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 #include <netinet6/in6_ifattach.h>
 122 #endif
 123 #if defined(INET) || defined(INET6)
 124 #include <netinet/ip_carp.h>
 125 #endif
 126 #include <machine/in_cksum.h>
 127 #include <netinet/if_ether.h> /* for struct arpcom */
 128 #include <net/bridgestp.h>
 129 #include <net/if_bridgevar.h>
 130 #include <net/if_llc.h>
 131 #include <net/if_vlan_var.h>
 132
 133 #include <net/route.h>
 134
 135 /*
 136  * Size of the route hash table.  Must be a power of two.
 137  */
 138 #ifndef BRIDGE_RTHASH_SIZE
 139 #define BRIDGE_RTHASH_SIZE              1024
 140 #endif
 141
 142 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 143
 144 /*
 145  * Default maximum number of addresses to cache.
 146  */
 147 #ifndef BRIDGE_RTABLE_MAX
 148 #define BRIDGE_RTABLE_MAX               2000
 149 #endif
 150
 151 /*
 152  * Timeout (in seconds) for entries learned dynamically.
 153  */
 154 #ifndef BRIDGE_RTABLE_TIMEOUT
 155 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 156 #endif
 157
 158 /*
 159  * Number of seconds between walks of the route list.
 160  */
 161 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 162 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 163 #endif
 164
 165 /*
 166  * List of capabilities to possibly mask on the member interface.
 167  */
 168 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 169                                          IFCAP_TXCSUM_IPV6)
 170
 171 /*
 172  * List of capabilities to strip
 173  */
 174 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 175
 176 /*
 177  * Bridge interface list entry.
 178  */
 179 struct bridge_iflist {
 180         LIST_ENTRY(bridge_iflist) bif_next;
 181         struct ifnet            *bif_ifp;       /* member if */
 182         struct bstp_port        bif_stp;        /* STP state */
 183         uint32_t                bif_flags;      /* member if flags */
 184         int                     bif_savedcaps;  /* saved capabilities */
 185         uint32_t                bif_addrmax;    /* max # of addresses */
 186         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 187         uint32_t                bif_addrexceeded;/* # of address violations */
 188 };
 189
 190 /*
 191  * Bridge route node.
 192  */
 193 struct bridge_rtnode {
 194         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 195         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 196         struct bridge_iflist    *brt_dst;       /* destination if */
 197         unsigned long           brt_expire;     /* expiration time */
 198         uint8_t                 brt_flags;      /* address flags */
 199         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 200         uint16_t                brt_vlan;       /* vlan id */
 201 };
 202 #define brt_ifp                 brt_dst->bif_ifp
 203
 204 /*
 205  * Software state for each bridge.
 206  */
 207 struct bridge_softc {
 208         struct ifnet            *sc_ifp;        /* make this an interface */
 209         LIST_ENTRY(bridge_softc) sc_list;
 210         struct mtx              sc_mtx;
 211         struct cv               sc_cv;
 212         uint32_t                sc_brtmax;      /* max # of addresses */
 213         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 214         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 215         struct callout          sc_brcallout;   /* bridge callout */
 216         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 217         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 218         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 219         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 220         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 221         uint32_t                sc_rthash_key;  /* key for hash */
 222         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 223         struct bstp_state       sc_stp;         /* STP state */
 224         uint32_t                sc_brtexceeded; /* # of cache drops */
 225         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 226         u_char                  sc_defaddr[6];  /* Default MAC address */
 227 };
 228
 229 static struct mtx       bridge_list_mtx;
 230 eventhandler_tag        bridge_detach_cookie = NULL;
 231
 232 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 233
 234 uma_zone_t bridge_rtnode_zone;
 235
 236 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 237 static void     bridge_clone_destroy(struct ifnet *);
 238
 239 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 240 static void     bridge_mutecaps(struct bridge_softc *);
 241 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 242                     int);
 243 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 244 static void     bridge_init(void *);
 245 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 246 static void     bridge_stop(struct ifnet *, int);
 247 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 248 static void     bridge_qflush(struct ifnet *);
 249 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 250 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 251                     struct rtentry *);
 252 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 253                     struct mbuf *);
 254 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 255
 256 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 257                     struct mbuf *m);
 258
 259 static void     bridge_timer(void *);
 260
 261 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 262                     struct mbuf *, int);
 263 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 264
 265 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 266                     uint16_t, struct bridge_iflist *, int, uint8_t);
 267 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 268                     uint16_t);
 269 static void     bridge_rttrim(struct bridge_softc *);
 270 static void     bridge_rtage(struct bridge_softc *);
 271 static void     bridge_rtflush(struct bridge_softc *, int);
 272 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 273                     uint16_t);
 274
 275 static void     bridge_rtable_init(struct bridge_softc *);
 276 static void     bridge_rtable_fini(struct bridge_softc *);
 277
 278 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 279 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 280                     const uint8_t *, uint16_t);
 281 static int      bridge_rtnode_insert(struct bridge_softc *,
 282                     struct bridge_rtnode *);
 283 static void     bridge_rtnode_destroy(struct bridge_softc *,
 284                     struct bridge_rtnode *);
 285 static void     bridge_rtable_expire(struct ifnet *, int);
 286 static void     bridge_state_change(struct ifnet *, int);
 287
 288 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 289                     const char *name);
 290 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 291                     struct ifnet *ifp);
 292 static void     bridge_delete_member(struct bridge_softc *,
 293                     struct bridge_iflist *, int);
 294 static void     bridge_delete_span(struct bridge_softc *,
 295                     struct bridge_iflist *);
 296
 297 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 298 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 328 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 329                     int);
 330 static int      bridge_ip_checkbasic(struct mbuf **mp);
 331 #ifdef INET6
 332 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 333 #endif /* INET6 */
 334 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 335                     struct ether_header *, int, struct llc *);
 336 static void     bridge_linkstate(struct ifnet *ifp);
 337 static void     bridge_linkcheck(struct bridge_softc *sc);
 338
 339 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
 340
 341 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 342 #define VLANTAGOF(_m)   \
 343     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 344
 345 static struct bstp_cb_ops bridge_ops = {
 346         .bcb_state = bridge_state_change,
 347         .bcb_rtage = bridge_rtable_expire
 348 };
 349
 350 SYSCTL_DECL(_net_link);
 351 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 352
 353 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 354 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 355 static int pfil_member = 1; /* run pfil hooks on the member interface */
 356 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 357 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 358 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 359                                    locally destined packets */
 360 static int log_stp   = 0;   /* log STP state changes */
 361 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
 362 TUNABLE_INT("net.link.bridge.pfil_onlyip", &pfil_onlyip);
 363 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 364     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 365 TUNABLE_INT("net.link.bridge.ipfw_arp", &pfil_ipfw_arp);
 366 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 367     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 368 TUNABLE_INT("net.link.bridge.pfil_bridge", &pfil_bridge);
 369 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 370     &pfil_bridge, 0, "Packet filter on the bridge interface");
 371 TUNABLE_INT("net.link.bridge.pfil_member", &pfil_member);
 372 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 373     &pfil_member, 0, "Packet filter on the member interface");
 374 TUNABLE_INT("net.link.bridge.pfil_local_phys", &pfil_local_phys);
 375 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 376     &pfil_local_phys, 0,
 377     "Packet filter on the physical interface for locally destined packets");
 378 TUNABLE_INT("net.link.bridge.log_stp", &log_stp);
 379 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 380     &log_stp, 0, "Log STP state changes");
 381 TUNABLE_INT("net.link.bridge.inherit_mac", &bridge_inherit_mac);
 382 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
 383     &bridge_inherit_mac, 0,
 384     "Inherit MAC address from the first bridge member");
 385
 386 static VNET_DEFINE(int, allow_llz_overlap) = 0;
 387 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 388 SYSCTL_VNET_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap, CTLFLAG_RW,
 389     &VNET_NAME(allow_llz_overlap), 0, "Allow overlap of link-local scope "
 390     "zones of a bridge interface and the member interfaces");
 391
 392 struct bridge_control {
 393         int     (*bc_func)(struct bridge_softc *, void *);
 394         int     bc_argsize;
 395         int     bc_flags;
 396 };
 397
 398 #define BC_F_COPYIN             0x01    /* copy arguments in */
 399 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 400 #define BC_F_SUSER              0x04    /* do super-user check */
 401
 402 const struct bridge_control bridge_control_table[] = {
 403         { bridge_ioctl_add,             sizeof(struct ifbreq),
 404           BC_F_COPYIN|BC_F_SUSER },
 405         { bridge_ioctl_del,             sizeof(struct ifbreq),
 406           BC_F_COPYIN|BC_F_SUSER },
 407
 408         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 409           BC_F_COPYIN|BC_F_COPYOUT },
 410         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 411           BC_F_COPYIN|BC_F_SUSER },
 412
 413         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 414           BC_F_COPYIN|BC_F_SUSER },
 415         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 416           BC_F_COPYOUT },
 417
 418         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 419           BC_F_COPYIN|BC_F_COPYOUT },
 420         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 421           BC_F_COPYIN|BC_F_COPYOUT },
 422
 423         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 424           BC_F_COPYIN|BC_F_SUSER },
 425
 426         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 427           BC_F_COPYIN|BC_F_SUSER },
 428         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 429           BC_F_COPYOUT },
 430
 431         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 432           BC_F_COPYIN|BC_F_SUSER },
 433
 434         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 435           BC_F_COPYIN|BC_F_SUSER },
 436
 437         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 438           BC_F_COPYOUT },
 439         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 440           BC_F_COPYIN|BC_F_SUSER },
 441
 442         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 443           BC_F_COPYOUT },
 444         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 445           BC_F_COPYIN|BC_F_SUSER },
 446
 447         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 448           BC_F_COPYOUT },
 449         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 450           BC_F_COPYIN|BC_F_SUSER },
 451
 452         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 453           BC_F_COPYOUT },
 454         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 455           BC_F_COPYIN|BC_F_SUSER },
 456
 457         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 458           BC_F_COPYIN|BC_F_SUSER },
 459
 460         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 461           BC_F_COPYIN|BC_F_SUSER },
 462
 463         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 464           BC_F_COPYIN|BC_F_SUSER },
 465         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 466           BC_F_COPYIN|BC_F_SUSER },
 467
 468         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 469           BC_F_COPYOUT },
 470
 471         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 472           BC_F_COPYOUT },
 473
 474         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 475           BC_F_COPYIN|BC_F_COPYOUT },
 476
 477         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 478           BC_F_COPYIN|BC_F_SUSER },
 479
 480         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 481           BC_F_COPYIN|BC_F_SUSER },
 482
 483         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 484           BC_F_COPYIN|BC_F_SUSER },
 485
 486 };
 487 const int bridge_control_table_size =
 488     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 489
 490 LIST_HEAD(, bridge_softc) bridge_list;
 491
 492 static struct if_clone *bridge_cloner;
 493 static const char bridge_name[] = "bridge";
 494
 495 static int
 496 bridge_modevent(module_t mod, int type, void *data)
 497 {
 498
 499         switch (type) {
 500         case MOD_LOAD:
 501                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 502                 bridge_cloner = if_clone_simple(bridge_name,
 503                     bridge_clone_create, bridge_clone_destroy, 0);
 504                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 505                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 506                     UMA_ALIGN_PTR, 0);
 507                 LIST_INIT(&bridge_list);
 508                 bridge_input_p = bridge_input;
 509                 bridge_output_p = bridge_output;
 510                 bridge_dn_p = bridge_dummynet;
 511                 bridge_linkstate_p = bridge_linkstate;
 512                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 513                     ifnet_departure_event, bridge_ifdetach, NULL,
 514                     EVENTHANDLER_PRI_ANY);
 515                 break;
 516         case MOD_UNLOAD:
 517                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 518                     bridge_detach_cookie);
 519                 if_clone_detach(bridge_cloner);
 520                 uma_zdestroy(bridge_rtnode_zone);
 521                 bridge_input_p = NULL;
 522                 bridge_output_p = NULL;
 523                 bridge_dn_p = NULL;
 524                 bridge_linkstate_p = NULL;
 525                 mtx_destroy(&bridge_list_mtx);
 526                 break;
 527         default:
 528                 return (EOPNOTSUPP);
 529         }
 530         return (0);
 531 }
 532
 533 static moduledata_t bridge_mod = {
 534         "if_bridge",
 535         bridge_modevent,
 536         0
 537 };
 538
 539 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 540 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 541
 542 /*
 543  * handler for net.link.bridge.ipfw
 544  */
 545 static int
 546 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 547 {
 548         int enable = pfil_ipfw;
 549         int error;
 550
 551         error = sysctl_handle_int(oidp, &enable, 0, req);
 552         enable = (enable) ? 1 : 0;
 553
 554         if (enable != pfil_ipfw) {
 555                 pfil_ipfw = enable;
 556
 557                 /*
 558                  * Disable pfil so that ipfw doesnt run twice, if the user
 559                  * really wants both then they can re-enable pfil_bridge and/or
 560                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 561                  * layer2 type.
