softc *);
 332
 333 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 334 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 335                     const uint8_t *, uint16_t);
 336 static int      bridge_rtnode_insert(struct bridge_softc *,
 337                     struct bridge_rtnode *);
 338 static void     bridge_rtnode_destroy(struct bridge_softc *,
 339                     struct bridge_rtnode *);
 340 static void     bridge_rtable_expire(struct ifnet *, int);
 341 static void     bridge_state_change(struct ifnet *, int);
 342
 343 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 344                     const char *name);
 345 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 346                     struct ifnet *ifp);
 347 static void     bridge_delete_member(struct bridge_softc *,
 348                     struct bridge_iflist *, int);
 349 static void     bridge_delete_span(struct bridge_softc *,
 350                     struct bridge_iflist *);
 351
 352 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 353 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 354 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 381 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 382 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 383 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 384                     int);
 385 static int      bridge_ip_checkbasic(struct mbuf **mp);
 386 #ifdef INET6
 387 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 388 #endif /* INET6 */
 389 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 390                     struct ether_header *, int, struct llc *);
 391 static void     bridge_linkstate(struct ifnet *ifp);
 392 static void     bridge_linkcheck(struct bridge_softc *sc);
 393
 394 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 395 #define VLANTAGOF(_m)   \
 396     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 397
 398 static struct bstp_cb_ops bridge_ops = {
 399         .bcb_state = bridge_state_change,
 400         .bcb_rtage = bridge_rtable_expire
 401 };
 402
 403 SYSCTL_DECL(_net_link);
 404 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 405     "Bridge");
 406
 407 /* only pass IP[46] packets when pfil is enabled */
 408 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 409 #define V_pfil_onlyip   VNET(pfil_onlyip)
 410 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 411     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 412     "Only pass IP packets when pfil is enabled");
 413
 414 /* run pfil hooks on the bridge interface */
 415 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 416 #define V_pfil_bridge   VNET(pfil_bridge)
 417 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 418     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 419     "Packet filter on the bridge interface");
 420
 421 /* layer2 filter with ipfw */
 422 VNET_DEFINE_STATIC(int, pfil_ipfw);
 423 #define V_pfil_ipfw     VNET(pfil_ipfw)
 424
 425 /* layer2 ARP filter with ipfw */
 426 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 427 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 428 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 429     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 430     "Filter ARP packets through IPFW layer2");
 431
 432 /* run pfil hooks on the member interface */
 433 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 434 #define V_pfil_member   VNET(pfil_member)
 435 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 436     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 437     "Packet filter on the member interface");
 438
 439 /* run pfil hooks on the physical interface for locally destined packets */
 440 VNET_DEFINE_STATIC(int, pfil_local_phys);
 441 #define V_pfil_local_phys       VNET(pfil_local_phys)
 442 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 443     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 444     "Packet filter on the physical interface for locally destined packets");
 445
 446 /* log STP state changes */
 447 VNET_DEFINE_STATIC(int, log_stp);
 448 #define V_log_stp       VNET(log_stp)
 449 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 450     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 451     "Log STP state changes");
 452
 453 /* share MAC with first bridge member */
 454 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 455 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 456 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 457     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 458     "Inherit MAC address from the first bridge member");
 459
 460 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 461 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 462 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 463     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 464     "Allow overlap of link-local scope "
 465     "zones of a bridge interface and the member interfaces");
 466
 467 /* log MAC address port flapping */
 468 VNET_DEFINE_STATIC(bool, log_mac_flap) = true;
 469 #define V_log_mac_flap  VNET(log_mac_flap)
 470 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
 471     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true,
 472     "Log MAC address port flapping");
 473
 474 VNET_DEFINE_STATIC(int, log_interval) = 5;
 475 VNET_DEFINE_STATIC(int, log_count) = 0;
 476 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 };
 477
 478 #define V_log_interval  VNET(log_interval)
 479 #define V_log_count     VNET(log_count)
 480 #define V_log_last      VNET(log_last)
 481
 482 struct bridge_control {
 483         int     (*bc_func)(struct bridge_softc *, void *);
 484         int     bc_argsize;
 485         int     bc_flags;
 486 };
 487
 488 #define BC_F_COPYIN             0x01    /* copy arguments in */
 489 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 490 #define BC_F_SUSER              0x04    /* do super-user check */
 491
 492 static const struct bridge_control bridge_control_table[] = {
 493         { bridge_ioctl_add,             sizeof(struct ifbreq),
 494           BC_F_COPYIN|BC_F_SUSER },
 495         { bridge_ioctl_del,             sizeof(struct ifbreq),
 496           BC_F_COPYIN|BC_F_SUSER },
 497
 498         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 499           BC_F_COPYIN|BC_F_COPYOUT },
 500         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 501           BC_F_COPYIN|BC_F_SUSER },
 502
 503         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 504           BC_F_COPYIN|BC_F_SUSER },
 505         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 506           BC_F_COPYOUT },
 507
 508         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 509           BC_F_COPYIN|BC_F_COPYOUT },
 510         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 511           BC_F_COPYIN|BC_F_COPYOUT },
 512
 513         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 514           BC_F_COPYIN|BC_F_SUSER },
 515
 516         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 517           BC_F_COPYIN|BC_F_SUSER },
 518         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 519           BC_F_COPYOUT },
 520
 521         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 522           BC_F_COPYIN|BC_F_SUSER },
 523
 524         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 525           BC_F_COPYIN|BC_F_SUSER },
 526
 527         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 528           BC_F_COPYOUT },
 529         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 530           BC_F_COPYIN|BC_F_SUSER },
 531
 532         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 533           BC_F_COPYOUT },
 534         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 535           BC_F_COPYIN|BC_F_SUSER },
 536
 537         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 538           BC_F_COPYOUT },
 539         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 540           BC_F_COPYIN|BC_F_SUSER },
 541
 542         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 543           BC_F_COPYOUT },
 544         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 545           BC_F_COPYIN|BC_F_SUSER },
 546
 547         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 548           BC_F_COPYIN|BC_F_SUSER },
 549
 550         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 551           BC_F_COPYIN|BC_F_SUSER },
 552
 553         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 554           BC_F_COPYIN|BC_F_SUSER },
 555         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 556           BC_F_COPYIN|BC_F_SUSER },
 557
 558         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 559           BC_F_COPYOUT },
 560
 561         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 562           BC_F_COPYOUT },
 563
 564         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 565           BC_F_COPYIN|BC_F_COPYOUT },
 566
 567         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 568           BC_F_COPYIN|BC_F_SUSER },
 569
 570         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 571           BC_F_COPYIN|BC_F_SUSER },
 572
 573         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 574           BC_F_COPYIN|BC_F_SUSER },
 575
 576 };
 577 static const int bridge_control_table_size = nitems(bridge_control_table);
 578
 579 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 580 #define V_bridge_list   VNET(bridge_list)
 581 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 582                                             "if_bridge list")
 583 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 584 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 585 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 586
 587 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 588 #define V_bridge_cloner VNET(bridge_cloner)
 589
 590 static const char bridge_name[] = "bridge";
 591
 592 static void
 593 vnet_bridge_init(const void *unused __unused)
 594 {
 595
 596         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 597             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 598             UMA_ALIGN_PTR, 0);
 599         BRIDGE_LIST_LOCK_INIT();
 600         LIST_INIT(&V_bridge_list);
 601         V_bridge_cloner = if_clone_simple(bridge_name,
 602             bridge_clone_create, bridge_clone_destroy, 0);
 603 }
 604 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 605     vnet_bridge_init, NULL);
 606
 607 static void
 608 vnet_bridge_uninit(const void *unused __unused)
 609 {
 610
 611         if_clone_detach(V_bridge_cloner);
 612         V_bridge_cloner = NULL;
 613         BRIDGE_LIST_LOCK_DESTROY();
 614
 615         /* Callbacks may use the UMA zone. */
 616         NET_EPOCH_DRAIN_CALLBACKS();
 617
 618         uma_zdestroy(V_bridge_rtnode_zone);
 619 }
 620 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 621     vnet_bridge_uninit, NULL);
 622
 623 static int
 624 bridge_modevent(module_t mod, int type, void *data)
 625 {
 626
 627         switch (type) {
 628         case MOD_LOAD:
 629                 bridge_dn_p = bridge_dummynet;
 630                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 631                     ifnet_departure_event, bridge_ifdetach, NULL,
 632                     EVENTHANDLER_PRI_ANY);
 633                 break;
 634         case MOD_UNLOAD:
 635                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 636                     bridge_detach_cookie);
 637                 bridge_dn_p = NULL;
 638                 break;
 639         default:
 640                 return (EOPNOTSUPP);
 641         }
 642         return (0);
 643 }
 644
 645 static moduledata_t bridge_mod = {
 646         "if_bridge",
 647         bridge_modevent,
 648         0
 649 };
 650
 651 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 652 MODULE_VERSION(if_bridge, 1);
 653 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 654
 655 /*
 656  * handler for net.link.bridge.ipfw
 657  */
 658 static int
 659 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 660 {
 661         int enable = V_pfil_ipfw;
 662         int error;
 663
 664         error = sysctl_handle_int(oidp, &enable, 0, req);
 665         enable &= 1;
 666
 667         if (enable != V_pfil_ipfw) {
 668                 V_pfil_ipfw = enable;
 669
 670                 /*
 671                  * Disable pfil so that ipfw doesnt run twice, if the user
 672                  * really wants both then they can re-enable pfil_bridge and/or
 673                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 674                  * layer2 type.
 675                  */
 676                 if (V_pfil_ipfw) {
 677                         V_pfil_onlyip = 0;
 678                         V_pfil_bridge = 0;
 679                         V_pfil_member = 0;
 680                 }
 681         }
 682
 683         return (error);
 684 }
 685 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 686     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
 687     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 688     "Layer2 filter with IPFW");
 689
 690 #ifdef VIMAGE
 691 static void
 692 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 693 {
 694         struct bridge_softc *sc = ifp->if_softc;
 695         struct bridge_iflist *bif;
 696
 697         BRIDGE_LOCK(sc);
 698
 699         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 700                 bridge_delete_member(sc, bif, 0);
 701
 702         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 703                 bridge_delete_span(sc, bif);
 704         }
 705
 706         BRIDGE_UNLOCK(sc);
 707
 708         ether_reassign(ifp, newvnet, arg);
 709 }
 710 #endif
 711
 712 /*
 713  * bridge_clone_create:
 714  *
 715  *      Create a new bridge instance.
 716  */
 717 static int
 718 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 719 {
 720         struct bridge_softc *sc;
 721         struct ifnet *ifp;
 722
 723         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 724         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 725         if (ifp == NULL) {
 726                 free(sc, M_DEVBUF);
 727                 return (ENOSPC);
 728         }
 729
 730         BRIDGE_LOCK_INIT(sc);
 731         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 732         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 733
 734         /* Initialize our routing table. */
 735         bridge_rtable_init(sc);
 736
 737         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 738
 739         CK_LIST_INIT(&sc->sc_iflist);
 740         CK_LIST_INIT(&sc->sc_spanlist);
 741
 742         ifp->if_softc = sc;
 743         if_initname(ifp, bridge_name, unit);
 744         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 745         ifp->if_ioctl = bridge_ioctl;
 746 #ifdef ALTQ
 747         ifp->if_start = bridge_altq_start;
 748         ifp->if_transmit = bridge_altq_transmit;
 749         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 750         ifp->if_snd.ifq_drv_maxlen = 0;
 751         IFQ_SET_READY(&ifp->if_snd);
 752 #else
 753         ifp->if_transmit = bridge_transmit;
 754 #endif
 755         ifp->if_qflush = bridge_qflush;
 756         ifp->if_init = bridge_init;
 757         ifp->if_type = IFT_BRIDGE;
 758
 759         ether_gen_addr(ifp, &sc->sc_defaddr);
 760
 761         bstp_attach(&sc->sc_stp, &bridge_ops);
 762         ether_ifattach(ifp, sc->sc_defaddr.octet);
 763         /* Now undo some of the damage... */
 764         ifp->if_baudrate = 0;
 765         ifp->if_type = IFT_BRIDGE;
 766 #ifdef VIMAGE
 767         ifp->if_reassign = bridge_reassign;
 768 #endif
 769
 770         BRIDGE_LIST_LOCK();
 771         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 772         BRIDGE_LIST_UNLOCK();
 773
 774         return (0);
 775 }
 776
 777 static void
 778 bridge_clone_destroy_cb(struct epoch_context *ctx)
 779 {
 780         struct bridge_softc *sc;
 781
 782         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 783
 784         BRIDGE_LOCK_DESTROY(sc);
 785         free(sc, M_DEVBUF);
 786 }
 787
 788 /*
 789  * bridge_clone_destroy:
 790  *
 791  *      Destroy a bridge instance.
