sys/net/bridge/if_bridge.c

   1 /*
   2  * Copyright 2001 Wasabi Systems, Inc.
   3  * All rights reserved.
   4  *
   5  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. All advertising materials mentioning features or use of this software
  16  *    must display the following acknowledgement:
  17  *      This product includes software developed for the NetBSD Project by
  18  *      Wasabi Systems, Inc.
  19  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  20  *    or promote products derived from this software without specific prior
  21  *    written permission.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  27  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  33  * POSSIBILITY OF SUCH DAMAGE.
  34  */
  35
  36 /*
  37  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  38  * All rights reserved.
  39  *
  40  * Redistribution and use in source and binary forms, with or without
  41  * modification, are permitted provided that the following conditions
  42  * are met:
  43  * 1. Redistributions of source code must retain the above copyright
  44  *    notice, this list of conditions and the following disclaimer.
  45  * 2. Redistributions in binary form must reproduce the above copyright
  46  *    notice, this list of conditions and the following disclaimer in the
  47  *    documentation and/or other materials provided with the distribution.
  48  * 3. All advertising materials mentioning features or use of this software
  49  *    must display the following acknowledgement:
  50  *      This product includes software developed by Jason L. Wright
  51  * 4. The name of the author may not be used to endorse or promote products
  52  *    derived from this software without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  55  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  56  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  57  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  58  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  59  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  60  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  62  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  63  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  64  * POSSIBILITY OF SUCH DAMAGE.
  65  *
  66  * $OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp $
  67  * $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $
  68  * $FreeBSD: src/sys/net/if_bridge.c,v 1.26 2005/10/13 23:05:55 thompsa Exp $
  69  * $DragonFly: src/sys/net/bridge/if_bridge.c,v 1.28 2008/01/11 11:59:40 sephe Exp $
  70  */
  71
  72 /*
  73  * Network interface bridge support.
  74  *
  75  * TODO:
  76  *
  77  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  78  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  79  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
  80  *        consider heterogenous bridges).
  81  */
  82
  83 #include <sys/cdefs.h>
  84
  85 #include "opt_inet.h"
  86 #include "opt_inet6.h"
  87
  88 #include <sys/param.h>
  89 #include <sys/mbuf.h>
  90 #include <sys/malloc.h>
  91 #include <sys/protosw.h>
  92 #include <sys/systm.h>
  93 #include <sys/time.h>
  94 #include <sys/socket.h> /* for net/if.h */
  95 #include <sys/sockio.h>
  96 #include <sys/ctype.h>  /* string functions */
  97 #include <sys/kernel.h>
  98 #include <sys/random.h>
  99 #include <sys/sysctl.h>
 100 #include <sys/module.h>
 101 #include <sys/proc.h>
 102 #include <sys/lock.h>
 103 #include <sys/thread.h>
 104 #include <sys/thread2.h>
 105 #include <sys/mpipe.h>
 106
 107 #include <net/bpf.h>
 108 #include <net/if.h>
 109 #include <net/if_dl.h>
 110 #include <net/if_types.h>
 111 #include <net/if_var.h>
 112 #include <net/pfil.h>
 113 #include <net/ifq_var.h>
 114 #include <net/if_clone.h>
 115
 116 #include <netinet/in.h> /* for struct arpcom */
 117 #include <netinet/in_systm.h>
 118 #include <netinet/in_var.h>
 119 #include <netinet/ip.h>
 120 #include <netinet/ip_var.h>
 121 #ifdef INET6
 122 #include <netinet/ip6.h>
 123 #include <netinet6/ip6_var.h>
 124 #endif
 125 #include <netinet/if_ether.h> /* for struct arpcom */
 126 #include <net/bridge/if_bridgevar.h>
 127 #include <net/if_llc.h>
 128 #include <net/netmsg2.h>
 129
 130 #include <net/route.h>
 131 #include <sys/in_cksum.h>
 132
 133 /*
 134  * Size of the route hash table.  Must be a power of two.
 135  */
 136 #ifndef BRIDGE_RTHASH_SIZE
 137 #define BRIDGE_RTHASH_SIZE              1024
 138 #endif
 139
 140 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 141
 142 /*
 143  * Maximum number of addresses to cache.
 144  */
 145 #ifndef BRIDGE_RTABLE_MAX
 146 #define BRIDGE_RTABLE_MAX               100
 147 #endif
 148
 149 /*
 150  * Spanning tree defaults.
 151  */
 152 #define BSTP_DEFAULT_MAX_AGE            (20 * 256)
 153 #define BSTP_DEFAULT_HELLO_TIME         (2 * 256)
 154 #define BSTP_DEFAULT_FORWARD_DELAY      (15 * 256)
 155 #define BSTP_DEFAULT_HOLD_TIME          (1 * 256)
 156 #define BSTP_DEFAULT_BRIDGE_PRIORITY    0x8000
 157 #define BSTP_DEFAULT_PORT_PRIORITY      0x80
 158 #define BSTP_DEFAULT_PATH_COST          55
 159
 160 /*
 161  * Timeout (in seconds) for entries learned dynamically.
 162  */
 163 #ifndef BRIDGE_RTABLE_TIMEOUT
 164 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 165 #endif
 166
 167 /*
 168  * Number of seconds between walks of the route list.
 169  */
 170 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 171 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 172 #endif
 173
 174 /*
 175  * List of capabilities to mask on the member interface.
 176  */
 177 #define BRIDGE_IFCAPS_MASK              IFCAP_TXCSUM
 178
 179 eventhandler_tag        bridge_detach_cookie = NULL;
 180
 181 extern  struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
 182 extern  int (*bridge_output_p)(struct ifnet *, struct mbuf *,
 183                 struct sockaddr *, struct rtentry *);
 184 extern  void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
 185
 186 static int      bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 187
 188 static int      bridge_clone_create(struct if_clone *, int);
 189 static void     bridge_clone_destroy(struct ifnet *);
 190
 191 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
 192 static void     bridge_mutecaps(struct bridge_iflist *, int);
 193 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 194 static void     bridge_init(void *);
 195 static void     bridge_stop(struct ifnet *);
 196 static void     bridge_start(struct ifnet *);
 197 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 198 static int      bridge_output_serialized(struct ifnet *, struct mbuf *,
 199                     struct sockaddr *, struct rtentry *);
 200
 201 static void     bridge_forward(struct bridge_softc *, struct mbuf *m);
 202
 203 static void     bridge_timer(void *);
 204
 205 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 206                     struct mbuf *, int);
 207 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 208
 209 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 210                     struct ifnet *, int, uint8_t);
 211 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
 212 static void     bridge_rttrim(struct bridge_softc *);
 213 static void     bridge_rtage(struct bridge_softc *);
 214 static void     bridge_rtflush(struct bridge_softc *, int);
 215 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
 216
 217 static int      bridge_rtable_init(struct bridge_softc *);
 218 static void     bridge_rtable_fini(struct bridge_softc *);
 219
 220 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 221 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 222                     const uint8_t *);
 223 static int      bridge_rtnode_insert(struct bridge_softc *,
 224                     struct bridge_rtnode *);
 225 static void     bridge_rtnode_destroy(struct bridge_softc *,
 226                     struct bridge_rtnode *);
 227
 228 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 229                     const char *name);
 230 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 231                     struct ifnet *ifp);
 232 static void     bridge_delete_member(struct bridge_softc *,
 233                     struct bridge_iflist *, int);
 234 static void     bridge_delete_span(struct bridge_softc *,
 235                     struct bridge_iflist *);
 236
 237 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 238 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 239 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 240 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 241 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 242 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 243 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 244 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 245 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 246 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 247 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 248 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 249 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 250 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 251 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 252 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 253 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 254 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 255 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 256 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 257 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 258 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 259 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 260 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 261 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 262 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 263                     int);
 264 static int      bridge_ip_checkbasic(struct mbuf **mp);
 265 #ifdef INET6
 266 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 267 #endif /* INET6 */
 268 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 269                     struct ether_header *, int, struct llc *);
 270 static void     bridge_enqueue_internal(struct ifnet *, struct mbuf *m,
 271                                         netisr_fn_t);
 272 static void     bridge_enqueue_handler(struct netmsg *);
 273 static void     bridge_pfil_enqueue_handler(struct netmsg *);
 274 static void     bridge_pfil_enqueue(struct ifnet *, struct mbuf *, int);
 275 static void     bridge_handoff_notags(struct ifnet *, struct mbuf *);
 276 static void     bridge_handoff(struct ifnet *, struct mbuf *);
 277
 278 SYSCTL_DECL(_net_link);
 279 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 280
 281 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 282 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 283 static int pfil_member = 1; /* run pfil hooks on the member interface */
 284 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 285     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 286 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 287     &pfil_bridge, 0, "Packet filter on the bridge interface");
 288 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 289     &pfil_member, 0, "Packet filter on the member interface");
 290
 291 struct bridge_control {
 292         int     (*bc_func)(struct bridge_softc *, void *);
 293         int     bc_argsize;
 294         int     bc_flags;
 295 };
 296
 297 #define BC_F_COPYIN             0x01    /* copy arguments in */
 298 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 299 #define BC_F_SUSER              0x04    /* do super-user check */
 300
 301 const struct bridge_control bridge_control_table[] = {
 302         { bridge_ioctl_add,             sizeof(struct ifbreq),
 303           BC_F_COPYIN|BC_F_SUSER },
 304         { bridge_ioctl_del,             sizeof(struct ifbreq),
 305           BC_F_COPYIN|BC_F_SUSER },
 306
 307         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 308           BC_F_COPYIN|BC_F_COPYOUT },
 309         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 310           BC_F_COPYIN|BC_F_SUSER },
 311
 312         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 313           BC_F_COPYIN|BC_F_SUSER },
 314         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 315           BC_F_COPYOUT },
 316
 317         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 318           BC_F_COPYIN|BC_F_COPYOUT },
 319         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 320           BC_F_COPYIN|BC_F_COPYOUT },
 321
 322         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 323           BC_F_COPYIN|BC_F_SUSER },
 324
 325         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 326           BC_F_COPYIN|BC_F_SUSER },
 327         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 328           BC_F_COPYOUT },
 329
 330         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 331           BC_F_COPYIN|BC_F_SUSER },
 332
 333         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 334           BC_F_COPYIN|BC_F_SUSER },
 335
 336         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 337           BC_F_COPYOUT },
 338         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 339           BC_F_COPYIN|BC_F_SUSER },
 340
 341         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 342           BC_F_COPYOUT },
 343         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 344           BC_F_COPYIN|BC_F_SUSER },
 345
 346         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 347           BC_F_COPYOUT },
 348         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 349           BC_F_COPYIN|BC_F_SUSER },
 350
 351         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 352           BC_F_COPYOUT },
 353         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 354           BC_F_COPYIN|BC_F_SUSER },
 355
 356         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 357           BC_F_COPYIN|BC_F_SUSER },
 358
 359         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 360           BC_F_COPYIN|BC_F_SUSER },
 361
 362         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 363           BC_F_COPYIN|BC_F_SUSER },
 364         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 365           BC_F_COPYIN|BC_F_SUSER },
 366 };
 367 const int bridge_control_table_size =
 368     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 369
 370 LIST_HEAD(, bridge_softc) bridge_list;
 371
 372 struct if_clone bridge_cloner = IF_CLONE_INITIALIZER("bridge",
 373                                 bridge_clone_create,
 374                                 bridge_clone_destroy, 0, IF_MAXUNIT);
 375
 376 static int
 377 bridge_modevent(module_t mod, int type, void *data)
 378 {
 379         switch (type) {
 380         case MOD_LOAD:
 381                 LIST_INIT(&bridge_list);
 382                 if_clone_attach(&bridge_cloner);
 383                 bridge_input_p = bridge_input;
 384                 bridge_output_p = bridge_output_serialized;
 385                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 386                     ifnet_detach_event, bridge_ifdetach, NULL,
 387                     EVENTHANDLER_PRI_ANY);
 388 #if notyet
 389                 bstp_linkstate_p = bstp_linkstate;
 390 #endif
 391                 break;
 392         case MOD_UNLOAD:
 393                 if (!LIST_EMPTY(&bridge_list))
 394                         return (EBUSY);
 395                 EVENTHANDLER_DEREGISTER(ifnet_detach_event,
 396                     bridge_detach_cookie);
 397                 if_clone_detach(&bridge_cloner);
 398                 bridge_input_p = NULL;
 399                 bridge_output_p = NULL;
 400 #if notyet
 401                 bstp_linkstate_p = NULL;
 402 #endif
 403                 break;
 404         default:
 405                 return (EOPNOTSUPP);
 406         }
 407         return (0);
 408 }
 409
 410 static moduledata_t bridge_mod = {
 411         "if_bridge",
 412         bridge_modevent,
 413         0
 414 };
 415
 416 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 417
 418
 419 /*
 420  * bridge_clone_create:
 421  *
 422  *      Create a new bridge instance.
