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  */
  70
  71 /*
  72  * Network interface bridge support.
  73  *
  74  * TODO:
  75  *
  76  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  77  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  78  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
  79  *        consider heterogenous bridges).
  80  *
  81  *
  82  * Bridge's route information is duplicated to each CPUs:
  83  *
  84  *      CPU0          CPU1          CPU2          CPU3
  85  * +-----------+ +-----------+ +-----------+ +-----------+
  86  * |  rtnode   | |  rtnode   | |  rtnode   | |  rtnode   |
  87  * |           | |           | |           | |           |
  88  * | dst eaddr | | dst eaddr | | dst eaddr | | dst eaddr |
  89  * +-----------+ +-----------+ +-----------+ +-----------+
  90  *       |         |                     |         |
  91  *       |         |                     |         |
  92  *       |         |     +----------+    |         |
  93  *       |         |     |  rtinfo  |    |         |
  94  *       |         +---->|          |<---+         |
  95  *       |               |  flags   |              |
  96  *       +-------------->|  timeout |<-------------+
  97  *                       |  dst_ifp |
  98  *                       +----------+
  99  *
 100  * We choose to put timeout and dst_ifp into shared part, so updating
 101  * them will be cheaper than using message forwarding.  Also there is
 102  * not need to use spinlock to protect the updating: timeout and dst_ifp
 103  * is not related and specific field's updating order has no importance.
 104  * The cache pollution by the share part should not be heavy: in a stable
 105  * setup, dst_ifp probably will be not changed in rtnode's life time,
 106  * while timeout is refreshed once per second; most of the time, timeout
 107  * and dst_ifp are read-only accessed.
 108  *
 109  *
 110  * Bridge route information installation on bridge_input path:
 111  *
 112  *      CPU0           CPU1         CPU2          CPU3
 113  *
 114  *                               tcp_thread2
 115  *                                    |
 116  *                                alloc nmsg
 117  *                    snd nmsg        |
 118  *                    w/o rtinfo      |
 119  *      ifnet0<-----------------------+
 120  *        |                           :
 121  *    lookup dst                      :
 122  *   rtnode exists?(Y)free nmsg       :
 123  *        |(N)                        :
 124  *        |
 125  *  alloc rtinfo
 126  *  alloc rtnode
 127  * install rtnode
 128  *        |
 129  *        +---------->ifnet1
 130  *        : fwd nmsg    |
 131  *        : w/ rtinfo   |
 132  *        :             |
 133  *        :             |
 134  *                 alloc rtnode
 135  *               (w/ nmsg's rtinfo)
 136  *                install rtnode
 137  *                      |
 138  *                      +---------->ifnet2
 139  *                      : fwd nmsg    |
 140  *                      : w/ rtinfo   |
 141  *                      :             |
 142  *                      :         same as ifnet1
 143  *                                    |
 144  *                                    +---------->ifnet3
 145  *                                    : fwd nmsg    |
 146  *                                    : w/ rtinfo   |
 147  *                                    :             |
 148  *                                    :         same as ifnet1
 149  *                                               free nmsg
 150  *                                                  :
 151  *                                                  :
 152  *
 153  * The netmsgs forwarded between protocol threads and ifnet threads are
 154  * allocated with (M_WAITOK|M_NULLOK), so it will not fail under most
 155  * cases (route information is too precious to be not installed :).
 156  * Since multiple threads may try to install route information for the
 157  * same dst eaddr, we look up route information in ifnet0.  However, this
 158  * looking up only need to be performed on ifnet0, which is the start
 159  * point of the route information installation process.
 160  *
 161  *
 162  * Bridge route information deleting/flushing:
 163  *
 164  *  CPU0            CPU1             CPU2             CPU3
 165  *
 166  * netisr0
 167  *   |
 168  * find suitable rtnodes,
 169  * mark their rtinfo dead
 170  *   |
 171  *   | domsg <------------------------------------------+
 172  *   |                                                  | replymsg
 173  *   |                                                  |
 174  *   V     fwdmsg           fwdmsg           fwdmsg     |
 175  * ifnet0 --------> ifnet1 --------> ifnet2 --------> ifnet3
 176  * delete rtnodes   delete rtnodes   delete rtnodes   delete rtnodes
 177  * w/ dead rtinfo   w/ dead rtinfo   w/ dead rtinfo   w/ dead rtinfo
 178  *                                                    free dead rtinfos
 179  *
 180  * All deleting/flushing operations are serialized by netisr0, so each
 181  * operation only reaps the route information marked dead by itself.
 182  *
 183  *
 184  * Bridge route information adding/deleting/flushing:
 185  * Since all operation is serialized by the fixed message flow between
 186  * ifnet threads, it is not possible to create corrupted per-cpu route
 187  * information.
 188  *
 189  *
 190  *
 191  * Percpu member interface list iteration with blocking operation:
 192  * Since one bridge could only delete one member interface at a time and
 193  * the deleted member interface is not freed after netmsg_service_sync(),
 194  * following way is used to make sure that even if the certain member
 195  * interface is ripped from the percpu list during the blocking operation,
 196  * the iteration still could keep going:
 197  *
 198  * TAILQ_FOREACH_MUTABLE(bif, sc->sc_iflists[mycpuid], bif_next, nbif) {
 199  *     blocking operation;
 200  *     blocking operation;
 201  *     ...
 202  *     ...
 203  *     if (nbif != NULL && !nbif->bif_onlist) {
 204  *         KKASSERT(bif->bif_onlist);
 205  *         nbif = TAILQ_NEXT(bif, bif_next);
 206  *     }
 207  * }
 208  *
 209  * As mentioned above only one member interface could be unlinked from the
 210  * percpu member interface list, so either bif or nbif may be not on the list,
 211  * but _not_ both.  To keep the list iteration, we don't care about bif, but
 212  * only nbif.  Since removed member interface will only be freed after we
 213  * finish our work, it is safe to access any field in an unlinked bif (here
 214  * bif_onlist).  If nbif is no longer on the list, then bif must be on the
 215  * list, so we change nbif to the next element of bif and keep going.
 216  */
 217
 218 #include "opt_inet.h"
 219 #include "opt_inet6.h"
 220
 221 #include <sys/param.h>
 222 #include <sys/mbuf.h>
 223 #include <sys/malloc.h>
 224 #include <sys/protosw.h>
 225 #include <sys/systm.h>
 226 #include <sys/time.h>
 227 #include <sys/socket.h> /* for net/if.h */
 228 #include <sys/sockio.h>
 229 #include <sys/ctype.h>  /* string functions */
 230 #include <sys/kernel.h>
 231 #include <sys/random.h>
 232 #include <sys/sysctl.h>
 233 #include <sys/module.h>
 234 #include <sys/proc.h>
 235 #include <sys/priv.h>
 236 #include <sys/lock.h>
 237 #include <sys/thread.h>
 238 #include <sys/thread2.h>
 239 #include <sys/mpipe.h>
 240
 241 #include <net/bpf.h>
 242 #include <net/if.h>
 243 #include <net/if_dl.h>
 244 #include <net/if_types.h>
 245 #include <net/if_var.h>
 246 #include <net/pfil.h>
 247 #include <net/ifq_var.h>
 248 #include <net/if_clone.h>
 249
 250 #include <netinet/in.h> /* for struct arpcom */
 251 #include <netinet/in_systm.h>
 252 #include <netinet/in_var.h>
 253 #include <netinet/ip.h>
 254 #include <netinet/ip_var.h>
 255 #ifdef INET6
 256 #include <netinet/ip6.h>
 257 #include <netinet6/ip6_var.h>
 258 #endif
 259 #include <netinet/if_ether.h> /* for struct arpcom */
 260 #include <net/bridge/if_bridgevar.h>
 261 #include <net/if_llc.h>
 262 #include <net/netmsg2.h>
 263
 264 #include <net/route.h>
 265 #include <sys/in_cksum.h>
 266
 267 /*
 268  * Size of the route hash table.  Must be a power of two.
 269  */
 270 #ifndef BRIDGE_RTHASH_SIZE
 271 #define BRIDGE_RTHASH_SIZE              1024
 272 #endif
 273
 274 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 275
 276 /*
 277  * Maximum number of addresses to cache.
 278  */
 279 #ifndef BRIDGE_RTABLE_MAX
 280 #define BRIDGE_RTABLE_MAX               100
 281 #endif
 282
 283 /*
 284  * Spanning tree defaults.
 285  */
 286 #define BSTP_DEFAULT_MAX_AGE            (20 * 256)
 287 #define BSTP_DEFAULT_HELLO_TIME         (2 * 256)
 288 #define BSTP_DEFAULT_FORWARD_DELAY      (15 * 256)
 289 #define BSTP_DEFAULT_HOLD_TIME          (1 * 256)
 290 #define BSTP_DEFAULT_BRIDGE_PRIORITY    0x8000
 291 #define BSTP_DEFAULT_PORT_PRIORITY      0x80
 292 #define BSTP_DEFAULT_PATH_COST          55
 293
 294 /*
 295  * Timeout (in seconds) for entries learned dynamically.
 296  */
 297 #ifndef BRIDGE_RTABLE_TIMEOUT
 298 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 299 #endif
 300
 301 /*
 302  * Number of seconds between walks of the route list.
 303  */
 304 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 305 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 306 #endif
 307
 308 /*
 309  * List of capabilities to mask on the member interface.
 310  */
 311 #define BRIDGE_IFCAPS_MASK              IFCAP_TXCSUM
 312
 313 typedef int     (*bridge_ctl_t)(struct bridge_softc *, void *);
 314
 315 struct netmsg_brctl {
 316         struct netmsg_base      base;
 317         bridge_ctl_t            bc_func;
 318         struct bridge_softc     *bc_sc;
 319         void                    *bc_arg;
 320 };
 321
 322 struct netmsg_brsaddr {
 323         struct netmsg_base      base;
 324         struct bridge_softc     *br_softc;
 325         struct ifnet            *br_dst_if;
 326         struct bridge_rtinfo    *br_rtinfo;
 327         int                     br_setflags;
 328         uint8_t                 br_dst[ETHER_ADDR_LEN];
 329         uint8_t                 br_flags;
 330 };
 331
 332 struct netmsg_braddbif {
 333         struct netmsg_base      base;
 334         struct bridge_softc     *br_softc;
 335         struct bridge_ifinfo    *br_bif_info;
 336         struct ifnet            *br_bif_ifp;
 337 };
 338
 339 struct netmsg_brdelbif {
 340         struct netmsg_base      base;
 341         struct bridge_softc     *br_softc;
 342         struct bridge_ifinfo    *br_bif_info;
 343         struct bridge_iflist_head *br_bif_list;
 344 };
 345
 346 struct netmsg_brsflags {
 347         struct netmsg_base      base;
 348         struct bridge_softc     *br_softc;
 349         struct bridge_ifinfo    *br_bif_info;
 350         uint32_t                br_bif_flags;
 351 };
 352
 353 eventhandler_tag        bridge_detach_cookie = NULL;
 354
 355 extern  struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
 356 extern  int (*bridge_output_p)(struct ifnet *, struct mbuf *);
 357 extern  void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
 358 extern  struct ifnet *(*bridge_interface_p)(void *if_bridge);
 359
 360 static int      bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 361
 362 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 363 static int      bridge_clone_destroy(struct ifnet *);
 364
 365 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
 366 static void     bridge_mutecaps(struct bridge_ifinfo *, struct ifnet *, int);
 367 static void     bridge_ifdetach(void *, struct ifnet *);
 368 static void     bridge_init(void *);
 369 static int      bridge_from_us(struct bridge_softc *, struct ether_header *);
 370 static void     bridge_stop(struct ifnet *);
 371 static void     bridge_start(struct ifnet *, struct ifaltq_subque *);
 372 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 373 static int      bridge_output(struct ifnet *, struct mbuf *);
 374 static struct ifnet *bridge_interface(void *if_bridge);
 375
 376 static void     bridge_forward(struct bridge_softc *, struct mbuf *m);
 377
 378 static void     bridge_timer_handler(netmsg_t);
 379 static void     bridge_timer(void *);
 380
 381 static void     bridge_start_bcast(struct bridge_softc *, struct mbuf *);
 382 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 383                     struct mbuf *);
 384 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 385
 386 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 387                     struct ifnet *, uint8_t);
 388 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
 389 static void     bridge_rtreap(struct bridge_softc *);
 390 static void     bridge_rtreap_async(struct bridge_softc *);
 391 static void     bridge_rttrim(struct bridge_softc *);
 392 static int      bridge_rtage_finddead(struct bridge_softc *);
 393 static void     bridge_rtage(struct bridge_softc *);
 394 static void     bridge_rtflush(struct bridge_softc *, int);
 395 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
 396 static int      bridge_rtsaddr(struct bridge_softc *, const uint8_t *,
 397                     struct ifnet *, uint8_t);
 398 static void     bridge_rtmsg_sync(struct bridge_softc *sc);
 399 static void     bridge_rtreap_handler(netmsg_t);
 400 static void     bridge_rtinstall_handler(netmsg_t);
 401 static int      bridge_rtinstall_oncpu(struct bridge_softc *, const uint8_t *,
 402                     struct ifnet *, int, uint8_t, struct bridge_rtinfo **);
 403
 404 static void     bridge_rtable_init(struct bridge_softc *);
 405 static void     bridge_rtable_fini(struct bridge_softc *);
 406
 407 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 408 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 409                     const uint8_t *);
 410 static void     bridge_rtnode_insert(struct bridge_softc *,
 411                     struct bridge_rtnode *);
 412 static void     bridge_rtnode_destroy(struct bridge_softc *,
 413                     struct bridge_rtnode *);
 414
 415 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 416                     const char *name);
 417 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 418                     struct ifnet *ifp);
 419 static struct bridge_iflist *bridge_lookup_member_ifinfo(struct bridge_softc *,
 420                     struct bridge_ifinfo *);
 421 static void     bridge_delete_member(struct bridge_softc *,
 422                     struct bridge_iflist *, int);
 423 static void     bridge_delete_span(struct bridge_softc *,
 424                     struct bridge_iflist *);
 425
 426 static int      bridge_control(struct bridge_softc *, u_long,
 427                                bridge_ctl_t, void *);
 428 static int      bridge_ioctl_init(struct bridge_softc *, void *);
 429 static int      bridge_ioctl_stop(struct bridge_softc *, void *);
 430 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 431 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 432 static void     bridge_ioctl_fillflags(struct bridge_softc *sc,
 433                                 struct bridge_iflist *bif, struct ifbreq *req);
 434 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 435 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 436 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 437 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 438 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 439 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 440 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 441 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 442 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 443 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 444 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 445 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 446 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 447 static int      bridge_ioctl_reinit(struct bridge_softc *, void *);
 448 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 449 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 450 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 451 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 452 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 453 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 454 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 455 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 456 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 457 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 458 static int      bridge_ioctl_sifbondwght(struct bridge_softc *, void *);
 459 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 460                     int);
 461 static int      bridge_ip_checkbasic(struct mbuf **mp);
 462 #ifdef INET6
 463 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 464 #endif /* INET6 */
 465 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 466                     struct ether_header *, int, struct llc *);
 467 static void     bridge_enqueue_handler(netmsg_t);
 468 static void     bridge_handoff(struct bridge_softc *, struct ifnet *,
 469                     struct mbuf *, int);
 470
 471 static void     bridge_del_bif_handler(netmsg_t);
 472 static void     bridge_add_bif_handler(netmsg_t);
 473 static void     bridge_del_bif(struct bridge_softc *, struct bridge_ifinfo *,
 474                     struct bridge_iflist_head *);
 475 static void     bridge_add_bif(struct bridge_softc *, struct bridge_ifinfo *,
 476                     struct ifnet *);
 477
 478 SYSCTL_DECL(_net_link);
 479 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 480
 481 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 482 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 483 static int pfil_member = 1; /* run pfil hooks on the member interface */
 484 static int bridge_debug;
 485 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 486     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 487 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 488     &pfil_bridge, 0, "Packet filter on the bridge interface");
 489 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 490     &pfil_member, 0, "Packet filter on the member interface");
 491 SYSCTL_INT(_net_link_bridge, OID_AUTO, debug, CTLFLAG_RW,
 492     &bridge_debug, 0, "Bridge debug mode");
 493
 494 struct bridge_control_arg {
 495         union {
 496                 struct ifbreq ifbreq;
 497                 struct ifbifconf ifbifconf;
 498                 struct ifbareq ifbareq;
 499                 struct ifbaconf ifbaconf;
 500                 struct ifbrparam ifbrparam;
 501         } bca_u;
 502         int     bca_len;
 503         void    *bca_uptr;
 504         void    *bca_kptr;
 505 };
 506
 507 struct bridge_control {
 508         bridge_ctl_t    bc_func;
 509         int             bc_argsize;
 510         int             bc_flags;
 511 };
 512
 513 #define BC_F_COPYIN             0x01    /* copy arguments in */
 514 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 515 #define BC_F_SUSER              0x04    /* do super-user check */
 516
 517 const struct bridge_control bridge_control_table[] = {
 518         { bridge_ioctl_add,             sizeof(struct ifbreq),
 519           BC_F_COPYIN|BC_F_SUSER },
 520         { bridge_ioctl_del,             sizeof(struct ifbreq),
 521           BC_F_COPYIN|BC_F_SUSER },
 522
 523         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 524           BC_F_COPYIN|BC_F_COPYOUT },
 525         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 526           BC_F_COPYIN|BC_F_SUSER },
 527
 528         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 529           BC_F_COPYIN|BC_F_SUSER },
 530         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 531           BC_F_COPYOUT },
 532
 533         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 534           BC_F_COPYIN|BC_F_COPYOUT },
 535         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 536           BC_F_COPYIN|BC_F_COPYOUT },
 537
 538         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 539           BC_F_COPYIN|BC_F_SUSER },
 540
 541         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 542           BC_F_COPYIN|BC_F_SUSER },
 543         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 544           BC_F_COPYOUT },
 545
 546         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 547           BC_F_COPYIN|BC_F_SUSER },
 548
 549         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 550           BC_F_COPYIN|BC_F_SUSER },
 551
 552         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 553           BC_F_COPYOUT },
 554         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 555           BC_F_COPYIN|BC_F_SUSER },
 556
 557         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 558           BC_F_COPYOUT },
 559         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 560           BC_F_COPYIN|BC_F_SUSER },
 561
 562         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 563           BC_F_COPYOUT },
 564         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 565           BC_F_COPYIN|BC_F_SUSER },
 566
 567         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 568           BC_F_COPYOUT },
 569         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 570           BC_F_COPYIN|BC_F_SUSER },
 571
 572         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 573           BC_F_COPYIN|BC_F_SUSER },
 574
 575         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 576           BC_F_COPYIN|BC_F_SUSER },
 577
 578         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 579           BC_F_COPYIN|BC_F_SUSER },
 580         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 581           BC_F_COPYIN|BC_F_SUSER },
 582
 583         { bridge_ioctl_sifbondwght,     sizeof(struct ifbreq),
 584           BC_F_COPYIN|BC_F_SUSER },
 585
 586 };
 587 static const int bridge_control_table_size = NELEM(bridge_control_table);
 588
 589 LIST_HEAD(, bridge_softc) bridge_list;
 590
 591 struct if_clone bridge_cloner = IF_CLONE_INITIALIZER("bridge",
 592                                 bridge_clone_create,
 593                                 bridge_clone_destroy, 0, IF_MAXUNIT);
 594
 595 static int
 596 bridge_modevent(module_t mod, int type, void *data)
 597 {
 598         switch (type) {
 599         case MOD_LOAD:
 600                 LIST_INIT(&bridge_list);
 601                 if_clone_attach(&bridge_cloner);
 602                 bridge_input_p = bridge_input;
 603                 bridge_output_p = bridge_output;
 604                 bridge_interface_p = bridge_interface;
 605                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 606                     ifnet_detach_event, bridge_ifdetach, NULL,
 607                     EVENTHANDLER_PRI_ANY);
 608 #if 0 /* notyet */
 609                 bstp_linkstate_p = bstp_linkstate;
 610 #endif
 611                 break;
 612         case MOD_UNLOAD:
 613                 if (!LIST_EMPTY(&bridge_list))
 614                         return (EBUSY);
 615                 EVENTHANDLER_DEREGISTER(ifnet_detach_event,
 616                     bridge_detach_cookie);
 617                 if_clone_detach(&bridge_cloner);
 618                 bridge_input_p = NULL;
 619                 bridge_output_p = NULL;
 620                 bridge_interface_p = NULL;
 621 #if 0 /* notyet */
 622                 bstp_linkstate_p = NULL;
 623 #endif
 624                 break;
 625         default:
 626                 return (EOPNOTSUPP);
 627         }
 628         return (0);
 629 }
 630
 631 static moduledata_t bridge_mod = {
 632         "if_bridge",
 633         bridge_modevent,
 634         0
 635 };
 636
 637 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 638
 639
 640 /*
 641  * bridge_clone_create:
 642  *
 643  *      Create a new bridge instance.