 562                  */
 563                 if (pfil_ipfw) {
 564                         pfil_onlyip = 0;
 565                         pfil_bridge = 0;
 566                         pfil_member = 0;
 567                 }
 568         }
 569
 570         return (error);
 571 }
 572 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 573             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 574
 575 /*
 576  * bridge_clone_create:
 577  *
 578  *      Create a new bridge instance.
 579  */
 580 static int
 581 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 582 {
 583         struct bridge_softc *sc, *sc2;
 584         struct ifnet *bifp, *ifp;
 585         int fb, retry;
 586         unsigned long hostid;
 587
 588         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 589         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 590         if (ifp == NULL) {
 591                 free(sc, M_DEVBUF);
 592                 return (ENOSPC);
 593         }
 594
 595         BRIDGE_LOCK_INIT(sc);
 596         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 597         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 598
 599         /* Initialize our routing table. */
 600         bridge_rtable_init(sc);
 601
 602         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 603
 604         LIST_INIT(&sc->sc_iflist);
 605         LIST_INIT(&sc->sc_spanlist);
 606
 607         ifp->if_softc = sc;
 608         if_initname(ifp, bridge_name, unit);
 609         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 610         ifp->if_ioctl = bridge_ioctl;
 611         ifp->if_transmit = bridge_transmit;
 612         ifp->if_qflush = bridge_qflush;
 613         ifp->if_init = bridge_init;
 614         ifp->if_type = IFT_BRIDGE;
 615
 616         /*
 617          * Generate an ethernet address with a locally administered address.
 618          *
 619          * Since we are using random ethernet addresses for the bridge, it is
 620          * possible that we might have address collisions, so make sure that
 621          * this hardware address isn't already in use on another bridge.
 622          * The first try uses the hostid and falls back to arc4rand().
 623          */
 624         fb = 0;
 625         getcredhostid(curthread->td_ucred, &hostid);
 626         do {
 627                 if (fb || hostid == 0) {
 628                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 629                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 630                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 631                 } else {
 632                         sc->sc_defaddr[0] = 0x2;
 633                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 634                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 635                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 636                         sc->sc_defaddr[4] =  hostid        & 0xff;
 637                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 638                 }
 639
 640                 fb = 1;
 641                 retry = 0;
 642                 mtx_lock(&bridge_list_mtx);
 643                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 644                         bifp = sc2->sc_ifp;
 645                         if (memcmp(sc->sc_defaddr,
 646                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 647                                 retry = 1;
 648                                 break;
 649                         }
 650                 }
 651                 mtx_unlock(&bridge_list_mtx);
 652         } while (retry == 1);
 653
 654         bstp_attach(&sc->sc_stp, &bridge_ops);
 655         ether_ifattach(ifp, sc->sc_defaddr);
 656         /* Now undo some of the damage... */
 657         ifp->if_baudrate = 0;
 658         ifp->if_type = IFT_BRIDGE;
 659
 660         mtx_lock(&bridge_list_mtx);
 661         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 662         mtx_unlock(&bridge_list_mtx);
 663
 664         return (0);
 665 }
 666
 667 /*
 668  * bridge_clone_destroy:
 669  *
 670  *      Destroy a bridge instance.
 671  */
 672 static void
 673 bridge_clone_destroy(struct ifnet *ifp)
 674 {
 675         struct bridge_softc *sc = ifp->if_softc;
 676         struct bridge_iflist *bif;
 677
 678         BRIDGE_LOCK(sc);
 679
 680         bridge_stop(ifp, 1);
 681         ifp->if_flags &= ~IFF_UP;
 682
 683         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 684                 bridge_delete_member(sc, bif, 0);
 685
 686         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 687                 bridge_delete_span(sc, bif);
 688         }
 689
 690         BRIDGE_UNLOCK(sc);
 691
 692         callout_drain(&sc->sc_brcallout);
 693
 694         mtx_lock(&bridge_list_mtx);
 695         LIST_REMOVE(sc, sc_list);
 696         mtx_unlock(&bridge_list_mtx);
 697
 698         bstp_detach(&sc->sc_stp);
 699         ether_ifdetach(ifp);
 700         if_free(ifp);
 701
 702         /* Tear down the routing table. */
 703         bridge_rtable_fini(sc);
 704
 705         BRIDGE_LOCK_DESTROY(sc);
 706         free(sc, M_DEVBUF);
 707 }
 708
 709 /*
 710  * bridge_ioctl:
 711  *
 712  *      Handle a control request from the operator.
 713  */
 714 static int
 715 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 716 {
 717         struct bridge_softc *sc = ifp->if_softc;
 718         struct ifreq *ifr = (struct ifreq *)data;
 719         struct bridge_iflist *bif;
 720         struct thread *td = curthread;
 721         union {
 722                 struct ifbreq ifbreq;
 723                 struct ifbifconf ifbifconf;
 724                 struct ifbareq ifbareq;
 725                 struct ifbaconf ifbaconf;
 726                 struct ifbrparam ifbrparam;
 727                 struct ifbropreq ifbropreq;
 728         } args;
 729         struct ifdrv *ifd = (struct ifdrv *) data;
 730         const struct bridge_control *bc;
 731         int error = 0;
 732
 733         switch (cmd) {
 734
 735         case SIOCADDMULTI:
 736         case SIOCDELMULTI:
 737                 break;
 738
 739         case SIOCGDRVSPEC:
 740         case SIOCSDRVSPEC:
 741                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 742                         error = EINVAL;
 743                         break;
 744                 }
 745                 bc = &bridge_control_table[ifd->ifd_cmd];
 746
 747                 if (cmd == SIOCGDRVSPEC &&
 748                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 749                         error = EINVAL;
 750                         break;
 751                 }
 752                 else if (cmd == SIOCSDRVSPEC &&
 753                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 754                         error = EINVAL;
 755                         break;
 756                 }
 757
 758                 if (bc->bc_flags & BC_F_SUSER) {
 759                         error = priv_check(td, PRIV_NET_BRIDGE);
 760                         if (error)
 761                                 break;
 762                 }
 763
 764                 if (ifd->ifd_len != bc->bc_argsize ||
 765                     ifd->ifd_len > sizeof(args)) {
 766                         error = EINVAL;
 767                         break;
 768                 }
 769
 770                 bzero(&args, sizeof(args));
 771                 if (bc->bc_flags & BC_F_COPYIN) {
 772                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 773                         if (error)
 774                                 break;
 775                 }
 776
 777                 BRIDGE_LOCK(sc);
 778                 error = (*bc->bc_func)(sc, &args);
 779                 BRIDGE_UNLOCK(sc);
 780                 if (error)
 781                         break;
 782
 783                 if (bc->bc_flags & BC_F_COPYOUT)
 784                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 785
 786                 break;
 787
 788         case SIOCSIFFLAGS:
 789                 if (!(ifp->if_flags & IFF_UP) &&
 790                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 791                         /*
 792                          * If interface is marked down and it is running,
 793                          * then stop and disable it.
 794                          */
 795                         BRIDGE_LOCK(sc);
 796                         bridge_stop(ifp, 1);
 797                         BRIDGE_UNLOCK(sc);
 798                 } else if ((ifp->if_flags & IFF_UP) &&
 799                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 800                         /*
 801                          * If interface is marked up and it is stopped, then
 802                          * start it.
 803                          */
 804                         (*ifp->if_init)(sc);
 805                 }
 806                 break;
 807
 808         case SIOCSIFMTU:
 809                 if (ifr->ifr_mtu < 576) {
 810                         error = EINVAL;
 811                         break;
 812                 }
 813                 if (LIST_EMPTY(&sc->sc_iflist)) {
 814                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 815                         break;
 816                 }
 817                 BRIDGE_LOCK(sc);
 818                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 819                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 820                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
 821                                     " != %d\n", sc->sc_ifp->if_xname,
 822                                     bif->bif_ifp->if_mtu,
 823                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 824                                 error = EINVAL;
 825                                 break;
 826                         }
 827                 }
 828                 if (!error)
 829                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 830                 BRIDGE_UNLOCK(sc);
 831                 break;
 832         default:
 833                 /*
 834                  * drop the lock as ether_ioctl() will call bridge_start() and
 835                  * cause the lock to be recursed.
 836                  */
 837                 error = ether_ioctl(ifp, cmd, data);
 838                 break;
 839         }
 840
 841         return (error);
 842 }
 843
 844 /*
 845  * bridge_mutecaps:
 846  *
 847  *      Clear or restore unwanted capabilities on the member interface
 848  */
 849 static void
 850 bridge_mutecaps(struct bridge_softc *sc)
 851 {
 852         struct bridge_iflist *bif;
 853         int enabled, mask;
 854
 855         /* Initial bitmask of capabilities to test */
 856         mask = BRIDGE_IFCAPS_MASK;
 857
 858         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 859                 /* Every member must support it or its disabled */
 860                 mask &= bif->bif_savedcaps;
 861         }
 862
 863         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 864                 enabled = bif->bif_ifp->if_capenable;
 865                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 866                 /* strip off mask bits and enable them again if allowed */
 867                 enabled &= ~BRIDGE_IFCAPS_MASK;
 868                 enabled |= mask;
 869                 bridge_set_ifcap(sc, bif, enabled);
 870         }
 871
 872 }
 873
 874 static void
 875 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 876 {
 877         struct ifnet *ifp = bif->bif_ifp;
 878         struct ifreq ifr;
 879         int error;
 880
 881         bzero(&ifr, sizeof(ifr));
 882         ifr.ifr_reqcap = set;
 883
 884         if (ifp->if_capenable != set) {
 885                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 886                 if (error)
 887                         if_printf(sc->sc_ifp,
 888                             "error setting interface capabilities on %s\n",
 889                             ifp->if_xname);
 890         }
 891 }
 892
 893 /*
 894  * bridge_lookup_member:
 895  *
 896  *      Lookup a bridge member interface.
 897  */
 898 static struct bridge_iflist *
 899 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 900 {
 901         struct bridge_iflist *bif;
 902         struct ifnet *ifp;
 903
 904         BRIDGE_LOCK_ASSERT(sc);
 905
 906         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 907                 ifp = bif->bif_ifp;
 908                 if (strcmp(ifp->if_xname, name) == 0)
 909                         return (bif);
 910         }
 911
 912         return (NULL);
 913 }
 914
 915 /*
 916  * bridge_lookup_member_if:
 917  *
 918  *      Lookup a bridge member interface by ifnet*.
 919  */
 920 static struct bridge_iflist *
 921 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 922 {
 923         struct bridge_iflist *bif;
 924
 925         BRIDGE_LOCK_ASSERT(sc);
 926
 927         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 928                 if (bif->bif_ifp == member_ifp)
 929                         return (bif);
 930         }
 931
 932         return (NULL);
 933 }
 934
 935 /*
 936  * bridge_delete_member:
 937  *
 938  *      Delete the specified member interface.