 423  */
 424 static int
 425 bridge_clone_create(struct if_clone *ifc, int unit)
 426 {
 427         struct bridge_softc *sc;
 428         struct ifnet *ifp;
 429         u_char eaddr[6];
 430
 431         sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 432         ifp = sc->sc_ifp = &sc->sc_if;
 433
 434         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 435         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 436         sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
 437         sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
 438         sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
 439         sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
 440         sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
 441
 442         /* Initialize our routing table. */
 443         bridge_rtable_init(sc);
 444
 445         callout_init(&sc->sc_brcallout);
 446         callout_init(&sc->sc_bstpcallout);
 447
 448         LIST_INIT(&sc->sc_iflist);
 449         LIST_INIT(&sc->sc_spanlist);
 450
 451         ifp->if_softc = sc;
 452         if_initname(ifp, ifc->ifc_name, unit);
 453         ifp->if_mtu = ETHERMTU;
 454         ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
 455         ifp->if_ioctl = bridge_ioctl;
 456         ifp->if_start = bridge_start;
 457         ifp->if_init = bridge_init;
 458         ifp->if_type = IFT_BRIDGE;
 459         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
 460         ifp->if_snd.ifq_maxlen = ifqmaxlen;
 461         ifq_set_ready(&ifp->if_snd);
 462         ifp->if_hdrlen = ETHER_HDR_LEN;
 463
 464         /*
 465          * Generate a random ethernet address and use the private AC:DE:48
 466          * OUI code.
 467          */
 468         {
 469                 int rnd = karc4random();
 470                 bcopy(&rnd, &eaddr[0], 4); /* ETHER_ADDR_LEN == 6 */
 471                 rnd = karc4random();
 472                 bcopy(&rnd, &eaddr[2], 4); /* ETHER_ADDR_LEN == 6 */
 473         }
 474         eaddr[0] &= ~1;         /* clear multicast bit */
 475         eaddr[0] |= 2;          /* set the LAA bit */
 476
 477         ether_ifattach(ifp, eaddr, NULL);
 478         /* Now undo some of the damage... */
 479         ifp->if_baudrate = 0;
 480         ifp->if_type = IFT_BRIDGE;
 481
 482         crit_enter();
 483         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 484         crit_exit();
 485
 486         return (0);
 487 }
 488
 489 /*
 490  * bridge_clone_destroy:
 491  *
 492  *      Destroy a bridge instance.
 493  */
 494 static void
 495 bridge_clone_destroy(struct ifnet *ifp)
 496 {
 497         struct bridge_softc *sc = ifp->if_softc;
 498         struct bridge_iflist *bif;
 499
 500         lwkt_serialize_enter(ifp->if_serializer);
 501
 502         bridge_stop(ifp);
 503         ifp->if_flags &= ~IFF_UP;
 504
 505         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 506                 bridge_delete_member(sc, bif, 0);
 507
 508         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 509                 bridge_delete_span(sc, bif);
 510         }
 511
 512         callout_stop(&sc->sc_brcallout);
 513         callout_stop(&sc->sc_bstpcallout);
 514
 515         lwkt_serialize_exit(ifp->if_serializer);
 516
 517         crit_enter();
 518         LIST_REMOVE(sc, sc_list);
 519         crit_exit();
 520
 521         ether_ifdetach(ifp);
 522
 523         /* Tear down the routing table. */
 524         bridge_rtable_fini(sc);
 525
 526         kfree(sc, M_DEVBUF);
 527 }
 528
 529 /*
 530  * bridge_ioctl:
 531  *
 532  *      Handle a control request from the operator.
 533  */
 534 static int
 535 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 536 {
 537         struct bridge_softc *sc = ifp->if_softc;
 538         union {
 539                 struct ifbreq ifbreq;
 540                 struct ifbifconf ifbifconf;
 541                 struct ifbareq ifbareq;
 542                 struct ifbaconf ifbaconf;
 543                 struct ifbrparam ifbrparam;
 544         } args;
 545         struct ifdrv *ifd = (struct ifdrv *) data;
 546         const struct bridge_control *bc;
 547         int error = 0;
 548
 549         ASSERT_SERIALIZED(ifp->if_serializer);
 550
 551         switch (cmd) {
 552         case SIOCADDMULTI:
 553         case SIOCDELMULTI:
 554                 break;
 555
 556         case SIOCGDRVSPEC:
 557         case SIOCSDRVSPEC:
 558                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 559                         error = EINVAL;
 560                         break;
 561                 }
 562                 bc = &bridge_control_table[ifd->ifd_cmd];
 563
 564                 if (cmd == SIOCGDRVSPEC &&
 565                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 566                         error = EINVAL;
 567                         break;
 568                 } else if (cmd == SIOCSDRVSPEC &&
 569                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 570                         error = EINVAL;
 571                         break;
 572                 }
 573
 574                 if (bc->bc_flags & BC_F_SUSER) {
 575                         error = suser_cred(cr, NULL_CRED_OKAY);
 576                         if (error)
 577                                 break;
 578                 }
 579
 580                 if (ifd->ifd_len != bc->bc_argsize ||
 581                     ifd->ifd_len > sizeof(args)) {
 582                         error = EINVAL;
 583                         break;
 584                 }
 585
 586                 memset(&args, 0, sizeof(args));
 587                 if (bc->bc_flags & BC_F_COPYIN) {
 588                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 589                         if (error)
 590                                 break;
 591                 }
 592
 593                 error = bc->bc_func(sc, &args);
 594                 if (error)
 595                         break;
 596
 597                 if (bc->bc_flags & BC_F_COPYOUT)
 598                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 599                 break;
 600
 601         case SIOCSIFFLAGS:
 602                 if (!(ifp->if_flags & IFF_UP) &&
 603                     (ifp->if_flags & IFF_RUNNING)) {
 604                         /*
 605                          * If interface is marked down and it is running,
 606                          * then stop it.
 607                          */
 608                         bridge_stop(ifp);
 609                 } else if ((ifp->if_flags & IFF_UP) &&
 610                     !(ifp->if_flags & IFF_RUNNING)) {
 611                         /*
 612                          * If interface is marked up and it is stopped, then
 613                          * start it.
 614                          */
 615                         ifp->if_init(sc);
 616                 }
 617                 break;
 618
 619         case SIOCSIFMTU:
 620                 /* Do not allow the MTU to be changed on the bridge */
 621                 error = EINVAL;
 622                 break;
 623
 624         default:
 625                 error = ether_ioctl(ifp, cmd, data);
 626                 break;
 627         }
 628         return (error);
 629 }
 630
 631 /*
 632  * bridge_mutecaps:
 633  *
 634  *      Clear or restore unwanted capabilities on the member interface
 635  */
 636 static void
 637 bridge_mutecaps(struct bridge_iflist *bif, int mute)
 638 {
 639         struct ifnet *ifp = bif->bif_ifp;
 640         struct ifreq ifr;
 641         int error;
 642
 643         if (ifp->if_ioctl == NULL)
 644                 return;
 645
 646         bzero(&ifr, sizeof(ifr));
 647         ifr.ifr_reqcap = ifp->if_capenable;
 648
 649         if (mute) {
 650                 /* mask off and save capabilities */
 651                 bif->bif_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
 652                 if (bif->bif_mutecap != 0)
 653                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
 654         } else
 655                 /* restore muted capabilities */
 656                 ifr.ifr_reqcap |= bif->bif_mutecap;
 657
 658         if (bif->bif_mutecap != 0) {
 659                 lwkt_serialize_enter(ifp->if_serializer);
 660                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL);
 661                 lwkt_serialize_exit(ifp->if_serializer);
 662         }
 663 }
 664
 665 /*
 666  * bridge_lookup_member:
 667  *
 668  *      Lookup a bridge member interface.
 669  */
 670 static struct bridge_iflist *
 671 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 672 {
 673         struct bridge_iflist *bif;
 674         struct ifnet *ifp;
 675
 676         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 677                 ifp = bif->bif_ifp;
 678                 if (strcmp(ifp->if_xname, name) == 0)
 679                         return (bif);
 680         }
 681
 682         return (NULL);
 683 }
 684
 685 /*
 686  * bridge_lookup_member_if:
 687  *
 688  *      Lookup a bridge member interface by ifnet*.