 644  */
 645 static int
 646 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused)
 647 {
 648         struct bridge_softc *sc;
 649         struct ifnet *ifp;
 650         u_char eaddr[6];
 651         int cpu, rnd;
 652
 653         sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
 654         ifp = sc->sc_ifp = &sc->sc_if;
 655
 656         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 657         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 658         sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
 659         sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
 660         sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
 661         sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
 662         sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
 663
 664         /* Initialize our routing table. */
 665         bridge_rtable_init(sc);
 666
 667         callout_init(&sc->sc_brcallout);
 668         netmsg_init(&sc->sc_brtimemsg, NULL, &netisr_adone_rport,
 669                     MSGF_DROPABLE, bridge_timer_handler);
 670         sc->sc_brtimemsg.lmsg.u.ms_resultp = sc;
 671
 672         callout_init(&sc->sc_bstpcallout);
 673         netmsg_init(&sc->sc_bstptimemsg, NULL, &netisr_adone_rport,
 674                     MSGF_DROPABLE, bstp_tick_handler);
 675         sc->sc_bstptimemsg.lmsg.u.ms_resultp = sc;
 676
 677         /* Initialize per-cpu member iface lists */
 678         sc->sc_iflists = kmalloc(sizeof(*sc->sc_iflists) * ncpus,
 679                                  M_DEVBUF, M_WAITOK);
 680         for (cpu = 0; cpu < ncpus; ++cpu)
 681                 TAILQ_INIT(&sc->sc_iflists[cpu]);
 682
 683         TAILQ_INIT(&sc->sc_spanlist);
 684
 685         ifp->if_softc = sc;
 686         if_initname(ifp, ifc->ifc_name, unit);
 687         ifp->if_mtu = ETHERMTU;
 688         ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
 689         ifp->if_ioctl = bridge_ioctl;
 690         ifp->if_start = bridge_start;
 691         ifp->if_init = bridge_init;
 692         ifp->if_type = IFT_ETHER;
 693         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
 694         ifq_set_ready(&ifp->if_snd);
 695         ifp->if_hdrlen = ETHER_HDR_LEN;
 696
 697         /*
 698          * Generate a random ethernet address and use the private AC:DE:48
 699          * OUI code.
 700          */
 701         rnd = karc4random();
 702         bcopy(&rnd, &eaddr[0], 4); /* ETHER_ADDR_LEN == 6 */
 703         rnd = karc4random();
 704         bcopy(&rnd, &eaddr[2], 4); /* ETHER_ADDR_LEN == 6 */
 705
 706         eaddr[0] &= ~1; /* clear multicast bit */
 707         eaddr[0] |= 2;  /* set the LAA bit */
 708
 709         ether_ifattach(ifp, eaddr, NULL);
 710         /* Now undo some of the damage... */
 711         ifp->if_baudrate = 0;
 712         /*ifp->if_type = IFT_BRIDGE;*/
 713
 714         crit_enter();   /* XXX MP */
 715         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 716         crit_exit();
 717
 718         return (0);
 719 }
 720
 721 static void
 722 bridge_delete_dispatch(netmsg_t msg)
 723 {
 724         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
 725         struct ifnet *bifp = sc->sc_ifp;
 726         struct bridge_iflist *bif;
 727
 728         ifnet_serialize_all(bifp);
 729
 730         while ((bif = TAILQ_FIRST(&sc->sc_iflists[mycpuid])) != NULL)
 731                 bridge_delete_member(sc, bif, 0);
 732
 733         while ((bif = TAILQ_FIRST(&sc->sc_spanlist)) != NULL)
 734                 bridge_delete_span(sc, bif);
 735
 736         ifnet_deserialize_all(bifp);
 737
 738         lwkt_replymsg(&msg->lmsg, 0);
 739 }
 740
 741 /*
 742  * bridge_clone_destroy:
 743  *
 744  *      Destroy a bridge instance.
 745  */
 746 static int
 747 bridge_clone_destroy(struct ifnet *ifp)
 748 {
 749         struct bridge_softc *sc = ifp->if_softc;
 750         struct netmsg_base msg;
 751
 752         ifnet_serialize_all(ifp);
 753
 754         bridge_stop(ifp);
 755         ifp->if_flags &= ~IFF_UP;
 756
 757         ifnet_deserialize_all(ifp);
 758
 759         netmsg_init(&msg, NULL, &curthread->td_msgport,
 760                     0, bridge_delete_dispatch);
 761         msg.lmsg.u.ms_resultp = sc;
 762         lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0);
 763
 764         crit_enter();   /* XXX MP */
 765         LIST_REMOVE(sc, sc_list);
 766         crit_exit();
 767
 768         ether_ifdetach(ifp);
 769
 770         /* Tear down the routing table. */
 771         bridge_rtable_fini(sc);
 772
 773         /* Free per-cpu member iface lists */
 774         kfree(sc->sc_iflists, M_DEVBUF);
 775
 776         kfree(sc, M_DEVBUF);
 777
 778         return 0;
 779 }
 780
 781 /*
 782  * bridge_ioctl:
 783  *
 784  *      Handle a control request from the operator.
 785  */
 786 static int
 787 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 788 {
 789         struct bridge_softc *sc = ifp->if_softc;
 790         struct bridge_control_arg args;
 791         struct ifdrv *ifd = (struct ifdrv *) data;
 792         const struct bridge_control *bc;
 793         int error = 0;
 794
 795         ASSERT_IFNET_SERIALIZED_ALL(ifp);
 796
 797         switch (cmd) {
 798         case SIOCADDMULTI:
 799         case SIOCDELMULTI:
 800                 break;
 801
 802         case SIOCGDRVSPEC:
 803         case SIOCSDRVSPEC:
 804                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 805                         error = EINVAL;
 806                         break;
 807                 }
 808                 bc = &bridge_control_table[ifd->ifd_cmd];
 809
 810                 if (cmd == SIOCGDRVSPEC &&
 811                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 812                         error = EINVAL;
 813                         break;
 814                 } else if (cmd == SIOCSDRVSPEC &&
 815                            (bc->bc_flags & BC_F_COPYOUT)) {
 816                         error = EINVAL;
 817                         break;
 818                 }
 819
 820                 if (bc->bc_flags & BC_F_SUSER) {
 821                         error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY);
 822                         if (error)
 823                                 break;
 824                 }
 825
 826                 if (ifd->ifd_len != bc->bc_argsize ||
 827                     ifd->ifd_len > sizeof(args.bca_u)) {
 828                         error = EINVAL;
 829                         break;
 830                 }
 831
 832                 memset(&args, 0, sizeof(args));
 833                 if (bc->bc_flags & BC_F_COPYIN) {
 834                         error = copyin(ifd->ifd_data, &args.bca_u,
 835                                        ifd->ifd_len);
 836                         if (error)
 837                                 break;
 838                 }
 839
 840                 error = bridge_control(sc, cmd, bc->bc_func, &args);
 841                 if (error) {
 842                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 843                         break;
 844                 }
 845
 846                 if (bc->bc_flags & BC_F_COPYOUT) {
 847                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 848                         if (args.bca_len != 0) {
 849                                 KKASSERT(args.bca_kptr != NULL);
 850                                 if (!error) {
 851                                         error = copyout(args.bca_kptr,
 852                                                 args.bca_uptr, args.bca_len);
 853                                 }
 854                                 kfree(args.bca_kptr, M_TEMP);
 855                         } else {
 856                                 KKASSERT(args.bca_kptr == NULL);
 857                         }
 858                 } else {
 859                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 860                 }
 861                 break;
 862
 863         case SIOCSIFFLAGS:
 864                 if (!(ifp->if_flags & IFF_UP) &&
 865                     (ifp->if_flags & IFF_RUNNING)) {
 866                         /*
 867                          * If interface is marked down and it is running,
 868                          * then stop it.
 869                          */
 870                         bridge_stop(ifp);
 871                 } else if ((ifp->if_flags & IFF_UP) &&
 872                     !(ifp->if_flags & IFF_RUNNING)) {
 873                         /*
 874                          * If interface is marked up and it is stopped, then
 875                          * start it.
 876                          */
 877                         ifp->if_init(sc);
 878                 }
 879
 880                 /*
 881                  * If running and link flag state change we have to
 882                  * reinitialize as well.
 883                  */
 884                 if ((ifp->if_flags & IFF_RUNNING) &&
 885                     (ifp->if_flags & (IFF_LINK0|IFF_LINK1|IFF_LINK2)) !=
 886                     sc->sc_copy_flags) {
 887                         sc->sc_copy_flags = ifp->if_flags &
 888                                         (IFF_LINK0|IFF_LINK1|IFF_LINK2);
 889                         bridge_control(sc, 0, bridge_ioctl_reinit, NULL);
 890                 }
 891
 892                 break;
 893
 894         case SIOCSIFMTU:
 895                 /* Do not allow the MTU to be changed on the bridge */
 896                 error = EINVAL;
 897                 break;
 898
 899         default:
 900                 error = ether_ioctl(ifp, cmd, data);
 901                 break;
 902         }
 903         return (error);
 904 }
 905
 906 /*
 907  * bridge_mutecaps:
 908  *
 909  *      Clear or restore unwanted capabilities on the member interface
 910  */
 911 static void
 912 bridge_mutecaps(struct bridge_ifinfo *bif_info, struct ifnet *ifp, int mute)
 913 {
 914         struct ifreq ifr;
 915         int error;
 916
 917         if (ifp->if_ioctl == NULL)
 918                 return;
 919
 920         bzero(&ifr, sizeof(ifr));
 921         ifr.ifr_reqcap = ifp->if_capenable;
 922
 923         if (mute) {
 924                 /* mask off and save capabilities */
 925                 bif_info->bifi_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
 926                 if (bif_info->bifi_mutecap != 0)
 927                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
 928         } else {
 929                 /* restore muted capabilities */
 930                 ifr.ifr_reqcap |= bif_info->bifi_mutecap;
 931         }
 932
 933         if (bif_info->bifi_mutecap != 0) {
 934                 ifnet_serialize_all(ifp);
 935                 error = ifp->if_ioctl(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL);
 936                 ifnet_deserialize_all(ifp);
 937         }
 938 }
 939
 940 /*
 941  * bridge_lookup_member:
 942  *
 943  *      Lookup a bridge member interface.
 944  */
 945 static struct bridge_iflist *
 946 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 947 {
 948         struct bridge_iflist *bif;
 949
 950         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 951                 if (strcmp(bif->bif_ifp->if_xname, name) == 0)
 952                         return (bif);
 953         }
 954         return (NULL);
 955 }
 956
 957 /*
 958  * bridge_lookup_member_if:
 959  *
 960  *      Lookup a bridge member interface by ifnet*.
 961  */
 962 static struct bridge_iflist *
 963 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 964 {
 965         struct bridge_iflist *bif;
 966
 967         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 968                 if (bif->bif_ifp == member_ifp)
 969                         return (bif);
 970         }
 971         return (NULL);
 972 }
 973
 974 /*
 975  * bridge_lookup_member_ifinfo:
 976  *
 977  *      Lookup a bridge member interface by bridge_ifinfo.
 978  */
 979 static struct bridge_iflist *
 980 bridge_lookup_member_ifinfo(struct bridge_softc *sc,
 981                             struct bridge_ifinfo *bif_info)
 982 {
 983         struct bridge_iflist *bif;
 984
 985         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 986                 if (bif->bif_info == bif_info)
 987                         return (bif);
 988         }
 989         return (NULL);
 990 }
 991
 992 /*
 993  * bridge_delete_member:
 994  *
 995  *      Delete the specified member interface.
 996  */
 997 static void
 998 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 999     int gone)
1000 {
1001         struct ifnet *ifs = bif->bif_ifp;
1002         struct ifnet *bifp = sc->sc_ifp;
1003         struct bridge_ifinfo *bif_info = bif->bif_info;
1004         struct bridge_iflist_head saved_bifs;
1005
1006         ASSERT_IFNET_SERIALIZED_ALL(bifp);
1007         KKASSERT(bif_info != NULL);
1008
1009         ifs->if_bridge = NULL;
1010
1011         /*
1012          * Release bridge interface's serializer:
1013          * - To avoid possible dead lock.
1014          * - Various sync operation will block the current thread.
1015          */
1016         ifnet_deserialize_all(bifp);
1017
1018         if (!gone) {
1019                 switch (ifs->if_type) {
1020                 case IFT_ETHER:
1021                 case IFT_L2VLAN:
1022                         /*
1023                          * Take the interface out of promiscuous mode.
1024                          */
1025                         ifpromisc(ifs, 0);
1026                         bridge_mutecaps(bif_info, ifs, 0);
1027                         break;
1028
1029                 case IFT_GIF:
1030                         break;
1031
1032                 default:
1033                         panic("bridge_delete_member: impossible");
1034                         break;
1035                 }
1036         }
1037
1038         /*
1039          * Remove bifs from percpu linked list.
1040          *
1041          * Removed bifs are not freed immediately, instead,
1042          * they are saved in saved_bifs.  They will be freed
1043          * after we make sure that no one is accessing them,
1044          * i.e. after following netmsg_service_sync()
1045          */
1046         TAILQ_INIT(&saved_bifs);
1047         bridge_del_bif(sc, bif_info, &saved_bifs);
1048
1049         /*
1050          * Make sure that all protocol threads:
1051          * o  see 'ifs' if_bridge is changed
1052          * o  know that bif is removed from the percpu linked list
1053          */
1054         netmsg_service_sync();
1055
1056         /*
1057          * Free the removed bifs
1058          */
1059         KKASSERT(!TAILQ_EMPTY(&saved_bifs));
1060         while ((bif = TAILQ_FIRST(&saved_bifs)) != NULL) {
1061                 TAILQ_REMOVE(&saved_bifs, bif, bif_next);
1062                 kfree(bif, M_DEVBUF);
1063         }
1064
1065         /* See the comment in bridge_ioctl_stop() */
1066         bridge_rtmsg_sync(sc);
1067         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL | IFBF_FLUSHSYNC);
1068
1069         ifnet_serialize_all(bifp);
1070
1071         if (bifp->if_flags & IFF_RUNNING)
1072                 bstp_initialization(sc);
1073
1074         /*
1075          * Free the bif_info after bstp_initialization(), so that
1076          * bridge_softc.sc_root_port will not reference a dangling
1077          * pointer.
1078          */
1079         kfree(bif_info, M_DEVBUF);
1080 }
1081
1082 /*
1083  * bridge_delete_span:
1084  *
1085  *      Delete the specified span interface.
