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 #include "opt_carp.h"
  83
  84 #include <sys/param.h>
  85 #include <sys/mbuf.h>
  86 #include <sys/malloc.h>
  87 #include <sys/protosw.h>
  88 #include <sys/systm.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
 112 #include <netinet/in.h> /* for struct arpcom */
 113 #include <netinet/in_systm.h>
 114 #include <netinet/in_var.h>
 115 #include <netinet/ip.h>
 116 #include <netinet/ip_var.h>
 117 #ifdef INET6
 118 #include <netinet/ip6.h>
 119 #include <netinet6/ip6_var.h>
 120 #endif
 121 #ifdef DEV_CARP
 122 #include <netinet/ip_carp.h>
 123 #endif
 124 #include <machine/in_cksum.h>
 125 #include <netinet/if_ether.h> /* for struct arpcom */
 126 #include <net/bridgestp.h>
 127 #include <net/if_bridgevar.h>
 128 #include <net/if_llc.h>
 129 #include <net/if_vlan_var.h>
 130
 131 #include <net/route.h>
 132 #include <netinet/ip_fw.h>
 133 #include <netinet/ip_dummynet.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  * Maximum number of addresses to cache.
 146  */
 147 #ifndef BRIDGE_RTABLE_MAX
 148 #define BRIDGE_RTABLE_MAX               100
 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 mask on the member interface.
 167  */
 168 #define BRIDGE_IFCAPS_MASK              IFCAP_TXCSUM
 169
 170 /*
 171  * Bridge interface list entry.
 172  */
 173 struct bridge_iflist {
 174         LIST_ENTRY(bridge_iflist) bif_next;
 175         struct ifnet            *bif_ifp;       /* member if */
 176         struct bstp_port        bif_stp;        /* STP state */
 177         uint32_t                bif_flags;      /* member if flags */
 178         int                     bif_mutecap;    /* member muted caps */
 179         uint32_t                bif_addrmax;    /* max # of addresses */
 180         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 181         uint32_t                bif_addrexceeded;/* # of address violations */
 182 };
 183
 184 /*
 185  * Bridge route node.
 186  */
 187 struct bridge_rtnode {
 188         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 189         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 190         struct bridge_iflist    *brt_dst;       /* destination if */
 191         unsigned long           brt_expire;     /* expiration time */
 192         uint8_t                 brt_flags;      /* address flags */
 193         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 194         uint16_t                brt_vlan;       /* vlan id */
 195 };
 196 #define brt_ifp                 brt_dst->bif_ifp
 197
 198 /*
 199  * Software state for each bridge.
 200  */
 201 struct bridge_softc {
 202         struct ifnet            *sc_ifp;        /* make this an interface */
 203         LIST_ENTRY(bridge_softc) sc_list;
 204         struct mtx              sc_mtx;
 205         struct cv               sc_cv;
 206         uint32_t                sc_brtmax;      /* max # of addresses */
 207         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 208         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 209         struct callout          sc_brcallout;   /* bridge callout */
 210         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 211         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 212         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 213         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 214         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 215         uint32_t                sc_rthash_key;  /* key for hash */
 216         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 217         struct bstp_state       sc_stp;         /* STP state */
 218         uint32_t                sc_brtexceeded; /* # of cache drops */
 219 };
 220
 221 static struct mtx       bridge_list_mtx;
 222 eventhandler_tag        bridge_detach_cookie = NULL;
 223
 224 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 225
 226 uma_zone_t bridge_rtnode_zone;
 227
 228 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 229 static void     bridge_clone_destroy(struct ifnet *);
 230
 231 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 232 static void     bridge_mutecaps(struct bridge_iflist *, int);
 233 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 234 static void     bridge_init(void *);
 235 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 236 static void     bridge_stop(struct ifnet *, int);
 237 static void     bridge_start(struct ifnet *);
 238 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 239 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 240                     struct rtentry *);
 241 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 242                     struct mbuf *);
 243 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 244
 245 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 246                     struct mbuf *m);
 247
 248 static void     bridge_timer(void *);
 249
 250 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 251                     struct mbuf *, int);
 252 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 253
 254 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 255                     uint16_t, struct bridge_iflist *, int, uint8_t);
 256 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 257                     uint16_t);
 258 static void     bridge_rttrim(struct bridge_softc *);
 259 static void     bridge_rtage(struct bridge_softc *);
 260 static void     bridge_rtflush(struct bridge_softc *, int);
 261 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 262                     uint16_t);
 263
 264 static int      bridge_rtable_init(struct bridge_softc *);
 265 static void     bridge_rtable_fini(struct bridge_softc *);
 266
 267 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 268 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 269                     const uint8_t *, uint16_t);
 270 static int      bridge_rtnode_insert(struct bridge_softc *,
 271                     struct bridge_rtnode *);
 272 static void     bridge_rtnode_destroy(struct bridge_softc *,
 273                     struct bridge_rtnode *);
 274 static void     bridge_rtable_expire(struct ifnet *, int);
 275 static void     bridge_state_change(struct ifnet *, int);
 276
 277 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 278                     const char *name);
 279 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 280                     struct ifnet *ifp);
 281 static void     bridge_delete_member(struct bridge_softc *,
 282                     struct bridge_iflist *, int);
 283 static void     bridge_delete_span(struct bridge_softc *,
 284                     struct bridge_iflist *);
 285
 286 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 287 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 288 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 289 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 290 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 291 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 292 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 293 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 294 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 295 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 296 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 297 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 298 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 317 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 318                     int);
 319 static int      bridge_ip_checkbasic(struct mbuf **mp);
 320 #ifdef INET6
 321 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 322 #endif /* INET6 */
 323 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 324                     struct ether_header *, int, struct llc *);
 325
 326 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 327 #define VLANTAGOF(_m)   \
 328     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 329
 330 static struct bstp_cb_ops bridge_ops = {
 331         .bcb_state = bridge_state_change,
 332         .bcb_rtage = bridge_rtable_expire
 333 };
 334
 335 SYSCTL_DECL(_net_link);
 336 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 337
 338 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 339 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 340 static int pfil_member = 1; /* run pfil hooks on the member interface */
 341 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 342 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 343 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 344                                    locally destined packets */
 345 static int log_stp   = 0;   /* log STP state changes */
 346 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 347     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 348 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 349     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 350 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 351     &pfil_bridge, 0, "Packet filter on the bridge interface");
 352 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 353     &pfil_member, 0, "Packet filter on the member interface");
 354 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 355     &pfil_local_phys, 0,
 356     "Packet filter on the physical interface for locally destined packets");
 357 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 358     &log_stp, 0, "Log STP state changes");
 359
 360 struct bridge_control {
 361         int     (*bc_func)(struct bridge_softc *, void *);
 362         int     bc_argsize;
 363         int     bc_flags;
 364 };
 365
 366 #define BC_F_COPYIN             0x01    /* copy arguments in */
 367 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 368 #define BC_F_SUSER              0x04    /* do super-user check */
 369
 370 const struct bridge_control bridge_control_table[] = {
 371         { bridge_ioctl_add,             sizeof(struct ifbreq),
 372           BC_F_COPYIN|BC_F_SUSER },
 373         { bridge_ioctl_del,             sizeof(struct ifbreq),
 374           BC_F_COPYIN|BC_F_SUSER },
 375
 376         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 377           BC_F_COPYIN|BC_F_COPYOUT },
 378         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 379           BC_F_COPYIN|BC_F_SUSER },
 380
 381         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 382           BC_F_COPYIN|BC_F_SUSER },
 383         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 384           BC_F_COPYOUT },
 385
 386         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 387           BC_F_COPYIN|BC_F_COPYOUT },
 388         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 389           BC_F_COPYIN|BC_F_COPYOUT },
 390
 391         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 392           BC_F_COPYIN|BC_F_SUSER },
 393
 394         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 395           BC_F_COPYIN|BC_F_SUSER },
 396         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 397           BC_F_COPYOUT },
 398
 399         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 400           BC_F_COPYIN|BC_F_SUSER },
 401
 402         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 403           BC_F_COPYIN|BC_F_SUSER },
 404
 405         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 406           BC_F_COPYOUT },
 407         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 408           BC_F_COPYIN|BC_F_SUSER },
 409
 410         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 411           BC_F_COPYOUT },
 412         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 413           BC_F_COPYIN|BC_F_SUSER },
 414
 415         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 416           BC_F_COPYOUT },
 417         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 418           BC_F_COPYIN|BC_F_SUSER },
 419
 420         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 421           BC_F_COPYOUT },
 422         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 423           BC_F_COPYIN|BC_F_SUSER },
 424
 425         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 426           BC_F_COPYIN|BC_F_SUSER },
 427
 428         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 429           BC_F_COPYIN|BC_F_SUSER },
 430
 431         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 432           BC_F_COPYIN|BC_F_SUSER },
 433         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 434           BC_F_COPYIN|BC_F_SUSER },
 435
 436         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 437           BC_F_COPYOUT },
 438
 439         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 440           BC_F_COPYOUT },
 441
 442         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 443           BC_F_COPYIN|BC_F_COPYOUT },
 444
 445         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 446           BC_F_COPYIN|BC_F_SUSER },
 447
 448         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 449           BC_F_COPYIN|BC_F_SUSER },
 450
 451         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 452           BC_F_COPYIN|BC_F_SUSER },
 453
 454 };
 455 const int bridge_control_table_size =
 456     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 457
 458 LIST_HEAD(, bridge_softc) bridge_list;
 459
 460 IFC_SIMPLE_DECLARE(bridge, 0);
 461
 462 static int
 463 bridge_modevent(module_t mod, int type, void *data)
 464 {
 465
 466         switch (type) {
 467         case MOD_LOAD:
 468                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 469                 if_clone_attach(&bridge_cloner);
 470                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 471                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 472                     UMA_ALIGN_PTR, 0);
 473                 LIST_INIT(&bridge_list);
 474                 bridge_input_p = bridge_input;
 475                 bridge_output_p = bridge_output;
 476                 bridge_dn_p = bridge_dummynet;
 477                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 478                     ifnet_departure_event, bridge_ifdetach, NULL,
 479                     EVENTHANDLER_PRI_ANY);
 480                 break;
 481         case MOD_UNLOAD:
 482                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 483                     bridge_detach_cookie);
 484                 if_clone_detach(&bridge_cloner);
 485                 uma_zdestroy(bridge_rtnode_zone);
 486                 bridge_input_p = NULL;
 487                 bridge_output_p = NULL;
 488                 bridge_dn_p = NULL;
 489                 mtx_destroy(&bridge_list_mtx);
 490                 break;
 491         default:
 492                 return (EOPNOTSUPP);
 493         }
 494         return (0);
 495 }
 496
 497 static moduledata_t bridge_mod = {
 498         "if_bridge",
 499         bridge_modevent,
 500         0
 501 };
 502
 503 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 504 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 505
 506 /*
 507  * handler for net.link.bridge.pfil_ipfw
 508  */
 509 static int
 510 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 511 {
 512         int enable = pfil_ipfw;
 513         int error;
 514
 515         error = sysctl_handle_int(oidp, &enable, 0, req);
 516         enable = (enable) ? 1 : 0;
 517
 518         if (enable != pfil_ipfw) {
 519                 pfil_ipfw = enable;
 520
 521                 /*
 522                  * Disable pfil so that ipfw doesnt run twice, if the user
 523                  * really wants both then they can re-enable pfil_bridge and/or
 524                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 525                  * layer2 type.