 939  */
 940 static void
 941 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 942     int gone)
 943 {
 944         struct ifnet *ifs = bif->bif_ifp;
 945         struct ifnet *fif = NULL;
 946
 947         BRIDGE_LOCK_ASSERT(sc);
 948
 949         if (bif->bif_flags & IFBIF_STP)
 950                 bstp_disable(&bif->bif_stp);
 951
 952         ifs->if_bridge = NULL;
 953         BRIDGE_XLOCK(sc);
 954         LIST_REMOVE(bif, bif_next);
 955         BRIDGE_XDROP(sc);
 956
 957         /*
 958          * If removing the interface that gave the bridge its mac address, set
 959          * the mac address of the bridge to the address of the next member, or
 960          * to its default address if no members are left.
 961          */
 962         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
 963                 if (LIST_EMPTY(&sc->sc_iflist)) {
 964                         bcopy(sc->sc_defaddr,
 965                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 966                         sc->sc_ifaddr = NULL;
 967                 } else {
 968                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
 969                         bcopy(IF_LLADDR(fif),
 970                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 971                         sc->sc_ifaddr = fif;
 972                 }
 973                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
 974         }
 975
 976         bridge_linkcheck(sc);
 977         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 978         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 979         KASSERT(bif->bif_addrcnt == 0,
 980             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 981
 982         BRIDGE_UNLOCK(sc);
 983         if (!gone) {
 984                 switch (ifs->if_type) {
 985                 case IFT_ETHER:
 986                 case IFT_L2VLAN:
 987                         /*
 988                          * Take the interface out of promiscuous mode, but only
 989                          * if it was promiscuous in the first place. It might
 990                          * not be if we're in the bridge_ioctl_add() error path.
 991                          */
 992                         if (ifs->if_flags & IFF_PROMISC)
 993                                 (void) ifpromisc(ifs, 0);
 994                         break;
 995
 996                 case IFT_GIF:
 997                         break;
 998
 999                 default:
1000 #ifdef DIAGNOSTIC
1001                         panic("bridge_delete_member: impossible");
1002 #endif
1003                         break;
1004                 }
1005                 /* reneable any interface capabilities */
1006                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1007         }
1008         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1009         BRIDGE_LOCK(sc);
1010         free(bif, M_DEVBUF);
1011 }
1012
1013 /*
1014  * bridge_delete_span:
1015  *
1016  *      Delete the specified span interface.
1017  */
1018 static void
1019 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1020 {
1021         BRIDGE_LOCK_ASSERT(sc);
1022
1023         KASSERT(bif->bif_ifp->if_bridge == NULL,
1024             ("%s: not a span interface", __func__));
1025
1026         LIST_REMOVE(bif, bif_next);
1027         free(bif, M_DEVBUF);
1028 }
1029
1030 static int
1031 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1032 {
1033         struct ifbreq *req = arg;
1034         struct bridge_iflist *bif = NULL;
1035         struct ifnet *ifs;
1036         int error = 0;
1037
1038         ifs = ifunit(req->ifbr_ifsname);
1039         if (ifs == NULL)
1040                 return (ENOENT);
1041         if (ifs->if_ioctl == NULL)      /* must be supported */
1042                 return (EINVAL);
1043
1044         /* If it's in the span list, it can't be a member. */
1045         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1046                 if (ifs == bif->bif_ifp)
1047                         return (EBUSY);
1048
1049         if (ifs->if_bridge == sc)
1050                 return (EEXIST);
1051
1052         if (ifs->if_bridge != NULL)
1053                 return (EBUSY);
1054
1055         switch (ifs->if_type) {
1056         case IFT_ETHER:
1057         case IFT_L2VLAN:
1058         case IFT_GIF:
1059                 /* permitted interface types */
1060                 break;
1061         default:
1062                 return (EINVAL);
1063         }
1064
1065 #ifdef INET6
1066         /*
1067          * Two valid inet6 addresses with link-local scope must not be
1068          * on the parent interface and the member interfaces at the
1069          * same time.  This restriction is needed to prevent violation
1070          * of link-local scope zone.  Attempts to add a member
1071          * interface which has inet6 addresses when the parent has
1072          * inet6 triggers removal of all inet6 addresses on the member
1073          * interface.
1074          */
1075
1076         /* Check if the parent interface has a link-local scope addr. */
1077         if (V_allow_llz_overlap == 0 &&
1078             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1079                 /*
1080                  * If any, remove all inet6 addresses from the member
1081                  * interfaces.
1082                  */
1083                 BRIDGE_XLOCK(sc);
1084                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1085                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1086                                 BRIDGE_UNLOCK(sc);
1087                                 in6_ifdetach(bif->bif_ifp);
1088                                 BRIDGE_LOCK(sc);
1089                                 if_printf(sc->sc_ifp,
1090                                     "IPv6 addresses on %s have been removed "
1091                                     "before adding it as a member to prevent "
1092                                     "IPv6 address scope violation.\n",
1093                                     bif->bif_ifp->if_xname);
1094                         }
1095                 }
1096                 BRIDGE_XDROP(sc);
1097                 if (in6ifa_llaonifp(ifs)) {
1098                         BRIDGE_UNLOCK(sc);
1099                         in6_ifdetach(ifs);
1100                         BRIDGE_LOCK(sc);
1101                         if_printf(sc->sc_ifp,
1102                             "IPv6 addresses on %s have been removed "
1103                             "before adding it as a member to prevent "
1104                             "IPv6 address scope violation.\n",
1105                             ifs->if_xname);
1106                 }
1107         }
1108 #endif
1109         /* Allow the first Ethernet member to define the MTU */
1110         if (LIST_EMPTY(&sc->sc_iflist))
1111                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1112         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1113                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1114                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1115                 return (EINVAL);
1116         }
1117
1118         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1119         if (bif == NULL)
1120                 return (ENOMEM);
1121
1122         bif->bif_ifp = ifs;
1123         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1124         bif->bif_savedcaps = ifs->if_capenable;
1125
1126         /*
1127          * Assign the interface's MAC address to the bridge if it's the first
1128          * member and the MAC address of the bridge has not been changed from
1129          * the default randomly generated one.
1130          */
1131         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1132             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1133                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1134                 sc->sc_ifaddr = ifs;
1135                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1136         }
1137
1138         ifs->if_bridge = sc;
1139         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1140         /*
1141          * XXX: XLOCK HERE!?!
1142          *
1143          * NOTE: insert_***HEAD*** should be safe for the traversals.
1144          */
1145         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1146
1147         /* Set interface capabilities to the intersection set of all members */
1148         bridge_mutecaps(sc);
1149         bridge_linkcheck(sc);
1150
1151         /* Place the interface into promiscuous mode */
1152         switch (ifs->if_type) {
1153                 case IFT_ETHER:
1154                 case IFT_L2VLAN:
1155                         BRIDGE_UNLOCK(sc);
1156                         error = ifpromisc(ifs, 1);
1157                         BRIDGE_LOCK(sc);
1158                         break;
1159         }
1160
1161         if (error)
1162                 bridge_delete_member(sc, bif, 0);
1163         return (error);
1164 }
1165
1166 static int
1167 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1168 {
1169         struct ifbreq *req = arg;
1170         struct bridge_iflist *bif;
1171
1172         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1173         if (bif == NULL)
1174                 return (ENOENT);
1175
1176         bridge_delete_member(sc, bif, 0);
1177
1178         return (0);
1179 }
1180
1181 static int
1182 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1183 {
1184         struct ifbreq *req = arg;
1185         struct bridge_iflist *bif;
1186         struct bstp_port *bp;
1187
1188         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1189         if (bif == NULL)
1190                 return (ENOENT);
1191
1192         bp = &bif->bif_stp;
1193         req->ifbr_ifsflags = bif->bif_flags;
1194         req->ifbr_state = bp->bp_state;
1195         req->ifbr_priority = bp->bp_priority;
1196         req->ifbr_path_cost = bp->bp_path_cost;
1197         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1198         req->ifbr_proto = bp->bp_protover;
1199         req->ifbr_role = bp->bp_role;
1200         req->ifbr_stpflags = bp->bp_flags;
1201         req->ifbr_addrcnt = bif->bif_addrcnt;
1202         req->ifbr_addrmax = bif->bif_addrmax;
1203         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1204
1205         /* Copy STP state options as flags */
1206         if (bp->bp_operedge)
1207                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1208         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1209                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1210         if (bp->bp_ptp_link)
1211                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1212         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1213                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1214         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1215                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1216         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1217                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1218         return (0);
1219 }
1220
1221 static int
1222 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1223 {
1224         struct ifbreq *req = arg;
1225         struct bridge_iflist *bif;
1226         struct bstp_port *bp;
1227         int error;
1228
1229         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1230         if (bif == NULL)
1231                 return (ENOENT);
1232         bp = &bif->bif_stp;
1233
1234         if (req->ifbr_ifsflags & IFBIF_SPAN)
1235                 /* SPAN is readonly */
1236                 return (EINVAL);
1237
1238         if (req->ifbr_ifsflags & IFBIF_STP) {
1239                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1240                         error = bstp_enable(&bif->bif_stp);
1241                         if (error)
1242                                 return (error);
1243                 }
1244         } else {
1245                 if ((bif->bif_flags & IFBIF_STP) != 0)
1246                         bstp_disable(&bif->bif_stp);
1247         }
1248
1249         /* Pass on STP flags */
1250         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1251         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1252         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1253         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1254
1255         /* Save the bits relating to the bridge */
1256         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1257
1258         return (0);
1259 }
1260
1261 static int
1262 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1263 {
1264         struct ifbrparam *param = arg;
1265
1266         sc->sc_brtmax = param->ifbrp_csize;
1267         bridge_rttrim(sc);
1268
1269         return (0);
1270 }
1271
1272 static int
1273 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1274 {
1275         struct ifbrparam *param = arg;
1276
1277         param->ifbrp_csize = sc->sc_brtmax;
1278
1279         return (0);
1280 }
1281
1282 static int
1283 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1284 {
1285         struct ifbifconf *bifc = arg;
1286         struct bridge_iflist *bif;
1287         struct ifbreq breq;
1288         char *buf, *outbuf;
1289         int count, buflen, len, error = 0;
1290
1291         count = 0;
1292         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1293                 count++;
1294         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1295                 count++;
1296
1297         buflen = sizeof(breq) * count;
1298         if (bifc->ifbic_len == 0) {
1299                 bifc->ifbic_len = buflen;
1300                 return (0);
1301         }
1302         BRIDGE_UNLOCK(sc);
1303         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1304         BRIDGE_LOCK(sc);
1305
1306         count = 0;
1307         buf = outbuf;
1308         len = min(bifc->ifbic_len, buflen);
1309         bzero(&breq, sizeof(breq));
1310         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1311                 if (len < sizeof(breq))
1312                         break;
1313
1314                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1315                     sizeof(breq.ifbr_ifsname));
1316                 /* Fill in the ifbreq structure */
1317                 error = bridge_ioctl_gifflags(sc, &breq);
1318                 if (error)
1319                         break;
1320                 memcpy(buf, &breq, sizeof(breq));
1321                 count++;
1322                 buf += sizeof(breq);
1323                 len -= sizeof(breq);
1324         }
1325         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1326                 if (len < sizeof(breq))
1327                         break;
1328
1329                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1330                     sizeof(breq.ifbr_ifsname));
1331                 breq.ifbr_ifsflags = bif->bif_flags;
1332                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1333                 memcpy(buf, &breq, sizeof(breq));
1334                 count++;
1335                 buf += sizeof(breq);
1336                 len -= sizeof(breq);
1337         }
1338
1339         BRIDGE_UNLOCK(sc);
1340         bifc->ifbic_len = sizeof(breq) * count;
1341         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1342         BRIDGE_LOCK(sc);
1343         free(outbuf, M_TEMP);
1344         return (error);
1345 }
1346
1347 static int
