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 #include <netinet6/in6_ifattach.h>
 123 #endif
 124 #if defined(INET) || defined(INET6)
 125 #include <netinet/ip_carp.h>
 126 #endif
 127 #include <machine/in_cksum.h>
 128 #include <netinet/if_ether.h> /* for struct arpcom */
 129 #include <net/bridgestp.h>
 130 #include <net/if_bridgevar.h>
 131 #include <net/if_llc.h>
 132 #include <net/if_vlan_var.h>
 133
 134 #include <net/route.h>
 135 #include <netinet/ip_fw.h>
 136 #include <netpfil/ipfw/ip_fw_private.h>
 137
 138 /*
 139  * Size of the route hash table.  Must be a power of two.
 140  */
 141 #ifndef BRIDGE_RTHASH_SIZE
 142 #define BRIDGE_RTHASH_SIZE              1024
 143 #endif
 144
 145 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 146
 147 /*
 148  * Default maximum number of addresses to cache.
 149  */
 150 #ifndef BRIDGE_RTABLE_MAX
 151 #define BRIDGE_RTABLE_MAX               2000
 152 #endif
 153
 154 /*
 155  * Timeout (in seconds) for entries learned dynamically.
 156  */
 157 #ifndef BRIDGE_RTABLE_TIMEOUT
 158 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 159 #endif
 160
 161 /*
 162  * Number of seconds between walks of the route list.
 163  */
 164 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 165 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 166 #endif
 167
 168 /*
 169  * List of capabilities to possibly mask on the member interface.
 170  */
 171 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 172
 173 /*
 174  * List of capabilities to strip
 175  */
 176 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 177
 178 /*
 179  * Bridge interface list entry.
 180  */
 181 struct bridge_iflist {
 182         LIST_ENTRY(bridge_iflist) bif_next;
 183         struct ifnet            *bif_ifp;       /* member if */
 184         struct bstp_port        bif_stp;        /* STP state */
 185         uint32_t                bif_flags;      /* member if flags */
 186         int                     bif_savedcaps;  /* saved capabilities */
 187         uint32_t                bif_addrmax;    /* max # of addresses */
 188         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 189         uint32_t                bif_addrexceeded;/* # of address violations */
 190 };
 191
 192 /*
 193  * Bridge route node.
 194  */
 195 struct bridge_rtnode {
 196         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 197         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 198         struct bridge_iflist    *brt_dst;       /* destination if */
 199         unsigned long           brt_expire;     /* expiration time */
 200         uint8_t                 brt_flags;      /* address flags */
 201         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 202         uint16_t                brt_vlan;       /* vlan id */
 203 };
 204 #define brt_ifp                 brt_dst->bif_ifp
 205
 206 /*
 207  * Software state for each bridge.
 208  */
 209 struct bridge_softc {
 210         struct ifnet            *sc_ifp;        /* make this an interface */
 211         LIST_ENTRY(bridge_softc) sc_list;
 212         struct mtx              sc_mtx;
 213         struct cv               sc_cv;
 214         uint32_t                sc_brtmax;      /* max # of addresses */
 215         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 216         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 217         struct callout          sc_brcallout;   /* bridge callout */
 218         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 219         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 220         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 221         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 222         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 223         uint32_t                sc_rthash_key;  /* key for hash */
 224         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 225         struct bstp_state       sc_stp;         /* STP state */
 226         uint32_t                sc_brtexceeded; /* # of cache drops */
 227         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 228         u_char                  sc_defaddr[6];  /* Default MAC address */
 229 };
 230
 231 static struct mtx       bridge_list_mtx;
 232 eventhandler_tag        bridge_detach_cookie = NULL;
 233
 234 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 235
 236 uma_zone_t bridge_rtnode_zone;
 237
 238 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 239 static void     bridge_clone_destroy(struct ifnet *);
 240
 241 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 242 static void     bridge_mutecaps(struct bridge_softc *);
 243 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 244                     int);
 245 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 246 static void     bridge_init(void *);
 247 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 248 static void     bridge_stop(struct ifnet *, int);
 249 static void     bridge_start(struct ifnet *);
 250 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 251 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 252                     struct rtentry *);
 253 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 254                     struct mbuf *);
 255 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 256
 257 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 258                     struct mbuf *m);
 259
 260 static void     bridge_timer(void *);
 261
 262 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 263                     struct mbuf *, int);
 264 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 265
 266 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 267                     uint16_t, struct bridge_iflist *, int, uint8_t);
 268 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 269                     uint16_t);
 270 static void     bridge_rttrim(struct bridge_softc *);
 271 static void     bridge_rtage(struct bridge_softc *);
 272 static void     bridge_rtflush(struct bridge_softc *, int);
 273 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 274                     uint16_t);
 275
 276 static int      bridge_rtable_init(struct bridge_softc *);
 277 static void     bridge_rtable_fini(struct bridge_softc *);
 278
 279 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 280 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 281                     const uint8_t *, uint16_t);
 282 static int      bridge_rtnode_insert(struct bridge_softc *,
 283                     struct bridge_rtnode *);
 284 static void     bridge_rtnode_destroy(struct bridge_softc *,
 285                     struct bridge_rtnode *);
 286 static void     bridge_rtable_expire(struct ifnet *, int);
 287 static void     bridge_state_change(struct ifnet *, int);
 288
 289 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 290                     const char *name);
 291 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 292                     struct ifnet *ifp);
 293 static void     bridge_delete_member(struct bridge_softc *,
 294                     struct bridge_iflist *, int);
 295 static void     bridge_delete_span(struct bridge_softc *,
 296                     struct bridge_iflist *);
 297
 298 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 328 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 329 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 330                     int);
 331 static int      bridge_ip_checkbasic(struct mbuf **mp);
 332 #ifdef INET6
 333 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 334 #endif /* INET6 */
 335 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 336                     struct ether_header *, int, struct llc *);
 337 static void     bridge_linkstate(struct ifnet *ifp);
 338 static void     bridge_linkcheck(struct bridge_softc *sc);
 339
 340 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
 341
 342 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 343 #define VLANTAGOF(_m)   \
 344     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 345
 346 static struct bstp_cb_ops bridge_ops = {
 347         .bcb_state = bridge_state_change,
 348         .bcb_rtage = bridge_rtable_expire
 349 };
 350
 351 SYSCTL_DECL(_net_link);
 352 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 353
 354 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 355 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 356 static int pfil_member = 1; /* run pfil hooks on the member interface */
 357 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 358 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 359 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 360                                    locally destined packets */
 361 static int log_stp   = 0;   /* log STP state changes */
 362 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
 363 TUNABLE_INT("net.link.bridge.pfil_onlyip", &pfil_onlyip);
 364 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 365     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 366 TUNABLE_INT("net.link.bridge.ipfw_arp", &pfil_ipfw_arp);
 367 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 368     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 369 TUNABLE_INT("net.link.bridge.pfil_bridge", &pfil_bridge);
 370 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 371     &pfil_bridge, 0, "Packet filter on the bridge interface");
 372 TUNABLE_INT("net.link.bridge.pfil_member", &pfil_member);
 373 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 374     &pfil_member, 0, "Packet filter on the member interface");
 375 TUNABLE_INT("net.link.bridge.pfil_local_phys", &pfil_local_phys);
 376 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 377     &pfil_local_phys, 0,
 378     "Packet filter on the physical interface for locally destined packets");
 379 TUNABLE_INT("net.link.bridge.log_stp", &log_stp);
 380 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 381     &log_stp, 0, "Log STP state changes");
 382 TUNABLE_INT("net.link.bridge.inherit_mac", &bridge_inherit_mac);
 383 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
 384     &bridge_inherit_mac, 0,
 385     "Inherit MAC address from the first bridge member");
 386
 387 static VNET_DEFINE(int, allow_llz_overlap) = 0;
 388 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 389 SYSCTL_VNET_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap, CTLFLAG_RW,
 390     &VNET_NAME(allow_llz_overlap), 0, "Allow overlap of link-local scope "
 391     "zones of a bridge interface and the member interfaces");
 392
 393 struct bridge_control {
 394         int     (*bc_func)(struct bridge_softc *, void *);
 395         int     bc_argsize;
 396         int     bc_flags;
 397 };
 398
 399 #define BC_F_COPYIN             0x01    /* copy arguments in */
 400 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 401 #define BC_F_SUSER              0x04    /* do super-user check */
 402
 403 const struct bridge_control bridge_control_table[] = {
 404         { bridge_ioctl_add,             sizeof(struct ifbreq),
 405           BC_F_COPYIN|BC_F_SUSER },
 406         { bridge_ioctl_del,             sizeof(struct ifbreq),
 407           BC_F_COPYIN|BC_F_SUSER },
 408
 409         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 410           BC_F_COPYIN|BC_F_COPYOUT },
 411         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 412           BC_F_COPYIN|BC_F_SUSER },
 413
 414         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 415           BC_F_COPYIN|BC_F_SUSER },
 416         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 417           BC_F_COPYOUT },
 418
 419         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 420           BC_F_COPYIN|BC_F_COPYOUT },
 421         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 422           BC_F_COPYIN|BC_F_COPYOUT },
 423
 424         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 425           BC_F_COPYIN|BC_F_SUSER },
 426
 427         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 428           BC_F_COPYIN|BC_F_SUSER },
 429         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 430           BC_F_COPYOUT },
 431
 432         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 433           BC_F_COPYIN|BC_F_SUSER },
 434
 435         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 436           BC_F_COPYIN|BC_F_SUSER },
 437
 438         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 439           BC_F_COPYOUT },
 440         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 441           BC_F_COPYIN|BC_F_SUSER },
 442
 443         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 444           BC_F_COPYOUT },
 445         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 446           BC_F_COPYIN|BC_F_SUSER },
 447
 448         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 449           BC_F_COPYOUT },
 450         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 451           BC_F_COPYIN|BC_F_SUSER },
 452
 453         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 454           BC_F_COPYOUT },
 455         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 456           BC_F_COPYIN|BC_F_SUSER },
 457
 458         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 459           BC_F_COPYIN|BC_F_SUSER },
 460
 461         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 462           BC_F_COPYIN|BC_F_SUSER },
 463
 464         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 465           BC_F_COPYIN|BC_F_SUSER },
 466         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 467           BC_F_COPYIN|BC_F_SUSER },
 468
 469         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 470           BC_F_COPYOUT },
 471
 472         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 473           BC_F_COPYOUT },
 474
 475         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 476           BC_F_COPYIN|BC_F_COPYOUT },
 477
 478         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 479           BC_F_COPYIN|BC_F_SUSER },
 480
 481         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 482           BC_F_COPYIN|BC_F_SUSER },
 483
 484         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 485           BC_F_COPYIN|BC_F_SUSER },
 486
 487 };
 488 const int bridge_control_table_size =
 489     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 490
 491 LIST_HEAD(, bridge_softc) bridge_list;
 492
 493 IFC_SIMPLE_DECLARE(bridge, 0);
 494
 495 static int
 496 bridge_modevent(module_t mod, int type, void *data)
 497 {
 498
 499         switch (type) {
 500         case MOD_LOAD:
 501                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 502                 if_clone_attach(&bridge_cloner);
 503                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 504                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 505                     UMA_ALIGN_PTR, 0);
 506                 LIST_INIT(&bridge_list);
 507                 bridge_input_p = bridge_input;
 508                 bridge_output_p = bridge_output;
 509                 bridge_dn_p = bridge_dummynet;
 510                 bridge_linkstate_p = bridge_linkstate;
 511                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 512                     ifnet_departure_event, bridge_ifdetach, NULL,
 513                     EVENTHANDLER_PRI_ANY);
 514                 break;
 515         case MOD_UNLOAD:
 516                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 517                     bridge_detach_cookie);
 518                 if_clone_detach(&bridge_cloner);
 519                 uma_zdestroy(bridge_rtnode_zone);
 520                 bridge_input_p = NULL;
 521                 bridge_output_p = NULL;
 522                 bridge_dn_p = NULL;
 523                 bridge_linkstate_p = NULL;
 524                 mtx_destroy(&bridge_list_mtx);
 525                 break;
 526         default:
 527                 return (EOPNOTSUPP);
 528         }
 529         return (0);
 530 }
 531
 532 static moduledata_t bridge_mod = {
 533         "if_bridge",
 534         bridge_modevent,
 535         0
 536 };
 537
 538 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 539 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 540
 541 /*
 542  * handler for net.link.bridge.pfil_ipfw
 543  */
 544 static int
 545 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 546 {
 547         int enable = pfil_ipfw;
 548         int error;
 549
 550         error = sysctl_handle_int(oidp, &enable, 0, req);
 551         enable = (enable) ? 1 : 0;
 552
 553         if (enable != pfil_ipfw) {
 554                 pfil_ipfw = enable;
 555
 556                 /*
 557                  * Disable pfil so that ipfw doesnt run twice, if the user
 558                  * really wants both then they can re-enable pfil_bridge and/or
 559                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 560                  * layer2 type.