 792  */
 793 static void
 794 bridge_clone_destroy(struct ifnet *ifp)
 795 {
 796         struct bridge_softc *sc = ifp->if_softc;
 797         struct bridge_iflist *bif;
 798         struct epoch_tracker et;
 799
 800         BRIDGE_LOCK(sc);
 801
 802         bridge_stop(ifp, 1);
 803         ifp->if_flags &= ~IFF_UP;
 804
 805         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 806                 bridge_delete_member(sc, bif, 0);
 807
 808         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 809                 bridge_delete_span(sc, bif);
 810         }
 811
 812         /* Tear down the routing table. */
 813         bridge_rtable_fini(sc);
 814
 815         BRIDGE_UNLOCK(sc);
 816
 817         NET_EPOCH_ENTER(et);
 818
 819         callout_drain(&sc->sc_brcallout);
 820
 821         BRIDGE_LIST_LOCK();
 822         LIST_REMOVE(sc, sc_list);
 823         BRIDGE_LIST_UNLOCK();
 824
 825         bstp_detach(&sc->sc_stp);
 826 #ifdef ALTQ
 827         IFQ_PURGE(&ifp->if_snd);
 828 #endif
 829         NET_EPOCH_EXIT(et);
 830
 831         ether_ifdetach(ifp);
 832         if_free(ifp);
 833
 834         NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
 835 }
 836
 837 /*
 838  * bridge_ioctl:
 839  *
 840  *      Handle a control request from the operator.
 841  */
 842 static int
 843 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 844 {
 845         struct bridge_softc *sc = ifp->if_softc;
 846         struct ifreq *ifr = (struct ifreq *)data;
 847         struct bridge_iflist *bif;
 848         struct thread *td = curthread;
 849         union {
 850                 struct ifbreq ifbreq;
 851                 struct ifbifconf ifbifconf;
 852                 struct ifbareq ifbareq;
 853                 struct ifbaconf ifbaconf;
 854                 struct ifbrparam ifbrparam;
 855                 struct ifbropreq ifbropreq;
 856         } args;
 857         struct ifdrv *ifd = (struct ifdrv *) data;
 858         const struct bridge_control *bc;
 859         int error = 0, oldmtu;
 860
 861         BRIDGE_LOCK(sc);
 862
 863         switch (cmd) {
 864         case SIOCADDMULTI:
 865         case SIOCDELMULTI:
 866                 break;
 867
 868         case SIOCGDRVSPEC:
 869         case SIOCSDRVSPEC:
 870                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 871                         error = EINVAL;
 872                         break;
 873                 }
 874                 bc = &bridge_control_table[ifd->ifd_cmd];
 875
 876                 if (cmd == SIOCGDRVSPEC &&
 877                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 878                         error = EINVAL;
 879                         break;
 880                 }
 881                 else if (cmd == SIOCSDRVSPEC &&
 882                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 883                         error = EINVAL;
 884                         break;
 885                 }
 886
 887                 if (bc->bc_flags & BC_F_SUSER) {
 888                         error = priv_check(td, PRIV_NET_BRIDGE);
 889                         if (error)
 890                                 break;
 891                 }
 892
 893                 if (ifd->ifd_len != bc->bc_argsize ||
 894                     ifd->ifd_len > sizeof(args)) {
 895                         error = EINVAL;
 896                         break;
 897                 }
 898
 899                 bzero(&args, sizeof(args));
 900                 if (bc->bc_flags & BC_F_COPYIN) {
 901                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 902                         if (error)
 903                                 break;
 904                 }
 905
 906                 oldmtu = ifp->if_mtu;
 907                 error = (*bc->bc_func)(sc, &args);
 908                 if (error)
 909                         break;
 910
 911                 /*
 912                  * Bridge MTU may change during addition of the first port.
 913                  * If it did, do network layer specific procedure.
 914                  */
 915                 if (ifp->if_mtu != oldmtu) {
 916 #ifdef INET6
 917                         nd6_setmtu(ifp);
 918 #endif
 919                         rt_updatemtu(ifp);
 920                 }
 921
 922                 if (bc->bc_flags & BC_F_COPYOUT)
 923                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 924
 925                 break;
 926
 927         case SIOCSIFFLAGS:
 928                 if (!(ifp->if_flags & IFF_UP) &&
 929                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 930                         /*
 931                          * If interface is marked down and it is running,
 932                          * then stop and disable it.
 933                          */
 934                         bridge_stop(ifp, 1);
 935                 } else if ((ifp->if_flags & IFF_UP) &&
 936                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 937                         /*
 938                          * If interface is marked up and it is stopped, then
 939                          * start it.
 940                          */
 941                         BRIDGE_UNLOCK(sc);
 942                         (*ifp->if_init)(sc);
 943                         BRIDGE_LOCK(sc);
 944                 }
 945                 break;
 946
 947         case SIOCSIFMTU:
 948                 if (ifr->ifr_mtu < 576) {
 949                         error = EINVAL;
 950                         break;
 951                 }
 952                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 953                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 954                         break;
 955                 }
 956                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 957                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 958                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 959                                     " != %d\n", sc->sc_ifp->if_xname,
 960                                     bif->bif_ifp->if_mtu,
 961                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 962                                 error = EINVAL;
 963                                 break;
 964                         }
 965                 }
 966                 if (!error)
 967                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 968                 break;
 969         default:
 970                 /*
 971                  * drop the lock as ether_ioctl() will call bridge_start() and
 972                  * cause the lock to be recursed.
 973                  */
 974                 BRIDGE_UNLOCK(sc);
 975                 error = ether_ioctl(ifp, cmd, data);
 976                 BRIDGE_LOCK(sc);
 977                 break;
 978         }
 979
 980         BRIDGE_UNLOCK(sc);
 981
 982         return (error);
 983 }
 984
 985 /*
 986  * bridge_mutecaps:
 987  *
 988  *      Clear or restore unwanted capabilities on the member interface
 989  */
 990 static void
 991 bridge_mutecaps(struct bridge_softc *sc)
 992 {
 993         struct bridge_iflist *bif;
 994         int enabled, mask;
 995
 996         BRIDGE_LOCK_ASSERT(sc);
 997
 998         /* Initial bitmask of capabilities to test */
 999         mask = BRIDGE_IFCAPS_MASK;
1000
1001         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1002                 /* Every member must support it or its disabled */
1003                 mask &= bif->bif_savedcaps;
1004         }
1005
1006         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1007                 enabled = bif->bif_ifp->if_capenable;
1008                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1009                 /* strip off mask bits and enable them again if allowed */
1010                 enabled &= ~BRIDGE_IFCAPS_MASK;
1011                 enabled |= mask;
1012                 bridge_set_ifcap(sc, bif, enabled);
1013         }
1014 }
1015
1016 static void
1017 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1018 {
1019         struct ifnet *ifp = bif->bif_ifp;
1020         struct ifreq ifr;
1021         int error, mask, stuck;
1022
1023         bzero(&ifr, sizeof(ifr));
1024         ifr.ifr_reqcap = set;
1025
1026         if (ifp->if_capenable != set) {
1027                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1028                 if (error)
1029                         if_printf(sc->sc_ifp,
1030                             "error setting capabilities on %s: %d\n",
1031                             ifp->if_xname, error);
1032                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1033                 stuck = ifp->if_capenable & mask & ~set;
1034                 if (stuck != 0)
1035                         if_printf(sc->sc_ifp,
1036                             "can't disable some capabilities on %s: 0x%x\n",
1037                             ifp->if_xname, stuck);
1038         }
1039 }
1040
1041 /*
1042  * bridge_lookup_member:
1043  *
1044  *      Lookup a bridge member interface.
1045  */
1046 static struct bridge_iflist *
1047 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1048 {
1049         struct bridge_iflist *bif;
1050         struct ifnet *ifp;
1051
1052         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1053
1054         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1055                 ifp = bif->bif_ifp;
1056                 if (strcmp(ifp->if_xname, name) == 0)
1057                         return (bif);
1058         }
1059
1060         return (NULL);
1061 }
1062
1063 /*
1064  * bridge_lookup_member_if:
1065  *
1066  *      Lookup a bridge member interface by ifnet*.
1067  */
1068 static struct bridge_iflist *
1069 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1070 {
1071         struct bridge_iflist *bif;
1072
1073         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1074
1075         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1076                 if (bif->bif_ifp == member_ifp)
1077                         return (bif);
1078         }
1079
1080         return (NULL);
1081 }
1082
1083 static void
1084 bridge_delete_member_cb(struct epoch_context *ctx)
1085 {
1086         struct bridge_iflist *bif;
1087
1088         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1089
1090         free(bif, M_DEVBUF);
1091 }
1092
1093 /*
1094  * bridge_delete_member:
1095  *
1096  *      Delete the specified member interface.
1097  */
1098 static void
1099 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1100     int gone)
1101 {
1102         struct ifnet *ifs = bif->bif_ifp;
1103         struct ifnet *fif = NULL;
1104         struct bridge_iflist *bifl;
1105
1106         BRIDGE_LOCK_ASSERT(sc);
1107
1108         if (bif->bif_flags & IFBIF_STP)
1109                 bstp_disable(&bif->bif_stp);
1110
1111         ifs->if_bridge = NULL;
1112         CK_LIST_REMOVE(bif, bif_next);
1113
1114         /*
1115          * If removing the interface that gave the bridge its mac address, set
1116          * the mac address of the bridge to the address of the next member, or
1117          * to its default address if no members are left.
1118          */
1119         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1120                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1121                         bcopy(&sc->sc_defaddr,
1122                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1123                         sc->sc_ifaddr = NULL;
1124                 } else {
1125                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1126                         fif = bifl->bif_ifp;
1127                         bcopy(IF_LLADDR(fif),
1128                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1129                         sc->sc_ifaddr = fif;
1130                 }
1131                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1132         }
1133
1134         bridge_linkcheck(sc);
1135         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1136         BRIDGE_RT_LOCK(sc);
1137         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1138         BRIDGE_RT_UNLOCK(sc);
1139         KASSERT(bif->bif_addrcnt == 0,
1140             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1141
1142         ifs->if_bridge_output = NULL;
1143         ifs->if_bridge_input = NULL;
1144         ifs->if_bridge_linkstate = NULL;
1145         if (!gone) {
1146                 switch (ifs->if_type) {
1147                 case IFT_ETHER:
1148                 case IFT_L2VLAN:
1149                         /*
1150                          * Take the interface out of promiscuous mode, but only
1151                          * if it was promiscuous in the first place. It might
1152                          * not be if we're in the bridge_ioctl_add() error path.
1153                          */
1154                         if (ifs->if_flags & IFF_PROMISC)
1155                                 (void) ifpromisc(ifs, 0);
1156                         break;
1157
1158                 case IFT_GIF:
1159                         break;
1160
1161                 default:
1162 #ifdef DIAGNOSTIC
1163                         panic("bridge_delete_member: impossible");
1164 #endif
1165                         break;
1166                 }
1167                 /* reneable any interface capabilities */
1168                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1169         }
1170         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1171
1172         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1173 }
1174
1175 /*
1176  * bridge_delete_span:
1177  *
1178  *      Delete the specified span interface.