1086  */
1087 static void
1088 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1089 {
1090         KASSERT(bif->bif_ifp->if_bridge == NULL,
1091             ("%s: not a span interface", __func__));
1092
1093         TAILQ_REMOVE(&sc->sc_iflists[mycpuid], bif, bif_next);
1094         kfree(bif, M_DEVBUF);
1095 }
1096
1097 static int
1098 bridge_ioctl_init(struct bridge_softc *sc, void *arg __unused)
1099 {
1100         struct ifnet *ifp = sc->sc_ifp;
1101
1102         if (ifp->if_flags & IFF_RUNNING)
1103                 return 0;
1104
1105         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1106             bridge_timer, sc);
1107
1108         ifp->if_flags |= IFF_RUNNING;
1109         bstp_initialization(sc);
1110         return 0;
1111 }
1112
1113 static int
1114 bridge_ioctl_stop(struct bridge_softc *sc, void *arg __unused)
1115 {
1116         struct ifnet *ifp = sc->sc_ifp;
1117         struct lwkt_msg *lmsg;
1118
1119         if ((ifp->if_flags & IFF_RUNNING) == 0)
1120                 return 0;
1121
1122         callout_stop(&sc->sc_brcallout);
1123
1124         crit_enter();
1125         lmsg = &sc->sc_brtimemsg.lmsg;
1126         if ((lmsg->ms_flags & MSGF_DONE) == 0) {
1127                 /* Pending to be processed; drop it */
1128                 lwkt_dropmsg(lmsg);
1129         }
1130         crit_exit();
1131
1132         bstp_stop(sc);
1133
1134         ifp->if_flags &= ~IFF_RUNNING;
1135
1136         ifnet_deserialize_all(ifp);
1137
1138         /* Let everyone know that we are stopped */
1139         netmsg_service_sync();
1140
1141         /*
1142          * Sync ifnetX msgports in the order we forward rtnode
1143          * installation message.  This is used to make sure that
1144          * all rtnode installation messages sent by bridge_rtupdate()
1145          * during above netmsg_service_sync() are flushed.
1146          */
1147         bridge_rtmsg_sync(sc);
1148         bridge_rtflush(sc, IFBF_FLUSHDYN | IFBF_FLUSHSYNC);
1149
1150         ifnet_serialize_all(ifp);
1151         return 0;
1152 }
1153
1154 static int
1155 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1156 {
1157         struct ifbreq *req = arg;
1158         struct bridge_iflist *bif;
1159         struct bridge_ifinfo *bif_info;
1160         struct ifnet *ifs, *bifp;
1161         int error = 0;
1162
1163         bifp = sc->sc_ifp;
1164         ASSERT_IFNET_SERIALIZED_ALL(bifp);
1165
1166         ifs = ifunit(req->ifbr_ifsname);
1167         if (ifs == NULL)
1168                 return (ENOENT);
1169
1170         /* If it's in the span list, it can't be a member. */
1171         TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
1172                 if (ifs == bif->bif_ifp)
1173                         return (EBUSY);
1174
1175         /* Allow the first Ethernet member to define the MTU */
1176         if (ifs->if_type != IFT_GIF) {
1177                 if (TAILQ_EMPTY(&sc->sc_iflists[mycpuid])) {
1178                         bifp->if_mtu = ifs->if_mtu;
1179                 } else if (bifp->if_mtu != ifs->if_mtu) {
1180                         if_printf(bifp, "invalid MTU for %s\n", ifs->if_xname);
1181                         return (EINVAL);
1182                 }
1183         }
1184
1185         if (ifs->if_bridge == sc)
1186                 return (EEXIST);
1187
1188         if (ifs->if_bridge != NULL)
1189                 return (EBUSY);
1190
1191         bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO);
1192         bif_info->bifi_priority = BSTP_DEFAULT_PORT_PRIORITY;
1193         bif_info->bifi_path_cost = BSTP_DEFAULT_PATH_COST;
1194         bif_info->bifi_ifp = ifs;
1195         bif_info->bifi_bond_weight = 1;
1196
1197         /*
1198          * Release bridge interface's serializer:
1199          * - To avoid possible dead lock.
1200          * - Various sync operation will block the current thread.
1201          */
1202         ifnet_deserialize_all(bifp);
1203
1204         switch (ifs->if_type) {
1205         case IFT_ETHER:
1206         case IFT_L2VLAN:
1207                 /*
1208                  * Place the interface into promiscuous mode.
1209                  */
1210                 error = ifpromisc(ifs, 1);
1211                 if (error) {
1212                         ifnet_serialize_all(bifp);
1213                         goto out;
1214                 }
1215                 bridge_mutecaps(bif_info, ifs, 1);
1216                 break;
1217
1218         case IFT_GIF: /* :^) */
1219                 break;
1220
1221         default:
1222                 error = EINVAL;
1223                 ifnet_serialize_all(bifp);
1224                 goto out;
1225         }
1226
1227         /*
1228          * Add bifs to percpu linked lists
1229          */
1230         bridge_add_bif(sc, bif_info, ifs);
1231
1232         ifnet_serialize_all(bifp);
1233
1234         if (bifp->if_flags & IFF_RUNNING)
1235                 bstp_initialization(sc);
1236         else
1237                 bstp_stop(sc);
1238
1239         /*
1240          * Everything has been setup, so let the member interface
1241          * deliver packets to this bridge on its input/output path.
1242          */
1243         ifs->if_bridge = sc;
1244 out:
1245         if (error) {
1246                 if (bif_info != NULL)
1247                         kfree(bif_info, M_DEVBUF);
1248         }
1249         return (error);
1250 }
1251
1252 static int
1253 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1254 {
1255         struct ifbreq *req = arg;
1256         struct bridge_iflist *bif;
1257
1258         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1259         if (bif == NULL)
1260                 return (ENOENT);
1261
1262         bridge_delete_member(sc, bif, 0);
1263
1264         return (0);
1265 }
1266
1267 static int
1268 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1269 {
1270         struct ifbreq *req = arg;
1271         struct bridge_iflist *bif;
1272
1273         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1274         if (bif == NULL)
1275                 return (ENOENT);
1276         bridge_ioctl_fillflags(sc, bif, req);
1277         return (0);
1278 }
1279
1280 static void
1281 bridge_ioctl_fillflags(struct bridge_softc *sc, struct bridge_iflist *bif,
1282                        struct ifbreq *req)
1283 {
1284         req->ifbr_ifsflags = bif->bif_flags;
1285         req->ifbr_state = bif->bif_state;
1286         req->ifbr_priority = bif->bif_priority;
1287         req->ifbr_path_cost = bif->bif_path_cost;
1288         req->ifbr_bond_weight = bif->bif_bond_weight;
1289         req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1290         if (bif->bif_flags & IFBIF_STP) {
1291                 req->ifbr_peer_root = bif->bif_peer_root;
1292                 req->ifbr_peer_bridge = bif->bif_peer_bridge;
1293                 req->ifbr_peer_cost = bif->bif_peer_cost;
1294                 req->ifbr_peer_port = bif->bif_peer_port;
1295                 if (bstp_supersedes_port_info(sc, bif)) {
1296                         req->ifbr_designated_root = bif->bif_peer_root;
1297                         req->ifbr_designated_bridge = bif->bif_peer_bridge;
1298                         req->ifbr_designated_cost = bif->bif_peer_cost;
1299                         req->ifbr_designated_port = bif->bif_peer_port;
1300                 } else {
1301                         req->ifbr_designated_root = sc->sc_bridge_id;
1302                         req->ifbr_designated_bridge = sc->sc_bridge_id;
1303                         req->ifbr_designated_cost = bif->bif_path_cost +
1304                                                     bif->bif_peer_cost;
1305                         req->ifbr_designated_port = bif->bif_port_id;
1306                 }
1307         } else {
1308                 req->ifbr_peer_root = 0;
1309                 req->ifbr_peer_bridge = 0;
1310                 req->ifbr_peer_cost = 0;
1311                 req->ifbr_peer_port = 0;
1312                 req->ifbr_designated_root = 0;
1313                 req->ifbr_designated_bridge = 0;
1314                 req->ifbr_designated_cost = 0;
1315                 req->ifbr_designated_port = 0;
1316         }
1317 }
1318
1319 static int
1320 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1321 {
1322         struct ifbreq *req = arg;
1323         struct bridge_iflist *bif;
1324         struct ifnet *bifp = sc->sc_ifp;
1325
1326         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1327         if (bif == NULL)
1328                 return (ENOENT);
1329
1330         if (req->ifbr_ifsflags & IFBIF_SPAN) {
1331                 /* SPAN is readonly */
1332                 return (EINVAL);
1333         }
1334
1335         if (req->ifbr_ifsflags & IFBIF_STP) {
1336                 switch (bif->bif_ifp->if_type) {
1337                 case IFT_ETHER:
1338                         /* These can do spanning tree. */
1339                         break;
1340
1341                 default:
1342                         /* Nothing else can. */
1343                         return (EINVAL);
1344                 }
1345         }
1346
1347         bif->bif_flags = (bif->bif_flags & IFBIF_KEEPMASK) |
1348                          (req->ifbr_ifsflags & ~IFBIF_KEEPMASK);
1349         if (bifp->if_flags & IFF_RUNNING)
1350                 bstp_initialization(sc);
1351
1352         return (0);
1353 }
1354
1355 static int
1356 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1357 {
1358         struct ifbrparam *param = arg;
1359         struct ifnet *ifp = sc->sc_ifp;
1360
1361         sc->sc_brtmax = param->ifbrp_csize;
1362
1363         ifnet_deserialize_all(ifp);
1364         bridge_rttrim(sc);
1365         ifnet_serialize_all(ifp);
1366
1367         return (0);
1368 }
1369
1370 static int
1371 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1372 {
1373         struct ifbrparam *param = arg;
1374
1375         param->ifbrp_csize = sc->sc_brtmax;
1376
1377         return (0);
1378 }
1379
1380 static int
1381 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1382 {
1383         struct bridge_control_arg *bc_arg = arg;
1384         struct ifbifconf *bifc = arg;
1385         struct bridge_iflist *bif;
1386         struct ifbreq *breq;
1387         int count, len;
1388
1389         count = 0;
1390         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next)
1391                 count++;
1392         TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
1393                 count++;
1394
1395         if (bifc->ifbic_len == 0) {
1396                 bifc->ifbic_len = sizeof(*breq) * count;
1397                 return 0;
1398         } else if (count == 0 || bifc->ifbic_len < sizeof(*breq)) {
1399                 bifc->ifbic_len = 0;
1400                 return 0;
1401         }
1402
1403         len = min(bifc->ifbic_len, sizeof(*breq) * count);
1404         KKASSERT(len >= sizeof(*breq));
1405
1406         breq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1407         if (breq == NULL) {
1408                 bifc->ifbic_len = 0;
1409                 return ENOMEM;
1410         }
1411         bc_arg->bca_kptr = breq;
1412
1413         count = 0;
1414         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1415                 if (len < sizeof(*breq))
1416                         break;
1417
1418                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1419                         sizeof(breq->ifbr_ifsname));
1420                 bridge_ioctl_fillflags(sc, bif, breq);
1421                 breq++;
1422                 count++;
1423                 len -= sizeof(*breq);
1424         }
1425         TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1426                 if (len < sizeof(*breq))
1427                         break;
1428
1429                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1430                         sizeof(breq->ifbr_ifsname));
1431                 breq->ifbr_ifsflags = bif->bif_flags;
1432                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1433                 breq++;
1434                 count++;
1435                 len -= sizeof(*breq);
1436         }
1437
1438         bifc->ifbic_len = sizeof(*breq) * count;
1439         KKASSERT(bifc->ifbic_len > 0);
1440
1441         bc_arg->bca_len = bifc->ifbic_len;
1442         bc_arg->bca_uptr = bifc->ifbic_req;
1443         return 0;
1444 }
1445
1446 static int
1447 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1448 {
1449         struct bridge_control_arg *bc_arg = arg;
1450         struct ifbaconf *bac = arg;
1451         struct bridge_rtnode *brt;
1452         struct ifbareq *bareq;
1453         int count, len;
1454
1455         count = 0;
1456         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list)
1457                 count++;
1458
1459         if (bac->ifbac_len == 0) {
1460                 bac->ifbac_len = sizeof(*bareq) * count;
1461                 return 0;
1462         } else if (count == 0 || bac->ifbac_len < sizeof(*bareq)) {
1463                 bac->ifbac_len = 0;
1464                 return 0;
1465         }
1466
1467         len = min(bac->ifbac_len, sizeof(*bareq) * count);
1468         KKASSERT(len >= sizeof(*bareq));
1469
1470         bareq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1471         if (bareq == NULL) {
1472                 bac->ifbac_len = 0;
1473                 return ENOMEM;
1474         }
1475         bc_arg->bca_kptr = bareq;
1476
1477         count = 0;
1478         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
1479                 struct bridge_rtinfo *bri = brt->brt_info;
1480                 unsigned long expire;
1481
1482                 if (len < sizeof(*bareq))
1483                         break;
1484
1485                 strlcpy(bareq->ifba_ifsname, bri->bri_ifp->if_xname,
1486                         sizeof(bareq->ifba_ifsname));
1487                 memcpy(bareq->ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1488                 expire = bri->bri_expire;
1489                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1490                     time_second < expire)
1491                         bareq->ifba_expire = expire - time_second;
1492                 else
1493                         bareq->ifba_expire = 0;
1494                 bareq->ifba_flags = bri->bri_flags;
1495                 bareq++;
1496                 count++;
1497                 len -= sizeof(*bareq);
1498         }
1499
1500         bac->ifbac_len = sizeof(*bareq) * count;
1501         KKASSERT(bac->ifbac_len > 0);
1502
1503         bc_arg->bca_len = bac->ifbac_len;
1504         bc_arg->bca_uptr = bac->ifbac_req;
1505         return 0;
1506 }
1507
1508 static int
1509 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1510 {
1511         struct ifbareq *req = arg;
1512         struct bridge_iflist *bif;
1513         struct ifnet *ifp = sc->sc_ifp;
1514         int error;
1515
1516         ASSERT_IFNET_SERIALIZED_ALL(ifp);
1517
1518         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1519         if (bif == NULL)
1520                 return (ENOENT);
1521
1522         ifnet_deserialize_all(ifp);
1523         error = bridge_rtsaddr(sc, req->ifba_dst, bif->bif_ifp,
1524                                req->ifba_flags);
1525         ifnet_serialize_all(ifp);
1526         return (error);
1527 }
1528
1529 static int
1530 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1531 {
1532         struct ifbrparam *param = arg;
1533
1534         sc->sc_brttimeout = param->ifbrp_ctime;
1535
1536         return (0);
1537 }
1538
1539 static int
1540 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1541 {
1542         struct ifbrparam *param = arg;
1543
1544         param->ifbrp_ctime = sc->sc_brttimeout;
1545
1546         return (0);
1547 }
1548
1549 static int
1550 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1551 {
1552         struct ifbareq *req = arg;
1553         struct ifnet *ifp = sc->sc_ifp;
1554         int error;
1555
1556         ifnet_deserialize_all(ifp);
1557         error = bridge_rtdaddr(sc, req->ifba_dst);
1558         ifnet_serialize_all(ifp);
1559         return error;
1560 }
1561
1562 static int
1563 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1564 {
1565         struct ifbreq *req = arg;
1566         struct ifnet *ifp = sc->sc_ifp;
1567
1568         ifnet_deserialize_all(ifp);
1569         bridge_rtflush(sc, req->ifbr_ifsflags | IFBF_FLUSHSYNC);
1570         ifnet_serialize_all(ifp);
1571
1572         return (0);
1573 }
1574
1575 static int
1576 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1577 {
1578         struct ifbrparam *param = arg;
1579
1580         param->ifbrp_prio = sc->sc_bridge_priority;
1581
1582         return (0);
1583 }
1584
1585 static int
1586 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1587 {
1588         struct ifbrparam *param = arg;
1589
1590         sc->sc_bridge_priority = param->ifbrp_prio;
1591
1592         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1593                 bstp_initialization(sc);
1594
1595         return (0);
1596 }
1597
1598 static int
1599 bridge_ioctl_reinit(struct bridge_softc *sc, void *arg __unused)
1600 {
1601         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1602                 bstp_initialization(sc);
1603         return (0);
1604 }
1605
1606 static int
1607 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1608 {
1609         struct ifbrparam *param = arg;
1610
1611         param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1612
1613         return (0);
1614 }
1615
1616 static int
1617 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1618 {
1619         struct ifbrparam *param = arg;
1620
1621         if (param->ifbrp_hellotime == 0)
1622                 return (EINVAL);
1623         sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1624
1625         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1626                 bstp_initialization(sc);
1627
1628         return (0);
1629 }
1630
1631 static int
1632 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1633 {
1634         struct ifbrparam *param = arg;
1635
1636         param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1637
1638         return (0);
1639 }
1640
1641 static int
1642 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1643 {
1644         struct ifbrparam *param = arg;
1645
1646         if (param->ifbrp_fwddelay == 0)
1647                 return (EINVAL);
1648         sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1649
1650         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1651                 bstp_initialization(sc);
1652
1653         return (0);
1654 }
1655
1656 static int
1657 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1658 {
1659         struct ifbrparam *param = arg;
1660
1661         param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1662
1663         return (0);
1664 }
1665
1666 static int
1667 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1668 {
1669         struct ifbrparam *param = arg;
1670
1671         if (param->ifbrp_maxage == 0)
1672                 return (EINVAL);
1673         sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1674
1675         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1676                 bstp_initialization(sc);
1677
1678         return (0);
1679 }
1680
1681 static int
1682 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1683 {
1684         struct ifbreq *req = arg;
1685         struct bridge_iflist *bif;
1686
1687         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1688         if (bif == NULL)
1689                 return (ENOENT);
1690
1691         bif->bif_priority = req->ifbr_priority;
1692
1693         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1694                 bstp_initialization(sc);
1695
1696         return (0);
1697 }
1698
1699 static int
1700 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1701 {
1702         struct ifbreq *req = arg;
1703         struct bridge_iflist *bif;
1704
1705         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1706         if (bif == NULL)
1707                 return (ENOENT);
1708
1709         bif->bif_path_cost = req->ifbr_path_cost;
1710
1711         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1712                 bstp_initialization(sc);
1713
1714         return (0);
1715 }
1716
1717 static int
1718 bridge_ioctl_sifbondwght(struct bridge_softc *sc, void *arg)
1719 {
1720         struct ifbreq *req = arg;
1721         struct bridge_iflist *bif;
1722
1723         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1724         if (bif == NULL)
1725                 return (ENOENT);
1726
1727         bif->bif_bond_weight = req->ifbr_bond_weight;
1728
1729         /* no reinit needed */
1730
1731         return (0);
1732 }
1733
1734 static int
1735 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1736 {
1737         struct ifbreq *req = arg;
1738         struct bridge_iflist *bif;
1739         struct ifnet *ifs;
1740         struct bridge_ifinfo *bif_info;
1741
1742         ifs = ifunit(req->ifbr_ifsname);
1743         if (ifs == NULL)
1744                 return (ENOENT);
1745
1746         TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
1747                 if (ifs == bif->bif_ifp)
1748                         return (EBUSY);
1749
1750         if (ifs->if_bridge != NULL)
1751                 return (EBUSY);
1752
1753         switch (ifs->if_type) {
1754         case IFT_ETHER:
1755         case IFT_GIF:
1756         case IFT_L2VLAN:
1757                 break;
1758
1759         default:
1760                 return (EINVAL);
1761         }
1762
1763         /*
1764          * bif_info is needed for bif_flags
1765          */
1766         bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO);
1767         bif_info->bifi_ifp = ifs;
1768
1769         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
1770         bif->bif_ifp = ifs;
1771         bif->bif_info = bif_info;
1772         bif->bif_flags = IFBIF_SPAN;
1773         /* NOTE: span bif does not need bridge_ifinfo */
1774
1775         TAILQ_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1776
1777         sc->sc_span = 1;
1778
1779         return (0);
1780 }
1781
1782 static int
1783 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1784 {
1785         struct ifbreq *req = arg;
1786         struct bridge_iflist *bif;
1787         struct ifnet *ifs;
1788
1789         ifs = ifunit(req->ifbr_ifsname);
1790         if (ifs == NULL)
1791                 return (ENOENT);
1792
1793         TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
1794                 if (ifs == bif->bif_ifp)
1795                         break;
1796
1797         if (bif == NULL)
1798                 return (ENOENT);
1799
1800         bridge_delete_span(sc, bif);
1801
1802         if (TAILQ_EMPTY(&sc->sc_spanlist))
1803                 sc->sc_span = 0;
1804
1805         return (0);
1806 }
1807
1808 static void
1809 bridge_ifdetach_dispatch(netmsg_t msg)
1810 {
1811         struct ifnet *ifp, *bifp;
1812         struct bridge_softc *sc;
1813         struct bridge_iflist *bif;
1814
1815         ifp = msg->lmsg.u.ms_resultp;
1816         sc = ifp->if_bridge;
1817
1818         /* Check if the interface is a bridge member */
1819         if (sc != NULL) {
1820                 bifp = sc->sc_ifp;
1821
1822                 ifnet_serialize_all(bifp);
1823
1824                 bif = bridge_lookup_member_if(sc, ifp);
1825                 if (bif != NULL) {
1826                         bridge_delete_member(sc, bif, 1);
1827                 } else {
1828                         /* XXX Why bif will be NULL? */
1829                 }
1830
1831                 ifnet_deserialize_all(bifp);
1832                 goto reply;
1833         }
1834
1835         crit_enter();   /* XXX MP */
1836
1837         /* Check if the interface is a span port */
1838         LIST_FOREACH(sc, &bridge_list, sc_list) {
1839                 bifp = sc->sc_ifp;
1840
1841                 ifnet_serialize_all(bifp);
1842
1843                 TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next)
1844                         if (ifp == bif->bif_ifp) {
1845                                 bridge_delete_span(sc, bif);
1846                                 break;
1847                         }
1848
1849                 ifnet_deserialize_all(bifp);
1850         }
1851
1852         crit_exit();
1853
1854 reply:
1855         lwkt_replymsg(&msg->lmsg, 0);
1856 }
1857
1858 /*
1859  * bridge_ifdetach:
1860  *
1861  *      Detach an interface from a bridge.  Called when a member
1862  *      interface is detaching.