 561                  */
 562                 if (pfil_ipfw) {
 563                         pfil_onlyip = 0;
 564                         pfil_bridge = 0;
 565                         pfil_member = 0;
 566                 }
 567         }
 568
 569         return (error);
 570 }
 571 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 572             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 573
 574 /*
 575  * bridge_clone_create:
 576  *
 577  *      Create a new bridge instance.
 578  */
 579 static int
 580 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 581 {
 582         struct bridge_softc *sc, *sc2;
 583         struct ifnet *bifp, *ifp;
 584         int fb, retry;
 585         unsigned long hostid;
 586
 587         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 588         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 589         if (ifp == NULL) {
 590                 free(sc, M_DEVBUF);
 591                 return (ENOSPC);
 592         }
 593
 594         BRIDGE_LOCK_INIT(sc);
 595         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 596         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 597
 598         /* Initialize our routing table. */
 599         bridge_rtable_init(sc);
 600
 601         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 602
 603         LIST_INIT(&sc->sc_iflist);
 604         LIST_INIT(&sc->sc_spanlist);
 605
 606         ifp->if_softc = sc;
 607         if_initname(ifp, ifc->ifc_name, unit);
 608         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 609         ifp->if_ioctl = bridge_ioctl;
 610         ifp->if_start = bridge_start;
 611         ifp->if_init = bridge_init;
 612         ifp->if_type = IFT_BRIDGE;
 613         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 614         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 615         IFQ_SET_READY(&ifp->if_snd);
 616
 617         /*
 618          * Generate an ethernet address with a locally administered address.
 619          *
 620          * Since we are using random ethernet addresses for the bridge, it is
 621          * possible that we might have address collisions, so make sure that
 622          * this hardware address isn't already in use on another bridge.
 623          * The first try uses the hostid and falls back to arc4rand().
 624          */
 625         fb = 0;
 626         getcredhostid(curthread->td_ucred, &hostid);
 627         for (retry = 1; retry != 0;) {
 628                 if (fb || hostid == 0) {
 629                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 630                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 631                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 632                 } else {
 633                         sc->sc_defaddr[0] = 0x2;
 634                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 635                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 636                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 637                         sc->sc_defaddr[4] =  hostid        & 0xff;
 638                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 639                 }
 640
 641                 fb = 1;
 642                 retry = 0;
 643                 mtx_lock(&bridge_list_mtx);
 644                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 645                         bifp = sc2->sc_ifp;
 646                         if (memcmp(sc->sc_defaddr,
 647                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 648                                 retry = 1;
 649                 }
 650                 mtx_unlock(&bridge_list_mtx);
 651         }
 652
 653         bstp_attach(&sc->sc_stp, &bridge_ops);
 654         ether_ifattach(ifp, sc->sc_defaddr);
 655         /* Now undo some of the damage... */
 656         ifp->if_baudrate = 0;
 657         ifp->if_type = IFT_BRIDGE;
 658
 659         mtx_lock(&bridge_list_mtx);
 660         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 661         mtx_unlock(&bridge_list_mtx);
 662
 663         return (0);
 664 }
 665
 666 /*
 667  * bridge_clone_destroy:
 668  *
 669  *      Destroy a bridge instance.
 670  */
 671 static void
 672 bridge_clone_destroy(struct ifnet *ifp)
 673 {
 674         struct bridge_softc *sc = ifp->if_softc;
 675         struct bridge_iflist *bif;
 676
 677         BRIDGE_LOCK(sc);
 678
 679         bridge_stop(ifp, 1);
 680         ifp->if_flags &= ~IFF_UP;
 681
 682         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 683                 bridge_delete_member(sc, bif, 0);
 684
 685         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 686                 bridge_delete_span(sc, bif);
 687         }
 688
 689         BRIDGE_UNLOCK(sc);
 690
 691         callout_drain(&sc->sc_brcallout);
 692
 693         mtx_lock(&bridge_list_mtx);
 694         LIST_REMOVE(sc, sc_list);
 695         mtx_unlock(&bridge_list_mtx);
 696
 697         bstp_detach(&sc->sc_stp);
 698         ether_ifdetach(ifp);
 699         if_free_type(ifp, IFT_ETHER);
 700
 701         /* Tear down the routing table. */
 702         bridge_rtable_fini(sc);
 703
 704         BRIDGE_LOCK_DESTROY(sc);
 705         free(sc, M_DEVBUF);
 706 }
 707
 708 /*
 709  * bridge_ioctl:
 710  *
 711  *      Handle a control request from the operator.
 712  */
 713 static int
 714 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 715 {
 716         struct bridge_softc *sc = ifp->if_softc;
 717         struct ifreq *ifr = (struct ifreq *)data;
 718         struct bridge_iflist *bif;
 719         struct thread *td = curthread;
 720         union {
 721                 struct ifbreq ifbreq;
 722                 struct ifbifconf ifbifconf;
 723                 struct ifbareq ifbareq;
 724                 struct ifbaconf ifbaconf;
 725                 struct ifbrparam ifbrparam;
 726                 struct ifbropreq ifbropreq;
 727         } args;
 728         struct ifdrv *ifd = (struct ifdrv *) data;
 729         const struct bridge_control *bc;
 730         int error = 0;
 731
 732         switch (cmd) {
 733
 734         case SIOCADDMULTI:
 735         case SIOCDELMULTI:
 736                 break;
 737
 738         case SIOCGDRVSPEC:
 739         case SIOCSDRVSPEC:
 740                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 741                         error = EINVAL;
 742                         break;
 743                 }
 744                 bc = &bridge_control_table[ifd->ifd_cmd];
 745
 746                 if (cmd == SIOCGDRVSPEC &&
 747                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 748                         error = EINVAL;
 749                         break;
 750                 }
 751                 else if (cmd == SIOCSDRVSPEC &&
 752                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 753                         error = EINVAL;
 754                         break;
 755                 }
 756
 757                 if (bc->bc_flags & BC_F_SUSER) {
 758                         error = priv_check(td, PRIV_NET_BRIDGE);
 759                         if (error)
 760                                 break;
 761                 }
 762
 763                 if (ifd->ifd_len != bc->bc_argsize ||
 764                     ifd->ifd_len > sizeof(args)) {
 765                         error = EINVAL;
 766                         break;
 767                 }
 768
 769                 bzero(&args, sizeof(args));
 770                 if (bc->bc_flags & BC_F_COPYIN) {
 771                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 772                         if (error)
 773                                 break;
 774                 }
 775
 776                 BRIDGE_LOCK(sc);
 777                 error = (*bc->bc_func)(sc, &args);
 778                 BRIDGE_UNLOCK(sc);
 779                 if (error)
 780                         break;
 781
 782                 if (bc->bc_flags & BC_F_COPYOUT)
 783                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 784
 785                 break;
 786
 787         case SIOCSIFFLAGS:
 788                 if (!(ifp->if_flags & IFF_UP) &&
 789                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 790                         /*
 791                          * If interface is marked down and it is running,
 792                          * then stop and disable it.
 793                          */
 794                         BRIDGE_LOCK(sc);
 795                         bridge_stop(ifp, 1);
 796                         BRIDGE_UNLOCK(sc);
 797                 } else if ((ifp->if_flags & IFF_UP) &&
 798                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 799                         /*
 800                          * If interface is marked up and it is stopped, then
 801                          * start it.
 802                          */
 803                         (*ifp->if_init)(sc);
 804                 }
 805                 break;
 806
 807         case SIOCSIFMTU:
 808                 if (ifr->ifr_mtu < 576) {
 809                         error = EINVAL;
 810                         break;
 811                 }
 812                 if (LIST_EMPTY(&sc->sc_iflist)) {
 813                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 814                         break;
 815                 }
 816                 BRIDGE_LOCK(sc);
 817                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 818                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 819                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
 820                                     " != %d\n", sc->sc_ifp->if_xname,
 821                                     bif->bif_ifp->if_mtu,
 822                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 823                                 error = EINVAL;
 824                                 break;
 825                         }
 826                 }
 827                 if (!error)
 828                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 829                 BRIDGE_UNLOCK(sc);
 830                 break;
 831         default:
 832                 /*
 833                  * drop the lock as ether_ioctl() will call bridge_start() and
 834                  * cause the lock to be recursed.
 835                  */
 836                 error = ether_ioctl(ifp, cmd, data);
 837                 break;
 838         }
 839
 840         return (error);
 841 }
 842
 843 /*
 844  * bridge_mutecaps:
 845  *
 846  *      Clear or restore unwanted capabilities on the member interface
 847  */
 848 static void
 849 bridge_mutecaps(struct bridge_softc *sc)
 850 {
 851         struct bridge_iflist *bif;
 852         int enabled, mask;
 853
 854         /* Initial bitmask of capabilities to test */
 855         mask = BRIDGE_IFCAPS_MASK;
 856
 857         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 858                 /* Every member must support it or its disabled */
 859                 mask &= bif->bif_savedcaps;
 860         }
 861
 862         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 863                 enabled = bif->bif_ifp->if_capenable;
 864                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 865                 /* strip off mask bits and enable them again if allowed */
 866                 enabled &= ~BRIDGE_IFCAPS_MASK;
 867                 enabled |= mask;
 868                 bridge_set_ifcap(sc, bif, enabled);
 869         }
 870
 871 }
 872
 873 static void
 874 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 875 {
 876         struct ifnet *ifp = bif->bif_ifp;
 877         struct ifreq ifr;
 878         int error;
 879
 880         bzero(&ifr, sizeof(ifr));
 881         ifr.ifr_reqcap = set;
 882
 883         if (ifp->if_capenable != set) {
 884                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 885                 if (error)
 886                         if_printf(sc->sc_ifp,
 887                             "error setting interface capabilities on %s\n",
 888                             ifp->if_xname);
 889         }
 890 }
 891
 892 /*
 893  * bridge_lookup_member:
 894  *
 895  *      Lookup a bridge member interface.
 896  */
 897 static struct bridge_iflist *
 898 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 899 {
 900         struct bridge_iflist *bif;
 901         struct ifnet *ifp;
 902
 903         BRIDGE_LOCK_ASSERT(sc);
 904
 905         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 906                 ifp = bif->bif_ifp;
 907                 if (strcmp(ifp->if_xname, name) == 0)
 908                         return (bif);
 909         }
 910
 911         return (NULL);
 912 }
 913
 914 /*
 915  * bridge_lookup_member_if:
 916  *
 917  *      Lookup a bridge member interface by ifnet*.
 918  */
 919 static struct bridge_iflist *
 920 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 921 {
 922         struct bridge_iflist *bif;
 923
 924         BRIDGE_LOCK_ASSERT(sc);
 925
 926         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 927                 if (bif->bif_ifp == member_ifp)
 928                         return (bif);
 929         }
 930
 931         return (NULL);
 932 }
 933
 934 /*
 935  * bridge_delete_member:
 936  *
 937  *      Delete the specified member interface.
 938  */
 939 static void
 940 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 941     int gone)
 942 {
 943         struct ifnet *ifs = bif->bif_ifp;
 944         struct ifnet *fif = NULL;
 945
 946         BRIDGE_LOCK_ASSERT(sc);
 947
 948         if (bif->bif_flags & IFBIF_STP)
 949                 bstp_disable(&bif->bif_stp);
 950
 951         ifs->if_bridge = NULL;
 952         BRIDGE_XLOCK(sc);
 953         LIST_REMOVE(bif, bif_next);
 954         BRIDGE_XDROP(sc);
 955
 956         /*
 957          * If removing the interface that gave the bridge its mac address, set
 958          * the mac address of the bridge to the address of the next member, or
 959          * to its default address if no members are left.