 689  */
 690 static struct bridge_iflist *
 691 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 692 {
 693         struct bridge_iflist *bif;
 694
 695         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 696                 if (bif->bif_ifp == member_ifp)
 697                         return (bif);
 698         }
 699
 700         return (NULL);
 701 }
 702
 703 /*
 704  * bridge_delete_member:
 705  *
 706  *      Delete the specified member interface.
 707  */
 708 static void
 709 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 710     int gone)
 711 {
 712         struct ifnet *ifs = bif->bif_ifp;
 713
 714         if (!gone) {
 715                 switch (ifs->if_type) {
 716                 case IFT_ETHER:
 717                 case IFT_L2VLAN:
 718                         /*
 719                          * Take the interface out of promiscuous mode.
 720                          */
 721                         (void) ifpromisc(ifs, 0);
 722                         bridge_mutecaps(bif, 0);
 723                         break;
 724
 725                 case IFT_GIF:
 726                         break;
 727
 728                 default:
 729 #ifdef DIAGNOSTIC
 730                         panic("bridge_delete_member: impossible");
 731 #endif
 732                         break;
 733                 }
 734         }
 735
 736         ifs->if_bridge = NULL;
 737         LIST_REMOVE(bif, bif_next);
 738
 739         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 740
 741         kfree(bif, M_DEVBUF);
 742
 743         if (sc->sc_ifp->if_flags & IFF_RUNNING)
 744                 bstp_initialization(sc);
 745 }
 746
 747 /*
 748  * bridge_delete_span:
 749  *
 750  *      Delete the specified span interface.
 751  */
 752 static void
 753 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
 754 {
 755         KASSERT(bif->bif_ifp->if_bridge == NULL,
 756             ("%s: not a span interface", __func__));
 757
 758         LIST_REMOVE(bif, bif_next);
 759         kfree(bif, M_DEVBUF);
 760 }
 761
 762 static int
 763 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
 764 {
 765         struct ifbreq *req = arg;
 766         struct bridge_iflist *bif = NULL;
 767         struct ifnet *ifs;
 768         int error = 0;
 769
 770         ifs = ifunit(req->ifbr_ifsname);
 771         if (ifs == NULL)
 772                 return (ENOENT);
 773
 774         /* If it's in the span list, it can't be a member. */
 775         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
 776                 if (ifs == bif->bif_ifp)
 777                         return (EBUSY);
 778
 779         /* Allow the first Ethernet member to define the MTU */
 780         if (ifs->if_type != IFT_GIF) {
 781                 if (LIST_EMPTY(&sc->sc_iflist))
 782                         sc->sc_ifp->if_mtu = ifs->if_mtu;
 783                 else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
 784                         if_printf(sc->sc_ifp, "invalid MTU for %s\n",
 785                             ifs->if_xname);
 786                         return (EINVAL);
 787                 }
 788         }
 789
 790         if (ifs->if_bridge == sc)
 791                 return (EEXIST);
 792
 793         if (ifs->if_bridge != NULL)
 794                 return (EBUSY);
 795
 796         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK|M_ZERO);
 797         bif->bif_ifp = ifs;
 798         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
 799         bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
 800         bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
 801
 802         switch (ifs->if_type) {
 803         case IFT_ETHER:
 804         case IFT_L2VLAN:
 805                 /*
 806                  * Place the interface into promiscuous mode.
 807                  */
 808                 error = ifpromisc(ifs, 1);
 809                 if (error)
 810                         goto out;
 811
 812                 bridge_mutecaps(bif, 1);
 813                 break;
 814
 815         case IFT_GIF: /* :^) */
 816                 break;
 817
 818         default:
 819                 error = EINVAL;
 820                 goto out;
 821         }
 822
 823         ifs->if_bridge = sc;
 824
 825         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
 826
 827         if (sc->sc_ifp->if_flags & IFF_RUNNING)
 828                 bstp_initialization(sc);
 829         else
 830                 bstp_stop(sc);
 831
 832 out:
 833         if (error) {
 834                 if (bif != NULL)
 835                         kfree(bif, M_DEVBUF);
 836         }
 837         return (error);
 838 }
 839
 840 static int
 841 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
 842 {
 843         struct ifbreq *req = arg;
 844         struct bridge_iflist *bif;
 845
 846         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
 847         if (bif == NULL)
 848                 return (ENOENT);
 849
 850         bridge_delete_member(sc, bif, 0);
 851
 852         return (0);
 853 }
 854
 855 static int
 856 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
 857 {
 858         struct ifbreq *req = arg;
 859         struct bridge_iflist *bif;
 860
 861         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
 862         if (bif == NULL)
 863                 return (ENOENT);
 864
 865         req->ifbr_ifsflags = bif->bif_flags;
 866         req->ifbr_state = bif->bif_state;
 867         req->ifbr_priority = bif->bif_priority;
 868         req->ifbr_path_cost = bif->bif_path_cost;
 869         req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
 870
 871         return (0);
 872 }
 873
 874 static int
 875 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
 876 {
 877         struct ifbreq *req = arg;
 878         struct bridge_iflist *bif;
 879
 880         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
 881         if (bif == NULL)
 882                 return (ENOENT);
 883
 884         if (req->ifbr_ifsflags & IFBIF_SPAN)
 885                 /* SPAN is readonly */
 886                 return (EINVAL);
 887
 888         if (req->ifbr_ifsflags & IFBIF_STP) {
 889                 switch (bif->bif_ifp->if_type) {
 890                 case IFT_ETHER:
 891                         /* These can do spanning tree. */
 892                         break;
 893
 894                 default:
 895                         /* Nothing else can. */
 896                         return (EINVAL);
 897                 }
 898         }
 899
 900         bif->bif_flags = req->ifbr_ifsflags;
 901
 902         if (sc->sc_ifp->if_flags & IFF_RUNNING)
 903                 bstp_initialization(sc);
 904
 905         return (0);
 906 }
 907
 908 static int
 909 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
 910 {
 911         struct ifbrparam *param = arg;
 912
 913         sc->sc_brtmax = param->ifbrp_csize;
 914         bridge_rttrim(sc);
 915
 916         return (0);
 917 }
 918
 919 static int
 920 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
 921 {
 922         struct ifbrparam *param = arg;
 923
 924         param->ifbrp_csize = sc->sc_brtmax;
 925
 926         return (0);
 927 }
 928
 929 static int
 930 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
 931 {
 932         struct ifbifconf *bifc = arg;
 933         struct bridge_iflist *bif;
 934         struct ifbreq breq;
 935         int count, len, error = 0;
 936
 937         count = 0;
 938         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
 939                 count++;
 940         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
 941                 count++;
 942
 943         if (bifc->ifbic_len == 0) {
 944                 bifc->ifbic_len = sizeof(breq) * count;
 945                 return (0);
 946         }
 947
 948         count = 0;
 949         len = bifc->ifbic_len;
 950         memset(&breq, 0, sizeof breq);
 951         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 952                 if (len < sizeof(breq))
 953                         break;
 954
 955                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
 956                     sizeof(breq.ifbr_ifsname));
 957                 breq.ifbr_ifsflags = bif->bif_flags;
 958                 breq.ifbr_state = bif->bif_state;
 959                 breq.ifbr_priority = bif->bif_priority;
 960                 breq.ifbr_path_cost = bif->bif_path_cost;
 961                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
 962                 error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
 963                 if (error)
 964                         break;
 965                 count++;
 966                 len -= sizeof(breq);
 967         }
 968         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
 969                 if (len < sizeof(breq))
 970                         break;
 971
 972                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
 973                     sizeof(breq.ifbr_ifsname));
 974                 breq.ifbr_ifsflags = bif->bif_flags;
 975                 breq.ifbr_state = bif->bif_state;
 976                 breq.ifbr_priority = bif->bif_priority;
 977                 breq.ifbr_path_cost = bif->bif_path_cost;
 978                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
 979                 error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
 980                 if (error)
 981                         break;
 982                 count++;
 983                 len -= sizeof(breq);
 984         }
 985
 986         bifc->ifbic_len = sizeof(breq) * count;
 987         return (error);
 988 }
 989
 990 static int
 991 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
 992 {
 993         struct ifbaconf *bac = arg;
 994         struct bridge_rtnode *brt;
 995         struct ifbareq bareq;
 996         int count = 0, error = 0, len;
 997
 998         if (bac->ifbac_len == 0)
 999                 return (0);
1000
1001         len = bac->ifbac_len;
1002         memset(&bareq, 0, sizeof(bareq));
1003         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1004                 if (len < sizeof(bareq))
1005                         goto out;
1006                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1007                     sizeof(bareq.ifba_ifsname));
1008                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1009                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1010                                 time_second < brt->brt_expire)
1011                         bareq.ifba_expire = brt->brt_expire - time_second;
1012                 else
1013                         bareq.ifba_expire = 0;
1014                 bareq.ifba_flags = brt->brt_flags;
1015
1016                 error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1017                 if (error)
1018                         goto out;
1019                 count++;
1020                 len -= sizeof(bareq);
1021         }
1022 out:
1023         bac->ifbac_len = sizeof(bareq) * count;
1024         return (error);
1025 }
1026
1027 static int
1028 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1029 {
1030         struct ifbareq *req = arg;
1031         struct bridge_iflist *bif;
1032         int error;
1033
1034         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1035         if (bif == NULL)
1036                 return (ENOENT);
1037
1038         error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1039             req->ifba_flags);
1040
1041         return (error);
1042 }
1043
1044 static int
1045 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1046 {
1047         struct ifbrparam *param = arg;
1048
1049         sc->sc_brttimeout = param->ifbrp_ctime;
1050
1051         return (0);
1052 }
1053
1054 static int
1055 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1056 {
1057         struct ifbrparam *param = arg;
1058
1059         param->ifbrp_ctime = sc->sc_brttimeout;
1060
1061         return (0);
1062 }
1063
1064 static int
1065 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1066 {
1067         struct ifbareq *req = arg;
1068
1069         return (bridge_rtdaddr(sc, req->ifba_dst));
1070 }
1071