1863  */
1864 static void
1865 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1866 {
1867         struct netmsg_base msg;
1868
1869         netmsg_init(&msg, NULL, &curthread->td_msgport,
1870                     0, bridge_ifdetach_dispatch);
1871         msg.lmsg.u.ms_resultp = ifp;
1872
1873         lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0);
1874 }
1875
1876 /*
1877  * bridge_init:
1878  *
1879  *      Initialize a bridge interface.
1880  */
1881 static void
1882 bridge_init(void *xsc)
1883 {
1884         bridge_control(xsc, SIOCSIFFLAGS, bridge_ioctl_init, NULL);
1885 }
1886
1887 /*
1888  * bridge_stop:
1889  *
1890  *      Stop the bridge interface.
1891  */
1892 static void
1893 bridge_stop(struct ifnet *ifp)
1894 {
1895         bridge_control(ifp->if_softc, SIOCSIFFLAGS, bridge_ioctl_stop, NULL);
1896 }
1897
1898 /*
1899  * Returns TRUE if the packet is being sent 'from us'... from our bridge
1900  * interface or from any member of our bridge interface.  This is used
1901  * later on to force the MAC to be the MAC of our bridge interface.
1902  */
1903 static int
1904 bridge_from_us(struct bridge_softc *sc, struct ether_header *eh)
1905 {
1906         struct bridge_iflist *bif;
1907
1908         if (memcmp(eh->ether_shost, IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN) == 0)
1909                 return (1);
1910
1911         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1912                 if (memcmp(eh->ether_shost, IF_LLADDR(bif->bif_ifp),
1913                            ETHER_ADDR_LEN) == 0) {
1914                         return (1);
1915                 }
1916         }
1917         return (0);
1918 }
1919
1920 /*
1921  * bridge_enqueue:
1922  *
1923  *      Enqueue a packet on a bridge member interface.
1924  *
1925  */
1926 void
1927 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m)
1928 {
1929         struct netmsg_packet *nmp;
1930
1931         mbuftrackid(m, 64);
1932
1933         nmp = &m->m_hdr.mh_netmsg;
1934         netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
1935                     0, bridge_enqueue_handler);
1936         nmp->nm_packet = m;
1937         nmp->base.lmsg.u.ms_resultp = dst_ifp;
1938
1939         lwkt_sendmsg(netisr_portfn(mycpu->gd_cpuid), &nmp->base.lmsg);
1940 }
1941
1942 /*
1943  * After looking up dst_if in our forwarding table we still have to
1944  * deal with channel bonding.  Find the best interface in the bonding set.
1945  */
1946 static struct ifnet *
1947 bridge_select_unicast(struct bridge_softc *sc, struct ifnet *dst_if,
1948                       int from_blocking, struct mbuf *m)
1949 {
1950         struct bridge_iflist *bif, *nbif;
1951         struct ifnet *alt_if;
1952         int alt_priority;
1953         int priority;
1954
1955         /*
1956          * Unicast, kinda replicates the output side of bridge_output().
1957          *
1958          * Even though this is a uni-cast packet we may have to select
1959          * an interface from a bonding set.
1960          */
1961         bif = bridge_lookup_member_if(sc, dst_if);
1962         if (bif == NULL) {
1963                 /* Not a member of the bridge (anymore?) */
1964                 return NULL;
1965         }
1966
1967         /*
1968          * If STP is enabled on the target we are an equal opportunity
1969          * employer and do not necessarily output to dst_if.  Instead
1970          * scan available links with the same MAC as the current dst_if
1971          * and choose the best one.
1972          *
1973          * We also need to do this because arp entries tag onto a particular
1974          * interface and if it happens to be dead then the packets will
1975          * go into a bit bucket.
1976          *
1977          * If LINK2 is set the matching links are bonded and we-round robin.
1978          * (the MAC address must be the same for the participating links).
1979          * In this case links in a STP FORWARDING or BONDED state are
1980          * allowed for unicast packets.
1981          */
1982         if (bif->bif_flags & IFBIF_STP) {
1983                 alt_if = NULL;
1984                 alt_priority = 0;
1985                 priority = 0;
1986
1987                 TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid],
1988                                      bif_next, nbif) {
1989                         /*
1990                          * dst_if may imply a bonding set so we must compare
1991                          * MAC addresses.
1992                          */
1993                         if (memcmp(IF_LLADDR(bif->bif_ifp),
1994                                    IF_LLADDR(dst_if),
1995                                    ETHER_ADDR_LEN) != 0) {
1996                                 continue;
1997                         }
1998
1999                         if ((bif->bif_ifp->if_flags & IFF_RUNNING) == 0)
2000                                 continue;
2001
2002                         /*
2003                          * NOTE: We allow tranmissions through a BLOCKING
2004                          *       or LEARNING interface only as a last resort.
2005                          *       We DISALLOW both cases if the receiving
2006                          *
2007                          * NOTE: If we send a packet through a learning
2008                          *       interface the receiving end (if also in
2009                          *       LEARNING) will throw it away, so this is
2010                          *       the ultimate last resort.
2011                          */
2012                         switch(bif->bif_state) {
2013                         case BSTP_IFSTATE_BLOCKING:
2014                                 if (from_blocking == 0 &&
2015                                     bif->bif_priority + 256 > alt_priority) {
2016                                         alt_priority = bif->bif_priority + 256;
2017                                         alt_if = bif->bif_ifp;
2018                                 }
2019                                 continue;
2020                         case BSTP_IFSTATE_LEARNING:
2021                                 if (from_blocking == 0 &&
2022                                     bif->bif_priority > alt_priority) {
2023                                         alt_priority = bif->bif_priority;
2024                                         alt_if = bif->bif_ifp;
2025                                 }
2026                                 continue;
2027                         case BSTP_IFSTATE_L1BLOCKING:
2028                         case BSTP_IFSTATE_LISTENING:
2029                         case BSTP_IFSTATE_DISABLED:
2030                                 continue;
2031                         default:
2032                                 /* FORWARDING, BONDED */
2033                                 break;
2034                         }
2035
2036                         /*
2037                          * XXX we need to use the toepliz hash or
2038                          *     something like that instead of
2039                          *     round-robining.
2040                          */
2041                         if (sc->sc_ifp->if_flags & IFF_LINK2) {
2042                                 dst_if = bif->bif_ifp;
2043                                 if (++bif->bif_bond_count >=
2044                                     bif->bif_bond_weight) {
2045                                         bif->bif_bond_count = 0;
2046                                         TAILQ_REMOVE(&sc->sc_iflists[mycpuid],
2047                                                      bif, bif_next);
2048                                         TAILQ_INSERT_TAIL(
2049                                                      &sc->sc_iflists[mycpuid],
2050                                                      bif, bif_next);
2051                                 }
2052                                 priority = 1;
2053                                 break;
2054                         }
2055
2056                         /*
2057                          * Select best interface in the FORWARDING or
2058                          * BONDED set.  Well, there shouldn't be any
2059                          * in a BONDED state if LINK2 is not set (they
2060                          * will all be in a BLOCKING) state, but there
2061                          * could be a transitory condition here.
2062                          */
2063                         if (bif->bif_priority > priority) {
2064                                 priority = bif->bif_priority;
2065                                 dst_if = bif->bif_ifp;
2066                         }
2067                 }
2068
2069                 /*
2070                  * If no suitable interfaces were found but a suitable
2071                  * alternative interface was found, use the alternative
2072                  * interface.
2073                  */
2074                 if (priority == 0 && alt_if)
2075                         dst_if = alt_if;
2076         }
2077
2078         /*
2079          * At this point, we're dealing with a unicast frame
2080          * going to a different interface.
2081          */
2082         if ((dst_if->if_flags & IFF_RUNNING) == 0)
2083                 dst_if = NULL;
2084         return (dst_if);
2085 }
2086
2087
2088 /*
2089  * bridge_output:
2090  *
2091  *      Send output from a bridge member interface.  This
2092  *      performs the bridging function for locally originated
2093  *      packets.
2094  *
2095  *      The mbuf has the Ethernet header already attached.  We must
2096  *      enqueue or free the mbuf before returning.
2097  */
2098 static int
2099 bridge_output(struct ifnet *ifp, struct mbuf *m)
2100 {
2101         struct bridge_softc *sc = ifp->if_bridge;
2102         struct bridge_iflist *bif, *nbif;
2103         struct ether_header *eh;
2104         struct ifnet *dst_if, *alt_if, *bifp;
2105         int from_us;
2106         int alt_priority;
2107
2108         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2109         mbuftrackid(m, 65);
2110
2111         /*
2112          * Make sure that we are still a member of a bridge interface.
2113          */
2114         if (sc == NULL) {
2115                 m_freem(m);
2116                 return (0);
2117         }
2118         bifp = sc->sc_ifp;
2119
2120         /*
2121          * Acquire header
2122          */
2123         if (m->m_len < ETHER_HDR_LEN) {
2124                 m = m_pullup(m, ETHER_HDR_LEN);
2125                 if (m == NULL) {
2126                         bifp->if_oerrors++;
2127                         return (0);
2128                 }
2129         }
2130         eh = mtod(m, struct ether_header *);
2131         from_us = bridge_from_us(sc, eh);
2132
2133         /*
2134          * If bridge is down, but the original output interface is up,
2135          * go ahead and send out that interface.  Otherwise, the packet
2136          * is dropped below.
2137          */
2138         if ((bifp->if_flags & IFF_RUNNING) == 0) {
2139                 dst_if = ifp;
2140                 goto sendunicast;
2141         }
2142
2143         /*
2144          * If the packet is a multicast, or we don't know a better way to
2145          * get there, send to all interfaces.
2146          */
2147         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2148                 dst_if = NULL;
2149         else
2150                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2151
2152         if (dst_if == NULL) {
2153                 struct mbuf *mc;
2154                 int used = 0;
2155                 int found = 0;
2156
2157                 if (sc->sc_span)
2158                         bridge_span(sc, m);
2159
2160                 alt_if = NULL;
2161                 alt_priority = 0;
2162                 TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid],
2163                                      bif_next, nbif) {
2164                         dst_if = bif->bif_ifp;
2165
2166                         if ((dst_if->if_flags & IFF_RUNNING) == 0)
2167                                 continue;
2168
2169                         /*
2170                          * If this is not the original output interface,
2171                          * and the interface is participating in spanning
2172                          * tree, make sure the port is in a state that
2173                          * allows forwarding.
2174                          *
2175                          * We keep track of a possible backup IF if we are
2176                          * unable to find any interfaces to forward through.
2177                          *
2178                          * NOTE: Currently round-robining is not implemented
2179                          *       across bonded interface groups (needs an
2180                          *       algorithm to track each group somehow).
2181                          *
2182                          *       Similarly we track only one alternative
2183                          *       interface if no suitable interfaces are
2184                          *       found.
2185                          */
2186                         if (dst_if != ifp &&
2187                             (bif->bif_flags & IFBIF_STP) != 0) {
2188                                 switch (bif->bif_state) {
2189                                 case BSTP_IFSTATE_BONDED:
2190                                         if (bif->bif_priority + 512 >
2191                                             alt_priority) {
2192                                                 alt_priority =
2193                                                     bif->bif_priority + 512;
2194                                                 alt_if = bif->bif_ifp;
2195                                         }
2196                                         continue;
2197                                 case BSTP_IFSTATE_BLOCKING:
2198                                         if (bif->bif_priority + 256 >
2199                                             alt_priority) {
2200                                                 alt_priority =
2201                                                     bif->bif_priority + 256;
2202                                                 alt_if = bif->bif_ifp;
2203                                         }
2204                                         continue;
2205                                 case BSTP_IFSTATE_LEARNING:
2206                                         if (bif->bif_priority > alt_priority) {
2207                                                 alt_priority =
2208                                                     bif->bif_priority;
2209                                                 alt_if = bif->bif_ifp;
2210                                         }
2211                                         continue;
2212                                 case BSTP_IFSTATE_L1BLOCKING:
2213                                 case BSTP_IFSTATE_LISTENING:
2214                                 case BSTP_IFSTATE_DISABLED:
2215                                         continue;
2216                                 default:
2217                                         /* FORWARDING */
2218                                         break;
2219                                 }
2220                         }
2221
2222                         KKASSERT(used == 0);
2223                         if (TAILQ_NEXT(bif, bif_next) == NULL) {
2224                                 used = 1;
2225                                 mc = m;
2226                         } else {
2227                                 mc = m_copypacket(m, MB_DONTWAIT);
2228                                 if (mc == NULL) {
2229                                         bifp->if_oerrors++;
2230                                         continue;
2231                                 }
2232                         }
2233
2234                         /*
2235                          * If the packet is 'from' us override ether_shost.
2236                          */
2237                         bridge_handoff(sc, dst_if, mc, from_us);
2238                         found = 1;
2239
2240                         if (nbif != NULL && !nbif->bif_onlist) {
2241                                 KKASSERT(bif->bif_onlist);
2242                                 nbif = TAILQ_NEXT(bif, bif_next);
2243                         }
2244                 }
2245
2246                 /*
2247                  * If we couldn't find anything use the backup interface
2248                  * if we have one.
