sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*-
   4  * SPDX-License-Identifier: BSD-4-Clause
   5  *
   6  * Copyright 2001 Wasabi Systems, Inc.
   7  * All rights reserved.
   8  *
   9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed for the NetBSD Project by
  22  *      Wasabi Systems, Inc.
  23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  24  *    or promote products derived from this software without specific prior
  25  *    written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 /*
  41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  42  * All rights reserved.
  43  *
  44  * Redistribution and use in source and binary forms, with or without
  45  * modification, are permitted provided that the following conditions
  46  * are met:
  47  * 1. Redistributions of source code must retain the above copyright
  48  *    notice, this list of conditions and the following disclaimer.
  49  * 2. Redistributions in binary form must reproduce the above copyright
  50  *    notice, this list of conditions and the following disclaimer in the
  51  *    documentation and/or other materials provided with the distribution.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  63  * POSSIBILITY OF SUCH DAMAGE.
  64  *
  65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  66  */
  67
  68 /*
  69  * Network interface bridge support.
  70  *
  71  * TODO:
  72  *
  73  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  74  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  75  *        to bridge other types of interfaces (maybe consider
  76  *        heterogeneous bridges).
  77  */
  78
  79 #include <sys/cdefs.h>
  80 __FBSDID("$FreeBSD$");
  81
  82 #include "opt_inet.h"
  83 #include "opt_inet6.h"
  84
  85 #include <sys/param.h>
  86 #include <sys/eventhandler.h>
  87 #include <sys/mbuf.h>
  88 #include <sys/malloc.h>
  89 #include <sys/protosw.h>
  90 #include <sys/systm.h>
  91 #include <sys/jail.h>
  92 #include <sys/time.h>
  93 #include <sys/socket.h> /* for net/if.h */
  94 #include <sys/sockio.h>
  95 #include <sys/ctype.h>  /* string functions */
  96 #include <sys/kernel.h>
  97 #include <sys/random.h>
  98 #include <sys/syslog.h>
  99 #include <sys/sysctl.h>
 100 #include <vm/uma.h>
 101 #include <sys/module.h>
 102 #include <sys/priv.h>
 103 #include <sys/proc.h>
 104 #include <sys/lock.h>
 105 #include <sys/mutex.h>
 106
 107 #include <net/bpf.h>
 108 #include <net/if.h>
 109 #include <net/if_clone.h>
 110 #include <net/if_dl.h>
 111 #include <net/if_types.h>
 112 #include <net/if_var.h>
 113 #include <net/pfil.h>
 114 #include <net/vnet.h>
 115
 116 #include <netinet/in.h>
 117 #include <netinet/in_systm.h>
 118 #include <netinet/in_var.h>
 119 #include <netinet/ip.h>
 120 #include <netinet/ip_var.h>
 121 #ifdef INET6
 122 #include <netinet/ip6.h>
 123 #include <netinet6/ip6_var.h>
 124 #include <netinet6/in6_ifattach.h>
 125 #endif
 126 #if defined(INET) || defined(INET6)
 127 #include <netinet/ip_carp.h>
 128 #endif
 129 #include <machine/in_cksum.h>
 130 #include <netinet/if_ether.h>
 131 #include <net/bridgestp.h>
 132 #include <net/if_bridgevar.h>
 133 #include <net/if_llc.h>
 134 #include <net/if_vlan_var.h>
 135
 136 #include <net/route.h>
 137
 138 #ifdef INET6
 139 /*
 140  * XXX: declare here to avoid to include many inet6 related files..
 141  * should be more generalized?
 142  */
 143 extern void     nd6_setmtu(struct ifnet *);
 144 #endif
 145
 146 /*
 147  * Size of the route hash table.  Must be a power of two.
 148  */
 149 #ifndef BRIDGE_RTHASH_SIZE
 150 #define BRIDGE_RTHASH_SIZE              1024
 151 #endif
 152
 153 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 154
 155 /*
 156  * Default maximum number of addresses to cache.
 157  */
 158 #ifndef BRIDGE_RTABLE_MAX
 159 #define BRIDGE_RTABLE_MAX               2000
 160 #endif
 161
 162 /*
 163  * Timeout (in seconds) for entries learned dynamically.
 164  */
 165 #ifndef BRIDGE_RTABLE_TIMEOUT
 166 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 167 #endif
 168
 169 /*
 170  * Number of seconds between walks of the route list.
 171  */
 172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 173 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 174 #endif
 175
 176 /*
 177  * List of capabilities to possibly mask on the member interface.
 178  */
 179 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 180                                          IFCAP_TXCSUM_IPV6)
 181
 182 /*
 183  * List of capabilities to strip
 184  */
 185 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 186
 187 /*
 188  * Bridge locking
 189  *
 190  * The bridge relies heavily on the epoch(9) system to protect its data
 191  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 192  * must ensure there is only one writer at a time.
 193  *
 194  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 195  * accesses we must hold:
 196  *
 197  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 198  *  - BRIDGE_LOCK, for any other change
 199  *
 200  * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
 201  * calls to bridge member interfaces and these ioctl()s can sleep.
 202  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 203  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 204  */
 205 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 206         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 207         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 208 } while (0)
 209 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 210         sx_destroy(&(_sc)->sc_sx);              \
 211         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 212 } while (0)
 213 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 214 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 215 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 216 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 217             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 218 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 219 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 220 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 221 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 222 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 223             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 224
 225 /*
 226  * Bridge interface list entry.
 227  */
 228 struct bridge_iflist {
 229         CK_LIST_ENTRY(bridge_iflist) bif_next;
 230         struct ifnet            *bif_ifp;       /* member if */
 231         struct bstp_port        bif_stp;        /* STP state */
 232         uint32_t                bif_flags;      /* member if flags */
 233         int                     bif_savedcaps;  /* saved capabilities */
 234         uint32_t                bif_addrmax;    /* max # of addresses */
 235         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 236         uint32_t                bif_addrexceeded;/* # of address violations */
 237         struct epoch_context    bif_epoch_ctx;
 238 };
 239
 240 /*
 241  * Bridge route node.
 242  */
 243 struct bridge_rtnode {
 244         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 245         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 246         struct bridge_iflist    *brt_dst;       /* destination if */
 247         unsigned long           brt_expire;     /* expiration time */
 248         uint8_t                 brt_flags;      /* address flags */
 249         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 250         uint16_t                brt_vlan;       /* vlan id */
 251         struct  vnet            *brt_vnet;
 252         struct  epoch_context   brt_epoch_ctx;
 253 };
 254 #define brt_ifp                 brt_dst->bif_ifp
 255
 256 /*
 257  * Software state for each bridge.
 258  */
 259 struct bridge_softc {
 260         struct ifnet            *sc_ifp;        /* make this an interface */
 261         LIST_ENTRY(bridge_softc) sc_list;
 262         struct sx               sc_sx;
 263         struct mtx              sc_rt_mtx;
 264         uint32_t                sc_brtmax;      /* max # of addresses */
 265         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 266         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 267         struct callout          sc_brcallout;   /* bridge callout */
 268         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 269         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 270         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 271         uint32_t                sc_rthash_key;  /* key for hash */
 272         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 273         struct bstp_state       sc_stp;         /* STP state */
 274         uint32_t                sc_brtexceeded; /* # of cache drops */
 275         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 276         struct ether_addr       sc_defaddr;     /* Default MAC address */
 277         struct epoch_context    sc_epoch_ctx;
 278 };
 279
 280 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 281 #define V_bridge_list_sx        VNET(bridge_list_sx)
 282 static eventhandler_tag bridge_detach_cookie;
 283
 284 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 285
 286 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 287 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 288
 289 static int      bridge_clone_create(struct if_clone *, char *, size_t,
 290                     struct ifc_data *, struct ifnet **);
 291 static int      bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
 292
 293 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 294 static void     bridge_mutecaps(struct bridge_softc *);
 295 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 296                     int);
 297 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 298 static void     bridge_init(void *);
 299 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 300 static void     bridge_stop(struct ifnet *, int);
 301 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 302 #ifdef ALTQ
 303 static void     bridge_altq_start(if_t);
 304 static int      bridge_altq_transmit(if_t, struct mbuf *);
 305 #endif
 306 static void     bridge_qflush(struct ifnet *);
 307 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 308 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 309                     struct rtentry *);
 310 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 311                     struct mbuf *);
 312 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 313
 314 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 315                     struct mbuf *m);
 316
 317 static void     bridge_timer(void *);
 318
 319 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 320                     struct mbuf *, int);
 321 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 322
 323 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 324                     uint16_t, struct bridge_iflist *, int, uint8_t);
 325 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 326                     uint16_t);
 327 static void     bridge_rttrim(struct bridge_softc *);
 328 static void     bridge_rtage(struct bridge_softc *);
 329 static void     bridge_rtflush(struct bridge_softc *, int);
 330 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 331                     uint16_t);
 332
 333 static void     bridge_rtable_init(struct bridge_softc *);
 334 static void     bridge_rtable_fini(struct bridge_softc *);
 335
 336 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 337 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 338                     const uint8_t *, uint16_t);
 339 static int      bridge_rtnode_insert(struct bridge_softc *,
 340                     struct bridge_rtnode *);
 341 static void     bridge_rtnode_destroy(struct bridge_softc *,
 342                     struct bridge_rtnode *);
 343 static void     bridge_rtable_expire(struct ifnet *, int);
 344 static void     bridge_state_change(struct ifnet *, int);
 345
 346 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 347                     const char *name);
 348 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 349                     struct ifnet *ifp);
 350 static void     bridge_delete_member(struct bridge_softc *,
 351                     struct bridge_iflist *, int);
 352 static void     bridge_delete_span(struct bridge_softc *,
 353                     struct bridge_iflist *);
 354
 355 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 381 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 382 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 383 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 384 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 385 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 386 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 387                     int);
 388 static int      bridge_ip_checkbasic(struct mbuf **mp);
 389 #ifdef INET6
 390 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 391 #endif /* INET6 */
 392 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 393                     struct ether_header *, int, struct llc *);
 394 static void     bridge_linkstate(struct ifnet *ifp);
 395 static void     bridge_linkcheck(struct bridge_softc *sc);
 396
 397 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 398 #define VLANTAGOF(_m)   \
 399     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 400
 401 static struct bstp_cb_ops bridge_ops = {
 402         .bcb_state = bridge_state_change,
 403         .bcb_rtage = bridge_rtable_expire
 404 };
 405
 406 SYSCTL_DECL(_net_link);
 407 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 408     "Bridge");
 409
 410 /* only pass IP[46] packets when pfil is enabled */
 411 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 412 #define V_pfil_onlyip   VNET(pfil_onlyip)
 413 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 414     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 415     "Only pass IP packets when pfil is enabled");
 416
 417 /* run pfil hooks on the bridge interface */
 418 VNET_DEFINE_STATIC(int, pfil_bridge) = 0;
 419 #define V_pfil_bridge   VNET(pfil_bridge)
 420 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 421     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 422     "Packet filter on the bridge interface");
 423
 424 /* layer2 filter with ipfw */
 425 VNET_DEFINE_STATIC(int, pfil_ipfw);
 426 #define V_pfil_ipfw     VNET(pfil_ipfw)
 427
 428 /* layer2 ARP filter with ipfw */
 429 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 430 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 431 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 432     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 433     "Filter ARP packets through IPFW layer2");
 434
 435 /* run pfil hooks on the member interface */
 436 VNET_DEFINE_STATIC(int, pfil_member) = 0;
 437 #define V_pfil_member   VNET(pfil_member)
 438 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 439     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 440     "Packet filter on the member interface");
 441
 442 /* run pfil hooks on the physical interface for locally destined packets */
 443 VNET_DEFINE_STATIC(int, pfil_local_phys);
 444 #define V_pfil_local_phys       VNET(pfil_local_phys)
 445 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 446     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 447     "Packet filter on the physical interface for locally destined packets");
 448
 449 /* log STP state changes */
 450 VNET_DEFINE_STATIC(int, log_stp);
 451 #define V_log_stp       VNET(log_stp)
 452 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 453     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 454     "Log STP state changes");
 455
 456 /* share MAC with first bridge member */
 457 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 458 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 459 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 460     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 461     "Inherit MAC address from the first bridge member");