1179  */
1180 static void
1181 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1182 {
1183         BRIDGE_LOCK_ASSERT(sc);
1184
1185         KASSERT(bif->bif_ifp->if_bridge == NULL,
1186             ("%s: not a span interface", __func__));
1187
1188         CK_LIST_REMOVE(bif, bif_next);
1189
1190         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1191 }
1192
1193 static int
1194 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1195 {
1196         struct ifbreq *req = arg;
1197         struct bridge_iflist *bif = NULL;
1198         struct ifnet *ifs;
1199         int error = 0;
1200
1201         ifs = ifunit(req->ifbr_ifsname);
1202         if (ifs == NULL)
1203                 return (ENOENT);
1204         if (ifs->if_ioctl == NULL)      /* must be supported */
1205                 return (EINVAL);
1206
1207         /* If it's in the span list, it can't be a member. */
1208         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1209                 if (ifs == bif->bif_ifp)
1210                         return (EBUSY);
1211
1212         if (ifs->if_bridge == sc)
1213                 return (EEXIST);
1214
1215         if (ifs->if_bridge != NULL)
1216                 return (EBUSY);
1217
1218         switch (ifs->if_type) {
1219         case IFT_ETHER:
1220         case IFT_L2VLAN:
1221         case IFT_GIF:
1222                 /* permitted interface types */
1223                 break;
1224         default:
1225                 return (EINVAL);
1226         }
1227
1228 #ifdef INET6
1229         /*
1230          * Two valid inet6 addresses with link-local scope must not be
1231          * on the parent interface and the member interfaces at the
1232          * same time.  This restriction is needed to prevent violation
1233          * of link-local scope zone.  Attempts to add a member
1234          * interface which has inet6 addresses when the parent has
1235          * inet6 triggers removal of all inet6 addresses on the member
1236          * interface.
1237          */
1238
1239         /* Check if the parent interface has a link-local scope addr. */
1240         if (V_allow_llz_overlap == 0 &&
1241             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1242                 /*
1243                  * If any, remove all inet6 addresses from the member
1244                  * interfaces.
1245                  */
1246                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1247                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1248                                 in6_ifdetach(bif->bif_ifp);
1249                                 if_printf(sc->sc_ifp,
1250                                     "IPv6 addresses on %s have been removed "
1251                                     "before adding it as a member to prevent "
1252                                     "IPv6 address scope violation.\n",
1253                                     bif->bif_ifp->if_xname);
1254                         }
1255                 }
1256                 if (in6ifa_llaonifp(ifs)) {
1257                         in6_ifdetach(ifs);
1258                         if_printf(sc->sc_ifp,
1259                             "IPv6 addresses on %s have been removed "
1260                             "before adding it as a member to prevent "
1261                             "IPv6 address scope violation.\n",
1262                             ifs->if_xname);
1263                 }
1264         }
1265 #endif
1266         /* Allow the first Ethernet member to define the MTU */
1267         if (CK_LIST_EMPTY(&sc->sc_iflist))
1268                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1269         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1270                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1271                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1272                 return (EINVAL);
1273         }
1274
1275         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1276         if (bif == NULL)
1277                 return (ENOMEM);
1278
1279         bif->bif_ifp = ifs;
1280         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1281         bif->bif_savedcaps = ifs->if_capenable;
1282
1283         /*
1284          * Assign the interface's MAC address to the bridge if it's the first
1285          * member and the MAC address of the bridge has not been changed from
1286          * the default randomly generated one.
1287          */
1288         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1289             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1290                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1291                 sc->sc_ifaddr = ifs;
1292                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1293         }
1294
1295         ifs->if_bridge = sc;
1296         ifs->if_bridge_output = bridge_output;
1297         ifs->if_bridge_input = bridge_input;
1298         ifs->if_bridge_linkstate = bridge_linkstate;
1299         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1300         /*
1301          * XXX: XLOCK HERE!?!
1302          *
1303          * NOTE: insert_***HEAD*** should be safe for the traversals.
1304          */
1305         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1306
1307         /* Set interface capabilities to the intersection set of all members */
1308         bridge_mutecaps(sc);
1309         bridge_linkcheck(sc);
1310
1311         /* Place the interface into promiscuous mode */
1312         switch (ifs->if_type) {
1313                 case IFT_ETHER:
1314                 case IFT_L2VLAN:
1315                         error = ifpromisc(ifs, 1);
1316                         break;
1317         }
1318
1319         if (error)
1320                 bridge_delete_member(sc, bif, 0);
1321         return (error);
1322 }
1323
1324 static int
1325 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1326 {
1327         struct ifbreq *req = arg;
1328         struct bridge_iflist *bif;
1329
1330         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1331         if (bif == NULL)
1332                 return (ENOENT);
1333
1334         bridge_delete_member(sc, bif, 0);
1335
1336         return (0);
1337 }
1338
1339 static int
1340 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1341 {
1342         struct ifbreq *req = arg;
1343         struct bridge_iflist *bif;
1344         struct bstp_port *bp;
1345
1346         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1347         if (bif == NULL)
1348                 return (ENOENT);
1349
1350         bp = &bif->bif_stp;
1351         req->ifbr_ifsflags = bif->bif_flags;
1352         req->ifbr_state = bp->bp_state;
1353         req->ifbr_priority = bp->bp_priority;
1354         req->ifbr_path_cost = bp->bp_path_cost;
1355         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1356         req->ifbr_proto = bp->bp_protover;
1357         req->ifbr_role = bp->bp_role;
1358         req->ifbr_stpflags = bp->bp_flags;
1359         req->ifbr_addrcnt = bif->bif_addrcnt;
1360         req->ifbr_addrmax = bif->bif_addrmax;
1361         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1362
1363         /* Copy STP state options as flags */
1364         if (bp->bp_operedge)
1365                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1366         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1367                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1368         if (bp->bp_ptp_link)
1369                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1370         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1371                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1372         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1373                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1374         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1375                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1376         return (0);
1377 }
1378
1379 static int
1380 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1381 {
1382         struct epoch_tracker et;
1383         struct ifbreq *req = arg;
1384         struct bridge_iflist *bif;
1385         struct bstp_port *bp;
1386         int error;
1387
1388         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1389         if (bif == NULL)
1390                 return (ENOENT);
1391         bp = &bif->bif_stp;
1392
1393         if (req->ifbr_ifsflags & IFBIF_SPAN)
1394                 /* SPAN is readonly */
1395                 return (EINVAL);
1396
1397         NET_EPOCH_ENTER(et);
1398
1399         if (req->ifbr_ifsflags & IFBIF_STP) {
1400                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1401                         error = bstp_enable(&bif->bif_stp);
1402                         if (error) {
1403                                 NET_EPOCH_EXIT(et);
1404                                 return (error);
1405                         }
1406                 }
1407         } else {
1408                 if ((bif->bif_flags & IFBIF_STP) != 0)
1409                         bstp_disable(&bif->bif_stp);
1410         }
1411
1412         /* Pass on STP flags */
1413         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1414         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1415         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1416         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1417
1418         /* Save the bits relating to the bridge */
1419         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1420
1421         NET_EPOCH_EXIT(et);
1422
1423         return (0);
1424 }
1425
1426 static int
1427 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1428 {
1429         struct ifbrparam *param = arg;
1430
1431         sc->sc_brtmax = param->ifbrp_csize;
1432         bridge_rttrim(sc);
1433
1434         return (0);
1435 }
1436
1437 static int
1438 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1439 {
1440         struct ifbrparam *param = arg;
1441
1442         param->ifbrp_csize = sc->sc_brtmax;
1443
1444         return (0);
1445 }
1446
1447 static int
1448 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1449 {
1450         struct ifbifconf *bifc = arg;
1451         struct bridge_iflist *bif;
1452         struct ifbreq breq;
1453         char *buf, *outbuf;
1454         int count, buflen, len, error = 0;
1455
1456         count = 0;
1457         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1458                 count++;
1459         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1460                 count++;
1461
1462         buflen = sizeof(breq) * count;
1463         if (bifc->ifbic_len == 0) {
1464                 bifc->ifbic_len = buflen;
1465                 return (0);
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(bifc->ifbic_len, buflen);
1474         bzero(&breq, sizeof(breq));
1475         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1476                 if (len < sizeof(breq))
1477                         break;
1478
1479                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1480                     sizeof(breq.ifbr_ifsname));
1481                 /* Fill in the ifbreq structure */
1482                 error = bridge_ioctl_gifflags(sc, &breq);
1483                 if (error)
1484                         break;
1485                 memcpy(buf, &breq, sizeof(breq));
1486                 count++;
1487                 buf += sizeof(breq);
1488                 len -= sizeof(breq);
1489         }
1490         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1491                 if (len < sizeof(breq))
1492                         break;
1493
1494                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1495                     sizeof(breq.ifbr_ifsname));
1496                 breq.ifbr_ifsflags = bif->bif_flags;
1497                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1498                 memcpy(buf, &breq, sizeof(breq));
1499                 count++;
1500                 buf += sizeof(breq);
1501                 len -= sizeof(breq);
1502         }
1503
1504         bifc->ifbic_len = sizeof(breq) * count;
1505         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1506         free(outbuf, M_TEMP);
1507         return (error);
1508 }
1509
1510 static int
1511 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1512 {
1513         struct ifbaconf *bac = arg;
1514         struct bridge_rtnode *brt;
1515         struct ifbareq bareq;
1516         char *buf, *outbuf;
1517         int count, buflen, len, error = 0;
1518
1519         if (bac->ifbac_len == 0)
1520                 return (0);
1521
1522         count = 0;
1523         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1524                 count++;
1525         buflen = sizeof(bareq) * count;
1526
1527         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1528         if (outbuf == NULL)
1529                 return (ENOMEM);
1530
1531         count = 0;
1532         buf = outbuf;
1533         len = min(bac->ifbac_len, buflen);
1534         bzero(&bareq, sizeof(bareq));
1535         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1536                 if (len < sizeof(bareq))
1537                         goto out;
1538                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1539                     sizeof(bareq.ifba_ifsname));
1540                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1541                 bareq.ifba_vlan = brt->brt_vlan;
1542                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1543                                 time_uptime < brt->brt_expire)
1544                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1545                 else
1546                         bareq.ifba_expire = 0;
1547                 bareq.ifba_flags = brt->brt_flags;
1548
1549                 memcpy(buf, &bareq, sizeof(bareq));
1550                 count++;
1551                 buf += sizeof(bareq);
1552                 len -= sizeof(bareq);
1553         }
1554 out:
1555         bac->ifbac_len = sizeof(bareq) * count;
1556         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1557         free(outbuf, M_TEMP);
1558         return (error);
1559 }
1560
1561 static int
1562 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1563 {
1564         struct ifbareq *req = arg;
1565         struct bridge_iflist *bif;
1566         struct epoch_tracker et;
1567         int error;
1568
1569         NET_EPOCH_ENTER(et);
1570         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1571         if (bif == NULL) {
1572                 NET_EPOCH_EXIT(et);
1573                 return (ENOENT);
1574         }
1575
1576         /* bridge_rtupdate() may acquire the lock. */
1577         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1578             req->ifba_flags);
1579         NET_EPOCH_EXIT(et);
1580
1581         return (error);
1582 }
1583
1584 static int
1585 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1586 {
1587         struct ifbrparam *param = arg;
1588
1589         sc->sc_brttimeout = param->ifbrp_ctime;
1590         return (0);
1591 }
1592
1593 static int
1594 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1595 {
1596         struct ifbrparam *param = arg;
1597
1598         param->ifbrp_ctime = sc->sc_brttimeout;
1599         return (0);
1600 }
1601
1602 static int
1603 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1604 {
1605         struct ifbareq *req = arg;
1606
1607         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1608 }
1609
1610 static int