2249                  */
2250                 if (found == 0 && alt_if) {
2251                         KKASSERT(used == 0);
2252                         mc = m;
2253                         used = 1;
2254                         bridge_handoff(sc, alt_if, mc, from_us);
2255                 }
2256
2257                 if (used == 0)
2258                         m_freem(m);
2259                 return (0);
2260         }
2261
2262         /*
2263          * Unicast
2264          */
2265 sendunicast:
2266         dst_if = bridge_select_unicast(sc, dst_if, 0, m);
2267
2268         if (sc->sc_span)
2269                 bridge_span(sc, m);
2270         if (dst_if == NULL)
2271                 m_freem(m);
2272         else
2273                 bridge_handoff(sc, dst_if, m, from_us);
2274         return (0);
2275 }
2276
2277 /*
2278  * Returns the bridge interface associated with an ifc.
2279  * Pass ifp->if_bridge (must not be NULL).  Used by the ARP
2280  * code to supply the bridge for the is-at info, making
2281  * the bridge responsible for matching local addresses.
2282  *
2283  * Without this the ARP code will supply bridge member interfaces
2284  * for the is-at which makes it difficult the bridge to fail-over
2285  * interfaces (amoung other things).
2286  */
2287 static struct ifnet *
2288 bridge_interface(void *if_bridge)
2289 {
2290         struct bridge_softc *sc = if_bridge;
2291         return (sc->sc_ifp);
2292 }
2293
2294 /*
2295  * bridge_start:
2296  *
2297  *      Start output on a bridge.
2298  */
2299 static void
2300 bridge_start(struct ifnet *ifp, struct ifaltq_subque *ifsq)
2301 {
2302         struct bridge_softc *sc = ifp->if_softc;
2303
2304         ASSERT_ALTQ_SQ_DEFAULT(ifp, ifsq);
2305         ASSERT_IFNET_SERIALIZED_TX(ifp);
2306
2307         ifsq_set_oactive(ifsq);
2308         for (;;) {
2309                 struct ifnet *dst_if = NULL;
2310                 struct ether_header *eh;
2311                 struct mbuf *m;
2312
2313                 m = ifsq_dequeue(ifsq, NULL);
2314                 if (m == NULL)
2315                         break;
2316                 mbuftrackid(m, 75);
2317
2318                 if (m->m_len < sizeof(*eh)) {
2319                         m = m_pullup(m, sizeof(*eh));
2320                         if (m == NULL) {
2321                                 ifp->if_oerrors++;
2322                                 continue;
2323                         }
2324                 }
2325                 eh = mtod(m, struct ether_header *);
2326
2327                 BPF_MTAP(ifp, m);
2328                 ifp->if_opackets++;
2329
2330                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0)
2331                         dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2332
2333                 /*
2334                  * Multicast or broadcast
2335                  */
2336                 if (dst_if == NULL) {
2337                         bridge_start_bcast(sc, m);
2338                         continue;
2339                 }
2340
2341                 /*
2342                  * Unicast
2343                  */
2344                 dst_if = bridge_select_unicast(sc, dst_if, 0, m);
2345
2346                 if (dst_if == NULL)
2347                         m_freem(m);
2348                 else
2349                         bridge_enqueue(dst_if, m);
2350         }
2351         ifsq_clr_oactive(ifsq);
2352 }
2353
2354 /*
2355  * bridge_forward:
2356  *
2357  *      Forward packets received on a bridge interface via the input
2358  *      path.
2359  *
2360  *      This implements the forwarding function of the bridge.
2361  */
2362 static void
2363 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
2364 {
2365         struct bridge_iflist *bif;
2366         struct ifnet *src_if, *dst_if, *ifp;
2367         struct ether_header *eh;
2368         int from_blocking;
2369
2370         mbuftrackid(m, 66);
2371         src_if = m->m_pkthdr.rcvif;
2372         ifp = sc->sc_ifp;
2373
2374         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2375
2376         ifp->if_ipackets++;
2377         ifp->if_ibytes += m->m_pkthdr.len;
2378
2379         /*
2380          * Look up the bridge_iflist.
2381          */
2382         bif = bridge_lookup_member_if(sc, src_if);
2383         if (bif == NULL) {
2384                 /* Interface is not a bridge member (anymore?) */
2385                 m_freem(m);
2386                 return;
2387         }
2388
2389         /*
2390          * In spanning tree mode receiving a packet from an interface
2391          * in a BLOCKING state is allowed, it could be a member of last
2392          * resort from the sender's point of view, but forwarding it is
2393          * not allowed.
2394          *
2395          * The sender's spanning tree will eventually sync up and the
2396          * sender will go into a BLOCKING state too (but this still may be
2397          * an interface of last resort during state changes).
2398          */
2399         if (bif->bif_flags & IFBIF_STP) {
2400                 switch (bif->bif_state) {
2401                 case BSTP_IFSTATE_L1BLOCKING:
2402                 case BSTP_IFSTATE_LISTENING:
2403                 case BSTP_IFSTATE_DISABLED:
2404                         m_freem(m);
2405                         return;
2406                 default:
2407                         /* learning, blocking, bonded, forwarding */
2408                         break;
2409                 }
2410         }
2411         from_blocking = (bif->bif_state == BSTP_IFSTATE_BLOCKING);
2412
2413         eh = mtod(m, struct ether_header *);
2414
2415         /*
2416          * If the interface is learning, and the source
2417          * address is valid and not multicast, record
2418          * the address.
2419          */
2420         if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
2421             from_blocking == 0 &&
2422             ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
2423             (eh->ether_shost[0] == 0 &&
2424              eh->ether_shost[1] == 0 &&
2425              eh->ether_shost[2] == 0 &&
2426              eh->ether_shost[3] == 0 &&
2427              eh->ether_shost[4] == 0 &&
2428              eh->ether_shost[5] == 0) == 0) {
2429                 bridge_rtupdate(sc, eh->ether_shost, src_if, IFBAF_DYNAMIC);
2430         }
2431
2432         /*
2433          * Don't forward from an interface in the listening or learning
2434          * state.  That is, in the learning state we learn information
2435          * but we throw away the packets.
2436          *
2437          * We let through packets on interfaces in the blocking state.
2438          * The blocking state is applicable to the send side, not the
2439          * receive side.
2440          */
2441         if ((bif->bif_flags & IFBIF_STP) != 0 &&
2442             (bif->bif_state == BSTP_IFSTATE_LISTENING ||
2443              bif->bif_state == BSTP_IFSTATE_LEARNING)) {
2444                 m_freem(m);
2445                 return;
2446         }
2447
2448         /*
2449          * At this point, the port either doesn't participate
2450          * in spanning tree or it is in the forwarding state.
2451          */
2452
2453         /*
2454          * If the packet is unicast, destined for someone on
2455          * "this" side of the bridge, drop it.
2456          *
2457          * src_if implies the entire bonding set so we have to compare MAC
2458          * addresses and not just if pointers.
2459          */
2460         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2461                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2462                 if (dst_if && memcmp(IF_LLADDR(src_if), IF_LLADDR(dst_if),
2463                                      ETHER_ADDR_LEN) == 0) {
2464                         m_freem(m);
2465                         return;
2466                 }
2467         } else {
2468                 /* ...forward it to all interfaces. */
2469                 ifp->if_imcasts++;
2470                 dst_if = NULL;
2471         }
2472
2473         /*
2474          * Brodcast if we do not have forwarding information.  However, if
2475          * we received the packet on a blocking interface we do not do this
2476          * (unless you really want to blow up your network).
2477          */
2478         if (dst_if == NULL) {
2479                 if (from_blocking)
2480                         m_freem(m);
2481                 else
2482                         bridge_broadcast(sc, src_if, m);
2483                 return;
2484         }
2485
2486         dst_if = bridge_select_unicast(sc, dst_if, from_blocking, m);
2487
2488         if (dst_if == NULL) {
2489                 m_freem(m);
2490                 return;
2491         }
2492
2493         if (inet_pfil_hook.ph_hashooks > 0
2494 #ifdef INET6
2495             || inet6_pfil_hook.ph_hashooks > 0
2496 #endif
2497             ) {
2498                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2499                         return;
2500                 if (m == NULL)
2501                         return;
2502
2503                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2504                         return;
2505                 if (m == NULL)
2506                         return;
2507         }
2508         bridge_handoff(sc, dst_if, m, 0);
2509 }
2510
2511 /*
2512  * bridge_input:
2513  *
2514  *      Receive input from a member interface.  Queue the packet for
2515  *      bridging if it is not for us.
2516  */
2517 static struct mbuf *
2518 bridge_input(struct ifnet *ifp, struct mbuf *m)
2519 {
2520         struct bridge_softc *sc = ifp->if_bridge;
2521         struct bridge_iflist *bif;
2522         struct ifnet *bifp, *new_ifp;
2523         struct ether_header *eh;
2524         struct mbuf *mc, *mc2;
2525
2526         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2527         mbuftrackid(m, 67);
2528
2529         /*
2530          * Make sure that we are still a member of a bridge interface.
2531          */
2532         if (sc == NULL)
2533                 return m;
2534
2535         new_ifp = NULL;
2536         bifp = sc->sc_ifp;
2537
2538         if ((bifp->if_flags & IFF_RUNNING) == 0)
2539                 goto out;
2540
2541         /*
2542          * Implement support for bridge monitoring.  If this flag has been
2543          * set on this interface, discard the packet once we push it through
2544          * the bpf(4) machinery, but before we do, increment various counters
2545          * associated with this bridge.
2546          */
2547         if (bifp->if_flags & IFF_MONITOR) {
2548                 /* Change input interface to this bridge */
2549                 m->m_pkthdr.rcvif = bifp;
2550
2551                 BPF_MTAP(bifp, m);
2552
2553                 /* Update bridge's ifnet statistics */
2554                 bifp->if_ipackets++;
2555                 bifp->if_ibytes += m->m_pkthdr.len;
2556                 if (m->m_flags & (M_MCAST | M_BCAST))
2557                         bifp->if_imcasts++;
2558
2559                 m_freem(m);
2560                 m = NULL;
2561                 goto out;
2562         }
2563
2564         /*
2565          * Handle the ether_header
2566          *
2567          * In all cases if the packet is destined for us via our MAC
2568          * we must clear BRIDGE_MBUF_TAGGED to ensure that we don't
2569          * repeat the source MAC out the same interface.
2570          *
2571          * This first test against our bridge MAC is the fast-path.
2572          *
2573          * NOTE!  The bridge interface can serve as an endpoint for
2574          *        communication but normally there are no IPs associated
2575          *        with it so you cannot route through it.  Instead what
2576          *        you do is point your default route *THROUGH* the bridge
2577          *        to the actual default router for one of the bridged spaces.
2578          *
2579          *        Another possibility is to put all your IP specifications
2580          *        on the bridge instead of on the individual interfaces.  If
2581          *        you do this it should be possible to use the bridge as an
2582          *        end point and route (rather than switch) through it using
2583          *        the default route or ipfw forwarding rules.
2584          */
2585
2586         /*
2587          * Acquire header
2588          */
2589         if (m->m_len < ETHER_HDR_LEN) {
2590                 m = m_pullup(m, ETHER_HDR_LEN);
2591                 if (m == NULL)
2592                         goto out;
2593         }
2594         eh = mtod(m, struct ether_header *);
2595         m->m_pkthdr.fw_flags |= BRIDGE_MBUF_TAGGED;
2596         bcopy(eh, &m->m_pkthdr.br.ether, sizeof(*eh));
2597
2598         if ((bridge_debug & 1) &&
2599             (ntohs(eh->ether_type) == ETHERTYPE_ARP ||
2600             ntohs(eh->ether_type) == ETHERTYPE_REVARP)) {
2601                 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
2602                         "%02x:%02x:%02x:%02x:%02x:%02x type %04x "
2603                         "lla %02x:%02x:%02x:%02x:%02x:%02x\n",
2604                         eh->ether_dhost[0],
2605                         eh->ether_dhost[1],
2606                         eh->ether_dhost[2],
2607                         eh->ether_dhost[3],
2608                         eh->ether_dhost[4],
2609                         eh->ether_dhost[5],
2610                         eh->ether_shost[0],
2611                         eh->ether_shost[1],
2612                         eh->ether_shost[2],
2613                         eh->ether_shost[3],
2614                         eh->ether_shost[4],
2615                         eh->ether_shost[5],
2616                         eh->ether_type,
2617                         ((u_char *)IF_LLADDR(bifp))[0],
2618                         ((u_char *)IF_LLADDR(bifp))[1],
2619                         ((u_char *)IF_LLADDR(bifp))[2],
2620                         ((u_char *)IF_LLADDR(bifp))[3],
2621                         ((u_char *)IF_LLADDR(bifp))[4],
2622                         ((u_char *)IF_LLADDR(bifp))[5]
2623                 );
2624         }
2625
2626         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
2627                 /*
2628                  * If the packet is for us, set the packets source as the
2629                  * bridge, and return the packet back to ifnet.if_input for
2630                  * local processing.
2631                  */
2632                 m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2633                 KASSERT(bifp->if_bridge == NULL,
2634                         ("loop created in bridge_input"));
2635                 if (pfil_member != 0) {
2636                         if (inet_pfil_hook.ph_hashooks > 0
2637 #ifdef INET6
2638                             || inet6_pfil_hook.ph_hashooks > 0
2639 #endif
2640                         ) {
2641                                 if (bridge_pfil(&m, NULL, ifp, PFIL_IN) != 0)
2642                                         goto out;
2643                                 if (m == NULL)
2644                                         goto out;
2645                         }
2646                 }
2647                 new_ifp = bifp;
2648                 goto out;
2649         }
2650
2651         /*
2652          * Tap all packets arriving on the bridge, no matter if
2653          * they are local destinations or not.  In is in.
2654          */
2655         BPF_MTAP(bifp, m);
2656
2657         bif = bridge_lookup_member_if(sc, ifp);
2658         if (bif == NULL)
2659                 goto out;
2660
2661         if (sc->sc_span)
2662                 bridge_span(sc, m);
2663
2664         if (m->m_flags & (M_BCAST | M_MCAST)) {
2665                 /*
2666                  * Tap off 802.1D packets; they do not get forwarded.
2667                  */
2668                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2669                             ETHER_ADDR_LEN) == 0) {
2670                         ifnet_serialize_all(bifp);
2671                         bstp_input(sc, bif, m);
2672                         ifnet_deserialize_all(bifp);
2673
2674                         /* m is freed by bstp_input */
2675                         m = NULL;
2676                         goto out;
2677                 }
2678
2679                 /*
2680                  * Other than 802.11d packets, ignore packets if the
2681                  * interface is not in a good state.
2682                  *
2683                  * NOTE: Broadcast/mcast packets received on a blocking or
2684                  *       learning interface are allowed for local processing.
2685                  *
2686                  *       The sending side of a blocked port will stop
2687                  *       transmitting when a better alternative is found.
2688                  *       However, later on we will disallow the forwarding
2689                  *       of bcast/mcsat packets over a blocking interface.
2690                  */
2691                 if (bif->bif_flags & IFBIF_STP) {
2692                         switch (bif->bif_state) {
2693                         case BSTP_IFSTATE_L1BLOCKING:
2694                         case BSTP_IFSTATE_LISTENING:
2695                         case BSTP_IFSTATE_DISABLED:
2696                                 goto out;
2697                         default:
2698                                 /* blocking, learning, bonded, forwarding */
2699                                 break;
2700                         }
2701                 }
2702
2703                 /*
2704                  * Make a deep copy of the packet and enqueue the copy
2705                  * for bridge processing; return the original packet for
2706                  * local processing.
2707                  */
2708                 mc = m_dup(m, MB_DONTWAIT);
2709                 if (mc == NULL)
2710                         goto out;
2711
2712                 /*
2713                  * It's just too dangerous to allow bcast/mcast over a
2714                  * blocked interface, eventually the network will sort
2715                  * itself out and a better path will be found.
2716                  */
2717                 if ((bif->bif_flags & IFBIF_STP) == 0 ||
2718                     bif->bif_state != BSTP_IFSTATE_BLOCKING) {
2719                         bridge_forward(sc, mc);
2720                 }
2721
2722                 /*
2723                  * Reinject the mbuf as arriving on the bridge so we have a
2724                  * chance at claiming multicast packets. We can not loop back
2725                  * here from ether_input as a bridge is never a member of a
2726                  * bridge.
2727                  */
2728                 KASSERT(bifp->if_bridge == NULL,
2729                         ("loop created in bridge_input"));
2730                 mc2 = m_dup(m, MB_DONTWAIT);
2731 #ifdef notyet
2732                 if (mc2 != NULL) {
2733                         /* Keep the layer3 header aligned */
2734                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2735                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2736                 }
2737 #endif
2738                 if (mc2 != NULL) {
2739                         /*
2740                          * Don't tap to bpf(4) again; we have already done
2741                          * the tapping.
2742                          *
2743                          * Leave m_pkthdr.rcvif alone, so ARP replies are
2744                          * processed as coming in on the correct interface.
2745                          *
2746                          * Clear the bridge flag for local processing in
2747                          * case the packet gets routed.
2748                          */
2749                         mc2->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2750                         ether_reinput_oncpu(bifp, mc2, 0);
2751                 }
2752
2753                 /* Return the original packet for local processing. */
2754                 goto out;
2755         }
2756
2757         /*
2758          * Input of a unicast packet.  We have to allow unicast packets
2759          * input from links in the BLOCKING state as this might be an
2760          * interface of last resort.
2761          *
2762          * NOTE: We explicitly ignore normal packets received on a link
2763          *       in the BLOCKING state.  The point of being in that state
2764          *       is to avoid getting duplicate packets.
2765          *
2766          *       HOWEVER, if LINK2 is set the normal spanning tree code
2767          *       will mark an interface BLOCKING to avoid multi-cast/broadcast
2768          *       loops.  Unicast packets CAN still loop if we allow the
2769          *       case (hence we only do it in LINK2), but it isn't quite as
2770          *       bad as a broadcast packet looping.
