net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*-
   4  * SPDX-License-Identifier: BSD-4-Clause
   5  *
   6  * Copyright 2001 Wasabi Systems, Inc.
   7  * All rights reserved.
   8  *
   9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed for the NetBSD Project by
  22  *      Wasabi Systems, Inc.
  23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  24  *    or promote products derived from this software without specific prior
  25  *    written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 /*
  41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  42  * All rights reserved.
  43  *
  44  * Redistribution and use in source and binary forms, with or without
  45  * modification, are permitted provided that the following conditions
  46  * are met:
  47  * 1. Redistributions of source code must retain the above copyright
  48  *    notice, this list of conditions and the following disclaimer.
  49  * 2. Redistributions in binary form must reproduce the above copyright
  50  *    notice, this list of conditions and the following disclaimer in the
  51  *    documentation and/or other materials provided with the distribution.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  63  * POSSIBILITY OF SUCH DAMAGE.
  64  *
  65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  66  */
  67
  68 /*
  69  * Network interface bridge support.
  70  *
  71  * TODO:
  72  *
  73  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  74  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  75  *        to bridge other types of interfaces (maybe consider
  76  *        heterogeneous bridges).
  77  */
  78
  79 #include <sys/cdefs.h>
  80 __FBSDID("$FreeBSD$");
  81
  82 #include "opt_inet.h"
  83 #include "opt_inet6.h"
  84
  85 #include <sys/param.h>
  86 #include <sys/eventhandler.h>
  87 #include <sys/mbuf.h>
  88 #include <sys/malloc.h>
  89 #include <sys/protosw.h>
  90 #include <sys/systm.h>
  91 #include <sys/jail.h>
  92 #include <sys/time.h>
  93 #include <sys/socket.h> /* for net/if.h */
  94 #include <sys/sockio.h>
  95 #include <sys/ctype.h>  /* string functions */
  96 #include <sys/kernel.h>
  97 #include <sys/random.h>
  98 #include <sys/syslog.h>
  99 #include <sys/sysctl.h>
 100 #include <vm/uma.h>
 101 #include <sys/module.h>
 102 #include <sys/priv.h>
 103 #include <sys/proc.h>
 104 #include <sys/lock.h>
 105 #include <sys/mutex.h>
 106
 107 #include <net/bpf.h>
 108 #include <net/if.h>
 109 #include <net/if_clone.h>
 110 #include <net/if_dl.h>
 111 #include <net/if_types.h>
 112 #include <net/if_var.h>
 113 #include <net/pfil.h>
 114 #include <net/vnet.h>
 115
 116 #include <netinet/in.h>
 117 #include <netinet/in_systm.h>
 118 #include <netinet/in_var.h>
 119 #include <netinet/ip.h>
 120 #include <netinet/ip_var.h>
 121 #ifdef INET6
 122 #include <netinet/ip6.h>
 123 #include <netinet6/ip6_var.h>
 124 #include <netinet6/in6_ifattach.h>
 125 #endif
 126 #if defined(INET) || defined(INET6)
 127 #include <netinet/ip_carp.h>
 128 #endif
 129 #include <machine/in_cksum.h>
 130 #include <netinet/if_ether.h>
 131 #include <net/bridgestp.h>
 132 #include <net/if_bridgevar.h>
 133 #include <net/if_llc.h>
 134 #include <net/if_vlan_var.h>
 135
 136 #include <net/route.h>
 137
 138 #ifdef INET6
 139 /*
 140  * XXX: declare here to avoid to include many inet6 related files..
 141  * should be more generalized?
 142  */
 143 extern void     nd6_setmtu(struct ifnet *);
 144 #endif
 145
 146 /*
 147  * Size of the route hash table.  Must be a power of two.
 148  */
 149 #ifndef BRIDGE_RTHASH_SIZE
 150 #define BRIDGE_RTHASH_SIZE              1024
 151 #endif
 152
 153 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 154
 155 /*
 156  * Default maximum number of addresses to cache.
 157  */
 158 #ifndef BRIDGE_RTABLE_MAX
 159 #define BRIDGE_RTABLE_MAX               2000
 160 #endif
 161
 162 /*
 163  * Timeout (in seconds) for entries learned dynamically.
 164  */
 165 #ifndef BRIDGE_RTABLE_TIMEOUT
 166 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 167 #endif
 168
 169 /*
 170  * Number of seconds between walks of the route list.
 171  */
 172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 173 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 174 #endif
 175
 176 /*
 177  * List of capabilities to possibly mask on the member interface.
 178  */
 179 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 180                                          IFCAP_TXCSUM_IPV6)
 181
 182 /*
 183  * List of capabilities to strip
 184  */
 185 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 186
 187 /*
 188  * Bridge locking
 189  *
 190  * The bridge relies heavily on the epoch(9) system to protect its data
 191  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 192  * must ensure there is only one writer at a time.
 193  *
 194  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 195  * accesses we must hold:
 196  *
 197  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 198  *  - BRIDGE_LOCK, for any other change
 199  *
 200  * The BRIDGE_LOCK is a sleepable lock, because it is held accross ioctl()
 201  * calls to bridge member interfaces and these ioctl()s can sleep.
 202  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 203  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 204  */
 205 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 206         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 207         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 208 } while (0)
 209 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 210         sx_destroy(&(_sc)->sc_sx);              \
 211         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 212 } while (0)
 213 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 214 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 215 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 216 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 217             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 218 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 219 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 220 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 221 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 222 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 223             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 224
 225 /*
 226  * Bridge interface list entry.
 227  */
 228 struct bridge_iflist {
 229         CK_LIST_ENTRY(bridge_iflist) bif_next;
 230         struct ifnet            *bif_ifp;       /* member if */
 231         struct bstp_port        bif_stp;        /* STP state */
 232         uint32_t                bif_flags;      /* member if flags */
 233         int                     bif_savedcaps;  /* saved capabilities */
 234         uint32_t                bif_addrmax;    /* max # of addresses */
 235         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 236         uint32_t                bif_addrexceeded;/* # of address violations */
 237
 238         struct epoch_context    bif_epoch_ctx;
 239 };
 240
 241 /*
 242  * Bridge route node.
 243  */
 244 struct bridge_rtnode {
 245         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 246         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 247         struct bridge_iflist    *brt_dst;       /* destination if */
 248         unsigned long           brt_expire;     /* expiration time */
 249         uint8_t                 brt_flags;      /* address flags */
 250         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 251         uint16_t                brt_vlan;       /* vlan id */
 252
 253         struct  vnet            *brt_vnet;
 254         struct  epoch_context   brt_epoch_ctx;
 255 };
 256 #define brt_ifp                 brt_dst->bif_ifp
 257
 258 /*
 259  * Software state for each bridge.
 260  */
 261 struct bridge_softc {
 262         struct ifnet            *sc_ifp;        /* make this an interface */
 263         LIST_ENTRY(bridge_softc) sc_list;
 264         struct sx               sc_sx;
 265         struct mtx              sc_rt_mtx;
 266         uint32_t                sc_brtmax;      /* max # of addresses */
 267         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 268         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 269         struct callout          sc_brcallout;   /* bridge callout */
 270         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 271         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 272         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 273         uint32_t                sc_rthash_key;  /* key for hash */
 274         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 275         struct bstp_state       sc_stp;         /* STP state */
 276         uint32_t                sc_brtexceeded; /* # of cache drops */
 277         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 278         struct ether_addr       sc_defaddr;     /* Default MAC address */
 279
 280         struct epoch_context    sc_epoch_ctx;
 281 };
 282
 283 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 284 #define V_bridge_list_sx        VNET(bridge_list_sx)
 285 static eventhandler_tag bridge_detach_cookie;
 286
 287 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 288
 289 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 290 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 291
 292 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 293 static void     bridge_clone_destroy(struct ifnet *);
 294
 295 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 296 static void     bridge_mutecaps(struct bridge_softc *);
 297 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 298                     int);
 299 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 300 static void     bridge_init(void *);
 301 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 302 static void     bridge_stop(struct ifnet *, int);
 303 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 304 static void     bridge_qflush(struct ifnet *);
 305 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 306 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 307                     struct rtentry *);
 308 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 309                     struct mbuf *);
 310 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 311
 312 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 313                     struct mbuf *m);
 314
 315 static void     bridge_timer(void *);
 316
 317 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 318                     struct mbuf *, int);
 319 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 320
 321 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 322                     uint16_t, struct bridge_iflist *, int, uint8_t);
 323 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 324                     uint16_t);
 325 static void     bridge_rttrim(struct bridge_softc *);
 326 static void     bridge_rtage(struct bridge_softc *);
 327 static void     bridge_rtflush(struct bridge_softc *, int);
 328 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 329                     uint16_t);
 330
 331 static void     bridge_rtable_init(struct bridge_softc *);
 332 static void     bridge_rtable_fini(struct bridge_softc *);
 333
 334 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 335 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 336                     const uint8_t *, uint16_t);
 337 static int      bridge_rtnode_insert(struct bridge_softc *,
 338                     struct bridge_rtnode *);
 339 static void     bridge_rtnode_destroy(struct bridge_softc *,
 340                     struct bridge_rtnode *);
 341 static void     bridge_rtable_expire(struct ifnet *, int);
 342 static void     bridge_state_change(struct ifnet *, int);
 343
 344 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 345                     const char *name);
 346 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 347                     struct ifnet *ifp);
 348 static void     bridge_delete_member(struct bridge_softc *,
 349                     struct bridge_iflist *, int);
 350 static void     bridge_delete_span(struct bridge_softc *,
 351                     struct bridge_iflist *);
 352
 353 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 354 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 381 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 382 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 383 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 384 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 385                     int);
 386 static int      bridge_ip_checkbasic(struct mbuf **mp);
 387 #ifdef INET6
 388 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 389 #endif /* INET6 */
 390 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 391                     struct ether_header *, int, struct llc *);
 392 static void     bridge_linkstate(struct ifnet *ifp);
 393 static void     bridge_linkcheck(struct bridge_softc *sc);
 394
 395
 396 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 397 #define VLANTAGOF(_m)   \
 398     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 399
 400 static struct bstp_cb_ops bridge_ops = {
 401         .bcb_state = bridge_state_change,
 402         .bcb_rtage = bridge_rtable_expire
 403 };
 404
 405 SYSCTL_DECL(_net_link);
 406 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 407
 408 /* only pass IP[46] packets when pfil is enabled */
 409 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 410 #define V_pfil_onlyip   VNET(pfil_onlyip)
 411 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 412     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 413     "Only pass IP packets when pfil is enabled");
 414
 415 /* run pfil hooks on the bridge interface */
 416 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 417 #define V_pfil_bridge   VNET(pfil_bridge)
 418 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 419     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 420     "Packet filter on the bridge interface");
 421
 422 /* layer2 filter with ipfw */
 423 VNET_DEFINE_STATIC(int, pfil_ipfw);
 424 #define V_pfil_ipfw     VNET(pfil_ipfw)
 425
 426 /* layer2 ARP filter with ipfw */
 427 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 428 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 429 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 430     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 431     "Filter ARP packets through IPFW layer2");
 432
 433 /* run pfil hooks on the member interface */
 434 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 435 #define V_pfil_member   VNET(pfil_member)
 436 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 437     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 438     "Packet filter on the member interface");
 439
 440 /* run pfil hooks on the physical interface for locally destined packets */
 441 VNET_DEFINE_STATIC(int, pfil_local_phys);
 442 #define V_pfil_local_phys       VNET(pfil_local_phys)
 443 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 444     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 445     "Packet filter on the physical interface for locally destined packets");
 446
 447 /* log STP state changes */
 448 VNET_DEFINE_STATIC(int, log_stp);
 449 #define V_log_stp       VNET(log_stp)
 450 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 451     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 452     "Log STP state changes");
 453
 454 /* share MAC with first bridge member */
 455 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 456 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 457 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 458     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 459     "Inherit MAC address from the first bridge member");
 460
 461 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 462 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 463 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 464     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 465     "Allow overlap of link-local scope "
 466     "zones of a bridge interface and the member interfaces");