 462
 463 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 464 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 465 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 466     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 467     "Allow overlap of link-local scope "
 468     "zones of a bridge interface and the member interfaces");
 469
 470 struct bridge_control {
 471         int     (*bc_func)(struct bridge_softc *, void *);
 472         int     bc_argsize;
 473         int     bc_flags;
 474 };
 475
 476 #define BC_F_COPYIN             0x01    /* copy arguments in */
 477 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 478 #define BC_F_SUSER              0x04    /* do super-user check */
 479
 480 const struct bridge_control bridge_control_table[] = {
 481         { bridge_ioctl_add,             sizeof(struct ifbreq),
 482           BC_F_COPYIN|BC_F_SUSER },
 483         { bridge_ioctl_del,             sizeof(struct ifbreq),
 484           BC_F_COPYIN|BC_F_SUSER },
 485
 486         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 487           BC_F_COPYIN|BC_F_COPYOUT },
 488         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 489           BC_F_COPYIN|BC_F_SUSER },
 490
 491         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 492           BC_F_COPYIN|BC_F_SUSER },
 493         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 494           BC_F_COPYOUT },
 495
 496         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 497           BC_F_COPYIN|BC_F_COPYOUT },
 498         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 499           BC_F_COPYIN|BC_F_COPYOUT },
 500
 501         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 502           BC_F_COPYIN|BC_F_SUSER },
 503
 504         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 505           BC_F_COPYIN|BC_F_SUSER },
 506         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 507           BC_F_COPYOUT },
 508
 509         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 510           BC_F_COPYIN|BC_F_SUSER },
 511
 512         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 513           BC_F_COPYIN|BC_F_SUSER },
 514
 515         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 516           BC_F_COPYOUT },
 517         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 518           BC_F_COPYIN|BC_F_SUSER },
 519
 520         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 521           BC_F_COPYOUT },
 522         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 523           BC_F_COPYIN|BC_F_SUSER },
 524
 525         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 526           BC_F_COPYOUT },
 527         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 528           BC_F_COPYIN|BC_F_SUSER },
 529
 530         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 531           BC_F_COPYOUT },
 532         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 533           BC_F_COPYIN|BC_F_SUSER },
 534
 535         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 536           BC_F_COPYIN|BC_F_SUSER },
 537
 538         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 539           BC_F_COPYIN|BC_F_SUSER },
 540
 541         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 542           BC_F_COPYIN|BC_F_SUSER },
 543         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 544           BC_F_COPYIN|BC_F_SUSER },
 545
 546         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 547           BC_F_COPYOUT },
 548
 549         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 550           BC_F_COPYOUT },
 551
 552         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 553           BC_F_COPYIN|BC_F_COPYOUT },
 554
 555         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 556           BC_F_COPYIN|BC_F_SUSER },
 557
 558         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 559           BC_F_COPYIN|BC_F_SUSER },
 560
 561         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 562           BC_F_COPYIN|BC_F_SUSER },
 563
 564 };
 565 const int bridge_control_table_size = nitems(bridge_control_table);
 566
 567 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 568 #define V_bridge_list   VNET(bridge_list)
 569 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 570                                             "if_bridge list")
 571 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 572 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 573 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 574
 575 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 576 #define V_bridge_cloner VNET(bridge_cloner)
 577
 578 static const char bridge_name[] = "bridge";
 579
 580 static void
 581 vnet_bridge_init(const void *unused __unused)
 582 {
 583
 584         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 585             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 586             UMA_ALIGN_PTR, 0);
 587         BRIDGE_LIST_LOCK_INIT();
 588         LIST_INIT(&V_bridge_list);
 589
 590         struct if_clone_addreq req = {
 591                 .create_f = bridge_clone_create,
 592                 .destroy_f = bridge_clone_destroy,
 593                 .flags = IFC_F_AUTOUNIT,
 594         };
 595         V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
 596 }
 597 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 598     vnet_bridge_init, NULL);
 599
 600 static void
 601 vnet_bridge_uninit(const void *unused __unused)
 602 {
 603
 604         ifc_detach_cloner(V_bridge_cloner);
 605         V_bridge_cloner = NULL;
 606         BRIDGE_LIST_LOCK_DESTROY();
 607
 608         /* Callbacks may use the UMA zone. */
 609         NET_EPOCH_DRAIN_CALLBACKS();
 610
 611         uma_zdestroy(V_bridge_rtnode_zone);
 612 }
 613 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 614     vnet_bridge_uninit, NULL);
 615
 616 static int
 617 bridge_modevent(module_t mod, int type, void *data)
 618 {
 619
 620         switch (type) {
 621         case MOD_LOAD:
 622                 bridge_dn_p = bridge_dummynet;
 623                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 624                     ifnet_departure_event, bridge_ifdetach, NULL,
 625                     EVENTHANDLER_PRI_ANY);
 626                 break;
 627         case MOD_UNLOAD:
 628                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 629                     bridge_detach_cookie);
 630                 bridge_dn_p = NULL;
 631                 break;
 632         default:
 633                 return (EOPNOTSUPP);
 634         }
 635         return (0);
 636 }
 637
 638 static moduledata_t bridge_mod = {
 639         "if_bridge",
 640         bridge_modevent,
 641         0
 642 };
 643
 644 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 645 MODULE_VERSION(if_bridge, 1);
 646 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 647
 648 /*
 649  * handler for net.link.bridge.ipfw
 650  */
 651 static int
 652 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 653 {
 654         int enable = V_pfil_ipfw;
 655         int error;
 656
 657         error = sysctl_handle_int(oidp, &enable, 0, req);
 658         enable &= 1;
 659
 660         if (enable != V_pfil_ipfw) {
 661                 V_pfil_ipfw = enable;
 662
 663                 /*
 664                  * Disable pfil so that ipfw doesnt run twice, if the user
 665                  * really wants both then they can re-enable pfil_bridge and/or
 666                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 667                  * layer2 type.
 668                  */
 669                 if (V_pfil_ipfw) {
 670                         V_pfil_onlyip = 0;
 671                         V_pfil_bridge = 0;
 672                         V_pfil_member = 0;
 673                 }
 674         }
 675
 676         return (error);
 677 }
 678 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 679     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
 680     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 681     "Layer2 filter with IPFW");
 682
 683 #ifdef VIMAGE
 684 static void
 685 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 686 {
 687         struct bridge_softc *sc = ifp->if_softc;
 688         struct bridge_iflist *bif;
 689
 690         BRIDGE_LOCK(sc);
 691
 692         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 693                 bridge_delete_member(sc, bif, 0);
 694
 695         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 696                 bridge_delete_span(sc, bif);
 697         }
 698
 699         BRIDGE_UNLOCK(sc);
 700
 701         ether_reassign(ifp, newvnet, arg);
 702 }
 703 #endif
 704
 705 /*
 706  * bridge_clone_create:
 707  *
 708  *      Create a new bridge instance.
 709  */
 710 static int
 711 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
 712     struct ifc_data *ifd, struct ifnet **ifpp)
 713 {
 714         struct bridge_softc *sc;
 715         struct ifnet *ifp;
 716
 717         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 718         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 719         if (ifp == NULL) {
 720                 free(sc, M_DEVBUF);
 721                 return (ENOSPC);
 722         }
 723
 724         BRIDGE_LOCK_INIT(sc);
 725         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 726         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 727
 728         /* Initialize our routing table. */
 729         bridge_rtable_init(sc);
 730
 731         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 732
 733         CK_LIST_INIT(&sc->sc_iflist);
 734         CK_LIST_INIT(&sc->sc_spanlist);
 735
 736         ifp->if_softc = sc;
 737         if_initname(ifp, bridge_name, ifd->unit);
 738         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 739         ifp->if_ioctl = bridge_ioctl;
 740 #ifdef ALTQ
 741         ifp->if_start = bridge_altq_start;
 742         ifp->if_transmit = bridge_altq_transmit;
 743         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 744         ifp->if_snd.ifq_drv_maxlen = 0;
 745         IFQ_SET_READY(&ifp->if_snd);
 746 #else
 747         ifp->if_transmit = bridge_transmit;
 748 #endif
 749         ifp->if_qflush = bridge_qflush;
 750         ifp->if_init = bridge_init;
 751         ifp->if_type = IFT_BRIDGE;
 752
 753         ether_gen_addr(ifp, &sc->sc_defaddr);
 754
 755         bstp_attach(&sc->sc_stp, &bridge_ops);
 756         ether_ifattach(ifp, sc->sc_defaddr.octet);
 757         /* Now undo some of the damage... */
 758         ifp->if_baudrate = 0;
 759         ifp->if_type = IFT_BRIDGE;
 760 #ifdef VIMAGE
 761         ifp->if_reassign = bridge_reassign;
 762 #endif
 763
 764         BRIDGE_LIST_LOCK();
 765         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 766         BRIDGE_LIST_UNLOCK();
 767         *ifpp = ifp;
 768
 769         return (0);
 770 }
 771
 772 static void
 773 bridge_clone_destroy_cb(struct epoch_context *ctx)
 774 {
 775         struct bridge_softc *sc;
 776
 777         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 778
 779         BRIDGE_LOCK_DESTROY(sc);
 780         free(sc, M_DEVBUF);
 781 }
 782
 783 /*
 784  * bridge_clone_destroy:
 785  *
 786  *      Destroy a bridge instance.
 787  */
 788 static int
 789 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
 790 {
 791         struct bridge_softc *sc = ifp->if_softc;
 792         struct bridge_iflist *bif;
 793         struct epoch_tracker et;
 794
 795         BRIDGE_LOCK(sc);
 796
 797         bridge_stop(ifp, 1);
 798         ifp->if_flags &= ~IFF_UP;
 799
 800         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 801                 bridge_delete_member(sc, bif, 0);
 802
 803         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 804                 bridge_delete_span(sc, bif);
 805         }
 806
 807         /* Tear down the routing table. */
 808         bridge_rtable_fini(sc);
 809
 810         BRIDGE_UNLOCK(sc);
 811
 812         NET_EPOCH_ENTER(et);
 813
 814         callout_drain(&sc->sc_brcallout);
 815
 816         BRIDGE_LIST_LOCK();
 817         LIST_REMOVE(sc, sc_list);
 818         BRIDGE_LIST_UNLOCK();
 819
 820         bstp_detach(&sc->sc_stp);
 821 #ifdef ALTQ
 822         IFQ_PURGE(&ifp->if_snd);
 823 #endif
 824         NET_EPOCH_EXIT(et);
 825
 826         ether_ifdetach(ifp);
 827         if_free(ifp);
 828
 829         NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
 830
 831         return (0);
 832 }
 833
 834 /*
 835  * bridge_ioctl:
 836  *
 837  *      Handle a control request from the operator.
 838  */
 839 static int
 840 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 841 {
 842         struct bridge_softc *sc = ifp->if_softc;
 843         struct ifreq *ifr = (struct ifreq *)data;
 844         struct bridge_iflist *bif;
 845         struct thread *td = curthread;
 846         union {
 847                 struct ifbreq ifbreq;
 848                 struct ifbifconf ifbifconf;
 849                 struct ifbareq ifbareq;
 850                 struct ifbaconf ifbaconf;
 851                 struct ifbrparam ifbrparam;
 852                 struct ifbropreq ifbropreq;
 853         } args;
 854         struct ifdrv *ifd = (struct ifdrv *) data;
 855         const struct bridge_control *bc;
 856         int error = 0, oldmtu;
 857
 858         BRIDGE_LOCK(sc);
 859
 860         switch (cmd) {
 861         case SIOCADDMULTI:
 862         case SIOCDELMULTI:
 863                 break;
 864
 865         case SIOCGDRVSPEC:
 866         case SIOCSDRVSPEC:
 867                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 868                         error = EINVAL;
 869                         break;
 870                 }
 871                 bc = &bridge_control_table[ifd->ifd_cmd];
 872
 873                 if (cmd == SIOCGDRVSPEC &&
 874                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 875                         error = EINVAL;
 876                         break;
 877                 }
 878                 else if (cmd == SIOCSDRVSPEC &&
 879                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 880                         error = EINVAL;
 881                         break;
 882                 }
 883
 884                 if (bc->bc_flags & BC_F_SUSER) {
 885                         error = priv_check(td, PRIV_NET_BRIDGE);
 886                         if (error)
 887                                 break;
 888                 }
 889
 890                 if (ifd->ifd_len != bc->bc_argsize ||
 891                     ifd->ifd_len > sizeof(args)) {
 892                         error = EINVAL;
 893                         break;
 894                 }
 895
 896                 bzero(&args, sizeof(args));
 897                 if (bc->bc_flags & BC_F_COPYIN) {
 898                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 899                         if (error)
 900                                 break;
 901                 }
 902
 903                 oldmtu = ifp->if_mtu;
 904                 error = (*bc->bc_func)(sc, &args);
 905                 if (error)
 906                         break;
 907
 908                 /*
 909                  * Bridge MTU may change during addition of the first port.
 910                  * If it did, do network layer specific procedure.
 911                  */
 912                 if (ifp->if_mtu != oldmtu) {
 913 #ifdef INET6
 914                         nd6_setmtu(ifp);
 915 #endif
 916                         rt_updatemtu(ifp);
 917                 }
 918
 919                 if (bc->bc_flags & BC_F_COPYOUT)
 920                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 921
 922                 break;
 923
 924         case SIOCSIFFLAGS:
 925                 if (!(ifp->if_flags & IFF_UP) &&
 926                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 927                         /*
 928                          * If interface is marked down and it is running,
 929                          * then stop and disable it.