1072 static int
1073 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1074 {
1075         struct ifbreq *req = arg;
1076
1077         bridge_rtflush(sc, req->ifbr_ifsflags);
1078
1079         return (0);
1080 }
1081
1082 static int
1083 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1084 {
1085         struct ifbrparam *param = arg;
1086
1087         param->ifbrp_prio = sc->sc_bridge_priority;
1088
1089         return (0);
1090 }
1091
1092 static int
1093 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1094 {
1095         struct ifbrparam *param = arg;
1096
1097         sc->sc_bridge_priority = param->ifbrp_prio;
1098
1099         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1100                 bstp_initialization(sc);
1101
1102         return (0);
1103 }
1104
1105 static int
1106 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1107 {
1108         struct ifbrparam *param = arg;
1109
1110         param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1111
1112         return (0);
1113 }
1114
1115 static int
1116 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1117 {
1118         struct ifbrparam *param = arg;
1119
1120         if (param->ifbrp_hellotime == 0)
1121                 return (EINVAL);
1122         sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1123
1124         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1125                 bstp_initialization(sc);
1126
1127         return (0);
1128 }
1129
1130 static int
1131 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1132 {
1133         struct ifbrparam *param = arg;
1134
1135         param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1136
1137         return (0);
1138 }
1139
1140 static int
1141 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1142 {
1143         struct ifbrparam *param = arg;
1144
1145         if (param->ifbrp_fwddelay == 0)
1146                 return (EINVAL);
1147         sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1148
1149         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1150                 bstp_initialization(sc);
1151
1152         return (0);
1153 }
1154
1155 static int
1156 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1157 {
1158         struct ifbrparam *param = arg;
1159
1160         param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1161
1162         return (0);
1163 }
1164
1165 static int
1166 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1167 {
1168         struct ifbrparam *param = arg;
1169
1170         if (param->ifbrp_maxage == 0)
1171                 return (EINVAL);
1172         sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1173
1174         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1175                 bstp_initialization(sc);
1176
1177         return (0);
1178 }
1179
1180 static int
1181 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1182 {
1183         struct ifbreq *req = arg;
1184         struct bridge_iflist *bif;
1185
1186         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1187         if (bif == NULL)
1188                 return (ENOENT);
1189
1190         bif->bif_priority = req->ifbr_priority;
1191
1192         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1193                 bstp_initialization(sc);
1194
1195         return (0);
1196 }
1197
1198 static int
1199 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1200 {
1201         struct ifbreq *req = arg;
1202         struct bridge_iflist *bif;
1203
1204         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1205         if (bif == NULL)
1206                 return (ENOENT);
1207
1208         bif->bif_path_cost = req->ifbr_path_cost;
1209
1210         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1211                 bstp_initialization(sc);
1212
1213         return (0);
1214 }
1215
1216 static int
1217 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1218 {
1219         struct ifbreq *req = arg;
1220         struct bridge_iflist *bif = NULL;
1221         struct ifnet *ifs;
1222
1223         ifs = ifunit(req->ifbr_ifsname);
1224         if (ifs == NULL)
1225                 return (ENOENT);
1226
1227         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1228                 if (ifs == bif->bif_ifp)
1229                         return (EBUSY);
1230
1231         if (ifs->if_bridge != NULL)
1232                 return (EBUSY);
1233
1234         switch (ifs->if_type) {
1235                 case IFT_ETHER:
1236                 case IFT_GIF:
1237                 case IFT_L2VLAN:
1238                         break;
1239                 default:
1240                         return (EINVAL);
1241         }
1242
1243         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK|M_ZERO);
1244
1245         bif->bif_ifp = ifs;
1246         bif->bif_flags = IFBIF_SPAN;
1247
1248         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1249
1250         return (0);
1251 }
1252
1253 static int
1254 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1255 {
1256         struct ifbreq *req = arg;
1257         struct bridge_iflist *bif;
1258         struct ifnet *ifs;
1259
1260         ifs = ifunit(req->ifbr_ifsname);
1261         if (ifs == NULL)
1262                 return (ENOENT);
1263
1264         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1265                 if (ifs == bif->bif_ifp)
1266                         break;
1267
1268         if (bif == NULL)
1269                 return (ENOENT);
1270
1271         bridge_delete_span(sc, bif);
1272
1273         return (0);
1274 }
1275
1276 /*
1277  * bridge_ifdetach:
1278  *
1279  *      Detach an interface from a bridge.  Called when a member
1280  *      interface is detaching.
1281  */
1282 static void
1283 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1284 {
1285         struct bridge_softc *sc = ifp->if_bridge;
1286         struct bridge_iflist *bif;
1287
1288         /* Check if the interface is a bridge member */
1289         if (sc != NULL) {
1290                 lwkt_serialize_enter(ifp->if_serializer);
1291
1292                 bif = bridge_lookup_member_if(sc, ifp);
1293                 if (bif != NULL)
1294                         bridge_delete_member(sc, bif, 1);
1295
1296                 lwkt_serialize_exit(ifp->if_serializer);
1297                 return;
1298         }
1299
1300         /* Check if the interface is a span port */
1301         LIST_FOREACH(sc, &bridge_list, sc_list) {
1302                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1303                         if (ifp == bif->bif_ifp) {
1304                                 bridge_delete_span(sc, bif);
1305                                 break;
1306                         }
1307         }
1308 }
1309
1310 /*
1311  * bridge_init:
1312  *
1313  *      Initialize a bridge interface.
1314  */
1315 static void
1316 bridge_init(void *xsc)
1317 {
1318         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1319         struct ifnet *ifp = sc->sc_ifp;
1320
1321         ASSERT_SERIALIZED(ifp->if_serializer);
1322
1323         if (ifp->if_flags & IFF_RUNNING)
1324                 return;
1325
1326         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1327             bridge_timer, sc);
1328
1329         ifp->if_flags |= IFF_RUNNING;
1330         bstp_initialization(sc);
1331         return;
1332 }
1333
1334 /*
1335  * bridge_stop:
1336  *
1337  *      Stop the bridge interface.
1338  */
1339 static void
1340 bridge_stop(struct ifnet *ifp)
1341 {
1342         struct bridge_softc *sc = ifp->if_softc;
1343
1344         ASSERT_SERIALIZED(ifp->if_serializer);
1345
1346         if ((ifp->if_flags & IFF_RUNNING) == 0)
1347                 return;
1348
1349         callout_stop(&sc->sc_brcallout);
1350         bstp_stop(sc);
1351
1352         bridge_rtflush(sc, IFBF_FLUSHDYN);
1353
1354         ifp->if_flags &= ~IFF_RUNNING;
1355 }
1356
1357 static void
1358 bridge_enqueue_internal(struct ifnet *dst_ifp, struct mbuf *m,
1359                         netisr_fn_t handler)
1360 {
1361         struct netmsg_packet *nmp;
1362         lwkt_port_t port;
1363         int cpu = mycpu->gd_cpuid;
1364
1365         while (m->m_type == MT_TAG) {
1366                 /* XXX see ether_output_frame for full rules check */
1367                 m = m->m_next;
1368         }
1369
1370         nmp = &m->m_hdr.mh_netmsg;
1371         netmsg_init(&nmp->nm_netmsg, &netisr_apanic_rport, 0, handler);
1372         nmp->nm_packet = m;
1373         nmp->nm_netmsg.nm_lmsg.u.ms_resultp = dst_ifp;
1374
1375         port = cpu_portfn(cpu);
1376         lwkt_sendmsg(port, &nmp->nm_netmsg.nm_lmsg);
1377 }
1378
1379 static void
1380 bridge_pfil_enqueue(struct ifnet *dst_ifp, struct mbuf *m,
1381                     int runfilt)
1382 {
1383         netisr_fn_t handler;
1384
1385         if (runfilt && (inet_pfil_hook.ph_hashooks > 0
1386 #ifdef INET6
1387             || inet6_pfil_hook.ph_hashooks > 0
1388 #endif
1389             )) {
1390                 handler = bridge_pfil_enqueue_handler;
1391         } else {
1392                 handler = bridge_enqueue_handler;
1393         }
1394         bridge_enqueue_internal(dst_ifp, m, handler);
1395 }
1396
1397 /*
1398  * bridge_enqueue:
1399  *
1400  *      Enqueue a packet on a bridge member interface.
1401  *
1402  */
1403 void
1404 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m)
1405 {
1406         bridge_enqueue_internal(dst_ifp, m, bridge_enqueue_handler);
1407 }
1408
1409 /*
1410  * bridge_output_serialized:
1411  *
1412  *      Send output from a bridge member interface.  This
1413  *      performs the bridging function for locally originated
1414  *      packets.
1415  *
1416  *      The mbuf has the Ethernet header already attached.  We must
1417  *      enqueue or free the mbuf before returning.
1418  */
1419 static int
1420 bridge_output_serialized(struct ifnet *ifp, struct mbuf *m,
1421     struct sockaddr *sa, struct rtentry *rt)
1422 {
1423         struct ether_header *eh;
1424         struct ifnet *dst_if;
1425         struct bridge_softc *sc;
1426
1427         sc = ifp->if_bridge;
1428
1429         ASSERT_SERIALIZED(ifp->if_serializer);
1430
1431         if (m->m_len < ETHER_HDR_LEN) {
1432                 m = m_pullup(m, ETHER_HDR_LEN);
1433                 if (m == NULL)
1434                         return (0);
1435         }
1436
1437         /*
1438          * Serialize our bridge interface.  We have to get rid of the
1439          * originating interface lock to avoid a deadlock.
1440          */
1441         lwkt_serialize_exit(ifp->if_serializer);
1442         lwkt_serialize_enter(sc->sc_ifp->if_serializer);
1443
1444         eh = mtod(m, struct ether_header *);
1445
1446         /*
1447          * If bridge is down, but the original output interface is up,
1448          * go ahead and send out that interface.  Otherwise, the packet
1449          * is dropped below.
1450          */
1451         if ((sc->sc_ifp->if_flags & IFF_RUNNING) == 0) {
1452                 dst_if = ifp;
1453                 goto sendunicast;
1454         }
1455
1456         /*
1457          * If the packet is a multicast, or we don't know a better way to
1458          * get there, send to all interfaces.
1459          */
1460         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1461                 dst_if = NULL;
1462         else
1463                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1464         if (dst_if == NULL) {
1465                 struct bridge_iflist *bif;
1466                 struct mbuf *mc;
1467                 int used = 0;
1468
1469                 bridge_span(sc, m);
1470
1471                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1472                         dst_if = bif->bif_ifp;
1473                         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1474                                 continue;
1475
1476                         /*
1477                          * If this is not the original output interface,
1478                          * and the interface is participating in spanning
1479                          * tree, make sure the port is in a state that
1480                          * allows forwarding.
