sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*
   4  * Copyright 2001 Wasabi Systems, Inc.
   5  * All rights reserved.
   6  *
   7  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. All advertising materials mentioning features or use of this software
  18  *    must display the following acknowledgement:
  19  *      This product includes software developed for the NetBSD Project by
  20  *      Wasabi Systems, Inc.
  21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  22  *    or promote products derived from this software without specific prior
  23  *    written permission.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35  * POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 /*
  39  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  40  * All rights reserved.
  41  *
  42  * Redistribution and use in source and binary forms, with or without
  43  * modification, are permitted provided that the following conditions
  44  * are met:
  45  * 1. Redistributions of source code must retain the above copyright
  46  *    notice, this list of conditions and the following disclaimer.
  47  * 2. Redistributions in binary form must reproduce the above copyright
  48  *    notice, this list of conditions and the following disclaimer in the
  49  *    documentation and/or other materials provided with the distribution.
  50  *
  51  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  52  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  53  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  54  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  55  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  56  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  57  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  59  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  60  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  61  * POSSIBILITY OF SUCH DAMAGE.
  62  *
  63  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  64  */
  65
  66 /*
  67  * Network interface bridge support.
  68  *
  69  * TODO:
  70  *
  71  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  72  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  73  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
  74  *        consider heterogenous bridges).
  75  */
  76
  77 #include <sys/cdefs.h>
  78 __FBSDID("$FreeBSD$");
  79
  80 #include "opt_inet.h"
  81 #include "opt_inet6.h"
  82
  83 #include <sys/param.h>
  84 #include <sys/mbuf.h>
  85 #include <sys/malloc.h>
  86 #include <sys/protosw.h>
  87 #include <sys/systm.h>
  88 #include <sys/jail.h>
  89 #include <sys/time.h>
  90 #include <sys/socket.h> /* for net/if.h */
  91 #include <sys/sockio.h>
  92 #include <sys/ctype.h>  /* string functions */
  93 #include <sys/kernel.h>
  94 #include <sys/random.h>
  95 #include <sys/syslog.h>
  96 #include <sys/sysctl.h>
  97 #include <vm/uma.h>
  98 #include <sys/module.h>
  99 #include <sys/priv.h>
 100 #include <sys/proc.h>
 101 #include <sys/lock.h>
 102 #include <sys/mutex.h>
 103 #include <sys/rwlock.h>
 104
 105 #include <net/bpf.h>
 106 #include <net/if.h>
 107 #include <net/if_clone.h>
 108 #include <net/if_dl.h>
 109 #include <net/if_types.h>
 110 #include <net/if_var.h>
 111 #include <net/pfil.h>
 112 #include <net/vnet.h>
 113
 114 #include <netinet/in.h> /* for struct arpcom */
 115 #include <netinet/in_systm.h>
 116 #include <netinet/in_var.h>
 117 #include <netinet/ip.h>
 118 #include <netinet/ip_var.h>
 119 #ifdef INET6
 120 #include <netinet/ip6.h>
 121 #include <netinet6/ip6_var.h>
 122 #endif
 123 #if defined(INET) || defined(INET6)
 124 #include <netinet/ip_carp.h>
 125 #endif
 126 #include <machine/in_cksum.h>
 127 #include <netinet/if_ether.h> /* for struct arpcom */
 128 #include <net/bridgestp.h>
 129 #include <net/if_bridgevar.h>
 130 #include <net/if_llc.h>
 131 #include <net/if_vlan_var.h>
 132
 133 #include <net/route.h>
 134 #include <netinet/ip_fw.h>
 135 #include <netinet/ipfw/ip_fw_private.h>
 136
 137 /*
 138  * Size of the route hash table.  Must be a power of two.
 139  */
 140 #ifndef BRIDGE_RTHASH_SIZE
 141 #define BRIDGE_RTHASH_SIZE              1024
 142 #endif
 143
 144 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 145
 146 /*
 147  * Default maximum number of addresses to cache.
 148  */
 149 #ifndef BRIDGE_RTABLE_MAX
 150 #define BRIDGE_RTABLE_MAX               2000
 151 #endif
 152
 153 /*
 154  * Timeout (in seconds) for entries learned dynamically.
 155  */
 156 #ifndef BRIDGE_RTABLE_TIMEOUT
 157 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 158 #endif
 159
 160 /*
 161  * Number of seconds between walks of the route list.
 162  */
 163 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 164 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 165 #endif
 166
 167 /*
 168  * List of capabilities to possibly mask on the member interface.
 169  */
 170 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 171
 172 /*
 173  * List of capabilities to strip
 174  */
 175 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 176
 177 /*
 178  * Bridge interface list entry.
 179  */
 180 struct bridge_iflist {
 181         LIST_ENTRY(bridge_iflist) bif_next;
 182         struct ifnet            *bif_ifp;       /* member if */
 183         struct bstp_port        bif_stp;        /* STP state */
 184         uint32_t                bif_flags;      /* member if flags */
 185         int                     bif_savedcaps;  /* saved capabilities */
 186         uint32_t                bif_addrmax;    /* max # of addresses */
 187         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 188         uint32_t                bif_addrexceeded;/* # of address violations */
 189 };
 190
 191 /*
 192  * Bridge route node.
 193  */
 194 struct bridge_rtnode {
 195         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 196         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 197         struct bridge_iflist    *brt_dst;       /* destination if */
 198         unsigned long           brt_expire;     /* expiration time */
 199         uint8_t                 brt_flags;      /* address flags */
 200         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 201         uint16_t                brt_vlan;       /* vlan id */
 202 };
 203 #define brt_ifp                 brt_dst->bif_ifp
 204
 205 /*
 206  * Software state for each bridge.
 207  */
 208 struct bridge_softc {
 209         struct ifnet            *sc_ifp;        /* make this an interface */
 210         LIST_ENTRY(bridge_softc) sc_list;
 211         struct mtx              sc_mtx;
 212         struct cv               sc_cv;
 213         uint32_t                sc_brtmax;      /* max # of addresses */
 214         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 215         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 216         struct callout          sc_brcallout;   /* bridge callout */
 217         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 218         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 219         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 220         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 221         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 222         uint32_t                sc_rthash_key;  /* key for hash */
 223         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 224         struct bstp_state       sc_stp;         /* STP state */
 225         uint32_t                sc_brtexceeded; /* # of cache drops */
 226         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 227         u_char                  sc_defaddr[6];  /* Default MAC address */
 228 };
 229
 230 static struct mtx       bridge_list_mtx;
 231 eventhandler_tag        bridge_detach_cookie = NULL;
 232
 233 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 234
 235 uma_zone_t bridge_rtnode_zone;
 236
 237 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 238 static void     bridge_clone_destroy(struct ifnet *);
 239
 240 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 241 static void     bridge_mutecaps(struct bridge_softc *);
 242 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 243                     int);
 244 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 245 static void     bridge_init(void *);
 246 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 247 static void     bridge_stop(struct ifnet *, int);
 248 static void     bridge_start(struct ifnet *);
 249 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 250 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 251                     struct rtentry *);
 252 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 253                     struct mbuf *);
 254 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 255
 256 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 257                     struct mbuf *m);
 258
 259 static void     bridge_timer(void *);
 260
 261 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 262                     struct mbuf *, int);
 263 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 264
 265 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 266                     uint16_t, struct bridge_iflist *, int, uint8_t);
 267 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 268                     uint16_t);
 269 static void     bridge_rttrim(struct bridge_softc *);
 270 static void     bridge_rtage(struct bridge_softc *);
 271 static void     bridge_rtflush(struct bridge_softc *, int);
 272 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 273                     uint16_t);
 274
 275 static int      bridge_rtable_init(struct bridge_softc *);
 276 static void     bridge_rtable_fini(struct bridge_softc *);
 277
 278 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 279 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 280                     const uint8_t *, uint16_t);
 281 static int      bridge_rtnode_insert(struct bridge_softc *,
 282                     struct bridge_rtnode *);
 283 static void     bridge_rtnode_destroy(struct bridge_softc *,
 284                     struct bridge_rtnode *);
 285 static void     bridge_rtable_expire(struct ifnet *, int);
 286 static void     bridge_state_change(struct ifnet *, int);
 287
 288 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 289                     const char *name);
 290 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 291                     struct ifnet *ifp);
 292 static void     bridge_delete_member(struct bridge_softc *,
 293                     struct bridge_iflist *, int);
 294 static void     bridge_delete_span(struct bridge_softc *,
 295                     struct bridge_iflist *);
 296
 297 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 298 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 328 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 329                     int);
 330 static int      bridge_ip_checkbasic(struct mbuf **mp);
 331 #ifdef INET6
 332 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 333 #endif /* INET6 */
 334 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 335                     struct ether_header *, int, struct llc *);
 336 static void     bridge_linkstate(struct ifnet *ifp);
 337 static void     bridge_linkcheck(struct bridge_softc *sc);
 338
 339 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
 340
 341 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 342 #define VLANTAGOF(_m)   \
 343     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 344
 345 static struct bstp_cb_ops bridge_ops = {
 346         .bcb_state = bridge_state_change,
 347         .bcb_rtage = bridge_rtable_expire
 348 };
 349
 350 SYSCTL_DECL(_net_link);
 351 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 352
 353 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 354 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 355 static int pfil_member = 1; /* run pfil hooks on the member interface */
 356 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 357 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 358 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 359                                    locally destined packets */
 360 static int log_stp   = 0;   /* log STP state changes */
 361 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
 362 TUNABLE_INT("net.link.bridge.pfil_onlyip", &pfil_onlyip);
 363 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 364     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 365 TUNABLE_INT("net.link.bridge.ipfw_arp", &pfil_ipfw_arp);
 366 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 367     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 368 TUNABLE_INT("net.link.bridge.pfil_bridge", &pfil_bridge);
 369 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 370     &pfil_bridge, 0, "Packet filter on the bridge interface");
 371 TUNABLE_INT("net.link.bridge.pfil_member", &pfil_member);
 372 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 373     &pfil_member, 0, "Packet filter on the member interface");
 374 TUNABLE_INT("net.link.bridge.pfil_local_phys", &pfil_local_phys);
 375 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 376     &pfil_local_phys, 0,
 377     "Packet filter on the physical interface for locally destined packets");
 378 TUNABLE_INT("net.link.bridge.log_stp", &log_stp);
 379 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 380     &log_stp, 0, "Log STP state changes");
 381 TUNABLE_INT("net.link.bridge.inherit_mac", &bridge_inherit_mac);
 382 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
 383     &bridge_inherit_mac, 0,
 384     "Inherit MAC address from the first bridge member");
 385
 386 struct bridge_control {
 387         int     (*bc_func)(struct bridge_softc *, void *);
 388         int     bc_argsize;
 389         int     bc_flags;
 390 };
 391
 392 #define BC_F_COPYIN             0x01    /* copy arguments in */
 393 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 394 #define BC_F_SUSER              0x04    /* do super-user check */
 395
 396 const struct bridge_control bridge_control_table[] = {
 397         { bridge_ioctl_add,             sizeof(struct ifbreq),
 398           BC_F_COPYIN|BC_F_SUSER },
 399         { bridge_ioctl_del,             sizeof(struct ifbreq),
 400           BC_F_COPYIN|BC_F_SUSER },
 401
 402         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 403           BC_F_COPYIN|BC_F_COPYOUT },
 404         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 405           BC_F_COPYIN|BC_F_SUSER },
 406
 407         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 408           BC_F_COPYIN|BC_F_SUSER },
 409         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 410           BC_F_COPYOUT },
 411
 412         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 413           BC_F_COPYIN|BC_F_COPYOUT },
 414         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 415           BC_F_COPYIN|BC_F_COPYOUT },
 416
 417         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 418           BC_F_COPYIN|BC_F_SUSER },
 419
 420         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 421           BC_F_COPYIN|BC_F_SUSER },
 422         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 423           BC_F_COPYOUT },
 424
 425         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 426           BC_F_COPYIN|BC_F_SUSER },
 427
 428         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 429           BC_F_COPYIN|BC_F_SUSER },
 430
 431         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 432           BC_F_COPYOUT },
 433         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 434           BC_F_COPYIN|BC_F_SUSER },
 435
 436         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 437           BC_F_COPYOUT },
 438         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 439           BC_F_COPYIN|BC_F_SUSER },
 440
 441         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 442           BC_F_COPYOUT },
 443         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 444           BC_F_COPYIN|BC_F_SUSER },
 445
 446         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 447           BC_F_COPYOUT },
 448         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 449           BC_F_COPYIN|BC_F_SUSER },
 450
 451         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 452           BC_F_COPYIN|BC_F_SUSER },
 453
 454         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 455           BC_F_COPYIN|BC_F_SUSER },
 456
 457         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 458           BC_F_COPYIN|BC_F_SUSER },
 459         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 460           BC_F_COPYIN|BC_F_SUSER },
 461
 462         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 463           BC_F_COPYOUT },
 464
 465         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 466           BC_F_COPYOUT },
 467
 468         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 469           BC_F_COPYIN|BC_F_COPYOUT },
 470
 471         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 472           BC_F_COPYIN|BC_F_SUSER },
 473
 474         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 475           BC_F_COPYIN|BC_F_SUSER },
 476
 477         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 478           BC_F_COPYIN|BC_F_SUSER },
 479
 480 };
 481 const int bridge_control_table_size =
 482     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 483
 484 LIST_HEAD(, bridge_softc) bridge_list;
 485
 486 IFC_SIMPLE_DECLARE(bridge, 0);
 487
 488 static int
 489 bridge_modevent(module_t mod, int type, void *data)
 490 {
 491
 492         switch (type) {
 493         case MOD_LOAD:
 494                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 495                 if_clone_attach(&bridge_cloner);
 496                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 497                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 498                     UMA_ALIGN_PTR, 0);
 499                 LIST_INIT(&bridge_list);
 500                 bridge_input_p = bridge_input;
 501                 bridge_output_p = bridge_output;
 502                 bridge_dn_p = bridge_dummynet;
 503                 bridge_linkstate_p = bridge_linkstate;
 504                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 505                     ifnet_departure_event, bridge_ifdetach, NULL,
 506                     EVENTHANDLER_PRI_ANY);
 507                 break;
 508         case MOD_UNLOAD:
 509                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 510                     bridge_detach_cookie);
 511                 if_clone_detach(&bridge_cloner);
 512                 uma_zdestroy(bridge_rtnode_zone);
 513                 bridge_input_p = NULL;
 514                 bridge_output_p = NULL;
 515                 bridge_dn_p = NULL;
 516                 bridge_linkstate_p = NULL;
 517                 mtx_destroy(&bridge_list_mtx);
 518                 break;
 519         default:
 520                 return (EOPNOTSUPP);
 521         }
 522         return (0);
 523 }
 524
 525 static moduledata_t bridge_mod = {
 526         "if_bridge",
 527         bridge_modevent,
 528         0
 529 };
 530
 531 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 532 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 533
 534 /*
 535  * handler for net.link.bridge.pfil_ipfw
 536  */
 537 static int
 538 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 539 {
 540         int enable = pfil_ipfw;
 541         int error;
 542
 543         error = sysctl_handle_int(oidp, &enable, 0, req);
 544         enable = (enable) ? 1 : 0;
 545
 546         if (enable != pfil_ipfw) {
 547                 pfil_ipfw = enable;
 548
 549                 /*
 550                  * Disable pfil so that ipfw doesnt run twice, if the user
 551                  * really wants both then they can re-enable pfil_bridge and/or
 552                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 553                  * layer2 type.