1348 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1349 {
1350         struct ifbaconf *bac = arg;
1351         struct bridge_rtnode *brt;
1352         struct ifbareq bareq;
1353         char *buf, *outbuf;
1354         int count, buflen, len, error = 0;
1355
1356         if (bac->ifbac_len == 0)
1357                 return (0);
1358
1359         count = 0;
1360         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1361                 count++;
1362         buflen = sizeof(bareq) * count;
1363
1364         BRIDGE_UNLOCK(sc);
1365         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1366         BRIDGE_LOCK(sc);
1367
1368         count = 0;
1369         buf = outbuf;
1370         len = min(bac->ifbac_len, buflen);
1371         bzero(&bareq, sizeof(bareq));
1372         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1373                 if (len < sizeof(bareq))
1374                         goto out;
1375                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1376                     sizeof(bareq.ifba_ifsname));
1377                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1378                 bareq.ifba_vlan = brt->brt_vlan;
1379                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1380                                 time_uptime < brt->brt_expire)
1381                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1382                 else
1383                         bareq.ifba_expire = 0;
1384                 bareq.ifba_flags = brt->brt_flags;
1385
1386                 memcpy(buf, &bareq, sizeof(bareq));
1387                 count++;
1388                 buf += sizeof(bareq);
1389                 len -= sizeof(bareq);
1390         }
1391 out:
1392         BRIDGE_UNLOCK(sc);
1393         bac->ifbac_len = sizeof(bareq) * count;
1394         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1395         BRIDGE_LOCK(sc);
1396         free(outbuf, M_TEMP);
1397         return (error);
1398 }
1399
1400 static int
1401 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1402 {
1403         struct ifbareq *req = arg;
1404         struct bridge_iflist *bif;
1405         int error;
1406
1407         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1408         if (bif == NULL)
1409                 return (ENOENT);
1410
1411         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1412             req->ifba_flags);
1413
1414         return (error);
1415 }
1416
1417 static int
1418 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1419 {
1420         struct ifbrparam *param = arg;
1421
1422         sc->sc_brttimeout = param->ifbrp_ctime;
1423         return (0);
1424 }
1425
1426 static int
1427 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1428 {
1429         struct ifbrparam *param = arg;
1430
1431         param->ifbrp_ctime = sc->sc_brttimeout;
1432         return (0);
1433 }
1434
1435 static int
1436 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1437 {
1438         struct ifbareq *req = arg;
1439
1440         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1441 }
1442
1443 static int
1444 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1445 {
1446         struct ifbreq *req = arg;
1447
1448         bridge_rtflush(sc, req->ifbr_ifsflags);
1449         return (0);
1450 }
1451
1452 static int
1453 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1454 {
1455         struct ifbrparam *param = arg;
1456         struct bstp_state *bs = &sc->sc_stp;
1457
1458         param->ifbrp_prio = bs->bs_bridge_priority;
1459         return (0);
1460 }
1461
1462 static int
1463 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1464 {
1465         struct ifbrparam *param = arg;
1466
1467         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1468 }
1469
1470 static int
1471 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1472 {
1473         struct ifbrparam *param = arg;
1474         struct bstp_state *bs = &sc->sc_stp;
1475
1476         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1477         return (0);
1478 }
1479
1480 static int
1481 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1482 {
1483         struct ifbrparam *param = arg;
1484
1485         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1486 }
1487
1488 static int
1489 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1490 {
1491         struct ifbrparam *param = arg;
1492         struct bstp_state *bs = &sc->sc_stp;
1493
1494         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1495         return (0);
1496 }
1497
1498 static int
1499 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1500 {
1501         struct ifbrparam *param = arg;
1502
1503         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1504 }
1505
1506 static int
1507 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1508 {
1509         struct ifbrparam *param = arg;
1510         struct bstp_state *bs = &sc->sc_stp;
1511
1512         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1513         return (0);
1514 }
1515
1516 static int
1517 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1518 {
1519         struct ifbrparam *param = arg;
1520
1521         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1522 }
1523
1524 static int
1525 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1526 {
1527         struct ifbreq *req = arg;
1528         struct bridge_iflist *bif;
1529
1530         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1531         if (bif == NULL)
1532                 return (ENOENT);
1533
1534         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1535 }
1536
1537 static int
1538 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1539 {
1540         struct ifbreq *req = arg;
1541         struct bridge_iflist *bif;
1542
1543         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1544         if (bif == NULL)
1545                 return (ENOENT);
1546
1547         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1548 }
1549
1550 static int
1551 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1552 {
1553         struct ifbreq *req = arg;
1554         struct bridge_iflist *bif;
1555
1556         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1557         if (bif == NULL)
1558                 return (ENOENT);
1559
1560         bif->bif_addrmax = req->ifbr_addrmax;
1561         return (0);
1562 }
1563
1564 static int
1565 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1566 {
1567         struct ifbreq *req = arg;
1568         struct bridge_iflist *bif = NULL;
1569         struct ifnet *ifs;
1570
1571         ifs = ifunit(req->ifbr_ifsname);
1572         if (ifs == NULL)
1573                 return (ENOENT);
1574
1575         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1576                 if (ifs == bif->bif_ifp)
1577                         return (EBUSY);
1578
1579         if (ifs->if_bridge != NULL)
1580                 return (EBUSY);
1581
1582         switch (ifs->if_type) {
1583                 case IFT_ETHER:
1584                 case IFT_GIF:
1585                 case IFT_L2VLAN:
1586                         break;
1587                 default:
1588                         return (EINVAL);
1589         }
1590
1591         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1592         if (bif == NULL)
1593                 return (ENOMEM);
1594
1595         bif->bif_ifp = ifs;
1596         bif->bif_flags = IFBIF_SPAN;
1597
1598         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1599
1600         return (0);
1601 }
1602
1603 static int
1604 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1605 {
1606         struct ifbreq *req = arg;
1607         struct bridge_iflist *bif;
1608         struct ifnet *ifs;
1609
1610         ifs = ifunit(req->ifbr_ifsname);
1611         if (ifs == NULL)
1612                 return (ENOENT);
1613
1614         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1615                 if (ifs == bif->bif_ifp)
1616                         break;
1617
1618         if (bif == NULL)
1619                 return (ENOENT);
1620
1621         bridge_delete_span(sc, bif);
1622
1623         return (0);
1624 }
1625
1626 static int
1627 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1628 {
1629         struct ifbropreq *req = arg;
1630         struct bstp_state *bs = &sc->sc_stp;
1631         struct bstp_port *root_port;
1632
1633         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1634         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1635         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1636
1637         root_port = bs->bs_root_port;
1638         if (root_port == NULL)
1639                 req->ifbop_root_port = 0;
1640         else
1641                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1642
1643         req->ifbop_holdcount = bs->bs_txholdcount;
1644         req->ifbop_priority = bs->bs_bridge_priority;
1645         req->ifbop_protocol = bs->bs_protover;
1646         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1647         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1648         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1649         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1650         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1651         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1652
1653         return (0);
1654 }
1655
1656 static int
1657 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1658 {
1659         struct ifbrparam *param = arg;
1660
1661         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1662         return (0);
1663 }
1664
1665 static int
1666 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1667 {
1668         struct ifbpstpconf *bifstp = arg;
1669         struct bridge_iflist *bif;
1670         struct bstp_port *bp;
1671         struct ifbpstpreq bpreq;
1672         char *buf, *outbuf;
1673         int count, buflen, len, error = 0;
1674
1675         count = 0;
1676         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1677                 if ((bif->bif_flags & IFBIF_STP) != 0)
1678                         count++;
1679         }
1680
1681         buflen = sizeof(bpreq) * count;
1682         if (bifstp->ifbpstp_len == 0) {
1683                 bifstp->ifbpstp_len = buflen;
1684                 return (0);
1685         }
1686
1687         BRIDGE_UNLOCK(sc);
1688         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1689         BRIDGE_LOCK(sc);
1690
1691         count = 0;
1692         buf = outbuf;
1693         len = min(bifstp->ifbpstp_len, buflen);
1694         bzero(&bpreq, sizeof(bpreq));
1695         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1696                 if (len < sizeof(bpreq))
1697                         break;
1698
1699                 if ((bif->bif_flags & IFBIF_STP) == 0)
1700                         continue;
1701
1702                 bp = &bif->bif_stp;
1703                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1704                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1705                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1706                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1707                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1708                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1709
1710                 memcpy(buf, &bpreq, sizeof(bpreq));
1711                 count++;
1712                 buf += sizeof(bpreq);
1713                 len -= sizeof(bpreq);
1714         }
1715
1716         BRIDGE_UNLOCK(sc);
1717         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1718         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1719         BRIDGE_LOCK(sc);
1720         free(outbuf, M_TEMP);
1721         return (error);
1722 }
1723
1724 static int
1725 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1726 {
1727         struct ifbrparam *param = arg;
1728
1729         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1730 }
1731
1732 static int
1733 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1734 {
1735         struct ifbrparam *param = arg;
1736
1737         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1738 }
1739
1740 /*
1741  * bridge_ifdetach:
1742  *
1743  *      Detach an interface from a bridge.  Called when a member
1744  *      interface is detaching.
1745  */
1746 static void
1747 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1748 {
1749         struct bridge_softc *sc = ifp->if_bridge;
1750         struct bridge_iflist *bif;
1751
1752         if (ifp->if_flags & IFF_RENAMING)
1753                 return;
1754
1755         /* Check if the interface is a bridge member */
1756         if (sc != NULL) {
1757                 BRIDGE_LOCK(sc);
1758
1759                 bif = bridge_lookup_member_if(sc, ifp);
1760                 if (bif != NULL)
1761                         bridge_delete_member(sc, bif, 1);
1762
1763                 BRIDGE_UNLOCK(sc);
1764                 return;
1765         }
1766
1767         /* Check if the interface is a span port */
1768         mtx_lock(&bridge_list_mtx);
1769         LIST_FOREACH(sc, &bridge_list, sc_list) {
1770                 BRIDGE_LOCK(sc);
1771                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1772                         if (ifp == bif->bif_ifp) {
1773                                 bridge_delete_span(sc, bif);
1774                                 break;
1775                         }
1776
1777                 BRIDGE_UNLOCK(sc);
1778         }
1779         mtx_unlock(&bridge_list_mtx);
1780 }
1781
1782 /*
1783  * bridge_init:
1784  *
1785  *      Initialize a bridge interface.
1786  */
1787 static void
1788 bridge_init(void *xsc)
1789 {
1790         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1791         struct ifnet *ifp = sc->sc_ifp;
1792
1793         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1794                 return;
1795
1796         BRIDGE_LOCK(sc);
1797         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1798             bridge_timer, sc);
1799
1800         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1801         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1802
1803         BRIDGE_UNLOCK(sc);
1804 }
1805
1806 /*
1807  * bridge_stop:
1808  *
1809  *      Stop the bridge interface.
1810  */
1811 static void
1812 bridge_stop(struct ifnet *ifp, int disable)
1813 {
1814         struct bridge_softc *sc = ifp->if_softc;
1815
1816         BRIDGE_LOCK_ASSERT(sc);
1817
1818         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1819                 return;
1820
1821         callout_stop(&sc->sc_brcallout);
1822         bstp_stop(&sc->sc_stp);
1823
1824         bridge_rtflush(sc, IFBF_FLUSHDYN);
1825
1826         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1827 }
1828
1829 /*
1830  * bridge_enqueue:
1831  *
1832  *      Enqueue a packet on a bridge member interface.
1833  *
1834  */
1835 static int
1836 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1837 {
1838         int len, err = 0;
1839         short mflags;
1840         struct mbuf *m0;
1841
1842         /* We may be sending a fragment so traverse the mbuf */
1843         for (; m; m = m0) {
1844                 m0 = m->m_nextpkt;
1845                 m->m_nextpkt = NULL;
1846                 len = m->m_pkthdr.len;
1847                 mflags = m->m_flags;
1848
1849                 /*
1850                  * If underlying interface can not do VLAN tag insertion itself
1851                  * then attach a packet tag that holds it.