1611 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1612 {
1613         struct ifbreq *req = arg;
1614
1615         BRIDGE_RT_LOCK(sc);
1616         bridge_rtflush(sc, req->ifbr_ifsflags);
1617         BRIDGE_RT_UNLOCK(sc);
1618
1619         return (0);
1620 }
1621
1622 static int
1623 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1624 {
1625         struct ifbrparam *param = arg;
1626         struct bstp_state *bs = &sc->sc_stp;
1627
1628         param->ifbrp_prio = bs->bs_bridge_priority;
1629         return (0);
1630 }
1631
1632 static int
1633 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1634 {
1635         struct ifbrparam *param = arg;
1636
1637         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1638 }
1639
1640 static int
1641 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1642 {
1643         struct ifbrparam *param = arg;
1644         struct bstp_state *bs = &sc->sc_stp;
1645
1646         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1647         return (0);
1648 }
1649
1650 static int
1651 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1652 {
1653         struct ifbrparam *param = arg;
1654
1655         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1656 }
1657
1658 static int
1659 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1660 {
1661         struct ifbrparam *param = arg;
1662         struct bstp_state *bs = &sc->sc_stp;
1663
1664         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1665         return (0);
1666 }
1667
1668 static int
1669 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1670 {
1671         struct ifbrparam *param = arg;
1672
1673         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1674 }
1675
1676 static int
1677 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1678 {
1679         struct ifbrparam *param = arg;
1680         struct bstp_state *bs = &sc->sc_stp;
1681
1682         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1683         return (0);
1684 }
1685
1686 static int
1687 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1688 {
1689         struct ifbrparam *param = arg;
1690
1691         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1692 }
1693
1694 static int
1695 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1696 {
1697         struct ifbreq *req = arg;
1698         struct bridge_iflist *bif;
1699
1700         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1701         if (bif == NULL)
1702                 return (ENOENT);
1703
1704         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1705 }
1706
1707 static int
1708 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1709 {
1710         struct ifbreq *req = arg;
1711         struct bridge_iflist *bif;
1712
1713         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1714         if (bif == NULL)
1715                 return (ENOENT);
1716
1717         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1718 }
1719
1720 static int
1721 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1722 {
1723         struct ifbreq *req = arg;
1724         struct bridge_iflist *bif;
1725
1726         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1727         if (bif == NULL)
1728                 return (ENOENT);
1729
1730         bif->bif_addrmax = req->ifbr_addrmax;
1731         return (0);
1732 }
1733
1734 static int
1735 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1736 {
1737         struct ifbreq *req = arg;
1738         struct bridge_iflist *bif = NULL;
1739         struct ifnet *ifs;
1740
1741         ifs = ifunit(req->ifbr_ifsname);
1742         if (ifs == NULL)
1743                 return (ENOENT);
1744
1745         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1746                 if (ifs == bif->bif_ifp)
1747                         return (EBUSY);
1748
1749         if (ifs->if_bridge != NULL)
1750                 return (EBUSY);
1751
1752         switch (ifs->if_type) {
1753                 case IFT_ETHER:
1754                 case IFT_GIF:
1755                 case IFT_L2VLAN:
1756                         break;
1757                 default:
1758                         return (EINVAL);
1759         }
1760
1761         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1762         if (bif == NULL)
1763                 return (ENOMEM);
1764
1765         bif->bif_ifp = ifs;
1766         bif->bif_flags = IFBIF_SPAN;
1767
1768         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1769
1770         return (0);
1771 }
1772
1773 static int
1774 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1775 {
1776         struct ifbreq *req = arg;
1777         struct bridge_iflist *bif;
1778         struct ifnet *ifs;
1779
1780         ifs = ifunit(req->ifbr_ifsname);
1781         if (ifs == NULL)
1782                 return (ENOENT);
1783
1784         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1785                 if (ifs == bif->bif_ifp)
1786                         break;
1787
1788         if (bif == NULL)
1789                 return (ENOENT);
1790
1791         bridge_delete_span(sc, bif);
1792
1793         return (0);
1794 }
1795
1796 static int
1797 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1798 {
1799         struct ifbropreq *req = arg;
1800         struct bstp_state *bs = &sc->sc_stp;
1801         struct bstp_port *root_port;
1802
1803         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1804         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1805         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1806
1807         root_port = bs->bs_root_port;
1808         if (root_port == NULL)
1809                 req->ifbop_root_port = 0;
1810         else
1811                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1812
1813         req->ifbop_holdcount = bs->bs_txholdcount;
1814         req->ifbop_priority = bs->bs_bridge_priority;
1815         req->ifbop_protocol = bs->bs_protover;
1816         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1817         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1818         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1819         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1820         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1821         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1822
1823         return (0);
1824 }
1825
1826 static int
1827 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1828 {
1829         struct ifbrparam *param = arg;
1830
1831         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1832         return (0);
1833 }
1834
1835 static int
1836 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1837 {
1838         struct ifbpstpconf *bifstp = arg;
1839         struct bridge_iflist *bif;
1840         struct bstp_port *bp;
1841         struct ifbpstpreq bpreq;
1842         char *buf, *outbuf;
1843         int count, buflen, len, error = 0;
1844
1845         count = 0;
1846         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1847                 if ((bif->bif_flags & IFBIF_STP) != 0)
1848                         count++;
1849         }
1850
1851         buflen = sizeof(bpreq) * count;
1852         if (bifstp->ifbpstp_len == 0) {
1853                 bifstp->ifbpstp_len = buflen;
1854                 return (0);
1855         }
1856
1857         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1858         if (outbuf == NULL)
1859                 return (ENOMEM);
1860
1861         count = 0;
1862         buf = outbuf;
1863         len = min(bifstp->ifbpstp_len, buflen);
1864         bzero(&bpreq, sizeof(bpreq));
1865         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1866                 if (len < sizeof(bpreq))
1867                         break;
1868
1869                 if ((bif->bif_flags & IFBIF_STP) == 0)
1870                         continue;
1871
1872                 bp = &bif->bif_stp;
1873                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1874                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1875                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1876                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1877                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1878                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1879
1880                 memcpy(buf, &bpreq, sizeof(bpreq));
1881                 count++;
1882                 buf += sizeof(bpreq);
1883                 len -= sizeof(bpreq);
1884         }
1885
1886         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1887         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1888         free(outbuf, M_TEMP);
1889         return (error);
1890 }
1891
1892 static int
1893 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1894 {
1895         struct ifbrparam *param = arg;
1896
1897         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1898 }
1899
1900 static int
1901 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1902 {
1903         struct ifbrparam *param = arg;
1904
1905         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1906 }
1907
1908 /*
1909  * bridge_ifdetach:
1910  *
1911  *      Detach an interface from a bridge.  Called when a member
1912  *      interface is detaching.
1913  */
1914 static void
1915 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1916 {
1917         struct bridge_softc *sc = ifp->if_bridge;
1918         struct bridge_iflist *bif;
1919
1920         if (ifp->if_flags & IFF_RENAMING)
1921                 return;
1922         if (V_bridge_cloner == NULL) {
1923                 /*
1924                  * This detach handler can be called after
1925                  * vnet_bridge_uninit().  Just return in that case.
1926                  */
1927                 return;
1928         }
1929         /* Check if the interface is a bridge member */
1930         if (sc != NULL) {
1931                 BRIDGE_LOCK(sc);
1932
1933                 bif = bridge_lookup_member_if(sc, ifp);
1934                 if (bif != NULL)
1935                         bridge_delete_member(sc, bif, 1);
1936
1937                 BRIDGE_UNLOCK(sc);
1938                 return;
1939         }
1940
1941         /* Check if the interface is a span port */
1942         BRIDGE_LIST_LOCK();
1943         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1944                 BRIDGE_LOCK(sc);
1945                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1946                         if (ifp == bif->bif_ifp) {
1947                                 bridge_delete_span(sc, bif);
1948                                 break;
1949                         }
1950
1951                 BRIDGE_UNLOCK(sc);
1952         }
1953         BRIDGE_LIST_UNLOCK();
1954 }
1955
1956 /*
1957  * bridge_init:
1958  *
1959  *      Initialize a bridge interface.
1960  */
1961 static void
1962 bridge_init(void *xsc)
1963 {
1964         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1965         struct ifnet *ifp = sc->sc_ifp;
1966
1967         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1968                 return;
1969
1970         BRIDGE_LOCK(sc);
1971         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1972             bridge_timer, sc);
1973
1974         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1975         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1976
1977         BRIDGE_UNLOCK(sc);
1978 }
1979
1980 /*
1981  * bridge_stop:
1982  *
1983  *      Stop the bridge interface.
1984  */
1985 static void
1986 bridge_stop(struct ifnet *ifp, int disable)
1987 {
1988         struct bridge_softc *sc = ifp->if_softc;
1989
1990         BRIDGE_LOCK_ASSERT(sc);
1991
1992         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1993                 return;
1994
1995         BRIDGE_RT_LOCK(sc);
1996         callout_stop(&sc->sc_brcallout);
1997
1998         bstp_stop(&sc->sc_stp);
1999
2000         bridge_rtflush(sc, IFBF_FLUSHDYN);
2001         BRIDGE_RT_UNLOCK(sc);
2002
2003         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2004 }
2005
2006 /*
2007  * bridge_enqueue:
2008  *
2009  *      Enqueue a packet on a bridge member interface.
2010  *
2011  */
2012 static int
2013 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2014 {
2015         int len, err = 0;
2016         short mflags;
2017         struct mbuf *m0;
2018
2019         /* We may be sending a fragment so traverse the mbuf */
2020         for (; m; m = m0) {
2021                 m0 = m->m_nextpkt;
2022                 m->m_nextpkt = NULL;
2023                 len = m->m_pkthdr.len;
2024                 mflags = m->m_flags;
2025
2026                 /*
2027                  * If underlying interface can not do VLAN tag insertion itself
2028                  * then attach a packet tag that holds it.
2029                  */
2030                 if ((m->m_flags & M_VLANTAG) &&
2031                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2032                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2033                         if (m == NULL) {
2034                                 if_printf(dst_ifp,
2035                                     "unable to prepend VLAN header\n");
2036                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2037                                 continue;
2038                         }
2039                         m->m_flags &= ~M_VLANTAG;
2040                 }
2041
2042                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2043                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2044                         int n;
2045
2046                         for (m = m0, n = 1; m != NULL; m = m0, n++) {
2047                                 m0 = m->m_nextpkt;
2048                                 m_freem(m);
2049                         }
2050                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2051                         break;
2052                 }
2053
2054                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2055                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2056                 if (mflags & M_MCAST)
2057                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2058         }
2059
2060         return (err);
2061 }
2062
2063 /*
2064  * bridge_dummynet:
2065  *
2066  *      Receive a queued packet from dummynet and pass it on to the output
2067  *      interface.
2068  *
2069  *      The mbuf has the Ethernet header already attached.
2070  */
2071 static void
2072 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2073 {
2074         struct bridge_softc *sc;
2075
2076         sc = ifp->if_bridge;
2077
2078         /*
2079          * The packet didnt originate from a member interface. This should only
2080          * ever happen if a member interface is removed while packets are
2081          * queued for it.
2082          */
2083         if (sc == NULL) {
2084                 m_freem(m);
2085                 return;
2086         }
2087
2088         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2089 #ifdef INET6
2090             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2091 #endif
2092             ) {
2093                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2094                         return;
2095                 if (m == NULL)
2096                         return;
2097         }
2098
2099         bridge_enqueue(sc, ifp, m);
2100 }
2101
2102 /*
2103  * bridge_output:
2104  *
2105  *      Send output from a bridge member interface.  This
2106  *      performs the bridging function for locally originated
2107  *      packets.