 930                          */
 931                         bridge_stop(ifp, 1);
 932                 } else if ((ifp->if_flags & IFF_UP) &&
 933                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 934                         /*
 935                          * If interface is marked up and it is stopped, then
 936                          * start it.
 937                          */
 938                         BRIDGE_UNLOCK(sc);
 939                         (*ifp->if_init)(sc);
 940                         BRIDGE_LOCK(sc);
 941                 }
 942                 break;
 943
 944         case SIOCSIFMTU:
 945                 oldmtu = sc->sc_ifp->if_mtu;
 946
 947                 if (ifr->ifr_mtu < 576) {
 948                         error = EINVAL;
 949                         break;
 950                 }
 951                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 952                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 953                         break;
 954                 }
 955                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 956                         error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 957                             SIOCSIFMTU, (caddr_t)ifr);
 958                         if (error != 0) {
 959                                 log(LOG_NOTICE, "%s: invalid MTU: %u for"
 960                                     " member %s\n", sc->sc_ifp->if_xname,
 961                                     ifr->ifr_mtu,
 962                                     bif->bif_ifp->if_xname);
 963                                 error = EINVAL;
 964                                 break;
 965                         }
 966                 }
 967                 if (error) {
 968                         /* Restore the previous MTU on all member interfaces. */
 969                         ifr->ifr_mtu = oldmtu;
 970                         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 971                                 (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 972                                     SIOCSIFMTU, (caddr_t)ifr);
 973                         }
 974                 } else {
 975                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 976                 }
 977                 break;
 978         default:
 979                 /*
 980                  * drop the lock as ether_ioctl() will call bridge_start() and
 981                  * cause the lock to be recursed.
 982                  */
 983                 BRIDGE_UNLOCK(sc);
 984                 error = ether_ioctl(ifp, cmd, data);
 985                 BRIDGE_LOCK(sc);
 986                 break;
 987         }
 988
 989         BRIDGE_UNLOCK(sc);
 990
 991         return (error);
 992 }
 993
 994 /*
 995  * bridge_mutecaps:
 996  *
 997  *      Clear or restore unwanted capabilities on the member interface
 998  */
 999 static void
1000 bridge_mutecaps(struct bridge_softc *sc)
1001 {
1002         struct bridge_iflist *bif;
1003         int enabled, mask;
1004
1005         BRIDGE_LOCK_ASSERT(sc);
1006
1007         /* Initial bitmask of capabilities to test */
1008         mask = BRIDGE_IFCAPS_MASK;
1009
1010         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1011                 /* Every member must support it or its disabled */
1012                 mask &= bif->bif_savedcaps;
1013         }
1014
1015         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1016                 enabled = bif->bif_ifp->if_capenable;
1017                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1018                 /* strip off mask bits and enable them again if allowed */
1019                 enabled &= ~BRIDGE_IFCAPS_MASK;
1020                 enabled |= mask;
1021                 bridge_set_ifcap(sc, bif, enabled);
1022         }
1023 }
1024
1025 static void
1026 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1027 {
1028         struct ifnet *ifp = bif->bif_ifp;
1029         struct ifreq ifr;
1030         int error, mask, stuck;
1031
1032         bzero(&ifr, sizeof(ifr));
1033         ifr.ifr_reqcap = set;
1034
1035         if (ifp->if_capenable != set) {
1036                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1037                 if (error)
1038                         if_printf(sc->sc_ifp,
1039                             "error setting capabilities on %s: %d\n",
1040                             ifp->if_xname, error);
1041                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1042                 stuck = ifp->if_capenable & mask & ~set;
1043                 if (stuck != 0)
1044                         if_printf(sc->sc_ifp,
1045                             "can't disable some capabilities on %s: 0x%x\n",
1046                             ifp->if_xname, stuck);
1047         }
1048 }
1049
1050 /*
1051  * bridge_lookup_member:
1052  *
1053  *      Lookup a bridge member interface.
1054  */
1055 static struct bridge_iflist *
1056 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1057 {
1058         struct bridge_iflist *bif;
1059         struct ifnet *ifp;
1060
1061         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1062
1063         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1064                 ifp = bif->bif_ifp;
1065                 if (strcmp(ifp->if_xname, name) == 0)
1066                         return (bif);
1067         }
1068
1069         return (NULL);
1070 }
1071
1072 /*
1073  * bridge_lookup_member_if:
1074  *
1075  *      Lookup a bridge member interface by ifnet*.
1076  */
1077 static struct bridge_iflist *
1078 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1079 {
1080         struct bridge_iflist *bif;
1081
1082         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1083
1084         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1085                 if (bif->bif_ifp == member_ifp)
1086                         return (bif);
1087         }
1088
1089         return (NULL);
1090 }
1091
1092 static void
1093 bridge_delete_member_cb(struct epoch_context *ctx)
1094 {
1095         struct bridge_iflist *bif;
1096
1097         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1098
1099         free(bif, M_DEVBUF);
1100 }
1101
1102 /*
1103  * bridge_delete_member:
1104  *
1105  *      Delete the specified member interface.
1106  */
1107 static void
1108 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1109     int gone)
1110 {
1111         struct ifnet *ifs = bif->bif_ifp;
1112         struct ifnet *fif = NULL;
1113         struct bridge_iflist *bifl;
1114
1115         BRIDGE_LOCK_ASSERT(sc);
1116
1117         if (bif->bif_flags & IFBIF_STP)
1118                 bstp_disable(&bif->bif_stp);
1119
1120         ifs->if_bridge = NULL;
1121         CK_LIST_REMOVE(bif, bif_next);
1122
1123         /*
1124          * If removing the interface that gave the bridge its mac address, set
1125          * the mac address of the bridge to the address of the next member, or
1126          * to its default address if no members are left.
1127          */
1128         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1129                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1130                         bcopy(&sc->sc_defaddr,
1131                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1132                         sc->sc_ifaddr = NULL;
1133                 } else {
1134                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1135                         fif = bifl->bif_ifp;
1136                         bcopy(IF_LLADDR(fif),
1137                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1138                         sc->sc_ifaddr = fif;
1139                 }
1140                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1141         }
1142
1143         bridge_linkcheck(sc);
1144         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1145         BRIDGE_RT_LOCK(sc);
1146         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1147         BRIDGE_RT_UNLOCK(sc);
1148         KASSERT(bif->bif_addrcnt == 0,
1149             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1150
1151         ifs->if_bridge_output = NULL;
1152         ifs->if_bridge_input = NULL;
1153         ifs->if_bridge_linkstate = NULL;
1154         if (!gone) {
1155                 switch (ifs->if_type) {
1156                 case IFT_ETHER:
1157                 case IFT_L2VLAN:
1158                         /*
1159                          * Take the interface out of promiscuous mode, but only
1160                          * if it was promiscuous in the first place. It might
1161                          * not be if we're in the bridge_ioctl_add() error path.
1162                          */
1163                         if (ifs->if_flags & IFF_PROMISC)
1164                                 (void) ifpromisc(ifs, 0);
1165                         break;
1166
1167                 case IFT_GIF:
1168                         break;
1169
1170                 default:
1171 #ifdef DIAGNOSTIC
1172                         panic("bridge_delete_member: impossible");
1173 #endif
1174                         break;
1175                 }
1176                 /* reneable any interface capabilities */
1177                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1178         }
1179         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1180
1181         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1182 }
1183
1184 /*
1185  * bridge_delete_span:
1186  *
1187  *      Delete the specified span interface.
1188  */
1189 static void
1190 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1191 {
1192         BRIDGE_LOCK_ASSERT(sc);
1193
1194         KASSERT(bif->bif_ifp->if_bridge == NULL,
1195             ("%s: not a span interface", __func__));
1196
1197         CK_LIST_REMOVE(bif, bif_next);
1198
1199         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1200 }
1201
1202 static int
1203 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1204 {
1205         struct ifbreq *req = arg;
1206         struct bridge_iflist *bif = NULL;
1207         struct ifnet *ifs;
1208         int error = 0;
1209
1210         ifs = ifunit(req->ifbr_ifsname);
1211         if (ifs == NULL)
1212                 return (ENOENT);
1213         if (ifs->if_ioctl == NULL)      /* must be supported */
1214                 return (EINVAL);
1215
1216         /* If it's in the span list, it can't be a member. */
1217         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1218                 if (ifs == bif->bif_ifp)
1219                         return (EBUSY);
1220
1221         if (ifs->if_bridge == sc)
1222                 return (EEXIST);
1223
1224         if (ifs->if_bridge != NULL)
1225                 return (EBUSY);
1226
1227         switch (ifs->if_type) {
1228         case IFT_ETHER:
1229         case IFT_L2VLAN:
1230         case IFT_GIF:
1231                 /* permitted interface types */
1232                 break;
1233         default:
1234                 return (EINVAL);
1235         }
1236
1237 #ifdef INET6
1238         /*
1239          * Two valid inet6 addresses with link-local scope must not be
1240          * on the parent interface and the member interfaces at the
1241          * same time.  This restriction is needed to prevent violation
1242          * of link-local scope zone.  Attempts to add a member
1243          * interface which has inet6 addresses when the parent has
1244          * inet6 triggers removal of all inet6 addresses on the member
1245          * interface.
1246          */
1247
1248         /* Check if the parent interface has a link-local scope addr. */
1249         if (V_allow_llz_overlap == 0 &&
1250             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1251                 /*
1252                  * If any, remove all inet6 addresses from the member
1253                  * interfaces.
1254                  */
1255                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1256                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1257                                 in6_ifdetach(bif->bif_ifp);
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                                     bif->bif_ifp->if_xname);
1263                         }
1264                 }
1265                 if (in6ifa_llaonifp(ifs)) {
1266                         in6_ifdetach(ifs);
1267                         if_printf(sc->sc_ifp,
1268                             "IPv6 addresses on %s have been removed "
1269                             "before adding it as a member to prevent "
1270                             "IPv6 address scope violation.\n",
1271                             ifs->if_xname);
1272                 }
1273         }
1274 #endif
1275         /* Allow the first Ethernet member to define the MTU */
1276         if (CK_LIST_EMPTY(&sc->sc_iflist))
1277                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1278         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1279                 struct ifreq ifr;
1280
1281                 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1282                     ifs->if_xname);
1283                 ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1284
1285                 error = (*ifs->if_ioctl)(ifs,
1286                     SIOCSIFMTU, (caddr_t)&ifr);
1287                 if (error != 0) {
1288                         log(LOG_NOTICE, "%s: invalid MTU: %u for"
1289                             " new member %s\n", sc->sc_ifp->if_xname,
1290                             ifr.ifr_mtu,
1291                             ifs->if_xname);
1292                         return (EINVAL);
1293                 }
1294         }
1295
1296         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1297         if (bif == NULL)
1298                 return (ENOMEM);
1299
1300         bif->bif_ifp = ifs;
1301         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1302         bif->bif_savedcaps = ifs->if_capenable;
1303
1304         /*
1305          * Assign the interface's MAC address to the bridge if it's the first
1306          * member and the MAC address of the bridge has not been changed from
1307          * the default randomly generated one.
1308          */
1309         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1310             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1311                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1312                 sc->sc_ifaddr = ifs;
1313                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1314         }
1315
1316         ifs->if_bridge = sc;
1317         ifs->if_bridge_output = bridge_output;
1318         ifs->if_bridge_input = bridge_input;
1319         ifs->if_bridge_linkstate = bridge_linkstate;
1320         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1321         /*
1322          * XXX: XLOCK HERE!?!
1323          *
1324          * NOTE: insert_***HEAD*** should be safe for the traversals.