 554                  */
 555                 if (pfil_ipfw) {
 556                         pfil_onlyip = 0;
 557                         pfil_bridge = 0;
 558                         pfil_member = 0;
 559                 }
 560         }
 561
 562         return (error);
 563 }
 564 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 565             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 566
 567 /*
 568  * bridge_clone_create:
 569  *
 570  *      Create a new bridge instance.
 571  */
 572 static int
 573 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 574 {
 575         struct bridge_softc *sc, *sc2;
 576         struct ifnet *bifp, *ifp;
 577         int fb, retry;
 578         unsigned long hostid;
 579
 580         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 581         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 582         if (ifp == NULL) {
 583                 free(sc, M_DEVBUF);
 584                 return (ENOSPC);
 585         }
 586
 587         BRIDGE_LOCK_INIT(sc);
 588         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 589         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 590
 591         /* Initialize our routing table. */
 592         bridge_rtable_init(sc);
 593
 594         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 595
 596         LIST_INIT(&sc->sc_iflist);
 597         LIST_INIT(&sc->sc_spanlist);
 598
 599         ifp->if_softc = sc;
 600         if_initname(ifp, ifc->ifc_name, unit);
 601         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 602         ifp->if_ioctl = bridge_ioctl;
 603         ifp->if_start = bridge_start;
 604         ifp->if_init = bridge_init;
 605         ifp->if_type = IFT_BRIDGE;
 606         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 607         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 608         IFQ_SET_READY(&ifp->if_snd);
 609
 610         /*
 611          * Generate an ethernet address with a locally administered address.
 612          *
 613          * Since we are using random ethernet addresses for the bridge, it is
 614          * possible that we might have address collisions, so make sure that
 615          * this hardware address isn't already in use on another bridge.
 616          * The first try uses the hostid and falls back to arc4rand().
 617          */
 618         fb = 0;
 619         getcredhostid(curthread->td_ucred, &hostid);
 620         for (retry = 1; retry != 0;) {
 621                 if (fb || hostid == 0) {
 622                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 623                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 624                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 625                 } else {
 626                         sc->sc_defaddr[0] = 0x2;
 627                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 628                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 629                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 630                         sc->sc_defaddr[4] =  hostid        & 0xff;
 631                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 632                 }
 633
 634                 fb = 1;
 635                 retry = 0;
 636                 mtx_lock(&bridge_list_mtx);
 637                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 638                         bifp = sc2->sc_ifp;
 639                         if (memcmp(sc->sc_defaddr,
 640                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 641                                 retry = 1;
 642                 }
 643                 mtx_unlock(&bridge_list_mtx);
 644         }
 645
 646         bstp_attach(&sc->sc_stp, &bridge_ops);
 647         ether_ifattach(ifp, sc->sc_defaddr);
 648         /* Now undo some of the damage... */
 649         ifp->if_baudrate = 0;
 650         ifp->if_type = IFT_BRIDGE;
 651
 652         mtx_lock(&bridge_list_mtx);
 653         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 654         mtx_unlock(&bridge_list_mtx);
 655
 656         return (0);
 657 }
 658
 659 /*
 660  * bridge_clone_destroy:
 661  *
 662  *      Destroy a bridge instance.
 663  */
 664 static void
 665 bridge_clone_destroy(struct ifnet *ifp)
 666 {
 667         struct bridge_softc *sc = ifp->if_softc;
 668         struct bridge_iflist *bif;
 669
 670         BRIDGE_LOCK(sc);
 671
 672         bridge_stop(ifp, 1);
 673         ifp->if_flags &= ~IFF_UP;
 674
 675         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 676                 bridge_delete_member(sc, bif, 0);
 677
 678         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 679                 bridge_delete_span(sc, bif);
 680         }
 681
 682         BRIDGE_UNLOCK(sc);
 683
 684         callout_drain(&sc->sc_brcallout);
 685
 686         mtx_lock(&bridge_list_mtx);
 687         LIST_REMOVE(sc, sc_list);
 688         mtx_unlock(&bridge_list_mtx);
 689
 690         bstp_detach(&sc->sc_stp);
 691         ether_ifdetach(ifp);
 692         if_free_type(ifp, IFT_ETHER);
 693
 694         /* Tear down the routing table. */
 695         bridge_rtable_fini(sc);
 696
 697         BRIDGE_LOCK_DESTROY(sc);
 698         free(sc, M_DEVBUF);
 699 }
 700
 701 /*
 702  * bridge_ioctl:
 703  *
 704  *      Handle a control request from the operator.
 705  */
 706 static int
 707 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 708 {
 709         struct bridge_softc *sc = ifp->if_softc;
 710         struct ifreq *ifr = (struct ifreq *)data;
 711         struct bridge_iflist *bif;
 712         struct thread *td = curthread;
 713         union {
 714                 struct ifbreq ifbreq;
 715                 struct ifbifconf ifbifconf;
 716                 struct ifbareq ifbareq;
 717                 struct ifbaconf ifbaconf;
 718                 struct ifbrparam ifbrparam;
 719                 struct ifbropreq ifbropreq;
 720         } args;
 721         struct ifdrv *ifd = (struct ifdrv *) data;
 722         const struct bridge_control *bc;
 723         int error = 0;
 724
 725         switch (cmd) {
 726
 727         case SIOCADDMULTI:
 728         case SIOCDELMULTI:
 729                 break;
 730
 731         case SIOCGDRVSPEC:
 732         case SIOCSDRVSPEC:
 733                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 734                         error = EINVAL;
 735                         break;
 736                 }
 737                 bc = &bridge_control_table[ifd->ifd_cmd];
 738
 739                 if (cmd == SIOCGDRVSPEC &&
 740                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 741                         error = EINVAL;
 742                         break;
 743                 }
 744                 else if (cmd == SIOCSDRVSPEC &&
 745                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 746                         error = EINVAL;
 747                         break;
 748                 }
 749
 750                 if (bc->bc_flags & BC_F_SUSER) {
 751                         error = priv_check(td, PRIV_NET_BRIDGE);
 752                         if (error)
 753                                 break;
 754                 }
 755
 756                 if (ifd->ifd_len != bc->bc_argsize ||
 757                     ifd->ifd_len > sizeof(args)) {
 758                         error = EINVAL;
 759                         break;
 760                 }
 761
 762                 bzero(&args, sizeof(args));
 763                 if (bc->bc_flags & BC_F_COPYIN) {
 764                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 765                         if (error)
 766                                 break;
 767                 }
 768
 769                 BRIDGE_LOCK(sc);
 770                 error = (*bc->bc_func)(sc, &args);
 771                 BRIDGE_UNLOCK(sc);
 772                 if (error)
 773                         break;
 774
 775                 if (bc->bc_flags & BC_F_COPYOUT)
 776                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 777
 778                 break;
 779
 780         case SIOCSIFFLAGS:
 781                 if (!(ifp->if_flags & IFF_UP) &&
 782                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 783                         /*
 784                          * If interface is marked down and it is running,
 785                          * then stop and disable it.
 786                          */
 787                         BRIDGE_LOCK(sc);
 788                         bridge_stop(ifp, 1);
 789                         BRIDGE_UNLOCK(sc);
 790                 } else if ((ifp->if_flags & IFF_UP) &&
 791                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 792                         /*
 793                          * If interface is marked up and it is stopped, then
 794                          * start it.
 795                          */
 796                         (*ifp->if_init)(sc);
 797                 }
 798                 break;
 799
 800         case SIOCSIFMTU:
 801                 if (ifr->ifr_mtu < 576) {
 802                         error = EINVAL;
 803                         break;
 804                 }
 805                 if (LIST_EMPTY(&sc->sc_iflist)) {
 806                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 807                         break;
 808                 }
 809                 BRIDGE_LOCK(sc);
 810                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 811                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 812                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
 813                                     " != %d\n", sc->sc_ifp->if_xname,
 814                                     bif->bif_ifp->if_mtu,
 815                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 816                                 error = EINVAL;
 817                                 break;
 818                         }
 819                 }
 820                 if (!error)
 821                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 822                 BRIDGE_UNLOCK(sc);
 823                 break;
 824         default:
 825                 /*
 826                  * drop the lock as ether_ioctl() will call bridge_start() and
 827                  * cause the lock to be recursed.