2771          */
2772         if (bif->bif_flags & IFBIF_STP) {
2773                 switch (bif->bif_state) {
2774                 case BSTP_IFSTATE_L1BLOCKING:
2775                 case BSTP_IFSTATE_LISTENING:
2776                 case BSTP_IFSTATE_DISABLED:
2777                         goto out;
2778                 default:
2779                         /* blocking, bonded, forwarding, learning */
2780                         break;
2781                 }
2782         }
2783
2784         /*
2785          * Unicast.  Make sure it's not for us.
2786          *
2787          * This loop is MPSAFE; the only blocking operation (bridge_rtupdate)
2788          * is followed by breaking out of the loop.
2789          */
2790         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2791                 if (bif->bif_ifp->if_type != IFT_ETHER)
2792                         continue;
2793
2794                 /*
2795                  * It is destined for an interface linked to the bridge.
2796                  * We want the bridge itself to take care of link level
2797                  * forwarding to member interfaces so reinput on the bridge.
2798                  * i.e. if you ping an IP on a target interface associated
2799                  * with the bridge, the arp is-at response should indicate
2800                  * the bridge MAC.
2801                  *
2802                  * Only update our addr list when learning if the port
2803                  * is not in a blocking state.  If it is we still allow
2804                  * the packet but we do not try to learn from it.
2805                  */
2806                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
2807                            ETHER_ADDR_LEN) == 0) {
2808                         if (bif->bif_ifp != ifp) {
2809                                 /* XXX loop prevention */
2810                                 m->m_flags |= M_ETHER_BRIDGED;
2811                         }
2812                         if ((bif->bif_flags & IFBIF_LEARNING) &&
2813                             bif->bif_state != BSTP_IFSTATE_BLOCKING) {
2814                                 bridge_rtupdate(sc, eh->ether_shost,
2815                                                 ifp, IFBAF_DYNAMIC);
2816                         }
2817                         new_ifp = bifp; /* not bif->bif_ifp */
2818                         m->m_pkthdr.fw_flags &= ~BRIDGE_MBUF_TAGGED;
2819                         goto out;
2820                 }
2821
2822                 /*
2823                  * Ignore received packets that were sent by us.
2824                  */
2825                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
2826                            ETHER_ADDR_LEN) == 0) {
2827                         m_freem(m);
2828                         m = NULL;
2829                         goto out;
2830                 }
2831         }
2832
2833         /*
2834          * It isn't for us.
2835          *
2836          * Perform the bridge forwarding function, but disallow bridging
2837          * to interfaces in the blocking state if the packet came in on
2838          * an interface in the blocking state.
2839          */
2840         bridge_forward(sc, m);
2841         m = NULL;
2842
2843         /*
2844          * ether_reinput_oncpu() will reprocess rcvif as
2845          * coming from new_ifp (since we do not specify
2846          * REINPUT_KEEPRCVIF).
2847          */
2848 out:
2849         if (new_ifp != NULL) {
2850                 /*
2851                  * Clear the bridge flag for local processing in
2852                  * case the packet gets routed.
2853                  */
2854                 ether_reinput_oncpu(new_ifp, m, REINPUT_RUNBPF);
2855                 m = NULL;
2856         }
2857         return (m);
2858 }
2859
2860 /*
2861  * bridge_start_bcast:
2862  *
2863  *      Broadcast the packet sent from bridge to all member
2864  *      interfaces.
2865  *      This is a simplified version of bridge_broadcast(), however,
2866  *      this function expects caller to hold bridge's serializer.
2867  */
2868 static void
2869 bridge_start_bcast(struct bridge_softc *sc, struct mbuf *m)
2870 {
2871         struct bridge_iflist *bif;
2872         struct mbuf *mc;
2873         struct ifnet *dst_if, *alt_if, *bifp;
2874         int used = 0;
2875         int found = 0;
2876         int alt_priority;
2877
2878         mbuftrackid(m, 68);
2879         bifp = sc->sc_ifp;
2880         ASSERT_IFNET_SERIALIZED_ALL(bifp);
2881
2882         /*
2883          * Following loop is MPSAFE; nothing is blocking
2884          * in the loop body.
2885          *
2886          * NOTE: We transmit through an member in the BLOCKING state only
2887          *       as a last resort.
2888          */
2889         alt_if = NULL;
2890         alt_priority = 0;
2891
2892         TAILQ_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2893                 dst_if = bif->bif_ifp;
2894
2895                 if (bif->bif_flags & IFBIF_STP) {
2896                         switch (bif->bif_state) {
2897                         case BSTP_IFSTATE_BLOCKING:
2898                                 if (bif->bif_priority > alt_priority) {
2899                                         alt_priority = bif->bif_priority;
2900                                         alt_if = bif->bif_ifp;
2901                                 }
2902                                 /* fall through */
2903                         case BSTP_IFSTATE_L1BLOCKING:
2904                         case BSTP_IFSTATE_DISABLED:
2905                                 continue;
2906                         default:
2907                                 /* listening, learning, bonded, forwarding */
2908                                 break;
2909                         }
2910                 }
2911
2912                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2913                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2914                         continue;
2915
2916                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2917                         continue;
2918
2919                 if (TAILQ_NEXT(bif, bif_next) == NULL) {
2920                         mc = m;
2921                         used = 1;
2922                 } else {
2923                         mc = m_copypacket(m, MB_DONTWAIT);
2924                         if (mc == NULL) {
2925                                 bifp->if_oerrors++;
2926                                 continue;
2927                         }
2928                 }
2929                 found = 1;
2930                 bridge_enqueue(dst_if, mc);
2931         }
2932
2933         if (found == 0 && alt_if) {
2934                 KKASSERT(used == 0);
2935                 mc = m;
2936                 used = 1;
2937                 bridge_enqueue(alt_if, mc);
2938         }
2939
2940         if (used == 0)
2941                 m_freem(m);
2942 }
2943
2944 /*
2945  * bridge_broadcast:
2946  *
2947  *      Send a frame to all interfaces that are members of
2948  *      the bridge, except for the one on which the packet
2949  *      arrived.
2950  */
2951 static void
2952 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2953                  struct mbuf *m)
2954 {
2955         struct bridge_iflist *bif, *nbif;
2956         struct ether_header *eh;
2957         struct mbuf *mc;
2958         struct ifnet *dst_if, *alt_if, *bifp;
2959         int used;
2960         int found;
2961         int alt_priority;
2962         int from_us;
2963
2964         mbuftrackid(m, 69);
2965         bifp = sc->sc_ifp;
2966         ASSERT_IFNET_NOT_SERIALIZED_ALL(bifp);
2967
2968         eh = mtod(m, struct ether_header *);
2969         from_us = bridge_from_us(sc, eh);
2970
2971         if (inet_pfil_hook.ph_hashooks > 0
2972 #ifdef INET6
2973             || inet6_pfil_hook.ph_hashooks > 0
2974 #endif
2975             ) {
2976                 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0)
2977                         return;
2978                 if (m == NULL)
2979                         return;
2980
2981                 /* Filter on the bridge interface before broadcasting */
2982                 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0)
2983                         return;
2984                 if (m == NULL)
2985                         return;
2986         }
2987
2988         alt_if = NULL;
2989         alt_priority = 0;
2990         found = 0;
2991         used = 0;
2992
2993         TAILQ_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
2994                 dst_if = bif->bif_ifp;
2995
2996                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2997                         continue;
2998
2999                 /*
3000                  * Don't bounce the packet out the same interface it came
3001                  * in on.  We have to test MAC addresses because a packet
3002                  * can come in a bonded interface and we don't want it to
3003                  * be echod out the forwarding interface for the same bonding
3004                  * set.
3005                  */
3006                 if (src_if && memcmp(IF_LLADDR(src_if), IF_LLADDR(dst_if),
3007                                      ETHER_ADDR_LEN) == 0) {
3008                         continue;
3009                 }
3010
3011                 /*
3012                  * Generally speaking we only broadcast through forwarding
3013                  * interfaces.  If no interfaces are available we select
3014                  * a BONDED, BLOCKING, or LEARNING interface to forward
3015                  * through.
3016                  */
3017                 if (bif->bif_flags & IFBIF_STP) {
3018                         switch (bif->bif_state) {
3019                         case BSTP_IFSTATE_BONDED:
3020                                 if (bif->bif_priority + 512 > alt_priority) {
3021                                         alt_priority = bif->bif_priority + 512;
3022                                         alt_if = bif->bif_ifp;
3023                                 }
3024                                 continue;
3025                         case BSTP_IFSTATE_BLOCKING:
3026                                 if (bif->bif_priority + 256 > alt_priority) {
3027                                         alt_priority = bif->bif_priority + 256;
3028                                         alt_if = bif->bif_ifp;
3029                                 }
3030                                 continue;
3031                         case BSTP_IFSTATE_LEARNING:
3032                                 if (bif->bif_priority > alt_priority) {
3033                                         alt_priority = bif->bif_priority;
3034                                         alt_if = bif->bif_ifp;
3035                                 }
3036                                 continue;
3037                         case BSTP_IFSTATE_L1BLOCKING:
3038                         case BSTP_IFSTATE_DISABLED:
3039                         case BSTP_IFSTATE_LISTENING:
3040                                 continue;
3041                         default:
3042                                 /* forwarding */
3043                                 break;
3044                         }
3045                 }
3046
3047                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
3048                     (m->m_flags & (M_BCAST|M_MCAST)) == 0) {
3049                         continue;
3050                 }
3051
3052                 if (TAILQ_NEXT(bif, bif_next) == NULL) {
3053                         mc = m;
3054                         used = 1;
3055                 } else {
3056                         mc = m_copypacket(m, MB_DONTWAIT);
3057                         if (mc == NULL) {
3058                                 sc->sc_ifp->if_oerrors++;
3059                                 continue;
3060                         }
3061                 }
3062                 found = 1;
3063
3064                 /*
3065                  * Filter on the output interface.  Pass a NULL bridge
3066                  * interface pointer so we do not redundantly filter on
3067                  * the bridge for each interface we broadcast on.
3068                  */
3069                 if (inet_pfil_hook.ph_hashooks > 0
3070 #ifdef INET6
3071                     || inet6_pfil_hook.ph_hashooks > 0
3072 #endif
3073                     ) {
3074                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
3075                                 continue;
3076                         if (mc == NULL)
3077                                 continue;
3078                 }
3079                 bridge_handoff(sc, dst_if, mc, from_us);
3080
3081                 if (nbif != NULL && !nbif->bif_onlist) {
3082                         KKASSERT(bif->bif_onlist);
3083                         nbif = TAILQ_NEXT(bif, bif_next);
3084                 }
3085         }
3086
3087         if (found == 0 && alt_if) {
3088                 KKASSERT(used == 0);
3089                 mc = m;
3090                 used = 1;
3091                 bridge_enqueue(alt_if, mc);
3092         }
3093
3094         if (used == 0)
3095                 m_freem(m);
3096 }
3097
3098 /*
3099  * bridge_span:
3100  *
3101  *      Duplicate a packet out one or more interfaces that are in span mode,
3102  *      the original mbuf is unmodified.
3103  */
3104 static void
3105 bridge_span(struct bridge_softc *sc, struct mbuf *m)
3106 {
3107         struct bridge_iflist *bif;
3108         struct ifnet *dst_if, *bifp;
3109         struct mbuf *mc;
3110
3111         mbuftrackid(m, 70);
3112         bifp = sc->sc_ifp;
3113         ifnet_serialize_all(bifp);
3114
3115         TAILQ_FOREACH(bif, &sc->sc_spanlist, bif_next) {
3116                 dst_if = bif->bif_ifp;
3117
3118                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
3119                         continue;
3120
3121                 mc = m_copypacket(m, MB_DONTWAIT);
3122                 if (mc == NULL) {
3123                         sc->sc_ifp->if_oerrors++;
3124                         continue;
3125                 }
3126                 bridge_enqueue(dst_if, mc);
3127         }
3128
3129         ifnet_deserialize_all(bifp);
3130 }
3131
3132 static void
3133 bridge_rtmsg_sync_handler(netmsg_t msg)
3134 {
3135         ifnet_forwardmsg(&msg->lmsg, mycpuid + 1);
3136 }
3137
3138 static void
3139 bridge_rtmsg_sync(struct bridge_softc *sc)
3140 {
3141         struct netmsg_base msg;
3142
3143         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3144
3145         netmsg_init(&msg, NULL, &curthread->td_msgport,
3146                     0, bridge_rtmsg_sync_handler);
3147         ifnet_domsg(&msg.lmsg, 0);
3148 }
3149
3150 static __inline void
3151 bridge_rtinfo_update(struct bridge_rtinfo *bri, struct ifnet *dst_if,
3152                      int setflags, uint8_t flags, uint32_t timeo)
3153 {
3154         if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
3155             bri->bri_ifp != dst_if)
3156                 bri->bri_ifp = dst_if;
3157         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
3158             bri->bri_expire != time_second + timeo)
3159                 bri->bri_expire = time_second + timeo;
3160         if (setflags)
3161                 bri->bri_flags = flags;
3162 }
3163
3164 static int
3165 bridge_rtinstall_oncpu(struct bridge_softc *sc, const uint8_t *dst,
3166                        struct ifnet *dst_if, int setflags, uint8_t flags,
3167                        struct bridge_rtinfo **bri0)
3168 {
3169         struct bridge_rtnode *brt;
3170         struct bridge_rtinfo *bri;
3171
3172         if (mycpuid == 0) {
3173                 brt = bridge_rtnode_lookup(sc, dst);
3174                 if (brt != NULL) {
3175                         /*
3176                          * rtnode for 'dst' already exists.  We inform the
3177                          * caller about this by leaving bri0 as NULL.  The
3178                          * caller will terminate the intallation upon getting
3179                          * NULL bri0.  However, we still need to update the
3180                          * rtinfo.
3181                          */
3182                         KKASSERT(*bri0 == NULL);
3183
3184                         /* Update rtinfo */
3185                         bridge_rtinfo_update(brt->brt_info, dst_if, setflags,
3186                                              flags, sc->sc_brttimeout);
3187                         return 0;
3188                 }
3189
3190                 /*
3191                  * We only need to check brtcnt on CPU0, since if limit
3192                  * is to be exceeded, ENOSPC is returned.  Caller knows
3193                  * this and will terminate the installation.
3194                  */
3195                 if (sc->sc_brtcnt >= sc->sc_brtmax)
3196                         return ENOSPC;
3197
3198                 KKASSERT(*bri0 == NULL);
3199                 bri = kmalloc(sizeof(struct bridge_rtinfo), M_DEVBUF,
3200                                   M_WAITOK | M_ZERO);
3201                 *bri0 = bri;
3202
3203                 /* Setup rtinfo */
3204                 bri->bri_flags = IFBAF_DYNAMIC;
3205                 bridge_rtinfo_update(bri, dst_if, setflags, flags,
3206                                      sc->sc_brttimeout);
3207         } else {
3208                 bri = *bri0;
3209                 KKASSERT(bri != NULL);
3210         }
3211
3212         brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF,
3213                       M_WAITOK | M_ZERO);
3214         memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
3215         brt->brt_info = bri;
3216
3217         bridge_rtnode_insert(sc, brt);
3218         return 0;
3219 }
3220
3221 static void
3222 bridge_rtinstall_handler(netmsg_t msg)
3223 {
3224         struct netmsg_brsaddr *brmsg = (struct netmsg_brsaddr *)msg;
3225         int error;
3226
3227         error = bridge_rtinstall_oncpu(brmsg->br_softc,
3228                                        brmsg->br_dst, brmsg->br_dst_if,
3229                                        brmsg->br_setflags, brmsg->br_flags,
3230                                        &brmsg->br_rtinfo);
3231         if (error) {
3232                 KKASSERT(mycpuid == 0 && brmsg->br_rtinfo == NULL);
3233                 lwkt_replymsg(&brmsg->base.lmsg, error);
3234                 return;
3235         } else if (brmsg->br_rtinfo == NULL) {
3236                 /* rtnode already exists for 'dst' */
3237                 KKASSERT(mycpuid == 0);
3238                 lwkt_replymsg(&brmsg->base.lmsg, 0);
3239                 return;
3240         }
3241         ifnet_forwardmsg(&brmsg->base.lmsg, mycpuid + 1);
3242 }
3243
3244 /*
3245  * bridge_rtupdate:
3246  *
3247  *      Add/Update a bridge routing entry.
3248  */
3249 static int
3250 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
3251                 struct ifnet *dst_if, uint8_t flags)
3252 {
3253         struct bridge_rtnode *brt;
3254
3255         /*
3256          * A route for this destination might already exist.  If so,
3257          * update it, otherwise create a new one.