1852                  */
1853                 if ((m->m_flags & M_VLANTAG) &&
1854                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1855                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1856                         if (m == NULL) {
1857                                 if_printf(dst_ifp,
1858                                     "unable to prepend VLAN header\n");
1859                                 dst_ifp->if_oerrors++;
1860                                 continue;
1861                         }
1862                         m->m_flags &= ~M_VLANTAG;
1863                 }
1864
1865                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
1866                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1867                         m_freem(m0);
1868                         sc->sc_ifp->if_oerrors++;
1869                         break;
1870                 }
1871
1872                 sc->sc_ifp->if_opackets++;
1873                 sc->sc_ifp->if_obytes += len;
1874                 if (mflags & M_MCAST)
1875                         sc->sc_ifp->if_omcasts++;
1876         }
1877
1878         return (err);
1879 }
1880
1881 /*
1882  * bridge_dummynet:
1883  *
1884  *      Receive a queued packet from dummynet and pass it on to the output
1885  *      interface.
1886  *
1887  *      The mbuf has the Ethernet header already attached.
1888  */
1889 static void
1890 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1891 {
1892         struct bridge_softc *sc;
1893
1894         sc = ifp->if_bridge;
1895
1896         /*
1897          * The packet didnt originate from a member interface. This should only
1898          * ever happen if a member interface is removed while packets are
1899          * queued for it.
1900          */
1901         if (sc == NULL) {
1902                 m_freem(m);
1903                 return;
1904         }
1905
1906         if (PFIL_HOOKED(&V_inet_pfil_hook)
1907 #ifdef INET6
1908             || PFIL_HOOKED(&V_inet6_pfil_hook)
1909 #endif
1910             ) {
1911                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1912                         return;
1913                 if (m == NULL)
1914                         return;
1915         }
1916
1917         bridge_enqueue(sc, ifp, m);
1918 }
1919
1920 /*
1921  * bridge_output:
1922  *
1923  *      Send output from a bridge member interface.  This
1924  *      performs the bridging function for locally originated
1925  *      packets.
1926  *
1927  *      The mbuf has the Ethernet header already attached.  We must
1928  *      enqueue or free the mbuf before returning.
1929  */
1930 static int
1931 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1932     struct rtentry *rt)
1933 {
1934         struct ether_header *eh;
1935         struct ifnet *dst_if;
1936         struct bridge_softc *sc;
1937         uint16_t vlan;
1938
1939         if (m->m_len < ETHER_HDR_LEN) {
1940                 m = m_pullup(m, ETHER_HDR_LEN);
1941                 if (m == NULL)
1942                         return (0);
1943         }
1944
1945         eh = mtod(m, struct ether_header *);
1946         sc = ifp->if_bridge;
1947         vlan = VLANTAGOF(m);
1948
1949         BRIDGE_LOCK(sc);
1950
1951         /*
1952          * If bridge is down, but the original output interface is up,
1953          * go ahead and send out that interface.  Otherwise, the packet
1954          * is dropped below.
1955          */
1956         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1957                 dst_if = ifp;
1958                 goto sendunicast;
1959         }
1960
1961         /*
1962          * If the packet is a multicast, or we don't know a better way to
1963          * get there, send to all interfaces.
1964          */
1965         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1966                 dst_if = NULL;
1967         else
1968                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1969         if (dst_if == NULL) {
1970                 struct bridge_iflist *bif;
1971                 struct mbuf *mc;
1972                 int error = 0, used = 0;
1973
1974                 bridge_span(sc, m);
1975
1976                 BRIDGE_LOCK2REF(sc, error);
1977                 if (error) {
1978                         m_freem(m);
1979                         return (0);
1980                 }
1981
1982                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1983                         dst_if = bif->bif_ifp;
1984
1985                         if (dst_if->if_type == IFT_GIF)
1986                                 continue;
1987                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1988                                 continue;
1989
1990                         /*
1991                          * If this is not the original output interface,
1992                          * and the interface is participating in spanning
1993                          * tree, make sure the port is in a state that
1994                          * allows forwarding.
1995                          */
1996                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1997                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1998                                 continue;
1999
2000                         if (LIST_NEXT(bif, bif_next) == NULL) {
2001                                 used = 1;
2002                                 mc = m;
2003                         } else {
2004                                 mc = m_copypacket(m, M_NOWAIT);
2005                                 if (mc == NULL) {
2006                                         sc->sc_ifp->if_oerrors++;
2007                                         continue;
2008                                 }
2009                         }
2010
2011                         bridge_enqueue(sc, dst_if, mc);
2012                 }
2013                 if (used == 0)
2014                         m_freem(m);
2015                 BRIDGE_UNREF(sc);
2016                 return (0);
2017         }
2018
2019 sendunicast:
2020         /*
2021          * XXX Spanning tree consideration here?
2022          */
2023
2024         bridge_span(sc, m);
2025         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2026                 m_freem(m);
2027                 BRIDGE_UNLOCK(sc);
2028                 return (0);
2029         }
2030
2031         BRIDGE_UNLOCK(sc);
2032         bridge_enqueue(sc, dst_if, m);
2033         return (0);
2034 }
2035
2036 /*
2037  * bridge_transmit:
2038  *
2039  *      Do output on a bridge.
2040  *
2041  */
2042 static int
2043 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2044 {
2045         struct bridge_softc *sc;
2046         struct ether_header *eh;
2047         struct ifnet *dst_if;
2048         int error = 0;
2049
2050         sc = ifp->if_softc;
2051
2052         ETHER_BPF_MTAP(ifp, m);
2053
2054         eh = mtod(m, struct ether_header *);
2055
2056         BRIDGE_LOCK(sc);
2057         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2058             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2059                 BRIDGE_UNLOCK(sc);
2060                 error = bridge_enqueue(sc, dst_if, m);
2061         } else
2062                 bridge_broadcast(sc, ifp, m, 0);
2063
2064         return (error);
2065 }
2066
2067 /*
2068  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2069  */
2070 static void
2071 bridge_qflush(struct ifnet *ifp __unused)
2072 {
2073 }
2074
2075 /*
2076  * bridge_forward:
2077  *
2078  *      The forwarding function of the bridge.
2079  *
2080  *      NOTE: Releases the lock on return.
2081  */
2082 static void
2083 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2084     struct mbuf *m)
2085 {
2086         struct bridge_iflist *dbif;
2087         struct ifnet *src_if, *dst_if, *ifp;
2088         struct ether_header *eh;
2089         uint16_t vlan;
2090         uint8_t *dst;
2091         int error;
2092
2093         src_if = m->m_pkthdr.rcvif;
2094         ifp = sc->sc_ifp;
2095
2096         ifp->if_ipackets++;
2097         ifp->if_ibytes += m->m_pkthdr.len;
2098         vlan = VLANTAGOF(m);
2099
2100         if ((sbif->bif_flags & IFBIF_STP) &&
2101             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2102                 goto drop;
2103
2104         eh = mtod(m, struct ether_header *);
2105         dst = eh->ether_dhost;
2106
2107         /* If the interface is learning, record the address. */
2108         if (sbif->bif_flags & IFBIF_LEARNING) {
2109                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2110                     sbif, 0, IFBAF_DYNAMIC);
2111                 /*
2112                  * If the interface has addresses limits then deny any source
2113                  * that is not in the cache.
2114                  */
2115                 if (error && sbif->bif_addrmax)
2116                         goto drop;
2117         }
2118
2119         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2120             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2121                 goto drop;
2122
2123         /*
2124          * At this point, the port either doesn't participate
2125          * in spanning tree or it is in the forwarding state.
2126          */
2127
2128         /*
2129          * If the packet is unicast, destined for someone on
2130          * "this" side of the bridge, drop it.
2131          */
2132         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2133                 dst_if = bridge_rtlookup(sc, dst, vlan);
2134                 if (src_if == dst_if)
2135                         goto drop;
2136         } else {
2137                 /*
2138                  * Check if its a reserved multicast address, any address
2139                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2140                  * bridge.
2141                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2142                  */
2143                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2144                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2145                     dst[4] == 0x00 && dst[5] <= 0x0f)
2146                         goto drop;
2147
2148                 /* ...forward it to all interfaces. */
2149                 ifp->if_imcasts++;
2150                 dst_if = NULL;
2151         }
2152
2153         /*
2154          * If we have a destination interface which is a member of our bridge,
2155          * OR this is a unicast packet, push it through the bpf(4) machinery.
2156          * For broadcast or multicast packets, don't bother because it will
2157          * be reinjected into ether_input. We do this before we pass the packets
2158          * through the pfil(9) framework, as it is possible that pfil(9) will
2159          * drop the packet, or possibly modify it, making it difficult to debug
2160          * firewall issues on the bridge.
2161          */
2162         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2163                 ETHER_BPF_MTAP(ifp, m);
2164
2165         /* run the packet filter */
2166         if (PFIL_HOOKED(&V_inet_pfil_hook)
2167 #ifdef INET6
2168             || PFIL_HOOKED(&V_inet6_pfil_hook)
2169 #endif
2170             ) {
2171                 BRIDGE_UNLOCK(sc);
2172                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2173                         return;
2174                 if (m == NULL)
2175                         return;
2176                 BRIDGE_LOCK(sc);
2177         }
2178
2179         if (dst_if == NULL) {
2180                 bridge_broadcast(sc, src_if, m, 1);
2181                 return;
2182         }
2183
2184         /*
2185          * At this point, we're dealing with a unicast frame
2186          * going to a different interface.
2187          */
2188         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2189                 goto drop;
2190
2191         dbif = bridge_lookup_member_if(sc, dst_if);
2192         if (dbif == NULL)
2193                 /* Not a member of the bridge (anymore?) */
2194                 goto drop;
2195
2196         /* Private segments can not talk to each other */
2197         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2198                 goto drop;
2199
2200         if ((dbif->bif_flags & IFBIF_STP) &&
2201             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2202                 goto drop;
2203
2204         BRIDGE_UNLOCK(sc);
2205
2206         if (PFIL_HOOKED(&V_inet_pfil_hook)
2207 #ifdef INET6
2208             || PFIL_HOOKED(&V_inet6_pfil_hook)
2209 #endif
2210             ) {
2211                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2212                         return;
2213                 if (m == NULL)
2214                         return;
2215         }
2216
2217         bridge_enqueue(sc, dst_if, m);
2218         return;
2219
2220 drop:
2221         BRIDGE_UNLOCK(sc);
2222         m_freem(m);
2223 }
2224
2225 /*
2226  * bridge_input:
2227  *
2228  *      Receive input from a member interface.  Queue the packet for
2229  *      bridging if it is not for us.
2230  */
2231 static struct mbuf *
2232 bridge_input(struct ifnet *ifp, struct mbuf *m)
2233 {
2234         struct bridge_softc *sc = ifp->if_bridge;
2235         struct bridge_iflist *bif, *bif2;
2236         struct ifnet *bifp;
2237         struct ether_header *eh;
2238         struct mbuf *mc, *mc2;
2239         uint16_t vlan;
2240         int error;
2241
2242         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2243                 return (m);
2244
2245         bifp = sc->sc_ifp;
2246         vlan = VLANTAGOF(m);
2247
2248         /*
2249          * Implement support for bridge monitoring. If this flag has been
2250          * set on this interface, discard the packet once we push it through
2251          * the bpf(4) machinery, but before we do, increment the byte and
2252          * packet counters associated with this interface.