 828                  */
 829                 error = ether_ioctl(ifp, cmd, data);
 830                 break;
 831         }
 832
 833         return (error);
 834 }
 835
 836 /*
 837  * bridge_mutecaps:
 838  *
 839  *      Clear or restore unwanted capabilities on the member interface
 840  */
 841 static void
 842 bridge_mutecaps(struct bridge_softc *sc)
 843 {
 844         struct bridge_iflist *bif;
 845         int enabled, mask;
 846
 847         /* Initial bitmask of capabilities to test */
 848         mask = BRIDGE_IFCAPS_MASK;
 849
 850         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 851                 /* Every member must support it or its disabled */
 852                 mask &= bif->bif_savedcaps;
 853         }
 854
 855         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 856                 enabled = bif->bif_ifp->if_capenable;
 857                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 858                 /* strip off mask bits and enable them again if allowed */
 859                 enabled &= ~BRIDGE_IFCAPS_MASK;
 860                 enabled |= mask;
 861                 bridge_set_ifcap(sc, bif, enabled);
 862         }
 863
 864 }
 865
 866 static void
 867 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 868 {
 869         struct ifnet *ifp = bif->bif_ifp;
 870         struct ifreq ifr;
 871         int error;
 872
 873         bzero(&ifr, sizeof(ifr));
 874         ifr.ifr_reqcap = set;
 875
 876         if (ifp->if_capenable != set) {
 877                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 878                 if (error)
 879                         if_printf(sc->sc_ifp,
 880                             "error setting interface capabilities on %s\n",
 881                             ifp->if_xname);
 882         }
 883 }
 884
 885 /*
 886  * bridge_lookup_member:
 887  *
 888  *      Lookup a bridge member interface.
 889  */
 890 static struct bridge_iflist *
 891 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 892 {
 893         struct bridge_iflist *bif;
 894         struct ifnet *ifp;
 895
 896         BRIDGE_LOCK_ASSERT(sc);
 897
 898         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 899                 ifp = bif->bif_ifp;
 900                 if (strcmp(ifp->if_xname, name) == 0)
 901                         return (bif);
 902         }
 903
 904         return (NULL);
 905 }
 906
 907 /*
 908  * bridge_lookup_member_if:
 909  *
 910  *      Lookup a bridge member interface by ifnet*.
 911  */
 912 static struct bridge_iflist *
 913 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 914 {
 915         struct bridge_iflist *bif;
 916
 917         BRIDGE_LOCK_ASSERT(sc);
 918
 919         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 920                 if (bif->bif_ifp == member_ifp)
 921                         return (bif);
 922         }
 923
 924         return (NULL);
 925 }
 926
 927 /*
 928  * bridge_delete_member:
 929  *
 930  *      Delete the specified member interface.
 931  */
 932 static void
 933 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 934     int gone)
 935 {
 936         struct ifnet *ifs = bif->bif_ifp;
 937         struct ifnet *fif = NULL;
 938
 939         BRIDGE_LOCK_ASSERT(sc);
 940
 941         if (bif->bif_flags & IFBIF_STP)
 942                 bstp_disable(&bif->bif_stp);
 943
 944         ifs->if_bridge = NULL;
 945         BRIDGE_XLOCK(sc);
 946         LIST_REMOVE(bif, bif_next);
 947         BRIDGE_XDROP(sc);
 948
 949         /*
 950          * If removing the interface that gave the bridge its mac address, set
 951          * the mac address of the bridge to the address of the next member, or
 952          * to its default address if no members are left.
 953          */
 954         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
 955                 if (LIST_EMPTY(&sc->sc_iflist)) {
 956                         bcopy(sc->sc_defaddr,
 957                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 958                         sc->sc_ifaddr = NULL;
 959                 } else {
 960                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
 961                         bcopy(IF_LLADDR(fif),
 962                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 963                         sc->sc_ifaddr = fif;
 964                 }
 965                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
 966         }
 967
 968         bridge_linkcheck(sc);
 969         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 970         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 971         KASSERT(bif->bif_addrcnt == 0,
 972             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 973
 974         BRIDGE_UNLOCK(sc);
 975         if (!gone) {
 976                 switch (ifs->if_type) {
 977                 case IFT_ETHER:
 978                 case IFT_L2VLAN:
 979                         /*
 980                          * Take the interface out of promiscuous mode.
 981                          */
 982                         (void) ifpromisc(ifs, 0);
 983                         break;
 984
 985                 case IFT_GIF:
 986                         break;
 987
 988                 default:
 989 #ifdef DIAGNOSTIC
 990                         panic("bridge_delete_member: impossible");
 991 #endif
 992                         break;
 993                 }
 994                 /* reneable any interface capabilities */
 995                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
 996         }
 997         bstp_destroy(&bif->bif_stp);    /* prepare to free */
 998         BRIDGE_LOCK(sc);
 999         free(bif, M_DEVBUF);
1000 }
1001
1002 /*
1003  * bridge_delete_span:
1004  *
1005  *      Delete the specified span interface.
1006  */
1007 static void
1008 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1009 {
1010         BRIDGE_LOCK_ASSERT(sc);
1011
1012         KASSERT(bif->bif_ifp->if_bridge == NULL,
1013             ("%s: not a span interface", __func__));
1014
1015         LIST_REMOVE(bif, bif_next);
1016         free(bif, M_DEVBUF);
1017 }
1018
1019 static int
1020 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1021 {
1022         struct ifbreq *req = arg;
1023         struct bridge_iflist *bif = NULL;
1024         struct ifnet *ifs;
1025         int error = 0;
1026
1027         ifs = ifunit(req->ifbr_ifsname);
1028         if (ifs == NULL)
1029                 return (ENOENT);
1030         if (ifs->if_ioctl == NULL)      /* must be supported */
1031                 return (EINVAL);
1032
1033         /* If it's in the span list, it can't be a member. */
1034         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1035                 if (ifs == bif->bif_ifp)
1036                         return (EBUSY);
1037
1038         if (ifs->if_bridge == sc)
1039                 return (EEXIST);
1040
1041         if (ifs->if_bridge != NULL)
1042                 return (EBUSY);
1043
1044         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1045         if (bif == NULL)
1046                 return (ENOMEM);
1047
1048         bif->bif_ifp = ifs;
1049         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1050         bif->bif_savedcaps = ifs->if_capenable;
1051
1052         switch (ifs->if_type) {
1053         case IFT_ETHER:
1054         case IFT_L2VLAN:
1055         case IFT_GIF:
1056                 /* permitted interface types */
1057                 break;
1058         default:
1059                 error = EINVAL;
1060                 goto out;
1061         }
1062
1063         /* Allow the first Ethernet member to define the MTU */
1064         if (LIST_EMPTY(&sc->sc_iflist))
1065                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1066         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1067                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1068                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1069                 error = EINVAL;
1070                 goto out;
1071         }
1072
1073         /*
1074          * Assign the interface's MAC address to the bridge if it's the first
1075          * member and the MAC address of the bridge has not been changed from
1076          * the default randomly generated one.
1077          */
1078         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1079             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1080                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1081                 sc->sc_ifaddr = ifs;
1082                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1083         }
1084
1085         ifs->if_bridge = sc;
1086         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1087         /*
1088          * XXX: XLOCK HERE!?!
1089          *
1090          * NOTE: insert_***HEAD*** should be safe for the traversals.
1091          */
1092         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1093
1094         /* Set interface capabilities to the intersection set of all members */
1095         bridge_mutecaps(sc);
1096         bridge_linkcheck(sc);
1097
1098         /* Place the interface into promiscuous mode */
1099         switch (ifs->if_type) {
1100                 case IFT_ETHER:
1101                 case IFT_L2VLAN:
1102                         BRIDGE_UNLOCK(sc);
1103                         error = ifpromisc(ifs, 1);
1104                         BRIDGE_LOCK(sc);
1105                         break;
1106         }
1107         if (error)
1108                 bridge_delete_member(sc, bif, 0);
1109 out:
1110         if (error) {
1111                 if (bif != NULL)
1112                         free(bif, M_DEVBUF);
1113         }
1114         return (error);
1115 }
1116
1117 static int
1118 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1119 {
1120         struct ifbreq *req = arg;
1121         struct bridge_iflist *bif;
1122
1123         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1124         if (bif == NULL)
1125                 return (ENOENT);
1126
1127         bridge_delete_member(sc, bif, 0);
1128
1129         return (0);
1130 }
1131
1132 static int
1133 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1134 {
1135         struct ifbreq *req = arg;
1136         struct bridge_iflist *bif;
1137         struct bstp_port *bp;
1138
1139         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1140         if (bif == NULL)
1141                 return (ENOENT);
1142
1143         bp = &bif->bif_stp;
1144         req->ifbr_ifsflags = bif->bif_flags;
1145         req->ifbr_state = bp->bp_state;
1146         req->ifbr_priority = bp->bp_priority;
1147         req->ifbr_path_cost = bp->bp_path_cost;
1148         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1149         req->ifbr_proto = bp->bp_protover;
1150         req->ifbr_role = bp->bp_role;
1151         req->ifbr_stpflags = bp->bp_flags;
1152         req->ifbr_addrcnt = bif->bif_addrcnt;
1153         req->ifbr_addrmax = bif->bif_addrmax;
1154         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1155
1156         /* Copy STP state options as flags */
1157         if (bp->bp_operedge)
1158                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1159         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1160                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1161         if (bp->bp_ptp_link)
1162                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1163         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1164                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1165         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1166                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1167         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1168                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1169         return (0);
1170 }
1171
1172 static int
1173 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1174 {
1175         struct ifbreq *req = arg;
1176         struct bridge_iflist *bif;
1177         struct bstp_port *bp;
1178         int error;
1179
1180         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1181         if (bif == NULL)
1182                 return (ENOENT);
1183         bp = &bif->bif_stp;
1184
1185         if (req->ifbr_ifsflags & IFBIF_SPAN)
1186                 /* SPAN is readonly */
1187                 return (EINVAL);
1188
1189         if (req->ifbr_ifsflags & IFBIF_STP) {
1190                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1191                         error = bstp_enable(&bif->bif_stp);
1192                         if (error)
1193                                 return (error);
1194                 }
1195         } else {
1196                 if ((bif->bif_flags & IFBIF_STP) != 0)
1197                         bstp_disable(&bif->bif_stp);
1198         }
1199
1200         /* Pass on STP flags */
1201         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1202         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1203         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1204         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1205
1206         /* Save the bits relating to the bridge */
1207         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1208
1209         return (0);
1210 }
1211
1212 static int
1213 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1214 {
1215         struct ifbrparam *param = arg;
1216
1217         sc->sc_brtmax = param->ifbrp_csize;
1218         bridge_rttrim(sc);
1219
1220         return (0);
1221 }
1222
1223 static int
1224 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1225 {
1226         struct ifbrparam *param = arg;
1227
1228         param->ifbrp_csize = sc->sc_brtmax;
1229
1230         return (0);
1231 }
1232
1233 static int
1234 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1235 {
1236         struct ifbifconf *bifc = arg;
1237         struct bridge_iflist *bif;
1238         struct ifbreq breq;
1239         char *buf, *outbuf;
1240         int count, buflen, len, error = 0;
1241
1242         count = 0;
1243         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1244                 count++;
1245         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1246                 count++;
1247
1248         buflen = sizeof(breq) * count;
1249         if (bifc->ifbic_len == 0) {
1250                 bifc->ifbic_len = buflen;
1251                 return (0);
1252         }
1253         BRIDGE_UNLOCK(sc);
1254         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1255         BRIDGE_LOCK(sc);
1256
1257         count = 0;
1258         buf = outbuf;
1259         len = min(bifc->ifbic_len, buflen);
1260         bzero(&breq, sizeof(breq));
1261         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1262                 if (len < sizeof(breq))
1263                         break;
1264
1265                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1266                     sizeof(breq.ifbr_ifsname));
1267                 /* Fill in the ifbreq structure */
1268                 error = bridge_ioctl_gifflags(sc, &breq);
1269                 if (error)
1270                         break;
1271                 memcpy(buf, &breq, sizeof(breq));
1272                 count++;
1273                 buf += sizeof(breq);
1274                 len -= sizeof(breq);
1275         }
1276         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1277                 if (len < sizeof(breq))
1278                         break;
1279
1280                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1281                     sizeof(breq.ifbr_ifsname));
1282                 breq.ifbr_ifsflags = bif->bif_flags;
1283                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1284                 memcpy(buf, &breq, sizeof(breq));
1285                 count++;