 467
 468 struct bridge_control {
 469         int     (*bc_func)(struct bridge_softc *, void *);
 470         int     bc_argsize;
 471         int     bc_flags;
 472 };
 473
 474 #define BC_F_COPYIN             0x01    /* copy arguments in */
 475 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 476 #define BC_F_SUSER              0x04    /* do super-user check */
 477
 478 const struct bridge_control bridge_control_table[] = {
 479         { bridge_ioctl_add,             sizeof(struct ifbreq),
 480           BC_F_COPYIN|BC_F_SUSER },
 481         { bridge_ioctl_del,             sizeof(struct ifbreq),
 482           BC_F_COPYIN|BC_F_SUSER },
 483
 484         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 485           BC_F_COPYIN|BC_F_COPYOUT },
 486         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 487           BC_F_COPYIN|BC_F_SUSER },
 488
 489         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 490           BC_F_COPYIN|BC_F_SUSER },
 491         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 492           BC_F_COPYOUT },
 493
 494         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 495           BC_F_COPYIN|BC_F_COPYOUT },
 496         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 497           BC_F_COPYIN|BC_F_COPYOUT },
 498
 499         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 500           BC_F_COPYIN|BC_F_SUSER },
 501
 502         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 503           BC_F_COPYIN|BC_F_SUSER },
 504         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 505           BC_F_COPYOUT },
 506
 507         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 508           BC_F_COPYIN|BC_F_SUSER },
 509
 510         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 511           BC_F_COPYIN|BC_F_SUSER },
 512
 513         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 514           BC_F_COPYOUT },
 515         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 516           BC_F_COPYIN|BC_F_SUSER },
 517
 518         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 519           BC_F_COPYOUT },
 520         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 521           BC_F_COPYIN|BC_F_SUSER },
 522
 523         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 524           BC_F_COPYOUT },
 525         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 526           BC_F_COPYIN|BC_F_SUSER },
 527
 528         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 529           BC_F_COPYOUT },
 530         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 531           BC_F_COPYIN|BC_F_SUSER },
 532
 533         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 534           BC_F_COPYIN|BC_F_SUSER },
 535
 536         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 537           BC_F_COPYIN|BC_F_SUSER },
 538
 539         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 540           BC_F_COPYIN|BC_F_SUSER },
 541         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 542           BC_F_COPYIN|BC_F_SUSER },
 543
 544         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 545           BC_F_COPYOUT },
 546
 547         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 548           BC_F_COPYOUT },
 549
 550         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 551           BC_F_COPYIN|BC_F_COPYOUT },
 552
 553         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 554           BC_F_COPYIN|BC_F_SUSER },
 555
 556         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 557           BC_F_COPYIN|BC_F_SUSER },
 558
 559         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 560           BC_F_COPYIN|BC_F_SUSER },
 561
 562 };
 563 const int bridge_control_table_size = nitems(bridge_control_table);
 564
 565 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 566 #define V_bridge_list   VNET(bridge_list)
 567 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 568                                             "if_bridge list")
 569 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 570 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 571 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 572
 573 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 574 #define V_bridge_cloner VNET(bridge_cloner)
 575
 576 static const char bridge_name[] = "bridge";
 577
 578 static void
 579 vnet_bridge_init(const void *unused __unused)
 580 {
 581
 582         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 583             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 584             UMA_ALIGN_PTR, 0);
 585         BRIDGE_LIST_LOCK_INIT();
 586         LIST_INIT(&V_bridge_list);
 587         V_bridge_cloner = if_clone_simple(bridge_name,
 588             bridge_clone_create, bridge_clone_destroy, 0);
 589 }
 590 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 591     vnet_bridge_init, NULL);
 592
 593 static void
 594 vnet_bridge_uninit(const void *unused __unused)
 595 {
 596
 597         if_clone_detach(V_bridge_cloner);
 598         V_bridge_cloner = NULL;
 599         BRIDGE_LIST_LOCK_DESTROY();
 600
 601         /* Before we can destroy the uma zone, because there are callbacks that
 602          * use it. */
 603         epoch_drain_callbacks(net_epoch_preempt);
 604
 605         uma_zdestroy(V_bridge_rtnode_zone);
 606 }
 607 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 608     vnet_bridge_uninit, NULL);
 609
 610 static int
 611 bridge_modevent(module_t mod, int type, void *data)
 612 {
 613
 614         switch (type) {
 615         case MOD_LOAD:
 616                 bridge_dn_p = bridge_dummynet;
 617                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 618                     ifnet_departure_event, bridge_ifdetach, NULL,
 619                     EVENTHANDLER_PRI_ANY);
 620                 break;
 621         case MOD_UNLOAD:
 622                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 623                     bridge_detach_cookie);
 624                 bridge_dn_p = NULL;
 625                 break;
 626         default:
 627                 return (EOPNOTSUPP);
 628         }
 629         return (0);
 630 }
 631
 632 static moduledata_t bridge_mod = {
 633         "if_bridge",
 634         bridge_modevent,
 635         0
 636 };
 637
 638 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 639 MODULE_VERSION(if_bridge, 1);
 640 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 641
 642 /*
 643  * handler for net.link.bridge.ipfw
 644  */
 645 static int
 646 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 647 {
 648         int enable = V_pfil_ipfw;
 649         int error;
 650
 651         error = sysctl_handle_int(oidp, &enable, 0, req);
 652         enable &= 1;
 653
 654         if (enable != V_pfil_ipfw) {
 655                 V_pfil_ipfw = enable;
 656
 657                 /*
 658                  * Disable pfil so that ipfw doesnt run twice, if the user
 659                  * really wants both then they can re-enable pfil_bridge and/or
 660                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 661                  * layer2 type.
 662                  */
 663                 if (V_pfil_ipfw) {
 664                         V_pfil_onlyip = 0;
 665                         V_pfil_bridge = 0;
 666                         V_pfil_member = 0;
 667                 }
 668         }
 669
 670         return (error);
 671 }
 672 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 673     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 674     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 675     "Layer2 filter with IPFW");
 676
 677 #ifdef VIMAGE
 678 static void
 679 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 680 {
 681         struct bridge_softc *sc = ifp->if_softc;
 682         struct bridge_iflist *bif;
 683
 684         BRIDGE_LOCK(sc);
 685
 686         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 687                 bridge_delete_member(sc, bif, 0);
 688
 689         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 690                 bridge_delete_span(sc, bif);
 691         }
 692
 693         BRIDGE_UNLOCK(sc);
 694
 695         ether_reassign(ifp, newvnet, arg);
 696 }
 697 #endif
 698
 699 /*
 700  * bridge_clone_create:
 701  *
 702  *      Create a new bridge instance.
 703  */
 704 static int
 705 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 706 {
 707         struct bridge_softc *sc, *sc2;
 708         struct ifnet *bifp, *ifp;
 709         int fb, retry;
 710         unsigned long hostid;
 711
 712         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 713         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 714         if (ifp == NULL) {
 715                 free(sc, M_DEVBUF);
 716                 return (ENOSPC);
 717         }
 718
 719         BRIDGE_LOCK_INIT(sc);
 720         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 721         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 722
 723         /* Initialize our routing table. */
 724         bridge_rtable_init(sc);
 725
 726         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 727
 728         CK_LIST_INIT(&sc->sc_iflist);
 729         CK_LIST_INIT(&sc->sc_spanlist);
 730
 731         ifp->if_softc = sc;
 732         if_initname(ifp, bridge_name, unit);
 733         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 734         ifp->if_ioctl = bridge_ioctl;
 735         ifp->if_transmit = bridge_transmit;
 736         ifp->if_qflush = bridge_qflush;
 737         ifp->if_init = bridge_init;
 738         ifp->if_type = IFT_BRIDGE;
 739
 740         /*
 741          * Generate an ethernet address with a locally administered address.
 742          *
 743          * Since we are using random ethernet addresses for the bridge, it is
 744          * possible that we might have address collisions, so make sure that
 745          * this hardware address isn't already in use on another bridge.
 746          * The first try uses the hostid and falls back to arc4rand().
 747          */
 748         fb = 0;
 749         getcredhostid(curthread->td_ucred, &hostid);
 750         do {
 751                 if (fb || hostid == 0) {
 752                         ether_gen_addr(ifp, &sc->sc_defaddr);
 753                 } else {
 754                         sc->sc_defaddr.octet[0] = 0x2;
 755                         sc->sc_defaddr.octet[1] = (hostid >> 24) & 0xff;
 756                         sc->sc_defaddr.octet[2] = (hostid >> 16) & 0xff;
 757                         sc->sc_defaddr.octet[3] = (hostid >> 8 ) & 0xff;
 758                         sc->sc_defaddr.octet[4] =  hostid        & 0xff;
 759                         sc->sc_defaddr.octet[5] = ifp->if_dunit & 0xff;
 760                 }
 761
 762                 fb = 1;
 763                 retry = 0;
 764                 BRIDGE_LIST_LOCK();
 765                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 766                         bifp = sc2->sc_ifp;
 767                         if (memcmp(sc->sc_defaddr.octet,
 768                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 769                                 retry = 1;
 770                                 break;
 771                         }
 772                 }
 773                 BRIDGE_LIST_UNLOCK();
 774         } while (retry == 1);
 775
 776         bstp_attach(&sc->sc_stp, &bridge_ops);
 777         ether_ifattach(ifp, sc->sc_defaddr.octet);
 778         /* Now undo some of the damage... */
 779         ifp->if_baudrate = 0;
 780         ifp->if_type = IFT_BRIDGE;
 781 #ifdef VIMAGE
 782         ifp->if_reassign = bridge_reassign;
 783 #endif
 784
 785         BRIDGE_LIST_LOCK();
 786         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 787         BRIDGE_LIST_UNLOCK();
 788
 789         return (0);
 790 }
 791
 792 static void
 793 bridge_clone_destroy_cb(struct epoch_context *ctx)
 794 {
 795         struct bridge_softc *sc;
 796
 797         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 798
 799         BRIDGE_LOCK_DESTROY(sc);
 800         free(sc, M_DEVBUF);
 801 }
 802
 803 /*
 804  * bridge_clone_destroy:
 805  *
 806  *      Destroy a bridge instance.
 807  */
 808 static void
 809 bridge_clone_destroy(struct ifnet *ifp)
 810 {
 811         struct bridge_softc *sc = ifp->if_softc;
 812         struct bridge_iflist *bif;
 813         struct epoch_tracker et;
 814
 815         BRIDGE_LOCK(sc);
 816
 817         bridge_stop(ifp, 1);
 818         ifp->if_flags &= ~IFF_UP;
 819
 820         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 821                 bridge_delete_member(sc, bif, 0);
 822
 823         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 824                 bridge_delete_span(sc, bif);
 825         }
 826
 827         /* Tear down the routing table. */
 828         bridge_rtable_fini(sc);
 829
 830         BRIDGE_UNLOCK(sc);
 831
 832         NET_EPOCH_ENTER_ET(et);
 833
 834         callout_drain(&sc->sc_brcallout);
 835
 836         BRIDGE_LIST_LOCK();
 837         LIST_REMOVE(sc, sc_list);
 838         BRIDGE_LIST_UNLOCK();
 839
 840         bstp_detach(&sc->sc_stp);
 841         NET_EPOCH_EXIT_ET(et);
 842
 843         ether_ifdetach(ifp);
 844         if_free(ifp);
 845
 846         epoch_call(net_epoch_preempt, &sc->sc_epoch_ctx, bridge_clone_destroy_cb);
 847 }
 848
 849 /*
 850  * bridge_ioctl:
 851  *
 852  *      Handle a control request from the operator.
 853  */
 854 static int
 855 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 856 {
 857         struct bridge_softc *sc = ifp->if_softc;
 858         struct ifreq *ifr = (struct ifreq *)data;
 859         struct bridge_iflist *bif;
 860         struct thread *td = curthread;
 861         union {
 862                 struct ifbreq ifbreq;
 863                 struct ifbifconf ifbifconf;
 864                 struct ifbareq ifbareq;
 865                 struct ifbaconf ifbaconf;
 866                 struct ifbrparam ifbrparam;
 867                 struct ifbropreq ifbropreq;
 868         } args;
 869         struct ifdrv *ifd = (struct ifdrv *) data;
 870         const struct bridge_control *bc;
 871         int error = 0, oldmtu;
 872
 873         BRIDGE_LOCK(sc);
 874
 875         switch (cmd) {
 876
 877         case SIOCADDMULTI:
 878         case SIOCDELMULTI:
 879                 break;
 880
 881         case SIOCGDRVSPEC:
 882         case SIOCSDRVSPEC:
 883                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 884                         error = EINVAL;
 885                         break;
 886                 }
 887                 bc = &bridge_control_table[ifd->ifd_cmd];
 888
 889                 if (cmd == SIOCGDRVSPEC &&
 890                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 891                         error = EINVAL;
 892                         break;
 893                 }
 894                 else if (cmd == SIOCSDRVSPEC &&
 895                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 896                         error = EINVAL;
 897                         break;
 898                 }
 899
 900                 if (bc->bc_flags & BC_F_SUSER) {
 901                         error = priv_check(td, PRIV_NET_BRIDGE);
 902                         if (error)
 903                                 break;
 904                 }
 905
 906                 if (ifd->ifd_len != bc->bc_argsize ||
 907                     ifd->ifd_len > sizeof(args)) {
 908                         error = EINVAL;
 909                         break;
 910                 }
 911
 912                 bzero(&args, sizeof(args));
 913                 if (bc->bc_flags & BC_F_COPYIN) {
 914                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 915                         if (error)
 916                                 break;
 917                 }
 918
 919                 oldmtu = ifp->if_mtu;
 920                 error = (*bc->bc_func)(sc, &args);
 921                 if (error)
 922                         break;
 923
 924                 /*
 925                  * Bridge MTU may change during addition of the first port.
 926                  * If it did, do network layer specific procedure.