1481                          */
1482                         if (dst_if != ifp &&
1483                             (bif->bif_flags & IFBIF_STP) != 0) {
1484                                 switch (bif->bif_state) {
1485                                 case BSTP_IFSTATE_BLOCKING:
1486                                 case BSTP_IFSTATE_LISTENING:
1487                                 case BSTP_IFSTATE_DISABLED:
1488                                         continue;
1489                                 }
1490                         }
1491
1492                         if (LIST_NEXT(bif, bif_next) == NULL) {
1493                                 used = 1;
1494                                 mc = m;
1495                         } else {
1496                                 mc = m_copypacket(m, MB_DONTWAIT);
1497                                 if (mc == NULL) {
1498                                         sc->sc_ifp->if_oerrors++;
1499                                         continue;
1500                                 }
1501                         }
1502                         bridge_enqueue(dst_if, mc);
1503                 }
1504                 if (used == 0)
1505                         m_freem(m);
1506                 lwkt_serialize_exit(sc->sc_ifp->if_serializer);
1507                 goto done;
1508         }
1509
1510 sendunicast:
1511         /*
1512          * XXX Spanning tree consideration here?
1513          */
1514
1515         bridge_span(sc, m);
1516         lwkt_serialize_exit(sc->sc_ifp->if_serializer);
1517         if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1518                 m_freem(m);
1519         } else {
1520                 bridge_enqueue(dst_if, m);
1521         }
1522 done:
1523         lwkt_serialize_enter(ifp->if_serializer);
1524         return (0);
1525 }
1526
1527 /*
1528  * bridge_start:
1529  *
1530  *      Start output on a bridge.
1531  *
1532  */
1533 static void
1534 bridge_start(struct ifnet *ifp)
1535 {
1536         struct bridge_softc *sc = ifp->if_softc;
1537
1538         ASSERT_SERIALIZED(ifp->if_serializer);
1539
1540         ifp->if_flags |= IFF_OACTIVE;
1541         for (;;) {
1542                 struct ifnet *dst_if = NULL;
1543                 struct ether_header *eh;
1544                 struct mbuf *m;
1545
1546                 m = ifq_dequeue(&ifp->if_snd, NULL);
1547                 if (m == NULL)
1548                         break;
1549
1550                 if (m->m_len < sizeof(*eh)) {
1551                         m = m_pullup(m, sizeof(*eh));
1552                         if (m == NULL) {
1553                                 ifp->if_oerrors++;
1554                                 continue;
1555                         }
1556                 }
1557                 eh = mtod(m, struct ether_header *);
1558
1559                 BPF_MTAP(ifp, m);
1560                 ifp->if_opackets++;
1561
1562                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0)
1563                         dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1564
1565                 if (dst_if == NULL)
1566                         bridge_broadcast(sc, ifp, m, 0);
1567                 else
1568                         bridge_enqueue(dst_if, m);
1569         }
1570         ifp->if_flags &= ~IFF_OACTIVE;
1571 }
1572
1573 /*
1574  * bridge_forward:
1575  *
1576  *      The forwarding function of the bridge.
1577  */
1578 static void
1579 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1580 {
1581         struct bridge_iflist *bif;
1582         struct ifnet *src_if, *dst_if, *ifp;
1583         struct ether_header *eh;
1584
1585         src_if = m->m_pkthdr.rcvif;
1586         ifp = sc->sc_ifp;
1587
1588         ASSERT_SERIALIZED(ifp->if_serializer);
1589
1590         ifp->if_ipackets++;
1591         ifp->if_ibytes += m->m_pkthdr.len;
1592
1593         /*
1594          * Look up the bridge_iflist.
1595          */
1596         bif = bridge_lookup_member_if(sc, src_if);
1597         if (bif == NULL) {
1598                 /* Interface is not a bridge member (anymore?) */
1599                 m_freem(m);
1600                 return;
1601         }
1602
1603         if (bif->bif_flags & IFBIF_STP) {
1604                 switch (bif->bif_state) {
1605                 case BSTP_IFSTATE_BLOCKING:
1606                 case BSTP_IFSTATE_LISTENING:
1607                 case BSTP_IFSTATE_DISABLED:
1608                         m_freem(m);
1609                         return;
1610                 }
1611         }
1612
1613         eh = mtod(m, struct ether_header *);
1614
1615         /*
1616          * If the interface is learning, and the source
1617          * address is valid and not multicast, record
1618          * the address.
1619          */
1620         if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1621             ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1622             (eh->ether_shost[0] == 0 &&
1623              eh->ether_shost[1] == 0 &&
1624              eh->ether_shost[2] == 0 &&
1625              eh->ether_shost[3] == 0 &&
1626              eh->ether_shost[4] == 0 &&
1627              eh->ether_shost[5] == 0) == 0) {
1628                 bridge_rtupdate(sc, eh->ether_shost, src_if, 0, IFBAF_DYNAMIC);
1629         }
1630
1631         if ((bif->bif_flags & IFBIF_STP) != 0 &&
1632             bif->bif_state == BSTP_IFSTATE_LEARNING) {
1633                 m_freem(m);
1634                 return;
1635         }
1636
1637         /*
1638          * At this point, the port either doesn't participate
1639          * in spanning tree or it is in the forwarding state.
1640          */
1641
1642         /*
1643          * If the packet is unicast, destined for someone on
1644          * "this" side of the bridge, drop it.
1645          */
1646         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1647                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1648                 if (src_if == dst_if) {
1649                         m_freem(m);
1650                         return;
1651                 }
1652         } else {
1653                 /* ...forward it to all interfaces. */
1654                 sc->sc_ifp->if_imcasts++;
1655                 dst_if = NULL;
1656         }
1657
1658         if (dst_if == NULL) {
1659                 bridge_broadcast(sc, src_if, m, 1);
1660                 return;
1661         }
1662
1663         /*
1664          * At this point, we're dealing with a unicast frame
1665          * going to a different interface.
1666          */
1667         if ((dst_if->if_flags & IFF_RUNNING) == 0) {
1668                 m_freem(m);
1669                 return;
1670         }
1671         bif = bridge_lookup_member_if(sc, dst_if);
1672         if (bif == NULL) {
1673                 /* Not a member of the bridge (anymore?) */
1674                 m_freem(m);
1675                 return;
1676         }
1677
1678         if (bif->bif_flags & IFBIF_STP) {
1679                 switch (bif->bif_state) {
1680                 case BSTP_IFSTATE_DISABLED:
1681                 case BSTP_IFSTATE_BLOCKING:
1682                         m_freem(m);
1683                         return;
1684                 }
1685         }
1686
1687         lwkt_serialize_exit(ifp->if_serializer);
1688
1689         /* run the packet filter */
1690         if (inet_pfil_hook.ph_hashooks > 0
1691 #ifdef INET6
1692             || inet6_pfil_hook.ph_hashooks > 0
1693 #endif
1694             ) {
1695                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
1696                         goto done;
1697                 if (m == NULL)
1698                         goto done;
1699
1700                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
1701                         goto done;
1702                 if (m == NULL)
1703                         goto done;
1704         }
1705         bridge_handoff(dst_if, m);
1706
1707         /*
1708          * ifp's serializer was held on entry and is expected to be held
1709          * on return.
1710          */
1711 done:
1712         lwkt_serialize_enter(ifp->if_serializer);
1713 }
1714
1715 /*
1716  * bridge_input:
1717  *
1718  *      Receive input from a member interface.  Queue the packet for
1719  *      bridging if it is not for us.
1720  */
1721 static struct mbuf *
1722 bridge_input(struct ifnet *ifp, struct mbuf *m)
1723 {
1724         struct bridge_softc *sc = ifp->if_bridge;
1725         struct bridge_iflist *bif;
1726         struct ifnet *bifp, *new_ifp;
1727         struct ether_header *eh;
1728         struct mbuf *mc, *mc2;
1729
1730         new_ifp = NULL;
1731         bifp = sc->sc_ifp;
1732
1733         lwkt_serialize_enter(bifp->if_serializer);
1734
1735         if ((bifp->if_flags & IFF_RUNNING) == 0)
1736                 goto out;
1737
1738         /*
1739          * Implement support for bridge monitoring. If this flag has been
1740          * set on this interface, discard the packet once we push it through
1741          * the bpf(4) machinery, but before we do, increment the byte and
1742          * packet counters associated with this interface.
1743          */
1744          if ((bifp->if_flags & IFF_MONITOR) != 0) {
1745                 m->m_pkthdr.rcvif = bifp;
1746                 BPF_MTAP(bifp, m);
1747                 bifp->if_ipackets++;
1748                 bifp->if_ibytes += m->m_pkthdr.len;
1749                 m_freem(m);
1750                 m = NULL;
1751                 goto out;
1752         }
1753
1754         eh = mtod(m, struct ether_header *);
1755
1756         m->m_flags &= ~M_PROTO1; /* XXX Hack - loop prevention */
1757
1758         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
1759                 /*
1760                  * If the packet is for us, set the packets source as the
1761                  * bridge, and return the packet back to ifnet.if_input for
1762                  * local processing.
1763                  */
1764                 KASSERT(bifp->if_bridge == NULL,
1765                         ("loop created in bridge_input"));
1766                 new_ifp = bifp;
1767                 goto out;
1768         }
1769
1770         /*
1771          * Tap all packets arriving on the bridge, no matter if
1772          * they are local destinations or not.  In is in.
1773          */
1774         BPF_MTAP(bifp, m);
1775
1776         bif = bridge_lookup_member_if(sc, ifp);
1777         if (bif == NULL)
1778                 goto out;
1779
1780         bridge_span(sc, m);
1781
1782         if (m->m_flags & (M_BCAST | M_MCAST)) {
1783                 /* Tap off 802.1D packets; they do not get forwarded. */
1784                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
1785                     ETHER_ADDR_LEN) == 0) {
1786                         m = bstp_input(ifp, m);
1787                         KASSERT(m == NULL,
1788                                 ("attempt to deliver 802.1D packet\n"));
1789                         goto out;
1790                 }
1791
1792                 if (bif->bif_flags & IFBIF_STP) {
1793                         switch (bif->bif_state) {
1794                         case BSTP_IFSTATE_BLOCKING:
1795                         case BSTP_IFSTATE_LISTENING:
1796                         case BSTP_IFSTATE_DISABLED:
1797                                 goto out;
1798                         }
1799                 }
1800
1801                 /*
1802                  * Make a deep copy of the packet and enqueue the copy
1803                  * for bridge processing; return the original packet for
1804                  * local processing.
1805                  */
1806                 mc = m_dup(m, MB_DONTWAIT);
1807                 if (mc == NULL)
1808                         goto out;
1809
1810                 bridge_forward(sc, mc);
1811
1812                 /*
1813                  * Reinject the mbuf as arriving on the bridge so we have a
1814                  * chance at claiming multicast packets. We can not loop back
1815                  * here from ether_input as a bridge is never a member of a
1816                  * bridge.