 960          */
 961         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
 962                 if (LIST_EMPTY(&sc->sc_iflist)) {
 963                         bcopy(sc->sc_defaddr,
 964                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 965                         sc->sc_ifaddr = NULL;
 966                 } else {
 967                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
 968                         bcopy(IF_LLADDR(fif),
 969                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 970                         sc->sc_ifaddr = fif;
 971                 }
 972                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
 973         }
 974
 975         bridge_linkcheck(sc);
 976         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 977         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 978         KASSERT(bif->bif_addrcnt == 0,
 979             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 980
 981         BRIDGE_UNLOCK(sc);
 982         if (!gone) {
 983                 switch (ifs->if_type) {
 984                 case IFT_ETHER:
 985                 case IFT_L2VLAN:
 986                         /*
 987                          * Take the interface out of promiscuous mode.
 988                          */
 989                         (void) ifpromisc(ifs, 0);
 990                         break;
 991
 992                 case IFT_GIF:
 993                         break;
 994
 995                 default:
 996 #ifdef DIAGNOSTIC
 997                         panic("bridge_delete_member: impossible");
 998 #endif
 999                         break;
1000                 }
1001                 /* reneable any interface capabilities */
1002                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1003         }
1004         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1005         BRIDGE_LOCK(sc);
1006         free(bif, M_DEVBUF);
1007 }
1008
1009 /*
1010  * bridge_delete_span:
1011  *
1012  *      Delete the specified span interface.
1013  */
1014 static void
1015 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1016 {
1017         BRIDGE_LOCK_ASSERT(sc);
1018
1019         KASSERT(bif->bif_ifp->if_bridge == NULL,
1020             ("%s: not a span interface", __func__));
1021
1022         LIST_REMOVE(bif, bif_next);
1023         free(bif, M_DEVBUF);
1024 }
1025
1026 static int
1027 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1028 {
1029         struct ifbreq *req = arg;
1030         struct bridge_iflist *bif = NULL;
1031         struct ifnet *ifs;
1032         int error = 0;
1033
1034         ifs = ifunit(req->ifbr_ifsname);
1035         if (ifs == NULL)
1036                 return (ENOENT);
1037         if (ifs->if_ioctl == NULL)      /* must be supported */
1038                 return (EINVAL);
1039
1040         /* If it's in the span list, it can't be a member. */
1041         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1042                 if (ifs == bif->bif_ifp)
1043                         return (EBUSY);
1044
1045         if (ifs->if_bridge == sc)
1046                 return (EEXIST);
1047
1048         if (ifs->if_bridge != NULL)
1049                 return (EBUSY);
1050
1051         switch (ifs->if_type) {
1052         case IFT_ETHER:
1053         case IFT_L2VLAN:
1054         case IFT_GIF:
1055                 /* permitted interface types */
1056                 break;
1057         default:
1058                 return (EINVAL);
1059         }
1060
1061 #ifdef INET6
1062         /*
1063          * Two valid inet6 addresses with link-local scope must not be
1064          * on the parent interface and the member interfaces at the
1065          * same time.  This restriction is needed to prevent violation
1066          * of link-local scope zone.  Attempts to add a member
1067          * interface which has inet6 addresses when the parent has
1068          * inet6 triggers removal of all inet6 addresses on the member
1069          * interface.
1070          */
1071
1072         /* Check if the parent interface has a link-local scope addr. */
1073         if (V_allow_llz_overlap == 0 &&
1074             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1075                 /*
1076                  * If any, remove all inet6 addresses from the member
1077                  * interfaces.
1078                  */
1079                 BRIDGE_XLOCK(sc);
1080                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1081                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1082                                 BRIDGE_UNLOCK(sc);
1083                                 in6_ifdetach(bif->bif_ifp);
1084                                 BRIDGE_LOCK(sc);
1085                                 if_printf(sc->sc_ifp,
1086                                     "IPv6 addresses on %s have been removed "
1087                                     "before adding it as a member to prevent "
1088                                     "IPv6 address scope violation.\n",
1089                                     bif->bif_ifp->if_xname);
1090                         }
1091                 }
1092                 BRIDGE_XDROP(sc);
1093                 if (in6ifa_llaonifp(ifs)) {
1094                         BRIDGE_UNLOCK(sc);
1095                         in6_ifdetach(ifs);
1096                         BRIDGE_LOCK(sc);
1097                         if_printf(sc->sc_ifp,
1098                             "IPv6 addresses on %s have been removed "
1099                             "before adding it as a member to prevent "
1100                             "IPv6 address scope violation.\n",
1101                             ifs->if_xname);
1102                 }
1103         }
1104 #endif
1105         /* Allow the first Ethernet member to define the MTU */
1106         if (LIST_EMPTY(&sc->sc_iflist))
1107                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1108         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1109                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1110                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1111                 return (EINVAL);
1112         }
1113
1114         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1115         if (bif == NULL)
1116                 return (ENOMEM);
1117
1118         bif->bif_ifp = ifs;
1119         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1120         bif->bif_savedcaps = ifs->if_capenable;
1121
1122         /*
1123          * Assign the interface's MAC address to the bridge if it's the first
1124          * member and the MAC address of the bridge has not been changed from
1125          * the default randomly generated one.
1126          */
1127         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1128             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1129                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1130                 sc->sc_ifaddr = ifs;
1131                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1132         }
1133
1134         ifs->if_bridge = sc;
1135         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1136         /*
1137          * XXX: XLOCK HERE!?!
1138          *
1139          * NOTE: insert_***HEAD*** should be safe for the traversals.
1140          */
1141         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1142
1143         /* Set interface capabilities to the intersection set of all members */
1144         bridge_mutecaps(sc);
1145         bridge_linkcheck(sc);
1146
1147         /* Place the interface into promiscuous mode */
1148         switch (ifs->if_type) {
1149                 case IFT_ETHER:
1150                 case IFT_L2VLAN:
1151                         BRIDGE_UNLOCK(sc);
1152                         error = ifpromisc(ifs, 1);
1153                         BRIDGE_LOCK(sc);
1154                         break;
1155         }
1156
1157         if (error) {
1158                 bridge_delete_member(sc, bif, 0);
1159                 free(bif, M_DEVBUF);
1160         }
1161         return (error);
1162 }
1163
1164 static int
1165 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1166 {
1167         struct ifbreq *req = arg;
1168         struct bridge_iflist *bif;
1169
1170         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1171         if (bif == NULL)
1172                 return (ENOENT);
1173
1174         bridge_delete_member(sc, bif, 0);
1175
1176         return (0);
1177 }
1178
1179 static int
1180 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1181 {
1182         struct ifbreq *req = arg;
1183         struct bridge_iflist *bif;
1184         struct bstp_port *bp;
1185
1186         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1187         if (bif == NULL)
1188                 return (ENOENT);
1189
1190         bp = &bif->bif_stp;
1191         req->ifbr_ifsflags = bif->bif_flags;
1192         req->ifbr_state = bp->bp_state;
1193         req->ifbr_priority = bp->bp_priority;
1194         req->ifbr_path_cost = bp->bp_path_cost;
1195         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1196         req->ifbr_proto = bp->bp_protover;
1197         req->ifbr_role = bp->bp_role;
1198         req->ifbr_stpflags = bp->bp_flags;
1199         req->ifbr_addrcnt = bif->bif_addrcnt;
1200         req->ifbr_addrmax = bif->bif_addrmax;
1201         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1202
1203         /* Copy STP state options as flags */
1204         if (bp->bp_operedge)
1205                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1206         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1207                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1208         if (bp->bp_ptp_link)
1209                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1210         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1211                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1212         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1213                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1214         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1215                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1216         return (0);
1217 }
1218
1219 static int
1220 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1221 {
1222         struct ifbreq *req = arg;
1223         struct bridge_iflist *bif;
1224         struct bstp_port *bp;
1225         int error;
1226
1227         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1228         if (bif == NULL)
1229                 return (ENOENT);
1230         bp = &bif->bif_stp;
1231
1232         if (req->ifbr_ifsflags & IFBIF_SPAN)
1233                 /* SPAN is readonly */
1234                 return (EINVAL);
1235
1236         if (req->ifbr_ifsflags & IFBIF_STP) {
1237                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1238                         error = bstp_enable(&bif->bif_stp);
1239                         if (error)
1240                                 return (error);
1241                 }
1242         } else {
1243                 if ((bif->bif_flags & IFBIF_STP) != 0)
1244                         bstp_disable(&bif->bif_stp);
1245         }
1246
1247         /* Pass on STP flags */
1248         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1249         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1250         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1251         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1252
1253         /* Save the bits relating to the bridge */
1254         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1255
1256         return (0);
1257 }
1258
1259 static int
1260 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1261 {
1262         struct ifbrparam *param = arg;
1263
1264         sc->sc_brtmax = param->ifbrp_csize;
1265         bridge_rttrim(sc);
1266
1267         return (0);
1268 }
1269
1270 static int
1271 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1272 {
1273         struct ifbrparam *param = arg;
1274
1275         param->ifbrp_csize = sc->sc_brtmax;
1276
1277         return (0);
1278 }
1279
1280 static int
1281 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1282 {
1283         struct ifbifconf *bifc = arg;
1284         struct bridge_iflist *bif;
1285         struct ifbreq breq;
1286         char *buf, *outbuf;
1287         int count, buflen, len, error = 0;
1288
1289         count = 0;
1290         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1291                 count++;
1292         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1293                 count++;
1294
1295         buflen = sizeof(breq) * count;
1296         if (bifc->ifbic_len == 0) {
1297                 bifc->ifbic_len = buflen;
1298                 return (0);
1299         }
1300         BRIDGE_UNLOCK(sc);
1301         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1302         BRIDGE_LOCK(sc);
1303
1304         count = 0;
1305         buf = outbuf;
1306         len = min(bifc->ifbic_len, buflen);
1307         bzero(&breq, sizeof(breq));
1308         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1309                 if (len < sizeof(breq))
1310                         break;
1311
1312                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1313                     sizeof(breq.ifbr_ifsname));
1314                 /* Fill in the ifbreq structure */
1315                 error = bridge_ioctl_gifflags(sc, &breq);
1316                 if (error)
1317                         break;
1318                 memcpy(buf, &breq, sizeof(breq));
1319                 count++;
1320                 buf += sizeof(breq);
1321                 len -= sizeof(breq);
1322         }
1323         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1324                 if (len < sizeof(breq))
1325                         break;
1326
1327                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1328                     sizeof(breq.ifbr_ifsname));
1329                 breq.ifbr_ifsflags = bif->bif_flags;
1330                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1331                 memcpy(buf, &breq, sizeof(breq));
1332                 count++;
1333                 buf += sizeof(breq);
1334                 len -= sizeof(breq);
1335         }
1336
1337         BRIDGE_UNLOCK(sc);
1338         bifc->ifbic_len = sizeof(breq) * count;
1339         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1340         BRIDGE_LOCK(sc);
1341         free(outbuf, M_TEMP);
1342         return (error);
1343 }
1344
1345 static int
1346 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1347 {
1348         struct ifbaconf *bac = arg;
1349         struct bridge_rtnode *brt;
1350         struct ifbareq bareq;
1351         char *buf, *outbuf;
1352         int count, buflen, len, error = 0;
1353
1354         if (bac->ifbac_len == 0)
1355                 return (0);
1356
1357         count = 0;
1358         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1359                 count++;
1360         buflen = sizeof(bareq) * count;
1361
1362         BRIDGE_UNLOCK(sc);
1363         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1364         BRIDGE_LOCK(sc);
1365
1366         count = 0;
1367         buf = outbuf;
1368         len = min(bac->ifbac_len, buflen);
1369         bzero(&bareq, sizeof(bareq));
1370         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1371                 if (len < sizeof(bareq))
1372                         goto out;
1373                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1374                     sizeof(bareq.ifba_ifsname));