2253          */
2254         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2255                 m->m_pkthdr.rcvif  = bifp;
2256                 ETHER_BPF_MTAP(bifp, m);
2257                 bifp->if_ipackets++;
2258                 bifp->if_ibytes += m->m_pkthdr.len;
2259                 m_freem(m);
2260                 return (NULL);
2261         }
2262         BRIDGE_LOCK(sc);
2263         bif = bridge_lookup_member_if(sc, ifp);
2264         if (bif == NULL) {
2265                 BRIDGE_UNLOCK(sc);
2266                 return (m);
2267         }
2268
2269         eh = mtod(m, struct ether_header *);
2270
2271         bridge_span(sc, m);
2272
2273         if (m->m_flags & (M_BCAST|M_MCAST)) {
2274                 /* Tap off 802.1D packets; they do not get forwarded. */
2275                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2276                     ETHER_ADDR_LEN) == 0) {
2277                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2278                         BRIDGE_UNLOCK(sc);
2279                         return (NULL);
2280                 }
2281
2282                 if ((bif->bif_flags & IFBIF_STP) &&
2283                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2284                         BRIDGE_UNLOCK(sc);
2285                         return (m);
2286                 }
2287
2288                 /*
2289                  * Make a deep copy of the packet and enqueue the copy
2290                  * for bridge processing; return the original packet for
2291                  * local processing.
2292                  */
2293                 mc = m_dup(m, M_NOWAIT);
2294                 if (mc == NULL) {
2295                         BRIDGE_UNLOCK(sc);
2296                         return (m);
2297                 }
2298
2299                 /* Perform the bridge forwarding function with the copy. */
2300                 bridge_forward(sc, bif, mc);
2301
2302                 /*
2303                  * Reinject the mbuf as arriving on the bridge so we have a
2304                  * chance at claiming multicast packets. We can not loop back
2305                  * here from ether_input as a bridge is never a member of a
2306                  * bridge.
2307                  */
2308                 KASSERT(bifp->if_bridge == NULL,
2309                     ("loop created in bridge_input"));
2310                 mc2 = m_dup(m, M_NOWAIT);
2311                 if (mc2 != NULL) {
2312                         /* Keep the layer3 header aligned */
2313                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2314                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2315                 }
2316                 if (mc2 != NULL) {
2317                         mc2->m_pkthdr.rcvif = bifp;
2318                         (*bifp->if_input)(bifp, mc2);
2319                 }
2320
2321                 /* Return the original packet for local processing. */
2322                 return (m);
2323         }
2324
2325         if ((bif->bif_flags & IFBIF_STP) &&
2326             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2327                 BRIDGE_UNLOCK(sc);
2328                 return (m);
2329         }
2330
2331 #if (defined(INET) || defined(INET6))
2332 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2333         || ((iface)->if_carp \
2334             && (*carp_forus_p)((iface), eh->ether_dhost))
2335 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2336         || ((iface)->if_carp \
2337             && (*carp_forus_p)((iface), eh->ether_shost))
2338 #else
2339 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2340 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2341 #endif
2342
2343 #ifdef INET6
2344 #   define OR_PFIL_HOOKED_INET6 \
2345         || PFIL_HOOKED(&V_inet6_pfil_hook)
2346 #else
2347 #   define OR_PFIL_HOOKED_INET6
2348 #endif
2349
2350 #define GRAB_OUR_PACKETS(iface) \
2351         if ((iface)->if_type == IFT_GIF) \
2352                 continue; \
2353         /* It is destined for us. */ \
2354         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2355             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2356             ) {                                                         \
2357                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2358                         ETHER_BPF_MTAP(iface, m);                       \
2359                         iface->if_ipackets++;                           \
2360                         iface->if_ibytes += m->m_pkthdr.len;            \
2361                         /* Filter on the physical interface. */         \
2362                         if (pfil_local_phys &&                          \
2363                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2364                              OR_PFIL_HOOKED_INET6)) {                   \
2365                                 if (bridge_pfil(&m, NULL, ifp,          \
2366                                     PFIL_IN) != 0 || m == NULL) {       \
2367                                         BRIDGE_UNLOCK(sc);              \
2368                                         return (NULL);                  \
2369                                 }                                       \
2370                                 eh = mtod(m, struct ether_header *);    \
2371                         }                                               \
2372                 }                                                       \
2373                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2374                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2375                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2376                         if (error && bif->bif_addrmax) {                \
2377                                 BRIDGE_UNLOCK(sc);                      \
2378                                 m_freem(m);                             \
2379                                 return (NULL);                          \
2380                         }                                               \
2381                 }                                                       \
2382                 m->m_pkthdr.rcvif = iface;                              \
2383                 BRIDGE_UNLOCK(sc);                                      \
2384                 return (m);                                             \
2385         }                                                               \
2386                                                                         \
2387         /* We just received a packet that we sent out. */               \
2388         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2389             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2390             ) {                                                         \
2391                 BRIDGE_UNLOCK(sc);                                      \
2392                 m_freem(m);                                             \
2393                 return (NULL);                                          \
2394         }
2395
2396         /*
2397          * Unicast.  Make sure it's not for the bridge.
2398          */
2399         do { GRAB_OUR_PACKETS(bifp) } while (0);
2400
2401         /*
2402          * Give a chance for ifp at first priority. This will help when the
2403          * packet comes through the interface like VLAN's with the same MACs
2404          * on several interfaces from the same bridge. This also will save
2405          * some CPU cycles in case the destination interface and the input
2406          * interface (eq ifp) are the same.
2407          */
2408         do { GRAB_OUR_PACKETS(ifp) } while (0);
2409
2410         /* Now check the all bridge members. */
2411         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2412                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2413         }
2414
2415 #undef OR_CARP_CHECK_WE_ARE_DST
2416 #undef OR_CARP_CHECK_WE_ARE_SRC
2417 #undef OR_PFIL_HOOKED_INET6
2418 #undef GRAB_OUR_PACKETS
2419
2420         /* Perform the bridge forwarding function. */
2421         bridge_forward(sc, bif, m);
2422
2423         return (NULL);
2424 }
2425
2426 /*
2427  * bridge_broadcast:
2428  *
2429  *      Send a frame to all interfaces that are members of
2430  *      the bridge, except for the one on which the packet
2431  *      arrived.
2432  *
2433  *      NOTE: Releases the lock on return.
2434  */
2435 static void
2436 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2437     struct mbuf *m, int runfilt)
2438 {
2439         struct bridge_iflist *dbif, *sbif;
2440         struct mbuf *mc;
2441         struct ifnet *dst_if;
2442         int error = 0, used = 0, i;
2443
2444         sbif = bridge_lookup_member_if(sc, src_if);
2445
2446         BRIDGE_LOCK2REF(sc, error);
2447         if (error) {
2448                 m_freem(m);
2449                 return;
2450         }
2451
2452         /* Filter on the bridge interface before broadcasting */
2453         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2454 #ifdef INET6
2455             || PFIL_HOOKED(&V_inet6_pfil_hook)
2456 #endif
2457             )) {
2458                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2459                         goto out;
2460                 if (m == NULL)
2461                         goto out;
2462         }
2463
2464         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2465                 dst_if = dbif->bif_ifp;
2466                 if (dst_if == src_if)
2467                         continue;
2468
2469                 /* Private segments can not talk to each other */
2470                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2471                         continue;
2472
2473                 if ((dbif->bif_flags & IFBIF_STP) &&
2474                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2475                         continue;
2476
2477                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2478                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2479                         continue;
2480
2481                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2482                         continue;
2483
2484                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2485                         mc = m;
2486                         used = 1;
2487                 } else {
2488                         mc = m_dup(m, M_NOWAIT);
2489                         if (mc == NULL) {
2490                                 sc->sc_ifp->if_oerrors++;
2491                                 continue;
2492                         }
2493                 }
2494
2495                 /*
2496                  * Filter on the output interface. Pass a NULL bridge interface
2497                  * pointer so we do not redundantly filter on the bridge for
2498                  * each interface we broadcast on.
2499                  */
2500                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2501 #ifdef INET6
2502                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2503 #endif
2504                     )) {
2505                         if (used == 0) {
2506                                 /* Keep the layer3 header aligned */
2507                                 i = min(mc->m_pkthdr.len, max_protohdr);
2508                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2509                                 if (mc == NULL) {
2510                                         sc->sc_ifp->if_oerrors++;
2511                                         continue;
2512                                 }
2513                         }
2514                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2515                                 continue;
2516                         if (mc == NULL)
2517                                 continue;
2518                 }
2519
2520                 bridge_enqueue(sc, dst_if, mc);
2521         }
2522         if (used == 0)
2523                 m_freem(m);
2524
2525 out:
2526         BRIDGE_UNREF(sc);
2527 }
2528
2529 /*
2530  * bridge_span:
2531  *
2532  *      Duplicate a packet out one or more interfaces that are in span mode,
2533  *      the original mbuf is unmodified.
2534  */
2535 static void
2536 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2537 {
2538         struct bridge_iflist *bif;
2539         struct ifnet *dst_if;
2540         struct mbuf *mc;
2541
2542         if (LIST_EMPTY(&sc->sc_spanlist))
2543                 return;
2544
2545         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2546                 dst_if = bif->bif_ifp;
2547
2548                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2549                         continue;
2550
2551                 mc = m_copypacket(m, M_NOWAIT);
2552                 if (mc == NULL) {
2553                         sc->sc_ifp->if_oerrors++;
2554                         continue;
2555                 }
2556
2557                 bridge_enqueue(sc, dst_if, mc);
2558         }
2559 }
2560
2561 /*
2562  * bridge_rtupdate:
2563  *
2564  *      Add a bridge routing entry.
2565  */
2566 static int
2567 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2568     struct bridge_iflist *bif, int setflags, uint8_t flags)
2569 {
2570         struct bridge_rtnode *brt;
2571         int error;
2572
2573         BRIDGE_LOCK_ASSERT(sc);
2574
2575         /* Check the source address is valid and not multicast. */
2576         if (ETHER_IS_MULTICAST(dst) ||
2577             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2578              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2579                 return (EINVAL);
2580
2581         /* 802.1p frames map to vlan 1 */
2582         if (vlan == 0)
2583                 vlan = 1;
2584
2585         /*
2586          * A route for this destination might already exist.  If so,
2587          * update it, otherwise create a new one.
2588          */
2589         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2590                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2591                         sc->sc_brtexceeded++;
2592                         return (ENOSPC);
2593                 }
2594                 /* Check per interface address limits (if enabled) */
2595                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2596                         bif->bif_addrexceeded++;
2597                         return (ENOSPC);
2598                 }
2599
2600                 /*
2601                  * Allocate a new bridge forwarding node, and
2602                  * initialize the expiration time and Ethernet
2603                  * address.
2604                  */
2605                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2606                 if (brt == NULL)
2607                         return (ENOMEM);
2608
2609                 if (bif->bif_flags & IFBIF_STICKY)
2610                         brt->brt_flags = IFBAF_STICKY;
2611                 else
2612                         brt->brt_flags = IFBAF_DYNAMIC;
2613
2614                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2615                 brt->brt_vlan = vlan;
2616
2617                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2618                         uma_zfree(bridge_rtnode_zone, brt);
2619                         return (error);
2620                 }
2621                 brt->brt_dst = bif;
2622                 bif->bif_addrcnt++;
2623         }
2624
2625         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2626             brt->brt_dst != bif) {
2627                 brt->brt_dst->bif_addrcnt--;
2628                 brt->brt_dst = bif;
2629                 brt->brt_dst->bif_addrcnt++;
2630         }
2631
2632         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2633                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2634         if (setflags)
2635                 brt->brt_flags = flags;
2636
2637         return (0);
2638 }
2639
2640 /*
2641  * bridge_rtlookup:
2642  *
2643  *      Lookup the destination interface for an address.
2644  */
2645 static struct ifnet *
2646 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2647 {
2648         struct bridge_rtnode *brt;
2649
2650         BRIDGE_LOCK_ASSERT(sc);
2651
2652         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2653                 return (NULL);
2654
2655         return (brt->brt_ifp);
2656 }
2657
2658 /*
2659  * bridge_rttrim:
2660  *
2661  *      Trim the routine table so that we have a number
2662  *      of routing entries less than or equal to the
2663  *      maximum number.
