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