1286                 buf += sizeof(breq);
1287                 len -= sizeof(breq);
1288         }
1289
1290         BRIDGE_UNLOCK(sc);
1291         bifc->ifbic_len = sizeof(breq) * count;
1292         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1293         BRIDGE_LOCK(sc);
1294         free(outbuf, M_TEMP);
1295         return (error);
1296 }
1297
1298 static int
1299 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1300 {
1301         struct ifbaconf *bac = arg;
1302         struct bridge_rtnode *brt;
1303         struct ifbareq bareq;
1304         char *buf, *outbuf;
1305         int count, buflen, len, error = 0;
1306
1307         if (bac->ifbac_len == 0)
1308                 return (0);
1309
1310         count = 0;
1311         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1312                 count++;
1313         buflen = sizeof(bareq) * count;
1314
1315         BRIDGE_UNLOCK(sc);
1316         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1317         BRIDGE_LOCK(sc);
1318
1319         count = 0;
1320         buf = outbuf;
1321         len = min(bac->ifbac_len, buflen);
1322         bzero(&bareq, sizeof(bareq));
1323         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1324                 if (len < sizeof(bareq))
1325                         goto out;
1326                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1327                     sizeof(bareq.ifba_ifsname));
1328                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1329                 bareq.ifba_vlan = brt->brt_vlan;
1330                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1331                                 time_uptime < brt->brt_expire)
1332                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1333                 else
1334                         bareq.ifba_expire = 0;
1335                 bareq.ifba_flags = brt->brt_flags;
1336
1337                 memcpy(buf, &bareq, sizeof(bareq));
1338                 count++;
1339                 buf += sizeof(bareq);
1340                 len -= sizeof(bareq);
1341         }
1342 out:
1343         BRIDGE_UNLOCK(sc);
1344         bac->ifbac_len = sizeof(bareq) * count;
1345         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1346         BRIDGE_LOCK(sc);
1347         free(outbuf, M_TEMP);
1348         return (error);
1349 }
1350
1351 static int
1352 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1353 {
1354         struct ifbareq *req = arg;
1355         struct bridge_iflist *bif;
1356         int error;
1357
1358         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1359         if (bif == NULL)
1360                 return (ENOENT);
1361
1362         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1363             req->ifba_flags);
1364
1365         return (error);
1366 }
1367
1368 static int
1369 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1370 {
1371         struct ifbrparam *param = arg;
1372
1373         sc->sc_brttimeout = param->ifbrp_ctime;
1374         return (0);
1375 }
1376
1377 static int
1378 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1379 {
1380         struct ifbrparam *param = arg;
1381
1382         param->ifbrp_ctime = sc->sc_brttimeout;
1383         return (0);
1384 }
1385
1386 static int
1387 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1388 {
1389         struct ifbareq *req = arg;
1390
1391         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1392 }
1393
1394 static int
1395 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1396 {
1397         struct ifbreq *req = arg;
1398
1399         bridge_rtflush(sc, req->ifbr_ifsflags);
1400         return (0);
1401 }
1402
1403 static int
1404 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1405 {
1406         struct ifbrparam *param = arg;
1407         struct bstp_state *bs = &sc->sc_stp;
1408
1409         param->ifbrp_prio = bs->bs_bridge_priority;
1410         return (0);
1411 }
1412
1413 static int
1414 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1415 {
1416         struct ifbrparam *param = arg;
1417
1418         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1419 }
1420
1421 static int
1422 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1423 {
1424         struct ifbrparam *param = arg;
1425         struct bstp_state *bs = &sc->sc_stp;
1426
1427         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1428         return (0);
1429 }
1430
1431 static int
1432 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1433 {
1434         struct ifbrparam *param = arg;
1435
1436         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1437 }
1438
1439 static int
1440 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1441 {
1442         struct ifbrparam *param = arg;
1443         struct bstp_state *bs = &sc->sc_stp;
1444
1445         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1446         return (0);
1447 }
1448
1449 static int
1450 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1451 {
1452         struct ifbrparam *param = arg;
1453
1454         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1455 }
1456
1457 static int
1458 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1459 {
1460         struct ifbrparam *param = arg;
1461         struct bstp_state *bs = &sc->sc_stp;
1462
1463         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1464         return (0);
1465 }
1466
1467 static int
1468 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1469 {
1470         struct ifbrparam *param = arg;
1471
1472         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1473 }
1474
1475 static int
1476 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1477 {
1478         struct ifbreq *req = arg;
1479         struct bridge_iflist *bif;
1480
1481         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1482         if (bif == NULL)
1483                 return (ENOENT);
1484
1485         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1486 }
1487
1488 static int
1489 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1490 {
1491         struct ifbreq *req = arg;
1492         struct bridge_iflist *bif;
1493
1494         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1495         if (bif == NULL)
1496                 return (ENOENT);
1497
1498         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1499 }
1500
1501 static int
1502 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1503 {
1504         struct ifbreq *req = arg;
1505         struct bridge_iflist *bif;
1506
1507         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1508         if (bif == NULL)
1509                 return (ENOENT);
1510
1511         bif->bif_addrmax = req->ifbr_addrmax;
1512         return (0);
1513 }
1514
1515 static int
1516 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1517 {
1518         struct ifbreq *req = arg;
1519         struct bridge_iflist *bif = NULL;
1520         struct ifnet *ifs;
1521
1522         ifs = ifunit(req->ifbr_ifsname);
1523         if (ifs == NULL)
1524                 return (ENOENT);
1525
1526         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1527                 if (ifs == bif->bif_ifp)
1528                         return (EBUSY);
1529
1530         if (ifs->if_bridge != NULL)
1531                 return (EBUSY);
1532
1533         switch (ifs->if_type) {
1534                 case IFT_ETHER:
1535                 case IFT_GIF:
1536                 case IFT_L2VLAN:
1537                         break;
1538                 default:
1539                         return (EINVAL);
1540         }
1541
1542         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1543         if (bif == NULL)
1544                 return (ENOMEM);
1545
1546         bif->bif_ifp = ifs;
1547         bif->bif_flags = IFBIF_SPAN;
1548
1549         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1550
1551         return (0);
1552 }
1553
1554 static int
1555 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1556 {
1557         struct ifbreq *req = arg;
1558         struct bridge_iflist *bif;
1559         struct ifnet *ifs;
1560
1561         ifs = ifunit(req->ifbr_ifsname);
1562         if (ifs == NULL)
1563                 return (ENOENT);
1564
1565         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1566                 if (ifs == bif->bif_ifp)
1567                         break;
1568
1569         if (bif == NULL)
1570                 return (ENOENT);
1571
1572         bridge_delete_span(sc, bif);
1573
1574         return (0);
1575 }
1576
1577 static int
1578 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1579 {
1580         struct ifbropreq *req = arg;
1581         struct bstp_state *bs = &sc->sc_stp;
1582         struct bstp_port *root_port;
1583
1584         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1585         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1586         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1587
1588         root_port = bs->bs_root_port;
1589         if (root_port == NULL)
1590                 req->ifbop_root_port = 0;
1591         else
1592                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1593
1594         req->ifbop_holdcount = bs->bs_txholdcount;
1595         req->ifbop_priority = bs->bs_bridge_priority;
1596         req->ifbop_protocol = bs->bs_protover;
1597         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1598         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1599         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1600         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1601         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1602         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1603
1604         return (0);
1605 }
1606
1607 static int
1608 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1609 {
1610         struct ifbrparam *param = arg;
1611
1612         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1613         return (0);
1614 }
1615
1616 static int
1617 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1618 {
1619         struct ifbpstpconf *bifstp = arg;
1620         struct bridge_iflist *bif;
1621         struct bstp_port *bp;
1622         struct ifbpstpreq bpreq;
1623         char *buf, *outbuf;
1624         int count, buflen, len, error = 0;
1625
1626         count = 0;
1627         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1628                 if ((bif->bif_flags & IFBIF_STP) != 0)
1629                         count++;
1630         }
1631
1632         buflen = sizeof(bpreq) * count;
1633         if (bifstp->ifbpstp_len == 0) {
1634                 bifstp->ifbpstp_len = buflen;
1635                 return (0);
1636         }
1637
1638         BRIDGE_UNLOCK(sc);
1639         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1640         BRIDGE_LOCK(sc);
1641
1642         count = 0;
1643         buf = outbuf;
1644         len = min(bifstp->ifbpstp_len, buflen);
1645         bzero(&bpreq, sizeof(bpreq));
1646         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1647                 if (len < sizeof(bpreq))
1648                         break;
1649
1650                 if ((bif->bif_flags & IFBIF_STP) == 0)
1651                         continue;
1652
1653                 bp = &bif->bif_stp;
1654                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1655                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1656                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1657                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1658                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1659                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1660
1661                 memcpy(buf, &bpreq, sizeof(bpreq));
1662                 count++;
1663                 buf += sizeof(bpreq);
1664                 len -= sizeof(bpreq);
1665         }
1666
1667         BRIDGE_UNLOCK(sc);
1668         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1669         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1670         BRIDGE_LOCK(sc);
1671         free(outbuf, M_TEMP);
1672         return (error);
1673 }
1674
1675 static int
1676 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1677 {
1678         struct ifbrparam *param = arg;
1679
1680         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1681 }
1682
1683 static int
1684 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1685 {
1686         struct ifbrparam *param = arg;
1687
1688         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1689 }
1690
1691 /*
1692  * bridge_ifdetach:
1693  *
1694  *      Detach an interface from a bridge.  Called when a member
1695  *      interface is detaching.
1696  */
1697 static void
1698 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1699 {
1700         struct bridge_softc *sc = ifp->if_bridge;
1701         struct bridge_iflist *bif;
1702
1703         if (ifp->if_flags & IFF_RENAMING)
1704                 return;
1705
1706         /* Check if the interface is a bridge member */
1707         if (sc != NULL) {
1708                 BRIDGE_LOCK(sc);
1709
1710                 bif = bridge_lookup_member_if(sc, ifp);
1711                 if (bif != NULL)
1712                         bridge_delete_member(sc, bif, 1);
1713
1714                 BRIDGE_UNLOCK(sc);
1715                 return;
1716         }
1717
1718         /* Check if the interface is a span port */
1719         mtx_lock(&bridge_list_mtx);
1720         LIST_FOREACH(sc, &bridge_list, sc_list) {
1721                 BRIDGE_LOCK(sc);
1722                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1723                         if (ifp == bif->bif_ifp) {
1724                                 bridge_delete_span(sc, bif);
1725                                 break;
1726                         }
1727
1728                 BRIDGE_UNLOCK(sc);
1729         }
1730         mtx_unlock(&bridge_list_mtx);
1731 }
1732
1733 /*
1734  * bridge_init:
1735  *
1736  *      Initialize a bridge interface.
1737  */
1738 static void
1739 bridge_init(void *xsc)
1740 {
1741         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1742         struct ifnet *ifp = sc->sc_ifp;
1743
1744         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1745                 return;
1746
1747         BRIDGE_LOCK(sc);
1748         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1749             bridge_timer, sc);
1750
1751         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1752         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1753
1754         BRIDGE_UNLOCK(sc);
1755 }
1756
1757 /*
1758  * bridge_stop:
1759  *
1760  *      Stop the bridge interface.
1761  */
1762 static void
1763 bridge_stop(struct ifnet *ifp, int disable)
1764 {
1765         struct bridge_softc *sc = ifp->if_softc;
1766
1767         BRIDGE_LOCK_ASSERT(sc);
1768
1769         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1770                 return;
1771
1772         callout_stop(&sc->sc_brcallout);
1773         bstp_stop(&sc->sc_stp);
1774
1775         bridge_rtflush(sc, IFBF_FLUSHDYN);
1776
1777         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1778 }
1779
1780 /*
1781  * bridge_enqueue:
1782  *
1783  *      Enqueue a packet on a bridge member interface.
1784  *
1785  */
1786 static void
1787 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1788 {
1789         int len, err = 0;
1790         short mflags;
1791         struct mbuf *m0;
1792
1793         len = m->m_pkthdr.len;
1794         mflags = m->m_flags;
1795
1796         /* We may be sending a fragment so traverse the mbuf */
1797         for (; m; m = m0) {
1798                 m0 = m->m_nextpkt;
1799                 m->m_nextpkt = NULL;
1800
1801                 /*
1802                  * If underlying interface can not do VLAN tag insertion itself
1803                  * then attach a packet tag that holds it.