 526                  */
 527                 if (pfil_ipfw) {
 528                         pfil_onlyip = 0;
 529                         pfil_bridge = 0;
 530                         pfil_member = 0;
 531                 }
 532         }
 533
 534         return (error);
 535 }
 536 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 537             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 538
 539 /*
 540  * bridge_clone_create:
 541  *
 542  *      Create a new bridge instance.
 543  */
 544 static int
 545 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 546 {
 547         struct bridge_softc *sc, *sc2;
 548         struct ifnet *bifp, *ifp;
 549         u_char eaddr[6];
 550         int retry;
 551
 552         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 553         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 554         if (ifp == NULL) {
 555                 free(sc, M_DEVBUF);
 556                 return (ENOSPC);
 557         }
 558
 559         BRIDGE_LOCK_INIT(sc);
 560         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 561         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 562
 563         /* Initialize our routing table. */
 564         bridge_rtable_init(sc);
 565
 566         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 567
 568         LIST_INIT(&sc->sc_iflist);
 569         LIST_INIT(&sc->sc_spanlist);
 570
 571         ifp->if_softc = sc;
 572         if_initname(ifp, ifc->ifc_name, unit);
 573         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 574         ifp->if_ioctl = bridge_ioctl;
 575         ifp->if_start = bridge_start;
 576         ifp->if_init = bridge_init;
 577         ifp->if_type = IFT_BRIDGE;
 578         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 579         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 580         IFQ_SET_READY(&ifp->if_snd);
 581
 582         /*
 583          * Generate a random ethernet address with a locally administered
 584          * address.
 585          *
 586          * Since we are using random ethernet addresses for the bridge, it is
 587          * possible that we might have address collisions, so make sure that
 588          * this hardware address isn't already in use on another bridge.
 589          */
 590         for (retry = 1; retry != 0;) {
 591                 arc4rand(eaddr, ETHER_ADDR_LEN, 1);
 592                 eaddr[0] &= ~1;         /* clear multicast bit */
 593                 eaddr[0] |= 2;          /* set the LAA bit */
 594                 retry = 0;
 595                 mtx_lock(&bridge_list_mtx);
 596                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 597                         bifp = sc2->sc_ifp;
 598                         if (memcmp(eaddr, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 599                                 retry = 1;
 600                 }
 601                 mtx_unlock(&bridge_list_mtx);
 602         }
 603
 604         bstp_attach(&sc->sc_stp, &bridge_ops);
 605         ether_ifattach(ifp, eaddr);
 606         /* Now undo some of the damage... */
 607         ifp->if_baudrate = 0;
 608         ifp->if_type = IFT_BRIDGE;
 609
 610         mtx_lock(&bridge_list_mtx);
 611         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 612         mtx_unlock(&bridge_list_mtx);
 613
 614         return (0);
 615 }
 616
 617 /*
 618  * bridge_clone_destroy:
 619  *
 620  *      Destroy a bridge instance.
 621  */
 622 static void
 623 bridge_clone_destroy(struct ifnet *ifp)
 624 {
 625         struct bridge_softc *sc = ifp->if_softc;
 626         struct bridge_iflist *bif;
 627
 628         BRIDGE_LOCK(sc);
 629
 630         bridge_stop(ifp, 1);
 631         ifp->if_flags &= ~IFF_UP;
 632
 633         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 634                 bridge_delete_member(sc, bif, 0);
 635
 636         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 637                 bridge_delete_span(sc, bif);
 638         }
 639
 640         BRIDGE_UNLOCK(sc);
 641
 642         callout_drain(&sc->sc_brcallout);
 643
 644         mtx_lock(&bridge_list_mtx);
 645         LIST_REMOVE(sc, sc_list);
 646         mtx_unlock(&bridge_list_mtx);
 647
 648         bstp_detach(&sc->sc_stp);
 649         ether_ifdetach(ifp);
 650         if_free_type(ifp, IFT_ETHER);
 651
 652         /* Tear down the routing table. */
 653         bridge_rtable_fini(sc);
 654
 655         BRIDGE_LOCK_DESTROY(sc);
 656         free(sc, M_DEVBUF);
 657 }
 658
 659 /*
 660  * bridge_ioctl:
 661  *
 662  *      Handle a control request from the operator.
 663  */
 664 static int
 665 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 666 {
 667         struct bridge_softc *sc = ifp->if_softc;
 668         struct thread *td = curthread;
 669         union {
 670                 struct ifbreq ifbreq;
 671                 struct ifbifconf ifbifconf;
 672                 struct ifbareq ifbareq;
 673                 struct ifbaconf ifbaconf;
 674                 struct ifbrparam ifbrparam;
 675                 struct ifbropreq ifbropreq;
 676         } args;
 677         struct ifdrv *ifd = (struct ifdrv *) data;
 678         const struct bridge_control *bc;
 679         int error = 0;
 680
 681         switch (cmd) {
 682
 683         case SIOCADDMULTI:
 684         case SIOCDELMULTI:
 685                 break;
 686
 687         case SIOCGDRVSPEC:
 688         case SIOCSDRVSPEC:
 689                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 690                         error = EINVAL;
 691                         break;
 692                 }
 693                 bc = &bridge_control_table[ifd->ifd_cmd];
 694
 695                 if (cmd == SIOCGDRVSPEC &&
 696                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 697                         error = EINVAL;
 698                         break;
 699                 }
 700                 else if (cmd == SIOCSDRVSPEC &&
 701                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 702                         error = EINVAL;
 703                         break;
 704                 }
 705
 706                 if (bc->bc_flags & BC_F_SUSER) {
 707                         error = priv_check(td, PRIV_NET_BRIDGE);
 708                         if (error)
 709                                 break;
 710                 }
 711
 712                 if (ifd->ifd_len != bc->bc_argsize ||
 713                     ifd->ifd_len > sizeof(args)) {
 714                         error = EINVAL;
 715                         break;
 716                 }
 717
 718                 bzero(&args, sizeof(args));
 719                 if (bc->bc_flags & BC_F_COPYIN) {
 720                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 721                         if (error)
 722                                 break;
 723                 }
 724
 725                 BRIDGE_LOCK(sc);
 726                 error = (*bc->bc_func)(sc, &args);
 727                 BRIDGE_UNLOCK(sc);
 728                 if (error)
 729                         break;
 730
 731                 if (bc->bc_flags & BC_F_COPYOUT)
 732                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 733
 734                 break;
 735
 736         case SIOCSIFFLAGS:
 737                 if (!(ifp->if_flags & IFF_UP) &&
 738                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 739                         /*
 740                          * If interface is marked down and it is running,
 741                          * then stop and disable it.
 742                          */
 743                         BRIDGE_LOCK(sc);
 744                         bridge_stop(ifp, 1);
 745                         BRIDGE_UNLOCK(sc);
 746                 } else if ((ifp->if_flags & IFF_UP) &&
 747                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 748                         /*
 749                          * If interface is marked up and it is stopped, then
 750                          * start it.
 751                          */
 752                         (*ifp->if_init)(sc);
 753                 }
 754                 break;
 755
 756         case SIOCSIFMTU:
 757                 /* Do not allow the MTU to be changed on the bridge */
 758                 error = EINVAL;
 759                 break;
 760
 761         default:
 762                 /*
 763                  * drop the lock as ether_ioctl() will call bridge_start() and
 764                  * cause the lock to be recursed.
 765                  */
 766                 error = ether_ioctl(ifp, cmd, data);
 767                 break;
 768         }
 769
 770         return (error);
 771 }
 772
 773 /*
 774  * bridge_mutecaps:
 775  *
 776  *      Clear or restore unwanted capabilities on the member interface
 777  */
 778 static void
 779 bridge_mutecaps(struct bridge_iflist *bif, int mute)
 780 {
 781         struct ifnet *ifp = bif->bif_ifp;
 782         struct ifreq ifr;
 783         int error;
 784
 785         if (ifp->if_ioctl == NULL)
 786                 return;
 787
 788         bzero(&ifr, sizeof(ifr));
 789         ifr.ifr_reqcap = ifp->if_capenable;
 790
 791         if (mute) {
 792                 /* mask off and save capabilities */
 793                 bif->bif_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
 794                 if (bif->bif_mutecap != 0)
 795                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
 796         } else
 797                 /* restore muted capabilities */
 798                 ifr.ifr_reqcap |= bif->bif_mutecap;
 799
 800
 801         if (bif->bif_mutecap != 0) {
 802                 IFF_LOCKGIANT(ifp);
 803                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 804                 IFF_UNLOCKGIANT(ifp);
 805         }
 806 }
 807
 808 /*
 809  * bridge_lookup_member:
 810  *
 811  *      Lookup a bridge member interface.
 812  */
 813 static struct bridge_iflist *
 814 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 815 {
 816         struct bridge_iflist *bif;
 817         struct ifnet *ifp;
 818
 819         BRIDGE_LOCK_ASSERT(sc);
 820
 821         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 822                 ifp = bif->bif_ifp;
 823                 if (strcmp(ifp->if_xname, name) == 0)
 824                         return (bif);
 825         }
 826
 827         return (NULL);
 828 }
 829
 830 /*
 831  * bridge_lookup_member_if:
 832  *
 833  *      Lookup a bridge member interface by ifnet*.
 834  */
 835 static struct bridge_iflist *
 836 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 837 {
 838         struct bridge_iflist *bif;
 839
 840         BRIDGE_LOCK_ASSERT(sc);
 841
 842         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 843                 if (bif->bif_ifp == member_ifp)
 844                         return (bif);
 845         }
 846
 847         return (NULL);
 848 }
 849
 850 /*
 851  * bridge_delete_member:
 852  *
 853  *      Delete the specified member interface.
 854  */
 855 static void
 856 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 857     int gone)
 858 {
 859         struct ifnet *ifs = bif->bif_ifp;
 860
 861         BRIDGE_LOCK_ASSERT(sc);
 862
 863         if (!gone) {
 864                 switch (ifs->if_type) {
 865                 case IFT_ETHER:
 866                 case IFT_L2VLAN:
 867                         /*
 868                          * Take the interface out of promiscuous mode.
 869                          */
 870                         (void) ifpromisc(ifs, 0);
 871                         bridge_mutecaps(bif, 0);
 872                         break;
 873
 874                 case IFT_GIF:
 875                         break;
 876
 877                 default:
 878 #ifdef DIAGNOSTIC
 879                         panic("bridge_delete_member: impossible");
 880 #endif
 881                         break;
 882                 }
 883         }
 884
 885         if (bif->bif_flags & IFBIF_STP)
 886                 bstp_disable(&bif->bif_stp);
 887
 888         ifs->if_bridge = NULL;
 889         BRIDGE_XLOCK(sc);
 890         LIST_REMOVE(bif, bif_next);
 891         BRIDGE_XDROP(sc);
 892
 893         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 894         KASSERT(bif->bif_addrcnt == 0,
 895             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 896
 897         BRIDGE_UNLOCK(sc);
 898         bstp_destroy(&bif->bif_stp);    /* prepare to free */
 899         BRIDGE_LOCK(sc);
 900         free(bif, M_DEVBUF);
 901 }
 902
 903 /*
 904  * bridge_delete_span:
 905  *
 906  *      Delete the specified span interface.