 927                  */
 928                 if (ifp->if_mtu != oldmtu) {
 929 #ifdef INET6
 930                         nd6_setmtu(ifp);
 931 #endif
 932                         rt_updatemtu(ifp);
 933                 }
 934
 935                 if (bc->bc_flags & BC_F_COPYOUT)
 936                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 937
 938                 break;
 939
 940         case SIOCSIFFLAGS:
 941                 if (!(ifp->if_flags & IFF_UP) &&
 942                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 943                         /*
 944                          * If interface is marked down and it is running,
 945                          * then stop and disable it.
 946                          */
 947                         bridge_stop(ifp, 1);
 948                 } else if ((ifp->if_flags & IFF_UP) &&
 949                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 950                         /*
 951                          * If interface is marked up and it is stopped, then
 952                          * start it.
 953                          */
 954                         BRIDGE_UNLOCK(sc);
 955                         (*ifp->if_init)(sc);
 956                         BRIDGE_LOCK(sc);
 957                 }
 958                 break;
 959
 960         case SIOCSIFMTU:
 961                 if (ifr->ifr_mtu < 576) {
 962                         error = EINVAL;
 963                         break;
 964                 }
 965                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 966                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 967                         break;
 968                 }
 969                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 970                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 971                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 972                                     " != %d\n", sc->sc_ifp->if_xname,
 973                                     bif->bif_ifp->if_mtu,
 974                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 975                                 error = EINVAL;
 976                                 break;
 977                         }
 978                 }
 979                 if (!error)
 980                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 981                 break;
 982         default:
 983                 /*
 984                  * drop the lock as ether_ioctl() will call bridge_start() and
 985                  * cause the lock to be recursed.
 986                  */
 987                 BRIDGE_UNLOCK(sc);
 988                 error = ether_ioctl(ifp, cmd, data);
 989                 BRIDGE_LOCK(sc);
 990                 break;
 991         }
 992
 993         BRIDGE_UNLOCK(sc);
 994
 995         return (error);
 996 }
 997
 998 /*
 999  * bridge_mutecaps:
1000  *
1001  *      Clear or restore unwanted capabilities on the member interface
1002  */
1003 static void
1004 bridge_mutecaps(struct bridge_softc *sc)
1005 {
1006         struct bridge_iflist *bif;
1007         int enabled, mask;
1008
1009         BRIDGE_LOCK_ASSERT(sc);
1010
1011         /* Initial bitmask of capabilities to test */
1012         mask = BRIDGE_IFCAPS_MASK;
1013
1014         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1015                 /* Every member must support it or its disabled */
1016                 mask &= bif->bif_savedcaps;
1017         }
1018
1019         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1020                 enabled = bif->bif_ifp->if_capenable;
1021                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1022                 /* strip off mask bits and enable them again if allowed */
1023                 enabled &= ~BRIDGE_IFCAPS_MASK;
1024                 enabled |= mask;
1025                 bridge_set_ifcap(sc, bif, enabled);
1026         }
1027 }
1028
1029 static void
1030 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1031 {
1032         struct ifnet *ifp = bif->bif_ifp;
1033         struct ifreq ifr;
1034         int error, mask, stuck;
1035
1036         bzero(&ifr, sizeof(ifr));
1037         ifr.ifr_reqcap = set;
1038
1039         if (ifp->if_capenable != set) {
1040                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1041                 if (error)
1042                         if_printf(sc->sc_ifp,
1043                             "error setting capabilities on %s: %d\n",
1044                             ifp->if_xname, error);
1045                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1046                 stuck = ifp->if_capenable & mask & ~set;
1047                 if (stuck != 0)
1048                         if_printf(sc->sc_ifp,
1049                             "can't disable some capabilities on %s: 0x%x\n",
1050                             ifp->if_xname, stuck);
1051         }
1052 }
1053
1054 /*
1055  * bridge_lookup_member:
1056  *
1057  *      Lookup a bridge member interface.
1058  */
1059 static struct bridge_iflist *
1060 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1061 {
1062         struct bridge_iflist *bif;
1063         struct ifnet *ifp;
1064
1065         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1066
1067         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1068                 ifp = bif->bif_ifp;
1069                 if (strcmp(ifp->if_xname, name) == 0)
1070                         return (bif);
1071         }
1072
1073         return (NULL);
1074 }
1075
1076 /*
1077  * bridge_lookup_member_if:
1078  *
1079  *      Lookup a bridge member interface by ifnet*.
1080  */
1081 static struct bridge_iflist *
1082 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1083 {
1084         struct bridge_iflist *bif;
1085
1086         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1087
1088         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1089                 if (bif->bif_ifp == member_ifp)
1090                         return (bif);
1091         }
1092
1093         return (NULL);
1094 }
1095
1096 static void
1097 bridge_delete_member_cb(struct epoch_context *ctx)
1098 {
1099         struct bridge_iflist *bif;
1100
1101         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1102
1103         free(bif, M_DEVBUF);
1104 }
1105
1106 /*
1107  * bridge_delete_member:
1108  *
1109  *      Delete the specified member interface.
1110  */
1111 static void
1112 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1113     int gone)
1114 {
1115         struct ifnet *ifs = bif->bif_ifp;
1116         struct ifnet *fif = NULL;
1117         struct bridge_iflist *bifl;
1118
1119         BRIDGE_LOCK_ASSERT(sc);
1120
1121         if (bif->bif_flags & IFBIF_STP)
1122                 bstp_disable(&bif->bif_stp);
1123
1124         ifs->if_bridge = NULL;
1125         CK_LIST_REMOVE(bif, bif_next);
1126
1127         /*
1128          * If removing the interface that gave the bridge its mac address, set
1129          * the mac address of the bridge to the address of the next member, or
1130          * to its default address if no members are left.
1131          */
1132         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1133                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1134                         bcopy(&sc->sc_defaddr,
1135                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1136                         sc->sc_ifaddr = NULL;
1137                 } else {
1138                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1139                         fif = bifl->bif_ifp;
1140                         bcopy(IF_LLADDR(fif),
1141                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1142                         sc->sc_ifaddr = fif;
1143                 }
1144                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1145         }
1146
1147         bridge_linkcheck(sc);
1148         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1149         BRIDGE_RT_LOCK(sc);
1150         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1151         BRIDGE_RT_UNLOCK(sc);
1152         KASSERT(bif->bif_addrcnt == 0,
1153             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1154
1155         ifs->if_bridge_output = NULL;
1156         ifs->if_bridge_input = NULL;
1157         ifs->if_bridge_linkstate = NULL;
1158         if (!gone) {
1159                 switch (ifs->if_type) {
1160                 case IFT_ETHER:
1161                 case IFT_L2VLAN:
1162                         /*
1163                          * Take the interface out of promiscuous mode, but only
1164                          * if it was promiscuous in the first place. It might
1165                          * not be if we're in the bridge_ioctl_add() error path.
1166                          */
1167                         if (ifs->if_flags & IFF_PROMISC)
1168                                 (void) ifpromisc(ifs, 0);
1169                         break;
1170
1171                 case IFT_GIF:
1172                         break;
1173
1174                 default:
1175 #ifdef DIAGNOSTIC
1176                         panic("bridge_delete_member: impossible");
1177 #endif
1178                         break;
1179                 }
1180                 /* reneable any interface capabilities */
1181                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1182         }
1183         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1184
1185         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1186             bridge_delete_member_cb);
1187 }
1188
1189 /*
1190  * bridge_delete_span:
1191  *
1192  *      Delete the specified span interface.
1193  */
1194 static void
1195 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1196 {
1197         BRIDGE_LOCK_ASSERT(sc);
1198
1199         KASSERT(bif->bif_ifp->if_bridge == NULL,
1200             ("%s: not a span interface", __func__));
1201
1202         CK_LIST_REMOVE(bif, bif_next);
1203
1204         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1205             bridge_delete_member_cb);
1206 }
1207
1208 static int
1209 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1210 {
1211         struct ifbreq *req = arg;
1212         struct bridge_iflist *bif = NULL;
1213         struct ifnet *ifs;
1214         int error = 0;
1215
1216         ifs = ifunit(req->ifbr_ifsname);
1217         if (ifs == NULL)
1218                 return (ENOENT);
1219         if (ifs->if_ioctl == NULL)      /* must be supported */
1220                 return (EINVAL);
1221
1222         /* If it's in the span list, it can't be a member. */
1223         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1224                 if (ifs == bif->bif_ifp)
1225                         return (EBUSY);
1226
1227         if (ifs->if_bridge == sc)
1228                 return (EEXIST);
1229
1230         if (ifs->if_bridge != NULL)
1231                 return (EBUSY);
1232
1233         switch (ifs->if_type) {
1234         case IFT_ETHER:
1235         case IFT_L2VLAN:
1236         case IFT_GIF:
1237                 /* permitted interface types */
1238                 break;
1239         default:
1240                 return (EINVAL);
1241         }
1242
1243 #ifdef INET6
1244         /*
1245          * Two valid inet6 addresses with link-local scope must not be
1246          * on the parent interface and the member interfaces at the
1247          * same time.  This restriction is needed to prevent violation
1248          * of link-local scope zone.  Attempts to add a member
1249          * interface which has inet6 addresses when the parent has
1250          * inet6 triggers removal of all inet6 addresses on the member
1251          * interface.
1252          */
1253
1254         /* Check if the parent interface has a link-local scope addr. */
1255         if (V_allow_llz_overlap == 0 &&
1256             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1257                 /*
1258                  * If any, remove all inet6 addresses from the member
1259                  * interfaces.
1260                  */
1261                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1262                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1263                                 in6_ifdetach(bif->bif_ifp);
1264                                 if_printf(sc->sc_ifp,
1265                                     "IPv6 addresses on %s have been removed "
1266                                     "before adding it as a member to prevent "
1267                                     "IPv6 address scope violation.\n",
1268                                     bif->bif_ifp->if_xname);
1269                         }
1270                 }
1271                 if (in6ifa_llaonifp(ifs)) {
1272                         in6_ifdetach(ifs);
1273                         if_printf(sc->sc_ifp,
1274                             "IPv6 addresses on %s have been removed "
1275                             "before adding it as a member to prevent "
1276                             "IPv6 address scope violation.\n",
1277                             ifs->if_xname);
1278                 }
1279         }
1280 #endif
1281         /* Allow the first Ethernet member to define the MTU */
1282         if (CK_LIST_EMPTY(&sc->sc_iflist))
1283                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1284         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1285                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1286                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1287                 return (EINVAL);
1288         }
1289
1290         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1291         if (bif == NULL)
1292                 return (ENOMEM);
1293
1294         bif->bif_ifp = ifs;
1295         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1296         bif->bif_savedcaps = ifs->if_capenable;
1297
1298         /*
1299          * Assign the interface's MAC address to the bridge if it's the first
1300          * member and the MAC address of the bridge has not been changed from
1301          * the default randomly generated one.
1302          */
1303         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1304             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1305                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1306                 sc->sc_ifaddr = ifs;
1307                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1308         }
1309
1310         ifs->if_bridge = sc;
1311         ifs->if_bridge_output = bridge_output;
1312         ifs->if_bridge_input = bridge_input;
1313         ifs->if_bridge_linkstate = bridge_linkstate;
1314         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1315         /*
1316          * XXX: XLOCK HERE!?!
1317          *
1318          * NOTE: insert_***HEAD*** should be safe for the traversals.