1817                  */
1818                 KASSERT(bifp->if_bridge == NULL,
1819                         ("loop created in bridge_input"));
1820                 mc2 = m_dup(m, MB_DONTWAIT);
1821 #ifdef notyet
1822                 if (mc2 != NULL) {
1823                         /* Keep the layer3 header aligned */
1824                         int i = min(mc2->m_pkthdr.len, max_protohdr);
1825                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
1826                 }
1827 #endif
1828                 if (mc2 != NULL) {
1829                         mc2->m_pkthdr.rcvif = bifp;
1830                         bifp->if_ipackets++;
1831                         bifp->if_input(bifp, mc2);
1832                 }
1833
1834                 /* Return the original packet for local processing. */
1835                 goto out;
1836         }
1837
1838         if (bif->bif_flags & IFBIF_STP) {
1839                 switch (bif->bif_state) {
1840                 case BSTP_IFSTATE_BLOCKING:
1841                 case BSTP_IFSTATE_LISTENING:
1842                 case BSTP_IFSTATE_DISABLED:
1843                         goto out;
1844                 }
1845         }
1846
1847         /*
1848          * Unicast.  Make sure it's not for us.
1849          */
1850         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1851                 if (bif->bif_ifp->if_type != IFT_ETHER)
1852                         continue;
1853
1854                 /* It is destined for us. */
1855                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
1856                     ETHER_ADDR_LEN) == 0) {
1857                         if (bif->bif_flags & IFBIF_LEARNING) {
1858                                 bridge_rtupdate(sc,
1859                                     eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
1860                         }
1861
1862                         if (bif->bif_ifp != ifp) {
1863                                 /* XXX loop prevention */
1864                                 m->m_flags |= M_PROTO1;
1865                                 new_ifp = bif->bif_ifp;
1866                         }
1867                         goto out;
1868                 }
1869
1870                 /* We just received a packet that we sent out. */
1871                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
1872                     ETHER_ADDR_LEN) == 0) {
1873                         m_freem(m);
1874                         m = NULL;
1875                         goto out;
1876                 }
1877         }
1878
1879         /* Perform the bridge forwarding function. */
1880         bridge_forward(sc, m);
1881         m = NULL;
1882 out:
1883         lwkt_serialize_exit(bifp->if_serializer);
1884
1885         if (new_ifp != NULL) {
1886                 lwkt_serialize_enter(new_ifp->if_serializer);
1887
1888                 m->m_pkthdr.rcvif = new_ifp;
1889                 new_ifp->if_ipackets++;
1890                 new_ifp->if_input(new_ifp, m);
1891                 m = NULL;
1892
1893                 lwkt_serialize_exit(new_ifp->if_serializer);
1894         }
1895         return (m);
1896 }
1897
1898 /*
1899  * bridge_broadcast:
1900  *
1901  *      Send a frame to all interfaces that are members of
1902  *      the bridge, except for the one on which the packet
1903  *      arrived.
1904  */
1905 static void
1906 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
1907     struct mbuf *m, int runfilt)
1908 {
1909         struct bridge_iflist *bif;
1910         struct mbuf *mc;
1911         struct ifnet *dst_if, *bifp;
1912         int used = 0;
1913
1914         bifp = sc->sc_ifp;
1915
1916         ASSERT_SERIALIZED(bifp->if_serializer);
1917
1918         /* run the packet filter */
1919         if (runfilt && (inet_pfil_hook.ph_hashooks > 0
1920 #ifdef INET6
1921             || inet6_pfil_hook.ph_hashooks > 0
1922 #endif
1923             )) {
1924                 lwkt_serialize_exit(bifp->if_serializer);
1925
1926                 /* Filter on the bridge interface before broadcasting */
1927
1928                 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0)
1929                         goto filt;
1930                 if (m == NULL)
1931                         goto filt;
1932
1933                 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0)
1934                         m = NULL;
1935 filt:
1936                 lwkt_serialize_enter(bifp->if_serializer);
1937                 if (m == NULL)
1938                         return;
1939         }
1940
1941         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1942                 dst_if = bif->bif_ifp;
1943                 if (dst_if == src_if)
1944                         continue;
1945
1946                 if (bif->bif_flags & IFBIF_STP) {
1947                         switch (bif->bif_state) {
1948                         case BSTP_IFSTATE_BLOCKING:
1949                         case BSTP_IFSTATE_DISABLED:
1950                                 continue;
1951                         }
1952                 }
1953
1954                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
1955                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
1956                         continue;
1957
1958                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1959                         continue;
1960
1961                 if (LIST_NEXT(bif, bif_next) == NULL) {
1962                         mc = m;
1963                         used = 1;
1964                 } else {
1965                         mc = m_copypacket(m, MB_DONTWAIT);
1966                         if (mc == NULL) {
1967                                 sc->sc_ifp->if_oerrors++;
1968                                 continue;
1969                         }
1970                 }
1971                 bridge_pfil_enqueue(dst_if, mc, runfilt);
1972         }
1973         if (used == 0)
1974                 m_freem(m);
1975 }
1976
1977 /*
1978  * bridge_span:
1979  *
1980  *      Duplicate a packet out one or more interfaces that are in span mode,
1981  *      the original mbuf is unmodified.
1982  */
1983 static void
1984 bridge_span(struct bridge_softc *sc, struct mbuf *m)
1985 {
1986         struct bridge_iflist *bif;
1987         struct ifnet *dst_if;
1988         struct mbuf *mc;
1989
1990         if (LIST_EMPTY(&sc->sc_spanlist))
1991                 return;
1992
1993         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1994                 dst_if = bif->bif_ifp;
1995
1996                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
1997                         continue;
1998
1999                 mc = m_copypacket(m, MB_DONTWAIT);
2000                 if (mc == NULL) {
2001                         sc->sc_ifp->if_oerrors++;
2002                         continue;
2003                 }
2004
2005                 bridge_enqueue(dst_if, mc);
2006         }
2007 }
2008
2009 /*
2010  * bridge_rtupdate:
2011  *
2012  *      Add a bridge routing entry.
2013  *      Can be called from interrupt context.
2014  */
2015 static int
2016 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2017     struct ifnet *dst_if, int setflags, uint8_t flags)
2018 {
2019         struct bridge_rtnode *brt;
2020         int error;
2021
2022         /*
2023          * A route for this destination might already exist.  If so,
2024          * update it, otherwise create a new one.
2025          */
2026         if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
2027                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2028                         return (ENOSPC);
2029
2030                 /*
2031                  * Allocate a new bridge forwarding node, and
2032                  * initialize the expiration time and Ethernet
2033                  * address.
2034                  */
2035                 brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF,
2036                               M_INTNOWAIT|M_ZERO);
2037                 if (brt == NULL)
2038                         return (ENOMEM);
2039
2040                 brt->brt_flags = IFBAF_DYNAMIC;
2041                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2042
2043                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2044                         kfree(brt, M_DEVBUF);
2045                         return (error);
2046                 }
2047         }
2048
2049         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2050                 brt->brt_ifp = dst_if;
2051         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2052                 brt->brt_expire = time_second + sc->sc_brttimeout;
2053         if (setflags)
2054                 brt->brt_flags = flags;
2055
2056         return (0);
2057 }
2058
2059 /*
2060  * bridge_rtlookup:
2061  *
2062  *      Lookup the destination interface for an address.
2063  */
2064 static struct ifnet *
2065 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2066 {
2067         struct bridge_rtnode *brt;
2068
2069         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2070                 return (NULL);
2071
2072         return (brt->brt_ifp);
2073 }
2074
2075 /*
2076  * bridge_rttrim:
2077  *
2078  *      Trim the routine table so that we have a number
2079  *      of routing entries less than or equal to the
2080  *      maximum number.
2081  */
2082 static void
2083 bridge_rttrim(struct bridge_softc *sc)
2084 {
2085         struct bridge_rtnode *brt, *nbrt;
2086
2087         /* Make sure we actually need to do this. */
2088         if (sc->sc_brtcnt <= sc->sc_brtmax)
2089                 return;
2090
2091         /* Force an aging cycle; this might trim enough addresses. */
2092         bridge_rtage(sc);
2093         if (sc->sc_brtcnt <= sc->sc_brtmax)
2094                 return;
2095
2096         for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2097                 nbrt = LIST_NEXT(brt, brt_list);
2098                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2099                         bridge_rtnode_destroy(sc, brt);
2100                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2101                                 return;
2102                 }
2103         }
2104 }
2105
2106 /*
2107  * bridge_timer:
2108  *
2109  *      Aging timer for the bridge.
2110  */
2111 static void
2112 bridge_timer(void *arg)
2113 {
2114         struct bridge_softc *sc = arg;
2115
2116         lwkt_serialize_enter(sc->sc_ifp->if_serializer);
2117
2118         bridge_rtage(sc);
2119
2120         if (sc->sc_ifp->if_flags & IFF_RUNNING)
2121                 callout_reset(&sc->sc_brcallout,
2122                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2123
2124         lwkt_serialize_exit(sc->sc_ifp->if_serializer);
2125 }
2126
2127 /*
2128  * bridge_rtage:
2129  *
2130  *      Perform an aging cycle.
2131  */
2132 static void
2133 bridge_rtage(struct bridge_softc *sc)
2134 {
2135         struct bridge_rtnode *brt, *nbrt;
2136
2137         for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2138                 nbrt = LIST_NEXT(brt, brt_list);
2139                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2140                         if (time_second >= brt->brt_expire)
2141                                 bridge_rtnode_destroy(sc, brt);
2142                 }
2143         }
2144 }
2145
2146 /*
2147  * bridge_rtflush:
2148  *
2149  *      Remove all dynamic addresses from the bridge.
2150  */
2151 static void
2152 bridge_rtflush(struct bridge_softc *sc, int full)
2153 {
2154         struct bridge_rtnode *brt, *nbrt;
2155
2156         for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2157                 nbrt = LIST_NEXT(brt, brt_list);
2158                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2159                         bridge_rtnode_destroy(sc, brt);
2160         }
2161 }
2162
2163 /*
2164  * bridge_rtdaddr:
2165  *
2166  *      Remove an address from the table.
2167  */
2168 static int
2169 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2170 {
2171         struct bridge_rtnode *brt;
2172
2173         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2174                 return (ENOENT);
2175
2176         bridge_rtnode_destroy(sc, brt);
2177         return (0);
2178 }
2179
2180 /*
2181  * bridge_rtdelete:
2182  *
2183  *      Delete routes to a speicifc member interface.