1804                  */
1805                 if ((m->m_flags & M_VLANTAG) &&
1806                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1807                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1808                         if (m == NULL) {
1809                                 if_printf(dst_ifp,
1810                                     "unable to prepend VLAN header\n");
1811                                 dst_ifp->if_oerrors++;
1812                                 continue;
1813                         }
1814                         m->m_flags &= ~M_VLANTAG;
1815                 }
1816
1817                 if (err == 0)
1818                         dst_ifp->if_transmit(dst_ifp, m);
1819         }
1820
1821         if (err == 0) {
1822                 sc->sc_ifp->if_opackets++;
1823                 sc->sc_ifp->if_obytes += len;
1824                 if (mflags & M_MCAST)
1825                         sc->sc_ifp->if_omcasts++;
1826         }
1827 }
1828
1829 /*
1830  * bridge_dummynet:
1831  *
1832  *      Receive a queued packet from dummynet and pass it on to the output
1833  *      interface.
1834  *
1835  *      The mbuf has the Ethernet header already attached.
1836  */
1837 static void
1838 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1839 {
1840         struct bridge_softc *sc;
1841
1842         sc = ifp->if_bridge;
1843
1844         /*
1845          * The packet didnt originate from a member interface. This should only
1846          * ever happen if a member interface is removed while packets are
1847          * queued for it.
1848          */
1849         if (sc == NULL) {
1850                 m_freem(m);
1851                 return;
1852         }
1853
1854         if (PFIL_HOOKED(&V_inet_pfil_hook)
1855 #ifdef INET6
1856             || PFIL_HOOKED(&V_inet6_pfil_hook)
1857 #endif
1858             ) {
1859                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1860                         return;
1861                 if (m == NULL)
1862                         return;
1863         }
1864
1865         bridge_enqueue(sc, ifp, m);
1866 }
1867
1868 /*
1869  * bridge_output:
1870  *
1871  *      Send output from a bridge member interface.  This
1872  *      performs the bridging function for locally originated
1873  *      packets.
1874  *
1875  *      The mbuf has the Ethernet header already attached.  We must
1876  *      enqueue or free the mbuf before returning.
1877  */
1878 static int
1879 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1880     struct rtentry *rt)
1881 {
1882         struct ether_header *eh;
1883         struct ifnet *dst_if;
1884         struct bridge_softc *sc;
1885         uint16_t vlan;
1886
1887         if (m->m_len < ETHER_HDR_LEN) {
1888                 m = m_pullup(m, ETHER_HDR_LEN);
1889                 if (m == NULL)
1890                         return (0);
1891         }
1892
1893         eh = mtod(m, struct ether_header *);
1894         sc = ifp->if_bridge;
1895         vlan = VLANTAGOF(m);
1896
1897         BRIDGE_LOCK(sc);
1898
1899         /*
1900          * If bridge is down, but the original output interface is up,
1901          * go ahead and send out that interface.  Otherwise, the packet
1902          * is dropped below.
1903          */
1904         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1905                 dst_if = ifp;
1906                 goto sendunicast;
1907         }
1908
1909         /*
1910          * If the packet is a multicast, or we don't know a better way to
1911          * get there, send to all interfaces.
1912          */
1913         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1914                 dst_if = NULL;
1915         else
1916                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1917         if (dst_if == NULL) {
1918                 struct bridge_iflist *bif;
1919                 struct mbuf *mc;
1920                 int error = 0, used = 0;
1921
1922                 bridge_span(sc, m);
1923
1924                 BRIDGE_LOCK2REF(sc, error);
1925                 if (error) {
1926                         m_freem(m);
1927                         return (0);
1928                 }
1929
1930                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1931                         dst_if = bif->bif_ifp;
1932
1933                         if (dst_if->if_type == IFT_GIF)
1934                                 continue;
1935                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1936                                 continue;
1937
1938                         /*
1939                          * If this is not the original output interface,
1940                          * and the interface is participating in spanning
1941                          * tree, make sure the port is in a state that
1942                          * allows forwarding.
1943                          */
1944                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1945                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1946                                 continue;
1947
1948                         if (LIST_NEXT(bif, bif_next) == NULL) {
1949                                 used = 1;
1950                                 mc = m;
1951                         } else {
1952                                 mc = m_copypacket(m, M_DONTWAIT);
1953                                 if (mc == NULL) {
1954                                         sc->sc_ifp->if_oerrors++;
1955                                         continue;
1956                                 }
1957                         }
1958
1959                         bridge_enqueue(sc, dst_if, mc);
1960                 }
1961                 if (used == 0)
1962                         m_freem(m);
1963                 BRIDGE_UNREF(sc);
1964                 return (0);
1965         }
1966
1967 sendunicast:
1968         /*
1969          * XXX Spanning tree consideration here?
1970          */
1971
1972         bridge_span(sc, m);
1973         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1974                 m_freem(m);
1975                 BRIDGE_UNLOCK(sc);
1976                 return (0);
1977         }
1978
1979         BRIDGE_UNLOCK(sc);
1980         bridge_enqueue(sc, dst_if, m);
1981         return (0);
1982 }
1983
1984 /*
1985  * bridge_start:
1986  *
1987  *      Start output on a bridge.
1988  *
1989  */
1990 static void
1991 bridge_start(struct ifnet *ifp)
1992 {
1993         struct bridge_softc *sc;
1994         struct mbuf *m;
1995         struct ether_header *eh;
1996         struct ifnet *dst_if;
1997
1998         sc = ifp->if_softc;
1999
2000         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
2001         for (;;) {
2002                 IFQ_DEQUEUE(&ifp->if_snd, m);
2003                 if (m == 0)
2004                         break;
2005                 ETHER_BPF_MTAP(ifp, m);
2006
2007                 eh = mtod(m, struct ether_header *);
2008                 dst_if = NULL;
2009
2010                 BRIDGE_LOCK(sc);
2011                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2012                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
2013                 }
2014
2015                 if (dst_if == NULL)
2016                         bridge_broadcast(sc, ifp, m, 0);
2017                 else {
2018                         BRIDGE_UNLOCK(sc);
2019                         bridge_enqueue(sc, dst_if, m);
2020                 }
2021         }
2022         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2023 }
2024
2025 /*
2026  * bridge_forward:
2027  *
2028  *      The forwarding function of the bridge.
2029  *
2030  *      NOTE: Releases the lock on return.
2031  */
2032 static void
2033 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2034     struct mbuf *m)
2035 {
2036         struct bridge_iflist *dbif;
2037         struct ifnet *src_if, *dst_if, *ifp;
2038         struct ether_header *eh;
2039         uint16_t vlan;
2040         uint8_t *dst;
2041         int error;
2042
2043         src_if = m->m_pkthdr.rcvif;
2044         ifp = sc->sc_ifp;
2045
2046         ifp->if_ipackets++;
2047         ifp->if_ibytes += m->m_pkthdr.len;
2048         vlan = VLANTAGOF(m);
2049
2050         if ((sbif->bif_flags & IFBIF_STP) &&
2051             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2052                 goto drop;
2053
2054         eh = mtod(m, struct ether_header *);
2055         dst = eh->ether_dhost;
2056
2057         /* If the interface is learning, record the address. */
2058         if (sbif->bif_flags & IFBIF_LEARNING) {
2059                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2060                     sbif, 0, IFBAF_DYNAMIC);
2061                 /*
2062                  * If the interface has addresses limits then deny any source
2063                  * that is not in the cache.
2064                  */
2065                 if (error && sbif->bif_addrmax)
2066                         goto drop;
2067         }
2068
2069         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2070             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2071                 goto drop;
2072
2073         /*
2074          * At this point, the port either doesn't participate
2075          * in spanning tree or it is in the forwarding state.
2076          */
2077
2078         /*
2079          * If the packet is unicast, destined for someone on
2080          * "this" side of the bridge, drop it.
2081          */
2082         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2083                 dst_if = bridge_rtlookup(sc, dst, vlan);
2084                 if (src_if == dst_if)
2085                         goto drop;
2086         } else {
2087                 /*
2088                  * Check if its a reserved multicast address, any address
2089                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2090                  * bridge.
2091                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2092                  */
2093                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2094                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2095                     dst[4] == 0x00 && dst[5] <= 0x0f)
2096                         goto drop;
2097
2098                 /* ...forward it to all interfaces. */
2099                 ifp->if_imcasts++;
2100                 dst_if = NULL;
2101         }
2102
2103         /*
2104          * If we have a destination interface which is a member of our bridge,
2105          * OR this is a unicast packet, push it through the bpf(4) machinery.
2106          * For broadcast or multicast packets, don't bother because it will
2107          * be reinjected into ether_input. We do this before we pass the packets
2108          * through the pfil(9) framework, as it is possible that pfil(9) will
2109          * drop the packet, or possibly modify it, making it difficult to debug
2110          * firewall issues on the bridge.
2111          */
2112         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2113                 ETHER_BPF_MTAP(ifp, m);
2114
2115         /* run the packet filter */
2116         if (PFIL_HOOKED(&V_inet_pfil_hook)
2117 #ifdef INET6
2118             || PFIL_HOOKED(&V_inet6_pfil_hook)
2119 #endif
2120             ) {
2121                 BRIDGE_UNLOCK(sc);
2122                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2123                         return;
2124                 if (m == NULL)
2125                         return;
2126                 BRIDGE_LOCK(sc);
2127         }
2128
2129         if (dst_if == NULL) {
2130                 bridge_broadcast(sc, src_if, m, 1);
2131                 return;
2132         }
2133
2134         /*
2135          * At this point, we're dealing with a unicast frame
2136          * going to a different interface.
2137          */
2138         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2139                 goto drop;
2140
2141         dbif = bridge_lookup_member_if(sc, dst_if);
2142         if (dbif == NULL)
2143                 /* Not a member of the bridge (anymore?) */
2144                 goto drop;
2145
2146         /* Private segments can not talk to each other */
2147         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2148                 goto drop;
2149
2150         if ((dbif->bif_flags & IFBIF_STP) &&
2151             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2152                 goto drop;
2153
2154         BRIDGE_UNLOCK(sc);
2155
2156         if (PFIL_HOOKED(&V_inet_pfil_hook)
2157 #ifdef INET6
2158             || PFIL_HOOKED(&V_inet6_pfil_hook)
2159 #endif
2160             ) {
2161                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2162                         return;
2163                 if (m == NULL)
2164                         return;
2165         }
2166
2167         bridge_enqueue(sc, dst_if, m);
2168         return;
2169
2170 drop:
2171         BRIDGE_UNLOCK(sc);
2172         m_freem(m);
2173 }
2174
2175 /*
2176  * bridge_input:
2177  *
2178  *      Receive input from a member interface.  Queue the packet for
2179  *      bridging if it is not for us.
2180  */
2181 static struct mbuf *
2182 bridge_input(struct ifnet *ifp, struct mbuf *m)
2183 {
2184         struct bridge_softc *sc = ifp->if_bridge;
2185         struct bridge_iflist *bif, *bif2;
2186         struct ifnet *bifp;
2187         struct ether_header *eh;
2188         struct mbuf *mc, *mc2;
2189         uint16_t vlan;
2190         int error;
2191
2192         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2193                 return (m);
2194
2195         bifp = sc->sc_ifp;
2196         vlan = VLANTAGOF(m);
2197
2198         /*
2199          * Implement support for bridge monitoring. If this flag has been
2200          * set on this interface, discard the packet once we push it through
2201          * the bpf(4) machinery, but before we do, increment the byte and
2202          * packet counters associated with this interface.
2203          */
2204         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2205                 m->m_pkthdr.rcvif  = bifp;
2206                 ETHER_BPF_MTAP(bifp, m);
2207                 bifp->if_ipackets++;
2208                 bifp->if_ibytes += m->m_pkthdr.len;
2209                 m_freem(m);
2210                 return (NULL);
2211         }
2212         BRIDGE_LOCK(sc);
2213         bif = bridge_lookup_member_if(sc, ifp);
2214         if (bif == NULL) {
2215                 BRIDGE_UNLOCK(sc);
2216                 return (m);
2217         }
2218
2219         eh = mtod(m, struct ether_header *);
2220
2221         bridge_span(sc, m);
2222
2223         if (m->m_flags & (M_BCAST|M_MCAST)) {
2224                 /* Tap off 802.1D packets; they do not get forwarded. */
2225                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2226                     ETHER_ADDR_LEN) == 0) {
2227                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2228                         BRIDGE_UNLOCK(sc);
2229                         return (NULL);
2230                 }
2231
2232                 if ((bif->bif_flags & IFBIF_STP) &&
2233                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2234                         BRIDGE_UNLOCK(sc);
2235                         return (m);
2236                 }
2237
2238                 /*
2239                  * Make a deep copy of the packet and enqueue the copy
2240                  * for bridge processing; return the original packet for
2241                  * local processing.