1375                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1376                 bareq.ifba_vlan = brt->brt_vlan;
1377                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1378                                 time_uptime < brt->brt_expire)
1379                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1380                 else
1381                         bareq.ifba_expire = 0;
1382                 bareq.ifba_flags = brt->brt_flags;
1383
1384                 memcpy(buf, &bareq, sizeof(bareq));
1385                 count++;
1386                 buf += sizeof(bareq);
1387                 len -= sizeof(bareq);
1388         }
1389 out:
1390         BRIDGE_UNLOCK(sc);
1391         bac->ifbac_len = sizeof(bareq) * count;
1392         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1393         BRIDGE_LOCK(sc);
1394         free(outbuf, M_TEMP);
1395         return (error);
1396 }
1397
1398 static int
1399 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1400 {
1401         struct ifbareq *req = arg;
1402         struct bridge_iflist *bif;
1403         int error;
1404
1405         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1406         if (bif == NULL)
1407                 return (ENOENT);
1408
1409         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1410             req->ifba_flags);
1411
1412         return (error);
1413 }
1414
1415 static int
1416 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1417 {
1418         struct ifbrparam *param = arg;
1419
1420         sc->sc_brttimeout = param->ifbrp_ctime;
1421         return (0);
1422 }
1423
1424 static int
1425 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1426 {
1427         struct ifbrparam *param = arg;
1428
1429         param->ifbrp_ctime = sc->sc_brttimeout;
1430         return (0);
1431 }
1432
1433 static int
1434 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1435 {
1436         struct ifbareq *req = arg;
1437
1438         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1439 }
1440
1441 static int
1442 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1443 {
1444         struct ifbreq *req = arg;
1445
1446         bridge_rtflush(sc, req->ifbr_ifsflags);
1447         return (0);
1448 }
1449
1450 static int
1451 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1452 {
1453         struct ifbrparam *param = arg;
1454         struct bstp_state *bs = &sc->sc_stp;
1455
1456         param->ifbrp_prio = bs->bs_bridge_priority;
1457         return (0);
1458 }
1459
1460 static int
1461 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1462 {
1463         struct ifbrparam *param = arg;
1464
1465         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1466 }
1467
1468 static int
1469 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1470 {
1471         struct ifbrparam *param = arg;
1472         struct bstp_state *bs = &sc->sc_stp;
1473
1474         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1475         return (0);
1476 }
1477
1478 static int
1479 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1480 {
1481         struct ifbrparam *param = arg;
1482
1483         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1484 }
1485
1486 static int
1487 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1488 {
1489         struct ifbrparam *param = arg;
1490         struct bstp_state *bs = &sc->sc_stp;
1491
1492         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1493         return (0);
1494 }
1495
1496 static int
1497 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1498 {
1499         struct ifbrparam *param = arg;
1500
1501         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1502 }
1503
1504 static int
1505 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1506 {
1507         struct ifbrparam *param = arg;
1508         struct bstp_state *bs = &sc->sc_stp;
1509
1510         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1511         return (0);
1512 }
1513
1514 static int
1515 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1516 {
1517         struct ifbrparam *param = arg;
1518
1519         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1520 }
1521
1522 static int
1523 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1524 {
1525         struct ifbreq *req = arg;
1526         struct bridge_iflist *bif;
1527
1528         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1529         if (bif == NULL)
1530                 return (ENOENT);
1531
1532         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1533 }
1534
1535 static int
1536 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1537 {
1538         struct ifbreq *req = arg;
1539         struct bridge_iflist *bif;
1540
1541         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1542         if (bif == NULL)
1543                 return (ENOENT);
1544
1545         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1546 }
1547
1548 static int
1549 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1550 {
1551         struct ifbreq *req = arg;
1552         struct bridge_iflist *bif;
1553
1554         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1555         if (bif == NULL)
1556                 return (ENOENT);
1557
1558         bif->bif_addrmax = req->ifbr_addrmax;
1559         return (0);
1560 }
1561
1562 static int
1563 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1564 {
1565         struct ifbreq *req = arg;
1566         struct bridge_iflist *bif = NULL;
1567         struct ifnet *ifs;
1568
1569         ifs = ifunit(req->ifbr_ifsname);
1570         if (ifs == NULL)
1571                 return (ENOENT);
1572
1573         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1574                 if (ifs == bif->bif_ifp)
1575                         return (EBUSY);
1576
1577         if (ifs->if_bridge != NULL)
1578                 return (EBUSY);
1579
1580         switch (ifs->if_type) {
1581                 case IFT_ETHER:
1582                 case IFT_GIF:
1583                 case IFT_L2VLAN:
1584                         break;
1585                 default:
1586                         return (EINVAL);
1587         }
1588
1589         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1590         if (bif == NULL)
1591                 return (ENOMEM);
1592
1593         bif->bif_ifp = ifs;
1594         bif->bif_flags = IFBIF_SPAN;
1595
1596         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1597
1598         return (0);
1599 }
1600
1601 static int
1602 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1603 {
1604         struct ifbreq *req = arg;
1605         struct bridge_iflist *bif;
1606         struct ifnet *ifs;
1607
1608         ifs = ifunit(req->ifbr_ifsname);
1609         if (ifs == NULL)
1610                 return (ENOENT);
1611
1612         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1613                 if (ifs == bif->bif_ifp)
1614                         break;
1615
1616         if (bif == NULL)
1617                 return (ENOENT);
1618
1619         bridge_delete_span(sc, bif);
1620
1621         return (0);
1622 }
1623
1624 static int
1625 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1626 {
1627         struct ifbropreq *req = arg;
1628         struct bstp_state *bs = &sc->sc_stp;
1629         struct bstp_port *root_port;
1630
1631         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1632         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1633         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1634
1635         root_port = bs->bs_root_port;
1636         if (root_port == NULL)
1637                 req->ifbop_root_port = 0;
1638         else
1639                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1640
1641         req->ifbop_holdcount = bs->bs_txholdcount;
1642         req->ifbop_priority = bs->bs_bridge_priority;
1643         req->ifbop_protocol = bs->bs_protover;
1644         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1645         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1646         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1647         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1648         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1649         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1650
1651         return (0);
1652 }
1653
1654 static int
1655 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1656 {
1657         struct ifbrparam *param = arg;
1658
1659         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1660         return (0);
1661 }
1662
1663 static int
1664 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1665 {
1666         struct ifbpstpconf *bifstp = arg;
1667         struct bridge_iflist *bif;
1668         struct bstp_port *bp;
1669         struct ifbpstpreq bpreq;
1670         char *buf, *outbuf;
1671         int count, buflen, len, error = 0;
1672
1673         count = 0;
1674         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1675                 if ((bif->bif_flags & IFBIF_STP) != 0)
1676                         count++;
1677         }
1678
1679         buflen = sizeof(bpreq) * count;
1680         if (bifstp->ifbpstp_len == 0) {
1681                 bifstp->ifbpstp_len = buflen;
1682                 return (0);
1683         }
1684
1685         BRIDGE_UNLOCK(sc);
1686         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1687         BRIDGE_LOCK(sc);
1688
1689         count = 0;
1690         buf = outbuf;
1691         len = min(bifstp->ifbpstp_len, buflen);
1692         bzero(&bpreq, sizeof(bpreq));
1693         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1694                 if (len < sizeof(bpreq))
1695                         break;
1696
1697                 if ((bif->bif_flags & IFBIF_STP) == 0)
1698                         continue;
1699
1700                 bp = &bif->bif_stp;
1701                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1702                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1703                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1704                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1705                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1706                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1707
1708                 memcpy(buf, &bpreq, sizeof(bpreq));
1709                 count++;
1710                 buf += sizeof(bpreq);
1711                 len -= sizeof(bpreq);
1712         }
1713
1714         BRIDGE_UNLOCK(sc);
1715         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1716         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1717         BRIDGE_LOCK(sc);
1718         free(outbuf, M_TEMP);
1719         return (error);
1720 }
1721
1722 static int
1723 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1724 {
1725         struct ifbrparam *param = arg;
1726
1727         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1728 }
1729
1730 static int
1731 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1732 {
1733         struct ifbrparam *param = arg;
1734
1735         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1736 }
1737
1738 /*
1739  * bridge_ifdetach:
1740  *
1741  *      Detach an interface from a bridge.  Called when a member
1742  *      interface is detaching.
1743  */
1744 static void
1745 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1746 {
1747         struct bridge_softc *sc = ifp->if_bridge;
1748         struct bridge_iflist *bif;
1749
1750         if (ifp->if_flags & IFF_RENAMING)
1751                 return;
1752
1753         /* Check if the interface is a bridge member */
1754         if (sc != NULL) {
1755                 BRIDGE_LOCK(sc);
1756
1757                 bif = bridge_lookup_member_if(sc, ifp);
1758                 if (bif != NULL)
1759                         bridge_delete_member(sc, bif, 1);
1760
1761                 BRIDGE_UNLOCK(sc);
1762                 return;
1763         }
1764
1765         /* Check if the interface is a span port */
1766         mtx_lock(&bridge_list_mtx);
1767         LIST_FOREACH(sc, &bridge_list, sc_list) {
1768                 BRIDGE_LOCK(sc);
1769                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1770                         if (ifp == bif->bif_ifp) {
1771                                 bridge_delete_span(sc, bif);
1772                                 break;
1773                         }
1774
1775                 BRIDGE_UNLOCK(sc);
1776         }
1777         mtx_unlock(&bridge_list_mtx);
1778 }
1779
1780 /*
1781  * bridge_init:
1782  *
1783  *      Initialize a bridge interface.
1784  */
1785 static void
1786 bridge_init(void *xsc)
1787 {
1788         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1789         struct ifnet *ifp = sc->sc_ifp;
1790
1791         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1792                 return;
1793
1794         BRIDGE_LOCK(sc);
1795         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1796             bridge_timer, sc);
1797
1798         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1799         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1800
1801         BRIDGE_UNLOCK(sc);
1802 }
1803
1804 /*
1805  * bridge_stop:
1806  *
1807  *      Stop the bridge interface.
1808  */
1809 static void
1810 bridge_stop(struct ifnet *ifp, int disable)
1811 {
1812         struct bridge_softc *sc = ifp->if_softc;
1813
1814         BRIDGE_LOCK_ASSERT(sc);
1815
1816         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1817                 return;
1818
1819         callout_stop(&sc->sc_brcallout);
1820         bstp_stop(&sc->sc_stp);
1821
1822         bridge_rtflush(sc, IFBF_FLUSHDYN);
1823
1824         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1825 }
1826
1827 /*
1828  * bridge_enqueue:
1829  *
1830  *      Enqueue a packet on a bridge member interface.
1831  *
1832  */
1833 static void
1834 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1835 {
1836         int len, err = 0;
1837         short mflags;
1838         struct mbuf *m0;
1839
1840         len = m->m_pkthdr.len;
1841         mflags = m->m_flags;
1842
1843         /* We may be sending a fragment so traverse the mbuf */
1844         for (; m; m = m0) {
1845                 m0 = m->m_nextpkt;
1846                 m->m_nextpkt = NULL;
1847
1848                 /*
1849                  * If underlying interface can not do VLAN tag insertion itself
1850                  * then attach a packet tag that holds it.
1851                  */
1852                 if ((m->m_flags & M_VLANTAG) &&
1853                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1854                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1855                         if (m == NULL) {
1856                                 if_printf(dst_ifp,
1857                                     "unable to prepend VLAN header\n");
1858                                 dst_ifp->if_oerrors++;
1859                                 continue;
1860                         }
1861                         m->m_flags &= ~M_VLANTAG;
1862                 }
1863
1864                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1865                         m_freem(m0);
1866                         break;
1867                 }
1868         }
1869
1870         if (err == 0) {
1871                 sc->sc_ifp->if_opackets++;
1872                 sc->sc_ifp->if_obytes += len;
1873                 if (mflags & M_MCAST)
1874                         sc->sc_ifp->if_omcasts++;
1875         }
1876 }
1877
1878 /*
1879  * bridge_dummynet:
1880  *
1881  *      Receive a queued packet from dummynet and pass it on to the output
1882  *      interface.