1319          */
1320         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1321
1322         /* Set interface capabilities to the intersection set of all members */
1323         bridge_mutecaps(sc);
1324         bridge_linkcheck(sc);
1325
1326         /* Place the interface into promiscuous mode */
1327         switch (ifs->if_type) {
1328                 case IFT_ETHER:
1329                 case IFT_L2VLAN:
1330                         error = ifpromisc(ifs, 1);
1331                         break;
1332         }
1333
1334         if (error)
1335                 bridge_delete_member(sc, bif, 0);
1336         return (error);
1337 }
1338
1339 static int
1340 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1341 {
1342         struct ifbreq *req = arg;
1343         struct bridge_iflist *bif;
1344
1345         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1346         if (bif == NULL)
1347                 return (ENOENT);
1348
1349         bridge_delete_member(sc, bif, 0);
1350
1351         return (0);
1352 }
1353
1354 static int
1355 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1356 {
1357         struct ifbreq *req = arg;
1358         struct bridge_iflist *bif;
1359         struct bstp_port *bp;
1360
1361         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1362         if (bif == NULL)
1363                 return (ENOENT);
1364
1365         bp = &bif->bif_stp;
1366         req->ifbr_ifsflags = bif->bif_flags;
1367         req->ifbr_state = bp->bp_state;
1368         req->ifbr_priority = bp->bp_priority;
1369         req->ifbr_path_cost = bp->bp_path_cost;
1370         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1371         req->ifbr_proto = bp->bp_protover;
1372         req->ifbr_role = bp->bp_role;
1373         req->ifbr_stpflags = bp->bp_flags;
1374         req->ifbr_addrcnt = bif->bif_addrcnt;
1375         req->ifbr_addrmax = bif->bif_addrmax;
1376         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1377
1378         /* Copy STP state options as flags */
1379         if (bp->bp_operedge)
1380                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1381         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1382                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1383         if (bp->bp_ptp_link)
1384                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1385         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1386                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1387         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1388                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1389         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1390                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1391         return (0);
1392 }
1393
1394 static int
1395 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1396 {
1397         struct ifbreq *req = arg;
1398         struct bridge_iflist *bif;
1399         struct bstp_port *bp;
1400         int error;
1401
1402         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1403         if (bif == NULL)
1404                 return (ENOENT);
1405         bp = &bif->bif_stp;
1406
1407         if (req->ifbr_ifsflags & IFBIF_SPAN)
1408                 /* SPAN is readonly */
1409                 return (EINVAL);
1410
1411         NET_EPOCH_ENTER();
1412
1413         if (req->ifbr_ifsflags & IFBIF_STP) {
1414                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1415                         error = bstp_enable(&bif->bif_stp);
1416                         if (error) {
1417                                 NET_EPOCH_EXIT();
1418                                 return (error);
1419                         }
1420                 }
1421         } else {
1422                 if ((bif->bif_flags & IFBIF_STP) != 0)
1423                         bstp_disable(&bif->bif_stp);
1424         }
1425
1426         /* Pass on STP flags */
1427         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1428         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1429         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1430         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1431
1432         /* Save the bits relating to the bridge */
1433         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1434
1435         NET_EPOCH_EXIT();
1436
1437         return (0);
1438 }
1439
1440 static int
1441 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1442 {
1443         struct ifbrparam *param = arg;
1444
1445         sc->sc_brtmax = param->ifbrp_csize;
1446         bridge_rttrim(sc);
1447
1448         return (0);
1449 }
1450
1451 static int
1452 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1453 {
1454         struct ifbrparam *param = arg;
1455
1456         param->ifbrp_csize = sc->sc_brtmax;
1457
1458         return (0);
1459 }
1460
1461 static int
1462 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1463 {
1464         struct ifbifconf *bifc = arg;
1465         struct bridge_iflist *bif;
1466         struct ifbreq breq;
1467         char *buf, *outbuf;
1468         int count, buflen, len, error = 0;
1469
1470         count = 0;
1471         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1472                 count++;
1473         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1474                 count++;
1475
1476         buflen = sizeof(breq) * count;
1477         if (bifc->ifbic_len == 0) {
1478                 bifc->ifbic_len = buflen;
1479                 return (0);
1480         }
1481         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1482         if (outbuf == NULL)
1483                 return (ENOMEM);
1484
1485         count = 0;
1486         buf = outbuf;
1487         len = min(bifc->ifbic_len, buflen);
1488         bzero(&breq, sizeof(breq));
1489         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1490                 if (len < sizeof(breq))
1491                         break;
1492
1493                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1494                     sizeof(breq.ifbr_ifsname));
1495                 /* Fill in the ifbreq structure */
1496                 error = bridge_ioctl_gifflags(sc, &breq);
1497                 if (error)
1498                         break;
1499                 memcpy(buf, &breq, sizeof(breq));
1500                 count++;
1501                 buf += sizeof(breq);
1502                 len -= sizeof(breq);
1503         }
1504         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1505                 if (len < sizeof(breq))
1506                         break;
1507
1508                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1509                     sizeof(breq.ifbr_ifsname));
1510                 breq.ifbr_ifsflags = bif->bif_flags;
1511                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1512                 memcpy(buf, &breq, sizeof(breq));
1513                 count++;
1514                 buf += sizeof(breq);
1515                 len -= sizeof(breq);
1516         }
1517
1518         bifc->ifbic_len = sizeof(breq) * count;
1519         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1520         free(outbuf, M_TEMP);
1521         return (error);
1522 }
1523
1524 static int
1525 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1526 {
1527         struct ifbaconf *bac = arg;
1528         struct bridge_rtnode *brt;
1529         struct ifbareq bareq;
1530         char *buf, *outbuf;
1531         int count, buflen, len, error = 0;
1532
1533         if (bac->ifbac_len == 0)
1534                 return (0);
1535
1536         count = 0;
1537         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1538                 count++;
1539         buflen = sizeof(bareq) * count;
1540
1541         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1542         if (outbuf == NULL)
1543                 return (ENOMEM);
1544
1545         count = 0;
1546         buf = outbuf;
1547         len = min(bac->ifbac_len, buflen);
1548         bzero(&bareq, sizeof(bareq));
1549         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1550                 if (len < sizeof(bareq))
1551                         goto out;
1552                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1553                     sizeof(bareq.ifba_ifsname));
1554                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1555                 bareq.ifba_vlan = brt->brt_vlan;
1556                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1557                                 time_uptime < brt->brt_expire)
1558                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1559                 else
1560                         bareq.ifba_expire = 0;
1561                 bareq.ifba_flags = brt->brt_flags;
1562
1563                 memcpy(buf, &bareq, sizeof(bareq));
1564                 count++;
1565                 buf += sizeof(bareq);
1566                 len -= sizeof(bareq);
1567         }
1568 out:
1569         bac->ifbac_len = sizeof(bareq) * count;
1570         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1571         free(outbuf, M_TEMP);
1572         return (error);
1573 }
1574
1575 static int
1576 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1577 {
1578         struct ifbareq *req = arg;
1579         struct bridge_iflist *bif;
1580         struct epoch_tracker et;
1581         int error;
1582
1583         NET_EPOCH_ENTER_ET(et);
1584
1585         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1586         if (bif == NULL) {
1587                 NET_EPOCH_EXIT_ET(et);
1588                 return (ENOENT);
1589         }
1590
1591         /* bridge_rtupdate() may acquire the lock. */
1592         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1593             req->ifba_flags);
1594         NET_EPOCH_EXIT_ET(et);
1595
1596         return (error);
1597 }
1598
1599 static int
1600 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1601 {
1602         struct ifbrparam *param = arg;
1603
1604         sc->sc_brttimeout = param->ifbrp_ctime;
1605         return (0);
1606 }
1607
1608 static int
1609 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1610 {
1611         struct ifbrparam *param = arg;
1612
1613         param->ifbrp_ctime = sc->sc_brttimeout;
1614         return (0);
1615 }
1616
1617 static int
1618 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1619 {
1620         struct ifbareq *req = arg;
1621
1622         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1623 }
1624
1625 static int
1626 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1627 {
1628         struct ifbreq *req = arg;
1629
1630         BRIDGE_RT_LOCK(sc);
1631         bridge_rtflush(sc, req->ifbr_ifsflags);
1632         BRIDGE_RT_UNLOCK(sc);
1633
1634         return (0);
1635 }
1636
1637 static int
1638 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1639 {
1640         struct ifbrparam *param = arg;
1641         struct bstp_state *bs = &sc->sc_stp;
1642
1643         param->ifbrp_prio = bs->bs_bridge_priority;
1644         return (0);
1645 }
1646
1647 static int
1648 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1649 {
1650         struct ifbrparam *param = arg;
1651
1652         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1653 }
1654
1655 static int
1656 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1657 {
1658         struct ifbrparam *param = arg;
1659         struct bstp_state *bs = &sc->sc_stp;
1660
1661         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1662         return (0);
1663 }
1664
1665 static int
1666 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1667 {
1668         struct ifbrparam *param = arg;
1669
1670         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1671 }
1672
1673 static int
1674 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1675 {
1676         struct ifbrparam *param = arg;
1677         struct bstp_state *bs = &sc->sc_stp;
1678
1679         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1680         return (0);
1681 }
1682
1683 static int
1684 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1685 {
1686         struct ifbrparam *param = arg;
1687
1688         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1689 }
1690
1691 static int
1692 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1693 {
1694         struct ifbrparam *param = arg;
1695         struct bstp_state *bs = &sc->sc_stp;
1696
1697         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1698         return (0);
1699 }
1700
1701 static int
1702 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1703 {
1704         struct ifbrparam *param = arg;
1705
1706         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1707 }
1708
1709 static int
1710 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1711 {
1712         struct ifbreq *req = arg;
1713         struct bridge_iflist *bif;
1714
1715         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1716         if (bif == NULL)
1717                 return (ENOENT);
1718
1719         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1720 }
1721
1722 static int
1723 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1724 {
1725         struct ifbreq *req = arg;
1726         struct bridge_iflist *bif;
1727
1728         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1729         if (bif == NULL)
1730                 return (ENOENT);
1731
1732         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1733 }
1734
1735 static int
1736 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1737 {
1738         struct ifbreq *req = arg;
1739         struct bridge_iflist *bif;
1740
1741         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1742         if (bif == NULL)
1743                 return (ENOENT);
1744
1745         bif->bif_addrmax = req->ifbr_addrmax;
1746         return (0);
1747 }
1748
1749 static int
1750 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1751 {
1752         struct ifbreq *req = arg;
1753         struct bridge_iflist *bif = NULL;
1754         struct ifnet *ifs;
1755
1756         ifs = ifunit(req->ifbr_ifsname);
1757         if (ifs == NULL)
1758                 return (ENOENT);
1759
1760         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1761                 if (ifs == bif->bif_ifp)
1762                         return (EBUSY);
1763
1764         if (ifs->if_bridge != NULL)
1765                 return (EBUSY);
1766
1767         switch (ifs->if_type) {
1768                 case IFT_ETHER:
1769                 case IFT_GIF:
1770                 case IFT_L2VLAN:
1771                         break;
1772                 default:
1773                         return (EINVAL);
1774         }
1775
1776         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1777         if (bif == NULL)
1778                 return (ENOMEM);
1779
1780         bif->bif_ifp = ifs;
1781         bif->bif_flags = IFBIF_SPAN;
1782
1783         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1784
1785         return (0);
1786 }
1787
1788 static int
1789 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1790 {
1791         struct ifbreq *req = arg;
1792         struct bridge_iflist *bif;
1793         struct ifnet *ifs;
1794
1795         ifs = ifunit(req->ifbr_ifsname);
1796         if (ifs == NULL)
1797                 return (ENOENT);
1798
1799         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1800                 if (ifs == bif->bif_ifp)
1801                         break;
1802
1803         if (bif == NULL)
1804                 return (ENOENT);
1805
1806         bridge_delete_span(sc, bif);
1807
1808         return (0);
1809 }
1810
1811 static int
1812 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1813 {
1814         struct ifbropreq *req = arg;
1815         struct bstp_state *bs = &sc->sc_stp;
1816         struct bstp_port *root_port;
1817
1818         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1819         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1820         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1821
1822         root_port = bs->bs_root_port;
1823         if (root_port == NULL)
1824                 req->ifbop_root_port = 0;
1825         else
1826                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1827
1828         req->ifbop_holdcount = bs->bs_txholdcount;
1829         req->ifbop_priority = bs->bs_bridge_priority;
1830         req->ifbop_protocol = bs->bs_protover;
1831         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1832         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1833         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1834         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1835         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1836         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1837
1838         return (0);
1839 }
1840
1841 static int
1842 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1843 {
1844         struct ifbrparam *param = arg;
1845
1846         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1847         return (0);
1848 }
1849
1850 static int
1851 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1852 {
1853         struct ifbpstpconf *bifstp = arg;
1854         struct bridge_iflist *bif;
1855         struct bstp_port *bp;
1856         struct ifbpstpreq bpreq;
1857         char *buf, *outbuf;
1858         int count, buflen, len, error = 0;
1859
1860         count = 0;
1861         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1862                 if ((bif->bif_flags & IFBIF_STP) != 0)
1863                         count++;
1864         }
1865
1866         buflen = sizeof(bpreq) * count;
1867         if (bifstp->ifbpstp_len == 0) {
1868                 bifstp->ifbpstp_len = buflen;
1869                 return (0);
1870         }
1871
1872         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1873         if (outbuf == NULL)
1874                 return (ENOMEM);
1875
1876         count = 0;
1877         buf = outbuf;
1878         len = min(bifstp->ifbpstp_len, buflen);
1879         bzero(&bpreq, sizeof(bpreq));
1880         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1881                 if (len < sizeof(bpreq))
1882                         break;
1883
1884                 if ((bif->bif_flags & IFBIF_STP) == 0)
1885                         continue;
1886
1887                 bp = &bif->bif_stp;
1888                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1889                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1890                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1891                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1892                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1893                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1894
1895                 memcpy(buf, &bpreq, sizeof(bpreq));
1896                 count++;
1897                 buf += sizeof(bpreq);
1898                 len -= sizeof(bpreq);
1899         }
1900
1901         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1902         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1903         free(outbuf, M_TEMP);
1904         return (error);
1905 }
1906
1907 static int
1908 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1909 {
1910         struct ifbrparam *param = arg;
1911
1912         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1913 }
1914
1915 static int
1916 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1917 {
1918         struct ifbrparam *param = arg;
1919
1920         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1921 }
1922
1923 /*
1924  * bridge_ifdetach:
1925  *
1926  *      Detach an interface from a bridge.  Called when a member
1927  *      interface is detaching.