3258          */
3259         if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
3260                 struct netmsg_brsaddr *brmsg;
3261
3262                 if (sc->sc_brtcnt >= sc->sc_brtmax)
3263                         return ENOSPC;
3264
3265                 brmsg = kmalloc(sizeof(*brmsg), M_LWKTMSG, M_WAITOK | M_NULLOK);
3266                 if (brmsg == NULL)
3267                         return ENOMEM;
3268
3269                 netmsg_init(&brmsg->base, NULL, &netisr_afree_rport,
3270                             0, bridge_rtinstall_handler);
3271                 memcpy(brmsg->br_dst, dst, ETHER_ADDR_LEN);
3272                 brmsg->br_dst_if = dst_if;
3273                 brmsg->br_flags = flags;
3274                 brmsg->br_setflags = 0;
3275                 brmsg->br_softc = sc;
3276                 brmsg->br_rtinfo = NULL;
3277
3278                 ifnet_sendmsg(&brmsg->base.lmsg, 0);
3279                 return 0;
3280         }
3281         bridge_rtinfo_update(brt->brt_info, dst_if, 0, flags,
3282                              sc->sc_brttimeout);
3283         return 0;
3284 }
3285
3286 static int
3287 bridge_rtsaddr(struct bridge_softc *sc, const uint8_t *dst,
3288                struct ifnet *dst_if, uint8_t flags)
3289 {
3290         struct netmsg_brsaddr brmsg;
3291
3292         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3293
3294         netmsg_init(&brmsg.base, NULL, &curthread->td_msgport,
3295                     0, bridge_rtinstall_handler);
3296         memcpy(brmsg.br_dst, dst, ETHER_ADDR_LEN);
3297         brmsg.br_dst_if = dst_if;
3298         brmsg.br_flags = flags;
3299         brmsg.br_setflags = 1;
3300         brmsg.br_softc = sc;
3301         brmsg.br_rtinfo = NULL;
3302
3303         return ifnet_domsg(&brmsg.base.lmsg, 0);
3304 }
3305
3306 /*
3307  * bridge_rtlookup:
3308  *
3309  *      Lookup the destination interface for an address.
3310  */
3311 static struct ifnet *
3312 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
3313 {
3314         struct bridge_rtnode *brt;
3315
3316         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
3317                 return NULL;
3318         return brt->brt_info->bri_ifp;
3319 }
3320
3321 static void
3322 bridge_rtreap_handler(netmsg_t msg)
3323 {
3324         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
3325         struct bridge_rtnode *brt, *nbrt;
3326
3327         LIST_FOREACH_MUTABLE(brt, &sc->sc_rtlists[mycpuid], brt_list, nbrt) {
3328                 if (brt->brt_info->bri_dead)
3329                         bridge_rtnode_destroy(sc, brt);
3330         }
3331         ifnet_forwardmsg(&msg->lmsg, mycpuid + 1);
3332 }
3333
3334 static void
3335 bridge_rtreap(struct bridge_softc *sc)
3336 {
3337         struct netmsg_base msg;
3338
3339         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3340
3341         netmsg_init(&msg, NULL, &curthread->td_msgport,
3342                     0, bridge_rtreap_handler);
3343         msg.lmsg.u.ms_resultp = sc;
3344
3345         ifnet_domsg(&msg.lmsg, 0);
3346 }
3347
3348 static void
3349 bridge_rtreap_async(struct bridge_softc *sc)
3350 {
3351         struct netmsg_base *msg;
3352
3353         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_WAITOK);
3354
3355         netmsg_init(msg, NULL, &netisr_afree_rport,
3356                     0, bridge_rtreap_handler);
3357         msg->lmsg.u.ms_resultp = sc;
3358
3359         ifnet_sendmsg(&msg->lmsg, 0);
3360 }
3361
3362 /*
3363  * bridge_rttrim:
3364  *
3365  *      Trim the routine table so that we have a number
3366  *      of routing entries less than or equal to the
3367  *      maximum number.
3368  */
3369 static void
3370 bridge_rttrim(struct bridge_softc *sc)
3371 {
3372         struct bridge_rtnode *brt;
3373         int dead;
3374
3375         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3376
3377         /* Make sure we actually need to do this. */
3378         if (sc->sc_brtcnt <= sc->sc_brtmax)
3379                 return;
3380
3381         /*
3382          * Find out how many rtnodes are dead
3383          */
3384         dead = bridge_rtage_finddead(sc);
3385         KKASSERT(dead <= sc->sc_brtcnt);
3386
3387         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
3388                 /* Enough dead rtnodes are found */
3389                 bridge_rtreap(sc);
3390                 return;
3391         }
3392
3393         /*
3394          * Kill some dynamic rtnodes to meet the brtmax
3395          */
3396         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
3397                 struct bridge_rtinfo *bri = brt->brt_info;
3398
3399                 if (bri->bri_dead) {
3400                         /*
3401                          * We have counted this rtnode in
3402                          * bridge_rtage_finddead()
3403                          */
3404                         continue;
3405                 }
3406
3407                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3408                         bri->bri_dead = 1;
3409                         ++dead;
3410                         KKASSERT(dead <= sc->sc_brtcnt);
3411
3412                         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
3413                                 /* Enough rtnodes are collected */
3414                                 break;
3415                         }
3416                 }
3417         }
3418         if (dead)
3419                 bridge_rtreap(sc);
3420 }
3421
3422 /*
3423  * bridge_timer:
3424  *
3425  *      Aging timer for the bridge.
3426  */
3427 static void
3428 bridge_timer(void *arg)
3429 {
3430         struct bridge_softc *sc = arg;
3431         struct netmsg_base *msg;
3432
3433         KKASSERT(mycpuid == BRIDGE_CFGCPU);
3434
3435         crit_enter();
3436
3437         if (callout_pending(&sc->sc_brcallout) ||
3438             !callout_active(&sc->sc_brcallout)) {
3439                 crit_exit();
3440                 return;
3441         }
3442         callout_deactivate(&sc->sc_brcallout);
3443
3444         msg = &sc->sc_brtimemsg;
3445         KKASSERT(msg->lmsg.ms_flags & MSGF_DONE);
3446         lwkt_sendmsg(BRIDGE_CFGPORT, &msg->lmsg);
3447
3448         crit_exit();
3449 }
3450
3451 static void
3452 bridge_timer_handler(netmsg_t msg)
3453 {
3454         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
3455
3456         KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
3457
3458         crit_enter();
3459         /* Reply ASAP */
3460         lwkt_replymsg(&msg->lmsg, 0);
3461         crit_exit();
3462
3463         bridge_rtage(sc);
3464         if (sc->sc_ifp->if_flags & IFF_RUNNING) {
3465                 callout_reset(&sc->sc_brcallout,
3466                     bridge_rtable_prune_period * hz, bridge_timer, sc);
3467         }
3468 }
3469
3470 static int
3471 bridge_rtage_finddead(struct bridge_softc *sc)
3472 {
3473         struct bridge_rtnode *brt;
3474         int dead = 0;
3475
3476         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
3477                 struct bridge_rtinfo *bri = brt->brt_info;
3478
3479                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
3480                     time_second >= bri->bri_expire) {
3481                         bri->bri_dead = 1;
3482                         ++dead;
3483                         KKASSERT(dead <= sc->sc_brtcnt);
3484                 }
3485         }
3486         return dead;
3487 }
3488
3489 /*
3490  * bridge_rtage:
3491  *
3492  *      Perform an aging cycle.
3493  */
3494 static void
3495 bridge_rtage(struct bridge_softc *sc)
3496 {
3497         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3498
3499         if (bridge_rtage_finddead(sc))
3500                 bridge_rtreap(sc);
3501 }
3502
3503 /*
3504  * bridge_rtflush:
3505  *
3506  *      Remove all dynamic addresses from the bridge.
3507  */
3508 static void
3509 bridge_rtflush(struct bridge_softc *sc, int bf)
3510 {
3511         struct bridge_rtnode *brt;
3512         int reap;
3513
3514         reap = 0;
3515         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
3516                 struct bridge_rtinfo *bri = brt->brt_info;
3517
3518                 if ((bf & IFBF_FLUSHALL) ||
3519                     (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3520                         bri->bri_dead = 1;
3521                         reap = 1;
3522                 }
3523         }
3524         if (reap) {
3525                 if (bf & IFBF_FLUSHSYNC)
3526                         bridge_rtreap(sc);
3527                 else
3528                         bridge_rtreap_async(sc);
3529         }
3530 }
3531
3532 /*
3533  * bridge_rtdaddr:
3534  *
3535  *      Remove an address from the table.
3536  */
3537 static int
3538 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
3539 {
3540         struct bridge_rtnode *brt;
3541
3542         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3543
3544         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
3545                 return (ENOENT);
3546
3547         /* TODO: add a cheaper delete operation */
3548         brt->brt_info->bri_dead = 1;
3549         bridge_rtreap(sc);
3550         return (0);
3551 }
3552
3553 /*
3554  * bridge_rtdelete:
3555  *
3556  *      Delete routes to a speicifc member interface.
3557  */
3558 void
3559 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int bf)
3560 {
3561         struct bridge_rtnode *brt;
3562         int reap;
3563
3564         reap = 0;
3565         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
3566                 struct bridge_rtinfo *bri = brt->brt_info;
3567
3568                 if (bri->bri_ifp == ifp &&
3569                     ((bf & IFBF_FLUSHALL) ||
3570                      (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
3571                         bri->bri_dead = 1;
3572                         reap = 1;
3573                 }
3574         }
3575         if (reap) {
3576                 if (bf & IFBF_FLUSHSYNC)
3577                         bridge_rtreap(sc);
3578                 else
3579                         bridge_rtreap_async(sc);
3580         }
3581 }
3582
3583 /*
3584  * bridge_rtable_init:
3585  *
3586  *      Initialize the route table for this bridge.
3587  */
3588 static void
3589 bridge_rtable_init(struct bridge_softc *sc)
3590 {
3591         int cpu;
3592
3593         /*
3594          * Initialize per-cpu hash tables
3595          */
3596         sc->sc_rthashs = kmalloc(sizeof(*sc->sc_rthashs) * ncpus,
3597                                  M_DEVBUF, M_WAITOK);
3598         for (cpu = 0; cpu < ncpus; ++cpu) {
3599                 int i;
3600
3601                 sc->sc_rthashs[cpu] =
3602                 kmalloc(sizeof(struct bridge_rtnode_head) * BRIDGE_RTHASH_SIZE,
3603                         M_DEVBUF, M_WAITOK);
3604
3605                 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3606                         LIST_INIT(&sc->sc_rthashs[cpu][i]);
3607         }
3608         sc->sc_rthash_key = karc4random();
3609
3610         /*
3611          * Initialize per-cpu lists
3612          */
3613         sc->sc_rtlists = kmalloc(sizeof(struct bridge_rtnode_head) * ncpus,
3614                                  M_DEVBUF, M_WAITOK);
3615         for (cpu = 0; cpu < ncpus; ++cpu)
3616                 LIST_INIT(&sc->sc_rtlists[cpu]);
3617 }
3618
3619 /*
3620  * bridge_rtable_fini:
3621  *
3622  *      Deconstruct the route table for this bridge.
3623  */
3624 static void
3625 bridge_rtable_fini(struct bridge_softc *sc)
3626 {
3627         int cpu;
3628
3629         /*
3630          * Free per-cpu hash tables
3631          */
3632         for (cpu = 0; cpu < ncpus; ++cpu)
3633                 kfree(sc->sc_rthashs[cpu], M_DEVBUF);
3634         kfree(sc->sc_rthashs, M_DEVBUF);
3635
3636         /*
3637          * Free per-cpu lists
3638          */
3639         kfree(sc->sc_rtlists, M_DEVBUF);
3640 }
3641
3642 /*
3643  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3644  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3645  */
3646 #define mix(a, b, c)                                                    \
3647 do {                                                                    \
3648         a -= b; a -= c; a ^= (c >> 13);                                 \
3649         b -= c; b -= a; b ^= (a << 8);                                  \
3650         c -= a; c -= b; c ^= (b >> 13);                                 \
3651         a -= b; a -= c; a ^= (c >> 12);                                 \
3652         b -= c; b -= a; b ^= (a << 16);                                 \
3653         c -= a; c -= b; c ^= (b >> 5);                                  \
3654         a -= b; a -= c; a ^= (c >> 3);                                  \
3655         b -= c; b -= a; b ^= (a << 10);                                 \
3656         c -= a; c -= b; c ^= (b >> 15);                                 \
3657 } while (/*CONSTCOND*/0)
3658
3659 static __inline uint32_t
3660 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3661 {
3662         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3663
3664         b += addr[5] << 8;
3665         b += addr[4];
3666         a += addr[3] << 24;
3667         a += addr[2] << 16;
3668         a += addr[1] << 8;
3669         a += addr[0];
3670
3671         mix(a, b, c);
3672
3673         return (c & BRIDGE_RTHASH_MASK);
3674 }
3675
3676 #undef mix
3677
3678 static int
3679 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3680 {
3681         int i, d;
3682
3683         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3684                 d = ((int)a[i]) - ((int)b[i]);
3685         }
3686
3687         return (d);
3688 }
3689
3690 /*
3691  * bridge_rtnode_lookup:
3692  *
3693  *      Look up a bridge route node for the specified destination.
3694  */
3695 static struct bridge_rtnode *
3696 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
3697 {
3698         struct bridge_rtnode *brt;
3699         uint32_t hash;
3700         int dir;
3701
3702         hash = bridge_rthash(sc, addr);
3703         LIST_FOREACH(brt, &sc->sc_rthashs[mycpuid][hash], brt_hash) {
3704                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3705                 if (dir == 0)
3706                         return (brt);
3707                 if (dir > 0)
3708                         return (NULL);
3709         }
3710
3711         return (NULL);
3712 }
3713
3714 /*
3715  * bridge_rtnode_insert:
3716  *
3717  *      Insert the specified bridge node into the route table.
3718  *      Caller has to make sure that rtnode does not exist.
3719  */
3720 static void
3721 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3722 {
3723         struct bridge_rtnode *lbrt;
3724         uint32_t hash;
3725         int dir;
3726
3727         hash = bridge_rthash(sc, brt->brt_addr);
3728
3729         lbrt = LIST_FIRST(&sc->sc_rthashs[mycpuid][hash]);
3730         if (lbrt == NULL) {
3731                 LIST_INSERT_HEAD(&sc->sc_rthashs[mycpuid][hash],
3732                                   brt, brt_hash);
3733                 goto out;
3734         }
3735
3736         do {
3737                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3738                 KASSERT(dir != 0, ("rtnode already exist"));
3739
3740                 if (dir > 0) {
3741                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3742                         goto out;
3743                 }
3744                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
3745                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3746                         goto out;
3747                 }
3748                 lbrt = LIST_NEXT(lbrt, brt_hash);
3749         } while (lbrt != NULL);
3750
3751         panic("no suitable position found for rtnode");
3752 out:
3753         LIST_INSERT_HEAD(&sc->sc_rtlists[mycpuid], brt, brt_list);
3754         if (mycpuid == 0) {
3755                 /*
3756                  * Update the brtcnt.
3757                  * We only need to do it once and we do it on CPU0.
3758                  */
3759                 sc->sc_brtcnt++;
3760         }
3761 }
3762
3763 /*
3764  * bridge_rtnode_destroy:
3765  *
3766  *      Destroy a bridge rtnode.
3767  */
3768 static void
3769 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3770 {
3771         LIST_REMOVE(brt, brt_hash);
3772         LIST_REMOVE(brt, brt_list);
3773
3774         if (mycpuid + 1 == ncpus) {
3775                 /* Free rtinfo associated with rtnode on the last cpu */
3776                 kfree(brt->brt_info, M_DEVBUF);
3777         }
3778         kfree(brt, M_DEVBUF);
3779
3780         if (mycpuid == 0) {
3781                 /* Update brtcnt only on CPU0 */
3782                 sc->sc_brtcnt--;
3783         }
3784 }
3785
3786 static __inline int
3787 bridge_post_pfil(struct mbuf *m)
3788 {
3789         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED)
3790                 return EOPNOTSUPP;
3791
3792         /* Not yet */
3793         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED)
3794                 return EOPNOTSUPP;
3795
3796         return 0;
3797 }
3798
3799 /*
3800  * Send bridge packets through pfil if they are one of the types pfil can deal
3801  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3802  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3803  * that interface.
3804  */
3805 static int
3806 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3807 {
3808         int snap, error, i, hlen;
3809         struct ether_header *eh1, eh2;
3810         struct ip *ip;
3811         struct llc llc1;
3812         u_int16_t ether_type;
3813
3814         snap = 0;
3815         error = -1;     /* Default error if not error == 0 */
3816
3817         if (pfil_bridge == 0 && pfil_member == 0)
3818                 return (0); /* filtering is disabled */
3819
3820         i = min((*mp)->m_pkthdr.len, max_protohdr);
3821         if ((*mp)->m_len < i) {
3822                 *mp = m_pullup(*mp, i);
3823                 if (*mp == NULL) {
3824                         kprintf("%s: m_pullup failed\n", __func__);
3825                         return (-1);
3826                 }
3827         }
3828
3829         eh1 = mtod(*mp, struct ether_header *);
3830         ether_type = ntohs(eh1->ether_type);
3831
3832         /*
3833          * Check for SNAP/LLC.
3834          */
3835         if (ether_type < ETHERMTU) {
3836                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3837
3838                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3839                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3840                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3841                     llc2->llc_control == LLC_UI) {
3842                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3843                         snap = 1;
3844                 }
3845         }
3846
3847         /*
3848          * If we're trying to filter bridge traffic, don't look at anything
3849          * other than IP and ARP traffic.  If the filter doesn't understand
3850          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3851          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3852          * but of course we don't have an AppleTalk filter to begin with.
3853          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3854          * ARP traffic.)
3855          */
3856         switch (ether_type) {
3857         case ETHERTYPE_ARP:
3858         case ETHERTYPE_REVARP:
3859                 return (0); /* Automatically pass */
3860
3861         case ETHERTYPE_IP:
3862 #ifdef INET6
3863         case ETHERTYPE_IPV6:
3864 #endif /* INET6 */
3865                 break;
3866
3867         default:
3868                 /*
3869                  * Check to see if the user wants to pass non-ip
3870                  * packets, these will not be checked by pfil(9)
3871                  * and passed unconditionally so the default is to drop.