1883  *
1884  *      The mbuf has the Ethernet header already attached.
1885  */
1886 static void
1887 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1888 {
1889         struct bridge_softc *sc;
1890
1891         sc = ifp->if_bridge;
1892
1893         /*
1894          * The packet didnt originate from a member interface. This should only
1895          * ever happen if a member interface is removed while packets are
1896          * queued for it.
1897          */
1898         if (sc == NULL) {
1899                 m_freem(m);
1900                 return;
1901         }
1902
1903         if (PFIL_HOOKED(&V_inet_pfil_hook)
1904 #ifdef INET6
1905             || PFIL_HOOKED(&V_inet6_pfil_hook)
1906 #endif
1907             ) {
1908                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1909                         return;
1910                 if (m == NULL)
1911                         return;
1912         }
1913
1914         bridge_enqueue(sc, ifp, m);
1915 }
1916
1917 /*
1918  * bridge_output:
1919  *
1920  *      Send output from a bridge member interface.  This
1921  *      performs the bridging function for locally originated
1922  *      packets.
1923  *
1924  *      The mbuf has the Ethernet header already attached.  We must
1925  *      enqueue or free the mbuf before returning.
1926  */
1927 static int
1928 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1929     struct rtentry *rt)
1930 {
1931         struct ether_header *eh;
1932         struct ifnet *dst_if;
1933         struct bridge_softc *sc;
1934         uint16_t vlan;
1935
1936         if (m->m_len < ETHER_HDR_LEN) {
1937                 m = m_pullup(m, ETHER_HDR_LEN);
1938                 if (m == NULL)
1939                         return (0);
1940         }
1941
1942         eh = mtod(m, struct ether_header *);
1943         sc = ifp->if_bridge;
1944         vlan = VLANTAGOF(m);
1945
1946         BRIDGE_LOCK(sc);
1947
1948         /*
1949          * If bridge is down, but the original output interface is up,
1950          * go ahead and send out that interface.  Otherwise, the packet
1951          * is dropped below.
1952          */
1953         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1954                 dst_if = ifp;
1955                 goto sendunicast;
1956         }
1957
1958         /*
1959          * If the packet is a multicast, or we don't know a better way to
1960          * get there, send to all interfaces.
1961          */
1962         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1963                 dst_if = NULL;
1964         else
1965                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1966         if (dst_if == NULL) {
1967                 struct bridge_iflist *bif;
1968                 struct mbuf *mc;
1969                 int error = 0, used = 0;
1970
1971                 bridge_span(sc, m);
1972
1973                 BRIDGE_LOCK2REF(sc, error);
1974                 if (error) {
1975                         m_freem(m);
1976                         return (0);
1977                 }
1978
1979                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1980                         dst_if = bif->bif_ifp;
1981
1982                         if (dst_if->if_type == IFT_GIF)
1983                                 continue;
1984                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1985                                 continue;
1986
1987                         /*
1988                          * If this is not the original output interface,
1989                          * and the interface is participating in spanning
1990                          * tree, make sure the port is in a state that
1991                          * allows forwarding.
1992                          */
1993                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1994                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1995                                 continue;
1996
1997                         if (LIST_NEXT(bif, bif_next) == NULL) {
1998                                 used = 1;
1999                                 mc = m;
2000                         } else {
2001                                 mc = m_copypacket(m, M_DONTWAIT);
2002                                 if (mc == NULL) {
2003                                         sc->sc_ifp->if_oerrors++;
2004                                         continue;
2005                                 }
2006                         }
2007
2008                         bridge_enqueue(sc, dst_if, mc);
2009                 }
2010                 if (used == 0)
2011                         m_freem(m);
2012                 BRIDGE_UNREF(sc);
2013                 return (0);
2014         }
2015
2016 sendunicast:
2017         /*
2018          * XXX Spanning tree consideration here?
2019          */
2020
2021         bridge_span(sc, m);
2022         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2023                 m_freem(m);
2024                 BRIDGE_UNLOCK(sc);
2025                 return (0);
2026         }
2027
2028         BRIDGE_UNLOCK(sc);
2029         bridge_enqueue(sc, dst_if, m);
2030         return (0);
2031 }
2032
2033 /*
2034  * bridge_start:
2035  *
2036  *      Start output on a bridge.
2037  *
2038  */
2039 static void
2040 bridge_start(struct ifnet *ifp)
2041 {
2042         struct bridge_softc *sc;
2043         struct mbuf *m;
2044         struct ether_header *eh;
2045         struct ifnet *dst_if;
2046
2047         sc = ifp->if_softc;
2048
2049         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
2050         for (;;) {
2051                 IFQ_DEQUEUE(&ifp->if_snd, m);
2052                 if (m == 0)
2053                         break;
2054                 ETHER_BPF_MTAP(ifp, m);
2055
2056                 eh = mtod(m, struct ether_header *);
2057                 dst_if = NULL;
2058
2059                 BRIDGE_LOCK(sc);
2060                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2061                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
2062                 }
2063
2064                 if (dst_if == NULL)
2065                         bridge_broadcast(sc, ifp, m, 0);
2066                 else {
2067                         BRIDGE_UNLOCK(sc);
2068                         bridge_enqueue(sc, dst_if, m);
2069                 }
2070         }
2071         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2072 }
2073
2074 /*
2075  * bridge_forward:
2076  *
2077  *      The forwarding function of the bridge.
2078  *
2079  *      NOTE: Releases the lock on return.
2080  */
2081 static void
2082 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2083     struct mbuf *m)
2084 {
2085         struct bridge_iflist *dbif;
2086         struct ifnet *src_if, *dst_if, *ifp;
2087         struct ether_header *eh;
2088         uint16_t vlan;
2089         uint8_t *dst;
2090         int error;
2091
2092         src_if = m->m_pkthdr.rcvif;
2093         ifp = sc->sc_ifp;
2094
2095         ifp->if_ipackets++;
2096         ifp->if_ibytes += m->m_pkthdr.len;
2097         vlan = VLANTAGOF(m);
2098
2099         if ((sbif->bif_flags & IFBIF_STP) &&
2100             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2101                 goto drop;
2102
2103         eh = mtod(m, struct ether_header *);
2104         dst = eh->ether_dhost;
2105
2106         /* If the interface is learning, record the address. */
2107         if (sbif->bif_flags & IFBIF_LEARNING) {
2108                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2109                     sbif, 0, IFBAF_DYNAMIC);
2110                 /*
2111                  * If the interface has addresses limits then deny any source
2112                  * that is not in the cache.
2113                  */
2114                 if (error && sbif->bif_addrmax)
2115                         goto drop;
2116         }
2117
2118         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2119             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2120                 goto drop;
2121
2122         /*
2123          * At this point, the port either doesn't participate
2124          * in spanning tree or it is in the forwarding state.
2125          */
2126
2127         /*
2128          * If the packet is unicast, destined for someone on
2129          * "this" side of the bridge, drop it.
2130          */
2131         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2132                 dst_if = bridge_rtlookup(sc, dst, vlan);
2133                 if (src_if == dst_if)
2134                         goto drop;
2135         } else {
2136                 /*
2137                  * Check if its a reserved multicast address, any address
2138                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2139                  * bridge.
2140                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2141                  */
2142                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2143                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2144                     dst[4] == 0x00 && dst[5] <= 0x0f)
2145                         goto drop;
2146
2147                 /* ...forward it to all interfaces. */
2148                 ifp->if_imcasts++;
2149                 dst_if = NULL;
2150         }
2151
2152         /*
2153          * If we have a destination interface which is a member of our bridge,
2154          * OR this is a unicast packet, push it through the bpf(4) machinery.
2155          * For broadcast or multicast packets, don't bother because it will
2156          * be reinjected into ether_input. We do this before we pass the packets
2157          * through the pfil(9) framework, as it is possible that pfil(9) will
2158          * drop the packet, or possibly modify it, making it difficult to debug
2159          * firewall issues on the bridge.
2160          */
2161         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2162                 ETHER_BPF_MTAP(ifp, m);
2163
2164         /* run the packet filter */
2165         if (PFIL_HOOKED(&V_inet_pfil_hook)
2166 #ifdef INET6
2167             || PFIL_HOOKED(&V_inet6_pfil_hook)
2168 #endif
2169             ) {
2170                 BRIDGE_UNLOCK(sc);
2171                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2172                         return;
2173                 if (m == NULL)
2174                         return;
2175                 BRIDGE_LOCK(sc);
2176         }
2177
2178         if (dst_if == NULL) {
2179                 bridge_broadcast(sc, src_if, m, 1);
2180                 return;
2181         }
2182
2183         /*
2184          * At this point, we're dealing with a unicast frame
2185          * going to a different interface.
2186          */
2187         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2188                 goto drop;
2189
2190         dbif = bridge_lookup_member_if(sc, dst_if);
2191         if (dbif == NULL)
2192                 /* Not a member of the bridge (anymore?) */
2193                 goto drop;
2194
2195         /* Private segments can not talk to each other */
2196         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2197                 goto drop;
2198
2199         if ((dbif->bif_flags & IFBIF_STP) &&
2200             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2201                 goto drop;
2202
2203         BRIDGE_UNLOCK(sc);
2204
2205         if (PFIL_HOOKED(&V_inet_pfil_hook)
2206 #ifdef INET6
2207             || PFIL_HOOKED(&V_inet6_pfil_hook)
2208 #endif
2209             ) {
2210                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2211                         return;
2212                 if (m == NULL)
2213                         return;
2214         }
2215
2216         bridge_enqueue(sc, dst_if, m);
2217         return;
2218
2219 drop:
2220         BRIDGE_UNLOCK(sc);
2221         m_freem(m);
2222 }
2223
2224 /*
2225  * bridge_input:
2226  *
2227  *      Receive input from a member interface.  Queue the packet for
2228  *      bridging if it is not for us.
2229  */
2230 static struct mbuf *
2231 bridge_input(struct ifnet *ifp, struct mbuf *m)
2232 {
2233         struct bridge_softc *sc = ifp->if_bridge;
2234         struct bridge_iflist *bif, *bif2;
2235         struct ifnet *bifp;
2236         struct ether_header *eh;
2237         struct mbuf *mc, *mc2;
2238         uint16_t vlan;
2239         int error;
2240
2241         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2242                 return (m);
2243
2244         bifp = sc->sc_ifp;
2245         vlan = VLANTAGOF(m);
2246
2247         /*
2248          * Implement support for bridge monitoring. If this flag has been
2249          * set on this interface, discard the packet once we push it through
2250          * the bpf(4) machinery, but before we do, increment the byte and
2251          * packet counters associated with this interface.
2252          */
2253         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2254                 m->m_pkthdr.rcvif  = bifp;
2255                 ETHER_BPF_MTAP(bifp, m);
2256                 bifp->if_ipackets++;
2257                 bifp->if_ibytes += m->m_pkthdr.len;
2258                 m_freem(m);
2259                 return (NULL);
2260         }
2261         BRIDGE_LOCK(sc);
2262         bif = bridge_lookup_member_if(sc, ifp);
2263         if (bif == NULL) {
2264                 BRIDGE_UNLOCK(sc);
2265                 return (m);
2266         }
2267
2268         eh = mtod(m, struct ether_header *);
2269
2270         bridge_span(sc, m);
2271
2272         if (m->m_flags & (M_BCAST|M_MCAST)) {
2273                 /* Tap off 802.1D packets; they do not get forwarded. */
2274                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2275                     ETHER_ADDR_LEN) == 0) {
2276                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2277                         BRIDGE_UNLOCK(sc);
2278                         return (NULL);
2279                 }
2280
2281                 if ((bif->bif_flags & IFBIF_STP) &&
2282                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2283                         BRIDGE_UNLOCK(sc);
2284                         return (m);
2285                 }
2286
2287                 /*
2288                  * Make a deep copy of the packet and enqueue the copy
2289                  * for bridge processing; return the original packet for
2290                  * local processing.