1928  */
1929 static void
1930 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1931 {
1932         struct bridge_softc *sc = ifp->if_bridge;
1933         struct bridge_iflist *bif;
1934
1935         if (ifp->if_flags & IFF_RENAMING)
1936                 return;
1937         if (V_bridge_cloner == NULL) {
1938                 /*
1939                  * This detach handler can be called after
1940                  * vnet_bridge_uninit().  Just return in that case.
1941                  */
1942                 return;
1943         }
1944         /* Check if the interface is a bridge member */
1945         if (sc != NULL) {
1946                 BRIDGE_LOCK(sc);
1947
1948                 bif = bridge_lookup_member_if(sc, ifp);
1949                 if (bif != NULL)
1950                         bridge_delete_member(sc, bif, 1);
1951
1952                 BRIDGE_UNLOCK(sc);
1953                 return;
1954         }
1955
1956         /* Check if the interface is a span port */
1957         BRIDGE_LIST_LOCK();
1958         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1959                 BRIDGE_LOCK(sc);
1960                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1961                         if (ifp == bif->bif_ifp) {
1962                                 bridge_delete_span(sc, bif);
1963                                 break;
1964                         }
1965
1966                 BRIDGE_UNLOCK(sc);
1967         }
1968         BRIDGE_LIST_UNLOCK();
1969 }
1970
1971 /*
1972  * bridge_init:
1973  *
1974  *      Initialize a bridge interface.
1975  */
1976 static void
1977 bridge_init(void *xsc)
1978 {
1979         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1980         struct ifnet *ifp = sc->sc_ifp;
1981
1982         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1983                 return;
1984
1985         BRIDGE_LOCK(sc);
1986         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1987             bridge_timer, sc);
1988
1989         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1990         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1991
1992         BRIDGE_UNLOCK(sc);
1993 }
1994
1995 /*
1996  * bridge_stop:
1997  *
1998  *      Stop the bridge interface.
1999  */
2000 static void
2001 bridge_stop(struct ifnet *ifp, int disable)
2002 {
2003         struct bridge_softc *sc = ifp->if_softc;
2004
2005         BRIDGE_LOCK_ASSERT(sc);
2006
2007         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2008                 return;
2009
2010         BRIDGE_RT_LOCK(sc);
2011         callout_stop(&sc->sc_brcallout);
2012
2013         bstp_stop(&sc->sc_stp);
2014
2015         bridge_rtflush(sc, IFBF_FLUSHDYN);
2016         BRIDGE_RT_UNLOCK(sc);
2017
2018         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2019 }
2020
2021 /*
2022  * bridge_enqueue:
2023  *
2024  *      Enqueue a packet on a bridge member interface.
2025  *
2026  */
2027 static int
2028 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2029 {
2030         int len, err = 0;
2031         short mflags;
2032         struct mbuf *m0;
2033
2034         /* We may be sending a fragment so traverse the mbuf */
2035         for (; m; m = m0) {
2036                 m0 = m->m_nextpkt;
2037                 m->m_nextpkt = NULL;
2038                 len = m->m_pkthdr.len;
2039                 mflags = m->m_flags;
2040
2041                 /*
2042                  * If underlying interface can not do VLAN tag insertion itself
2043                  * then attach a packet tag that holds it.
2044                  */
2045                 if ((m->m_flags & M_VLANTAG) &&
2046                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2047                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2048                         if (m == NULL) {
2049                                 if_printf(dst_ifp,
2050                                     "unable to prepend VLAN header\n");
2051                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2052                                 continue;
2053                         }
2054                         m->m_flags &= ~M_VLANTAG;
2055                 }
2056
2057                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2058                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2059                         m_freem(m0);
2060                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2061                         break;
2062                 }
2063
2064                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2065                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2066                 if (mflags & M_MCAST)
2067                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2068         }
2069
2070         return (err);
2071 }
2072
2073 /*
2074  * bridge_dummynet:
2075  *
2076  *      Receive a queued packet from dummynet and pass it on to the output
2077  *      interface.
2078  *
2079  *      The mbuf has the Ethernet header already attached.
2080  */
2081 static void
2082 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2083 {
2084         struct bridge_softc *sc;
2085
2086         sc = ifp->if_bridge;
2087
2088         /*
2089          * The packet didnt originate from a member interface. This should only
2090          * ever happen if a member interface is removed while packets are
2091          * queued for it.
2092          */
2093         if (sc == NULL) {
2094                 m_freem(m);
2095                 return;
2096         }
2097
2098         if (PFIL_HOOKED(&V_inet_pfil_hook)
2099 #ifdef INET6
2100             || PFIL_HOOKED(&V_inet6_pfil_hook)
2101 #endif
2102             ) {
2103                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2104                         return;
2105                 if (m == NULL)
2106                         return;
2107         }
2108
2109         bridge_enqueue(sc, ifp, m);
2110 }
2111
2112 /*
2113  * bridge_output:
2114  *
2115  *      Send output from a bridge member interface.  This
2116  *      performs the bridging function for locally originated
2117  *      packets.
2118  *
2119  *      The mbuf has the Ethernet header already attached.  We must
2120  *      enqueue or free the mbuf before returning.
2121  */
2122 static int
2123 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2124     struct rtentry *rt)
2125 {
2126         struct epoch_tracker et;
2127         struct ether_header *eh;
2128         struct ifnet *dst_if;
2129         struct bridge_softc *sc;
2130         uint16_t vlan;
2131
2132         NET_EPOCH_ENTER_ET(et);
2133
2134         if (m->m_len < ETHER_HDR_LEN) {
2135                 m = m_pullup(m, ETHER_HDR_LEN);
2136                 if (m == NULL) {
2137                         NET_EPOCH_EXIT_ET(et);
2138                         return (0);
2139                 }
2140         }
2141
2142         eh = mtod(m, struct ether_header *);
2143         sc = ifp->if_bridge;
2144         vlan = VLANTAGOF(m);
2145
2146         /*
2147          * If bridge is down, but the original output interface is up,
2148          * go ahead and send out that interface.  Otherwise, the packet
2149          * is dropped below.
2150          */
2151         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2152                 dst_if = ifp;
2153                 goto sendunicast;
2154         }
2155
2156         /*
2157          * If the packet is a multicast, or we don't know a better way to
2158          * get there, send to all interfaces.
2159          */
2160         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2161                 dst_if = NULL;
2162         else
2163                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2164         if (dst_if == NULL) {
2165                 struct bridge_iflist *bif;
2166                 struct mbuf *mc;
2167                 int used = 0;
2168
2169                 bridge_span(sc, m);
2170
2171                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2172                         dst_if = bif->bif_ifp;
2173
2174                         if (dst_if->if_type == IFT_GIF)
2175                                 continue;
2176                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2177                                 continue;
2178
2179                         /*
2180                          * If this is not the original output interface,
2181                          * and the interface is participating in spanning
2182                          * tree, make sure the port is in a state that
2183                          * allows forwarding.
2184                          */
2185                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2186                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2187                                 continue;
2188
2189                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2190                                 used = 1;
2191                                 mc = m;
2192                         } else {
2193                                 mc = m_copypacket(m, M_NOWAIT);
2194                                 if (mc == NULL) {
2195                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2196                                         continue;
2197                                 }
2198                         }
2199
2200                         bridge_enqueue(sc, dst_if, mc);
2201                 }
2202                 if (used == 0)
2203                         m_freem(m);
2204                 NET_EPOCH_EXIT_ET(et);
2205                 return (0);
2206         }
2207
2208 sendunicast:
2209         /*
2210          * XXX Spanning tree consideration here?
2211          */
2212
2213         bridge_span(sc, m);
2214         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2215                 m_freem(m);
2216                 NET_EPOCH_EXIT_ET(et);
2217                 return (0);
2218         }
2219
2220         bridge_enqueue(sc, dst_if, m);
2221         NET_EPOCH_EXIT_ET(et);
2222         return (0);
2223 }
2224
2225 /*
2226  * bridge_transmit:
2227  *
2228  *      Do output on a bridge.
2229  *
2230  */
2231 static int
2232 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2233 {
2234         struct epoch_tracker et;
2235         struct bridge_softc *sc;
2236         struct ether_header *eh;
2237         struct ifnet *dst_if;
2238         int error = 0;
2239
2240         NET_EPOCH_ENTER_ET(et);
2241
2242         sc = ifp->if_softc;
2243
2244         ETHER_BPF_MTAP(ifp, m);
2245
2246         eh = mtod(m, struct ether_header *);
2247
2248         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2249             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2250                 error = bridge_enqueue(sc, dst_if, m);
2251         } else
2252                 bridge_broadcast(sc, ifp, m, 0);
2253
2254         NET_EPOCH_EXIT_ET(et);
2255
2256         return (error);
2257 }
2258
2259 /*
2260  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2261  */
2262 static void
2263 bridge_qflush(struct ifnet *ifp __unused)
2264 {
2265 }
2266
2267 /*
2268  * bridge_forward:
2269  *
2270  *      The forwarding function of the bridge.
2271  *
2272  *      NOTE: Releases the lock on return.
2273  */
2274 static void
2275 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2276     struct mbuf *m)
2277 {
2278         struct bridge_iflist *dbif;
2279         struct ifnet *src_if, *dst_if, *ifp;
2280         struct ether_header *eh;
2281         uint16_t vlan;
2282         uint8_t *dst;
2283         int error;
2284
2285         MPASS(in_epoch(net_epoch_preempt));
2286
2287         src_if = m->m_pkthdr.rcvif;
2288         ifp = sc->sc_ifp;
2289
2290         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2291         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2292         vlan = VLANTAGOF(m);
2293
2294         if ((sbif->bif_flags & IFBIF_STP) &&
2295             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2296                 goto drop;
2297
2298         eh = mtod(m, struct ether_header *);
2299         dst = eh->ether_dhost;
2300
2301         /* If the interface is learning, record the address. */
2302         if (sbif->bif_flags & IFBIF_LEARNING) {
2303                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2304                     sbif, 0, IFBAF_DYNAMIC);
2305                 /*
2306                  * If the interface has addresses limits then deny any source
2307                  * that is not in the cache.
2308                  */
2309                 if (error && sbif->bif_addrmax)
2310                         goto drop;
2311         }
2312
2313         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2314             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2315                 goto drop;
2316
2317         /*
2318          * At this point, the port either doesn't participate
2319          * in spanning tree or it is in the forwarding state.
2320          */
2321
2322         /*
2323          * If the packet is unicast, destined for someone on
2324          * "this" side of the bridge, drop it.
2325          */
2326         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2327                 dst_if = bridge_rtlookup(sc, dst, vlan);
2328                 if (src_if == dst_if)
2329                         goto drop;
2330         } else {
2331                 /*
2332                  * Check if its a reserved multicast address, any address
2333                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2334                  * bridge.
2335                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2336                  */
2337                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2338                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2339                     dst[4] == 0x00 && dst[5] <= 0x0f)
2340                         goto drop;
2341
2342                 /* ...forward it to all interfaces. */
2343                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2344                 dst_if = NULL;
2345         }
2346
2347         /*
2348          * If we have a destination interface which is a member of our bridge,
2349          * OR this is a unicast packet, push it through the bpf(4) machinery.
2350          * For broadcast or multicast packets, don't bother because it will
2351          * be reinjected into ether_input. We do this before we pass the packets
2352          * through the pfil(9) framework, as it is possible that pfil(9) will
2353          * drop the packet, or possibly modify it, making it difficult to debug
2354          * firewall issues on the bridge.
2355          */
2356         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2357                 ETHER_BPF_MTAP(ifp, m);
2358
2359         /* run the packet filter */
2360         if (PFIL_HOOKED(&V_inet_pfil_hook)
2361 #ifdef INET6
2362             || PFIL_HOOKED(&V_inet6_pfil_hook)
2363 #endif
2364             ) {
2365                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2366                         return;
2367                 if (m == NULL)
2368                         return;
2369         }
2370
2371         if (dst_if == NULL) {
2372                 bridge_broadcast(sc, src_if, m, 1);
2373                 return;
2374         }
2375
2376         /*
2377          * At this point, we're dealing with a unicast frame
2378          * going to a different interface.
2379          */
2380         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2381                 goto drop;
2382
2383         dbif = bridge_lookup_member_if(sc, dst_if);
2384         if (dbif == NULL)
2385                 /* Not a member of the bridge (anymore?) */
2386                 goto drop;
2387
2388         /* Private segments can not talk to each other */
2389         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2390                 goto drop;
2391
2392         if ((dbif->bif_flags & IFBIF_STP) &&
2393             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2394                 goto drop;
2395
2396         if (PFIL_HOOKED(&V_inet_pfil_hook)
2397 #ifdef INET6
2398             || PFIL_HOOKED(&V_inet6_pfil_hook)
2399 #endif
2400             ) {
2401                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2402                         return;
2403                 if (m == NULL)
2404                         return;
2405         }
2406
2407         bridge_enqueue(sc, dst_if, m);
2408         return;
2409
2410 drop:
2411         m_freem(m);
2412 }
2413
2414 /*
2415  * bridge_input:
2416  *
2417  *      Receive input from a member interface.  Queue the packet for
2418  *      bridging if it is not for us.