3872                  */
3873                 if (pfil_onlyip)
3874                         goto bad;
3875         }
3876
3877         /* Strip off the Ethernet header and keep a copy. */
3878         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3879         m_adj(*mp, ETHER_HDR_LEN);
3880
3881         /* Strip off snap header, if present */
3882         if (snap) {
3883                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3884                 m_adj(*mp, sizeof(struct llc));
3885         }
3886
3887         /*
3888          * Check the IP header for alignment and errors
3889          */
3890         if (dir == PFIL_IN) {
3891                 switch (ether_type) {
3892                 case ETHERTYPE_IP:
3893                         error = bridge_ip_checkbasic(mp);
3894                         break;
3895 #ifdef INET6
3896                 case ETHERTYPE_IPV6:
3897                         error = bridge_ip6_checkbasic(mp);
3898                         break;
3899 #endif /* INET6 */
3900                 default:
3901                         error = 0;
3902                 }
3903                 if (error)
3904                         goto bad;
3905         }
3906
3907         error = 0;
3908
3909         /*
3910          * Run the packet through pfil
3911          */
3912         switch (ether_type) {
3913         case ETHERTYPE_IP:
3914                 /*
3915                  * before calling the firewall, swap fields the same as
3916                  * IP does. here we assume the header is contiguous
3917                  */
3918                 ip = mtod(*mp, struct ip *);
3919
3920                 ip->ip_len = ntohs(ip->ip_len);
3921                 ip->ip_off = ntohs(ip->ip_off);
3922
3923                 /*
3924                  * Run pfil on the member interface and the bridge, both can
3925                  * be skipped by clearing pfil_member or pfil_bridge.
3926                  *
3927                  * Keep the order:
3928                  *   in_if -> bridge_if -> out_if
3929                  */
3930                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) {
3931                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3932                         if (*mp == NULL || error != 0) /* filter may consume */
3933                                 break;
3934                         error = bridge_post_pfil(*mp);
3935                         if (error)
3936                                 break;
3937                 }
3938
3939                 if (pfil_member && ifp != NULL) {
3940                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp, dir);
3941                         if (*mp == NULL || error != 0) /* filter may consume */
3942                                 break;
3943                         error = bridge_post_pfil(*mp);
3944                         if (error)
3945                                 break;
3946                 }
3947
3948                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) {
3949                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3950                         if (*mp == NULL || error != 0) /* filter may consume */
3951                                 break;
3952                         error = bridge_post_pfil(*mp);
3953                         if (error)
3954                                 break;
3955                 }
3956
3957                 /* check if we need to fragment the packet */
3958                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3959                         i = (*mp)->m_pkthdr.len;
3960                         if (i > ifp->if_mtu) {
3961                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3962                                             &llc1);
3963                                 return (error);
3964                         }
3965                 }
3966
3967                 /* Recalculate the ip checksum and restore byte ordering */
3968                 ip = mtod(*mp, struct ip *);
3969                 hlen = ip->ip_hl << 2;
3970                 if (hlen < sizeof(struct ip))
3971                         goto bad;
3972                 if (hlen > (*mp)->m_len) {
3973                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3974                                 goto bad;
3975                         ip = mtod(*mp, struct ip *);
3976                         if (ip == NULL)
3977                                 goto bad;
3978                 }
3979                 ip->ip_len = htons(ip->ip_len);
3980                 ip->ip_off = htons(ip->ip_off);
3981                 ip->ip_sum = 0;
3982                 if (hlen == sizeof(struct ip))
3983                         ip->ip_sum = in_cksum_hdr(ip);
3984                 else
3985                         ip->ip_sum = in_cksum(*mp, hlen);
3986
3987                 break;
3988 #ifdef INET6
3989         case ETHERTYPE_IPV6:
3990                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3991                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3992                                         dir);
3993
3994                 if (*mp == NULL || error != 0) /* filter may consume */
3995                         break;
3996
3997                 if (pfil_member && ifp != NULL)
3998                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3999                                         dir);
4000
4001                 if (*mp == NULL || error != 0) /* filter may consume */
4002                         break;
4003
4004                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
4005                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
4006                                         dir);
4007                 break;
4008 #endif
4009         default:
4010                 error = 0;
4011                 break;
4012         }
4013
4014         if (*mp == NULL)
4015                 return (error);
4016         if (error != 0)
4017                 goto bad;
4018
4019         error = -1;
4020
4021         /*
4022          * Finally, put everything back the way it was and return
4023          */
4024         if (snap) {
4025                 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT);
4026                 if (*mp == NULL)
4027                         return (error);
4028                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
4029         }
4030
4031         M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT);
4032         if (*mp == NULL)
4033                 return (error);
4034         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
4035
4036         return (0);
4037
4038 bad:
4039         m_freem(*mp);
4040         *mp = NULL;
4041         return (error);
4042 }
4043
4044 /*
4045  * Perform basic checks on header size since
4046  * pfil assumes ip_input has already processed
4047  * it for it.  Cut-and-pasted from ip_input.c.
4048  * Given how simple the IPv6 version is,
4049  * does the IPv4 version really need to be
4050  * this complicated?
4051  *
4052  * XXX Should we update ipstat here, or not?
4053  * XXX Right now we update ipstat but not
4054  * XXX csum_counter.
4055  */
4056 static int
4057 bridge_ip_checkbasic(struct mbuf **mp)
4058 {
4059         struct mbuf *m = *mp;
4060         struct ip *ip;
4061         int len, hlen;
4062         u_short sum;
4063
4064         if (*mp == NULL)
4065                 return (-1);
4066 #if 0 /* notyet */
4067         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
4068                 if ((m = m_copyup(m, sizeof(struct ip),
4069                         (max_linkhdr + 3) & ~3)) == NULL) {
4070                         /* XXXJRT new stat, please */
4071                         ipstat.ips_toosmall++;
4072                         goto bad;
4073                 }
4074         } else
4075 #endif
4076 #ifndef __predict_false
4077 #define __predict_false(x) x
4078 #endif
4079          if (__predict_false(m->m_len < sizeof (struct ip))) {
4080                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
4081                         ipstat.ips_toosmall++;
4082                         goto bad;
4083                 }
4084         }
4085         ip = mtod(m, struct ip *);
4086         if (ip == NULL) goto bad;
4087
4088         if (ip->ip_v != IPVERSION) {
4089                 ipstat.ips_badvers++;
4090                 goto bad;
4091         }
4092         hlen = ip->ip_hl << 2;
4093         if (hlen < sizeof(struct ip)) { /* minimum header length */
4094                 ipstat.ips_badhlen++;
4095                 goto bad;
4096         }
4097         if (hlen > m->m_len) {
4098                 if ((m = m_pullup(m, hlen)) == NULL) {
4099                         ipstat.ips_badhlen++;
4100                         goto bad;
4101                 }
4102                 ip = mtod(m, struct ip *);
4103                 if (ip == NULL) goto bad;
4104         }
4105
4106         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
4107                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
4108         } else {
4109                 if (hlen == sizeof(struct ip)) {
4110                         sum = in_cksum_hdr(ip);
4111                 } else {
4112                         sum = in_cksum(m, hlen);
4113                 }
4114         }
4115         if (sum) {
4116                 ipstat.ips_badsum++;
4117                 goto bad;
4118         }
4119
4120         /* Retrieve the packet length. */
4121         len = ntohs(ip->ip_len);
4122
4123         /*
4124          * Check for additional length bogosity
4125          */
4126         if (len < hlen) {
4127                 ipstat.ips_badlen++;
4128                 goto bad;
4129         }
4130
4131         /*
4132          * Check that the amount of data in the buffers
4133          * is as at least much as the IP header would have us expect.
4134          * Drop packet if shorter than we expect.
4135          */
4136         if (m->m_pkthdr.len < len) {
4137                 ipstat.ips_tooshort++;
4138                 goto bad;
4139         }
4140
4141         /* Checks out, proceed */
4142         *mp = m;
4143         return (0);
4144
4145 bad:
4146         *mp = m;
4147         return (-1);
4148 }
4149
4150 #ifdef INET6
4151 /*
4152  * Same as above, but for IPv6.
4153  * Cut-and-pasted from ip6_input.c.
4154  * XXX Should we update ip6stat, or not?
4155  */
4156 static int
4157 bridge_ip6_checkbasic(struct mbuf **mp)
4158 {
4159         struct mbuf *m = *mp;
4160         struct ip6_hdr *ip6;
4161
4162         /*
4163          * If the IPv6 header is not aligned, slurp it up into a new
4164          * mbuf with space for link headers, in the event we forward
4165          * it.  Otherwise, if it is aligned, make sure the entire base
4166          * IPv6 header is in the first mbuf of the chain.
4167          */
4168 #if 0 /* notyet */
4169         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
4170                 struct ifnet *inifp = m->m_pkthdr.rcvif;
4171                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
4172                             (max_linkhdr + 3) & ~3)) == NULL) {
4173                         /* XXXJRT new stat, please */
4174                         ip6stat.ip6s_toosmall++;
4175                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
4176                         goto bad;
4177                 }
4178         } else
4179 #endif
4180         if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
4181                 struct ifnet *inifp = m->m_pkthdr.rcvif;
4182                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
4183                         ip6stat.ip6s_toosmall++;
4184                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
4185                         goto bad;
4186                 }
4187         }
4188
4189         ip6 = mtod(m, struct ip6_hdr *);
4190
4191         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
4192                 ip6stat.ip6s_badvers++;
4193                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
4194                 goto bad;
4195         }
4196
4197         /* Checks out, proceed */
4198         *mp = m;
4199         return (0);
4200
4201 bad:
4202         *mp = m;
4203         return (-1);
4204 }
4205 #endif /* INET6 */
4206
4207 /*
4208  * bridge_fragment:
4209  *
4210  *      Return a fragmented mbuf chain.
4211  */
4212 static int
4213 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
4214     int snap, struct llc *llc)
4215 {
4216         struct mbuf *m0;
4217         struct ip *ip;
4218         int error = -1;
4219
4220         if (m->m_len < sizeof(struct ip) &&
4221             (m = m_pullup(m, sizeof(struct ip))) == NULL)
4222                 goto out;
4223         ip = mtod(m, struct ip *);
4224
4225         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
4226                     CSUM_DELAY_IP);
4227         if (error)
4228                 goto out;
4229
4230         /* walk the chain and re-add the Ethernet header */
4231         for (m0 = m; m0; m0 = m0->m_nextpkt) {
4232                 if (error == 0) {
4233                         if (snap) {
4234                                 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT);
4235                                 if (m0 == NULL) {
4236                                         error = ENOBUFS;
4237                                         continue;
4238                                 }
4239                                 bcopy(llc, mtod(m0, caddr_t),
4240                                     sizeof(struct llc));
4241                         }
4242                         M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT);
4243                         if (m0 == NULL) {
4244                                 error = ENOBUFS;
4245                                 continue;
4246                         }
4247                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
4248                 } else
4249                         m_freem(m);
4250         }
4251
4252         if (error == 0)
4253                 ipstat.ips_fragmented++;
4254
4255         return (error);
4256
4257 out:
4258         if (m != NULL)
4259                 m_freem(m);
4260         return (error);
4261 }
4262
4263 static void
4264 bridge_enqueue_handler(netmsg_t msg)
4265 {
4266         struct netmsg_packet *nmp;
4267         struct ifnet *dst_ifp;
4268         struct mbuf *m;
4269
4270         nmp = &msg->packet;
4271         m = nmp->nm_packet;
4272         dst_ifp = nmp->base.lmsg.u.ms_resultp;
4273         mbuftrackid(m, 71);
4274
4275         bridge_handoff(dst_ifp->if_bridge, dst_ifp, m, 1);
4276 }
4277
4278 static void
4279 bridge_handoff(struct bridge_softc *sc, struct ifnet *dst_ifp,
4280                struct mbuf *m, int from_us)
4281 {
4282         struct mbuf *m0;
4283         struct ifnet *bifp;
4284
4285         bifp = sc->sc_ifp;
4286         mbuftrackid(m, 72);
4287
4288         /* We may be sending a fragment so traverse the mbuf */
4289         for (; m; m = m0) {
4290                 struct altq_pktattr pktattr;
4291
4292                 m0 = m->m_nextpkt;
4293                 m->m_nextpkt = NULL;
4294
4295                 /*
4296                  * If being sent from our host override ether_shost
4297                  * with the bridge MAC.  This is mandatory for ARP
4298                  * so things don't get confused.  In particular we
4299                  * don't want ARPs to get associated with link interfaces
4300                  * under the bridge which might or might not stay valid.
4301                  *
4302                  * Also override ether_shost when relaying a packet out
4303                  * the same interface it came in on, due to multi-homed
4304                  * addresses & default routes, otherwise switches will
4305                  * get very confused.
4306                  *
4307                  * Otherwise if we are in transparent mode.
4308                  */
4309                 if (from_us || m->m_pkthdr.rcvif == dst_ifp) {
4310                         m_copyback(m,
4311                                    offsetof(struct ether_header, ether_shost),
4312                                    ETHER_ADDR_LEN, IF_LLADDR(sc->sc_ifp));
4313                 } else if ((bifp->if_flags & IFF_LINK0) &&
4314                            (m->m_pkthdr.fw_flags & BRIDGE_MBUF_TAGGED)) {
4315                         m_copyback(m,
4316                                    offsetof(struct ether_header, ether_shost),
4317                                    ETHER_ADDR_LEN,
4318                                    m->m_pkthdr.br.ether.ether_shost);
4319                 } /* else retain shost */
4320
4321                 if (ifq_is_enabled(&dst_ifp->if_snd))
4322                         altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
4323
4324                 ifq_dispatch(dst_ifp, m, &pktattr);
4325         }
4326 }
4327
4328 static void
4329 bridge_control_dispatch(netmsg_t msg)
4330 {
4331         struct netmsg_brctl *bc_msg = (struct netmsg_brctl *)msg;
4332         struct ifnet *bifp = bc_msg->bc_sc->sc_ifp;
4333         int error;
4334
4335         ifnet_serialize_all(bifp);
4336         error = bc_msg->bc_func(bc_msg->bc_sc, bc_msg->bc_arg);
4337         ifnet_deserialize_all(bifp);
4338
4339         lwkt_replymsg(&bc_msg->base.lmsg, error);
4340 }
4341
4342 static int
4343 bridge_control(struct bridge_softc *sc, u_long cmd,
4344                bridge_ctl_t bc_func, void *bc_arg)
4345 {
4346         struct ifnet *bifp = sc->sc_ifp;
4347         struct netmsg_brctl bc_msg;
4348         int error;
4349
4350         ASSERT_IFNET_SERIALIZED_ALL(bifp);
4351
4352         bzero(&bc_msg, sizeof(bc_msg));
4353
4354         netmsg_init(&bc_msg.base, NULL, &curthread->td_msgport,
4355                     0, bridge_control_dispatch);
4356         bc_msg.bc_func = bc_func;
4357         bc_msg.bc_sc = sc;
4358         bc_msg.bc_arg = bc_arg;
4359
4360         ifnet_deserialize_all(bifp);
4361         error = lwkt_domsg(BRIDGE_CFGPORT, &bc_msg.base.lmsg, 0);
4362         ifnet_serialize_all(bifp);
4363         return error;
4364 }
4365
4366 static void
4367 bridge_add_bif_handler(netmsg_t msg)
4368 {
4369         struct netmsg_braddbif *amsg = (struct netmsg_braddbif *)msg;
4370         struct bridge_softc *sc;
4371         struct bridge_iflist *bif;
4372
4373         sc = amsg->br_softc;
4374
4375         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
4376         bif->bif_ifp = amsg->br_bif_ifp;
4377         bif->bif_onlist = 1;
4378         bif->bif_info = amsg->br_bif_info;
4379
4380         /*
4381          * runs through bif_info
4382          */
4383         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
4384
4385         TAILQ_INSERT_HEAD(&sc->sc_iflists[mycpuid], bif, bif_next);
4386
4387         ifnet_forwardmsg(&amsg->base.lmsg, mycpuid + 1);
4388 }
4389
4390 static void
4391 bridge_add_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
4392                struct ifnet *ifp)
4393 {
4394         struct netmsg_braddbif amsg;
4395
4396         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
4397
4398         netmsg_init(&amsg.base, NULL, &curthread->td_msgport,
4399                     0, bridge_add_bif_handler);
4400         amsg.br_softc = sc;
4401         amsg.br_bif_info = bif_info;
4402         amsg.br_bif_ifp = ifp;
4403
4404         ifnet_domsg(&amsg.base.lmsg, 0);
4405 }
4406
4407 static void
4408 bridge_del_bif_handler(netmsg_t msg)
4409 {
4410         struct netmsg_brdelbif *dmsg = (struct netmsg_brdelbif *)msg;
4411         struct bridge_softc *sc;
4412         struct bridge_iflist *bif;
4413
4414         sc = dmsg->br_softc;
4415
4416         /*
4417          * Locate the bif associated with the br_bif_info
4418          * on the current CPU
4419          */
4420         bif = bridge_lookup_member_ifinfo(sc, dmsg->br_bif_info);
4421         KKASSERT(bif != NULL && bif->bif_onlist);
4422
4423         /* Remove the bif from the current CPU's iflist */
4424         bif->bif_onlist = 0;
4425         TAILQ_REMOVE(dmsg->br_bif_list, bif, bif_next);
4426
4427         /* Save the removed bif for later freeing */
4428         TAILQ_INSERT_HEAD(dmsg->br_bif_list, bif, bif_next);
4429
4430         ifnet_forwardmsg(&dmsg->base.lmsg, mycpuid + 1);
4431 }
4432
4433 static void
4434 bridge_del_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
4435                struct bridge_iflist_head *saved_bifs)
4436 {
4437         struct netmsg_brdelbif dmsg;
4438
4439         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
4440
4441         netmsg_init(&dmsg.base, NULL, &curthread->td_msgport,
4442                     0, bridge_del_bif_handler);
4443         dmsg.br_softc = sc;
4444         dmsg.br_bif_info = bif_info;
4445         dmsg.br_bif_list = saved_bifs;
4446
4447         ifnet_domsg(&dmsg.base.lmsg, 0);
4448 }