2291                  */
2292                 mc = m_dup(m, M_DONTWAIT);
2293                 if (mc == NULL) {
2294                         BRIDGE_UNLOCK(sc);
2295                         return (m);
2296                 }
2297
2298                 /* Perform the bridge forwarding function with the copy. */
2299                 bridge_forward(sc, bif, mc);
2300
2301                 /*
2302                  * Reinject the mbuf as arriving on the bridge so we have a
2303                  * chance at claiming multicast packets. We can not loop back
2304                  * here from ether_input as a bridge is never a member of a
2305                  * bridge.
2306                  */
2307                 KASSERT(bifp->if_bridge == NULL,
2308                     ("loop created in bridge_input"));
2309                 mc2 = m_dup(m, M_DONTWAIT);
2310                 if (mc2 != NULL) {
2311                         /* Keep the layer3 header aligned */
2312                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2313                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2314                 }
2315                 if (mc2 != NULL) {
2316                         mc2->m_pkthdr.rcvif = bifp;
2317                         (*bifp->if_input)(bifp, mc2);
2318                 }
2319
2320                 /* Return the original packet for local processing. */
2321                 return (m);
2322         }
2323
2324         if ((bif->bif_flags & IFBIF_STP) &&
2325             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2326                 BRIDGE_UNLOCK(sc);
2327                 return (m);
2328         }
2329
2330 #if (defined(INET) || defined(INET6))
2331 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2332         || ((iface)->if_carp \
2333             && (*carp_forus_p)((iface), eh->ether_dhost))
2334 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2335         || ((iface)->if_carp \
2336             && (*carp_forus_p)((iface), eh->ether_shost))
2337 #else
2338 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2339 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2340 #endif
2341
2342 #ifdef INET6
2343 #   define OR_PFIL_HOOKED_INET6 \
2344         || PFIL_HOOKED(&V_inet6_pfil_hook)
2345 #else
2346 #   define OR_PFIL_HOOKED_INET6
2347 #endif
2348
2349 #define GRAB_OUR_PACKETS(iface) \
2350         if ((iface)->if_type == IFT_GIF) \
2351                 continue; \
2352         /* It is destined for us. */ \
2353         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2354             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2355             ) {                                                         \
2356                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2357                         ETHER_BPF_MTAP(iface, m);                       \
2358                         iface->if_ipackets++;                           \
2359                         /* Filter on the physical interface. */         \
2360                         if (pfil_local_phys &&                          \
2361                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2362                              OR_PFIL_HOOKED_INET6)) {                   \
2363                                 if (bridge_pfil(&m, NULL, ifp,          \
2364                                     PFIL_IN) != 0 || m == NULL) {       \
2365                                         BRIDGE_UNLOCK(sc);              \
2366                                         return (NULL);                  \
2367                                 }                                       \
2368                                 eh = mtod(m, struct ether_header *);    \
2369                         }                                               \
2370                 }                                                       \
2371                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2372                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2373                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2374                         if (error && bif->bif_addrmax) {                \
2375                                 BRIDGE_UNLOCK(sc);                      \
2376                                 m_freem(m);                             \
2377                                 return (NULL);                          \
2378                         }                                               \
2379                 }                                                       \
2380                 m->m_pkthdr.rcvif = iface;                              \
2381                 BRIDGE_UNLOCK(sc);                                      \
2382                 return (m);                                             \
2383         }                                                               \
2384                                                                         \
2385         /* We just received a packet that we sent out. */               \
2386         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2387             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2388             ) {                                                         \
2389                 BRIDGE_UNLOCK(sc);                                      \
2390                 m_freem(m);                                             \
2391                 return (NULL);                                          \
2392         }
2393
2394         /*
2395          * Unicast.  Make sure it's not for the bridge.
2396          */
2397         do { GRAB_OUR_PACKETS(bifp) } while (0);
2398
2399         /*
2400          * Give a chance for ifp at first priority. This will help when the
2401          * packet comes through the interface like VLAN's with the same MACs
2402          * on several interfaces from the same bridge. This also will save
2403          * some CPU cycles in case the destination interface and the input
2404          * interface (eq ifp) are the same.
2405          */
2406         do { GRAB_OUR_PACKETS(ifp) } while (0);
2407
2408         /* Now check the all bridge members. */
2409         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2410                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2411         }
2412
2413 #undef OR_CARP_CHECK_WE_ARE_DST
2414 #undef OR_CARP_CHECK_WE_ARE_SRC
2415 #undef OR_PFIL_HOOKED_INET6
2416 #undef GRAB_OUR_PACKETS
2417
2418         /* Perform the bridge forwarding function. */
2419         bridge_forward(sc, bif, m);
2420
2421         return (NULL);
2422 }
2423
2424 /*
2425  * bridge_broadcast:
2426  *
2427  *      Send a frame to all interfaces that are members of
2428  *      the bridge, except for the one on which the packet
2429  *      arrived.
2430  *
2431  *      NOTE: Releases the lock on return.
2432  */
2433 static void
2434 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2435     struct mbuf *m, int runfilt)
2436 {
2437         struct bridge_iflist *dbif, *sbif;
2438         struct mbuf *mc;
2439         struct ifnet *dst_if;
2440         int error = 0, used = 0, i;
2441
2442         sbif = bridge_lookup_member_if(sc, src_if);
2443
2444         BRIDGE_LOCK2REF(sc, error);
2445         if (error) {
2446                 m_freem(m);
2447                 return;
2448         }
2449
2450         /* Filter on the bridge interface before broadcasting */
2451         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2452 #ifdef INET6
2453             || PFIL_HOOKED(&V_inet6_pfil_hook)
2454 #endif
2455             )) {
2456                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2457                         goto out;
2458                 if (m == NULL)
2459                         goto out;
2460         }
2461
2462         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2463                 dst_if = dbif->bif_ifp;
2464                 if (dst_if == src_if)
2465                         continue;
2466
2467                 /* Private segments can not talk to each other */
2468                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2469                         continue;
2470
2471                 if ((dbif->bif_flags & IFBIF_STP) &&
2472                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2473                         continue;
2474
2475                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2476                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2477                         continue;
2478
2479                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2480                         continue;
2481
2482                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2483                         mc = m;
2484                         used = 1;
2485                 } else {
2486                         mc = m_dup(m, M_DONTWAIT);
2487                         if (mc == NULL) {
2488                                 sc->sc_ifp->if_oerrors++;
2489                                 continue;
2490                         }
2491                 }
2492
2493                 /*
2494                  * Filter on the output interface. Pass a NULL bridge interface
2495                  * pointer so we do not redundantly filter on the bridge for
2496                  * each interface we broadcast on.
2497                  */
2498                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2499 #ifdef INET6
2500                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2501 #endif
2502                     )) {
2503                         if (used == 0) {
2504                                 /* Keep the layer3 header aligned */
2505                                 i = min(mc->m_pkthdr.len, max_protohdr);
2506                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2507                                 if (mc == NULL) {
2508                                         sc->sc_ifp->if_oerrors++;
2509                                         continue;
2510                                 }
2511                         }
2512                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2513                                 continue;
2514                         if (mc == NULL)
2515                                 continue;
2516                 }
2517
2518                 bridge_enqueue(sc, dst_if, mc);
2519         }
2520         if (used == 0)
2521                 m_freem(m);
2522
2523 out:
2524         BRIDGE_UNREF(sc);
2525 }
2526
2527 /*
2528  * bridge_span:
2529  *
2530  *      Duplicate a packet out one or more interfaces that are in span mode,
2531  *      the original mbuf is unmodified.
2532  */
2533 static void
2534 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2535 {
2536         struct bridge_iflist *bif;
2537         struct ifnet *dst_if;
2538         struct mbuf *mc;
2539
2540         if (LIST_EMPTY(&sc->sc_spanlist))
2541                 return;
2542
2543         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2544                 dst_if = bif->bif_ifp;
2545
2546                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2547                         continue;
2548
2549                 mc = m_copypacket(m, M_DONTWAIT);
2550                 if (mc == NULL) {
2551                         sc->sc_ifp->if_oerrors++;
2552                         continue;
2553                 }
2554
2555                 bridge_enqueue(sc, dst_if, mc);
2556         }
2557 }
2558
2559 /*
2560  * bridge_rtupdate:
2561  *
2562  *      Add a bridge routing entry.
2563  */
2564 static int
2565 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2566     struct bridge_iflist *bif, int setflags, uint8_t flags)
2567 {
2568         struct bridge_rtnode *brt;
2569         int error;
2570
2571         BRIDGE_LOCK_ASSERT(sc);
2572
2573         /* Check the source address is valid and not multicast. */
2574         if (ETHER_IS_MULTICAST(dst) ||
2575             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2576              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2577                 return (EINVAL);
2578
2579         /* 802.1p frames map to vlan 1 */
2580         if (vlan == 0)
2581                 vlan = 1;
2582
2583         /*
2584          * A route for this destination might already exist.  If so,
2585          * update it, otherwise create a new one.
2586          */
2587         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2588                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2589                         sc->sc_brtexceeded++;
2590                         return (ENOSPC);
2591                 }
2592                 /* Check per interface address limits (if enabled) */
2593                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2594                         bif->bif_addrexceeded++;
2595                         return (ENOSPC);
2596                 }
2597
2598                 /*
2599                  * Allocate a new bridge forwarding node, and
2600                  * initialize the expiration time and Ethernet
2601                  * address.
2602                  */
2603                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2604                 if (brt == NULL)
2605                         return (ENOMEM);
2606
2607                 if (bif->bif_flags & IFBIF_STICKY)
2608                         brt->brt_flags = IFBAF_STICKY;
2609                 else
2610                         brt->brt_flags = IFBAF_DYNAMIC;
2611
2612                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2613                 brt->brt_vlan = vlan;
2614
2615                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2616                         uma_zfree(bridge_rtnode_zone, brt);
2617                         return (error);
2618                 }
2619                 brt->brt_dst = bif;
2620                 bif->bif_addrcnt++;
2621         }
2622
2623         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2624             brt->brt_dst != bif) {
2625                 brt->brt_dst->bif_addrcnt--;
2626                 brt->brt_dst = bif;
2627                 brt->brt_dst->bif_addrcnt++;
2628         }
2629
2630         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2631                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2632         if (setflags)
2633                 brt->brt_flags = flags;
2634
2635         return (0);
2636 }
2637
2638 /*
2639  * bridge_rtlookup:
2640  *
2641  *      Lookup the destination interface for an address.
2642  */
2643 static struct ifnet *
2644 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2645 {
2646         struct bridge_rtnode *brt;
2647
2648         BRIDGE_LOCK_ASSERT(sc);
2649
2650         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2651                 return (NULL);
2652
2653         return (brt->brt_ifp);
2654 }
2655
2656 /*
2657  * bridge_rttrim:
2658  *
2659  *      Trim the routine table so that we have a number
2660  *      of routing entries less than or equal to the
2661  *      maximum number.
2662  */
2663 static void
2664 bridge_rttrim(struct bridge_softc *sc)
2665 {
2666         struct bridge_rtnode *brt, *nbrt;
2667
2668         BRIDGE_LOCK_ASSERT(sc);
2669
2670         /* Make sure we actually need to do this. */
2671         if (sc->sc_brtcnt <= sc->sc_brtmax)
2672                 return;
2673
2674         /* Force an aging cycle; this might trim enough addresses. */
2675         bridge_rtage(sc);
2676         if (sc->sc_brtcnt <= sc->sc_brtmax)
2677                 return;
2678
2679         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2680                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2681                         bridge_rtnode_destroy(sc, brt);
2682                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2683                                 return;
2684                 }
2685         }
2686 }
2687
2688 /*
2689  * bridge_timer:
2690  *
2691  *      Aging timer for the bridge.
2692  */
2693 static void
2694 bridge_timer(void *arg)
2695 {
2696         struct bridge_softc *sc = arg;
2697
2698         BRIDGE_LOCK_ASSERT(sc);
2699
2700         bridge_rtage(sc);
2701
2702         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2703                 callout_reset(&sc->sc_brcallout,
2704                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2705 }
2706
2707 /*
2708  * bridge_rtage:
2709  *
2710  *      Perform an aging cycle.