2242                  */
2243                 mc = m_dup(m, M_DONTWAIT);
2244                 if (mc == NULL) {
2245                         BRIDGE_UNLOCK(sc);
2246                         return (m);
2247                 }
2248
2249                 /* Perform the bridge forwarding function with the copy. */
2250                 bridge_forward(sc, bif, mc);
2251
2252                 /*
2253                  * Reinject the mbuf as arriving on the bridge so we have a
2254                  * chance at claiming multicast packets. We can not loop back
2255                  * here from ether_input as a bridge is never a member of a
2256                  * bridge.
2257                  */
2258                 KASSERT(bifp->if_bridge == NULL,
2259                     ("loop created in bridge_input"));
2260                 mc2 = m_dup(m, M_DONTWAIT);
2261                 if (mc2 != NULL) {
2262                         /* Keep the layer3 header aligned */
2263                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2264                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2265                 }
2266                 if (mc2 != NULL) {
2267                         mc2->m_pkthdr.rcvif = bifp;
2268                         (*bifp->if_input)(bifp, mc2);
2269                 }
2270
2271                 /* Return the original packet for local processing. */
2272                 return (m);
2273         }
2274
2275         if ((bif->bif_flags & IFBIF_STP) &&
2276             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2277                 BRIDGE_UNLOCK(sc);
2278                 return (m);
2279         }
2280
2281 #if (defined(INET) || defined(INET6))
2282 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2283         || ((iface)->if_carp \
2284             && (*carp_forus_p)((iface), eh->ether_dhost))
2285 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2286         || ((iface)->if_carp \
2287             && (*carp_forus_p)((iface), eh->ether_shost))
2288 #else
2289 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2290 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2291 #endif
2292
2293 #ifdef INET6
2294 #   define OR_PFIL_HOOKED_INET6 \
2295         || PFIL_HOOKED(&V_inet6_pfil_hook)
2296 #else
2297 #   define OR_PFIL_HOOKED_INET6
2298 #endif
2299
2300 #define GRAB_OUR_PACKETS(iface) \
2301         if ((iface)->if_type == IFT_GIF) \
2302                 continue; \
2303         /* It is destined for us. */ \
2304         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2305             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2306             ) {                                                         \
2307                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2308                         ETHER_BPF_MTAP(iface, m);                       \
2309                         iface->if_ipackets++;                           \
2310                         /* Filter on the physical interface. */         \
2311                         if (pfil_local_phys &&                          \
2312                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2313                              OR_PFIL_HOOKED_INET6)) {                   \
2314                                 if (bridge_pfil(&m, NULL, ifp,          \
2315                                     PFIL_IN) != 0 || m == NULL) {       \
2316                                         BRIDGE_UNLOCK(sc);              \
2317                                         return (NULL);                  \
2318                                 }                                       \
2319                         }                                               \
2320                 }                                                       \
2321                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2322                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2323                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2324                         if (error && bif->bif_addrmax) {                \
2325                                 BRIDGE_UNLOCK(sc);                      \
2326                                 m_freem(m);                             \
2327                                 return (NULL);                          \
2328                         }                                               \
2329                 }                                                       \
2330                 m->m_pkthdr.rcvif = iface;                              \
2331                 BRIDGE_UNLOCK(sc);                                      \
2332                 return (m);                                             \
2333         }                                                               \
2334                                                                         \
2335         /* We just received a packet that we sent out. */               \
2336         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2337             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2338             ) {                                                         \
2339                 BRIDGE_UNLOCK(sc);                                      \
2340                 m_freem(m);                                             \
2341                 return (NULL);                                          \
2342         }
2343
2344         /*
2345          * Unicast.  Make sure it's not for the bridge.
2346          */
2347         do { GRAB_OUR_PACKETS(bifp) } while (0);
2348
2349         /*
2350          * Give a chance for ifp at first priority. This will help when the
2351          * packet comes through the interface like VLAN's with the same MACs
2352          * on several interfaces from the same bridge. This also will save
2353          * some CPU cycles in case the destination interface and the input
2354          * interface (eq ifp) are the same.
2355          */
2356         do { GRAB_OUR_PACKETS(ifp) } while (0);
2357
2358         /* Now check the all bridge members. */
2359         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2360                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2361         }
2362
2363 #undef OR_CARP_CHECK_WE_ARE_DST
2364 #undef OR_CARP_CHECK_WE_ARE_SRC
2365 #undef OR_PFIL_HOOKED_INET6
2366 #undef GRAB_OUR_PACKETS
2367
2368         /* Perform the bridge forwarding function. */
2369         bridge_forward(sc, bif, m);
2370
2371         return (NULL);
2372 }
2373
2374 /*
2375  * bridge_broadcast:
2376  *
2377  *      Send a frame to all interfaces that are members of
2378  *      the bridge, except for the one on which the packet
2379  *      arrived.
2380  *
2381  *      NOTE: Releases the lock on return.
2382  */
2383 static void
2384 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2385     struct mbuf *m, int runfilt)
2386 {
2387         struct bridge_iflist *dbif, *sbif;
2388         struct mbuf *mc;
2389         struct ifnet *dst_if;
2390         int error = 0, used = 0, i;
2391
2392         sbif = bridge_lookup_member_if(sc, src_if);
2393
2394         BRIDGE_LOCK2REF(sc, error);
2395         if (error) {
2396                 m_freem(m);
2397                 return;
2398         }
2399
2400         /* Filter on the bridge interface before broadcasting */
2401         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2402 #ifdef INET6
2403             || PFIL_HOOKED(&V_inet6_pfil_hook)
2404 #endif
2405             )) {
2406                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2407                         goto out;
2408                 if (m == NULL)
2409                         goto out;
2410         }
2411
2412         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2413                 dst_if = dbif->bif_ifp;
2414                 if (dst_if == src_if)
2415                         continue;
2416
2417                 /* Private segments can not talk to each other */
2418                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2419                         continue;
2420
2421                 if ((dbif->bif_flags & IFBIF_STP) &&
2422                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2423                         continue;
2424
2425                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2426                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2427                         continue;
2428
2429                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2430                         continue;
2431
2432                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2433                         mc = m;
2434                         used = 1;
2435                 } else {
2436                         mc = m_dup(m, M_DONTWAIT);
2437                         if (mc == NULL) {
2438                                 sc->sc_ifp->if_oerrors++;
2439                                 continue;
2440                         }
2441                 }
2442
2443                 /*
2444                  * Filter on the output interface. Pass a NULL bridge interface
2445                  * pointer so we do not redundantly filter on the bridge for
2446                  * each interface we broadcast on.
2447                  */
2448                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2449 #ifdef INET6
2450                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2451 #endif
2452                     )) {
2453                         if (used == 0) {
2454                                 /* Keep the layer3 header aligned */
2455                                 i = min(mc->m_pkthdr.len, max_protohdr);
2456                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2457                                 if (mc == NULL) {
2458                                         sc->sc_ifp->if_oerrors++;
2459                                         continue;
2460                                 }
2461                         }
2462                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2463                                 continue;
2464                         if (mc == NULL)
2465                                 continue;
2466                 }
2467
2468                 bridge_enqueue(sc, dst_if, mc);
2469         }
2470         if (used == 0)
2471                 m_freem(m);
2472
2473 out:
2474         BRIDGE_UNREF(sc);
2475 }
2476
2477 /*
2478  * bridge_span:
2479  *
2480  *      Duplicate a packet out one or more interfaces that are in span mode,
2481  *      the original mbuf is unmodified.
2482  */
2483 static void
2484 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2485 {
2486         struct bridge_iflist *bif;
2487         struct ifnet *dst_if;
2488         struct mbuf *mc;
2489
2490         if (LIST_EMPTY(&sc->sc_spanlist))
2491                 return;
2492
2493         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2494                 dst_if = bif->bif_ifp;
2495
2496                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2497                         continue;
2498
2499                 mc = m_copypacket(m, M_DONTWAIT);
2500                 if (mc == NULL) {
2501                         sc->sc_ifp->if_oerrors++;
2502                         continue;
2503                 }
2504
2505                 bridge_enqueue(sc, dst_if, mc);
2506         }
2507 }
2508
2509 /*
2510  * bridge_rtupdate:
2511  *
2512  *      Add a bridge routing entry.
2513  */
2514 static int
2515 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2516     struct bridge_iflist *bif, int setflags, uint8_t flags)
2517 {
2518         struct bridge_rtnode *brt;
2519         int error;
2520
2521         BRIDGE_LOCK_ASSERT(sc);
2522
2523         /* Check the source address is valid and not multicast. */
2524         if (ETHER_IS_MULTICAST(dst) ||
2525             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2526              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2527                 return (EINVAL);
2528
2529         /* 802.1p frames map to vlan 1 */
2530         if (vlan == 0)
2531                 vlan = 1;
2532
2533         /*
2534          * A route for this destination might already exist.  If so,
2535          * update it, otherwise create a new one.
2536          */
2537         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2538                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2539                         sc->sc_brtexceeded++;
2540                         return (ENOSPC);
2541                 }
2542                 /* Check per interface address limits (if enabled) */
2543                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2544                         bif->bif_addrexceeded++;
2545                         return (ENOSPC);
2546                 }
2547
2548                 /*
2549                  * Allocate a new bridge forwarding node, and
2550                  * initialize the expiration time and Ethernet
2551                  * address.
2552                  */
2553                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2554                 if (brt == NULL)
2555                         return (ENOMEM);
2556
2557                 if (bif->bif_flags & IFBIF_STICKY)
2558                         brt->brt_flags = IFBAF_STICKY;
2559                 else
2560                         brt->brt_flags = IFBAF_DYNAMIC;
2561
2562                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2563                 brt->brt_vlan = vlan;
2564
2565                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2566                         uma_zfree(bridge_rtnode_zone, brt);
2567                         return (error);
2568                 }
2569                 brt->brt_dst = bif;
2570                 bif->bif_addrcnt++;
2571         }
2572
2573         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2574             brt->brt_dst != bif) {
2575                 brt->brt_dst->bif_addrcnt--;
2576                 brt->brt_dst = bif;
2577                 brt->brt_dst->bif_addrcnt++;
2578         }
2579
2580         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2581                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2582         if (setflags)
2583                 brt->brt_flags = flags;
2584
2585         return (0);
2586 }
2587
2588 /*
2589  * bridge_rtlookup:
2590  *
2591  *      Lookup the destination interface for an address.
2592  */
2593 static struct ifnet *
2594 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2595 {
2596         struct bridge_rtnode *brt;
2597
2598         BRIDGE_LOCK_ASSERT(sc);
2599
2600         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2601                 return (NULL);
2602
2603         return (brt->brt_ifp);
2604 }
2605
2606 /*
2607  * bridge_rttrim:
2608  *
2609  *      Trim the routine table so that we have a number
2610  *      of routing entries less than or equal to the
2611  *      maximum number.
2612  */
2613 static void
2614 bridge_rttrim(struct bridge_softc *sc)
2615 {
2616         struct bridge_rtnode *brt, *nbrt;
2617
2618         BRIDGE_LOCK_ASSERT(sc);
2619
2620         /* Make sure we actually need to do this. */
2621         if (sc->sc_brtcnt <= sc->sc_brtmax)
2622                 return;
2623
2624         /* Force an aging cycle; this might trim enough addresses. */
2625         bridge_rtage(sc);
2626         if (sc->sc_brtcnt <= sc->sc_brtmax)
2627                 return;
2628
2629         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2630                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2631                         bridge_rtnode_destroy(sc, brt);
2632                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2633                                 return;
2634                 }
2635         }
2636 }
2637
2638 /*
2639  * bridge_timer:
2640  *
2641  *      Aging timer for the bridge.
2642  */
2643 static void
2644 bridge_timer(void *arg)
2645 {
2646         struct bridge_softc *sc = arg;
2647
2648         BRIDGE_LOCK_ASSERT(sc);
2649
2650         bridge_rtage(sc);
2651
2652         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2653                 callout_reset(&sc->sc_brcallout,
2654                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2655 }
2656
2657 /*
2658  * bridge_rtage:
2659  *
2660  *      Perform an aging cycle.