1325          */
1326         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1327
1328         /* Set interface capabilities to the intersection set of all members */
1329         bridge_mutecaps(sc);
1330         bridge_linkcheck(sc);
1331
1332         /* Place the interface into promiscuous mode */
1333         switch (ifs->if_type) {
1334                 case IFT_ETHER:
1335                 case IFT_L2VLAN:
1336                         error = ifpromisc(ifs, 1);
1337                         break;
1338         }
1339
1340         if (error)
1341                 bridge_delete_member(sc, bif, 0);
1342         return (error);
1343 }
1344
1345 static int
1346 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1347 {
1348         struct ifbreq *req = arg;
1349         struct bridge_iflist *bif;
1350
1351         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1352         if (bif == NULL)
1353                 return (ENOENT);
1354
1355         bridge_delete_member(sc, bif, 0);
1356
1357         return (0);
1358 }
1359
1360 static int
1361 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1362 {
1363         struct ifbreq *req = arg;
1364         struct bridge_iflist *bif;
1365         struct bstp_port *bp;
1366
1367         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1368         if (bif == NULL)
1369                 return (ENOENT);
1370
1371         bp = &bif->bif_stp;
1372         req->ifbr_ifsflags = bif->bif_flags;
1373         req->ifbr_state = bp->bp_state;
1374         req->ifbr_priority = bp->bp_priority;
1375         req->ifbr_path_cost = bp->bp_path_cost;
1376         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1377         req->ifbr_proto = bp->bp_protover;
1378         req->ifbr_role = bp->bp_role;
1379         req->ifbr_stpflags = bp->bp_flags;
1380         req->ifbr_addrcnt = bif->bif_addrcnt;
1381         req->ifbr_addrmax = bif->bif_addrmax;
1382         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1383
1384         /* Copy STP state options as flags */
1385         if (bp->bp_operedge)
1386                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1387         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1388                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1389         if (bp->bp_ptp_link)
1390                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1391         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1392                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1393         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1394                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1395         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1396                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1397         return (0);
1398 }
1399
1400 static int
1401 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1402 {
1403         struct epoch_tracker et;
1404         struct ifbreq *req = arg;
1405         struct bridge_iflist *bif;
1406         struct bstp_port *bp;
1407         int error;
1408
1409         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1410         if (bif == NULL)
1411                 return (ENOENT);
1412         bp = &bif->bif_stp;
1413
1414         if (req->ifbr_ifsflags & IFBIF_SPAN)
1415                 /* SPAN is readonly */
1416                 return (EINVAL);
1417
1418         NET_EPOCH_ENTER(et);
1419
1420         if (req->ifbr_ifsflags & IFBIF_STP) {
1421                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1422                         error = bstp_enable(&bif->bif_stp);
1423                         if (error) {
1424                                 NET_EPOCH_EXIT(et);
1425                                 return (error);
1426                         }
1427                 }
1428         } else {
1429                 if ((bif->bif_flags & IFBIF_STP) != 0)
1430                         bstp_disable(&bif->bif_stp);
1431         }
1432
1433         /* Pass on STP flags */
1434         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1435         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1436         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1437         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1438
1439         /* Save the bits relating to the bridge */
1440         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1441
1442         NET_EPOCH_EXIT(et);
1443
1444         return (0);
1445 }
1446
1447 static int
1448 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1449 {
1450         struct ifbrparam *param = arg;
1451
1452         sc->sc_brtmax = param->ifbrp_csize;
1453         bridge_rttrim(sc);
1454
1455         return (0);
1456 }
1457
1458 static int
1459 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1460 {
1461         struct ifbrparam *param = arg;
1462
1463         param->ifbrp_csize = sc->sc_brtmax;
1464
1465         return (0);
1466 }
1467
1468 static int
1469 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1470 {
1471         struct ifbifconf *bifc = arg;
1472         struct bridge_iflist *bif;
1473         struct ifbreq breq;
1474         char *buf, *outbuf;
1475         int count, buflen, len, error = 0;
1476
1477         count = 0;
1478         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1479                 count++;
1480         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1481                 count++;
1482
1483         buflen = sizeof(breq) * count;
1484         if (bifc->ifbic_len == 0) {
1485                 bifc->ifbic_len = buflen;
1486                 return (0);
1487         }
1488         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1489         if (outbuf == NULL)
1490                 return (ENOMEM);
1491
1492         count = 0;
1493         buf = outbuf;
1494         len = min(bifc->ifbic_len, buflen);
1495         bzero(&breq, sizeof(breq));
1496         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1497                 if (len < sizeof(breq))
1498                         break;
1499
1500                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1501                     sizeof(breq.ifbr_ifsname));
1502                 /* Fill in the ifbreq structure */
1503                 error = bridge_ioctl_gifflags(sc, &breq);
1504                 if (error)
1505                         break;
1506                 memcpy(buf, &breq, sizeof(breq));
1507                 count++;
1508                 buf += sizeof(breq);
1509                 len -= sizeof(breq);
1510         }
1511         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1512                 if (len < sizeof(breq))
1513                         break;
1514
1515                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1516                     sizeof(breq.ifbr_ifsname));
1517                 breq.ifbr_ifsflags = bif->bif_flags;
1518                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1519                 memcpy(buf, &breq, sizeof(breq));
1520                 count++;
1521                 buf += sizeof(breq);
1522                 len -= sizeof(breq);
1523         }
1524
1525         bifc->ifbic_len = sizeof(breq) * count;
1526         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1527         free(outbuf, M_TEMP);
1528         return (error);
1529 }
1530
1531 static int
1532 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1533 {
1534         struct ifbaconf *bac = arg;
1535         struct bridge_rtnode *brt;
1536         struct ifbareq bareq;
1537         char *buf, *outbuf;
1538         int count, buflen, len, error = 0;
1539
1540         if (bac->ifbac_len == 0)
1541                 return (0);
1542
1543         count = 0;
1544         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1545                 count++;
1546         buflen = sizeof(bareq) * count;
1547
1548         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1549         if (outbuf == NULL)
1550                 return (ENOMEM);
1551
1552         count = 0;
1553         buf = outbuf;
1554         len = min(bac->ifbac_len, buflen);
1555         bzero(&bareq, sizeof(bareq));
1556         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1557                 if (len < sizeof(bareq))
1558                         goto out;
1559                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1560                     sizeof(bareq.ifba_ifsname));
1561                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1562                 bareq.ifba_vlan = brt->brt_vlan;
1563                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1564                                 time_uptime < brt->brt_expire)
1565                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1566                 else
1567                         bareq.ifba_expire = 0;
1568                 bareq.ifba_flags = brt->brt_flags;
1569
1570                 memcpy(buf, &bareq, sizeof(bareq));
1571                 count++;
1572                 buf += sizeof(bareq);
1573                 len -= sizeof(bareq);
1574         }
1575 out:
1576         bac->ifbac_len = sizeof(bareq) * count;
1577         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1578         free(outbuf, M_TEMP);
1579         return (error);
1580 }
1581
1582 static int
1583 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1584 {
1585         struct ifbareq *req = arg;
1586         struct bridge_iflist *bif;
1587         struct epoch_tracker et;
1588         int error;
1589
1590         NET_EPOCH_ENTER(et);
1591         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1592         if (bif == NULL) {
1593                 NET_EPOCH_EXIT(et);
1594                 return (ENOENT);
1595         }
1596
1597         /* bridge_rtupdate() may acquire the lock. */
1598         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1599             req->ifba_flags);
1600         NET_EPOCH_EXIT(et);
1601
1602         return (error);
1603 }
1604
1605 static int
1606 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1607 {
1608         struct ifbrparam *param = arg;
1609
1610         sc->sc_brttimeout = param->ifbrp_ctime;
1611         return (0);
1612 }
1613
1614 static int
1615 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1616 {
1617         struct ifbrparam *param = arg;
1618
1619         param->ifbrp_ctime = sc->sc_brttimeout;
1620         return (0);
1621 }
1622
1623 static int
1624 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1625 {
1626         struct ifbareq *req = arg;
1627
1628         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1629 }
1630
1631 static int
1632 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1633 {
1634         struct ifbreq *req = arg;
1635
1636         BRIDGE_RT_LOCK(sc);
1637         bridge_rtflush(sc, req->ifbr_ifsflags);
1638         BRIDGE_RT_UNLOCK(sc);
1639
1640         return (0);
1641 }
1642
1643 static int
1644 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1645 {
1646         struct ifbrparam *param = arg;
1647         struct bstp_state *bs = &sc->sc_stp;
1648
1649         param->ifbrp_prio = bs->bs_bridge_priority;
1650         return (0);
1651 }
1652
1653 static int
1654 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1655 {
1656         struct ifbrparam *param = arg;
1657
1658         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1659 }
1660
1661 static int
1662 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1663 {
1664         struct ifbrparam *param = arg;
1665         struct bstp_state *bs = &sc->sc_stp;
1666
1667         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1668         return (0);
1669 }
1670
1671 static int
1672 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1673 {
1674         struct ifbrparam *param = arg;
1675
1676         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1677 }
1678
1679 static int
1680 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1681 {
1682         struct ifbrparam *param = arg;
1683         struct bstp_state *bs = &sc->sc_stp;
1684
1685         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1686         return (0);
1687 }
1688
1689 static int
1690 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1691 {
1692         struct ifbrparam *param = arg;
1693
1694         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1695 }
1696
1697 static int
1698 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1699 {
1700         struct ifbrparam *param = arg;
1701         struct bstp_state *bs = &sc->sc_stp;
1702
1703         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1704         return (0);
1705 }
1706
1707 static int
1708 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1709 {
1710         struct ifbrparam *param = arg;
1711
1712         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1713 }
1714
1715 static int
1716 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1717 {
1718         struct ifbreq *req = arg;
1719         struct bridge_iflist *bif;
1720
1721         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1722         if (bif == NULL)
1723                 return (ENOENT);
1724
1725         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1726 }
1727
1728 static int
1729 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1730 {
1731         struct ifbreq *req = arg;
1732         struct bridge_iflist *bif;
1733
1734         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1735         if (bif == NULL)
1736                 return (ENOENT);
1737
1738         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1739 }
1740
1741 static int
1742 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1743 {
1744         struct ifbreq *req = arg;
1745         struct bridge_iflist *bif;
1746
1747         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1748         if (bif == NULL)
1749                 return (ENOENT);
1750
1751         bif->bif_addrmax = req->ifbr_addrmax;
1752         return (0);
1753 }
1754
1755 static int
1756 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1757 {
1758         struct ifbreq *req = arg;
1759         struct bridge_iflist *bif = NULL;
1760         struct ifnet *ifs;
1761
1762         ifs = ifunit(req->ifbr_ifsname);
1763         if (ifs == NULL)
1764                 return (ENOENT);
1765
1766         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1767                 if (ifs == bif->bif_ifp)
1768                         return (EBUSY);
1769
1770         if (ifs->if_bridge != NULL)
1771                 return (EBUSY);
1772
1773         switch (ifs->if_type) {
1774                 case IFT_ETHER:
1775                 case IFT_GIF:
1776                 case IFT_L2VLAN:
1777                         break;
1778                 default:
1779                         return (EINVAL);
1780         }
1781
1782         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1783         if (bif == NULL)
1784                 return (ENOMEM);
1785
1786         bif->bif_ifp = ifs;
1787         bif->bif_flags = IFBIF_SPAN;
1788
1789         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1790
1791         return (0);
1792 }
1793
1794 static int
1795 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1796 {
1797         struct ifbreq *req = arg;
1798         struct bridge_iflist *bif;
1799         struct ifnet *ifs;
1800
1801         ifs = ifunit(req->ifbr_ifsname);
1802         if (ifs == NULL)
1803                 return (ENOENT);
1804
1805         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1806                 if (ifs == bif->bif_ifp)
1807                         break;
1808
1809         if (bif == NULL)
1810                 return (ENOENT);
1811
1812         bridge_delete_span(sc, bif);
1813
1814         return (0);
1815 }
1816
1817 static int
1818 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1819 {
1820         struct ifbropreq *req = arg;
1821         struct bstp_state *bs = &sc->sc_stp;
1822         struct bstp_port *root_port;
1823
1824         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1825         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1826         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1827
1828         root_port = bs->bs_root_port;
1829         if (root_port == NULL)
1830                 req->ifbop_root_port = 0;
1831         else
1832                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1833
1834         req->ifbop_holdcount = bs->bs_txholdcount;
1835         req->ifbop_priority = bs->bs_bridge_priority;
1836         req->ifbop_protocol = bs->bs_protover;
1837         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1838         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1839         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1840         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1841         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1842         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1843
1844         return (0);
1845 }
1846
1847 static int
1848 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1849 {
1850         struct ifbrparam *param = arg;
1851
1852         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1853         return (0);
1854 }
1855
1856 static int
1857 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1858 {
1859         struct ifbpstpconf *bifstp = arg;
1860         struct bridge_iflist *bif;
1861         struct bstp_port *bp;
1862         struct ifbpstpreq bpreq;
1863         char *buf, *outbuf;
1864         int count, buflen, len, error = 0;
1865
1866         count = 0;
1867         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1868                 if ((bif->bif_flags & IFBIF_STP) != 0)
1869                         count++;
1870         }
1871
1872         buflen = sizeof(bpreq) * count;
1873         if (bifstp->ifbpstp_len == 0) {
1874                 bifstp->ifbpstp_len = buflen;
1875                 return (0);
1876         }
1877
1878         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1879         if (outbuf == NULL)
1880                 return (ENOMEM);
1881
1882         count = 0;
1883         buf = outbuf;
1884         len = min(bifstp->ifbpstp_len, buflen);
1885         bzero(&bpreq, sizeof(bpreq));
1886         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1887                 if (len < sizeof(bpreq))
1888                         break;
1889
1890                 if ((bif->bif_flags & IFBIF_STP) == 0)
1891                         continue;
1892
1893                 bp = &bif->bif_stp;
1894                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1895                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1896                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1897                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1898                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1899                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1900
1901                 memcpy(buf, &bpreq, sizeof(bpreq));
1902                 count++;
1903                 buf += sizeof(bpreq);
1904                 len -= sizeof(bpreq);
1905         }
1906
1907         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1908         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1909         free(outbuf, M_TEMP);
1910         return (error);
1911 }
1912
1913 static int
1914 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1915 {
1916         struct ifbrparam *param = arg;
1917
1918         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1919 }
1920
1921 static int
1922 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1923 {
1924         struct ifbrparam *param = arg;
1925
1926         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1927 }
1928
1929 /*
1930  * bridge_ifdetach:
1931  *
1932  *      Detach an interface from a bridge.  Called when a member
1933  *      interface is detaching.