2711  */
2712 static void
2713 bridge_rtage(struct bridge_softc *sc)
2714 {
2715         struct bridge_rtnode *brt, *nbrt;
2716
2717         BRIDGE_LOCK_ASSERT(sc);
2718
2719         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2720                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2721                         if (time_uptime >= brt->brt_expire)
2722                                 bridge_rtnode_destroy(sc, brt);
2723                 }
2724         }
2725 }
2726
2727 /*
2728  * bridge_rtflush:
2729  *
2730  *      Remove all dynamic addresses from the bridge.
2731  */
2732 static void
2733 bridge_rtflush(struct bridge_softc *sc, int full)
2734 {
2735         struct bridge_rtnode *brt, *nbrt;
2736
2737         BRIDGE_LOCK_ASSERT(sc);
2738
2739         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2740                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2741                         bridge_rtnode_destroy(sc, brt);
2742         }
2743 }
2744
2745 /*
2746  * bridge_rtdaddr:
2747  *
2748  *      Remove an address from the table.
2749  */
2750 static int
2751 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2752 {
2753         struct bridge_rtnode *brt;
2754         int found = 0;
2755
2756         BRIDGE_LOCK_ASSERT(sc);
2757
2758         /*
2759          * If vlan is zero then we want to delete for all vlans so the lookup
2760          * may return more than one.
2761          */
2762         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2763                 bridge_rtnode_destroy(sc, brt);
2764                 found = 1;
2765         }
2766
2767         return (found ? 0 : ENOENT);
2768 }
2769
2770 /*
2771  * bridge_rtdelete:
2772  *
2773  *      Delete routes to a speicifc member interface.
2774  */
2775 static void
2776 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2777 {
2778         struct bridge_rtnode *brt, *nbrt;
2779
2780         BRIDGE_LOCK_ASSERT(sc);
2781
2782         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2783                 if (brt->brt_ifp == ifp && (full ||
2784                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2785                         bridge_rtnode_destroy(sc, brt);
2786         }
2787 }
2788
2789 /*
2790  * bridge_rtable_init:
2791  *
2792  *      Initialize the route table for this bridge.
2793  */
2794 static int
2795 bridge_rtable_init(struct bridge_softc *sc)
2796 {
2797         int i;
2798
2799         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2800             M_DEVBUF, M_NOWAIT);
2801         if (sc->sc_rthash == NULL)
2802                 return (ENOMEM);
2803
2804         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2805                 LIST_INIT(&sc->sc_rthash[i]);
2806
2807         sc->sc_rthash_key = arc4random();
2808
2809         LIST_INIT(&sc->sc_rtlist);
2810
2811         return (0);
2812 }
2813
2814 /*
2815  * bridge_rtable_fini:
2816  *
2817  *      Deconstruct the route table for this bridge.
2818  */
2819 static void
2820 bridge_rtable_fini(struct bridge_softc *sc)
2821 {
2822
2823         KASSERT(sc->sc_brtcnt == 0,
2824             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2825         free(sc->sc_rthash, M_DEVBUF);
2826 }
2827
2828 /*
2829  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2830  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2831  */
2832 #define mix(a, b, c)                                                    \
2833 do {                                                                    \
2834         a -= b; a -= c; a ^= (c >> 13);                                 \
2835         b -= c; b -= a; b ^= (a << 8);                                  \
2836         c -= a; c -= b; c ^= (b >> 13);                                 \
2837         a -= b; a -= c; a ^= (c >> 12);                                 \
2838         b -= c; b -= a; b ^= (a << 16);                                 \
2839         c -= a; c -= b; c ^= (b >> 5);                                  \
2840         a -= b; a -= c; a ^= (c >> 3);                                  \
2841         b -= c; b -= a; b ^= (a << 10);                                 \
2842         c -= a; c -= b; c ^= (b >> 15);                                 \
2843 } while (/*CONSTCOND*/0)
2844
2845 static __inline uint32_t
2846 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2847 {
2848         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2849
2850         b += addr[5] << 8;
2851         b += addr[4];
2852         a += addr[3] << 24;
2853         a += addr[2] << 16;
2854         a += addr[1] << 8;
2855         a += addr[0];
2856
2857         mix(a, b, c);
2858
2859         return (c & BRIDGE_RTHASH_MASK);
2860 }
2861
2862 #undef mix
2863
2864 static int
2865 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2866 {
2867         int i, d;
2868
2869         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2870                 d = ((int)a[i]) - ((int)b[i]);
2871         }
2872
2873         return (d);
2874 }
2875
2876 /*
2877  * bridge_rtnode_lookup:
2878  *
2879  *      Look up a bridge route node for the specified destination. Compare the
2880  *      vlan id or if zero then just return the first match.
2881  */
2882 static struct bridge_rtnode *
2883 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2884 {
2885         struct bridge_rtnode *brt;
2886         uint32_t hash;
2887         int dir;
2888
2889         BRIDGE_LOCK_ASSERT(sc);
2890
2891         hash = bridge_rthash(sc, addr);
2892         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2893                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2894                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2895                         return (brt);
2896                 if (dir > 0)
2897                         return (NULL);
2898         }
2899
2900         return (NULL);
2901 }
2902
2903 /*
2904  * bridge_rtnode_insert:
2905  *
2906  *      Insert the specified bridge node into the route table.  We
2907  *      assume the entry is not already in the table.
2908  */
2909 static int
2910 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2911 {
2912         struct bridge_rtnode *lbrt;
2913         uint32_t hash;
2914         int dir;
2915
2916         BRIDGE_LOCK_ASSERT(sc);
2917
2918         hash = bridge_rthash(sc, brt->brt_addr);
2919
2920         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2921         if (lbrt == NULL) {
2922                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2923                 goto out;
2924         }
2925
2926         do {
2927                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2928                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2929                         return (EEXIST);
2930                 if (dir > 0) {
2931                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2932                         goto out;
2933                 }
2934                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2935                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2936                         goto out;
2937                 }
2938                 lbrt = LIST_NEXT(lbrt, brt_hash);
2939         } while (lbrt != NULL);
2940
2941 #ifdef DIAGNOSTIC
2942         panic("bridge_rtnode_insert: impossible");
2943 #endif
2944
2945 out:
2946         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2947         sc->sc_brtcnt++;
2948
2949         return (0);
2950 }
2951
2952 /*
2953  * bridge_rtnode_destroy:
2954  *
2955  *      Destroy a bridge rtnode.
2956  */
2957 static void
2958 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2959 {
2960         BRIDGE_LOCK_ASSERT(sc);
2961
2962         LIST_REMOVE(brt, brt_hash);
2963
2964         LIST_REMOVE(brt, brt_list);
2965         sc->sc_brtcnt--;
2966         brt->brt_dst->bif_addrcnt--;
2967         uma_zfree(bridge_rtnode_zone, brt);
2968 }
2969
2970 /*
2971  * bridge_rtable_expire:
2972  *
2973  *      Set the expiry time for all routes on an interface.
2974  */
2975 static void
2976 bridge_rtable_expire(struct ifnet *ifp, int age)
2977 {
2978         struct bridge_softc *sc = ifp->if_bridge;
2979         struct bridge_rtnode *brt;
2980
2981         BRIDGE_LOCK(sc);
2982
2983         /*
2984          * If the age is zero then flush, otherwise set all the expiry times to
2985          * age for the interface
2986          */
2987         if (age == 0)
2988                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2989         else {
2990                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2991                         /* Cap the expiry time to 'age' */
2992                         if (brt->brt_ifp == ifp &&
2993                             brt->brt_expire > time_uptime + age &&
2994                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2995                                 brt->brt_expire = time_uptime + age;
2996                 }
2997         }
2998         BRIDGE_UNLOCK(sc);
2999 }
3000
3001 /*
3002  * bridge_state_change:
3003  *
3004  *      Callback from the bridgestp code when a port changes states.
3005  */
3006 static void
3007 bridge_state_change(struct ifnet *ifp, int state)
3008 {
3009         struct bridge_softc *sc = ifp->if_bridge;
3010         static const char *stpstates[] = {
3011                 "disabled",
3012                 "listening",
3013                 "learning",
3014                 "forwarding",
3015                 "blocking",
3016                 "discarding"
3017         };
3018
3019         if (log_stp)
3020                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3021                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3022 }
3023
3024 /*
3025  * Send bridge packets through pfil if they are one of the types pfil can deal
3026  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3027  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3028  * that interface.
3029  */
3030 static int
3031 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3032 {
3033         int snap, error, i, hlen;
3034         struct ether_header *eh1, eh2;
3035         struct ip_fw_args args;
3036         struct ip *ip;
3037         struct llc llc1;
3038         u_int16_t ether_type;
3039
3040         snap = 0;
3041         error = -1;     /* Default error if not error == 0 */
3042
3043 #if 0
3044         /* we may return with the IP fields swapped, ensure its not shared */
3045         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3046 #endif
3047
3048         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
3049                 return (0); /* filtering is disabled */
3050
3051         i = min((*mp)->m_pkthdr.len, max_protohdr);
3052         if ((*mp)->m_len < i) {
3053             *mp = m_pullup(*mp, i);
3054             if (*mp == NULL) {
3055                 printf("%s: m_pullup failed\n", __func__);
3056                 return (-1);
3057             }
3058         }
3059
3060         eh1 = mtod(*mp, struct ether_header *);
3061         ether_type = ntohs(eh1->ether_type);
3062
3063         /*
3064          * Check for SNAP/LLC.
3065          */
3066         if (ether_type < ETHERMTU) {
3067                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3068
3069                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3070                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3071                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3072                     llc2->llc_control == LLC_UI) {
3073                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3074                         snap = 1;
3075                 }
3076         }
3077
3078         /*
3079          * If we're trying to filter bridge traffic, don't look at anything
3080          * other than IP and ARP traffic.  If the filter doesn't understand
3081          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3082          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3083          * but of course we don't have an AppleTalk filter to begin with.
3084          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3085          * ARP traffic.)
3086          */
3087         switch (ether_type) {
3088                 case ETHERTYPE_ARP:
3089                 case ETHERTYPE_REVARP:
3090                         if (pfil_ipfw_arp == 0)
3091                                 return (0); /* Automatically pass */
3092                         break;
3093
3094                 case ETHERTYPE_IP:
3095 #ifdef INET6
3096                 case ETHERTYPE_IPV6:
3097 #endif /* INET6 */
3098                         break;
3099                 default:
3100                         /*
3101                          * Check to see if the user wants to pass non-ip
3102                          * packets, these will not be checked by pfil(9) and
3103                          * passed unconditionally so the default is to drop.
3104                          */
3105                         if (pfil_onlyip)
3106                                 goto bad;
3107         }
3108
3109         /* Strip off the Ethernet header and keep a copy. */
3110         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3111         m_adj(*mp, ETHER_HDR_LEN);
3112
3113         /* Strip off snap header, if present */
3114         if (snap) {
3115                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3116                 m_adj(*mp, sizeof(struct llc));
3117         }
3118
3119         /*
3120          * Check the IP header for alignment and errors
3121          */
3122         if (dir == PFIL_IN) {
3123                 switch (ether_type) {
3124                         case ETHERTYPE_IP:
3125                                 error = bridge_ip_checkbasic(mp);
3126                                 break;
3127 #ifdef INET6
3128                         case ETHERTYPE_IPV6:
3129                                 error = bridge_ip6_checkbasic(mp);
3130                                 break;
3131 #endif /* INET6 */
3132                         default:
3133                                 error = 0;
3134                 }
3135                 if (error)
3136                         goto bad;
3137         }
3138
3139         /* XXX this section is also in if_ethersubr.c */
3140         // XXX PFIL_OUT or DIR_OUT ?