2419  */
2420 static struct mbuf *
2421 bridge_input(struct ifnet *ifp, struct mbuf *m)
2422 {
2423         struct epoch_tracker et;
2424         struct bridge_softc *sc = ifp->if_bridge;
2425         struct bridge_iflist *bif, *bif2;
2426         struct ifnet *bifp;
2427         struct ether_header *eh;
2428         struct mbuf *mc, *mc2;
2429         uint16_t vlan;
2430         int error;
2431
2432         NET_EPOCH_ENTER_ET(et);
2433
2434         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2435                 NET_EPOCH_EXIT_ET(et);
2436                 return (m);
2437         }
2438
2439         bifp = sc->sc_ifp;
2440         vlan = VLANTAGOF(m);
2441
2442         /*
2443          * Implement support for bridge monitoring. If this flag has been
2444          * set on this interface, discard the packet once we push it through
2445          * the bpf(4) machinery, but before we do, increment the byte and
2446          * packet counters associated with this interface.
2447          */
2448         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2449                 m->m_pkthdr.rcvif  = bifp;
2450                 ETHER_BPF_MTAP(bifp, m);
2451                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2452                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2453                 m_freem(m);
2454                 NET_EPOCH_EXIT_ET(et);
2455                 return (NULL);
2456         }
2457         bif = bridge_lookup_member_if(sc, ifp);
2458         if (bif == NULL) {
2459                 NET_EPOCH_EXIT_ET(et);
2460                 return (m);
2461         }
2462
2463         eh = mtod(m, struct ether_header *);
2464
2465         bridge_span(sc, m);
2466
2467         if (m->m_flags & (M_BCAST|M_MCAST)) {
2468                 /* Tap off 802.1D packets; they do not get forwarded. */
2469                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2470                     ETHER_ADDR_LEN) == 0) {
2471                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2472                         NET_EPOCH_EXIT_ET(et);
2473                         return (NULL);
2474                 }
2475
2476                 if ((bif->bif_flags & IFBIF_STP) &&
2477                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2478                         NET_EPOCH_EXIT_ET(et);
2479                         return (m);
2480                 }
2481
2482                 /*
2483                  * Make a deep copy of the packet and enqueue the copy
2484                  * for bridge processing; return the original packet for
2485                  * local processing.
2486                  */
2487                 mc = m_dup(m, M_NOWAIT);
2488                 if (mc == NULL) {
2489                         NET_EPOCH_EXIT_ET(et);
2490                         return (m);
2491                 }
2492
2493                 /* Perform the bridge forwarding function with the copy. */
2494                 bridge_forward(sc, bif, mc);
2495
2496                 /*
2497                  * Reinject the mbuf as arriving on the bridge so we have a
2498                  * chance at claiming multicast packets. We can not loop back
2499                  * here from ether_input as a bridge is never a member of a
2500                  * bridge.
2501                  */
2502                 KASSERT(bifp->if_bridge == NULL,
2503                     ("loop created in bridge_input"));
2504                 mc2 = m_dup(m, M_NOWAIT);
2505                 if (mc2 != NULL) {
2506                         /* Keep the layer3 header aligned */
2507                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2508                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2509                 }
2510                 if (mc2 != NULL) {
2511                         mc2->m_pkthdr.rcvif = bifp;
2512                         (*bifp->if_input)(bifp, mc2);
2513                 }
2514
2515                 /* Return the original packet for local processing. */
2516                 NET_EPOCH_EXIT_ET(et);
2517                 return (m);
2518         }
2519
2520         if ((bif->bif_flags & IFBIF_STP) &&
2521             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2522                 NET_EPOCH_EXIT_ET(et);
2523                 return (m);
2524         }
2525
2526 #if (defined(INET) || defined(INET6))
2527 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2528         || ((iface)->if_carp \
2529             && (*carp_forus_p)((iface), eh->ether_dhost))
2530 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2531         || ((iface)->if_carp \
2532             && (*carp_forus_p)((iface), eh->ether_shost))
2533 #else
2534 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2535 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2536 #endif
2537
2538 #ifdef INET6
2539 #   define OR_PFIL_HOOKED_INET6 \
2540         || PFIL_HOOKED(&V_inet6_pfil_hook)
2541 #else
2542 #   define OR_PFIL_HOOKED_INET6
2543 #endif
2544
2545 #define GRAB_OUR_PACKETS(iface) \
2546         if ((iface)->if_type == IFT_GIF) \
2547                 continue; \
2548         /* It is destined for us. */ \
2549         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2550             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2551             ) {                                                         \
2552                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2553                         ETHER_BPF_MTAP(iface, m);                       \
2554                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2555                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2556                         /* Filter on the physical interface. */         \
2557                         if (V_pfil_local_phys &&                        \
2558                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2559                              OR_PFIL_HOOKED_INET6)) {                   \
2560                                 if (bridge_pfil(&m, NULL, ifp,          \
2561                                     PFIL_IN) != 0 || m == NULL) {       \
2562                                         NET_EPOCH_EXIT_ET(et);  \
2563                                         return (NULL);                  \
2564                                 }                                       \
2565                                 eh = mtod(m, struct ether_header *);    \
2566                         }                                               \
2567                 }                                                       \
2568                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2569                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2570                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2571                         if (error && bif->bif_addrmax) {                \
2572                                 m_freem(m);                             \
2573                                 NET_EPOCH_EXIT_ET(et);                  \
2574                                 return (NULL);                          \
2575                         }                                               \
2576                 }                                                       \
2577                 m->m_pkthdr.rcvif = iface;                              \
2578                 NET_EPOCH_EXIT_ET(et);                                  \
2579                 return (m);                                             \
2580         }                                                               \
2581                                                                         \
2582         /* We just received a packet that we sent out. */               \
2583         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2584             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2585             ) {                                                         \
2586                 m_freem(m);                                             \
2587                 NET_EPOCH_EXIT_ET(et);                                  \
2588                 return (NULL);                                          \
2589         }
2590
2591         /*
2592          * Unicast.  Make sure it's not for the bridge.
2593          */
2594         do { GRAB_OUR_PACKETS(bifp) } while (0);
2595
2596         /*
2597          * Give a chance for ifp at first priority. This will help when the
2598          * packet comes through the interface like VLAN's with the same MACs
2599          * on several interfaces from the same bridge. This also will save
2600          * some CPU cycles in case the destination interface and the input
2601          * interface (eq ifp) are the same.
2602          */
2603         do { GRAB_OUR_PACKETS(ifp) } while (0);
2604
2605         /* Now check the all bridge members. */
2606         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2607                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2608         }
2609
2610 #undef OR_CARP_CHECK_WE_ARE_DST
2611 #undef OR_CARP_CHECK_WE_ARE_SRC
2612 #undef OR_PFIL_HOOKED_INET6
2613 #undef GRAB_OUR_PACKETS
2614
2615         /* Perform the bridge forwarding function. */
2616         bridge_forward(sc, bif, m);
2617
2618         NET_EPOCH_EXIT_ET(et);
2619         return (NULL);
2620 }
2621
2622 /*
2623  * bridge_broadcast:
2624  *
2625  *      Send a frame to all interfaces that are members of
2626  *      the bridge, except for the one on which the packet
2627  *      arrived.
2628  *
2629  *      NOTE: Releases the lock on return.
2630  */
2631 static void
2632 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2633     struct mbuf *m, int runfilt)
2634 {
2635         struct bridge_iflist *dbif, *sbif;
2636         struct mbuf *mc;
2637         struct ifnet *dst_if;
2638         int used = 0, i;
2639
2640         MPASS(in_epoch(net_epoch_preempt));
2641
2642         sbif = bridge_lookup_member_if(sc, src_if);
2643
2644         /* Filter on the bridge interface before broadcasting */
2645         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2646 #ifdef INET6
2647             || PFIL_HOOKED(&V_inet6_pfil_hook)
2648 #endif
2649             )) {
2650                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2651                         return;
2652                 if (m == NULL)
2653                         return;
2654         }
2655
2656         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2657                 dst_if = dbif->bif_ifp;
2658                 if (dst_if == src_if)
2659                         continue;
2660
2661                 /* Private segments can not talk to each other */
2662                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2663                         continue;
2664
2665                 if ((dbif->bif_flags & IFBIF_STP) &&
2666                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2667                         continue;
2668
2669                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2670                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2671                         continue;
2672
2673                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2674                         continue;
2675
2676                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2677                         mc = m;
2678                         used = 1;
2679                 } else {
2680                         mc = m_dup(m, M_NOWAIT);
2681                         if (mc == NULL) {
2682                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2683                                 continue;
2684                         }
2685                 }
2686
2687                 /*
2688                  * Filter on the output interface. Pass a NULL bridge interface
2689                  * pointer so we do not redundantly filter on the bridge for
2690                  * each interface we broadcast on.
2691                  */
2692                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2693 #ifdef INET6
2694                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2695 #endif
2696                     )) {
2697                         if (used == 0) {
2698                                 /* Keep the layer3 header aligned */
2699                                 i = min(mc->m_pkthdr.len, max_protohdr);
2700                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2701                                 if (mc == NULL) {
2702                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2703                                         continue;
2704                                 }
2705                         }
2706                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2707                                 continue;
2708                         if (mc == NULL)
2709                                 continue;
2710                 }
2711
2712                 bridge_enqueue(sc, dst_if, mc);
2713         }
2714         if (used == 0)
2715                 m_freem(m);
2716 }
2717
2718 /*
2719  * bridge_span:
2720  *
2721  *      Duplicate a packet out one or more interfaces that are in span mode,
2722  *      the original mbuf is unmodified.
2723  */
2724 static void
2725 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2726 {
2727         struct bridge_iflist *bif;
2728         struct ifnet *dst_if;
2729         struct mbuf *mc;
2730
2731         MPASS(in_epoch(net_epoch_preempt));
2732
2733         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2734                 return;
2735
2736         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2737                 dst_if = bif->bif_ifp;
2738
2739                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2740                         continue;
2741
2742                 mc = m_copypacket(m, M_NOWAIT);
2743                 if (mc == NULL) {
2744                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2745                         continue;
2746                 }
2747
2748                 bridge_enqueue(sc, dst_if, mc);
2749         }
2750 }
2751
2752 /*
2753  * bridge_rtupdate:
2754  *
2755  *      Add a bridge routing entry.
2756  */
2757 static int
2758 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2759     struct bridge_iflist *bif, int setflags, uint8_t flags)
2760 {
2761         struct bridge_rtnode *brt;
2762         int error;
2763
2764         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2765
2766         /* Check the source address is valid and not multicast. */
2767         if (ETHER_IS_MULTICAST(dst) ||
2768             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2769              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2770                 return (EINVAL);
2771
2772         /* 802.1p frames map to vlan 1 */
2773         if (vlan == 0)
2774                 vlan = 1;
2775
2776         /*
2777          * A route for this destination might already exist.  If so,
2778          * update it, otherwise create a new one.
2779          */
2780         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2781                 BRIDGE_RT_LOCK(sc);
2782
2783                 /* Check again, now that we have the lock. There could have
2784                  * been a race and we only want to insert this once. */
2785                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2786                         BRIDGE_RT_UNLOCK(sc);
2787                         return (0);
2788                 }
2789
2790                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2791                         sc->sc_brtexceeded++;
2792                         BRIDGE_RT_UNLOCK(sc);
2793                         return (ENOSPC);
2794                 }
2795                 /* Check per interface address limits (if enabled) */
2796                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2797                         bif->bif_addrexceeded++;
2798                         BRIDGE_RT_UNLOCK(sc);
2799                         return (ENOSPC);
2800                 }
2801
2802                 /*
2803                  * Allocate a new bridge forwarding node, and
2804                  * initialize the expiration time and Ethernet
2805                  * address.
2806                  */
2807                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2808                 if (brt == NULL) {
2809                         BRIDGE_RT_UNLOCK(sc);
2810                         return (ENOMEM);
2811                 }
2812                 brt->brt_vnet = curvnet;
2813
2814                 if (bif->bif_flags & IFBIF_STICKY)
2815                         brt->brt_flags = IFBAF_STICKY;
2816                 else
2817                         brt->brt_flags = IFBAF_DYNAMIC;
2818
2819                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2820                 brt->brt_vlan = vlan;
2821
2822                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2823                         uma_zfree(V_bridge_rtnode_zone, brt);
2824                         BRIDGE_RT_UNLOCK(sc);
2825                         return (error);
2826                 }
2827                 brt->brt_dst = bif;
2828                 bif->bif_addrcnt++;
2829
2830                 BRIDGE_RT_UNLOCK(sc);
2831         }
2832
2833         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2834             brt->brt_dst != bif) {
2835                 BRIDGE_RT_LOCK(sc);
2836                 brt->brt_dst->bif_addrcnt--;
2837                 brt->brt_dst = bif;
2838                 brt->brt_dst->bif_addrcnt++;
2839                 BRIDGE_RT_UNLOCK(sc);
2840         }
2841
2842         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2843                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2844         if (setflags)
2845                 brt->brt_flags = flags;
2846
2847         return (0);
2848 }
2849
2850 /*
2851  * bridge_rtlookup:
2852  *
2853  *      Lookup the destination interface for an address.