2664  */
2665 static void
2666 bridge_rttrim(struct bridge_softc *sc)
2667 {
2668         struct bridge_rtnode *brt, *nbrt;
2669
2670         BRIDGE_LOCK_ASSERT(sc);
2671
2672         /* Make sure we actually need to do this. */
2673         if (sc->sc_brtcnt <= sc->sc_brtmax)
2674                 return;
2675
2676         /* Force an aging cycle; this might trim enough addresses. */
2677         bridge_rtage(sc);
2678         if (sc->sc_brtcnt <= sc->sc_brtmax)
2679                 return;
2680
2681         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2682                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2683                         bridge_rtnode_destroy(sc, brt);
2684                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2685                                 return;
2686                 }
2687         }
2688 }
2689
2690 /*
2691  * bridge_timer:
2692  *
2693  *      Aging timer for the bridge.
2694  */
2695 static void
2696 bridge_timer(void *arg)
2697 {
2698         struct bridge_softc *sc = arg;
2699
2700         BRIDGE_LOCK_ASSERT(sc);
2701
2702         bridge_rtage(sc);
2703
2704         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2705                 callout_reset(&sc->sc_brcallout,
2706                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2707 }
2708
2709 /*
2710  * bridge_rtage:
2711  *
2712  *      Perform an aging cycle.
2713  */
2714 static void
2715 bridge_rtage(struct bridge_softc *sc)
2716 {
2717         struct bridge_rtnode *brt, *nbrt;
2718
2719         BRIDGE_LOCK_ASSERT(sc);
2720
2721         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2722                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2723                         if (time_uptime >= brt->brt_expire)
2724                                 bridge_rtnode_destroy(sc, brt);
2725                 }
2726         }
2727 }
2728
2729 /*
2730  * bridge_rtflush:
2731  *
2732  *      Remove all dynamic addresses from the bridge.
2733  */
2734 static void
2735 bridge_rtflush(struct bridge_softc *sc, int full)
2736 {
2737         struct bridge_rtnode *brt, *nbrt;
2738
2739         BRIDGE_LOCK_ASSERT(sc);
2740
2741         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2742                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2743                         bridge_rtnode_destroy(sc, brt);
2744         }
2745 }
2746
2747 /*
2748  * bridge_rtdaddr:
2749  *
2750  *      Remove an address from the table.
2751  */
2752 static int
2753 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2754 {
2755         struct bridge_rtnode *brt;
2756         int found = 0;
2757
2758         BRIDGE_LOCK_ASSERT(sc);
2759
2760         /*
2761          * If vlan is zero then we want to delete for all vlans so the lookup
2762          * may return more than one.
2763          */
2764         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2765                 bridge_rtnode_destroy(sc, brt);
2766                 found = 1;
2767         }
2768
2769         return (found ? 0 : ENOENT);
2770 }
2771
2772 /*
2773  * bridge_rtdelete:
2774  *
2775  *      Delete routes to a speicifc member interface.
2776  */
2777 static void
2778 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2779 {
2780         struct bridge_rtnode *brt, *nbrt;
2781
2782         BRIDGE_LOCK_ASSERT(sc);
2783
2784         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2785                 if (brt->brt_ifp == ifp && (full ||
2786                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2787                         bridge_rtnode_destroy(sc, brt);
2788         }
2789 }
2790
2791 /*
2792  * bridge_rtable_init:
2793  *
2794  *      Initialize the route table for this bridge.
2795  */
2796 static void
2797 bridge_rtable_init(struct bridge_softc *sc)
2798 {
2799         int i;
2800
2801         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2802             M_DEVBUF, M_WAITOK);
2803
2804         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2805                 LIST_INIT(&sc->sc_rthash[i]);
2806
2807         sc->sc_rthash_key = arc4random();
2808         LIST_INIT(&sc->sc_rtlist);
2809 }
2810
2811 /*
2812  * bridge_rtable_fini:
2813  *
2814  *      Deconstruct the route table for this bridge.
2815  */
2816 static void
2817 bridge_rtable_fini(struct bridge_softc *sc)
2818 {
2819
2820         KASSERT(sc->sc_brtcnt == 0,
2821             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2822         free(sc->sc_rthash, M_DEVBUF);
2823 }
2824
2825 /*
2826  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2827  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2828  */
2829 #define mix(a, b, c)                                                    \
2830 do {                                                                    \
2831         a -= b; a -= c; a ^= (c >> 13);                                 \
2832         b -= c; b -= a; b ^= (a << 8);                                  \
2833         c -= a; c -= b; c ^= (b >> 13);                                 \
2834         a -= b; a -= c; a ^= (c >> 12);                                 \
2835         b -= c; b -= a; b ^= (a << 16);                                 \
2836         c -= a; c -= b; c ^= (b >> 5);                                  \
2837         a -= b; a -= c; a ^= (c >> 3);                                  \
2838         b -= c; b -= a; b ^= (a << 10);                                 \
2839         c -= a; c -= b; c ^= (b >> 15);                                 \
2840 } while (/*CONSTCOND*/0)
2841
2842 static __inline uint32_t
2843 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2844 {
2845         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2846
2847         b += addr[5] << 8;
2848         b += addr[4];
2849         a += addr[3] << 24;
2850         a += addr[2] << 16;
2851         a += addr[1] << 8;
2852         a += addr[0];
2853
2854         mix(a, b, c);
2855
2856         return (c & BRIDGE_RTHASH_MASK);
2857 }
2858
2859 #undef mix
2860
2861 static int
2862 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2863 {
2864         int i, d;
2865
2866         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2867                 d = ((int)a[i]) - ((int)b[i]);
2868         }
2869
2870         return (d);
2871 }
2872
2873 /*
2874  * bridge_rtnode_lookup:
2875  *
2876  *      Look up a bridge route node for the specified destination. Compare the
2877  *      vlan id or if zero then just return the first match.
2878  */
2879 static struct bridge_rtnode *
2880 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2881 {
2882         struct bridge_rtnode *brt;
2883         uint32_t hash;
2884         int dir;
2885
2886         BRIDGE_LOCK_ASSERT(sc);
2887
2888         hash = bridge_rthash(sc, addr);
2889         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2890                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2891                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2892                         return (brt);
2893                 if (dir > 0)
2894                         return (NULL);
2895         }
2896
2897         return (NULL);
2898 }
2899
2900 /*
2901  * bridge_rtnode_insert:
2902  *
2903  *      Insert the specified bridge node into the route table.  We
2904  *      assume the entry is not already in the table.
2905  */
2906 static int
2907 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2908 {
2909         struct bridge_rtnode *lbrt;
2910         uint32_t hash;
2911         int dir;
2912
2913         BRIDGE_LOCK_ASSERT(sc);
2914
2915         hash = bridge_rthash(sc, brt->brt_addr);
2916
2917         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2918         if (lbrt == NULL) {
2919                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2920                 goto out;
2921         }
2922
2923         do {
2924                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2925                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2926                         return (EEXIST);
2927                 if (dir > 0) {
2928                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2929                         goto out;
2930                 }
2931                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2932                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2933                         goto out;
2934                 }
2935                 lbrt = LIST_NEXT(lbrt, brt_hash);
2936         } while (lbrt != NULL);
2937
2938 #ifdef DIAGNOSTIC
2939         panic("bridge_rtnode_insert: impossible");
2940 #endif
2941
2942 out:
2943         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2944         sc->sc_brtcnt++;
2945
2946         return (0);
2947 }
2948
2949 /*
2950  * bridge_rtnode_destroy:
2951  *
2952  *      Destroy a bridge rtnode.
2953  */
2954 static void
2955 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2956 {
2957         BRIDGE_LOCK_ASSERT(sc);
2958
2959         LIST_REMOVE(brt, brt_hash);
2960
2961         LIST_REMOVE(brt, brt_list);
2962         sc->sc_brtcnt--;
2963         brt->brt_dst->bif_addrcnt--;
2964         uma_zfree(bridge_rtnode_zone, brt);
2965 }
2966
2967 /*
2968  * bridge_rtable_expire:
2969  *
2970  *      Set the expiry time for all routes on an interface.
2971  */
2972 static void
2973 bridge_rtable_expire(struct ifnet *ifp, int age)
2974 {
2975         struct bridge_softc *sc = ifp->if_bridge;
2976         struct bridge_rtnode *brt;
2977
2978         BRIDGE_LOCK(sc);
2979
2980         /*
2981          * If the age is zero then flush, otherwise set all the expiry times to
2982          * age for the interface
2983          */
2984         if (age == 0)
2985                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2986         else {
2987                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2988                         /* Cap the expiry time to 'age' */
2989                         if (brt->brt_ifp == ifp &&
2990                             brt->brt_expire > time_uptime + age &&
2991                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2992                                 brt->brt_expire = time_uptime + age;
2993                 }
2994         }
2995         BRIDGE_UNLOCK(sc);
2996 }
2997
2998 /*
2999  * bridge_state_change:
3000  *
3001  *      Callback from the bridgestp code when a port changes states.
3002  */
3003 static void
3004 bridge_state_change(struct ifnet *ifp, int state)
3005 {
3006         struct bridge_softc *sc = ifp->if_bridge;
3007         static const char *stpstates[] = {
3008                 "disabled",
3009                 "listening",
3010                 "learning",
3011                 "forwarding",
3012                 "blocking",
3013                 "discarding"
3014         };
3015
3016         if (log_stp)
3017                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3018                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3019 }
3020
3021 /*
3022  * Send bridge packets through pfil if they are one of the types pfil can deal
3023  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3024  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3025  * that interface.
3026  */
3027 static int
3028 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3029 {
3030         int snap, error, i, hlen;
3031         struct ether_header *eh1, eh2;
3032         struct ip *ip;
3033         struct llc llc1;
3034         u_int16_t ether_type;
3035
3036         snap = 0;
3037         error = -1;     /* Default error if not error == 0 */
3038
3039 #if 0
3040         /* we may return with the IP fields swapped, ensure its not shared */
3041         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3042 #endif
3043
3044         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
3045                 return (0); /* filtering is disabled */
3046
3047         i = min((*mp)->m_pkthdr.len, max_protohdr);
3048         if ((*mp)->m_len < i) {
3049             *mp = m_pullup(*mp, i);
3050             if (*mp == NULL) {
3051                 printf("%s: m_pullup failed\n", __func__);
3052                 return (-1);
3053             }
3054         }
3055
3056         eh1 = mtod(*mp, struct ether_header *);
3057         ether_type = ntohs(eh1->ether_type);
3058
3059         /*
3060          * Check for SNAP/LLC.
3061          */
3062         if (ether_type < ETHERMTU) {
3063                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3064
3065                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3066                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3067                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3068                     llc2->llc_control == LLC_UI) {
3069                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3070                         snap = 1;
3071                 }
3072         }
3073
3074         /*
3075          * If we're trying to filter bridge traffic, don't look at anything
3076          * other than IP and ARP traffic.  If the filter doesn't understand
3077          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3078          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3079          * but of course we don't have an AppleTalk filter to begin with.
3080          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3081          * ARP traffic.)
3082          */
3083         switch (ether_type) {
3084                 case ETHERTYPE_ARP:
3085                 case ETHERTYPE_REVARP:
3086                         if (pfil_ipfw_arp == 0)
3087                                 return (0); /* Automatically pass */
3088                         break;
3089
3090                 case ETHERTYPE_IP:
3091 #ifdef INET6
3092                 case ETHERTYPE_IPV6:
3093 #endif /* INET6 */
3094                         break;
3095                 default:
3096                         /*
3097                          * Check to see if the user wants to pass non-ip
3098                          * packets, these will not be checked by pfil(9) and
3099                          * passed unconditionally so the default is to drop.