2108  *
2109  *      The mbuf has the Ethernet header already attached.  We must
2110  *      enqueue or free the mbuf before returning.
2111  */
2112 static int
2113 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2114     struct rtentry *rt)
2115 {
2116         struct ether_header *eh;
2117         struct ifnet *bifp, *dst_if;
2118         struct bridge_softc *sc;
2119         uint16_t vlan;
2120
2121         NET_EPOCH_ASSERT();
2122
2123         if (m->m_len < ETHER_HDR_LEN) {
2124                 m = m_pullup(m, ETHER_HDR_LEN);
2125                 if (m == NULL)
2126                         return (0);
2127         }
2128
2129         eh = mtod(m, struct ether_header *);
2130         sc = ifp->if_bridge;
2131         vlan = VLANTAGOF(m);
2132
2133         bifp = sc->sc_ifp;
2134
2135         /*
2136          * If bridge is down, but the original output interface is up,
2137          * go ahead and send out that interface.  Otherwise, the packet
2138          * is dropped below.
2139          */
2140         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2141                 dst_if = ifp;
2142                 goto sendunicast;
2143         }
2144
2145         /*
2146          * If the packet is a multicast, or we don't know a better way to
2147          * get there, send to all interfaces.
2148          */
2149         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2150                 dst_if = NULL;
2151         else
2152                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2153         /* Tap any traffic not passing back out the originating interface */
2154         if (dst_if != ifp)
2155                 ETHER_BPF_MTAP(bifp, m);
2156         if (dst_if == NULL) {
2157                 struct bridge_iflist *bif;
2158                 struct mbuf *mc;
2159                 int used = 0;
2160
2161                 bridge_span(sc, m);
2162
2163                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2164                         dst_if = bif->bif_ifp;
2165
2166                         if (dst_if->if_type == IFT_GIF)
2167                                 continue;
2168                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2169                                 continue;
2170
2171                         /*
2172                          * If this is not the original output interface,
2173                          * and the interface is participating in spanning
2174                          * tree, make sure the port is in a state that
2175                          * allows forwarding.
2176                          */
2177                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2178                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2179                                 continue;
2180
2181                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2182                                 used = 1;
2183                                 mc = m;
2184                         } else {
2185                                 mc = m_dup(m, M_NOWAIT);
2186                                 if (mc == NULL) {
2187                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2188                                         continue;
2189                                 }
2190                         }
2191
2192                         bridge_enqueue(sc, dst_if, mc);
2193                 }
2194                 if (used == 0)
2195                         m_freem(m);
2196                 return (0);
2197         }
2198
2199 sendunicast:
2200         /*
2201          * XXX Spanning tree consideration here?
2202          */
2203
2204         bridge_span(sc, m);
2205         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2206                 m_freem(m);
2207                 return (0);
2208         }
2209
2210         bridge_enqueue(sc, dst_if, m);
2211         return (0);
2212 }
2213
2214 /*
2215  * bridge_transmit:
2216  *
2217  *      Do output on a bridge.
2218  *
2219  */
2220 static int
2221 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2222 {
2223         struct bridge_softc *sc;
2224         struct ether_header *eh;
2225         struct ifnet *dst_if;
2226         int error = 0;
2227
2228         sc = ifp->if_softc;
2229
2230         ETHER_BPF_MTAP(ifp, m);
2231
2232         eh = mtod(m, struct ether_header *);
2233
2234         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2235             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2236                 error = bridge_enqueue(sc, dst_if, m);
2237         } else
2238                 bridge_broadcast(sc, ifp, m, 0);
2239
2240         return (error);
2241 }
2242
2243 #ifdef ALTQ
2244 static void
2245 bridge_altq_start(if_t ifp)
2246 {
2247         struct ifaltq *ifq = &ifp->if_snd;
2248         struct mbuf *m;
2249
2250         IFQ_LOCK(ifq);
2251         IFQ_DEQUEUE_NOLOCK(ifq, m);
2252         while (m != NULL) {
2253                 bridge_transmit(ifp, m);
2254                 IFQ_DEQUEUE_NOLOCK(ifq, m);
2255         }
2256         IFQ_UNLOCK(ifq);
2257 }
2258
2259 static int
2260 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2261 {
2262         int err;
2263
2264         if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2265                 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2266                 if (err == 0)
2267                         bridge_altq_start(ifp);
2268         } else
2269                 err = bridge_transmit(ifp, m);
2270
2271         return (err);
2272 }
2273 #endif  /* ALTQ */
2274
2275 /*
2276  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2277  */
2278 static void
2279 bridge_qflush(struct ifnet *ifp __unused)
2280 {
2281 }
2282
2283 /*
2284  * bridge_forward:
2285  *
2286  *      The forwarding function of the bridge.
2287  *
2288  *      NOTE: Releases the lock on return.
2289  */
2290 static void
2291 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2292     struct mbuf *m)
2293 {
2294         struct bridge_iflist *dbif;
2295         struct ifnet *src_if, *dst_if, *ifp;
2296         struct ether_header *eh;
2297         uint16_t vlan;
2298         uint8_t *dst;
2299         int error;
2300
2301         NET_EPOCH_ASSERT();
2302
2303         src_if = m->m_pkthdr.rcvif;
2304         ifp = sc->sc_ifp;
2305
2306         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2307         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2308         vlan = VLANTAGOF(m);
2309
2310         if ((sbif->bif_flags & IFBIF_STP) &&
2311             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2312                 goto drop;
2313
2314         eh = mtod(m, struct ether_header *);
2315         dst = eh->ether_dhost;
2316
2317         /* If the interface is learning, record the address. */
2318         if (sbif->bif_flags & IFBIF_LEARNING) {
2319                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2320                     sbif, 0, IFBAF_DYNAMIC);
2321                 /*
2322                  * If the interface has addresses limits then deny any source
2323                  * that is not in the cache.
2324                  */
2325                 if (error && sbif->bif_addrmax)
2326                         goto drop;
2327         }
2328
2329         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2330             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2331                 goto drop;
2332
2333         /*
2334          * At this point, the port either doesn't participate
2335          * in spanning tree or it is in the forwarding state.
2336          */
2337
2338         /*
2339          * If the packet is unicast, destined for someone on
2340          * "this" side of the bridge, drop it.
2341          */
2342         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2343                 dst_if = bridge_rtlookup(sc, dst, vlan);
2344                 if (src_if == dst_if)
2345                         goto drop;
2346         } else {
2347                 /*
2348                  * Check if its a reserved multicast address, any address
2349                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2350                  * bridge.
2351                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2352                  */
2353                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2354                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2355                     dst[4] == 0x00 && dst[5] <= 0x0f)
2356                         goto drop;
2357
2358                 /* ...forward it to all interfaces. */
2359                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2360                 dst_if = NULL;
2361         }
2362
2363         /*
2364          * If we have a destination interface which is a member of our bridge,
2365          * OR this is a unicast packet, push it through the bpf(4) machinery.
2366          * For broadcast or multicast packets, don't bother because it will
2367          * be reinjected into ether_input. We do this before we pass the packets
2368          * through the pfil(9) framework, as it is possible that pfil(9) will
2369          * drop the packet, or possibly modify it, making it difficult to debug
2370          * firewall issues on the bridge.
2371          */
2372         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2373                 ETHER_BPF_MTAP(ifp, m);
2374
2375         /* run the packet filter */
2376         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2377 #ifdef INET6
2378             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2379 #endif
2380             ) {
2381                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2382                         return;
2383                 if (m == NULL)
2384                         return;
2385         }
2386
2387         if (dst_if == NULL) {
2388                 bridge_broadcast(sc, src_if, m, 1);
2389                 return;
2390         }
2391
2392         /*
2393          * At this point, we're dealing with a unicast frame
2394          * going to a different interface.
2395          */
2396         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2397                 goto drop;
2398
2399         dbif = bridge_lookup_member_if(sc, dst_if);
2400         if (dbif == NULL)
2401                 /* Not a member of the bridge (anymore?) */
2402                 goto drop;
2403
2404         /* Private segments can not talk to each other */
2405         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2406                 goto drop;
2407
2408         if ((dbif->bif_flags & IFBIF_STP) &&
2409             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2410                 goto drop;
2411
2412         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2413 #ifdef INET6
2414             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2415 #endif
2416             ) {
2417                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2418                         return;
2419                 if (m == NULL)
2420                         return;
2421         }
2422
2423         bridge_enqueue(sc, dst_if, m);
2424         return;
2425
2426 drop:
2427         m_freem(m);
2428 }
2429
2430 /*
2431  * bridge_input:
2432  *
2433  *      Receive input from a member interface.  Queue the packet for
2434  *      bridging if it is not for us.
2435  */
2436 static struct mbuf *
2437 bridge_input(struct ifnet *ifp, struct mbuf *m)
2438 {
2439         struct bridge_softc *sc = ifp->if_bridge;
2440         struct bridge_iflist *bif, *bif2;
2441         struct ifnet *bifp;
2442         struct ether_header *eh;
2443         struct mbuf *mc, *mc2;
2444         uint16_t vlan;
2445         int error;
2446
2447         NET_EPOCH_ASSERT();
2448
2449         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2450                 return (m);
2451
2452         bifp = sc->sc_ifp;
2453         vlan = VLANTAGOF(m);
2454
2455         /*
2456          * Implement support for bridge monitoring. If this flag has been
2457          * set on this interface, discard the packet once we push it through
2458          * the bpf(4) machinery, but before we do, increment the byte and
2459          * packet counters associated with this interface.
2460          */
2461         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2462                 m->m_pkthdr.rcvif  = bifp;
2463                 ETHER_BPF_MTAP(bifp, m);
2464                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2465                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2466                 m_freem(m);
2467                 return (NULL);
2468         }
2469         bif = bridge_lookup_member_if(sc, ifp);
2470         if (bif == NULL) {
2471                 return (m);
2472         }
2473
2474         eh = mtod(m, struct ether_header *);
2475
2476         bridge_span(sc, m);
2477
2478         if (m->m_flags & (M_BCAST|M_MCAST)) {
2479                 /* Tap off 802.1D packets; they do not get forwarded. */
2480                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2481                     ETHER_ADDR_LEN) == 0) {
2482                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2483                         return (NULL);
2484                 }
2485
2486                 if ((bif->bif_flags & IFBIF_STP) &&
2487                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2488                         return (m);
2489                 }
2490
2491                 /*
2492                  * Make a deep copy of the packet and enqueue the copy
2493                  * for bridge processing; return the original packet for
2494                  * local processing.
2495                  */
2496                 mc = m_dup(m, M_NOWAIT);
2497                 if (mc == NULL) {
2498                         return (m);
2499                 }
2500
2501                 /* Perform the bridge forwarding function with the copy. */
2502                 bridge_forward(sc, bif, mc);
2503
2504                 /*
2505                  * Reinject the mbuf as arriving on the bridge so we have a
2506                  * chance at claiming multicast packets. We can not loop back
2507                  * here from ether_input as a bridge is never a member of a
2508                  * bridge.