2854  */
2855 static struct ifnet *
2856 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2857 {
2858         struct bridge_rtnode *brt;
2859
2860         MPASS(in_epoch(net_epoch_preempt));
2861
2862         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2863                 return (NULL);
2864
2865         return (brt->brt_ifp);
2866 }
2867
2868 /*
2869  * bridge_rttrim:
2870  *
2871  *      Trim the routine table so that we have a number
2872  *      of routing entries less than or equal to the
2873  *      maximum number.
2874  */
2875 static void
2876 bridge_rttrim(struct bridge_softc *sc)
2877 {
2878         struct bridge_rtnode *brt, *nbrt;
2879
2880         MPASS(in_epoch(net_epoch_preempt));
2881         BRIDGE_RT_LOCK_ASSERT(sc);
2882
2883         /* Make sure we actually need to do this. */
2884         if (sc->sc_brtcnt <= sc->sc_brtmax)
2885                 return;
2886
2887         /* Force an aging cycle; this might trim enough addresses. */
2888         bridge_rtage(sc);
2889         if (sc->sc_brtcnt <= sc->sc_brtmax)
2890                 return;
2891
2892         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2893                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2894                         bridge_rtnode_destroy(sc, brt);
2895                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2896                                 return;
2897                 }
2898         }
2899 }
2900
2901 /*
2902  * bridge_timer:
2903  *
2904  *      Aging timer for the bridge.
2905  */
2906 static void
2907 bridge_timer(void *arg)
2908 {
2909         struct bridge_softc *sc = arg;
2910
2911         BRIDGE_RT_LOCK_ASSERT(sc);
2912
2913         /* Destruction of rtnodes requires a proper vnet context */
2914         CURVNET_SET(sc->sc_ifp->if_vnet);
2915         bridge_rtage(sc);
2916
2917         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2918                 callout_reset(&sc->sc_brcallout,
2919                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2920         CURVNET_RESTORE();
2921 }
2922
2923 /*
2924  * bridge_rtage:
2925  *
2926  *      Perform an aging cycle.
2927  */
2928 static void
2929 bridge_rtage(struct bridge_softc *sc)
2930 {
2931         struct bridge_rtnode *brt, *nbrt;
2932
2933         BRIDGE_RT_LOCK_ASSERT(sc);
2934
2935         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2936                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2937                         if (time_uptime >= brt->brt_expire)
2938                                 bridge_rtnode_destroy(sc, brt);
2939                 }
2940         }
2941 }
2942
2943 /*
2944  * bridge_rtflush:
2945  *
2946  *      Remove all dynamic addresses from the bridge.
2947  */
2948 static void
2949 bridge_rtflush(struct bridge_softc *sc, int full)
2950 {
2951         struct bridge_rtnode *brt, *nbrt;
2952
2953         BRIDGE_RT_LOCK_ASSERT(sc);
2954
2955         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2956                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2957                         bridge_rtnode_destroy(sc, brt);
2958         }
2959 }
2960
2961 /*
2962  * bridge_rtdaddr:
2963  *
2964  *      Remove an address from the table.
2965  */
2966 static int
2967 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2968 {
2969         struct bridge_rtnode *brt;
2970         int found = 0;
2971
2972         BRIDGE_RT_LOCK(sc);
2973
2974         /*
2975          * If vlan is zero then we want to delete for all vlans so the lookup
2976          * may return more than one.
2977          */
2978         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2979                 bridge_rtnode_destroy(sc, brt);
2980                 found = 1;
2981         }
2982
2983         BRIDGE_RT_UNLOCK(sc);
2984
2985         return (found ? 0 : ENOENT);
2986 }
2987
2988 /*
2989  * bridge_rtdelete:
2990  *
2991  *      Delete routes to a speicifc member interface.
2992  */
2993 static void
2994 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2995 {
2996         struct bridge_rtnode *brt, *nbrt;
2997
2998         BRIDGE_RT_LOCK_ASSERT(sc);
2999
3000         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3001                 if (brt->brt_ifp == ifp && (full ||
3002                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3003                         bridge_rtnode_destroy(sc, brt);
3004         }
3005 }
3006
3007 /*
3008  * bridge_rtable_init:
3009  *
3010  *      Initialize the route table for this bridge.
3011  */
3012 static void
3013 bridge_rtable_init(struct bridge_softc *sc)
3014 {
3015         int i;
3016
3017         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3018             M_DEVBUF, M_WAITOK);
3019
3020         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3021                 CK_LIST_INIT(&sc->sc_rthash[i]);
3022
3023         sc->sc_rthash_key = arc4random();
3024         CK_LIST_INIT(&sc->sc_rtlist);
3025 }
3026
3027 /*
3028  * bridge_rtable_fini:
3029  *
3030  *      Deconstruct the route table for this bridge.
3031  */
3032 static void
3033 bridge_rtable_fini(struct bridge_softc *sc)
3034 {
3035
3036         KASSERT(sc->sc_brtcnt == 0,
3037             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3038         free(sc->sc_rthash, M_DEVBUF);
3039 }
3040
3041 /*
3042  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3043  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3044  */
3045 #define mix(a, b, c)                                                    \
3046 do {                                                                    \
3047         a -= b; a -= c; a ^= (c >> 13);                                 \
3048         b -= c; b -= a; b ^= (a << 8);                                  \
3049         c -= a; c -= b; c ^= (b >> 13);                                 \
3050         a -= b; a -= c; a ^= (c >> 12);                                 \
3051         b -= c; b -= a; b ^= (a << 16);                                 \
3052         c -= a; c -= b; c ^= (b >> 5);                                  \
3053         a -= b; a -= c; a ^= (c >> 3);                                  \
3054         b -= c; b -= a; b ^= (a << 10);                                 \
3055         c -= a; c -= b; c ^= (b >> 15);                                 \
3056 } while (/*CONSTCOND*/0)
3057
3058 static __inline uint32_t
3059 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3060 {
3061         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3062
3063         b += addr[5] << 8;
3064         b += addr[4];
3065         a += addr[3] << 24;
3066         a += addr[2] << 16;
3067         a += addr[1] << 8;
3068         a += addr[0];
3069
3070         mix(a, b, c);
3071
3072         return (c & BRIDGE_RTHASH_MASK);
3073 }
3074
3075 #undef mix
3076
3077 static int
3078 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3079 {
3080         int i, d;
3081
3082         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3083                 d = ((int)a[i]) - ((int)b[i]);
3084         }
3085
3086         return (d);
3087 }
3088
3089 /*
3090  * bridge_rtnode_lookup:
3091  *
3092  *      Look up a bridge route node for the specified destination. Compare the
3093  *      vlan id or if zero then just return the first match.
3094  */
3095 static struct bridge_rtnode *
3096 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3097 {
3098         struct bridge_rtnode *brt;
3099         uint32_t hash;
3100         int dir;
3101
3102         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3103
3104         hash = bridge_rthash(sc, addr);
3105         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3106                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3107                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3108                         return (brt);
3109                 if (dir > 0)
3110                         return (NULL);
3111         }
3112
3113         return (NULL);
3114 }
3115
3116 /*
3117  * bridge_rtnode_insert:
3118  *
3119  *      Insert the specified bridge node into the route table.  We
3120  *      assume the entry is not already in the table.
3121  */
3122 static int
3123 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3124 {
3125         struct bridge_rtnode *lbrt;
3126         uint32_t hash;
3127         int dir;
3128
3129         BRIDGE_RT_LOCK_ASSERT(sc);
3130
3131         hash = bridge_rthash(sc, brt->brt_addr);
3132
3133         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3134         if (lbrt == NULL) {
3135                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3136                 goto out;
3137         }
3138
3139         do {
3140                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3141                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3142                         return (EEXIST);
3143                 if (dir > 0) {
3144                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3145                         goto out;
3146                 }
3147                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3148                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3149                         goto out;
3150                 }
3151                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3152         } while (lbrt != NULL);
3153
3154 #ifdef DIAGNOSTIC
3155         panic("bridge_rtnode_insert: impossible");
3156 #endif
3157
3158 out:
3159         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3160         sc->sc_brtcnt++;
3161
3162         return (0);
3163 }
3164
3165 static void
3166 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3167 {
3168         struct bridge_rtnode *brt;
3169
3170         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3171
3172         CURVNET_SET(brt->brt_vnet);
3173         uma_zfree(V_bridge_rtnode_zone, brt);
3174         CURVNET_RESTORE();
3175 }
3176
3177 /*
3178  * bridge_rtnode_destroy:
3179  *
3180  *      Destroy a bridge rtnode.
3181  */
3182 static void
3183 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3184 {
3185
3186         BRIDGE_RT_LOCK_ASSERT(sc);
3187
3188         CK_LIST_REMOVE(brt, brt_hash);
3189
3190         CK_LIST_REMOVE(brt, brt_list);
3191         sc->sc_brtcnt--;
3192         brt->brt_dst->bif_addrcnt--;
3193
3194         epoch_call(net_epoch_preempt, &brt->brt_epoch_ctx,
3195             bridge_rtnode_destroy_cb);
3196 }
3197
3198 /*
3199  * bridge_rtable_expire:
3200  *
3201  *      Set the expiry time for all routes on an interface.
3202  */
3203 static void
3204 bridge_rtable_expire(struct ifnet *ifp, int age)
3205 {
3206         struct bridge_softc *sc = ifp->if_bridge;
3207         struct bridge_rtnode *brt;
3208
3209         CURVNET_SET(ifp->if_vnet);
3210         BRIDGE_RT_LOCK(sc);
3211
3212         /*
3213          * If the age is zero then flush, otherwise set all the expiry times to
3214          * age for the interface
3215          */
3216         if (age == 0)
3217                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3218         else {
3219                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3220                         /* Cap the expiry time to 'age' */
3221                         if (brt->brt_ifp == ifp &&
3222                             brt->brt_expire > time_uptime + age &&
3223                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3224                                 brt->brt_expire = time_uptime + age;
3225                 }
3226         }
3227         BRIDGE_RT_UNLOCK(sc);
3228         CURVNET_RESTORE();
3229 }
3230
3231 /*
3232  * bridge_state_change:
3233  *
3234  *      Callback from the bridgestp code when a port changes states.
3235  */
3236 static void
3237 bridge_state_change(struct ifnet *ifp, int state)
3238 {
3239         struct bridge_softc *sc = ifp->if_bridge;
3240         static const char *stpstates[] = {
3241                 "disabled",
3242                 "listening",
3243                 "learning",
3244                 "forwarding",
3245                 "blocking",
3246                 "discarding"
3247         };
3248
3249         CURVNET_SET(ifp->if_vnet);
3250         if (V_log_stp)
3251                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3252                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3253         CURVNET_RESTORE();
3254 }
3255
3256 /*
3257  * Send bridge packets through pfil if they are one of the types pfil can deal
3258  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3259  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3260  * that interface.
3261  */
3262 static int
3263 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3264 {
3265         int snap, error, i, hlen;
3266         struct ether_header *eh1, eh2;
3267         struct ip *ip;
3268         struct llc llc1;
3269         u_int16_t ether_type;
3270
3271         snap = 0;
3272         error = -1;     /* Default error if not error == 0 */
3273
3274 #if 0
3275         /* we may return with the IP fields swapped, ensure its not shared */
3276         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3277 #endif
3278
3279         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3280                 return (0); /* filtering is disabled */
3281
3282         i = min((*mp)->m_pkthdr.len, max_protohdr);
3283         if ((*mp)->m_len < i) {
3284             *mp = m_pullup(*mp, i);
3285             if (*mp == NULL) {
3286                 printf("%s: m_pullup failed\n", __func__);
3287                 return (-1);
3288             }
3289         }
3290
3291         eh1 = mtod(*mp, struct ether_header *);
3292         ether_type = ntohs(eh1->ether_type);
3293
3294         /*
3295          * Check for SNAP/LLC.
3296          */
3297         if (ether_type < ETHERMTU) {
3298                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3299
3300                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3301                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3302                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3303                     llc2->llc_control == LLC_UI) {
3304                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3305                         snap = 1;
3306                 }
3307         }
3308
3309         /*
3310          * If we're trying to filter bridge traffic, don't look at anything
3311          * other than IP and ARP traffic.  If the filter doesn't understand
3312          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3313          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3314          * but of course we don't have an AppleTalk filter to begin with.
3315          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3316          * ARP traffic.)