 907  */
 908 static void
 909 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
 910 {
 911         BRIDGE_LOCK_ASSERT(sc);
 912
 913         KASSERT(bif->bif_ifp->if_bridge == NULL,
 914             ("%s: not a span interface", __func__));
 915
 916         LIST_REMOVE(bif, bif_next);
 917         free(bif, M_DEVBUF);
 918 }
 919
 920 static int
 921 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
 922 {
 923         struct ifbreq *req = arg;
 924         struct bridge_iflist *bif = NULL;
 925         struct ifnet *ifs;
 926         int error = 0;
 927
 928         ifs = ifunit(req->ifbr_ifsname);
 929         if (ifs == NULL)
 930                 return (ENOENT);
 931
 932         /* If it's in the span list, it can't be a member. */
 933         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
 934                 if (ifs == bif->bif_ifp)
 935                         return (EBUSY);
 936
 937         /* Allow the first Ethernet member to define the MTU */
 938         if (ifs->if_type != IFT_GIF) {
 939                 if (LIST_EMPTY(&sc->sc_iflist))
 940                         sc->sc_ifp->if_mtu = ifs->if_mtu;
 941                 else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
 942                         if_printf(sc->sc_ifp, "invalid MTU for %s\n",
 943                             ifs->if_xname);
 944                         return (EINVAL);
 945                 }
 946         }
 947
 948         if (ifs->if_bridge == sc)
 949                 return (EEXIST);
 950
 951         if (ifs->if_bridge != NULL)
 952                 return (EBUSY);
 953
 954         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
 955         if (bif == NULL)
 956                 return (ENOMEM);
 957
 958         bif->bif_ifp = ifs;
 959         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
 960
 961         switch (ifs->if_type) {
 962         case IFT_ETHER:
 963         case IFT_L2VLAN:
 964                 /*
 965                  * Place the interface into promiscuous mode.
 966                  */
 967                 error = ifpromisc(ifs, 1);
 968                 if (error)
 969                         goto out;
 970
 971                 bridge_mutecaps(bif, 1);
 972                 break;
 973
 974         case IFT_GIF:
 975                 break;
 976
 977         default:
 978                 error = EINVAL;
 979                 goto out;
 980         }
 981
 982         ifs->if_bridge = sc;
 983         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
 984         /*
 985          * XXX: XLOCK HERE!?!
 986          *
 987          * NOTE: insert_***HEAD*** should be safe for the traversals.
 988          */
 989         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
 990
 991 out:
 992         if (error) {
 993                 if (bif != NULL)
 994                         free(bif, M_DEVBUF);
 995         }
 996         return (error);
 997 }
 998
 999 static int
1000 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1001 {
1002         struct ifbreq *req = arg;
1003         struct bridge_iflist *bif;
1004
1005         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1006         if (bif == NULL)
1007                 return (ENOENT);
1008
1009         bridge_delete_member(sc, bif, 0);
1010
1011         return (0);
1012 }
1013
1014 static int
1015 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1016 {
1017         struct ifbreq *req = arg;
1018         struct bridge_iflist *bif;
1019         struct bstp_port *bp;
1020
1021         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1022         if (bif == NULL)
1023                 return (ENOENT);
1024
1025         bp = &bif->bif_stp;
1026         req->ifbr_ifsflags = bif->bif_flags;
1027         req->ifbr_state = bp->bp_state;
1028         req->ifbr_priority = bp->bp_priority;
1029         req->ifbr_path_cost = bp->bp_path_cost;
1030         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1031         req->ifbr_proto = bp->bp_protover;
1032         req->ifbr_role = bp->bp_role;
1033         req->ifbr_stpflags = bp->bp_flags;
1034         req->ifbr_addrcnt = bif->bif_addrcnt;
1035         req->ifbr_addrmax = bif->bif_addrmax;
1036         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1037
1038         /* Copy STP state options as flags */
1039         if (bp->bp_operedge)
1040                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1041         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1042                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1043         if (bp->bp_ptp_link)
1044                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1045         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1046                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1047         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1048                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1049         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1050                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1051         return (0);
1052 }
1053
1054 static int
1055 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1056 {
1057         struct ifbreq *req = arg;
1058         struct bridge_iflist *bif;
1059         struct bstp_port *bp;
1060         int error;
1061
1062         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1063         if (bif == NULL)
1064                 return (ENOENT);
1065         bp = &bif->bif_stp;
1066
1067         if (req->ifbr_ifsflags & IFBIF_SPAN)
1068                 /* SPAN is readonly */
1069                 return (EINVAL);
1070
1071         if (req->ifbr_ifsflags & IFBIF_STP) {
1072                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1073                         error = bstp_enable(&bif->bif_stp);
1074                         if (error)
1075                                 return (error);
1076                 }
1077         } else {
1078                 if ((bif->bif_flags & IFBIF_STP) != 0)
1079                         bstp_disable(&bif->bif_stp);
1080         }
1081
1082         /* Pass on STP flags */
1083         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1084         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1085         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1086         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1087
1088         /* Save the bits relating to the bridge */
1089         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1090
1091         return (0);
1092 }
1093
1094 static int
1095 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1096 {
1097         struct ifbrparam *param = arg;
1098
1099         sc->sc_brtmax = param->ifbrp_csize;
1100         bridge_rttrim(sc);
1101
1102         return (0);
1103 }
1104
1105 static int
1106 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1107 {
1108         struct ifbrparam *param = arg;
1109
1110         param->ifbrp_csize = sc->sc_brtmax;
1111
1112         return (0);
1113 }
1114
1115 static int
1116 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1117 {
1118         struct ifbifconf *bifc = arg;
1119         struct bridge_iflist *bif;
1120         struct ifbreq breq;
1121         char *buf, *outbuf;
1122         int count, buflen, len, error = 0;
1123
1124         count = 0;
1125         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1126                 count++;
1127         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1128                 count++;
1129
1130         buflen = sizeof(breq) * count;
1131         if (bifc->ifbic_len == 0) {
1132                 bifc->ifbic_len = buflen;
1133                 return (0);
1134         }
1135         BRIDGE_UNLOCK(sc);
1136         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1137         BRIDGE_LOCK(sc);
1138
1139         count = 0;
1140         buf = outbuf;
1141         len = min(bifc->ifbic_len, buflen);
1142         bzero(&breq, sizeof(breq));
1143         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1144                 if (len < sizeof(breq))
1145                         break;
1146
1147                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1148                     sizeof(breq.ifbr_ifsname));
1149                 /* Fill in the ifbreq structure */
1150                 error = bridge_ioctl_gifflags(sc, &breq);
1151                 if (error)
1152                         break;
1153                 memcpy(buf, &breq, sizeof(breq));
1154                 count++;
1155                 buf += sizeof(breq);
1156                 len -= sizeof(breq);
1157         }
1158         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1159                 if (len < sizeof(breq))
1160                         break;
1161
1162                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1163                     sizeof(breq.ifbr_ifsname));
1164                 breq.ifbr_ifsflags = bif->bif_flags;
1165                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1166                 memcpy(buf, &breq, sizeof(breq));
1167                 count++;
1168                 buf += sizeof(breq);
1169                 len -= sizeof(breq);
1170         }
1171
1172         BRIDGE_UNLOCK(sc);
1173         bifc->ifbic_len = sizeof(breq) * count;
1174         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1175         BRIDGE_LOCK(sc);
1176         free(outbuf, M_TEMP);
1177         return (error);
1178 }
1179
1180 static int
1181 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1182 {
1183         struct ifbaconf *bac = arg;
1184         struct bridge_rtnode *brt;
1185         struct ifbareq bareq;
1186         char *buf, *outbuf;
1187         int count, buflen, len, error = 0;
1188
1189         if (bac->ifbac_len == 0)
1190                 return (0);
1191
1192         count = 0;
1193         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1194                 count++;
1195         buflen = sizeof(bareq) * count;
1196
1197         BRIDGE_UNLOCK(sc);
1198         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1199         BRIDGE_LOCK(sc);
1200
1201         count = 0;
1202         buf = outbuf;
1203         len = min(bac->ifbac_len, buflen);
1204         bzero(&bareq, sizeof(bareq));
1205         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1206                 if (len < sizeof(bareq))
1207                         goto out;
1208                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1209                     sizeof(bareq.ifba_ifsname));
1210                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1211                 bareq.ifba_vlan = brt->brt_vlan;
1212                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1213                                 time_uptime < brt->brt_expire)
1214                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1215                 else
1216                         bareq.ifba_expire = 0;
1217                 bareq.ifba_flags = brt->brt_flags;
1218
1219                 memcpy(buf, &bareq, sizeof(bareq));
1220                 count++;
1221                 buf += sizeof(bareq);
1222                 len -= sizeof(bareq);
1223         }
1224 out:
1225         BRIDGE_UNLOCK(sc);
1226         bac->ifbac_len = sizeof(bareq) * count;
1227         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1228         BRIDGE_LOCK(sc);
1229         free(outbuf, M_TEMP);
1230         return (error);
1231 }
1232
1233 static int
1234 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1235 {
1236         struct ifbareq *req = arg;
1237         struct bridge_iflist *bif;
1238         int error;
1239
1240         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1241         if (bif == NULL)
1242                 return (ENOENT);
1243
1244         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1245             req->ifba_flags);
1246
1247         return (error);
1248 }
1249
1250 static int
1251 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1252 {
1253         struct ifbrparam *param = arg;
1254
1255         sc->sc_brttimeout = param->ifbrp_ctime;
1256         return (0);
1257 }
1258
1259 static int
1260 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1261 {
1262         struct ifbrparam *param = arg;
1263
1264         param->ifbrp_ctime = sc->sc_brttimeout;
1265         return (0);
1266 }
1267
1268 static int
1269 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1270 {
1271         struct ifbareq *req = arg;
1272
1273         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1274 }
1275
1276 static int
1277 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1278 {
1279         struct ifbreq *req = arg;
1280
1281         bridge_rtflush(sc, req->ifbr_ifsflags);
1282         return (0);
1283 }
1284
1285 static int
1286 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1287 {
1288         struct ifbrparam *param = arg;
1289         struct bstp_state *bs = &sc->sc_stp;
1290
1291         param->ifbrp_prio = bs->bs_bridge_priority;
1292         return (0);
1293 }
1294
1295 static int
1296 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1297 {
1298         struct ifbrparam *param = arg;
1299
1300         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1301 }
1302
1303 static int
1304 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1305 {
1306         struct ifbrparam *param = arg;
1307         struct bstp_state *bs = &sc->sc_stp;
1308
1309         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1310         return (0);
1311 }
1312
1313 static int
1314 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1315 {
1316         struct ifbrparam *param = arg;
1317
1318         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1319 }
1320
1321 static int
1322 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1323 {
1324         struct ifbrparam *param = arg;
1325         struct bstp_state *bs = &sc->sc_stp;
1326
1327         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1328         return (0);
1329 }
1330
1331 static int
1332 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1333 {
1334         struct ifbrparam *param = arg;
1335
1336         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1337 }
1338
1339 static int