2509                  */
2510                 KASSERT(bifp->if_bridge == NULL,
2511                     ("loop created in bridge_input"));
2512                 mc2 = m_dup(m, M_NOWAIT);
2513                 if (mc2 != NULL) {
2514                         /* Keep the layer3 header aligned */
2515                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2516                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2517                 }
2518                 if (mc2 != NULL) {
2519                         mc2->m_pkthdr.rcvif = bifp;
2520                         (*bifp->if_input)(bifp, mc2);
2521                 }
2522
2523                 /* Return the original packet for local processing. */
2524                 return (m);
2525         }
2526
2527         if ((bif->bif_flags & IFBIF_STP) &&
2528             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2529                 return (m);
2530         }
2531
2532 #if defined(INET) || defined(INET6)
2533 #define CARP_CHECK_WE_ARE_DST(iface) \
2534         ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost))
2535 #define CARP_CHECK_WE_ARE_SRC(iface) \
2536         ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost))
2537 #else
2538 #define CARP_CHECK_WE_ARE_DST(iface)    false
2539 #define CARP_CHECK_WE_ARE_SRC(iface)    false
2540 #endif
2541
2542 #ifdef INET6
2543 #define PFIL_HOOKED_INET6       PFIL_HOOKED_IN(V_inet6_pfil_head)
2544 #else
2545 #define PFIL_HOOKED_INET6       false
2546 #endif
2547
2548 #define GRAB_OUR_PACKETS(iface)                                         \
2549         if ((iface)->if_type == IFT_GIF)                                \
2550                 continue;                                               \
2551         /* It is destined for us. */                                    \
2552         if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \
2553             CARP_CHECK_WE_ARE_DST(iface)) {                             \
2554                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2555                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2556                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2557                         if (error && bif->bif_addrmax) {                \
2558                                 m_freem(m);                             \
2559                                 return (NULL);                          \
2560                         }                                               \
2561                 }                                                       \
2562                 m->m_pkthdr.rcvif = iface;                              \
2563                 if ((iface) == ifp) {                                   \
2564                         /* Skip bridge processing... src == dest */     \
2565                         return (m);                                     \
2566                 }                                                       \
2567                 /* It's passing over or to the bridge, locally. */      \
2568                 ETHER_BPF_MTAP(bifp, m);                                \
2569                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2570                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); \
2571                 /* Filter on the physical interface. */                 \
2572                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) || \
2573                     PFIL_HOOKED_INET6)) {                               \
2574                         if (bridge_pfil(&m, NULL, ifp,                  \
2575                             PFIL_IN) != 0 || m == NULL) {               \
2576                                 return (NULL);                          \
2577                         }                                               \
2578                 }                                                       \
2579                 if ((iface) != bifp)                                    \
2580                         ETHER_BPF_MTAP(iface, m);                       \
2581                 return (m);                                             \
2582         }                                                               \
2583                                                                         \
2584         /* We just received a packet that we sent out. */               \
2585         if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \
2586             CARP_CHECK_WE_ARE_SRC(iface)) {                             \
2587                 m_freem(m);                                             \
2588                 return (NULL);                                          \
2589         }
2590
2591         /*
2592          * Unicast.  Make sure it's not for the bridge.
2593          */
2594         do { GRAB_OUR_PACKETS(bifp) } while (0);
2595
2596         /*
2597          * Give a chance for ifp at first priority. This will help when the
2598          * packet comes through the interface like VLAN's with the same MACs
2599          * on several interfaces from the same bridge. This also will save
2600          * some CPU cycles in case the destination interface and the input
2601          * interface (eq ifp) are the same.
2602          */
2603         do { GRAB_OUR_PACKETS(ifp) } while (0);
2604
2605         /* Now check the all bridge members. */
2606         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2607                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2608         }
2609
2610 #undef CARP_CHECK_WE_ARE_DST
2611 #undef CARP_CHECK_WE_ARE_SRC
2612 #undef PFIL_HOOKED_INET6
2613 #undef GRAB_OUR_PACKETS
2614
2615         /* Perform the bridge forwarding function. */
2616         bridge_forward(sc, bif, m);
2617
2618         return (NULL);
2619 }
2620
2621 /*
2622  * bridge_broadcast:
2623  *
2624  *      Send a frame to all interfaces that are members of
2625  *      the bridge, except for the one on which the packet
2626  *      arrived.
2627  *
2628  *      NOTE: Releases the lock on return.
2629  */
2630 static void
2631 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2632     struct mbuf *m, int runfilt)
2633 {
2634         struct bridge_iflist *dbif, *sbif;
2635         struct mbuf *mc;
2636         struct ifnet *dst_if;
2637         int used = 0, i;
2638
2639         NET_EPOCH_ASSERT();
2640
2641         sbif = bridge_lookup_member_if(sc, src_if);
2642
2643         /* Filter on the bridge interface before broadcasting */
2644         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2645 #ifdef INET6
2646             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2647 #endif
2648             )) {
2649                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2650                         return;
2651                 if (m == NULL)
2652                         return;
2653         }
2654
2655         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2656                 dst_if = dbif->bif_ifp;
2657                 if (dst_if == src_if)
2658                         continue;
2659
2660                 /* Private segments can not talk to each other */
2661                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2662                         continue;
2663
2664                 if ((dbif->bif_flags & IFBIF_STP) &&
2665                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2666                         continue;
2667
2668                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2669                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2670                         continue;
2671
2672                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2673                         continue;
2674
2675                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2676                         mc = m;
2677                         used = 1;
2678                 } else {
2679                         mc = m_dup(m, M_NOWAIT);
2680                         if (mc == NULL) {
2681                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2682                                 continue;
2683                         }
2684                 }
2685
2686                 /*
2687                  * Filter on the output interface. Pass a NULL bridge interface
2688                  * pointer so we do not redundantly filter on the bridge for
2689                  * each interface we broadcast on.
2690                  */
2691                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2692 #ifdef INET6
2693                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2694 #endif
2695                     )) {
2696                         if (used == 0) {
2697                                 /* Keep the layer3 header aligned */
2698                                 i = min(mc->m_pkthdr.len, max_protohdr);
2699                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2700                                 if (mc == NULL) {
2701                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2702                                         continue;
2703                                 }
2704                         }
2705                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2706                                 continue;
2707                         if (mc == NULL)
2708                                 continue;
2709                 }
2710
2711                 bridge_enqueue(sc, dst_if, mc);
2712         }
2713         if (used == 0)
2714                 m_freem(m);
2715 }
2716
2717 /*
2718  * bridge_span:
2719  *
2720  *      Duplicate a packet out one or more interfaces that are in span mode,
2721  *      the original mbuf is unmodified.
2722  */
2723 static void
2724 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2725 {
2726         struct bridge_iflist *bif;
2727         struct ifnet *dst_if;
2728         struct mbuf *mc;
2729
2730         NET_EPOCH_ASSERT();
2731
2732         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2733                 return;
2734
2735         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2736                 dst_if = bif->bif_ifp;
2737
2738                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2739                         continue;
2740
2741                 mc = m_dup(m, M_NOWAIT);
2742                 if (mc == NULL) {
2743                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2744                         continue;
2745                 }
2746
2747                 bridge_enqueue(sc, dst_if, mc);
2748         }
2749 }
2750
2751 /*
2752  * bridge_rtupdate:
2753  *
2754  *      Add a bridge routing entry.
2755  */
2756 static int
2757 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2758     struct bridge_iflist *bif, int setflags, uint8_t flags)
2759 {
2760         struct bridge_rtnode *brt;
2761         struct bridge_iflist *obif;
2762         int error;
2763
2764         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2765
2766         /* Check the source address is valid and not multicast. */
2767         if (ETHER_IS_MULTICAST(dst) ||
2768             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2769              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2770                 return (EINVAL);
2771
2772         /* 802.1p frames map to vlan 1 */
2773         if (vlan == 0)
2774                 vlan = 1;
2775
2776         /*
2777          * A route for this destination might already exist.  If so,
2778          * update it, otherwise create a new one.
2779          */
2780         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2781                 BRIDGE_RT_LOCK(sc);
2782
2783                 /* Check again, now that we have the lock. There could have
2784                  * been a race and we only want to insert this once. */
2785                 if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) {
2786                         BRIDGE_RT_UNLOCK(sc);
2787                         return (0);
2788                 }
2789
2790                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2791                         sc->sc_brtexceeded++;
2792                         BRIDGE_RT_UNLOCK(sc);
2793                         return (ENOSPC);
2794                 }
2795                 /* Check per interface address limits (if enabled) */
2796                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2797                         bif->bif_addrexceeded++;
2798                         BRIDGE_RT_UNLOCK(sc);
2799                         return (ENOSPC);
2800                 }
2801
2802                 /*
2803                  * Allocate a new bridge forwarding node, and
2804                  * initialize the expiration time and Ethernet
2805                  * address.
2806                  */
2807                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2808                 if (brt == NULL) {
2809                         BRIDGE_RT_UNLOCK(sc);
2810                         return (ENOMEM);
2811                 }
2812                 brt->brt_vnet = curvnet;
2813
2814                 if (bif->bif_flags & IFBIF_STICKY)
2815                         brt->brt_flags = IFBAF_STICKY;
2816                 else
2817                         brt->brt_flags = IFBAF_DYNAMIC;
2818
2819                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2820                 brt->brt_vlan = vlan;
2821
2822                 brt->brt_dst = bif;
2823                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2824                         uma_zfree(V_bridge_rtnode_zone, brt);
2825                         BRIDGE_RT_UNLOCK(sc);
2826                         return (error);
2827                 }
2828                 bif->bif_addrcnt++;
2829
2830                 BRIDGE_RT_UNLOCK(sc);
2831         }
2832
2833         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2834             (obif = brt->brt_dst) != bif) {
2835                 MPASS(obif != NULL);
2836
2837                 BRIDGE_RT_LOCK(sc);
2838                 brt->brt_dst->bif_addrcnt--;
2839                 brt->brt_dst = bif;
2840                 brt->brt_dst->bif_addrcnt++;
2841                 BRIDGE_RT_UNLOCK(sc);
2842
2843                 if (V_log_mac_flap &&
2844                     ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
2845                         log(LOG_NOTICE,
2846                             "%s: mac address %6D vlan %d moved from %s to %s\n",
2847                             sc->sc_ifp->if_xname,
2848                             &brt->brt_addr[0], ":",
2849                             brt->brt_vlan,
2850                             obif->bif_ifp->if_xname,
2851                             bif->bif_ifp->if_xname);
2852                 }
2853         }
2854
2855         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2856                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2857         if (setflags)
2858                 brt->brt_flags = flags;
2859
2860         return (0);
2861 }
2862
2863 /*
2864  * bridge_rtlookup:
2865  *
2866  *      Lookup the destination interface for an address.
2867  */
2868 static struct ifnet *
2869 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2870 {
2871         struct bridge_rtnode *brt;
2872
2873         NET_EPOCH_ASSERT();
2874
2875         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2876                 return (NULL);
2877
2878         return (brt->brt_ifp);
2879 }
2880
2881 /*
2882  * bridge_rttrim:
2883  *
2884  *      Trim the routine table so that we have a number
2885  *      of routing entries less than or equal to the
2886  *      maximum number.
2887  */
2888 static void
2889 bridge_rttrim(struct bridge_softc *sc)
2890 {
2891         struct bridge_rtnode *brt, *nbrt;
2892
2893         NET_EPOCH_ASSERT();
2894         BRIDGE_RT_LOCK_ASSERT(sc);
2895
2896         /* Make sure we actually need to do this. */
2897         if (sc->sc_brtcnt <= sc->sc_brtmax)
2898                 return;
2899
2900         /* Force an aging cycle; this might trim enough addresses. */
2901         bridge_rtage(sc);
2902         if (sc->sc_brtcnt <= sc->sc_brtmax)
2903                 return;
2904
2905         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2906                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2907                         bridge_rtnode_destroy(sc, brt);
2908                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2909                                 return;
2910                 }
2911         }
2912 }
2913
2914 /*
2915  * bridge_timer:
2916  *
2917  *      Aging timer for the bridge.
2918  */
2919 static void
2920 bridge_timer(void *arg)
2921 {
2922         struct bridge_softc *sc = arg;
2923
2924         BRIDGE_RT_LOCK_ASSERT(sc);
2925
2926         /* Destruction of rtnodes requires a proper vnet context */
2927         CURVNET_SET(sc->sc_ifp->if_vnet);
2928         bridge_rtage(sc);
2929
2930         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2931                 callout_reset(&sc->sc_brcallout,
2932                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2933         CURVNET_RESTORE();
2934 }
2935
2936 /*
2937  * bridge_rtage:
2938  *
2939  *      Perform an aging cycle.