3317          */
3318         switch (ether_type) {
3319                 case ETHERTYPE_ARP:
3320                 case ETHERTYPE_REVARP:
3321                         if (V_pfil_ipfw_arp == 0)
3322                                 return (0); /* Automatically pass */
3323                         break;
3324
3325                 case ETHERTYPE_IP:
3326 #ifdef INET6
3327                 case ETHERTYPE_IPV6:
3328 #endif /* INET6 */
3329                         break;
3330                 default:
3331                         /*
3332                          * Check to see if the user wants to pass non-ip
3333                          * packets, these will not be checked by pfil(9) and
3334                          * passed unconditionally so the default is to drop.
3335                          */
3336                         if (V_pfil_onlyip)
3337                                 goto bad;
3338         }
3339
3340         /* Run the packet through pfil before stripping link headers */
3341         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3342                         dir == PFIL_OUT && ifp != NULL) {
3343
3344                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, 0,
3345                     NULL);
3346
3347                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3348                         return (error);
3349         }
3350
3351         /* Strip off the Ethernet header and keep a copy. */
3352         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3353         m_adj(*mp, ETHER_HDR_LEN);
3354
3355         /* Strip off snap header, if present */
3356         if (snap) {
3357                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3358                 m_adj(*mp, sizeof(struct llc));
3359         }
3360
3361         /*
3362          * Check the IP header for alignment and errors
3363          */
3364         if (dir == PFIL_IN) {
3365                 switch (ether_type) {
3366                         case ETHERTYPE_IP:
3367                                 error = bridge_ip_checkbasic(mp);
3368                                 break;
3369 #ifdef INET6
3370                         case ETHERTYPE_IPV6:
3371                                 error = bridge_ip6_checkbasic(mp);
3372                                 break;
3373 #endif /* INET6 */
3374                         default:
3375                                 error = 0;
3376                 }
3377                 if (error)
3378                         goto bad;
3379         }
3380
3381         error = 0;
3382
3383         /*
3384          * Run the packet through pfil
3385          */
3386         switch (ether_type) {
3387         case ETHERTYPE_IP:
3388                 /*
3389                  * Run pfil on the member interface and the bridge, both can
3390                  * be skipped by clearing pfil_member or pfil_bridge.
3391                  *
3392                  * Keep the order:
3393                  *   in_if -> bridge_if -> out_if
3394                  */
3395                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3396                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3397                                         dir, 0, NULL);
3398
3399                 if (*mp == NULL || error != 0) /* filter may consume */
3400                         break;
3401
3402                 if (V_pfil_member && ifp != NULL)
3403                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3404                                         dir, 0, NULL);
3405
3406                 if (*mp == NULL || error != 0) /* filter may consume */
3407                         break;
3408
3409                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3410                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3411                                         dir, 0, NULL);
3412
3413                 if (*mp == NULL || error != 0) /* filter may consume */
3414                         break;
3415
3416                 /* check if we need to fragment the packet */
3417                 /* bridge_fragment generates a mbuf chain of packets */
3418                 /* that already include eth headers */
3419                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3420                         i = (*mp)->m_pkthdr.len;
3421                         if (i > ifp->if_mtu) {
3422                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3423                                             &llc1);
3424                                 return (error);
3425                         }
3426                 }
3427
3428                 /* Recalculate the ip checksum. */
3429                 ip = mtod(*mp, struct ip *);
3430                 hlen = ip->ip_hl << 2;
3431                 if (hlen < sizeof(struct ip))
3432                         goto bad;
3433                 if (hlen > (*mp)->m_len) {
3434                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3435                                 goto bad;
3436                         ip = mtod(*mp, struct ip *);
3437                         if (ip == NULL)
3438                                 goto bad;
3439                 }
3440                 ip->ip_sum = 0;
3441                 if (hlen == sizeof(struct ip))
3442                         ip->ip_sum = in_cksum_hdr(ip);
3443                 else
3444                         ip->ip_sum = in_cksum(*mp, hlen);
3445
3446                 break;
3447 #ifdef INET6
3448         case ETHERTYPE_IPV6:
3449                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3450                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3451                                         dir, 0, NULL);
3452
3453                 if (*mp == NULL || error != 0) /* filter may consume */
3454                         break;
3455
3456                 if (V_pfil_member && ifp != NULL)
3457                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3458                                         dir, 0, NULL);
3459
3460                 if (*mp == NULL || error != 0) /* filter may consume */
3461                         break;
3462
3463                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3464                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3465                                         dir, 0, NULL);
3466                 break;
3467 #endif
3468         default:
3469                 error = 0;
3470                 break;
3471         }
3472
3473         if (*mp == NULL)
3474                 return (error);
3475         if (error != 0)
3476                 goto bad;
3477
3478         error = -1;
3479
3480         /*
3481          * Finally, put everything back the way it was and return
3482          */
3483         if (snap) {
3484                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3485                 if (*mp == NULL)
3486                         return (error);
3487                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3488         }
3489
3490         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3491         if (*mp == NULL)
3492                 return (error);
3493         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3494
3495         return (0);
3496
3497 bad:
3498         m_freem(*mp);
3499         *mp = NULL;
3500         return (error);
3501 }
3502
3503 /*
3504  * Perform basic checks on header size since
3505  * pfil assumes ip_input has already processed
3506  * it for it.  Cut-and-pasted from ip_input.c.
3507  * Given how simple the IPv6 version is,
3508  * does the IPv4 version really need to be
3509  * this complicated?
3510  *
3511  * XXX Should we update ipstat here, or not?
3512  * XXX Right now we update ipstat but not
3513  * XXX csum_counter.
3514  */
3515 static int
3516 bridge_ip_checkbasic(struct mbuf **mp)
3517 {
3518         struct mbuf *m = *mp;
3519         struct ip *ip;
3520         int len, hlen;
3521         u_short sum;
3522
3523         if (*mp == NULL)
3524                 return (-1);
3525
3526         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3527                 if ((m = m_copyup(m, sizeof(struct ip),
3528                         (max_linkhdr + 3) & ~3)) == NULL) {
3529                         /* XXXJRT new stat, please */
3530                         KMOD_IPSTAT_INC(ips_toosmall);
3531                         goto bad;
3532                 }
3533         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3534                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3535                         KMOD_IPSTAT_INC(ips_toosmall);
3536                         goto bad;
3537                 }
3538         }
3539         ip = mtod(m, struct ip *);
3540         if (ip == NULL) goto bad;
3541
3542         if (ip->ip_v != IPVERSION) {
3543                 KMOD_IPSTAT_INC(ips_badvers);
3544                 goto bad;
3545         }
3546         hlen = ip->ip_hl << 2;
3547         if (hlen < sizeof(struct ip)) { /* minimum header length */
3548                 KMOD_IPSTAT_INC(ips_badhlen);
3549                 goto bad;
3550         }
3551         if (hlen > m->m_len) {
3552                 if ((m = m_pullup(m, hlen)) == NULL) {
3553                         KMOD_IPSTAT_INC(ips_badhlen);
3554                         goto bad;
3555                 }
3556                 ip = mtod(m, struct ip *);
3557                 if (ip == NULL) goto bad;
3558         }
3559
3560         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3561                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3562         } else {
3563                 if (hlen == sizeof(struct ip)) {
3564                         sum = in_cksum_hdr(ip);
3565                 } else {
3566                         sum = in_cksum(m, hlen);
3567                 }
3568         }
3569         if (sum) {
3570                 KMOD_IPSTAT_INC(ips_badsum);
3571                 goto bad;
3572         }
3573
3574         /* Retrieve the packet length. */
3575         len = ntohs(ip->ip_len);
3576
3577         /*
3578          * Check for additional length bogosity
3579          */
3580         if (len < hlen) {
3581                 KMOD_IPSTAT_INC(ips_badlen);
3582                 goto bad;
3583         }
3584
3585         /*
3586          * Check that the amount of data in the buffers
3587          * is as at least much as the IP header would have us expect.
3588          * Drop packet if shorter than we expect.
3589          */
3590         if (m->m_pkthdr.len < len) {
3591                 KMOD_IPSTAT_INC(ips_tooshort);
3592                 goto bad;
3593         }
3594
3595         /* Checks out, proceed */
3596         *mp = m;
3597         return (0);
3598
3599 bad:
3600         *mp = m;
3601         return (-1);
3602 }
3603
3604 #ifdef INET6
3605 /*
3606  * Same as above, but for IPv6.
3607  * Cut-and-pasted from ip6_input.c.
3608  * XXX Should we update ip6stat, or not?
3609  */
3610 static int
3611 bridge_ip6_checkbasic(struct mbuf **mp)
3612 {
3613         struct mbuf *m = *mp;
3614         struct ip6_hdr *ip6;
3615
3616         /*
3617          * If the IPv6 header is not aligned, slurp it up into a new
3618          * mbuf with space for link headers, in the event we forward
3619          * it.  Otherwise, if it is aligned, make sure the entire base
3620          * IPv6 header is in the first mbuf of the chain.
3621          */
3622         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3623                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3624                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3625                             (max_linkhdr + 3) & ~3)) == NULL) {
3626                         /* XXXJRT new stat, please */
3627                         IP6STAT_INC(ip6s_toosmall);
3628                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3629                         goto bad;
3630                 }
3631         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3632                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3633                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3634                         IP6STAT_INC(ip6s_toosmall);
3635                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3636                         goto bad;
3637                 }
3638         }
3639
3640         ip6 = mtod(m, struct ip6_hdr *);
3641
3642         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3643                 IP6STAT_INC(ip6s_badvers);
3644                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3645                 goto bad;
3646         }
3647
3648         /* Checks out, proceed */
3649         *mp = m;
3650         return (0);
3651
3652 bad:
3653         *mp = m;
3654         return (-1);
3655 }
3656 #endif /* INET6 */
3657
3658 /*
3659  * bridge_fragment:
3660  *
3661  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3662  */
3663 static int
3664 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3665     int snap, struct llc *llc)
3666 {
3667         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3668         struct ip *ip;
3669         int error = -1;
3670
3671         if (m->m_len < sizeof(struct ip) &&
3672             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3673                 goto dropit;
3674         ip = mtod(m, struct ip *);
3675
3676         m->m_pkthdr.csum_flags |= CSUM_IP;
3677         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3678         if (error)
3679                 goto dropit;
3680
3681         /*
3682          * Walk the chain and re-add the Ethernet header for
3683          * each mbuf packet.
3684          */
3685         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3686                 nextpkt = mcur->m_nextpkt;
3687                 mcur->m_nextpkt = NULL;
3688                 if (snap) {
3689                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3690                         if (mcur == NULL) {
3691                                 error = ENOBUFS;
3692                                 if (mprev != NULL)
3693                                         mprev->m_nextpkt = nextpkt;
3694                                 goto dropit;
3695                         }
3696                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3697                 }
3698
3699                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3700                 if (mcur == NULL) {
3701                         error = ENOBUFS;
3702                         if (mprev != NULL)
3703                                 mprev->m_nextpkt = nextpkt;
3704                         goto dropit;
3705                 }
3706                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3707
3708                 /*
3709                  * The previous two M_PREPEND could have inserted one or two
3710                  * mbufs in front so we have to update the previous packet's
3711                  * m_nextpkt.
3712                  */
3713                 mcur->m_nextpkt = nextpkt;
3714                 if (mprev != NULL)
3715                         mprev->m_nextpkt = mcur;
3716                 else {
3717                         /* The first mbuf in the original chain needs to be
3718                          * updated. */
3719                         *mp = mcur;
3720                 }
3721                 mprev = mcur;
3722         }
3723
3724         KMOD_IPSTAT_INC(ips_fragmented);
3725         return (error);
3726
3727 dropit:
3728         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3729                 m = mcur->m_nextpkt;
3730                 m_freem(mcur);
3731         }
3732         return (error);
3733 }
3734
3735 static void
3736 bridge_linkstate(struct ifnet *ifp)
3737 {
3738         struct bridge_softc *sc = ifp->if_bridge;
3739         struct bridge_iflist *bif;
3740         struct epoch_tracker et;
3741
3742         NET_EPOCH_ENTER_ET(et);
3743
3744         bif = bridge_lookup_member_if(sc, ifp);
3745         if (bif == NULL) {
3746                 NET_EPOCH_EXIT_ET(et);
3747                 return;
3748         }
3749         bridge_linkcheck(sc);
3750
3751         bstp_linkstate(&bif->bif_stp);
3752
3753         NET_EPOCH_EXIT_ET(et);
3754 }
3755
3756 static void
3757 bridge_linkcheck(struct bridge_softc *sc)
3758 {
3759         struct bridge_iflist *bif;
3760         int new_link, hasls;
3761
3762         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3763
3764         new_link = LINK_STATE_DOWN;
3765         hasls = 0;
3766         /* Our link is considered up if at least one of our ports is active */
3767         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3768                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3769                         hasls++;
3770                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3771                         new_link = LINK_STATE_UP;
3772                         break;
3773                 }
3774         }
3775         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3776                 /* If no interfaces support link-state then we default to up */
3777                 new_link = LINK_STATE_UP;
3778         }
3779         if_link_state_change(sc->sc_ifp, new_link);
3780 }