1340 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1341 {
1342         struct ifbrparam *param = arg;
1343         struct bstp_state *bs = &sc->sc_stp;
1344
1345         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1346         return (0);
1347 }
1348
1349 static int
1350 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1351 {
1352         struct ifbrparam *param = arg;
1353
1354         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1355 }
1356
1357 static int
1358 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1359 {
1360         struct ifbreq *req = arg;
1361         struct bridge_iflist *bif;
1362
1363         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1364         if (bif == NULL)
1365                 return (ENOENT);
1366
1367         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1368 }
1369
1370 static int
1371 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1372 {
1373         struct ifbreq *req = arg;
1374         struct bridge_iflist *bif;
1375
1376         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1377         if (bif == NULL)
1378                 return (ENOENT);
1379
1380         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1381 }
1382
1383 static int
1384 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1385 {
1386         struct ifbreq *req = arg;
1387         struct bridge_iflist *bif;
1388
1389         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1390         if (bif == NULL)
1391                 return (ENOENT);
1392
1393         bif->bif_addrmax = req->ifbr_addrmax;
1394         return (0);
1395 }
1396
1397 static int
1398 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1399 {
1400         struct ifbreq *req = arg;
1401         struct bridge_iflist *bif = NULL;
1402         struct ifnet *ifs;
1403
1404         ifs = ifunit(req->ifbr_ifsname);
1405         if (ifs == NULL)
1406                 return (ENOENT);
1407
1408         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1409                 if (ifs == bif->bif_ifp)
1410                         return (EBUSY);
1411
1412         if (ifs->if_bridge != NULL)
1413                 return (EBUSY);
1414
1415         switch (ifs->if_type) {
1416                 case IFT_ETHER:
1417                 case IFT_GIF:
1418                 case IFT_L2VLAN:
1419                         break;
1420                 default:
1421                         return (EINVAL);
1422         }
1423
1424         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1425         if (bif == NULL)
1426                 return (ENOMEM);
1427
1428         bif->bif_ifp = ifs;
1429         bif->bif_flags = IFBIF_SPAN;
1430
1431         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1432
1433         return (0);
1434 }
1435
1436 static int
1437 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1438 {
1439         struct ifbreq *req = arg;
1440         struct bridge_iflist *bif;
1441         struct ifnet *ifs;
1442
1443         ifs = ifunit(req->ifbr_ifsname);
1444         if (ifs == NULL)
1445                 return (ENOENT);
1446
1447         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1448                 if (ifs == bif->bif_ifp)
1449                         break;
1450
1451         if (bif == NULL)
1452                 return (ENOENT);
1453
1454         bridge_delete_span(sc, bif);
1455
1456         return (0);
1457 }
1458
1459 static int
1460 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1461 {
1462         struct ifbropreq *req = arg;
1463         struct bstp_state *bs = &sc->sc_stp;
1464         struct bstp_port *root_port;
1465
1466         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1467         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1468         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1469
1470         root_port = bs->bs_root_port;
1471         if (root_port == NULL)
1472                 req->ifbop_root_port = 0;
1473         else
1474                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1475
1476         req->ifbop_holdcount = bs->bs_txholdcount;
1477         req->ifbop_priority = bs->bs_bridge_priority;
1478         req->ifbop_protocol = bs->bs_protover;
1479         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1480         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1481         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1482         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1483         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1484         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1485
1486         return (0);
1487 }
1488
1489 static int
1490 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1491 {
1492         struct ifbrparam *param = arg;
1493
1494         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1495         return (0);
1496 }
1497
1498 static int
1499 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1500 {
1501         struct ifbpstpconf *bifstp = arg;
1502         struct bridge_iflist *bif;
1503         struct bstp_port *bp;
1504         struct ifbpstpreq bpreq;
1505         char *buf, *outbuf;
1506         int count, buflen, len, error = 0;
1507
1508         count = 0;
1509         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1510                 if ((bif->bif_flags & IFBIF_STP) != 0)
1511                         count++;
1512         }
1513
1514         buflen = sizeof(bpreq) * count;
1515         if (bifstp->ifbpstp_len == 0) {
1516                 bifstp->ifbpstp_len = buflen;
1517                 return (0);
1518         }
1519
1520         BRIDGE_UNLOCK(sc);
1521         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1522         BRIDGE_LOCK(sc);
1523
1524         count = 0;
1525         buf = outbuf;
1526         len = min(bifstp->ifbpstp_len, buflen);
1527         bzero(&bpreq, sizeof(bpreq));
1528         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1529                 if (len < sizeof(bpreq))
1530                         break;
1531
1532                 if ((bif->bif_flags & IFBIF_STP) == 0)
1533                         continue;
1534
1535                 bp = &bif->bif_stp;
1536                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1537                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1538                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1539                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1540                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1541                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1542
1543                 memcpy(buf, &bpreq, sizeof(bpreq));
1544                 count++;
1545                 buf += sizeof(bpreq);
1546                 len -= sizeof(bpreq);
1547         }
1548
1549         BRIDGE_UNLOCK(sc);
1550         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1551         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1552         BRIDGE_LOCK(sc);
1553         free(outbuf, M_TEMP);
1554         return (error);
1555 }
1556
1557 static int
1558 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1559 {
1560         struct ifbrparam *param = arg;
1561
1562         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1563 }
1564
1565 static int
1566 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1567 {
1568         struct ifbrparam *param = arg;
1569
1570         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1571 }
1572
1573 /*
1574  * bridge_ifdetach:
1575  *
1576  *      Detach an interface from a bridge.  Called when a member
1577  *      interface is detaching.
1578  */
1579 static void
1580 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1581 {
1582         struct bridge_softc *sc = ifp->if_bridge;
1583         struct bridge_iflist *bif;
1584
1585         /* Check if the interface is a bridge member */
1586         if (sc != NULL) {
1587                 BRIDGE_LOCK(sc);
1588
1589                 bif = bridge_lookup_member_if(sc, ifp);
1590                 if (bif != NULL)
1591                         bridge_delete_member(sc, bif, 1);
1592
1593                 BRIDGE_UNLOCK(sc);
1594                 return;
1595         }
1596
1597         /* Check if the interface is a span port */
1598         mtx_lock(&bridge_list_mtx);
1599         LIST_FOREACH(sc, &bridge_list, sc_list) {
1600                 BRIDGE_LOCK(sc);
1601                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1602                         if (ifp == bif->bif_ifp) {
1603                                 bridge_delete_span(sc, bif);
1604                                 break;
1605                         }
1606
1607                 BRIDGE_UNLOCK(sc);
1608         }
1609         mtx_unlock(&bridge_list_mtx);
1610 }
1611
1612 /*
1613  * bridge_init:
1614  *
1615  *      Initialize a bridge interface.
1616  */
1617 static void
1618 bridge_init(void *xsc)
1619 {
1620         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1621         struct ifnet *ifp = sc->sc_ifp;
1622
1623         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1624                 return;
1625
1626         BRIDGE_LOCK(sc);
1627         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1628             bridge_timer, sc);
1629
1630         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1631         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1632
1633         BRIDGE_UNLOCK(sc);
1634 }
1635
1636 /*
1637  * bridge_stop:
1638  *
1639  *      Stop the bridge interface.
1640  */
1641 static void
1642 bridge_stop(struct ifnet *ifp, int disable)
1643 {
1644         struct bridge_softc *sc = ifp->if_softc;
1645
1646         BRIDGE_LOCK_ASSERT(sc);
1647
1648         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1649                 return;
1650
1651         callout_stop(&sc->sc_brcallout);
1652         bstp_stop(&sc->sc_stp);
1653
1654         bridge_rtflush(sc, IFBF_FLUSHDYN);
1655
1656         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1657 }
1658
1659 /*
1660  * bridge_enqueue:
1661  *
1662  *      Enqueue a packet on a bridge member interface.
1663  *
1664  */
1665 static void
1666 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1667 {
1668         int len, err = 0;
1669         short mflags;
1670         struct mbuf *m0;
1671
1672         len = m->m_pkthdr.len;
1673         mflags = m->m_flags;
1674
1675         /* We may be sending a fragment so traverse the mbuf */
1676         for (; m; m = m0) {
1677                 m0 = m->m_nextpkt;
1678                 m->m_nextpkt = NULL;
1679
1680                 /*
1681                  * If underlying interface can not do VLAN tag insertion itself
1682                  * then attach a packet tag that holds it.
1683                  */
1684                 if ((m->m_flags & M_VLANTAG) &&
1685                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1686                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1687                         if (m == NULL) {
1688                                 if_printf(dst_ifp,
1689                                     "unable to prepend VLAN header\n");
1690                                 dst_ifp->if_oerrors++;
1691                                 continue;
1692                         }
1693                         m->m_flags &= ~M_VLANTAG;
1694                 }
1695
1696                 if (err == 0)
1697                         IFQ_ENQUEUE(&dst_ifp->if_snd, m, err);
1698         }
1699
1700         if (err == 0) {
1701
1702                 sc->sc_ifp->if_opackets++;
1703                 sc->sc_ifp->if_obytes += len;
1704
1705                 dst_ifp->if_obytes += len;
1706
1707                 if (mflags & M_MCAST) {
1708                         sc->sc_ifp->if_omcasts++;
1709                         dst_ifp->if_omcasts++;
1710                 }
1711         }
1712
1713         if ((dst_ifp->if_drv_flags & IFF_DRV_OACTIVE) == 0)
1714                 (*dst_ifp->if_start)(dst_ifp);
1715 }
1716
1717 /*
1718  * bridge_dummynet:
1719  *
1720  *      Receive a queued packet from dummynet and pass it on to the output
1721  *      interface.
1722  *
1723  *      The mbuf has the Ethernet header already attached.
1724  */
1725 static void
1726 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1727 {
1728         struct bridge_softc *sc;
1729
1730         sc = ifp->if_bridge;
1731
1732         /*
1733          * The packet didnt originate from a member interface. This should only
1734          * ever happen if a member interface is removed while packets are
1735          * queued for it.
1736          */
1737         if (sc == NULL) {
1738                 m_freem(m);
1739                 return;
1740         }
1741
1742         if (PFIL_HOOKED(&inet_pfil_hook)
1743 #ifdef INET6
1744             || PFIL_HOOKED(&inet6_pfil_hook)
1745 #endif
1746             ) {
1747                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1748                         return;
1749                 if (m == NULL)
1750                         return;
1751         }
1752
1753         bridge_enqueue(sc, ifp, m);
1754 }
1755
1756 /*
1757  * bridge_output:
1758  *
1759  *      Send output from a bridge member interface.  This
1760  *      performs the bridging function for locally originated
1761  *      packets.
1762  *
1763  *      The mbuf has the Ethernet header already attached.  We must
1764  *      enqueue or free the mbuf before returning.
1765  */
1766 static int
1767 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1768     struct rtentry *rt)
1769 {
1770         struct ether_header *eh;
1771         struct ifnet *dst_if;
1772         struct bridge_softc *sc;
1773         uint16_t vlan;
1774
1775         if (m->m_len < ETHER_HDR_LEN) {
1776                 m = m_pullup(m, ETHER_HDR_LEN);
1777                 if (m == NULL)
1778                         return (0);
1779         }
1780
1781         eh = mtod(m, struct ether_header *);
1782         sc = ifp->if_bridge;
1783         vlan = VLANTAGOF(m);
1784
1785         BRIDGE_LOCK(sc);
1786
1787         /*
1788          * If bridge is down, but the original output interface is up,
1789          * go ahead and send out that interface.  Otherwise, the packet
1790          * is dropped below.