2184  */
2185 void
2186 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2187 {
2188         struct bridge_rtnode *brt, *nbrt;
2189
2190         for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2191                 nbrt = LIST_NEXT(brt, brt_list);
2192                 if (brt->brt_ifp == ifp && (full ||
2193                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2194                         bridge_rtnode_destroy(sc, brt);
2195         }
2196 }
2197
2198 /*
2199  * bridge_rtable_init:
2200  *
2201  *      Initialize the route table for this bridge.
2202  */
2203 static int
2204 bridge_rtable_init(struct bridge_softc *sc)
2205 {
2206         int i;
2207
2208         sc->sc_rthash = kmalloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2209             M_DEVBUF, M_WAITOK);
2210
2211         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2212                 LIST_INIT(&sc->sc_rthash[i]);
2213
2214         sc->sc_rthash_key = karc4random();
2215
2216         LIST_INIT(&sc->sc_rtlist);
2217
2218         return (0);
2219 }
2220
2221 /*
2222  * bridge_rtable_fini:
2223  *
2224  *      Deconstruct the route table for this bridge.
2225  */
2226 static void
2227 bridge_rtable_fini(struct bridge_softc *sc)
2228 {
2229
2230         kfree(sc->sc_rthash, M_DEVBUF);
2231 }
2232
2233 /*
2234  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2235  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2236  */
2237 #define mix(a, b, c)                                                    \
2238 do {                                                                    \
2239         a -= b; a -= c; a ^= (c >> 13);                                 \
2240         b -= c; b -= a; b ^= (a << 8);                                  \
2241         c -= a; c -= b; c ^= (b >> 13);                                 \
2242         a -= b; a -= c; a ^= (c >> 12);                                 \
2243         b -= c; b -= a; b ^= (a << 16);                                 \
2244         c -= a; c -= b; c ^= (b >> 5);                                  \
2245         a -= b; a -= c; a ^= (c >> 3);                                  \
2246         b -= c; b -= a; b ^= (a << 10);                                 \
2247         c -= a; c -= b; c ^= (b >> 15);                                 \
2248 } while (/*CONSTCOND*/0)
2249
2250 static __inline uint32_t
2251 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2252 {
2253         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2254
2255         b += addr[5] << 8;
2256         b += addr[4];
2257         a += addr[3] << 24;
2258         a += addr[2] << 16;
2259         a += addr[1] << 8;
2260         a += addr[0];
2261
2262         mix(a, b, c);
2263
2264         return (c & BRIDGE_RTHASH_MASK);
2265 }
2266
2267 #undef mix
2268
2269 static int
2270 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2271 {
2272         int i, d;
2273
2274         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2275                 d = ((int)a[i]) - ((int)b[i]);
2276         }
2277
2278         return (d);
2279 }
2280
2281 /*
2282  * bridge_rtnode_lookup:
2283  *
2284  *      Look up a bridge route node for the specified destination.
2285  */
2286 static struct bridge_rtnode *
2287 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2288 {
2289         struct bridge_rtnode *brt;
2290         uint32_t hash;
2291         int dir;
2292
2293         hash = bridge_rthash(sc, addr);
2294         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2295                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2296                 if (dir == 0)
2297                         return (brt);
2298                 if (dir > 0)
2299                         return (NULL);
2300         }
2301
2302         return (NULL);
2303 }
2304
2305 /*
2306  * bridge_rtnode_insert:
2307  *
2308  *      Insert the specified bridge node into the route table.  We
2309  *      assume the entry is not already in the table.
2310  */
2311 static int
2312 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2313 {
2314         struct bridge_rtnode *lbrt;
2315         uint32_t hash;
2316         int dir;
2317
2318         hash = bridge_rthash(sc, brt->brt_addr);
2319
2320         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2321         if (lbrt == NULL) {
2322                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2323                 goto out;
2324         }
2325
2326         do {
2327                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2328                 if (dir == 0)
2329                         return (EEXIST);
2330                 if (dir > 0) {
2331                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2332                         goto out;
2333                 }
2334                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2335                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2336                         goto out;
2337                 }
2338                 lbrt = LIST_NEXT(lbrt, brt_hash);
2339         } while (lbrt != NULL);
2340
2341 #ifdef DIAGNOSTIC
2342         panic("bridge_rtnode_insert: impossible");
2343 #endif
2344
2345 out:
2346         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2347         sc->sc_brtcnt++;
2348
2349         return (0);
2350 }
2351
2352 /*
2353  * bridge_rtnode_destroy:
2354  *
2355  *      Destroy a bridge rtnode.
2356  */
2357 static void
2358 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2359 {
2360
2361         LIST_REMOVE(brt, brt_hash);
2362
2363         LIST_REMOVE(brt, brt_list);
2364         sc->sc_brtcnt--;
2365         kfree(brt, M_DEVBUF);
2366 }
2367
2368 /*
2369  * Send bridge packets through pfil if they are one of the types pfil can deal
2370  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2371  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2372  * that interface.
2373  */
2374 static int
2375 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2376 {
2377         int snap, error, i, hlen;
2378         struct ether_header *eh1, eh2;
2379         struct ip *ip;
2380         struct llc llc1;
2381         u_int16_t ether_type;
2382
2383         snap = 0;
2384         error = -1;     /* Default error if not error == 0 */
2385
2386         if (pfil_bridge == 0 && pfil_member == 0)
2387                 return (0); /* filtering is disabled */
2388
2389         i = min((*mp)->m_pkthdr.len, max_protohdr);
2390         if ((*mp)->m_len < i) {
2391             *mp = m_pullup(*mp, i);
2392             if (*mp == NULL) {
2393                 kprintf("%s: m_pullup failed\n", __func__);
2394                 return (-1);
2395             }
2396         }
2397
2398         eh1 = mtod(*mp, struct ether_header *);
2399         ether_type = ntohs(eh1->ether_type);
2400
2401         /*
2402          * Check for SNAP/LLC.
2403          */
2404         if (ether_type < ETHERMTU) {
2405                 struct llc *llc2 = (struct llc *)(eh1 + 1);
2406
2407                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2408                     llc2->llc_dsap == LLC_SNAP_LSAP &&
2409                     llc2->llc_ssap == LLC_SNAP_LSAP &&
2410                     llc2->llc_control == LLC_UI) {
2411                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
2412                         snap = 1;
2413                 }
2414         }
2415
2416         /*
2417          * If we're trying to filter bridge traffic, don't look at anything
2418          * other than IP and ARP traffic.  If the filter doesn't understand
2419          * IPv6, don't allow IPv6 through the bridge either.  This is lame
2420          * since if we really wanted, say, an AppleTalk filter, we are hosed,
2421          * but of course we don't have an AppleTalk filter to begin with.
2422          * (Note that since pfil doesn't understand ARP it will pass *ALL*
2423          * ARP traffic.)
2424          */
2425         switch (ether_type) {
2426                 case ETHERTYPE_ARP:
2427                 case ETHERTYPE_REVARP:
2428                         return (0); /* Automatically pass */
2429                 case ETHERTYPE_IP:
2430 #ifdef INET6
2431                 case ETHERTYPE_IPV6:
2432 #endif /* INET6 */
2433                         break;
2434                 default:
2435                         /*
2436                          * Check to see if the user wants to pass non-ip
2437                          * packets, these will not be checked by pfil(9) and
2438                          * passed unconditionally so the default is to drop.
2439                          */
2440                         if (pfil_onlyip)
2441                                 goto bad;
2442         }
2443
2444         /* Strip off the Ethernet header and keep a copy. */
2445         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
2446         m_adj(*mp, ETHER_HDR_LEN);
2447
2448         /* Strip off snap header, if present */
2449         if (snap) {
2450                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
2451                 m_adj(*mp, sizeof(struct llc));
2452         }
2453
2454         /*
2455          * Check the IP header for alignment and errors
2456          */
2457         if (dir == PFIL_IN) {
2458                 switch (ether_type) {
2459                         case ETHERTYPE_IP:
2460                                 error = bridge_ip_checkbasic(mp);
2461                                 break;
2462 #ifdef INET6
2463                         case ETHERTYPE_IPV6:
2464                                 error = bridge_ip6_checkbasic(mp);
2465                                 break;
2466 #endif /* INET6 */
2467                         default:
2468                                 error = 0;
2469                 }
2470                 if (error)
2471                         goto bad;
2472         }
2473
2474         error = 0;
2475
2476         /*
2477          * Run the packet through pfil
2478          */
2479         switch (ether_type)
2480         {
2481         case ETHERTYPE_IP :
2482                 /*
2483                  * before calling the firewall, swap fields the same as
2484                  * IP does. here we assume the header is contiguous
2485                  */
2486                 ip = mtod(*mp, struct ip *);
2487
2488                 ip->ip_len = ntohs(ip->ip_len);
2489                 ip->ip_off = ntohs(ip->ip_off);
2490
2491                 /*
2492                  * Run pfil on the member interface and the bridge, both can
2493                  * be skipped by clearing pfil_member or pfil_bridge.
2494                  *
2495                  * Keep the order:
2496                  *   in_if -> bridge_if -> out_if
2497                  */
2498                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
2499                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
2500                                         dir);
2501
2502                 if (*mp == NULL || error != 0) /* filter may consume */
2503                         break;
2504
2505                 if (pfil_member && ifp != NULL)
2506                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp,
2507                                         dir);
2508
2509                 if (*mp == NULL || error != 0) /* filter may consume */
2510                         break;
2511
2512                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
2513                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
2514                                         dir);
2515
2516                 if (*mp == NULL || error != 0) /* filter may consume */
2517                         break;
2518
2519                 /* check if we need to fragment the packet */
2520                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
2521                         i = (*mp)->m_pkthdr.len;
2522                         if (i > ifp->if_mtu) {
2523                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
2524                                             &llc1);
2525                                 return (error);
2526                         }
2527                 }
2528
2529                 /* Recalculate the ip checksum and restore byte ordering */
2530                 ip = mtod(*mp, struct ip *);
2531                 hlen = ip->ip_hl << 2;
2532                 if (hlen < sizeof(struct ip))
2533                         goto bad;
2534                 if (hlen > (*mp)->m_len) {
2535                         if ((*mp = m_pullup(*mp, hlen)) == 0)
2536                                 goto bad;
2537                         ip = mtod(*mp, struct ip *);
2538                         if (ip == NULL)
2539                                 goto bad;
2540                 }
2541                 ip->ip_len = htons(ip->ip_len);
2542                 ip->ip_off = htons(ip->ip_off);
2543                 ip->ip_sum = 0;
2544                 if (hlen == sizeof(struct ip))
2545                         ip->ip_sum = in_cksum_hdr(ip);
2546                 else
2547                         ip->ip_sum = in_cksum(*mp, hlen);
2548
2549                 break;
2550 #ifdef INET6
2551         case ETHERTYPE_IPV6 :
2552                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
2553                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
2554                                         dir);
2555
2556                 if (*mp == NULL || error != 0) /* filter may consume */
2557                         break;
2558
2559                 if (pfil_member && ifp != NULL)
2560                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
2561                                         dir);
2562
2563                 if (*mp == NULL || error != 0) /* filter may consume */
2564                         break;
2565
2566                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
2567                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
2568                                         dir);
2569                 break;
2570 #endif
2571         default :
2572                 error = 0;
2573                 break;
2574         }
2575
2576         if (*mp == NULL)
2577                 return (error);
2578         if (error != 0)
2579                 goto bad;
2580
2581         error = -1;
2582
2583         /*
2584          * Finally, put everything back the way it was and return
2585          */
2586         if (snap) {
2587                 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT);
2588                 if (*mp == NULL)
2589                         return (error);
2590                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
2591         }
2592
2593         M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT);
2594         if (*mp == NULL)
2595                 return (error);
2596         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
2597
2598         return (0);
2599
2600 bad:
2601         m_freem(*mp);
2602         *mp = NULL;
2603         return (error);
2604 }
2605
2606 /*
2607  * Perform basic checks on header size since
2608  * pfil assumes ip_input has already processed
2609  * it for it.  Cut-and-pasted from ip_input.c.