3141         if (V_ip_fw_chk_ptr && pfil_ipfw != 0 &&
3142                         dir == PFIL_OUT && ifp != NULL) {
3143                 struct m_tag *mtag;
3144
3145                 error = -1;
3146                 /* fetch the start point from existing tags, if any */
3147                 mtag = m_tag_locate(*mp, MTAG_IPFW_RULE, 0, NULL);
3148                 if (mtag == NULL) {
3149                         args.rule.slot = 0;
3150                 } else {
3151                         struct ipfw_rule_ref *r;
3152
3153                         /* XXX can we free the tag after use ? */
3154                         mtag->m_tag_id = PACKET_TAG_NONE;
3155                         r = (struct ipfw_rule_ref *)(mtag + 1);
3156                         /* packet already partially processed ? */
3157                         if (r->info & IPFW_ONEPASS)
3158                                 goto ipfwpass;
3159                         args.rule = *r;
3160                 }
3161
3162                 args.m = *mp;
3163                 args.oif = ifp;
3164                 args.next_hop = NULL;
3165                 args.next_hop6 = NULL;
3166                 args.eh = &eh2;
3167                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3168                 i = V_ip_fw_chk_ptr(&args);
3169                 *mp = args.m;
3170
3171                 if (*mp == NULL)
3172                         return (error);
3173
3174                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3175
3176                         /* put the Ethernet header back on */
3177                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3178                         if (*mp == NULL)
3179                                 return (error);
3180                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3181
3182                         /*
3183                          * Pass the pkt to dummynet, which consumes it. The
3184                          * packet will return to us via bridge_dummynet().
3185                          */
3186                         args.oif = ifp;
3187                         ip_dn_io_ptr(mp, DIR_FWD | PROTO_IFB, &args);
3188                         return (error);
3189                 }
3190
3191                 if (i != IP_FW_PASS) /* drop */
3192                         goto bad;
3193         }
3194
3195 ipfwpass:
3196         error = 0;
3197
3198         /*
3199          * Run the packet through pfil
3200          */
3201         switch (ether_type) {
3202         case ETHERTYPE_IP:
3203                 /*
3204                  * before calling the firewall, swap fields the same as
3205                  * IP does. here we assume the header is contiguous
3206                  */
3207                 ip = mtod(*mp, struct ip *);
3208
3209                 ip->ip_len = ntohs(ip->ip_len);
3210                 ip->ip_off = ntohs(ip->ip_off);
3211
3212                 /*
3213                  * Run pfil on the member interface and the bridge, both can
3214                  * be skipped by clearing pfil_member or pfil_bridge.
3215                  *
3216                  * Keep the order:
3217                  *   in_if -> bridge_if -> out_if
3218                  */
3219                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3220                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3221                                         dir, NULL);
3222
3223                 if (*mp == NULL || error != 0) /* filter may consume */
3224                         break;
3225
3226                 if (pfil_member && ifp != NULL)
3227                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3228                                         dir, NULL);
3229
3230                 if (*mp == NULL || error != 0) /* filter may consume */
3231                         break;
3232
3233                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3234                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3235                                         dir, NULL);
3236
3237                 if (*mp == NULL || error != 0) /* filter may consume */
3238                         break;
3239
3240                 /* check if we need to fragment the packet */
3241                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3242                         i = (*mp)->m_pkthdr.len;
3243                         if (i > ifp->if_mtu) {
3244                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3245                                             &llc1);
3246                                 return (error);
3247                         }
3248                 }
3249
3250                 /* Recalculate the ip checksum and restore byte ordering */
3251                 ip = mtod(*mp, struct ip *);
3252                 hlen = ip->ip_hl << 2;
3253                 if (hlen < sizeof(struct ip))
3254                         goto bad;
3255                 if (hlen > (*mp)->m_len) {
3256                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3257                                 goto bad;
3258                         ip = mtod(*mp, struct ip *);
3259                         if (ip == NULL)
3260                                 goto bad;
3261                 }
3262                 ip->ip_len = htons(ip->ip_len);
3263                 ip->ip_off = htons(ip->ip_off);
3264                 ip->ip_sum = 0;
3265                 if (hlen == sizeof(struct ip))
3266                         ip->ip_sum = in_cksum_hdr(ip);
3267                 else
3268                         ip->ip_sum = in_cksum(*mp, hlen);
3269
3270                 break;
3271 #ifdef INET6
3272         case ETHERTYPE_IPV6:
3273                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3274                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3275                                         dir, NULL);
3276
3277                 if (*mp == NULL || error != 0) /* filter may consume */
3278                         break;
3279
3280                 if (pfil_member && ifp != NULL)
3281                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3282                                         dir, NULL);
3283
3284                 if (*mp == NULL || error != 0) /* filter may consume */
3285                         break;
3286
3287                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3288                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3289                                         dir, NULL);
3290                 break;
3291 #endif
3292         default:
3293                 error = 0;
3294                 break;
3295         }
3296
3297         if (*mp == NULL)
3298                 return (error);
3299         if (error != 0)
3300                 goto bad;
3301
3302         error = -1;
3303
3304         /*
3305          * Finally, put everything back the way it was and return
3306          */
3307         if (snap) {
3308                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3309                 if (*mp == NULL)
3310                         return (error);
3311                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3312         }
3313
3314         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3315         if (*mp == NULL)
3316                 return (error);
3317         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3318
3319         return (0);
3320
3321 bad:
3322         m_freem(*mp);
3323         *mp = NULL;
3324         return (error);
3325 }
3326
3327 /*
3328  * Perform basic checks on header size since
3329  * pfil assumes ip_input has already processed
3330  * it for it.  Cut-and-pasted from ip_input.c.
3331  * Given how simple the IPv6 version is,
3332  * does the IPv4 version really need to be
3333  * this complicated?
3334  *
3335  * XXX Should we update ipstat here, or not?
3336  * XXX Right now we update ipstat but not
3337  * XXX csum_counter.
3338  */
3339 static int
3340 bridge_ip_checkbasic(struct mbuf **mp)
3341 {
3342         struct mbuf *m = *mp;
3343         struct ip *ip;
3344         int len, hlen;
3345         u_short sum;
3346
3347         if (*mp == NULL)
3348                 return (-1);
3349
3350         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3351                 if ((m = m_copyup(m, sizeof(struct ip),
3352                         (max_linkhdr + 3) & ~3)) == NULL) {
3353                         /* XXXJRT new stat, please */
3354                         KMOD_IPSTAT_INC(ips_toosmall);
3355                         goto bad;
3356                 }
3357         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3358                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3359                         KMOD_IPSTAT_INC(ips_toosmall);
3360                         goto bad;
3361                 }
3362         }
3363         ip = mtod(m, struct ip *);
3364         if (ip == NULL) goto bad;
3365
3366         if (ip->ip_v != IPVERSION) {
3367                 KMOD_IPSTAT_INC(ips_badvers);
3368                 goto bad;
3369         }
3370         hlen = ip->ip_hl << 2;
3371         if (hlen < sizeof(struct ip)) { /* minimum header length */
3372                 KMOD_IPSTAT_INC(ips_badhlen);
3373                 goto bad;
3374         }
3375         if (hlen > m->m_len) {
3376                 if ((m = m_pullup(m, hlen)) == 0) {
3377                         KMOD_IPSTAT_INC(ips_badhlen);
3378                         goto bad;
3379                 }
3380                 ip = mtod(m, struct ip *);
3381                 if (ip == NULL) goto bad;
3382         }
3383
3384         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3385                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3386         } else {
3387                 if (hlen == sizeof(struct ip)) {
3388                         sum = in_cksum_hdr(ip);
3389                 } else {
3390                         sum = in_cksum(m, hlen);
3391                 }
3392         }
3393         if (sum) {
3394                 KMOD_IPSTAT_INC(ips_badsum);
3395                 goto bad;
3396         }
3397
3398         /* Retrieve the packet length. */
3399         len = ntohs(ip->ip_len);
3400
3401         /*
3402          * Check for additional length bogosity
3403          */
3404         if (len < hlen) {
3405                 KMOD_IPSTAT_INC(ips_badlen);
3406                 goto bad;
3407         }
3408
3409         /*
3410          * Check that the amount of data in the buffers
3411          * is as at least much as the IP header would have us expect.
3412          * Drop packet if shorter than we expect.
3413          */
3414         if (m->m_pkthdr.len < len) {
3415                 KMOD_IPSTAT_INC(ips_tooshort);
3416                 goto bad;
3417         }
3418
3419         /* Checks out, proceed */
3420         *mp = m;
3421         return (0);
3422
3423 bad:
3424         *mp = m;
3425         return (-1);
3426 }
3427
3428 #ifdef INET6
3429 /*
3430  * Same as above, but for IPv6.
3431  * Cut-and-pasted from ip6_input.c.
3432  * XXX Should we update ip6stat, or not?
3433  */
3434 static int
3435 bridge_ip6_checkbasic(struct mbuf **mp)
3436 {
3437         struct mbuf *m = *mp;
3438         struct ip6_hdr *ip6;
3439
3440         /*
3441          * If the IPv6 header is not aligned, slurp it up into a new
3442          * mbuf with space for link headers, in the event we forward
3443          * it.  Otherwise, if it is aligned, make sure the entire base
3444          * IPv6 header is in the first mbuf of the chain.
3445          */
3446         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3447                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3448                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3449                             (max_linkhdr + 3) & ~3)) == NULL) {
3450                         /* XXXJRT new stat, please */
3451                         IP6STAT_INC(ip6s_toosmall);
3452                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3453                         goto bad;
3454                 }
3455         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3456                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3457                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3458                         IP6STAT_INC(ip6s_toosmall);
3459                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3460                         goto bad;
3461                 }
3462         }
3463
3464         ip6 = mtod(m, struct ip6_hdr *);
3465
3466         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3467                 IP6STAT_INC(ip6s_badvers);
3468                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3469                 goto bad;
3470         }
3471
3472         /* Checks out, proceed */
3473         *mp = m;
3474         return (0);
3475
3476 bad:
3477         *mp = m;
3478         return (-1);
3479 }
3480 #endif /* INET6 */
3481
3482 /*
3483  * bridge_fragment:
3484  *
3485  *      Return a fragmented mbuf chain.
3486  */
3487 static int
3488 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3489     int snap, struct llc *llc)
3490 {
3491         struct mbuf *m0;
3492         struct ip *ip;
3493         int error = -1;
3494
3495         if (m->m_len < sizeof(struct ip) &&
3496             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3497                 goto out;
3498         ip = mtod(m, struct ip *);
3499
3500         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3501                     CSUM_DELAY_IP);
3502         if (error)
3503                 goto out;
3504
3505         /* walk the chain and re-add the Ethernet header */
3506         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3507                 if (error == 0) {
3508                         if (snap) {
3509                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3510                                 if (m0 == NULL) {
3511                                         error = ENOBUFS;
3512                                         continue;
3513                                 }
3514                                 bcopy(llc, mtod(m0, caddr_t),
3515                                     sizeof(struct llc));
3516                         }
3517                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3518                         if (m0 == NULL) {
3519                                 error = ENOBUFS;
3520                                 continue;
3521                         }
3522                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3523                 } else
3524                         m_freem(m);
3525         }
3526
3527         if (error == 0)
3528                 KMOD_IPSTAT_INC(ips_fragmented);
3529
3530         return (error);
3531
3532 out:
3533         if (m != NULL)
3534                 m_freem(m);
3535         return (error);
3536 }
3537
3538 static void
3539 bridge_linkstate(struct ifnet *ifp)
3540 {
3541         struct bridge_softc *sc = ifp->if_bridge;
3542         struct bridge_iflist *bif;
3543
3544         BRIDGE_LOCK(sc);
3545         bif = bridge_lookup_member_if(sc, ifp);
3546         if (bif == NULL) {
3547                 BRIDGE_UNLOCK(sc);
3548                 return;
3549         }
3550         bridge_linkcheck(sc);
3551         BRIDGE_UNLOCK(sc);
3552
3553         bstp_linkstate(&bif->bif_stp);
3554 }
3555
3556 static void
3557 bridge_linkcheck(struct bridge_softc *sc)
3558 {
3559         struct bridge_iflist *bif;
3560         int new_link, hasls;
3561
3562         BRIDGE_LOCK_ASSERT(sc);
3563         new_link = LINK_STATE_DOWN;
3564         hasls = 0;
3565         /* Our link is considered up if at least one of our ports is active */
3566         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3567                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3568                         hasls++;
3569                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3570                         new_link = LINK_STATE_UP;
3571                         break;
3572                 }
3573         }
3574         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3575                 /* If no interfaces support link-state then we default to up */
3576                 new_link = LINK_STATE_UP;
3577         }
3578         if_link_state_change(sc->sc_ifp, new_link);
3579 }