1791          */
1792         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1793                 dst_if = ifp;
1794                 goto sendunicast;
1795         }
1796
1797         /*
1798          * If the packet is a multicast, or we don't know a better way to
1799          * get there, send to all interfaces.
1800          */
1801         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1802                 dst_if = NULL;
1803         else
1804                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1805         if (dst_if == NULL) {
1806                 struct bridge_iflist *bif;
1807                 struct mbuf *mc;
1808                 int error = 0, used = 0;
1809
1810                 bridge_span(sc, m);
1811
1812                 BRIDGE_LOCK2REF(sc, error);
1813                 if (error) {
1814                         m_freem(m);
1815                         return (0);
1816                 }
1817
1818                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1819                         dst_if = bif->bif_ifp;
1820
1821                         if (dst_if->if_type == IFT_GIF)
1822                                 continue;
1823                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1824                                 continue;
1825
1826                         /*
1827                          * If this is not the original output interface,
1828                          * and the interface is participating in spanning
1829                          * tree, make sure the port is in a state that
1830                          * allows forwarding.
1831                          */
1832                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1833                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1834                                 continue;
1835
1836                         if (LIST_NEXT(bif, bif_next) == NULL) {
1837                                 used = 1;
1838                                 mc = m;
1839                         } else {
1840                                 mc = m_copypacket(m, M_DONTWAIT);
1841                                 if (mc == NULL) {
1842                                         sc->sc_ifp->if_oerrors++;
1843                                         continue;
1844                                 }
1845                         }
1846
1847                         bridge_enqueue(sc, dst_if, mc);
1848                 }
1849                 if (used == 0)
1850                         m_freem(m);
1851                 BRIDGE_UNREF(sc);
1852                 return (0);
1853         }
1854
1855 sendunicast:
1856         /*
1857          * XXX Spanning tree consideration here?
1858          */
1859
1860         bridge_span(sc, m);
1861         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1862                 m_freem(m);
1863                 BRIDGE_UNLOCK(sc);
1864                 return (0);
1865         }
1866
1867         BRIDGE_UNLOCK(sc);
1868         bridge_enqueue(sc, dst_if, m);
1869         return (0);
1870 }
1871
1872 /*
1873  * bridge_start:
1874  *
1875  *      Start output on a bridge.
1876  *
1877  */
1878 static void
1879 bridge_start(struct ifnet *ifp)
1880 {
1881         struct bridge_softc *sc;
1882         struct mbuf *m;
1883         struct ether_header *eh;
1884         struct ifnet *dst_if;
1885
1886         sc = ifp->if_softc;
1887
1888         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1889         for (;;) {
1890                 IFQ_DEQUEUE(&ifp->if_snd, m);
1891                 if (m == 0)
1892                         break;
1893                 ETHER_BPF_MTAP(ifp, m);
1894
1895                 eh = mtod(m, struct ether_header *);
1896                 dst_if = NULL;
1897
1898                 BRIDGE_LOCK(sc);
1899                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1900                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
1901                 }
1902
1903                 if (dst_if == NULL)
1904                         bridge_broadcast(sc, ifp, m, 0);
1905                 else {
1906                         BRIDGE_UNLOCK(sc);
1907                         bridge_enqueue(sc, dst_if, m);
1908                 }
1909         }
1910         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
1911 }
1912
1913 /*
1914  * bridge_forward:
1915  *
1916  *      The forwarding function of the bridge.
1917  *
1918  *      NOTE: Releases the lock on return.
1919  */
1920 static void
1921 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
1922     struct mbuf *m)
1923 {
1924         struct bridge_iflist *dbif;
1925         struct ifnet *src_if, *dst_if, *ifp;
1926         struct ether_header *eh;
1927         uint16_t vlan;
1928         uint8_t *dst;
1929         int error;
1930
1931         src_if = m->m_pkthdr.rcvif;
1932         ifp = sc->sc_ifp;
1933
1934         ifp->if_ipackets++;
1935         ifp->if_ibytes += m->m_pkthdr.len;
1936         vlan = VLANTAGOF(m);
1937
1938         if ((sbif->bif_flags & IFBIF_STP) &&
1939             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1940                 goto drop;
1941
1942         eh = mtod(m, struct ether_header *);
1943         dst = eh->ether_dhost;
1944
1945         /* If the interface is learning, record the address. */
1946         if (sbif->bif_flags & IFBIF_LEARNING) {
1947                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
1948                     sbif, 0, IFBAF_DYNAMIC);
1949                 /*
1950                  * If the interface has addresses limits then deny any source
1951                  * that is not in the cache.
1952                  */
1953                 if (error && sbif->bif_addrmax)
1954                         goto drop;
1955         }
1956
1957         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
1958             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
1959                 goto drop;
1960
1961         /*
1962          * At this point, the port either doesn't participate
1963          * in spanning tree or it is in the forwarding state.
1964          */
1965
1966         /*
1967          * If the packet is unicast, destined for someone on
1968          * "this" side of the bridge, drop it.
1969          */
1970         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1971                 dst_if = bridge_rtlookup(sc, dst, vlan);
1972                 if (src_if == dst_if)
1973                         goto drop;
1974         } else {
1975                 /*
1976                  * Check if its a reserved multicast address, any address
1977                  * listed in 802.1D section 7.12.6 may not be forwarded by the
1978                  * bridge.
1979                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
1980                  */
1981                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
1982                     dst[2] == 0xc2 && dst[3] == 0x00 &&
1983                     dst[4] == 0x00 && dst[5] <= 0x0f)
1984                         goto drop;
1985
1986                 /* ...forward it to all interfaces. */
1987                 ifp->if_imcasts++;
1988                 dst_if = NULL;
1989         }
1990
1991         /*
1992          * If we have a destination interface which is a member of our bridge,
1993          * OR this is a unicast packet, push it through the bpf(4) machinery.
1994          * For broadcast or multicast packets, don't bother because it will
1995          * be reinjected into ether_input. We do this before we pass the packets
1996          * through the pfil(9) framework, as it is possible that pfil(9) will
1997          * drop the packet, or possibly modify it, making it difficult to debug
1998          * firewall issues on the bridge.
1999          */
2000         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2001                 ETHER_BPF_MTAP(ifp, m);
2002
2003         /* run the packet filter */
2004         if (PFIL_HOOKED(&inet_pfil_hook)
2005 #ifdef INET6
2006             || PFIL_HOOKED(&inet6_pfil_hook)
2007 #endif
2008             ) {
2009                 BRIDGE_UNLOCK(sc);
2010                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2011                         return;
2012                 if (m == NULL)
2013                         return;
2014                 BRIDGE_LOCK(sc);
2015         }
2016
2017         if (dst_if == NULL) {
2018                 bridge_broadcast(sc, src_if, m, 1);
2019                 return;
2020         }
2021
2022         /*
2023          * At this point, we're dealing with a unicast frame
2024          * going to a different interface.
2025          */
2026         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2027                 goto drop;
2028
2029         dbif = bridge_lookup_member_if(sc, dst_if);
2030         if (dbif == NULL)
2031                 /* Not a member of the bridge (anymore?) */
2032                 goto drop;
2033
2034         /* Private segments can not talk to each other */
2035         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2036                 goto drop;
2037
2038         if ((dbif->bif_flags & IFBIF_STP) &&
2039             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2040                 goto drop;
2041
2042         BRIDGE_UNLOCK(sc);
2043
2044         if (PFIL_HOOKED(&inet_pfil_hook)
2045 #ifdef INET6
2046             || PFIL_HOOKED(&inet6_pfil_hook)
2047 #endif
2048             ) {
2049                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2050                         return;
2051                 if (m == NULL)
2052                         return;
2053         }
2054
2055         bridge_enqueue(sc, dst_if, m);
2056         return;
2057
2058 drop:
2059         BRIDGE_UNLOCK(sc);
2060         m_freem(m);
2061 }
2062
2063 /*
2064  * bridge_input:
2065  *
2066  *      Receive input from a member interface.  Queue the packet for
2067  *      bridging if it is not for us.
2068  */
2069 static struct mbuf *
2070 bridge_input(struct ifnet *ifp, struct mbuf *m)
2071 {
2072         struct bridge_softc *sc = ifp->if_bridge;
2073         struct bridge_iflist *bif, *bif2;
2074         struct ifnet *bifp;
2075         struct ether_header *eh;
2076         struct mbuf *mc, *mc2;
2077         uint16_t vlan;
2078         int error;
2079
2080         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2081                 return (m);
2082
2083         bifp = sc->sc_ifp;
2084         vlan = VLANTAGOF(m);
2085
2086         /*
2087          * Implement support for bridge monitoring. If this flag has been
2088          * set on this interface, discard the packet once we push it through
2089          * the bpf(4) machinery, but before we do, increment the byte and
2090          * packet counters associated with this interface.
2091          */
2092         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2093                 m->m_pkthdr.rcvif  = bifp;
2094                 ETHER_BPF_MTAP(bifp, m);
2095                 bifp->if_ipackets++;
2096                 bifp->if_ibytes += m->m_pkthdr.len;
2097                 m_freem(m);
2098                 return (NULL);
2099         }
2100         BRIDGE_LOCK(sc);
2101         bif = bridge_lookup_member_if(sc, ifp);
2102         if (bif == NULL) {
2103                 BRIDGE_UNLOCK(sc);
2104                 return (m);
2105         }
2106
2107         eh = mtod(m, struct ether_header *);
2108
2109         bridge_span(sc, m);
2110
2111         if (m->m_flags & (M_BCAST|M_MCAST)) {
2112                 /* Tap off 802.1D packets; they do not get forwarded. */
2113                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2114                     ETHER_ADDR_LEN) == 0) {
2115                         m = bstp_input(&bif->bif_stp, ifp, m);
2116                         if (m == NULL) {
2117                                 BRIDGE_UNLOCK(sc);
2118                                 return (NULL);
2119                         }
2120                 }
2121
2122                 if ((bif->bif_flags & IFBIF_STP) &&
2123                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2124                         BRIDGE_UNLOCK(sc);
2125                         return (m);
2126                 }
2127
2128                 /*
2129                  * Make a deep copy of the packet and enqueue the copy
2130                  * for bridge processing; return the original packet for
2131                  * local processing.
2132                  */
2133                 mc = m_dup(m, M_DONTWAIT);
2134                 if (mc == NULL) {
2135                         BRIDGE_UNLOCK(sc);
2136                         return (m);
2137                 }
2138
2139                 /* Perform the bridge forwarding function with the copy. */
2140                 bridge_forward(sc, bif, mc);
2141
2142                 /*
2143                  * Reinject the mbuf as arriving on the bridge so we have a
2144                  * chance at claiming multicast packets. We can not loop back
2145                  * here from ether_input as a bridge is never a member of a
2146                  * bridge.