2940  */
2941 static void
2942 bridge_rtage(struct bridge_softc *sc)
2943 {
2944         struct bridge_rtnode *brt, *nbrt;
2945
2946         BRIDGE_RT_LOCK_ASSERT(sc);
2947
2948         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2949                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2950                         if (time_uptime >= brt->brt_expire)
2951                                 bridge_rtnode_destroy(sc, brt);
2952                 }
2953         }
2954 }
2955
2956 /*
2957  * bridge_rtflush:
2958  *
2959  *      Remove all dynamic addresses from the bridge.
2960  */
2961 static void
2962 bridge_rtflush(struct bridge_softc *sc, int full)
2963 {
2964         struct bridge_rtnode *brt, *nbrt;
2965
2966         BRIDGE_RT_LOCK_ASSERT(sc);
2967
2968         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2969                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2970                         bridge_rtnode_destroy(sc, brt);
2971         }
2972 }
2973
2974 /*
2975  * bridge_rtdaddr:
2976  *
2977  *      Remove an address from the table.
2978  */
2979 static int
2980 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2981 {
2982         struct bridge_rtnode *brt;
2983         int found = 0;
2984
2985         BRIDGE_RT_LOCK(sc);
2986
2987         /*
2988          * If vlan is zero then we want to delete for all vlans so the lookup
2989          * may return more than one.
2990          */
2991         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2992                 bridge_rtnode_destroy(sc, brt);
2993                 found = 1;
2994         }
2995
2996         BRIDGE_RT_UNLOCK(sc);
2997
2998         return (found ? 0 : ENOENT);
2999 }
3000
3001 /*
3002  * bridge_rtdelete:
3003  *
3004  *      Delete routes to a speicifc member interface.
3005  */
3006 static void
3007 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3008 {
3009         struct bridge_rtnode *brt, *nbrt;
3010
3011         BRIDGE_RT_LOCK_ASSERT(sc);
3012
3013         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3014                 if (brt->brt_ifp == ifp && (full ||
3015                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3016                         bridge_rtnode_destroy(sc, brt);
3017         }
3018 }
3019
3020 /*
3021  * bridge_rtable_init:
3022  *
3023  *      Initialize the route table for this bridge.
3024  */
3025 static void
3026 bridge_rtable_init(struct bridge_softc *sc)
3027 {
3028         int i;
3029
3030         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3031             M_DEVBUF, M_WAITOK);
3032
3033         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3034                 CK_LIST_INIT(&sc->sc_rthash[i]);
3035
3036         sc->sc_rthash_key = arc4random();
3037         CK_LIST_INIT(&sc->sc_rtlist);
3038 }
3039
3040 /*
3041  * bridge_rtable_fini:
3042  *
3043  *      Deconstruct the route table for this bridge.
3044  */
3045 static void
3046 bridge_rtable_fini(struct bridge_softc *sc)
3047 {
3048
3049         KASSERT(sc->sc_brtcnt == 0,
3050             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3051         free(sc->sc_rthash, M_DEVBUF);
3052 }
3053
3054 /*
3055  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3056  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3057  */
3058 #define mix(a, b, c)                                                    \
3059 do {                                                                    \
3060         a -= b; a -= c; a ^= (c >> 13);                                 \
3061         b -= c; b -= a; b ^= (a << 8);                                  \
3062         c -= a; c -= b; c ^= (b >> 13);                                 \
3063         a -= b; a -= c; a ^= (c >> 12);                                 \
3064         b -= c; b -= a; b ^= (a << 16);                                 \
3065         c -= a; c -= b; c ^= (b >> 5);                                  \
3066         a -= b; a -= c; a ^= (c >> 3);                                  \
3067         b -= c; b -= a; b ^= (a << 10);                                 \
3068         c -= a; c -= b; c ^= (b >> 15);                                 \
3069 } while (/*CONSTCOND*/0)
3070
3071 static __inline uint32_t
3072 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3073 {
3074         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3075
3076         b += addr[5] << 8;
3077         b += addr[4];
3078         a += addr[3] << 24;
3079         a += addr[2] << 16;
3080         a += addr[1] << 8;
3081         a += addr[0];
3082
3083         mix(a, b, c);
3084
3085         return (c & BRIDGE_RTHASH_MASK);
3086 }
3087
3088 #undef mix
3089
3090 static int
3091 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3092 {
3093         int i, d;
3094
3095         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3096                 d = ((int)a[i]) - ((int)b[i]);
3097         }
3098
3099         return (d);
3100 }
3101
3102 /*
3103  * bridge_rtnode_lookup:
3104  *
3105  *      Look up a bridge route node for the specified destination. Compare the
3106  *      vlan id or if zero then just return the first match.
3107  */
3108 static struct bridge_rtnode *
3109 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3110 {
3111         struct bridge_rtnode *brt;
3112         uint32_t hash;
3113         int dir;
3114
3115         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3116
3117         hash = bridge_rthash(sc, addr);
3118         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3119                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3120                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3121                         return (brt);
3122                 if (dir > 0)
3123                         return (NULL);
3124         }
3125
3126         return (NULL);
3127 }
3128
3129 /*
3130  * bridge_rtnode_insert:
3131  *
3132  *      Insert the specified bridge node into the route table.  We
3133  *      assume the entry is not already in the table.
3134  */
3135 static int
3136 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3137 {
3138         struct bridge_rtnode *lbrt;
3139         uint32_t hash;
3140         int dir;
3141
3142         BRIDGE_RT_LOCK_ASSERT(sc);
3143
3144         hash = bridge_rthash(sc, brt->brt_addr);
3145
3146         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3147         if (lbrt == NULL) {
3148                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3149                 goto out;
3150         }
3151
3152         do {
3153                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3154                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3155                         return (EEXIST);
3156                 if (dir > 0) {
3157                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3158                         goto out;
3159                 }
3160                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3161                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3162                         goto out;
3163                 }
3164                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3165         } while (lbrt != NULL);
3166
3167 #ifdef DIAGNOSTIC
3168         panic("bridge_rtnode_insert: impossible");
3169 #endif
3170
3171 out:
3172         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3173         sc->sc_brtcnt++;
3174
3175         return (0);
3176 }
3177
3178 static void
3179 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3180 {
3181         struct bridge_rtnode *brt;
3182
3183         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3184
3185         CURVNET_SET(brt->brt_vnet);
3186         uma_zfree(V_bridge_rtnode_zone, brt);
3187         CURVNET_RESTORE();
3188 }
3189
3190 /*
3191  * bridge_rtnode_destroy:
3192  *
3193  *      Destroy a bridge rtnode.
3194  */
3195 static void
3196 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3197 {
3198         BRIDGE_RT_LOCK_ASSERT(sc);
3199
3200         CK_LIST_REMOVE(brt, brt_hash);
3201
3202         CK_LIST_REMOVE(brt, brt_list);
3203         sc->sc_brtcnt--;
3204         brt->brt_dst->bif_addrcnt--;
3205
3206         NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3207 }
3208
3209 /*
3210  * bridge_rtable_expire:
3211  *
3212  *      Set the expiry time for all routes on an interface.
3213  */
3214 static void
3215 bridge_rtable_expire(struct ifnet *ifp, int age)
3216 {
3217         struct bridge_softc *sc = ifp->if_bridge;
3218         struct bridge_rtnode *brt;
3219
3220         CURVNET_SET(ifp->if_vnet);
3221         BRIDGE_RT_LOCK(sc);
3222
3223         /*
3224          * If the age is zero then flush, otherwise set all the expiry times to
3225          * age for the interface
3226          */
3227         if (age == 0)
3228                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3229         else {
3230                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3231                         /* Cap the expiry time to 'age' */
3232                         if (brt->brt_ifp == ifp &&
3233                             brt->brt_expire > time_uptime + age &&
3234                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3235                                 brt->brt_expire = time_uptime + age;
3236                 }
3237         }
3238         BRIDGE_RT_UNLOCK(sc);
3239         CURVNET_RESTORE();
3240 }
3241
3242 /*
3243  * bridge_state_change:
3244  *
3245  *      Callback from the bridgestp code when a port changes states.
3246  */
3247 static void
3248 bridge_state_change(struct ifnet *ifp, int state)
3249 {
3250         struct bridge_softc *sc = ifp->if_bridge;
3251         static const char *stpstates[] = {
3252                 "disabled",
3253                 "listening",
3254                 "learning",
3255                 "forwarding",
3256                 "blocking",
3257                 "discarding"
3258         };
3259
3260         CURVNET_SET(ifp->if_vnet);
3261         if (V_log_stp)
3262                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3263                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3264         CURVNET_RESTORE();
3265 }
3266
3267 /*
3268  * Send bridge packets through pfil if they are one of the types pfil can deal
3269  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3270  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3271  * that interface.
3272  */
3273 static int
3274 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3275 {
3276         int snap, error, i, hlen;
3277         struct ether_header *eh1, eh2;
3278         struct ip *ip;
3279         struct llc llc1;
3280         u_int16_t ether_type;
3281         pfil_return_t rv;
3282
3283         snap = 0;
3284         error = -1;     /* Default error if not error == 0 */
3285
3286 #if 0
3287         /* we may return with the IP fields swapped, ensure its not shared */
3288         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3289 #endif
3290
3291         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3292                 return (0); /* filtering is disabled */
3293
3294         i = min((*mp)->m_pkthdr.len, max_protohdr);
3295         if ((*mp)->m_len < i) {
3296             *mp = m_pullup(*mp, i);
3297             if (*mp == NULL) {
3298                 printf("%s: m_pullup failed\n", __func__);
3299                 return (-1);
3300             }
3301         }
3302
3303         eh1 = mtod(*mp, struct ether_header *);
3304         ether_type = ntohs(eh1->ether_type);
3305
3306         /*
3307          * Check for SNAP/LLC.
3308          */
3309         if (ether_type < ETHERMTU) {
3310                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3311
3312                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3313                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3314                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3315                     llc2->llc_control == LLC_UI) {
3316                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3317                         snap = 1;
3318                 }
3319         }
3320
3321         /*
3322          * If we're trying to filter bridge traffic, don't look at anything
3323          * other than IP and ARP traffic.  If the filter doesn't understand
3324          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3325          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3326          * but of course we don't have an AppleTalk filter to begin with.
3327          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3328          * ARP traffic.)
3329          */
3330         switch (ether_type) {
3331                 case ETHERTYPE_ARP:
3332                 case ETHERTYPE_REVARP:
3333                         if (V_pfil_ipfw_arp == 0)
3334                                 return (0); /* Automatically pass */
3335                         break;
3336
3337                 case ETHERTYPE_IP:
3338 #ifdef INET6
3339                 case ETHERTYPE_IPV6:
3340 #endif /* INET6 */
3341                         break;
3342                 default:
3343                         /*
3344                          * Check to see if the user wants to pass non-ip
3345                          * packets, these will not be checked by pfil(9) and
3346                          * passed unconditionally so the default is to drop.