2610  * Given how simple the IPv6 version is,
2611  * does the IPv4 version really need to be
2612  * this complicated?
2613  *
2614  * XXX Should we update ipstat here, or not?
2615  * XXX Right now we update ipstat but not
2616  * XXX csum_counter.
2617  */
2618 static int
2619 bridge_ip_checkbasic(struct mbuf **mp)
2620 {
2621         struct mbuf *m = *mp;
2622         struct ip *ip;
2623         int len, hlen;
2624         u_short sum;
2625
2626         if (*mp == NULL)
2627                 return (-1);
2628 #if notyet
2629         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
2630                 if ((m = m_copyup(m, sizeof(struct ip),
2631                         (max_linkhdr + 3) & ~3)) == NULL) {
2632                         /* XXXJRT new stat, please */
2633                         ipstat.ips_toosmall++;
2634                         goto bad;
2635                 }
2636         } else
2637 #endif
2638 #ifndef __predict_false
2639 #define __predict_false(x) x
2640 #endif
2641          if (__predict_false(m->m_len < sizeof (struct ip))) {
2642                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
2643                         ipstat.ips_toosmall++;
2644                         goto bad;
2645                 }
2646         }
2647         ip = mtod(m, struct ip *);
2648         if (ip == NULL) goto bad;
2649
2650         if (ip->ip_v != IPVERSION) {
2651                 ipstat.ips_badvers++;
2652                 goto bad;
2653         }
2654         hlen = ip->ip_hl << 2;
2655         if (hlen < sizeof(struct ip)) { /* minimum header length */
2656                 ipstat.ips_badhlen++;
2657                 goto bad;
2658         }
2659         if (hlen > m->m_len) {
2660                 if ((m = m_pullup(m, hlen)) == 0) {
2661                         ipstat.ips_badhlen++;
2662                         goto bad;
2663                 }
2664                 ip = mtod(m, struct ip *);
2665                 if (ip == NULL) goto bad;
2666         }
2667
2668         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
2669                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
2670         } else {
2671                 if (hlen == sizeof(struct ip)) {
2672                         sum = in_cksum_hdr(ip);
2673                 } else {
2674                         sum = in_cksum(m, hlen);
2675                 }
2676         }
2677         if (sum) {
2678                 ipstat.ips_badsum++;
2679                 goto bad;
2680         }
2681
2682         /* Retrieve the packet length. */
2683         len = ntohs(ip->ip_len);
2684
2685         /*
2686          * Check for additional length bogosity
2687          */
2688         if (len < hlen) {
2689                 ipstat.ips_badlen++;
2690                 goto bad;
2691         }
2692
2693         /*
2694          * Check that the amount of data in the buffers
2695          * is as at least much as the IP header would have us expect.
2696          * Drop packet if shorter than we expect.
2697          */
2698         if (m->m_pkthdr.len < len) {
2699                 ipstat.ips_tooshort++;
2700                 goto bad;
2701         }
2702
2703         /* Checks out, proceed */
2704         *mp = m;
2705         return (0);
2706
2707 bad:
2708         *mp = m;
2709         return (-1);
2710 }
2711
2712 #ifdef INET6
2713 /*
2714  * Same as above, but for IPv6.
2715  * Cut-and-pasted from ip6_input.c.
2716  * XXX Should we update ip6stat, or not?
2717  */
2718 static int
2719 bridge_ip6_checkbasic(struct mbuf **mp)
2720 {
2721         struct mbuf *m = *mp;
2722         struct ip6_hdr *ip6;
2723
2724         /*
2725          * If the IPv6 header is not aligned, slurp it up into a new
2726          * mbuf with space for link headers, in the event we forward
2727          * it.  Otherwise, if it is aligned, make sure the entire base
2728          * IPv6 header is in the first mbuf of the chain.
2729          */
2730 #if notyet
2731         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
2732                 struct ifnet *inifp = m->m_pkthdr.rcvif;
2733                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
2734                             (max_linkhdr + 3) & ~3)) == NULL) {
2735                         /* XXXJRT new stat, please */
2736                         ip6stat.ip6s_toosmall++;
2737                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2738                         goto bad;
2739                 }
2740         } else
2741 #endif
2742         if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
2743                 struct ifnet *inifp = m->m_pkthdr.rcvif;
2744                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
2745                         ip6stat.ip6s_toosmall++;
2746                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
2747                         goto bad;
2748                 }
2749         }
2750
2751         ip6 = mtod(m, struct ip6_hdr *);
2752
2753         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
2754                 ip6stat.ip6s_badvers++;
2755                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
2756                 goto bad;
2757         }
2758
2759         /* Checks out, proceed */
2760         *mp = m;
2761         return (0);
2762
2763 bad:
2764         *mp = m;
2765         return (-1);
2766 }
2767 #endif /* INET6 */
2768
2769 /*
2770  * bridge_fragment:
2771  *
2772  *      Return a fragmented mbuf chain.
2773  */
2774 static int
2775 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
2776     int snap, struct llc *llc)
2777 {
2778         struct mbuf *m0;
2779         struct ip *ip;
2780         int error = -1;
2781
2782         if (m->m_len < sizeof(struct ip) &&
2783             (m = m_pullup(m, sizeof(struct ip))) == NULL)
2784                 goto out;
2785         ip = mtod(m, struct ip *);
2786
2787         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
2788                     CSUM_DELAY_IP);
2789         if (error)
2790                 goto out;
2791
2792         /* walk the chain and re-add the Ethernet header */
2793         for (m0 = m; m0; m0 = m0->m_nextpkt) {
2794                 if (error == 0) {
2795                         if (snap) {
2796                                 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT);
2797                                 if (m0 == NULL) {
2798                                         error = ENOBUFS;
2799                                         continue;
2800                                 }
2801                                 bcopy(llc, mtod(m0, caddr_t),
2802                                     sizeof(struct llc));
2803                         }
2804                         M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT);
2805                         if (m0 == NULL) {
2806                                 error = ENOBUFS;
2807                                 continue;
2808                         }
2809                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
2810                 } else
2811                         m_freem(m);
2812         }
2813
2814         if (error == 0)
2815                 ipstat.ips_fragmented++;
2816
2817         return (error);
2818
2819 out:
2820         if (m != NULL)
2821                 m_freem(m);
2822         return (error);
2823 }
2824
2825 static void
2826 bridge_enqueue_handler(struct netmsg *nmsg)
2827 {
2828         struct netmsg_packet *nmp;
2829         struct ifnet *dst_ifp;
2830         struct mbuf *m;
2831
2832         nmp = (struct netmsg_packet *)nmsg;
2833         m = nmp->nm_packet;
2834         dst_ifp = nmp->nm_netmsg.nm_lmsg.u.ms_resultp;
2835
2836         bridge_handoff_notags(dst_ifp, m);
2837 }
2838
2839 static void
2840 bridge_pfil_enqueue_handler(struct netmsg *nmsg)
2841 {
2842         struct netmsg_packet *nmp;
2843         struct ifnet *dst_ifp;
2844         struct mbuf *m;
2845
2846         nmp = (struct netmsg_packet *)nmsg;
2847         m = nmp->nm_packet;
2848         dst_ifp = nmp->nm_netmsg.nm_lmsg.u.ms_resultp;
2849
2850         /*
2851          * Filter on the output interface. Pass a NULL bridge interface
2852          * pointer so we do not redundantly filter on the bridge for
2853          * each interface we broadcast on.
2854          */
2855         if (inet_pfil_hook.ph_hashooks > 0
2856 #ifdef INET6
2857             || inet6_pfil_hook.ph_hashooks > 0
2858 #endif
2859             ) {
2860                 if (bridge_pfil(&m, NULL, dst_ifp, PFIL_OUT) != 0)
2861                         return;
2862                 if (m == NULL)
2863                         return;
2864         }
2865         bridge_handoff_notags(dst_ifp, m);
2866 }
2867
2868 static void
2869 bridge_handoff(struct ifnet *dst_ifp, struct mbuf *m)
2870 {
2871         while (m->m_type == MT_TAG) {
2872                 /* XXX see ether_output_frame for full rules check */
2873                 m = m->m_next;
2874         }
2875         bridge_handoff_notags(dst_ifp, m);
2876 }
2877
2878 static void
2879 bridge_handoff_notags(struct ifnet *dst_ifp, struct mbuf *m)
2880 {
2881         struct mbuf *m0;
2882
2883         KKASSERT(m->m_type != MT_TAG);
2884
2885         lwkt_serialize_enter(dst_ifp->if_serializer);
2886
2887         /* We may be sending a fragment so traverse the mbuf */
2888         for (; m; m = m0) {
2889                 struct altq_pktattr pktattr;
2890
2891                 m0 = m->m_nextpkt;
2892                 m->m_nextpkt = NULL;
2893
2894                 if (ifq_is_enabled(&dst_ifp->if_snd))
2895                         altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
2896
2897                 ifq_handoff(dst_ifp, m, &pktattr);
2898         }
2899
2900         lwkt_serialize_exit(dst_ifp->if_serializer);
2901 }