2147                  */
2148                 KASSERT(bifp->if_bridge == NULL,
2149                     ("loop created in bridge_input"));
2150                 mc2 = m_dup(m, M_DONTWAIT);
2151                 if (mc2 != NULL) {
2152                         /* Keep the layer3 header aligned */
2153                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2154                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2155                 }
2156                 if (mc2 != NULL) {
2157                         mc2->m_pkthdr.rcvif = bifp;
2158                         (*bifp->if_input)(bifp, mc2);
2159                 }
2160
2161                 /* Return the original packet for local processing. */
2162                 return (m);
2163         }
2164
2165         if ((bif->bif_flags & IFBIF_STP) &&
2166             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2167                 BRIDGE_UNLOCK(sc);
2168                 return (m);
2169         }
2170
2171 #ifdef DEV_CARP
2172 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2173         || ((iface)->if_carp \
2174             && carp_forus((iface)->if_carp, eh->ether_dhost))
2175 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2176         || ((iface)->if_carp \
2177             && carp_forus((iface)->if_carp, eh->ether_shost))
2178 #else
2179 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2180 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2181 #endif
2182
2183 #ifdef INET6
2184 #   define OR_PFIL_HOOKED_INET6 \
2185         || PFIL_HOOKED(&inet6_pfil_hook)
2186 #else
2187 #   define OR_PFIL_HOOKED_INET6
2188 #endif
2189
2190 #define GRAB_OUR_PACKETS(iface) \
2191         if ((iface)->if_type == IFT_GIF) \
2192                 continue; \
2193         /* It is destined for us. */ \
2194         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2195             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2196             ) {                                                         \
2197                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2198                         ETHER_BPF_MTAP(iface, m);                       \
2199                         iface->if_ipackets++;                           \
2200                         /* Filter on the physical interface. */         \
2201                         if (pfil_local_phys &&                          \
2202                             (PFIL_HOOKED(&inet_pfil_hook)               \
2203                              OR_PFIL_HOOKED_INET6)) {                   \
2204                                 if (bridge_pfil(&m, NULL, ifp,          \
2205                                     PFIL_IN) != 0 || m == NULL) {       \
2206                                         BRIDGE_UNLOCK(sc);              \
2207                                         return (NULL);                  \
2208                                 }                                       \
2209                         }                                               \
2210                 }                                                       \
2211                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2212                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2213                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2214                         if (error && bif->bif_addrmax) {                \
2215                                 BRIDGE_UNLOCK(sc);                      \
2216                                 m_freem(m);                             \
2217                                 return (NULL);                          \
2218                         }                                               \
2219                 }                                                       \
2220                 m->m_pkthdr.rcvif = iface;                              \
2221                 BRIDGE_UNLOCK(sc);                                      \
2222                 return (m);                                             \
2223         }                                                               \
2224                                                                         \
2225         /* We just received a packet that we sent out. */               \
2226         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2227             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2228             ) {                                                         \
2229                 BRIDGE_UNLOCK(sc);                                      \
2230                 m_freem(m);                                             \
2231                 return (NULL);                                          \
2232         }
2233
2234         /*
2235          * Unicast.  Make sure it's not for the bridge.
2236          */
2237         do { GRAB_OUR_PACKETS(bifp) } while (0);
2238
2239         /*
2240          * Give a chance for ifp at first priority. This will help when the
2241          * packet comes through the interface like VLAN's with the same MACs
2242          * on several interfaces from the same bridge. This also will save
2243          * some CPU cycles in case the destination interface and the input
2244          * interface (eq ifp) are the same.
2245          */
2246         do { GRAB_OUR_PACKETS(ifp) } while (0);
2247
2248         /* Now check the all bridge members. */
2249         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2250                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2251         }
2252
2253 #undef OR_CARP_CHECK_WE_ARE_DST
2254 #undef OR_CARP_CHECK_WE_ARE_SRC
2255 #undef OR_PFIL_HOOKED_INET6
2256 #undef GRAB_OUR_PACKETS
2257
2258         /* Perform the bridge forwarding function. */
2259         bridge_forward(sc, bif, m);
2260
2261         return (NULL);
2262 }
2263
2264 /*
2265  * bridge_broadcast:
2266  *
2267  *      Send a frame to all interfaces that are members of
2268  *      the bridge, except for the one on which the packet
2269  *      arrived.
2270  *
2271  *      NOTE: Releases the lock on return.
2272  */
2273 static void
2274 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2275     struct mbuf *m, int runfilt)
2276 {
2277         struct bridge_iflist *dbif, *sbif;
2278         struct mbuf *mc;
2279         struct ifnet *dst_if;
2280         int error = 0, used = 0, i;
2281
2282         sbif = bridge_lookup_member_if(sc, src_if);
2283
2284         BRIDGE_LOCK2REF(sc, error);
2285         if (error) {
2286                 m_freem(m);
2287                 return;
2288         }
2289
2290         /* Filter on the bridge interface before broadcasting */
2291         if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2292 #ifdef INET6
2293             || PFIL_HOOKED(&inet6_pfil_hook)
2294 #endif
2295             )) {
2296                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2297                         goto out;
2298                 if (m == NULL)
2299                         goto out;
2300         }
2301
2302         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2303                 dst_if = dbif->bif_ifp;
2304                 if (dst_if == src_if)
2305                         continue;
2306
2307                 /* Private segments can not talk to each other */
2308                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2309                         continue;
2310
2311                 if ((dbif->bif_flags & IFBIF_STP) &&
2312                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2313                         continue;
2314
2315                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2316                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2317                         continue;
2318
2319                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2320                         continue;
2321
2322                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2323                         mc = m;
2324                         used = 1;
2325                 } else {
2326                         mc = m_dup(m, M_DONTWAIT);
2327                         if (mc == NULL) {
2328                                 sc->sc_ifp->if_oerrors++;
2329                                 continue;
2330                         }
2331                 }
2332
2333                 /*
2334                  * Filter on the output interface. Pass a NULL bridge interface
2335                  * pointer so we do not redundantly filter on the bridge for
2336                  * each interface we broadcast on.
2337                  */
2338                 if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2339 #ifdef INET6
2340                     || PFIL_HOOKED(&inet6_pfil_hook)
2341 #endif
2342                     )) {
2343                         if (used == 0) {
2344                                 /* Keep the layer3 header aligned */
2345                                 i = min(mc->m_pkthdr.len, max_protohdr);
2346                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2347                                 if (mc == NULL) {
2348                                         sc->sc_ifp->if_oerrors++;
2349                                         continue;
2350                                 }
2351                         }
2352                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2353                                 continue;
2354                         if (mc == NULL)
2355                                 continue;
2356                 }
2357
2358                 bridge_enqueue(sc, dst_if, mc);
2359         }
2360         if (used == 0)
2361                 m_freem(m);
2362
2363 out:
2364         BRIDGE_UNREF(sc);
2365 }
2366
2367 /*
2368  * bridge_span:
2369  *
2370  *      Duplicate a packet out one or more interfaces that are in span mode,
2371  *      the original mbuf is unmodified.
2372  */
2373 static void
2374 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2375 {
2376         struct bridge_iflist *bif;
2377         struct ifnet *dst_if;
2378         struct mbuf *mc;
2379
2380         if (LIST_EMPTY(&sc->sc_spanlist))
2381                 return;
2382
2383         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2384                 dst_if = bif->bif_ifp;
2385
2386                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2387                         continue;
2388
2389                 mc = m_copypacket(m, M_DONTWAIT);
2390                 if (mc == NULL) {
2391                         sc->sc_ifp->if_oerrors++;
2392                         continue;
2393                 }
2394
2395                 bridge_enqueue(sc, dst_if, mc);
2396         }
2397 }
2398
2399 /*
2400  * bridge_rtupdate:
2401  *
2402  *      Add a bridge routing entry.
2403  */
2404 static int
2405 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2406     struct bridge_iflist *bif, int setflags, uint8_t flags)
2407 {
2408         struct bridge_rtnode *brt;
2409         int error;
2410
2411         BRIDGE_LOCK_ASSERT(sc);
2412
2413         /* Check the source address is valid and not multicast. */
2414         if (ETHER_IS_MULTICAST(dst) ||
2415             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2416              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2417                 return (EINVAL);
2418
2419         /* 802.1p frames map to vlan 1 */
2420         if (vlan == 0)
2421                 vlan = 1;
2422
2423         /*
2424          * A route for this destination might already exist.  If so,
2425          * update it, otherwise create a new one.
2426          */
2427         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2428                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2429                         sc->sc_brtexceeded++;
2430                         return (ENOSPC);
2431                 }
2432                 /* Check per interface address limits (if enabled) */
2433                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2434                         bif->bif_addrexceeded++;
2435                         return (ENOSPC);
2436                 }
2437
2438                 /*
2439                  * Allocate a new bridge forwarding node, and
2440                  * initialize the expiration time and Ethernet
2441                  * address.
2442                  */
2443                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2444                 if (brt == NULL)
2445                         return (ENOMEM);
2446
2447                 if (bif->bif_flags & IFBIF_STICKY)
2448                         brt->brt_flags = IFBAF_STICKY;
2449                 else
2450                         brt->brt_flags = IFBAF_DYNAMIC;
2451
2452                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2453                 brt->brt_vlan = vlan;
2454
2455                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2456                         uma_zfree(bridge_rtnode_zone, brt);
2457                         return (error);
2458                 }
2459                 brt->brt_dst = bif;
2460                 bif->bif_addrcnt++;
2461         }
2462
2463         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2464             brt->brt_dst != bif) {
2465                 brt->brt_dst->bif_addrcnt--;
2466                 brt->brt_dst = bif;
2467                 brt->brt_dst->bif_addrcnt++;
2468         }
2469
2470         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2471                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2472         if (setflags)
2473                 brt->brt_flags = flags;
2474
2475         return (0);
2476 }
2477
2478 /*
2479  * bridge_rtlookup:
2480  *
2481  *      Lookup the destination interface for an address.
2482  */
2483 static struct ifnet *
2484 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2485 {
2486         struct bridge_rtnode *brt;
2487
2488         BRIDGE_LOCK_ASSERT(sc);
2489
2490         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2491                 return (NULL);
2492
2493         return (brt->brt_ifp);
2494 }
2495
2496 /*
2497  * bridge_rttrim:
2498  *
2499  *      Trim the routine table so that we have a number
2500  *      of routing entries less than or equal to the
2501  *      maximum number.
2502  */
2503 static void
2504 bridge_rttrim(struct bridge_softc *sc)
2505 {
2506         struct bridge_rtnode *brt, *nbrt;
2507
2508         BRIDGE_LOCK_ASSERT(sc);
2509
2510         /* Make sure we actually need to do this. */
2511         if (sc->sc_brtcnt <= sc->sc_brtmax)
2512                 return;
2513
2514         /* Force an aging cycle; this might trim enough addresses. */
2515         bridge_rtage(sc);
2516         if (sc->sc_brtcnt <= sc->sc_brtmax)
2517                 return;
2518
2519         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2520                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2521                         bridge_rtnode_destroy(sc, brt);
2522                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2523                                 return;
2524                 }
2525         }
2526 }
2527
2528 /*
2529  * bridge_timer:
2530  *
2531  *      Aging timer for the bridge.
2532  */
2533 static void
2534 bridge_timer(void *arg)
2535 {
2536         struct bridge_softc *sc = arg;
2537
2538         BRIDGE_LOCK_ASSERT(sc);
2539
2540         bridge_rtage(sc);
2541
2542         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2543                 callout_reset(&sc->sc_brcallout,
2544                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2545 }
2546
2547 /*
2548  * bridge_rtage:
2549  *
2550  *      Perform an aging cycle.