1934  */
1935 static void
1936 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1937 {
1938         struct bridge_softc *sc = ifp->if_bridge;
1939         struct bridge_iflist *bif;
1940
1941         if (ifp->if_flags & IFF_RENAMING)
1942                 return;
1943         if (V_bridge_cloner == NULL) {
1944                 /*
1945                  * This detach handler can be called after
1946                  * vnet_bridge_uninit().  Just return in that case.
1947                  */
1948                 return;
1949         }
1950         /* Check if the interface is a bridge member */
1951         if (sc != NULL) {
1952                 BRIDGE_LOCK(sc);
1953
1954                 bif = bridge_lookup_member_if(sc, ifp);
1955                 if (bif != NULL)
1956                         bridge_delete_member(sc, bif, 1);
1957
1958                 BRIDGE_UNLOCK(sc);
1959                 return;
1960         }
1961
1962         /* Check if the interface is a span port */
1963         BRIDGE_LIST_LOCK();
1964         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1965                 BRIDGE_LOCK(sc);
1966                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1967                         if (ifp == bif->bif_ifp) {
1968                                 bridge_delete_span(sc, bif);
1969                                 break;
1970                         }
1971
1972                 BRIDGE_UNLOCK(sc);
1973         }
1974         BRIDGE_LIST_UNLOCK();
1975 }
1976
1977 /*
1978  * bridge_init:
1979  *
1980  *      Initialize a bridge interface.
1981  */
1982 static void
1983 bridge_init(void *xsc)
1984 {
1985         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1986         struct ifnet *ifp = sc->sc_ifp;
1987
1988         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1989                 return;
1990
1991         BRIDGE_LOCK(sc);
1992         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1993             bridge_timer, sc);
1994
1995         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1996         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1997
1998         BRIDGE_UNLOCK(sc);
1999 }
2000
2001 /*
2002  * bridge_stop:
2003  *
2004  *      Stop the bridge interface.
2005  */
2006 static void
2007 bridge_stop(struct ifnet *ifp, int disable)
2008 {
2009         struct bridge_softc *sc = ifp->if_softc;
2010
2011         BRIDGE_LOCK_ASSERT(sc);
2012
2013         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2014                 return;
2015
2016         BRIDGE_RT_LOCK(sc);
2017         callout_stop(&sc->sc_brcallout);
2018
2019         bstp_stop(&sc->sc_stp);
2020
2021         bridge_rtflush(sc, IFBF_FLUSHDYN);
2022         BRIDGE_RT_UNLOCK(sc);
2023
2024         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2025 }
2026
2027 /*
2028  * bridge_enqueue:
2029  *
2030  *      Enqueue a packet on a bridge member interface.
2031  *
2032  */
2033 static int
2034 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2035 {
2036         int len, err = 0;
2037         short mflags;
2038         struct mbuf *m0;
2039
2040         /* We may be sending a fragment so traverse the mbuf */
2041         for (; m; m = m0) {
2042                 m0 = m->m_nextpkt;
2043                 m->m_nextpkt = NULL;
2044                 len = m->m_pkthdr.len;
2045                 mflags = m->m_flags;
2046
2047                 /*
2048                  * If underlying interface can not do VLAN tag insertion itself
2049                  * then attach a packet tag that holds it.
2050                  */
2051                 if ((m->m_flags & M_VLANTAG) &&
2052                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2053                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2054                         if (m == NULL) {
2055                                 if_printf(dst_ifp,
2056                                     "unable to prepend VLAN header\n");
2057                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2058                                 continue;
2059                         }
2060                         m->m_flags &= ~M_VLANTAG;
2061                 }
2062
2063                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2064                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2065                         int n;
2066
2067                         for (m = m0, n = 1; m != NULL; m = m0, n++) {
2068                                 m0 = m->m_nextpkt;
2069                                 m_freem(m);
2070                         }
2071                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2072                         break;
2073                 }
2074
2075                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2076                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2077                 if (mflags & M_MCAST)
2078                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2079         }
2080
2081         return (err);
2082 }
2083
2084 /*
2085  * bridge_dummynet:
2086  *
2087  *      Receive a queued packet from dummynet and pass it on to the output
2088  *      interface.
2089  *
2090  *      The mbuf has the Ethernet header already attached.
2091  */
2092 static void
2093 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2094 {
2095         struct bridge_softc *sc;
2096
2097         sc = ifp->if_bridge;
2098
2099         /*
2100          * The packet didnt originate from a member interface. This should only
2101          * ever happen if a member interface is removed while packets are
2102          * queued for it.
2103          */
2104         if (sc == NULL) {
2105                 m_freem(m);
2106                 return;
2107         }
2108
2109         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2110 #ifdef INET6
2111             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2112 #endif
2113             ) {
2114                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2115                         return;
2116                 if (m == NULL)
2117                         return;
2118         }
2119
2120         bridge_enqueue(sc, ifp, m);
2121 }
2122
2123 /*
2124  * bridge_output:
2125  *
2126  *      Send output from a bridge member interface.  This
2127  *      performs the bridging function for locally originated
2128  *      packets.
2129  *
2130  *      The mbuf has the Ethernet header already attached.  We must
2131  *      enqueue or free the mbuf before returning.
2132  */
2133 static int
2134 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2135     struct rtentry *rt)
2136 {
2137         struct ether_header *eh;
2138         struct ifnet *bifp, *dst_if;
2139         struct bridge_softc *sc;
2140         uint16_t vlan;
2141
2142         NET_EPOCH_ASSERT();
2143
2144         if (m->m_len < ETHER_HDR_LEN) {
2145                 m = m_pullup(m, ETHER_HDR_LEN);
2146                 if (m == NULL)
2147                         return (0);
2148         }
2149
2150         eh = mtod(m, struct ether_header *);
2151         sc = ifp->if_bridge;
2152         vlan = VLANTAGOF(m);
2153
2154         bifp = sc->sc_ifp;
2155
2156         /*
2157          * If bridge is down, but the original output interface is up,
2158          * go ahead and send out that interface.  Otherwise, the packet
2159          * is dropped below.
2160          */
2161         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2162                 dst_if = ifp;
2163                 goto sendunicast;
2164         }
2165
2166         /*
2167          * If the packet is a multicast, or we don't know a better way to
2168          * get there, send to all interfaces.
2169          */
2170         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2171                 dst_if = NULL;
2172         else
2173                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2174         /* Tap any traffic not passing back out the originating interface */
2175         if (dst_if != ifp)
2176                 ETHER_BPF_MTAP(bifp, m);
2177         if (dst_if == NULL) {
2178                 struct bridge_iflist *bif;
2179                 struct mbuf *mc;
2180                 int used = 0;
2181
2182                 bridge_span(sc, m);
2183
2184                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2185                         dst_if = bif->bif_ifp;
2186
2187                         if (dst_if->if_type == IFT_GIF)
2188                                 continue;
2189                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2190                                 continue;
2191
2192                         /*
2193                          * If this is not the original output interface,
2194                          * and the interface is participating in spanning
2195                          * tree, make sure the port is in a state that
2196                          * allows forwarding.
2197                          */
2198                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2199                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2200                                 continue;
2201
2202                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2203                                 used = 1;
2204                                 mc = m;
2205                         } else {
2206                                 mc = m_dup(m, M_NOWAIT);
2207                                 if (mc == NULL) {
2208                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2209                                         continue;
2210                                 }
2211                         }
2212
2213                         bridge_enqueue(sc, dst_if, mc);
2214                 }
2215                 if (used == 0)
2216                         m_freem(m);
2217                 return (0);
2218         }
2219
2220 sendunicast:
2221         /*
2222          * XXX Spanning tree consideration here?
2223          */
2224
2225         bridge_span(sc, m);
2226         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2227                 m_freem(m);
2228                 return (0);
2229         }
2230
2231         bridge_enqueue(sc, dst_if, m);
2232         return (0);
2233 }
2234
2235 /*
2236  * bridge_transmit:
2237  *
2238  *      Do output on a bridge.
2239  *
2240  */
2241 static int
2242 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2243 {
2244         struct bridge_softc *sc;
2245         struct ether_header *eh;
2246         struct ifnet *dst_if;
2247         int error = 0;
2248
2249         sc = ifp->if_softc;
2250
2251         ETHER_BPF_MTAP(ifp, m);
2252
2253         eh = mtod(m, struct ether_header *);
2254
2255         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2256             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2257                 error = bridge_enqueue(sc, dst_if, m);
2258         } else
2259                 bridge_broadcast(sc, ifp, m, 0);
2260
2261         return (error);
2262 }
2263
2264 #ifdef ALTQ
2265 static void
2266 bridge_altq_start(if_t ifp)
2267 {
2268         struct ifaltq *ifq = &ifp->if_snd;
2269         struct mbuf *m;
2270
2271         IFQ_LOCK(ifq);
2272         IFQ_DEQUEUE_NOLOCK(ifq, m);
2273         while (m != NULL) {
2274                 bridge_transmit(ifp, m);
2275                 IFQ_DEQUEUE_NOLOCK(ifq, m);
2276         }
2277         IFQ_UNLOCK(ifq);
2278 }
2279
2280 static int
2281 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2282 {
2283         int err;
2284
2285         if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2286                 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2287                 if (err == 0)
2288                         bridge_altq_start(ifp);
2289         } else
2290                 err = bridge_transmit(ifp, m);
2291
2292         return (err);
2293 }
2294 #endif  /* ALTQ */
2295
2296 /*
2297  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2298  */
2299 static void
2300 bridge_qflush(struct ifnet *ifp __unused)
2301 {
2302 }
2303
2304 /*
2305  * bridge_forward:
2306  *
2307  *      The forwarding function of the bridge.
2308  *
2309  *      NOTE: Releases the lock on return.
2310  */
2311 static void
2312 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2313     struct mbuf *m)
2314 {
2315         struct bridge_iflist *dbif;
2316         struct ifnet *src_if, *dst_if, *ifp;
2317         struct ether_header *eh;
2318         uint16_t vlan;
2319         uint8_t *dst;
2320         int error;
2321
2322         NET_EPOCH_ASSERT();
2323
2324         src_if = m->m_pkthdr.rcvif;
2325         ifp = sc->sc_ifp;
2326
2327         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2328         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2329         vlan = VLANTAGOF(m);
2330
2331         if ((sbif->bif_flags & IFBIF_STP) &&
2332             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2333                 goto drop;
2334
2335         eh = mtod(m, struct ether_header *);
2336         dst = eh->ether_dhost;
2337
2338         /* If the interface is learning, record the address. */
2339         if (sbif->bif_flags & IFBIF_LEARNING) {
2340                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2341                     sbif, 0, IFBAF_DYNAMIC);
2342                 /*
2343                  * If the interface has addresses limits then deny any source
2344                  * that is not in the cache.
2345                  */
2346                 if (error && sbif->bif_addrmax)
2347                         goto drop;
2348         }
2349
2350         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2351             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2352                 goto drop;
2353
2354         /*
2355          * At this point, the port either doesn't participate
2356          * in spanning tree or it is in the forwarding state.
2357          */
2358
2359         /*
2360          * If the packet is unicast, destined for someone on
2361          * "this" side of the bridge, drop it.
2362          */
2363         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2364                 dst_if = bridge_rtlookup(sc, dst, vlan);
2365                 if (src_if == dst_if)
2366                         goto drop;
2367         } else {
2368                 /*
2369                  * Check if its a reserved multicast address, any address
2370                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2371                  * bridge.
2372                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2373                  */
2374                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2375                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2376                     dst[4] == 0x00 && dst[5] <= 0x0f)
2377                         goto drop;
2378
2379                 /* ...forward it to all interfaces. */
2380                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2381                 dst_if = NULL;
2382         }
2383
2384         /*
2385          * If we have a destination interface which is a member of our bridge,
2386          * OR this is a unicast packet, push it through the bpf(4) machinery.
2387          * For broadcast or multicast packets, don't bother because it will
2388          * be reinjected into ether_input. We do this before we pass the packets
2389          * through the pfil(9) framework, as it is possible that pfil(9) will
2390          * drop the packet, or possibly modify it, making it difficult to debug
2391          * firewall issues on the bridge.
2392          */
2393         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2394                 ETHER_BPF_MTAP(ifp, m);
2395
2396         /* run the packet filter */
2397         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2398 #ifdef INET6
2399             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2400 #endif
2401             ) {
2402                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2403                         return;
2404                 if (m == NULL)
2405                         return;
2406         }
2407
2408         if (dst_if == NULL) {
2409                 bridge_broadcast(sc, src_if, m, 1);
2410                 return;
2411         }
2412
2413         /*
2414          * At this point, we're dealing with a unicast frame
2415          * going to a different interface.