2661  */
2662 static void
2663 bridge_rtage(struct bridge_softc *sc)
2664 {
2665         struct bridge_rtnode *brt, *nbrt;
2666
2667         BRIDGE_LOCK_ASSERT(sc);
2668
2669         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2670                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2671                         if (time_uptime >= brt->brt_expire)
2672                                 bridge_rtnode_destroy(sc, brt);
2673                 }
2674         }
2675 }
2676
2677 /*
2678  * bridge_rtflush:
2679  *
2680  *      Remove all dynamic addresses from the bridge.
2681  */
2682 static void
2683 bridge_rtflush(struct bridge_softc *sc, int full)
2684 {
2685         struct bridge_rtnode *brt, *nbrt;
2686
2687         BRIDGE_LOCK_ASSERT(sc);
2688
2689         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2690                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2691                         bridge_rtnode_destroy(sc, brt);
2692         }
2693 }
2694
2695 /*
2696  * bridge_rtdaddr:
2697  *
2698  *      Remove an address from the table.
2699  */
2700 static int
2701 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2702 {
2703         struct bridge_rtnode *brt;
2704         int found = 0;
2705
2706         BRIDGE_LOCK_ASSERT(sc);
2707
2708         /*
2709          * If vlan is zero then we want to delete for all vlans so the lookup
2710          * may return more than one.
2711          */
2712         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2713                 bridge_rtnode_destroy(sc, brt);
2714                 found = 1;
2715         }
2716
2717         return (found ? 0 : ENOENT);
2718 }
2719
2720 /*
2721  * bridge_rtdelete:
2722  *
2723  *      Delete routes to a speicifc member interface.
2724  */
2725 static void
2726 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2727 {
2728         struct bridge_rtnode *brt, *nbrt;
2729
2730         BRIDGE_LOCK_ASSERT(sc);
2731
2732         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2733                 if (brt->brt_ifp == ifp && (full ||
2734                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2735                         bridge_rtnode_destroy(sc, brt);
2736         }
2737 }
2738
2739 /*
2740  * bridge_rtable_init:
2741  *
2742  *      Initialize the route table for this bridge.
2743  */
2744 static int
2745 bridge_rtable_init(struct bridge_softc *sc)
2746 {
2747         int i;
2748
2749         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2750             M_DEVBUF, M_NOWAIT);
2751         if (sc->sc_rthash == NULL)
2752                 return (ENOMEM);
2753
2754         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2755                 LIST_INIT(&sc->sc_rthash[i]);
2756
2757         sc->sc_rthash_key = arc4random();
2758
2759         LIST_INIT(&sc->sc_rtlist);
2760
2761         return (0);
2762 }
2763
2764 /*
2765  * bridge_rtable_fini:
2766  *
2767  *      Deconstruct the route table for this bridge.
2768  */
2769 static void
2770 bridge_rtable_fini(struct bridge_softc *sc)
2771 {
2772
2773         KASSERT(sc->sc_brtcnt == 0,
2774             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2775         free(sc->sc_rthash, M_DEVBUF);
2776 }
2777
2778 /*
2779  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2780  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2781  */
2782 #define mix(a, b, c)                                                    \
2783 do {                                                                    \
2784         a -= b; a -= c; a ^= (c >> 13);                                 \
2785         b -= c; b -= a; b ^= (a << 8);                                  \
2786         c -= a; c -= b; c ^= (b >> 13);                                 \
2787         a -= b; a -= c; a ^= (c >> 12);                                 \
2788         b -= c; b -= a; b ^= (a << 16);                                 \
2789         c -= a; c -= b; c ^= (b >> 5);                                  \
2790         a -= b; a -= c; a ^= (c >> 3);                                  \
2791         b -= c; b -= a; b ^= (a << 10);                                 \
2792         c -= a; c -= b; c ^= (b >> 15);                                 \
2793 } while (/*CONSTCOND*/0)
2794
2795 static __inline uint32_t
2796 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2797 {
2798         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2799
2800         b += addr[5] << 8;
2801         b += addr[4];
2802         a += addr[3] << 24;
2803         a += addr[2] << 16;
2804         a += addr[1] << 8;
2805         a += addr[0];
2806
2807         mix(a, b, c);
2808
2809         return (c & BRIDGE_RTHASH_MASK);
2810 }
2811
2812 #undef mix
2813
2814 static int
2815 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2816 {
2817         int i, d;
2818
2819         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2820                 d = ((int)a[i]) - ((int)b[i]);
2821         }
2822
2823         return (d);
2824 }
2825
2826 /*
2827  * bridge_rtnode_lookup:
2828  *
2829  *      Look up a bridge route node for the specified destination. Compare the
2830  *      vlan id or if zero then just return the first match.
2831  */
2832 static struct bridge_rtnode *
2833 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2834 {
2835         struct bridge_rtnode *brt;
2836         uint32_t hash;
2837         int dir;
2838
2839         BRIDGE_LOCK_ASSERT(sc);
2840
2841         hash = bridge_rthash(sc, addr);
2842         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2843                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2844                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2845                         return (brt);
2846                 if (dir > 0)
2847                         return (NULL);
2848         }
2849
2850         return (NULL);
2851 }
2852
2853 /*
2854  * bridge_rtnode_insert:
2855  *
2856  *      Insert the specified bridge node into the route table.  We
2857  *      assume the entry is not already in the table.
2858  */
2859 static int
2860 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2861 {
2862         struct bridge_rtnode *lbrt;
2863         uint32_t hash;
2864         int dir;
2865
2866         BRIDGE_LOCK_ASSERT(sc);
2867
2868         hash = bridge_rthash(sc, brt->brt_addr);
2869
2870         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2871         if (lbrt == NULL) {
2872                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2873                 goto out;
2874         }
2875
2876         do {
2877                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2878                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2879                         return (EEXIST);
2880                 if (dir > 0) {
2881                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2882                         goto out;
2883                 }
2884                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2885                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2886                         goto out;
2887                 }
2888                 lbrt = LIST_NEXT(lbrt, brt_hash);
2889         } while (lbrt != NULL);
2890
2891 #ifdef DIAGNOSTIC
2892         panic("bridge_rtnode_insert: impossible");
2893 #endif
2894
2895 out:
2896         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2897         sc->sc_brtcnt++;
2898
2899         return (0);
2900 }
2901
2902 /*
2903  * bridge_rtnode_destroy:
2904  *
2905  *      Destroy a bridge rtnode.
2906  */
2907 static void
2908 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2909 {
2910         BRIDGE_LOCK_ASSERT(sc);
2911
2912         LIST_REMOVE(brt, brt_hash);
2913
2914         LIST_REMOVE(brt, brt_list);
2915         sc->sc_brtcnt--;
2916         brt->brt_dst->bif_addrcnt--;
2917         uma_zfree(bridge_rtnode_zone, brt);
2918 }
2919
2920 /*
2921  * bridge_rtable_expire:
2922  *
2923  *      Set the expiry time for all routes on an interface.
2924  */
2925 static void
2926 bridge_rtable_expire(struct ifnet *ifp, int age)
2927 {
2928         struct bridge_softc *sc = ifp->if_bridge;
2929         struct bridge_rtnode *brt;
2930
2931         BRIDGE_LOCK(sc);
2932
2933         /*
2934          * If the age is zero then flush, otherwise set all the expiry times to
2935          * age for the interface
2936          */
2937         if (age == 0)
2938                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2939         else {
2940                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2941                         /* Cap the expiry time to 'age' */
2942                         if (brt->brt_ifp == ifp &&
2943                             brt->brt_expire > time_uptime + age &&
2944                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2945                                 brt->brt_expire = time_uptime + age;
2946                 }
2947         }
2948         BRIDGE_UNLOCK(sc);
2949 }
2950
2951 /*
2952  * bridge_state_change:
2953  *
2954  *      Callback from the bridgestp code when a port changes states.
2955  */
2956 static void
2957 bridge_state_change(struct ifnet *ifp, int state)
2958 {
2959         struct bridge_softc *sc = ifp->if_bridge;
2960         static const char *stpstates[] = {
2961                 "disabled",
2962                 "listening",
2963                 "learning",
2964                 "forwarding",
2965                 "blocking",
2966                 "discarding"
2967         };
2968
2969         if (log_stp)
2970                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2971                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2972 }
2973
2974 /*
2975  * Send bridge packets through pfil if they are one of the types pfil can deal
2976  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2977  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2978  * that interface.
2979  */
2980 static int
2981 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2982 {
2983         int snap, error, i, hlen;
2984         struct ether_header *eh1, eh2;
2985         struct ip_fw_args args;
2986         struct ip *ip;
2987         struct llc llc1;
2988         u_int16_t ether_type;
2989
2990         snap = 0;
2991         error = -1;     /* Default error if not error == 0 */
2992
2993 #if 0
2994         /* we may return with the IP fields swapped, ensure its not shared */
2995         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2996 #endif
2997
2998         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2999                 return (0); /* filtering is disabled */
3000
3001         i = min((*mp)->m_pkthdr.len, max_protohdr);
3002         if ((*mp)->m_len < i) {
3003             *mp = m_pullup(*mp, i);
3004             if (*mp == NULL) {
3005                 printf("%s: m_pullup failed\n", __func__);
3006                 return (-1);
3007             }
3008         }
3009
3010         eh1 = mtod(*mp, struct ether_header *);
3011         ether_type = ntohs(eh1->ether_type);
3012
3013         /*
3014          * Check for SNAP/LLC.
3015          */
3016         if (ether_type < ETHERMTU) {
3017                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3018
3019                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3020                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3021                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3022                     llc2->llc_control == LLC_UI) {
3023                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3024                         snap = 1;
3025                 }
3026         }
3027
3028         /*
3029          * If we're trying to filter bridge traffic, don't look at anything
3030          * other than IP and ARP traffic.  If the filter doesn't understand
3031          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3032          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3033          * but of course we don't have an AppleTalk filter to begin with.
3034          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3035          * ARP traffic.)
3036          */
3037         switch (ether_type) {
3038                 case ETHERTYPE_ARP:
3039                 case ETHERTYPE_REVARP:
3040                         if (pfil_ipfw_arp == 0)
3041                                 return (0); /* Automatically pass */
3042                         break;
3043
3044                 case ETHERTYPE_IP:
3045 #ifdef INET6
3046                 case ETHERTYPE_IPV6:
3047 #endif /* INET6 */
3048                         break;
3049                 default:
3050                         /*
3051                          * Check to see if the user wants to pass non-ip
3052                          * packets, these will not be checked by pfil(9) and
3053                          * passed unconditionally so the default is to drop.
3054                          */
3055                         if (pfil_onlyip)
3056                                 goto bad;
3057         }
3058
3059         /* Strip off the Ethernet header and keep a copy. */
3060         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3061         m_adj(*mp, ETHER_HDR_LEN);
3062
3063         /* Strip off snap header, if present */
3064         if (snap) {
3065                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3066                 m_adj(*mp, sizeof(struct llc));
3067         }
3068
3069         /*
3070          * Check the IP header for alignment and errors
3071          */
3072         if (dir == PFIL_IN) {
3073                 switch (ether_type) {
3074                         case ETHERTYPE_IP:
3075                                 error = bridge_ip_checkbasic(mp);
3076                                 break;
3077 #ifdef INET6
3078                         case ETHERTYPE_IPV6:
3079                                 error = bridge_ip6_checkbasic(mp);
3080                                 break;
3081 #endif /* INET6 */
3082                         default:
3083                                 error = 0;
3084                 }
3085                 if (error)
3086                         goto bad;
3087         }
3088
3089         /* XXX this section is also in if_ethersubr.c */
3090         // XXX PFIL_OUT or DIR_OUT ?