2551  */
2552 static void
2553 bridge_rtage(struct bridge_softc *sc)
2554 {
2555         struct bridge_rtnode *brt, *nbrt;
2556
2557         BRIDGE_LOCK_ASSERT(sc);
2558
2559         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2560                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2561                         if (time_uptime >= brt->brt_expire)
2562                                 bridge_rtnode_destroy(sc, brt);
2563                 }
2564         }
2565 }
2566
2567 /*
2568  * bridge_rtflush:
2569  *
2570  *      Remove all dynamic addresses from the bridge.
2571  */
2572 static void
2573 bridge_rtflush(struct bridge_softc *sc, int full)
2574 {
2575         struct bridge_rtnode *brt, *nbrt;
2576
2577         BRIDGE_LOCK_ASSERT(sc);
2578
2579         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2580                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2581                         bridge_rtnode_destroy(sc, brt);
2582         }
2583 }
2584
2585 /*
2586  * bridge_rtdaddr:
2587  *
2588  *      Remove an address from the table.
2589  */
2590 static int
2591 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2592 {
2593         struct bridge_rtnode *brt;
2594         int found = 0;
2595
2596         BRIDGE_LOCK_ASSERT(sc);
2597
2598         /*
2599          * If vlan is zero then we want to delete for all vlans so the lookup
2600          * may return more than one.
2601          */
2602         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2603                 bridge_rtnode_destroy(sc, brt);
2604                 found = 1;
2605         }
2606
2607         return (found ? 0 : ENOENT);
2608 }
2609
2610 /*
2611  * bridge_rtdelete:
2612  *
2613  *      Delete routes to a speicifc member interface.
2614  */
2615 static void
2616 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2617 {
2618         struct bridge_rtnode *brt, *nbrt;
2619
2620         BRIDGE_LOCK_ASSERT(sc);
2621
2622         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2623                 if (brt->brt_ifp == ifp && (full ||
2624                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2625                         bridge_rtnode_destroy(sc, brt);
2626         }
2627 }
2628
2629 /*
2630  * bridge_rtable_init:
2631  *
2632  *      Initialize the route table for this bridge.
2633  */
2634 static int
2635 bridge_rtable_init(struct bridge_softc *sc)
2636 {
2637         int i;
2638
2639         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2640             M_DEVBUF, M_NOWAIT);
2641         if (sc->sc_rthash == NULL)
2642                 return (ENOMEM);
2643
2644         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2645                 LIST_INIT(&sc->sc_rthash[i]);
2646
2647         sc->sc_rthash_key = arc4random();
2648
2649         LIST_INIT(&sc->sc_rtlist);
2650
2651         return (0);
2652 }
2653
2654 /*
2655  * bridge_rtable_fini:
2656  *
2657  *      Deconstruct the route table for this bridge.
2658  */
2659 static void
2660 bridge_rtable_fini(struct bridge_softc *sc)
2661 {
2662
2663         KASSERT(sc->sc_brtcnt == 0,
2664             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2665         free(sc->sc_rthash, M_DEVBUF);
2666 }
2667
2668 /*
2669  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2670  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2671  */
2672 #define mix(a, b, c)                                                    \
2673 do {                                                                    \
2674         a -= b; a -= c; a ^= (c >> 13);                                 \
2675         b -= c; b -= a; b ^= (a << 8);                                  \
2676         c -= a; c -= b; c ^= (b >> 13);                                 \
2677         a -= b; a -= c; a ^= (c >> 12);                                 \
2678         b -= c; b -= a; b ^= (a << 16);                                 \
2679         c -= a; c -= b; c ^= (b >> 5);                                  \
2680         a -= b; a -= c; a ^= (c >> 3);                                  \
2681         b -= c; b -= a; b ^= (a << 10);                                 \
2682         c -= a; c -= b; c ^= (b >> 15);                                 \
2683 } while (/*CONSTCOND*/0)
2684
2685 static __inline uint32_t
2686 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2687 {
2688         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2689
2690         b += addr[5] << 8;
2691         b += addr[4];
2692         a += addr[3] << 24;
2693         a += addr[2] << 16;
2694         a += addr[1] << 8;
2695         a += addr[0];
2696
2697         mix(a, b, c);
2698
2699         return (c & BRIDGE_RTHASH_MASK);
2700 }
2701
2702 #undef mix
2703
2704 static int
2705 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2706 {
2707         int i, d;
2708
2709         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2710                 d = ((int)a[i]) - ((int)b[i]);
2711         }
2712
2713         return (d);
2714 }
2715
2716 /*
2717  * bridge_rtnode_lookup:
2718  *
2719  *      Look up a bridge route node for the specified destination. Compare the
2720  *      vlan id or if zero then just return the first match.
2721  */
2722 static struct bridge_rtnode *
2723 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2724 {
2725         struct bridge_rtnode *brt;
2726         uint32_t hash;
2727         int dir;
2728
2729         BRIDGE_LOCK_ASSERT(sc);
2730
2731         hash = bridge_rthash(sc, addr);
2732         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2733                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2734                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2735                         return (brt);
2736                 if (dir > 0)
2737                         return (NULL);
2738         }
2739
2740         return (NULL);
2741 }
2742
2743 /*
2744  * bridge_rtnode_insert:
2745  *
2746  *      Insert the specified bridge node into the route table.  We
2747  *      assume the entry is not already in the table.
2748  */
2749 static int
2750 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2751 {
2752         struct bridge_rtnode *lbrt;
2753         uint32_t hash;
2754         int dir;
2755
2756         BRIDGE_LOCK_ASSERT(sc);
2757
2758         hash = bridge_rthash(sc, brt->brt_addr);
2759
2760         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2761         if (lbrt == NULL) {
2762                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2763                 goto out;
2764         }
2765
2766         do {
2767                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2768                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2769                         return (EEXIST);
2770                 if (dir > 0) {
2771                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2772                         goto out;
2773                 }
2774                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2775                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2776                         goto out;
2777                 }
2778                 lbrt = LIST_NEXT(lbrt, brt_hash);
2779         } while (lbrt != NULL);
2780
2781 #ifdef DIAGNOSTIC
2782         panic("bridge_rtnode_insert: impossible");
2783 #endif
2784
2785 out:
2786         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2787         sc->sc_brtcnt++;
2788
2789         return (0);
2790 }
2791
2792 /*
2793  * bridge_rtnode_destroy:
2794  *
2795  *      Destroy a bridge rtnode.
2796  */
2797 static void
2798 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2799 {
2800         BRIDGE_LOCK_ASSERT(sc);
2801
2802         LIST_REMOVE(brt, brt_hash);
2803
2804         LIST_REMOVE(brt, brt_list);
2805         sc->sc_brtcnt--;
2806         brt->brt_dst->bif_addrcnt--;
2807         uma_zfree(bridge_rtnode_zone, brt);
2808 }
2809
2810 /*
2811  * bridge_rtable_expire:
2812  *
2813  *      Set the expiry time for all routes on an interface.
2814  */
2815 static void
2816 bridge_rtable_expire(struct ifnet *ifp, int age)
2817 {
2818         struct bridge_softc *sc = ifp->if_bridge;
2819         struct bridge_rtnode *brt;
2820
2821         BRIDGE_LOCK(sc);
2822
2823         /*
2824          * If the age is zero then flush, otherwise set all the expiry times to
2825          * age for the interface
2826          */
2827         if (age == 0)
2828                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2829         else {
2830                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2831                         /* Cap the expiry time to 'age' */
2832                         if (brt->brt_ifp == ifp &&
2833                             brt->brt_expire > time_uptime + age &&
2834                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2835                                 brt->brt_expire = time_uptime + age;
2836                 }
2837         }
2838         BRIDGE_UNLOCK(sc);
2839 }
2840
2841 /*
2842  * bridge_state_change:
2843  *
2844  *      Callback from the bridgestp code when a port changes states.
2845  */
2846 static void
2847 bridge_state_change(struct ifnet *ifp, int state)
2848 {
2849         struct bridge_softc *sc = ifp->if_bridge;
2850         static const char *stpstates[] = {
2851                 "disabled",
2852                 "listening",
2853                 "learning",
2854                 "forwarding",
2855                 "blocking",
2856                 "discarding"
2857         };
2858
2859         if (log_stp)
2860                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2861                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2862 }
2863
2864 /*
2865  * Send bridge packets through pfil if they are one of the types pfil can deal
2866  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2867  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2868  * that interface.
2869  */
2870 static int
2871 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2872 {
2873         int snap, error, i, hlen;
2874         struct ether_header *eh1, eh2;
2875         struct ip_fw_args args;
2876         struct ip *ip;
2877         struct llc llc1;
2878         u_int16_t ether_type;
2879
2880         snap = 0;
2881         error = -1;     /* Default error if not error == 0 */
2882
2883 #if 0
2884         /* we may return with the IP fields swapped, ensure its not shared */
2885         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2886 #endif
2887
2888         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2889                 return (0); /* filtering is disabled */
2890
2891         i = min((*mp)->m_pkthdr.len, max_protohdr);
2892         if ((*mp)->m_len < i) {
2893             *mp = m_pullup(*mp, i);
2894             if (*mp == NULL) {
2895                 printf("%s: m_pullup failed\n", __func__);
2896                 return (-1);
2897             }
2898         }
2899
2900         eh1 = mtod(*mp, struct ether_header *);
2901         ether_type = ntohs(eh1->ether_type);
2902
2903         /*
2904          * Check for SNAP/LLC.
2905          */
2906         if (ether_type < ETHERMTU) {
2907                 struct llc *llc2 = (struct llc *)(eh1 + 1);
2908
2909                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2910                     llc2->llc_dsap == LLC_SNAP_LSAP &&
2911                     llc2->llc_ssap == LLC_SNAP_LSAP &&
2912                     llc2->llc_control == LLC_UI) {
2913                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
2914                         snap = 1;
2915                 }
2916         }
2917
2918         /*
2919          * If we're trying to filter bridge traffic, don't look at anything
2920          * other than IP and ARP traffic.  If the filter doesn't understand
2921          * IPv6, don't allow IPv6 through the bridge either.  This is lame
2922          * since if we really wanted, say, an AppleTalk filter, we are hosed,
2923          * but of course we don't have an AppleTalk filter to begin with.
2924          * (Note that since pfil doesn't understand ARP it will pass *ALL*
2925          * ARP traffic.)
2926          */
2927         switch (ether_type) {
2928                 case ETHERTYPE_ARP:
2929                 case ETHERTYPE_REVARP:
2930                         if (pfil_ipfw_arp == 0)
2931                                 return (0); /* Automatically pass */
2932                         break;
2933
2934                 case ETHERTYPE_IP:
2935 #ifdef INET6
2936                 case ETHERTYPE_IPV6:
2937 #endif /* INET6 */
2938                         break;
2939                 default:
2940                         /*
2941                          * Check to see if the user wants to pass non-ip
2942                          * packets, these will not be checked by pfil(9) and
2943                          * passed unconditionally so the default is to drop.