2416          */
2417         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2418                 goto drop;
2419
2420         dbif = bridge_lookup_member_if(sc, dst_if);
2421         if (dbif == NULL)
2422                 /* Not a member of the bridge (anymore?) */
2423                 goto drop;
2424
2425         /* Private segments can not talk to each other */
2426         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2427                 goto drop;
2428
2429         if ((dbif->bif_flags & IFBIF_STP) &&
2430             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2431                 goto drop;
2432
2433         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2434 #ifdef INET6
2435             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2436 #endif
2437             ) {
2438                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2439                         return;
2440                 if (m == NULL)
2441                         return;
2442         }
2443
2444         bridge_enqueue(sc, dst_if, m);
2445         return;
2446
2447 drop:
2448         m_freem(m);
2449 }
2450
2451 /*
2452  * bridge_input:
2453  *
2454  *      Receive input from a member interface.  Queue the packet for
2455  *      bridging if it is not for us.
2456  */
2457 static struct mbuf *
2458 bridge_input(struct ifnet *ifp, struct mbuf *m)
2459 {
2460         struct bridge_softc *sc = ifp->if_bridge;
2461         struct bridge_iflist *bif, *bif2;
2462         struct ifnet *bifp;
2463         struct ether_header *eh;
2464         struct mbuf *mc, *mc2;
2465         uint16_t vlan;
2466         int error;
2467
2468         NET_EPOCH_ASSERT();
2469
2470         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2471                 return (m);
2472
2473         bifp = sc->sc_ifp;
2474         vlan = VLANTAGOF(m);
2475
2476         /*
2477          * Implement support for bridge monitoring. If this flag has been
2478          * set on this interface, discard the packet once we push it through
2479          * the bpf(4) machinery, but before we do, increment the byte and
2480          * packet counters associated with this interface.
2481          */
2482         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2483                 m->m_pkthdr.rcvif  = bifp;
2484                 ETHER_BPF_MTAP(bifp, m);
2485                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2486                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2487                 m_freem(m);
2488                 return (NULL);
2489         }
2490         bif = bridge_lookup_member_if(sc, ifp);
2491         if (bif == NULL) {
2492                 return (m);
2493         }
2494
2495         eh = mtod(m, struct ether_header *);
2496
2497         bridge_span(sc, m);
2498
2499         if (m->m_flags & (M_BCAST|M_MCAST)) {
2500                 /* Tap off 802.1D packets; they do not get forwarded. */
2501                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2502                     ETHER_ADDR_LEN) == 0) {
2503                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2504                         return (NULL);
2505                 }
2506
2507                 if ((bif->bif_flags & IFBIF_STP) &&
2508                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2509                         return (m);
2510                 }
2511
2512                 /*
2513                  * Make a deep copy of the packet and enqueue the copy
2514                  * for bridge processing; return the original packet for
2515                  * local processing.
2516                  */
2517                 mc = m_dup(m, M_NOWAIT);
2518                 if (mc == NULL) {
2519                         return (m);
2520                 }
2521
2522                 /* Perform the bridge forwarding function with the copy. */
2523                 bridge_forward(sc, bif, mc);
2524
2525                 /*
2526                  * Reinject the mbuf as arriving on the bridge so we have a
2527                  * chance at claiming multicast packets. We can not loop back
2528                  * here from ether_input as a bridge is never a member of a
2529                  * bridge.
2530                  */
2531                 KASSERT(bifp->if_bridge == NULL,
2532                     ("loop created in bridge_input"));
2533                 mc2 = m_dup(m, M_NOWAIT);
2534                 if (mc2 != NULL) {
2535                         /* Keep the layer3 header aligned */
2536                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2537                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2538                 }
2539                 if (mc2 != NULL) {
2540                         mc2->m_pkthdr.rcvif = bifp;
2541                         (*bifp->if_input)(bifp, mc2);
2542                 }
2543
2544                 /* Return the original packet for local processing. */
2545                 return (m);
2546         }
2547
2548         if ((bif->bif_flags & IFBIF_STP) &&
2549             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2550                 return (m);
2551         }
2552
2553 #if (defined(INET) || defined(INET6))
2554 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2555         || ((iface)->if_carp \
2556             && (*carp_forus_p)((iface), eh->ether_dhost))
2557 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2558         || ((iface)->if_carp \
2559             && (*carp_forus_p)((iface), eh->ether_shost))
2560 #else
2561 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2562 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2563 #endif
2564
2565 #ifdef INET6
2566 #   define OR_PFIL_HOOKED_INET6 \
2567         || PFIL_HOOKED_IN(V_inet6_pfil_head)
2568 #else
2569 #   define OR_PFIL_HOOKED_INET6
2570 #endif
2571
2572 #define GRAB_OUR_PACKETS(iface) \
2573         if ((iface)->if_type == IFT_GIF) \
2574                 continue; \
2575         /* It is destined for us. */ \
2576         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2577             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2578             ) {                                                         \
2579                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2580                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2581                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2582                         if (error && bif->bif_addrmax) {                \
2583                                 m_freem(m);                             \
2584                                 return (NULL);                          \
2585                         }                                               \
2586                 }                                                       \
2587                 m->m_pkthdr.rcvif = iface;                              \
2588                 if ((iface) == ifp) {                                   \
2589                         /* Skip bridge processing... src == dest */     \
2590                         return (m);                                     \
2591                 }                                                       \
2592                 /* It's passing over or to the bridge, locally. */      \
2593                 ETHER_BPF_MTAP(bifp, m);                                \
2594                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2595                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); \
2596                 /* Filter on the physical interface. */                 \
2597                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) \
2598                      OR_PFIL_HOOKED_INET6)) {                           \
2599                         if (bridge_pfil(&m, NULL, ifp,                  \
2600                             PFIL_IN) != 0 || m == NULL) {               \
2601                                 return (NULL);                          \
2602                         }                                               \
2603                 }                                                       \
2604                 if ((iface) != bifp)                                    \
2605                         ETHER_BPF_MTAP(iface, m);                       \
2606                 return (m);                                             \
2607         }                                                               \
2608                                                                         \
2609         /* We just received a packet that we sent out. */               \
2610         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2611             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2612             ) {                                                         \
2613                 m_freem(m);                                             \
2614                 return (NULL);                                          \
2615         }
2616
2617         /*
2618          * Unicast.  Make sure it's not for the bridge.
2619          */
2620         do { GRAB_OUR_PACKETS(bifp) } while (0);
2621
2622         /*
2623          * Give a chance for ifp at first priority. This will help when the
2624          * packet comes through the interface like VLAN's with the same MACs
2625          * on several interfaces from the same bridge. This also will save
2626          * some CPU cycles in case the destination interface and the input
2627          * interface (eq ifp) are the same.
2628          */
2629         do { GRAB_OUR_PACKETS(ifp) } while (0);
2630
2631         /* Now check the all bridge members. */
2632         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2633                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2634         }
2635
2636 #undef OR_CARP_CHECK_WE_ARE_DST
2637 #undef OR_CARP_CHECK_WE_ARE_SRC
2638 #undef OR_PFIL_HOOKED_INET6
2639 #undef GRAB_OUR_PACKETS
2640
2641         /* Perform the bridge forwarding function. */
2642         bridge_forward(sc, bif, m);
2643
2644         return (NULL);
2645 }
2646
2647 /*
2648  * bridge_broadcast:
2649  *
2650  *      Send a frame to all interfaces that are members of
2651  *      the bridge, except for the one on which the packet
2652  *      arrived.
2653  *
2654  *      NOTE: Releases the lock on return.
2655  */
2656 static void
2657 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2658     struct mbuf *m, int runfilt)
2659 {
2660         struct bridge_iflist *dbif, *sbif;
2661         struct mbuf *mc;
2662         struct ifnet *dst_if;
2663         int used = 0, i;
2664
2665         NET_EPOCH_ASSERT();
2666
2667         sbif = bridge_lookup_member_if(sc, src_if);
2668
2669         /* Filter on the bridge interface before broadcasting */
2670         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2671 #ifdef INET6
2672             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2673 #endif
2674             )) {
2675                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2676                         return;
2677                 if (m == NULL)
2678                         return;
2679         }
2680
2681         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2682                 dst_if = dbif->bif_ifp;
2683                 if (dst_if == src_if)
2684                         continue;
2685
2686                 /* Private segments can not talk to each other */
2687                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2688                         continue;
2689
2690                 if ((dbif->bif_flags & IFBIF_STP) &&
2691                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2692                         continue;
2693
2694                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2695                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2696                         continue;
2697
2698                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2699                         continue;
2700
2701                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2702                         mc = m;
2703                         used = 1;
2704                 } else {
2705                         mc = m_dup(m, M_NOWAIT);
2706                         if (mc == NULL) {
2707                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2708                                 continue;
2709                         }
2710                 }
2711
2712                 /*
2713                  * Filter on the output interface. Pass a NULL bridge interface
2714                  * pointer so we do not redundantly filter on the bridge for
2715                  * each interface we broadcast on.
2716                  */
2717                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2718 #ifdef INET6
2719                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2720 #endif
2721                     )) {
2722                         if (used == 0) {
2723                                 /* Keep the layer3 header aligned */
2724                                 i = min(mc->m_pkthdr.len, max_protohdr);
2725                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2726                                 if (mc == NULL) {
2727                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2728                                         continue;
2729                                 }
2730                         }
2731                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2732                                 continue;
2733                         if (mc == NULL)
2734                                 continue;
2735                 }
2736
2737                 bridge_enqueue(sc, dst_if, mc);
2738         }
2739         if (used == 0)
2740                 m_freem(m);
2741 }
2742
2743 /*
2744  * bridge_span:
2745  *
2746  *      Duplicate a packet out one or more interfaces that are in span mode,
2747  *      the original mbuf is unmodified.
2748  */
2749 static void
2750 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2751 {
2752         struct bridge_iflist *bif;
2753         struct ifnet *dst_if;
2754         struct mbuf *mc;
2755
2756         NET_EPOCH_ASSERT();
2757
2758         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2759                 return;
2760
2761         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2762                 dst_if = bif->bif_ifp;
2763
2764                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2765                         continue;
2766
2767                 mc = m_dup(m, M_NOWAIT);
2768                 if (mc == NULL) {
2769                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2770                         continue;
2771                 }
2772
2773                 bridge_enqueue(sc, dst_if, mc);
2774         }
2775 }
2776
2777 /*
2778  * bridge_rtupdate:
2779  *
2780  *      Add a bridge routing entry.
2781  */
2782 static int
2783 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2784     struct bridge_iflist *bif, int setflags, uint8_t flags)
2785 {
2786         struct bridge_rtnode *brt;
2787         int error;
2788
2789         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2790
2791         /* Check the source address is valid and not multicast. */
2792         if (ETHER_IS_MULTICAST(dst) ||
2793             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2794              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2795                 return (EINVAL);
2796
2797         /* 802.1p frames map to vlan 1 */
2798         if (vlan == 0)
2799                 vlan = 1;
2800
2801         /*
2802          * A route for this destination might already exist.  If so,
2803          * update it, otherwise create a new one.
2804          */
2805         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2806                 BRIDGE_RT_LOCK(sc);
2807
2808                 /* Check again, now that we have the lock. There could have
2809                  * been a race and we only want to insert this once. */
2810                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2811                         BRIDGE_RT_UNLOCK(sc);
2812                         return (0);
2813                 }
2814
2815                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2816                         sc->sc_brtexceeded++;
2817                         BRIDGE_RT_UNLOCK(sc);
2818                         return (ENOSPC);
2819                 }
2820                 /* Check per interface address limits (if enabled) */
2821                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2822                         bif->bif_addrexceeded++;
2823                         BRIDGE_RT_UNLOCK(sc);
2824                         return (ENOSPC);
2825                 }
2826
2827                 /*
2828                  * Allocate a new bridge forwarding node, and
2829                  * initialize the expiration time and Ethernet
2830                  * address.
2831                  */
2832                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2833                 if (brt == NULL) {
2834                         BRIDGE_RT_UNLOCK(sc);
2835                         return (ENOMEM);
2836                 }
2837                 brt->brt_vnet = curvnet;
2838
2839                 if (bif->bif_flags & IFBIF_STICKY)
2840                         brt->brt_flags = IFBAF_STICKY;
2841                 else
2842                         brt->brt_flags = IFBAF_DYNAMIC;
2843
2844                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2845                 brt->brt_vlan = vlan;
2846
2847                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2848                         uma_zfree(V_bridge_rtnode_zone, brt);
2849                         BRIDGE_RT_UNLOCK(sc);
2850                         return (error);
2851                 }
2852                 brt->brt_dst = bif;
2853                 bif->bif_addrcnt++;
2854
2855                 BRIDGE_RT_UNLOCK(sc);
2856         }
2857
2858         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2859             brt->brt_dst != bif) {
2860                 BRIDGE_RT_LOCK(sc);
2861                 brt->brt_dst->bif_addrcnt--;
2862                 brt->brt_dst = bif;
2863                 brt->brt_dst->bif_addrcnt++;
2864                 BRIDGE_RT_UNLOCK(sc);
2865         }
2866
2867         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2868                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2869         if (setflags)
2870                 brt->brt_flags = flags;
2871
2872         return (0);
2873 }
2874
2875 /*
2876  * bridge_rtlookup:
2877  *
2878  *      Lookup the destination interface for an address.