3091         if (V_ip_fw_chk_ptr && pfil_ipfw != 0 &&
3092                         dir == PFIL_OUT && ifp != NULL) {
3093                 struct m_tag *mtag;
3094
3095                 error = -1;
3096                 /* fetch the start point from existing tags, if any */
3097                 mtag = m_tag_locate(*mp, MTAG_IPFW_RULE, 0, NULL);
3098                 if (mtag == NULL) {
3099                         args.rule.slot = 0;
3100                 } else {
3101                         struct ipfw_rule_ref *r;
3102
3103                         /* XXX can we free the tag after use ? */
3104                         mtag->m_tag_id = PACKET_TAG_NONE;
3105                         r = (struct ipfw_rule_ref *)(mtag + 1);
3106                         /* packet already partially processed ? */
3107                         if (r->info & IPFW_ONEPASS)
3108                                 goto ipfwpass;
3109                         args.rule = *r;
3110                 }
3111
3112                 args.m = *mp;
3113                 args.oif = ifp;
3114                 args.next_hop = NULL;
3115                 args.eh = &eh2;
3116                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3117                 i = V_ip_fw_chk_ptr(&args);
3118                 *mp = args.m;
3119
3120                 if (*mp == NULL)
3121                         return (error);
3122
3123                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3124
3125                         /* put the Ethernet header back on */
3126                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3127                         if (*mp == NULL)
3128                                 return (error);
3129                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3130
3131                         /*
3132                          * Pass the pkt to dummynet, which consumes it. The
3133                          * packet will return to us via bridge_dummynet().
3134                          */
3135                         args.oif = ifp;
3136                         ip_dn_io_ptr(mp, DIR_FWD | PROTO_IFB, &args);
3137                         return (error);
3138                 }
3139
3140                 if (i != IP_FW_PASS) /* drop */
3141                         goto bad;
3142         }
3143
3144 ipfwpass:
3145         error = 0;
3146
3147         /*
3148          * Run the packet through pfil
3149          */
3150         switch (ether_type) {
3151         case ETHERTYPE_IP:
3152                 /*
3153                  * before calling the firewall, swap fields the same as
3154                  * IP does. here we assume the header is contiguous
3155                  */
3156                 ip = mtod(*mp, struct ip *);
3157
3158                 ip->ip_len = ntohs(ip->ip_len);
3159                 ip->ip_off = ntohs(ip->ip_off);
3160
3161                 /*
3162                  * Run pfil on the member interface and the bridge, both can
3163                  * be skipped by clearing pfil_member or pfil_bridge.
3164                  *
3165                  * Keep the order:
3166                  *   in_if -> bridge_if -> out_if
3167                  */
3168                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3169                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3170                                         dir, NULL);
3171
3172                 if (*mp == NULL || error != 0) /* filter may consume */
3173                         break;
3174
3175                 if (pfil_member && ifp != NULL)
3176                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3177                                         dir, NULL);
3178
3179                 if (*mp == NULL || error != 0) /* filter may consume */
3180                         break;
3181
3182                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3183                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3184                                         dir, NULL);
3185
3186                 if (*mp == NULL || error != 0) /* filter may consume */
3187                         break;
3188
3189                 /* check if we need to fragment the packet */
3190                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3191                         i = (*mp)->m_pkthdr.len;
3192                         if (i > ifp->if_mtu) {
3193                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3194                                             &llc1);
3195                                 return (error);
3196                         }
3197                 }
3198
3199                 /* Recalculate the ip checksum and restore byte ordering */
3200                 ip = mtod(*mp, struct ip *);
3201                 hlen = ip->ip_hl << 2;
3202                 if (hlen < sizeof(struct ip))
3203                         goto bad;
3204                 if (hlen > (*mp)->m_len) {
3205                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3206                                 goto bad;
3207                         ip = mtod(*mp, struct ip *);
3208                         if (ip == NULL)
3209                                 goto bad;
3210                 }
3211                 ip->ip_len = htons(ip->ip_len);
3212                 ip->ip_off = htons(ip->ip_off);
3213                 ip->ip_sum = 0;
3214                 if (hlen == sizeof(struct ip))
3215                         ip->ip_sum = in_cksum_hdr(ip);
3216                 else
3217                         ip->ip_sum = in_cksum(*mp, hlen);
3218
3219                 break;
3220 #ifdef INET6
3221         case ETHERTYPE_IPV6:
3222                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3223                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3224                                         dir, NULL);
3225
3226                 if (*mp == NULL || error != 0) /* filter may consume */
3227                         break;
3228
3229                 if (pfil_member && ifp != NULL)
3230                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3231                                         dir, NULL);
3232
3233                 if (*mp == NULL || error != 0) /* filter may consume */
3234                         break;
3235
3236                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3237                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3238                                         dir, NULL);
3239                 break;
3240 #endif
3241         default:
3242                 error = 0;
3243                 break;
3244         }
3245
3246         if (*mp == NULL)
3247                 return (error);
3248         if (error != 0)
3249                 goto bad;
3250
3251         error = -1;
3252
3253         /*
3254          * Finally, put everything back the way it was and return
3255          */
3256         if (snap) {
3257                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3258                 if (*mp == NULL)
3259                         return (error);
3260                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3261         }
3262
3263         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3264         if (*mp == NULL)
3265                 return (error);
3266         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3267
3268         return (0);
3269
3270 bad:
3271         m_freem(*mp);
3272         *mp = NULL;
3273         return (error);
3274 }
3275
3276 /*
3277  * Perform basic checks on header size since
3278  * pfil assumes ip_input has already processed
3279  * it for it.  Cut-and-pasted from ip_input.c.
3280  * Given how simple the IPv6 version is,
3281  * does the IPv4 version really need to be
3282  * this complicated?
3283  *
3284  * XXX Should we update ipstat here, or not?
3285  * XXX Right now we update ipstat but not
3286  * XXX csum_counter.
3287  */
3288 static int
3289 bridge_ip_checkbasic(struct mbuf **mp)
3290 {
3291         struct mbuf *m = *mp;
3292         struct ip *ip;
3293         int len, hlen;
3294         u_short sum;
3295
3296         if (*mp == NULL)
3297                 return (-1);
3298
3299         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3300                 if ((m = m_copyup(m, sizeof(struct ip),
3301                         (max_linkhdr + 3) & ~3)) == NULL) {
3302                         /* XXXJRT new stat, please */
3303                         KMOD_IPSTAT_INC(ips_toosmall);
3304                         goto bad;
3305                 }
3306         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3307                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3308                         KMOD_IPSTAT_INC(ips_toosmall);
3309                         goto bad;
3310                 }
3311         }
3312         ip = mtod(m, struct ip *);
3313         if (ip == NULL) goto bad;
3314
3315         if (ip->ip_v != IPVERSION) {
3316                 KMOD_IPSTAT_INC(ips_badvers);
3317                 goto bad;
3318         }
3319         hlen = ip->ip_hl << 2;
3320         if (hlen < sizeof(struct ip)) { /* minimum header length */
3321                 KMOD_IPSTAT_INC(ips_badhlen);
3322                 goto bad;
3323         }
3324         if (hlen > m->m_len) {
3325                 if ((m = m_pullup(m, hlen)) == 0) {
3326                         KMOD_IPSTAT_INC(ips_badhlen);
3327                         goto bad;
3328                 }
3329                 ip = mtod(m, struct ip *);
3330                 if (ip == NULL) goto bad;
3331         }
3332
3333         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3334                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3335         } else {
3336                 if (hlen == sizeof(struct ip)) {
3337                         sum = in_cksum_hdr(ip);
3338                 } else {
3339                         sum = in_cksum(m, hlen);
3340                 }
3341         }
3342         if (sum) {
3343                 KMOD_IPSTAT_INC(ips_badsum);
3344                 goto bad;
3345         }
3346
3347         /* Retrieve the packet length. */
3348         len = ntohs(ip->ip_len);
3349
3350         /*
3351          * Check for additional length bogosity
3352          */
3353         if (len < hlen) {
3354                 KMOD_IPSTAT_INC(ips_badlen);
3355                 goto bad;
3356         }
3357
3358         /*
3359          * Check that the amount of data in the buffers
3360          * is as at least much as the IP header would have us expect.
3361          * Drop packet if shorter than we expect.
3362          */
3363         if (m->m_pkthdr.len < len) {
3364                 KMOD_IPSTAT_INC(ips_tooshort);
3365                 goto bad;
3366         }
3367
3368         /* Checks out, proceed */
3369         *mp = m;
3370         return (0);
3371
3372 bad:
3373         *mp = m;
3374         return (-1);
3375 }
3376
3377 #ifdef INET6
3378 /*
3379  * Same as above, but for IPv6.
3380  * Cut-and-pasted from ip6_input.c.
3381  * XXX Should we update ip6stat, or not?
3382  */
3383 static int
3384 bridge_ip6_checkbasic(struct mbuf **mp)
3385 {
3386         struct mbuf *m = *mp;
3387         struct ip6_hdr *ip6;
3388
3389         /*
3390          * If the IPv6 header is not aligned, slurp it up into a new
3391          * mbuf with space for link headers, in the event we forward
3392          * it.  Otherwise, if it is aligned, make sure the entire base
3393          * IPv6 header is in the first mbuf of the chain.
3394          */
3395         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3396                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3397                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3398                             (max_linkhdr + 3) & ~3)) == NULL) {
3399                         /* XXXJRT new stat, please */
3400                         V_ip6stat.ip6s_toosmall++;
3401                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3402                         goto bad;
3403                 }
3404         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3405                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3406                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3407                         V_ip6stat.ip6s_toosmall++;
3408                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3409                         goto bad;
3410                 }
3411         }
3412
3413         ip6 = mtod(m, struct ip6_hdr *);
3414
3415         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3416                 V_ip6stat.ip6s_badvers++;
3417                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3418                 goto bad;
3419         }
3420
3421         /* Checks out, proceed */
3422         *mp = m;
3423         return (0);
3424
3425 bad:
3426         *mp = m;
3427         return (-1);
3428 }
3429 #endif /* INET6 */
3430
3431 /*
3432  * bridge_fragment:
3433  *
3434  *      Return a fragmented mbuf chain.
3435  */
3436 static int
3437 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3438     int snap, struct llc *llc)
3439 {
3440         struct mbuf *m0;
3441         struct ip *ip;
3442         int error = -1;
3443
3444         if (m->m_len < sizeof(struct ip) &&
3445             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3446                 goto out;
3447         ip = mtod(m, struct ip *);
3448
3449         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3450                     CSUM_DELAY_IP);
3451         if (error)
3452                 goto out;
3453
3454         /* walk the chain and re-add the Ethernet header */
3455         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3456                 if (error == 0) {
3457                         if (snap) {
3458                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3459                                 if (m0 == NULL) {
3460                                         error = ENOBUFS;
3461                                         continue;
3462                                 }
3463                                 bcopy(llc, mtod(m0, caddr_t),
3464                                     sizeof(struct llc));
3465                         }
3466                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3467                         if (m0 == NULL) {
3468                                 error = ENOBUFS;
3469                                 continue;
3470                         }
3471                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3472                 } else
3473                         m_freem(m);
3474         }
3475
3476         if (error == 0)
3477                 KMOD_IPSTAT_INC(ips_fragmented);
3478
3479         return (error);
3480
3481 out:
3482         if (m != NULL)
3483                 m_freem(m);
3484         return (error);
3485 }
3486
3487 static void
3488 bridge_linkstate(struct ifnet *ifp)
3489 {
3490         struct bridge_softc *sc = ifp->if_bridge;
3491         struct bridge_iflist *bif;
3492
3493         BRIDGE_LOCK(sc);
3494         bif = bridge_lookup_member_if(sc, ifp);
3495         if (bif == NULL) {
3496                 BRIDGE_UNLOCK(sc);
3497                 return;
3498         }
3499         bridge_linkcheck(sc);
3500         BRIDGE_UNLOCK(sc);
3501
3502         bstp_linkstate(&bif->bif_stp);
3503 }
3504
3505 static void
3506 bridge_linkcheck(struct bridge_softc *sc)
3507 {
3508         struct bridge_iflist *bif;
3509         int new_link, hasls;
3510
3511         BRIDGE_LOCK_ASSERT(sc);
3512         new_link = LINK_STATE_DOWN;
3513         hasls = 0;
3514         /* Our link is considered up if at least one of our ports is active */
3515         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3516                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3517                         hasls++;
3518                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3519                         new_link = LINK_STATE_UP;
3520                         break;
3521                 }
3522         }
3523         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3524                 /* If no interfaces support link-state then we default to up */
3525                 new_link = LINK_STATE_UP;
3526         }
3527         if_link_state_change(sc->sc_ifp, new_link);
3528 }