3347                          */
3348                         if (V_pfil_onlyip)
3349                                 goto bad;
3350         }
3351
3352         /* Run the packet through pfil before stripping link headers */
3353         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3354             dir == PFIL_OUT && ifp != NULL) {
3355                 switch (pfil_run_hooks(V_link_pfil_head, mp, ifp, dir, NULL)) {
3356                 case PFIL_DROPPED:
3357                         return (EACCES);
3358                 case PFIL_CONSUMED:
3359                         return (0);
3360                 }
3361         }
3362
3363         /* Strip off the Ethernet header and keep a copy. */
3364         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3365         m_adj(*mp, ETHER_HDR_LEN);
3366
3367         /* Strip off snap header, if present */
3368         if (snap) {
3369                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3370                 m_adj(*mp, sizeof(struct llc));
3371         }
3372
3373         /*
3374          * Check the IP header for alignment and errors
3375          */
3376         if (dir == PFIL_IN) {
3377                 switch (ether_type) {
3378                         case ETHERTYPE_IP:
3379                                 error = bridge_ip_checkbasic(mp);
3380                                 break;
3381 #ifdef INET6
3382                         case ETHERTYPE_IPV6:
3383                                 error = bridge_ip6_checkbasic(mp);
3384                                 break;
3385 #endif /* INET6 */
3386                         default:
3387                                 error = 0;
3388                 }
3389                 if (error)
3390                         goto bad;
3391         }
3392
3393         error = 0;
3394
3395         /*
3396          * Run the packet through pfil
3397          */
3398         rv = PFIL_PASS;
3399         switch (ether_type) {
3400         case ETHERTYPE_IP:
3401                 /*
3402                  * Run pfil on the member interface and the bridge, both can
3403                  * be skipped by clearing pfil_member or pfil_bridge.
3404                  *
3405                  * Keep the order:
3406                  *   in_if -> bridge_if -> out_if
3407                  */
3408                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3409                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3410                     PFIL_PASS)
3411                         break;
3412
3413                 if (V_pfil_member && ifp != NULL && (rv =
3414                     pfil_run_hooks(V_inet_pfil_head, mp, ifp, dir, NULL)) !=
3415                     PFIL_PASS)
3416                         break;
3417
3418                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3419                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3420                     PFIL_PASS)
3421                         break;
3422
3423                 /* check if we need to fragment the packet */
3424                 /* bridge_fragment generates a mbuf chain of packets */
3425                 /* that already include eth headers */
3426                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3427                         i = (*mp)->m_pkthdr.len;
3428                         if (i > ifp->if_mtu) {
3429                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3430                                             &llc1);
3431                                 return (error);
3432                         }
3433                 }
3434
3435                 /* Recalculate the ip checksum. */
3436                 ip = mtod(*mp, struct ip *);
3437                 hlen = ip->ip_hl << 2;
3438                 if (hlen < sizeof(struct ip))
3439                         goto bad;
3440                 if (hlen > (*mp)->m_len) {
3441                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3442                                 goto bad;
3443                         ip = mtod(*mp, struct ip *);
3444                         if (ip == NULL)
3445                                 goto bad;
3446                 }
3447                 ip->ip_sum = 0;
3448                 if (hlen == sizeof(struct ip))
3449                         ip->ip_sum = in_cksum_hdr(ip);
3450                 else
3451                         ip->ip_sum = in_cksum(*mp, hlen);
3452
3453                 break;
3454 #ifdef INET6
3455         case ETHERTYPE_IPV6:
3456                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3457                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3458                     PFIL_PASS)
3459                         break;
3460
3461                 if (V_pfil_member && ifp != NULL && (rv =
3462                     pfil_run_hooks(V_inet6_pfil_head, mp, ifp, dir, NULL)) !=
3463                     PFIL_PASS)
3464                         break;
3465
3466                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3467                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3468                     PFIL_PASS)
3469                         break;
3470                 break;
3471 #endif
3472         }
3473
3474         switch (rv) {
3475         case PFIL_CONSUMED:
3476                 return (0);
3477         case PFIL_DROPPED:
3478                 return (EACCES);
3479         default:
3480                 break;
3481         }
3482
3483         error = -1;
3484
3485         /*
3486          * Finally, put everything back the way it was and return
3487          */
3488         if (snap) {
3489                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3490                 if (*mp == NULL)
3491                         return (error);
3492                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3493         }
3494
3495         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3496         if (*mp == NULL)
3497                 return (error);
3498         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3499
3500         return (0);
3501
3502 bad:
3503         m_freem(*mp);
3504         *mp = NULL;
3505         return (error);
3506 }
3507
3508 /*
3509  * Perform basic checks on header size since
3510  * pfil assumes ip_input has already processed
3511  * it for it.  Cut-and-pasted from ip_input.c.
3512  * Given how simple the IPv6 version is,
3513  * does the IPv4 version really need to be
3514  * this complicated?
3515  *
3516  * XXX Should we update ipstat here, or not?
3517  * XXX Right now we update ipstat but not
3518  * XXX csum_counter.
3519  */
3520 static int
3521 bridge_ip_checkbasic(struct mbuf **mp)
3522 {
3523         struct mbuf *m = *mp;
3524         struct ip *ip;
3525         int len, hlen;
3526         u_short sum;
3527
3528         if (*mp == NULL)
3529                 return (-1);
3530
3531         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3532                 if ((m = m_copyup(m, sizeof(struct ip),
3533                         (max_linkhdr + 3) & ~3)) == NULL) {
3534                         /* XXXJRT new stat, please */
3535                         KMOD_IPSTAT_INC(ips_toosmall);
3536                         goto bad;
3537                 }
3538         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3539                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3540                         KMOD_IPSTAT_INC(ips_toosmall);
3541                         goto bad;
3542                 }
3543         }
3544         ip = mtod(m, struct ip *);
3545         if (ip == NULL) goto bad;
3546
3547         if (ip->ip_v != IPVERSION) {
3548                 KMOD_IPSTAT_INC(ips_badvers);
3549                 goto bad;
3550         }
3551         hlen = ip->ip_hl << 2;
3552         if (hlen < sizeof(struct ip)) { /* minimum header length */
3553                 KMOD_IPSTAT_INC(ips_badhlen);
3554                 goto bad;
3555         }
3556         if (hlen > m->m_len) {
3557                 if ((m = m_pullup(m, hlen)) == NULL) {
3558                         KMOD_IPSTAT_INC(ips_badhlen);
3559                         goto bad;
3560                 }
3561                 ip = mtod(m, struct ip *);
3562                 if (ip == NULL) goto bad;
3563         }
3564
3565         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3566                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3567         } else {
3568                 if (hlen == sizeof(struct ip)) {
3569                         sum = in_cksum_hdr(ip);
3570                 } else {
3571                         sum = in_cksum(m, hlen);
3572                 }
3573         }
3574         if (sum) {
3575                 KMOD_IPSTAT_INC(ips_badsum);
3576                 goto bad;
3577         }
3578
3579         /* Retrieve the packet length. */
3580         len = ntohs(ip->ip_len);
3581
3582         /*
3583          * Check for additional length bogosity
3584          */
3585         if (len < hlen) {
3586                 KMOD_IPSTAT_INC(ips_badlen);
3587                 goto bad;
3588         }
3589
3590         /*
3591          * Check that the amount of data in the buffers
3592          * is as at least much as the IP header would have us expect.
3593          * Drop packet if shorter than we expect.
3594          */
3595         if (m->m_pkthdr.len < len) {
3596                 KMOD_IPSTAT_INC(ips_tooshort);
3597                 goto bad;
3598         }
3599
3600         /* Checks out, proceed */
3601         *mp = m;
3602         return (0);
3603
3604 bad:
3605         *mp = m;
3606         return (-1);
3607 }
3608
3609 #ifdef INET6
3610 /*
3611  * Same as above, but for IPv6.
3612  * Cut-and-pasted from ip6_input.c.
3613  * XXX Should we update ip6stat, or not?
3614  */
3615 static int
3616 bridge_ip6_checkbasic(struct mbuf **mp)
3617 {
3618         struct mbuf *m = *mp;
3619         struct ip6_hdr *ip6;
3620
3621         /*
3622          * If the IPv6 header is not aligned, slurp it up into a new
3623          * mbuf with space for link headers, in the event we forward
3624          * it.  Otherwise, if it is aligned, make sure the entire base
3625          * IPv6 header is in the first mbuf of the chain.
3626          */
3627         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3628                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3629                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3630                             (max_linkhdr + 3) & ~3)) == NULL) {
3631                         /* XXXJRT new stat, please */
3632                         IP6STAT_INC(ip6s_toosmall);
3633                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3634                         goto bad;
3635                 }
3636         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3637                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3638                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3639                         IP6STAT_INC(ip6s_toosmall);
3640                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3641                         goto bad;
3642                 }
3643         }
3644
3645         ip6 = mtod(m, struct ip6_hdr *);
3646
3647         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3648                 IP6STAT_INC(ip6s_badvers);
3649                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3650                 goto bad;
3651         }
3652
3653         /* Checks out, proceed */
3654         *mp = m;
3655         return (0);
3656
3657 bad:
3658         *mp = m;
3659         return (-1);
3660 }
3661 #endif /* INET6 */
3662
3663 /*
3664  * bridge_fragment:
3665  *
3666  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3667  */
3668 static int
3669 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3670     int snap, struct llc *llc)
3671 {
3672         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3673         struct ip *ip;
3674         int error = -1;
3675
3676         if (m->m_len < sizeof(struct ip) &&
3677             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3678                 goto dropit;
3679         ip = mtod(m, struct ip *);
3680
3681         m->m_pkthdr.csum_flags |= CSUM_IP;
3682         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3683         if (error)
3684                 goto dropit;
3685
3686         /*
3687          * Walk the chain and re-add the Ethernet header for
3688          * each mbuf packet.
3689          */
3690         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3691                 nextpkt = mcur->m_nextpkt;
3692                 mcur->m_nextpkt = NULL;
3693                 if (snap) {
3694                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3695                         if (mcur == NULL) {
3696                                 error = ENOBUFS;
3697                                 if (mprev != NULL)
3698                                         mprev->m_nextpkt = nextpkt;
3699                                 goto dropit;
3700                         }
3701                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3702                 }
3703
3704                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3705                 if (mcur == NULL) {
3706                         error = ENOBUFS;
3707                         if (mprev != NULL)
3708                                 mprev->m_nextpkt = nextpkt;
3709                         goto dropit;
3710                 }
3711                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3712
3713                 /*
3714                  * The previous two M_PREPEND could have inserted one or two
3715                  * mbufs in front so we have to update the previous packet's
3716                  * m_nextpkt.
3717                  */
3718                 mcur->m_nextpkt = nextpkt;
3719                 if (mprev != NULL)
3720                         mprev->m_nextpkt = mcur;
3721                 else {
3722                         /* The first mbuf in the original chain needs to be
3723                          * updated. */
3724                         *mp = mcur;
3725                 }
3726                 mprev = mcur;
3727         }
3728
3729         KMOD_IPSTAT_INC(ips_fragmented);
3730         return (error);
3731
3732 dropit:
3733         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3734                 m = mcur->m_nextpkt;
3735                 m_freem(mcur);
3736         }
3737         return (error);
3738 }
3739
3740 static void
3741 bridge_linkstate(struct ifnet *ifp)
3742 {
3743         struct bridge_softc *sc = ifp->if_bridge;
3744         struct bridge_iflist *bif;
3745         struct epoch_tracker et;
3746
3747         NET_EPOCH_ENTER(et);
3748
3749         bif = bridge_lookup_member_if(sc, ifp);
3750         if (bif == NULL) {
3751                 NET_EPOCH_EXIT(et);
3752                 return;
3753         }
3754         bridge_linkcheck(sc);
3755
3756         bstp_linkstate(&bif->bif_stp);
3757
3758         NET_EPOCH_EXIT(et);
3759 }
3760
3761 static void
3762 bridge_linkcheck(struct bridge_softc *sc)
3763 {
3764         struct bridge_iflist *bif;
3765         int new_link, hasls;
3766
3767         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3768
3769         new_link = LINK_STATE_DOWN;
3770         hasls = 0;
3771         /* Our link is considered up if at least one of our ports is active */
3772         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3773                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3774                         hasls++;
3775                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3776                         new_link = LINK_STATE_UP;
3777                         break;
3778                 }
3779         }
3780         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3781                 /* If no interfaces support link-state then we default to up */
3782                 new_link = LINK_STATE_UP;
3783         }
3784         if_link_state_change(sc->sc_ifp, new_link);
3785 }