2879  */
2880 static struct ifnet *
2881 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2882 {
2883         struct bridge_rtnode *brt;
2884
2885         NET_EPOCH_ASSERT();
2886
2887         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2888                 return (NULL);
2889
2890         return (brt->brt_ifp);
2891 }
2892
2893 /*
2894  * bridge_rttrim:
2895  *
2896  *      Trim the routine table so that we have a number
2897  *      of routing entries less than or equal to the
2898  *      maximum number.
2899  */
2900 static void
2901 bridge_rttrim(struct bridge_softc *sc)
2902 {
2903         struct bridge_rtnode *brt, *nbrt;
2904
2905         NET_EPOCH_ASSERT();
2906         BRIDGE_RT_LOCK_ASSERT(sc);
2907
2908         /* Make sure we actually need to do this. */
2909         if (sc->sc_brtcnt <= sc->sc_brtmax)
2910                 return;
2911
2912         /* Force an aging cycle; this might trim enough addresses. */
2913         bridge_rtage(sc);
2914         if (sc->sc_brtcnt <= sc->sc_brtmax)
2915                 return;
2916
2917         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2918                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2919                         bridge_rtnode_destroy(sc, brt);
2920                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2921                                 return;
2922                 }
2923         }
2924 }
2925
2926 /*
2927  * bridge_timer:
2928  *
2929  *      Aging timer for the bridge.
2930  */
2931 static void
2932 bridge_timer(void *arg)
2933 {
2934         struct bridge_softc *sc = arg;
2935
2936         BRIDGE_RT_LOCK_ASSERT(sc);
2937
2938         /* Destruction of rtnodes requires a proper vnet context */
2939         CURVNET_SET(sc->sc_ifp->if_vnet);
2940         bridge_rtage(sc);
2941
2942         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2943                 callout_reset(&sc->sc_brcallout,
2944                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2945         CURVNET_RESTORE();
2946 }
2947
2948 /*
2949  * bridge_rtage:
2950  *
2951  *      Perform an aging cycle.
2952  */
2953 static void
2954 bridge_rtage(struct bridge_softc *sc)
2955 {
2956         struct bridge_rtnode *brt, *nbrt;
2957
2958         BRIDGE_RT_LOCK_ASSERT(sc);
2959
2960         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2961                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2962                         if (time_uptime >= brt->brt_expire)
2963                                 bridge_rtnode_destroy(sc, brt);
2964                 }
2965         }
2966 }
2967
2968 /*
2969  * bridge_rtflush:
2970  *
2971  *      Remove all dynamic addresses from the bridge.
2972  */
2973 static void
2974 bridge_rtflush(struct bridge_softc *sc, int full)
2975 {
2976         struct bridge_rtnode *brt, *nbrt;
2977
2978         BRIDGE_RT_LOCK_ASSERT(sc);
2979
2980         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2981                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2982                         bridge_rtnode_destroy(sc, brt);
2983         }
2984 }
2985
2986 /*
2987  * bridge_rtdaddr:
2988  *
2989  *      Remove an address from the table.
2990  */
2991 static int
2992 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2993 {
2994         struct bridge_rtnode *brt;
2995         int found = 0;
2996
2997         BRIDGE_RT_LOCK(sc);
2998
2999         /*
3000          * If vlan is zero then we want to delete for all vlans so the lookup
3001          * may return more than one.
3002          */
3003         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3004                 bridge_rtnode_destroy(sc, brt);
3005                 found = 1;
3006         }
3007
3008         BRIDGE_RT_UNLOCK(sc);
3009
3010         return (found ? 0 : ENOENT);
3011 }
3012
3013 /*
3014  * bridge_rtdelete:
3015  *
3016  *      Delete routes to a speicifc member interface.
3017  */
3018 static void
3019 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3020 {
3021         struct bridge_rtnode *brt, *nbrt;
3022
3023         BRIDGE_RT_LOCK_ASSERT(sc);
3024
3025         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3026                 if (brt->brt_ifp == ifp && (full ||
3027                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3028                         bridge_rtnode_destroy(sc, brt);
3029         }
3030 }
3031
3032 /*
3033  * bridge_rtable_init:
3034  *
3035  *      Initialize the route table for this bridge.
3036  */
3037 static void
3038 bridge_rtable_init(struct bridge_softc *sc)
3039 {
3040         int i;
3041
3042         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3043             M_DEVBUF, M_WAITOK);
3044
3045         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3046                 CK_LIST_INIT(&sc->sc_rthash[i]);
3047
3048         sc->sc_rthash_key = arc4random();
3049         CK_LIST_INIT(&sc->sc_rtlist);
3050 }
3051
3052 /*
3053  * bridge_rtable_fini:
3054  *
3055  *      Deconstruct the route table for this bridge.
3056  */
3057 static void
3058 bridge_rtable_fini(struct bridge_softc *sc)
3059 {
3060
3061         KASSERT(sc->sc_brtcnt == 0,
3062             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3063         free(sc->sc_rthash, M_DEVBUF);
3064 }
3065
3066 /*
3067  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3068  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3069  */
3070 #define mix(a, b, c)                                                    \
3071 do {                                                                    \
3072         a -= b; a -= c; a ^= (c >> 13);                                 \
3073         b -= c; b -= a; b ^= (a << 8);                                  \
3074         c -= a; c -= b; c ^= (b >> 13);                                 \
3075         a -= b; a -= c; a ^= (c >> 12);                                 \
3076         b -= c; b -= a; b ^= (a << 16);                                 \
3077         c -= a; c -= b; c ^= (b >> 5);                                  \
3078         a -= b; a -= c; a ^= (c >> 3);                                  \
3079         b -= c; b -= a; b ^= (a << 10);                                 \
3080         c -= a; c -= b; c ^= (b >> 15);                                 \
3081 } while (/*CONSTCOND*/0)
3082
3083 static __inline uint32_t
3084 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3085 {
3086         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3087
3088         b += addr[5] << 8;
3089         b += addr[4];
3090         a += addr[3] << 24;
3091         a += addr[2] << 16;
3092         a += addr[1] << 8;
3093         a += addr[0];
3094
3095         mix(a, b, c);
3096
3097         return (c & BRIDGE_RTHASH_MASK);
3098 }
3099
3100 #undef mix
3101
3102 static int
3103 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3104 {
3105         int i, d;
3106
3107         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3108                 d = ((int)a[i]) - ((int)b[i]);
3109         }
3110
3111         return (d);
3112 }
3113
3114 /*
3115  * bridge_rtnode_lookup:
3116  *
3117  *      Look up a bridge route node for the specified destination. Compare the
3118  *      vlan id or if zero then just return the first match.
3119  */
3120 static struct bridge_rtnode *
3121 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3122 {
3123         struct bridge_rtnode *brt;
3124         uint32_t hash;
3125         int dir;
3126
3127         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3128
3129         hash = bridge_rthash(sc, addr);
3130         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3131                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3132                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3133                         return (brt);
3134                 if (dir > 0)
3135                         return (NULL);
3136         }
3137
3138         return (NULL);
3139 }
3140
3141 /*
3142  * bridge_rtnode_insert:
3143  *
3144  *      Insert the specified bridge node into the route table.  We
3145  *      assume the entry is not already in the table.
3146  */
3147 static int
3148 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3149 {
3150         struct bridge_rtnode *lbrt;
3151         uint32_t hash;
3152         int dir;
3153
3154         BRIDGE_RT_LOCK_ASSERT(sc);
3155
3156         hash = bridge_rthash(sc, brt->brt_addr);
3157
3158         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3159         if (lbrt == NULL) {
3160                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3161                 goto out;
3162         }
3163
3164         do {
3165                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3166                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3167                         return (EEXIST);
3168                 if (dir > 0) {
3169                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3170                         goto out;
3171                 }
3172                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3173                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3174                         goto out;
3175                 }
3176                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3177         } while (lbrt != NULL);
3178
3179 #ifdef DIAGNOSTIC
3180         panic("bridge_rtnode_insert: impossible");
3181 #endif
3182
3183 out:
3184         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3185         sc->sc_brtcnt++;
3186
3187         return (0);
3188 }
3189
3190 static void
3191 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3192 {
3193         struct bridge_rtnode *brt;
3194
3195         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3196
3197         CURVNET_SET(brt->brt_vnet);
3198         uma_zfree(V_bridge_rtnode_zone, brt);
3199         CURVNET_RESTORE();
3200 }
3201
3202 /*
3203  * bridge_rtnode_destroy:
3204  *
3205  *      Destroy a bridge rtnode.
3206  */
3207 static void
3208 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3209 {
3210         BRIDGE_RT_LOCK_ASSERT(sc);
3211
3212         CK_LIST_REMOVE(brt, brt_hash);
3213
3214         CK_LIST_REMOVE(brt, brt_list);
3215         sc->sc_brtcnt--;
3216         brt->brt_dst->bif_addrcnt--;
3217
3218         NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3219 }
3220
3221 /*
3222  * bridge_rtable_expire:
3223  *
3224  *      Set the expiry time for all routes on an interface.
3225  */
3226 static void
3227 bridge_rtable_expire(struct ifnet *ifp, int age)
3228 {
3229         struct bridge_softc *sc = ifp->if_bridge;
3230         struct bridge_rtnode *brt;
3231
3232         CURVNET_SET(ifp->if_vnet);
3233         BRIDGE_RT_LOCK(sc);
3234
3235         /*
3236          * If the age is zero then flush, otherwise set all the expiry times to
3237          * age for the interface
3238          */
3239         if (age == 0)
3240                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3241         else {
3242                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3243                         /* Cap the expiry time to 'age' */
3244                         if (brt->brt_ifp == ifp &&
3245                             brt->brt_expire > time_uptime + age &&
3246                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3247                                 brt->brt_expire = time_uptime + age;
3248                 }
3249         }
3250         BRIDGE_RT_UNLOCK(sc);
3251         CURVNET_RESTORE();
3252 }
3253
3254 /*
3255  * bridge_state_change:
3256  *
3257  *      Callback from the bridgestp code when a port changes states.
3258  */
3259 static void
3260 bridge_state_change(struct ifnet *ifp, int state)
3261 {
3262         struct bridge_softc *sc = ifp->if_bridge;
3263         static const char *stpstates[] = {
3264                 "disabled",
3265                 "listening",
3266                 "learning",
3267                 "forwarding",
3268                 "blocking",
3269                 "discarding"
3270         };
3271
3272         CURVNET_SET(ifp->if_vnet);
3273         if (V_log_stp)
3274                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3275                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3276         CURVNET_RESTORE();
3277 }
3278
3279 /*
3280  * Send bridge packets through pfil if they are one of the types pfil can deal
3281  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3282  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3283  * that interface.
3284  */
3285 static int
3286 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3287 {
3288         int snap, error, i, hlen;
3289         struct ether_header *eh1, eh2;
3290         struct ip *ip;
3291         struct llc llc1;
3292         u_int16_t ether_type;
3293         pfil_return_t rv;
3294
3295         snap = 0;
3296         error = -1;     /* Default error if not error == 0 */
3297
3298 #if 0
3299         /* we may return with the IP fields swapped, ensure its not shared */
3300         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3301 #endif
3302
3303         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3304                 return (0); /* filtering is disabled */
3305
3306         i = min((*mp)->m_pkthdr.len, max_protohdr);
3307         if ((*mp)->m_len < i) {
3308             *mp = m_pullup(*mp, i);
3309             if (*mp == NULL) {
3310                 printf("%s: m_pullup failed\n", __func__);
3311                 return (-1);
3312             }
3313         }
3314
3315         eh1 = mtod(*mp, struct ether_header *);
3316         ether_type = ntohs(eh1->ether_type);
3317
3318         /*
3319          * Check for SNAP/LLC.
3320          */
3321         if (ether_type < ETHERMTU) {
3322                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3323
3324                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3325                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3326                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3327                     llc2->llc_control == LLC_UI) {
3328                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3329                         snap = 1;
3330                 }
3331         }
3332
3333         /*
3334          * If we're trying to filter bridge traffic, don't look at anything
3335          * other than IP and ARP traffic.  If the filter doesn't understand
3336          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3337          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3338          * but of course we don't have an AppleTalk filter to begin with.
3339          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3340          * ARP traffic.)