2944                          */
2945                         if (pfil_onlyip)
2946                                 goto bad;
2947         }
2948
2949         /* Strip off the Ethernet header and keep a copy. */
2950         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
2951         m_adj(*mp, ETHER_HDR_LEN);
2952
2953         /* Strip off snap header, if present */
2954         if (snap) {
2955                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
2956                 m_adj(*mp, sizeof(struct llc));
2957         }
2958
2959         /*
2960          * Check the IP header for alignment and errors
2961          */
2962         if (dir == PFIL_IN) {
2963                 switch (ether_type) {
2964                         case ETHERTYPE_IP:
2965                                 error = bridge_ip_checkbasic(mp);
2966                                 break;
2967 #ifdef INET6
2968                         case ETHERTYPE_IPV6:
2969                                 error = bridge_ip6_checkbasic(mp);
2970                                 break;
2971 #endif /* INET6 */
2972                         default:
2973                                 error = 0;
2974                 }
2975                 if (error)
2976                         goto bad;
2977         }
2978
2979         if (IPFW_LOADED && pfil_ipfw != 0 && dir == PFIL_OUT && ifp != NULL) {
2980                 error = -1;
2981                 args.rule = ip_dn_claim_rule(*mp);
2982                 if (args.rule != NULL && fw_one_pass)
2983                         goto ipfwpass; /* packet already partially processed */
2984
2985                 args.m = *mp;
2986                 args.oif = ifp;
2987                 args.next_hop = NULL;
2988                 args.eh = &eh2;
2989                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
2990                 i = ip_fw_chk_ptr(&args);
2991                 *mp = args.m;
2992
2993                 if (*mp == NULL)
2994                         return (error);
2995
2996                 if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
2997
2998                         /* put the Ethernet header back on */
2999                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3000                         if (*mp == NULL)
3001                                 return (error);
3002                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3003
3004                         /*
3005                          * Pass the pkt to dummynet, which consumes it. The
3006                          * packet will return to us via bridge_dummynet().
3007                          */
3008                         args.oif = ifp;
3009                         ip_dn_io_ptr(mp, DN_TO_IFB_FWD, &args);
3010                         return (error);
3011                 }
3012
3013                 if (i != IP_FW_PASS) /* drop */
3014                         goto bad;
3015         }
3016
3017 ipfwpass:
3018         error = 0;
3019
3020         /*
3021          * Run the packet through pfil
3022          */
3023         switch (ether_type) {
3024         case ETHERTYPE_IP:
3025                 /*
3026                  * before calling the firewall, swap fields the same as
3027                  * IP does. here we assume the header is contiguous
3028                  */
3029                 ip = mtod(*mp, struct ip *);
3030
3031                 ip->ip_len = ntohs(ip->ip_len);
3032                 ip->ip_off = ntohs(ip->ip_off);
3033
3034                 /*
3035                  * Run pfil on the member interface and the bridge, both can
3036                  * be skipped by clearing pfil_member or pfil_bridge.
3037                  *
3038                  * Keep the order:
3039                  *   in_if -> bridge_if -> out_if
3040                  */
3041                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3042                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3043                                         dir, NULL);
3044
3045                 if (*mp == NULL || error != 0) /* filter may consume */
3046                         break;
3047
3048                 if (pfil_member && ifp != NULL)
3049                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp,
3050                                         dir, NULL);
3051
3052                 if (*mp == NULL || error != 0) /* filter may consume */
3053                         break;
3054
3055                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3056                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3057                                         dir, NULL);
3058
3059                 if (*mp == NULL || error != 0) /* filter may consume */
3060                         break;
3061
3062                 /* check if we need to fragment the packet */
3063                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3064                         i = (*mp)->m_pkthdr.len;
3065                         if (i > ifp->if_mtu) {
3066                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3067                                             &llc1);
3068                                 return (error);
3069                         }
3070                 }
3071
3072                 /* Recalculate the ip checksum and restore byte ordering */
3073                 ip = mtod(*mp, struct ip *);
3074                 hlen = ip->ip_hl << 2;
3075                 if (hlen < sizeof(struct ip))
3076                         goto bad;
3077                 if (hlen > (*mp)->m_len) {
3078                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3079                                 goto bad;
3080                         ip = mtod(*mp, struct ip *);
3081                         if (ip == NULL)
3082                                 goto bad;
3083                 }
3084                 ip->ip_len = htons(ip->ip_len);
3085                 ip->ip_off = htons(ip->ip_off);
3086                 ip->ip_sum = 0;
3087                 if (hlen == sizeof(struct ip))
3088                         ip->ip_sum = in_cksum_hdr(ip);
3089                 else
3090                         ip->ip_sum = in_cksum(*mp, hlen);
3091
3092                 break;
3093 #ifdef INET6
3094         case ETHERTYPE_IPV6:
3095                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3096                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3097                                         dir, NULL);
3098
3099                 if (*mp == NULL || error != 0) /* filter may consume */
3100                         break;
3101
3102                 if (pfil_member && ifp != NULL)
3103                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3104                                         dir, NULL);
3105
3106                 if (*mp == NULL || error != 0) /* filter may consume */
3107                         break;
3108
3109                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3110                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3111                                         dir, NULL);
3112                 break;
3113 #endif
3114         default:
3115                 error = 0;
3116                 break;
3117         }
3118
3119         if (*mp == NULL)
3120                 return (error);
3121         if (error != 0)
3122                 goto bad;
3123
3124         error = -1;
3125
3126         /*
3127          * Finally, put everything back the way it was and return
3128          */
3129         if (snap) {
3130                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3131                 if (*mp == NULL)
3132                         return (error);
3133                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3134         }
3135
3136         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3137         if (*mp == NULL)
3138                 return (error);
3139         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3140
3141         return (0);
3142
3143 bad:
3144         m_freem(*mp);
3145         *mp = NULL;
3146         return (error);
3147 }
3148
3149 /*
3150  * Perform basic checks on header size since
3151  * pfil assumes ip_input has already processed
3152  * it for it.  Cut-and-pasted from ip_input.c.
3153  * Given how simple the IPv6 version is,
3154  * does the IPv4 version really need to be
3155  * this complicated?
3156  *
3157  * XXX Should we update ipstat here, or not?
3158  * XXX Right now we update ipstat but not
3159  * XXX csum_counter.
3160  */
3161 static int
3162 bridge_ip_checkbasic(struct mbuf **mp)
3163 {
3164         struct mbuf *m = *mp;
3165         struct ip *ip;
3166         int len, hlen;
3167         u_short sum;
3168
3169         if (*mp == NULL)
3170                 return (-1);
3171
3172         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3173                 if ((m = m_copyup(m, sizeof(struct ip),
3174                         (max_linkhdr + 3) & ~3)) == NULL) {
3175                         /* XXXJRT new stat, please */
3176                         ipstat.ips_toosmall++;
3177                         goto bad;
3178                 }
3179         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3180                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3181                         ipstat.ips_toosmall++;
3182                         goto bad;
3183                 }
3184         }
3185         ip = mtod(m, struct ip *);
3186         if (ip == NULL) goto bad;
3187
3188         if (ip->ip_v != IPVERSION) {
3189                 ipstat.ips_badvers++;
3190                 goto bad;
3191         }
3192         hlen = ip->ip_hl << 2;
3193         if (hlen < sizeof(struct ip)) { /* minimum header length */
3194                 ipstat.ips_badhlen++;
3195                 goto bad;
3196         }
3197         if (hlen > m->m_len) {
3198                 if ((m = m_pullup(m, hlen)) == 0) {
3199                         ipstat.ips_badhlen++;
3200                         goto bad;
3201                 }
3202                 ip = mtod(m, struct ip *);
3203                 if (ip == NULL) goto bad;
3204         }
3205
3206         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3207                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3208         } else {
3209                 if (hlen == sizeof(struct ip)) {
3210                         sum = in_cksum_hdr(ip);
3211                 } else {
3212                         sum = in_cksum(m, hlen);
3213                 }
3214         }
3215         if (sum) {
3216                 ipstat.ips_badsum++;
3217                 goto bad;
3218         }
3219
3220         /* Retrieve the packet length. */
3221         len = ntohs(ip->ip_len);
3222
3223         /*
3224          * Check for additional length bogosity
3225          */
3226         if (len < hlen) {
3227                 ipstat.ips_badlen++;
3228                 goto bad;
3229         }
3230
3231         /*
3232          * Check that the amount of data in the buffers
3233          * is as at least much as the IP header would have us expect.
3234          * Drop packet if shorter than we expect.
3235          */
3236         if (m->m_pkthdr.len < len) {
3237                 ipstat.ips_tooshort++;
3238                 goto bad;
3239         }
3240
3241         /* Checks out, proceed */
3242         *mp = m;
3243         return (0);
3244
3245 bad:
3246         *mp = m;
3247         return (-1);
3248 }
3249
3250 #ifdef INET6
3251 /*
3252  * Same as above, but for IPv6.
3253  * Cut-and-pasted from ip6_input.c.
3254  * XXX Should we update ip6stat, or not?
3255  */
3256 static int
3257 bridge_ip6_checkbasic(struct mbuf **mp)
3258 {
3259         struct mbuf *m = *mp;
3260         struct ip6_hdr *ip6;
3261
3262         /*
3263          * If the IPv6 header is not aligned, slurp it up into a new
3264          * mbuf with space for link headers, in the event we forward
3265          * it.  Otherwise, if it is aligned, make sure the entire base
3266          * IPv6 header is in the first mbuf of the chain.
3267          */
3268         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3269                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3270                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3271                             (max_linkhdr + 3) & ~3)) == NULL) {
3272                         /* XXXJRT new stat, please */
3273                         ip6stat.ip6s_toosmall++;
3274                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3275                         goto bad;
3276                 }
3277         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3278                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3279                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3280                         ip6stat.ip6s_toosmall++;
3281                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3282                         goto bad;
3283                 }
3284         }
3285
3286         ip6 = mtod(m, struct ip6_hdr *);
3287
3288         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3289                 ip6stat.ip6s_badvers++;
3290                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3291                 goto bad;
3292         }
3293
3294         /* Checks out, proceed */
3295         *mp = m;
3296         return (0);
3297
3298 bad:
3299         *mp = m;
3300         return (-1);
3301 }
3302 #endif /* INET6 */
3303
3304 /*
3305  * bridge_fragment:
3306  *
3307  *      Return a fragmented mbuf chain.
3308  */
3309 static int
3310 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3311     int snap, struct llc *llc)
3312 {
3313         struct mbuf *m0;
3314         struct ip *ip;
3315         int error = -1;
3316
3317         if (m->m_len < sizeof(struct ip) &&
3318             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3319                 goto out;
3320         ip = mtod(m, struct ip *);
3321
3322         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3323                     CSUM_DELAY_IP);
3324         if (error)
3325                 goto out;
3326
3327         /* walk the chain and re-add the Ethernet header */
3328         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3329                 if (error == 0) {
3330                         if (snap) {
3331                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3332                                 if (m0 == NULL) {
3333                                         error = ENOBUFS;
3334                                         continue;
3335                                 }
3336                                 bcopy(llc, mtod(m0, caddr_t),
3337                                     sizeof(struct llc));
3338                         }
3339                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3340                         if (m0 == NULL) {
3341                                 error = ENOBUFS;
3342                                 continue;
3343                         }
3344                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3345                 } else
3346                         m_freem(m);
3347         }
3348
3349         if (error == 0)
3350                 ipstat.ips_fragmented++;
3351
3352         return (error);
3353
3354 out:
3355         if (m != NULL)
3356                 m_freem(m);
3357         return (error);
3358 }