3100                          */
3101                         if (pfil_onlyip)
3102                                 goto bad;
3103         }
3104
3105         /* Run the packet through pfil before stripping link headers */
3106         if (PFIL_HOOKED(&V_link_pfil_hook) && pfil_ipfw != 0 &&
3107                         dir == PFIL_OUT && ifp != NULL) {
3108
3109                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, NULL);
3110
3111                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3112                         return (error);
3113         }
3114
3115         /* Strip off the Ethernet header and keep a copy. */
3116         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3117         m_adj(*mp, ETHER_HDR_LEN);
3118
3119         /* Strip off snap header, if present */
3120         if (snap) {
3121                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3122                 m_adj(*mp, sizeof(struct llc));
3123         }
3124
3125         /*
3126          * Check the IP header for alignment and errors
3127          */
3128         if (dir == PFIL_IN) {
3129                 switch (ether_type) {
3130                         case ETHERTYPE_IP:
3131                                 error = bridge_ip_checkbasic(mp);
3132                                 break;
3133 #ifdef INET6
3134                         case ETHERTYPE_IPV6:
3135                                 error = bridge_ip6_checkbasic(mp);
3136                                 break;
3137 #endif /* INET6 */
3138                         default:
3139                                 error = 0;
3140                 }
3141                 if (error)
3142                         goto bad;
3143         }
3144
3145         error = 0;
3146
3147         /*
3148          * Run the packet through pfil
3149          */
3150         switch (ether_type) {
3151         case ETHERTYPE_IP:
3152                 /*
3153                  * Run pfil on the member interface and the bridge, both can
3154                  * be skipped by clearing pfil_member or pfil_bridge.
3155                  *
3156                  * Keep the order:
3157                  *   in_if -> bridge_if -> out_if
3158                  */
3159                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3160                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3161                                         dir, NULL);
3162
3163                 if (*mp == NULL || error != 0) /* filter may consume */
3164                         break;
3165
3166                 if (pfil_member && ifp != NULL)
3167                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3168                                         dir, NULL);
3169
3170                 if (*mp == NULL || error != 0) /* filter may consume */
3171                         break;
3172
3173                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3174                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3175                                         dir, NULL);
3176
3177                 if (*mp == NULL || error != 0) /* filter may consume */
3178                         break;
3179
3180                 /* check if we need to fragment the packet */
3181                 /* bridge_fragment generates a mbuf chain of packets */
3182                 /* that already include eth headers */
3183                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3184                         i = (*mp)->m_pkthdr.len;
3185                         if (i > ifp->if_mtu) {
3186                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3187                                             &llc1);
3188                                 return (error);
3189                         }
3190                 }
3191
3192                 /* Recalculate the ip checksum. */
3193                 ip = mtod(*mp, struct ip *);
3194                 hlen = ip->ip_hl << 2;
3195                 if (hlen < sizeof(struct ip))
3196                         goto bad;
3197                 if (hlen > (*mp)->m_len) {
3198                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3199                                 goto bad;
3200                         ip = mtod(*mp, struct ip *);
3201                         if (ip == NULL)
3202                                 goto bad;
3203                 }
3204                 ip->ip_sum = 0;
3205                 if (hlen == sizeof(struct ip))
3206                         ip->ip_sum = in_cksum_hdr(ip);
3207                 else
3208                         ip->ip_sum = in_cksum(*mp, hlen);
3209
3210                 break;
3211 #ifdef INET6
3212         case ETHERTYPE_IPV6:
3213                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3214                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3215                                         dir, NULL);
3216
3217                 if (*mp == NULL || error != 0) /* filter may consume */
3218                         break;
3219
3220                 if (pfil_member && ifp != NULL)
3221                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3222                                         dir, NULL);
3223
3224                 if (*mp == NULL || error != 0) /* filter may consume */
3225                         break;
3226
3227                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3228                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3229                                         dir, NULL);
3230                 break;
3231 #endif
3232         default:
3233                 error = 0;
3234                 break;
3235         }
3236
3237         if (*mp == NULL)
3238                 return (error);
3239         if (error != 0)
3240                 goto bad;
3241
3242         error = -1;
3243
3244         /*
3245          * Finally, put everything back the way it was and return
3246          */
3247         if (snap) {
3248                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3249                 if (*mp == NULL)
3250                         return (error);
3251                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3252         }
3253
3254         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3255         if (*mp == NULL)
3256                 return (error);
3257         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3258
3259         return (0);
3260
3261 bad:
3262         m_freem(*mp);
3263         *mp = NULL;
3264         return (error);
3265 }
3266
3267 /*
3268  * Perform basic checks on header size since
3269  * pfil assumes ip_input has already processed
3270  * it for it.  Cut-and-pasted from ip_input.c.
3271  * Given how simple the IPv6 version is,
3272  * does the IPv4 version really need to be
3273  * this complicated?
3274  *
3275  * XXX Should we update ipstat here, or not?
3276  * XXX Right now we update ipstat but not
3277  * XXX csum_counter.
3278  */
3279 static int
3280 bridge_ip_checkbasic(struct mbuf **mp)
3281 {
3282         struct mbuf *m = *mp;
3283         struct ip *ip;
3284         int len, hlen;
3285         u_short sum;
3286
3287         if (*mp == NULL)
3288                 return (-1);
3289
3290         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3291                 if ((m = m_copyup(m, sizeof(struct ip),
3292                         (max_linkhdr + 3) & ~3)) == NULL) {
3293                         /* XXXJRT new stat, please */
3294                         KMOD_IPSTAT_INC(ips_toosmall);
3295                         goto bad;
3296                 }
3297         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3298                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3299                         KMOD_IPSTAT_INC(ips_toosmall);
3300                         goto bad;
3301                 }
3302         }
3303         ip = mtod(m, struct ip *);
3304         if (ip == NULL) goto bad;
3305
3306         if (ip->ip_v != IPVERSION) {
3307                 KMOD_IPSTAT_INC(ips_badvers);
3308                 goto bad;
3309         }
3310         hlen = ip->ip_hl << 2;
3311         if (hlen < sizeof(struct ip)) { /* minimum header length */
3312                 KMOD_IPSTAT_INC(ips_badhlen);
3313                 goto bad;
3314         }
3315         if (hlen > m->m_len) {
3316                 if ((m = m_pullup(m, hlen)) == 0) {
3317                         KMOD_IPSTAT_INC(ips_badhlen);
3318                         goto bad;
3319                 }
3320                 ip = mtod(m, struct ip *);
3321                 if (ip == NULL) goto bad;
3322         }
3323
3324         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3325                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3326         } else {
3327                 if (hlen == sizeof(struct ip)) {
3328                         sum = in_cksum_hdr(ip);
3329                 } else {
3330                         sum = in_cksum(m, hlen);
3331                 }
3332         }
3333         if (sum) {
3334                 KMOD_IPSTAT_INC(ips_badsum);
3335                 goto bad;
3336         }
3337
3338         /* Retrieve the packet length. */
3339         len = ntohs(ip->ip_len);
3340
3341         /*
3342          * Check for additional length bogosity
3343          */
3344         if (len < hlen) {
3345                 KMOD_IPSTAT_INC(ips_badlen);
3346                 goto bad;
3347         }
3348
3349         /*
3350          * Check that the amount of data in the buffers
3351          * is as at least much as the IP header would have us expect.
3352          * Drop packet if shorter than we expect.
3353          */
3354         if (m->m_pkthdr.len < len) {
3355                 KMOD_IPSTAT_INC(ips_tooshort);
3356                 goto bad;
3357         }
3358
3359         /* Checks out, proceed */
3360         *mp = m;
3361         return (0);
3362
3363 bad:
3364         *mp = m;
3365         return (-1);
3366 }
3367
3368 #ifdef INET6
3369 /*
3370  * Same as above, but for IPv6.
3371  * Cut-and-pasted from ip6_input.c.
3372  * XXX Should we update ip6stat, or not?
3373  */
3374 static int
3375 bridge_ip6_checkbasic(struct mbuf **mp)
3376 {
3377         struct mbuf *m = *mp;
3378         struct ip6_hdr *ip6;
3379
3380         /*
3381          * If the IPv6 header is not aligned, slurp it up into a new
3382          * mbuf with space for link headers, in the event we forward
3383          * it.  Otherwise, if it is aligned, make sure the entire base
3384          * IPv6 header is in the first mbuf of the chain.
3385          */
3386         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3387                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3388                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3389                             (max_linkhdr + 3) & ~3)) == NULL) {
3390                         /* XXXJRT new stat, please */
3391                         IP6STAT_INC(ip6s_toosmall);
3392                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3393                         goto bad;
3394                 }
3395         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3396                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3397                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3398                         IP6STAT_INC(ip6s_toosmall);
3399                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3400                         goto bad;
3401                 }
3402         }
3403
3404         ip6 = mtod(m, struct ip6_hdr *);
3405
3406         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3407                 IP6STAT_INC(ip6s_badvers);
3408                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3409                 goto bad;
3410         }
3411
3412         /* Checks out, proceed */
3413         *mp = m;
3414         return (0);
3415
3416 bad:
3417         *mp = m;
3418         return (-1);
3419 }
3420 #endif /* INET6 */
3421
3422 /*
3423  * bridge_fragment:
3424  *
3425  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3426  */
3427 static int
3428 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3429     int snap, struct llc *llc)
3430 {
3431         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3432         struct ip *ip;
3433         int error = -1;
3434
3435         if (m->m_len < sizeof(struct ip) &&
3436             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3437                 goto dropit;
3438         ip = mtod(m, struct ip *);
3439
3440         m->m_pkthdr.csum_flags |= CSUM_IP;
3441         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3442         if (error)
3443                 goto dropit;
3444
3445         /*
3446          * Walk the chain and re-add the Ethernet header for
3447          * each mbuf packet.
3448          */
3449         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3450                 nextpkt = mcur->m_nextpkt;
3451                 mcur->m_nextpkt = NULL;
3452                 if (snap) {
3453                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3454                         if (mcur == NULL) {
3455                                 error = ENOBUFS;
3456                                 if (mprev != NULL)
3457                                         mprev->m_nextpkt = nextpkt;
3458                                 goto dropit;
3459                         }
3460                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3461                 }
3462
3463                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3464                 if (mcur == NULL) {
3465                         error = ENOBUFS;
3466                         if (mprev != NULL)
3467                                 mprev->m_nextpkt = nextpkt;
3468                         goto dropit;
3469                 }
3470                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3471
3472                 /*
3473                  * The previous two M_PREPEND could have inserted one or two
3474                  * mbufs in front so we have to update the previous packet's
3475                  * m_nextpkt.
3476                  */
3477                 mcur->m_nextpkt = nextpkt;
3478                 if (mprev != NULL)
3479                         mprev->m_nextpkt = mcur;
3480                 else {
3481                         /* The first mbuf in the original chain needs to be
3482                          * updated. */
3483                         *mp = mcur;
3484                 }
3485                 mprev = mcur;
3486         }
3487
3488         KMOD_IPSTAT_INC(ips_fragmented);
3489         return (error);
3490
3491 dropit:
3492         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3493                 m = mcur->m_nextpkt;
3494                 m_freem(mcur);
3495         }
3496         return (error);
3497 }
3498
3499 static void
3500 bridge_linkstate(struct ifnet *ifp)
3501 {
3502         struct bridge_softc *sc = ifp->if_bridge;
3503         struct bridge_iflist *bif;
3504
3505         BRIDGE_LOCK(sc);
3506         bif = bridge_lookup_member_if(sc, ifp);
3507         if (bif == NULL) {
3508                 BRIDGE_UNLOCK(sc);
3509                 return;
3510         }
3511         bridge_linkcheck(sc);
3512         BRIDGE_UNLOCK(sc);
3513
3514         bstp_linkstate(&bif->bif_stp);
3515 }
3516
3517 static void
3518 bridge_linkcheck(struct bridge_softc *sc)
3519 {
3520         struct bridge_iflist *bif;
3521         int new_link, hasls;
3522
3523         BRIDGE_LOCK_ASSERT(sc);
3524         new_link = LINK_STATE_DOWN;
3525         hasls = 0;
3526         /* Our link is considered up if at least one of our ports is active */
3527         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3528                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3529                         hasls++;
3530                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3531                         new_link = LINK_STATE_UP;
3532                         break;
3533                 }
3534         }
3535         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3536                 /* If no interfaces support link-state then we default to up */
3537                 new_link = LINK_STATE_UP;
3538         }
3539         if_link_state_change(sc->sc_ifp, new_link);
3540 }