3341          */
3342         switch (ether_type) {
3343                 case ETHERTYPE_ARP:
3344                 case ETHERTYPE_REVARP:
3345                         if (V_pfil_ipfw_arp == 0)
3346                                 return (0); /* Automatically pass */
3347                         break;
3348
3349                 case ETHERTYPE_IP:
3350 #ifdef INET6
3351                 case ETHERTYPE_IPV6:
3352 #endif /* INET6 */
3353                         break;
3354                 default:
3355                         /*
3356                          * Check to see if the user wants to pass non-ip
3357                          * packets, these will not be checked by pfil(9) and
3358                          * passed unconditionally so the default is to drop.
3359                          */
3360                         if (V_pfil_onlyip)
3361                                 goto bad;
3362         }
3363
3364         /* Run the packet through pfil before stripping link headers */
3365         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3366             dir == PFIL_OUT && ifp != NULL) {
3367                 switch (pfil_run_hooks(V_link_pfil_head, mp, ifp, dir, NULL)) {
3368                 case PFIL_DROPPED:
3369                         return (EACCES);
3370                 case PFIL_CONSUMED:
3371                         return (0);
3372                 }
3373         }
3374
3375         /* Strip off the Ethernet header and keep a copy. */
3376         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3377         m_adj(*mp, ETHER_HDR_LEN);
3378
3379         /* Strip off snap header, if present */
3380         if (snap) {
3381                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3382                 m_adj(*mp, sizeof(struct llc));
3383         }
3384
3385         /*
3386          * Check the IP header for alignment and errors
3387          */
3388         if (dir == PFIL_IN) {
3389                 switch (ether_type) {
3390                         case ETHERTYPE_IP:
3391                                 error = bridge_ip_checkbasic(mp);
3392                                 break;
3393 #ifdef INET6
3394                         case ETHERTYPE_IPV6:
3395                                 error = bridge_ip6_checkbasic(mp);
3396                                 break;
3397 #endif /* INET6 */
3398                         default:
3399                                 error = 0;
3400                 }
3401                 if (error)
3402                         goto bad;
3403         }
3404
3405         error = 0;
3406
3407         /*
3408          * Run the packet through pfil
3409          */
3410         rv = PFIL_PASS;
3411         switch (ether_type) {
3412         case ETHERTYPE_IP:
3413                 /*
3414                  * Run pfil on the member interface and the bridge, both can
3415                  * be skipped by clearing pfil_member or pfil_bridge.
3416                  *
3417                  * Keep the order:
3418                  *   in_if -> bridge_if -> out_if
3419                  */
3420                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3421                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3422                     PFIL_PASS)
3423                         break;
3424
3425                 if (V_pfil_member && ifp != NULL && (rv =
3426                     pfil_run_hooks(V_inet_pfil_head, mp, ifp, dir, NULL)) !=
3427                     PFIL_PASS)
3428                         break;
3429
3430                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3431                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3432                     PFIL_PASS)
3433                         break;
3434
3435                 /* check if we need to fragment the packet */
3436                 /* bridge_fragment generates a mbuf chain of packets */
3437                 /* that already include eth headers */
3438                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3439                         i = (*mp)->m_pkthdr.len;
3440                         if (i > ifp->if_mtu) {
3441                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3442                                             &llc1);
3443                                 return (error);
3444                         }
3445                 }
3446
3447                 /* Recalculate the ip checksum. */
3448                 ip = mtod(*mp, struct ip *);
3449                 hlen = ip->ip_hl << 2;
3450                 if (hlen < sizeof(struct ip))
3451                         goto bad;
3452                 if (hlen > (*mp)->m_len) {
3453                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3454                                 goto bad;
3455                         ip = mtod(*mp, struct ip *);
3456                         if (ip == NULL)
3457                                 goto bad;
3458                 }
3459                 ip->ip_sum = 0;
3460                 if (hlen == sizeof(struct ip))
3461                         ip->ip_sum = in_cksum_hdr(ip);
3462                 else
3463                         ip->ip_sum = in_cksum(*mp, hlen);
3464
3465                 break;
3466 #ifdef INET6
3467         case ETHERTYPE_IPV6:
3468                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3469                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3470                     PFIL_PASS)
3471                         break;
3472
3473                 if (V_pfil_member && ifp != NULL && (rv =
3474                     pfil_run_hooks(V_inet6_pfil_head, mp, ifp, dir, NULL)) !=
3475                     PFIL_PASS)
3476                         break;
3477
3478                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3479                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3480                     PFIL_PASS)
3481                         break;
3482                 break;
3483 #endif
3484         }
3485
3486         switch (rv) {
3487         case PFIL_CONSUMED:
3488                 return (0);
3489         case PFIL_DROPPED:
3490                 return (EACCES);
3491         default:
3492                 break;
3493         }
3494
3495         error = -1;
3496
3497         /*
3498          * Finally, put everything back the way it was and return
3499          */
3500         if (snap) {
3501                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3502                 if (*mp == NULL)
3503                         return (error);
3504                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3505         }
3506
3507         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3508         if (*mp == NULL)
3509                 return (error);
3510         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3511
3512         return (0);
3513
3514 bad:
3515         m_freem(*mp);
3516         *mp = NULL;
3517         return (error);
3518 }
3519
3520 /*
3521  * Perform basic checks on header size since
3522  * pfil assumes ip_input has already processed
3523  * it for it.  Cut-and-pasted from ip_input.c.
3524  * Given how simple the IPv6 version is,
3525  * does the IPv4 version really need to be
3526  * this complicated?
3527  *
3528  * XXX Should we update ipstat here, or not?
3529  * XXX Right now we update ipstat but not
3530  * XXX csum_counter.
3531  */
3532 static int
3533 bridge_ip_checkbasic(struct mbuf **mp)
3534 {
3535         struct mbuf *m = *mp;
3536         struct ip *ip;
3537         int len, hlen;
3538         u_short sum;
3539
3540         if (*mp == NULL)
3541                 return (-1);
3542
3543         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3544                 if ((m = m_copyup(m, sizeof(struct ip),
3545                         (max_linkhdr + 3) & ~3)) == NULL) {
3546                         /* XXXJRT new stat, please */
3547                         KMOD_IPSTAT_INC(ips_toosmall);
3548                         goto bad;
3549                 }
3550         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3551                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3552                         KMOD_IPSTAT_INC(ips_toosmall);
3553                         goto bad;
3554                 }
3555         }
3556         ip = mtod(m, struct ip *);
3557         if (ip == NULL) goto bad;
3558
3559         if (ip->ip_v != IPVERSION) {
3560                 KMOD_IPSTAT_INC(ips_badvers);
3561                 goto bad;
3562         }
3563         hlen = ip->ip_hl << 2;
3564         if (hlen < sizeof(struct ip)) { /* minimum header length */
3565                 KMOD_IPSTAT_INC(ips_badhlen);
3566                 goto bad;
3567         }
3568         if (hlen > m->m_len) {
3569                 if ((m = m_pullup(m, hlen)) == NULL) {
3570                         KMOD_IPSTAT_INC(ips_badhlen);
3571                         goto bad;
3572                 }
3573                 ip = mtod(m, struct ip *);
3574                 if (ip == NULL) goto bad;
3575         }
3576
3577         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3578                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3579         } else {
3580                 if (hlen == sizeof(struct ip)) {
3581                         sum = in_cksum_hdr(ip);
3582                 } else {
3583                         sum = in_cksum(m, hlen);
3584                 }
3585         }
3586         if (sum) {
3587                 KMOD_IPSTAT_INC(ips_badsum);
3588                 goto bad;
3589         }
3590
3591         /* Retrieve the packet length. */
3592         len = ntohs(ip->ip_len);
3593
3594         /*
3595          * Check for additional length bogosity
3596          */
3597         if (len < hlen) {
3598                 KMOD_IPSTAT_INC(ips_badlen);
3599                 goto bad;
3600         }
3601
3602         /*
3603          * Check that the amount of data in the buffers
3604          * is as at least much as the IP header would have us expect.
3605          * Drop packet if shorter than we expect.
3606          */
3607         if (m->m_pkthdr.len < len) {
3608                 KMOD_IPSTAT_INC(ips_tooshort);
3609                 goto bad;
3610         }
3611
3612         /* Checks out, proceed */
3613         *mp = m;
3614         return (0);
3615
3616 bad:
3617         *mp = m;
3618         return (-1);
3619 }
3620
3621 #ifdef INET6
3622 /*
3623  * Same as above, but for IPv6.
3624  * Cut-and-pasted from ip6_input.c.
3625  * XXX Should we update ip6stat, or not?
3626  */
3627 static int
3628 bridge_ip6_checkbasic(struct mbuf **mp)
3629 {
3630         struct mbuf *m = *mp;
3631         struct ip6_hdr *ip6;
3632
3633         /*
3634          * If the IPv6 header is not aligned, slurp it up into a new
3635          * mbuf with space for link headers, in the event we forward
3636          * it.  Otherwise, if it is aligned, make sure the entire base
3637          * IPv6 header is in the first mbuf of the chain.
3638          */
3639         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3640                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3641                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3642                             (max_linkhdr + 3) & ~3)) == NULL) {
3643                         /* XXXJRT new stat, please */
3644                         IP6STAT_INC(ip6s_toosmall);
3645                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3646                         goto bad;
3647                 }
3648         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3649                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3650                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3651                         IP6STAT_INC(ip6s_toosmall);
3652                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3653                         goto bad;
3654                 }
3655         }
3656
3657         ip6 = mtod(m, struct ip6_hdr *);
3658
3659         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3660                 IP6STAT_INC(ip6s_badvers);
3661                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3662                 goto bad;
3663         }
3664
3665         /* Checks out, proceed */
3666         *mp = m;
3667         return (0);
3668
3669 bad:
3670         *mp = m;
3671         return (-1);
3672 }
3673 #endif /* INET6 */
3674
3675 /*
3676  * bridge_fragment:
3677  *
3678  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3679  */
3680 static int
3681 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3682     int snap, struct llc *llc)
3683 {
3684         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3685         struct ip *ip;
3686         int error = -1;
3687
3688         if (m->m_len < sizeof(struct ip) &&
3689             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3690                 goto dropit;
3691         ip = mtod(m, struct ip *);
3692
3693         m->m_pkthdr.csum_flags |= CSUM_IP;
3694         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3695         if (error)
3696                 goto dropit;
3697
3698         /*
3699          * Walk the chain and re-add the Ethernet header for
3700          * each mbuf packet.
3701          */
3702         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3703                 nextpkt = mcur->m_nextpkt;
3704                 mcur->m_nextpkt = NULL;
3705                 if (snap) {
3706                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3707                         if (mcur == NULL) {
3708                                 error = ENOBUFS;
3709                                 if (mprev != NULL)
3710                                         mprev->m_nextpkt = nextpkt;
3711                                 goto dropit;
3712                         }
3713                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3714                 }
3715
3716                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3717                 if (mcur == NULL) {
3718                         error = ENOBUFS;
3719                         if (mprev != NULL)
3720                                 mprev->m_nextpkt = nextpkt;
3721                         goto dropit;
3722                 }
3723                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3724
3725                 /*
3726                  * The previous two M_PREPEND could have inserted one or two
3727                  * mbufs in front so we have to update the previous packet's
3728                  * m_nextpkt.
3729                  */
3730                 mcur->m_nextpkt = nextpkt;
3731                 if (mprev != NULL)
3732                         mprev->m_nextpkt = mcur;
3733                 else {
3734                         /* The first mbuf in the original chain needs to be
3735                          * updated. */
3736                         *mp = mcur;
3737                 }
3738                 mprev = mcur;
3739         }
3740
3741         KMOD_IPSTAT_INC(ips_fragmented);
3742         return (error);
3743
3744 dropit:
3745         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3746                 m = mcur->m_nextpkt;
3747                 m_freem(mcur);
3748         }
3749         return (error);
3750 }
3751
3752 static void
3753 bridge_linkstate(struct ifnet *ifp)
3754 {
3755         struct bridge_softc *sc = ifp->if_bridge;
3756         struct bridge_iflist *bif;
3757         struct epoch_tracker et;
3758
3759         NET_EPOCH_ENTER(et);
3760
3761         bif = bridge_lookup_member_if(sc, ifp);
3762         if (bif == NULL) {
3763                 NET_EPOCH_EXIT(et);
3764                 return;
3765         }
3766         bridge_linkcheck(sc);
3767
3768         bstp_linkstate(&bif->bif_stp);
3769
3770         NET_EPOCH_EXIT(et);
3771 }
3772
3773 static void
3774 bridge_linkcheck(struct bridge_softc *sc)
3775 {
3776         struct bridge_iflist *bif;
3777         int new_link, hasls;
3778
3779         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3780
3781         new_link = LINK_STATE_DOWN;
3782         hasls = 0;
3783         /* Our link is considered up if at least one of our ports is active */
3784         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3785                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3786                         hasls++;
3787                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3788                         new_link = LINK_STATE_UP;
3789                         break;
3790                 }
3791         }
3792         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3793                 /* If no interfaces support link-state then we default to up */
3794                 new_link = LINK_STATE_UP;
3795         }
3796         if_link_state_change(sc->sc_ifp, new_link);
3797 }