sys/net/bridge/if_bridge.c

   1 /*
   2  * Copyright 2001 Wasabi Systems, Inc.
   3  * All rights reserved.
   4  *
   5  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  * 3. All advertising materials mentioning features or use of this software
  16  *    must display the following acknowledgement:
  17  *      This product includes software developed for the NetBSD Project by
  18  *      Wasabi Systems, Inc.
  19  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  20  *    or promote products derived from this software without specific prior
  21  *    written permission.
  22  *
  23  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  25  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  26  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  27  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  33  * POSSIBILITY OF SUCH DAMAGE.
  34  */
  35
  36 /*
  37  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  38  * All rights reserved.
  39  *
  40  * Redistribution and use in source and binary forms, with or without
  41  * modification, are permitted provided that the following conditions
  42  * are met:
  43  * 1. Redistributions of source code must retain the above copyright
  44  *    notice, this list of conditions and the following disclaimer.
  45  * 2. Redistributions in binary form must reproduce the above copyright
  46  *    notice, this list of conditions and the following disclaimer in the
  47  *    documentation and/or other materials provided with the distribution.
  48  * 3. All advertising materials mentioning features or use of this software
  49  *    must display the following acknowledgement:
  50  *      This product includes software developed by Jason L. Wright
  51  * 4. The name of the author may not be used to endorse or promote products
  52  *    derived from this software without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  55  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  56  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  57  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  58  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  59  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  60  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  62  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  63  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  64  * POSSIBILITY OF SUCH DAMAGE.
  65  *
  66  * $OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp $
  67  * $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $
  68  * $FreeBSD: src/sys/net/if_bridge.c,v 1.26 2005/10/13 23:05:55 thompsa Exp $
  69  * $DragonFly: src/sys/net/bridge/if_bridge.c,v 1.60 2008/11/26 12:49:43 sephe Exp $
  70  */
  71
  72 /*
  73  * Network interface bridge support.
  74  *
  75  * TODO:
  76  *
  77  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  78  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  79  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
  80  *        consider heterogenous bridges).
  81  *
  82  *
  83  * Bridge's route information is duplicated to each CPUs:
  84  *
  85  *      CPU0          CPU1          CPU2          CPU3
  86  * +-----------+ +-----------+ +-----------+ +-----------+
  87  * |  rtnode   | |  rtnode   | |  rtnode   | |  rtnode   |
  88  * |           | |           | |           | |           |
  89  * | dst eaddr | | dst eaddr | | dst eaddr | | dst eaddr |
  90  * +-----------+ +-----------+ +-----------+ +-----------+
  91  *       |         |                     |         |
  92  *       |         |                     |         |
  93  *       |         |     +----------+    |         |
  94  *       |         |     |  rtinfo  |    |         |
  95  *       |         +---->|          |<---+         |
  96  *       |               |  flags   |              |
  97  *       +-------------->|  timeout |<-------------+
  98  *                       |  dst_ifp |
  99  *                       +----------+
 100  *
 101  * We choose to put timeout and dst_ifp into shared part, so updating
 102  * them will be cheaper than using message forwarding.  Also there is
 103  * not need to use spinlock to protect the updating: timeout and dst_ifp
 104  * is not related and specific field's updating order has no importance.
 105  * The cache pollution by the share part should not be heavy: in a stable
 106  * setup, dst_ifp probably will be not changed in rtnode's life time,
 107  * while timeout is refreshed once per second; most of the time, timeout
 108  * and dst_ifp are read-only accessed.
 109  *
 110  *
 111  * Bridge route information installation on bridge_input path:
 112  *
 113  *      CPU0           CPU1         CPU2          CPU3
 114  *
 115  *                               tcp_thread2
 116  *                                    |
 117  *                                alloc nmsg
 118  *                    snd nmsg        |
 119  *                    w/o rtinfo      |
 120  *      ifnet0<-----------------------+
 121  *        |                           :
 122  *    lookup dst                      :
 123  *   rtnode exists?(Y)free nmsg       :
 124  *        |(N)                        :
 125  *        |
 126  *  alloc rtinfo
 127  *  alloc rtnode
 128  * install rtnode
 129  *        |
 130  *        +---------->ifnet1
 131  *        : fwd nmsg    |
 132  *        : w/ rtinfo   |
 133  *        :             |
 134  *        :             |
 135  *                 alloc rtnode
 136  *               (w/ nmsg's rtinfo)
 137  *                install rtnode
 138  *                      |
 139  *                      +---------->ifnet2
 140  *                      : fwd nmsg    |
 141  *                      : w/ rtinfo   |
 142  *                      :             |
 143  *                      :         same as ifnet1
 144  *                                    |
 145  *                                    +---------->ifnet3
 146  *                                    : fwd nmsg    |
 147  *                                    : w/ rtinfo   |
 148  *                                    :             |
 149  *                                    :         same as ifnet1
 150  *                                               free nmsg
 151  *                                                  :
 152  *                                                  :
 153  *
 154  * The netmsgs forwarded between protocol threads and ifnet threads are
 155  * allocated with (M_WAITOK|M_NULLOK), so it will not fail under most
 156  * cases (route information is too precious to be not installed :).
 157  * Since multiple threads may try to install route information for the
 158  * same dst eaddr, we look up route information in ifnet0.  However, this
 159  * looking up only need to be performed on ifnet0, which is the start
 160  * point of the route information installation process.
 161  *
 162  *
 163  * Bridge route information deleting/flushing:
 164  *
 165  *  CPU0            CPU1             CPU2             CPU3
 166  *
 167  * netisr0
 168  *   |
 169  * find suitable rtnodes,
 170  * mark their rtinfo dead
 171  *   |
 172  *   | domsg <------------------------------------------+
 173  *   |                                                  | replymsg
 174  *   |                                                  |
 175  *   V     fwdmsg           fwdmsg           fwdmsg     |
 176  * ifnet0 --------> ifnet1 --------> ifnet2 --------> ifnet3
 177  * delete rtnodes   delete rtnodes   delete rtnodes   delete rtnodes
 178  * w/ dead rtinfo   w/ dead rtinfo   w/ dead rtinfo   w/ dead rtinfo
 179  *                                                    free dead rtinfos
 180  *
 181  * All deleting/flushing operations are serialized by netisr0, so each
 182  * operation only reaps the route information marked dead by itself.
 183  *
 184  *
 185  * Bridge route information adding/deleting/flushing:
 186  * Since all operation is serialized by the fixed message flow between
 187  * ifnet threads, it is not possible to create corrupted per-cpu route
 188  * information.
 189  *
 190  *
 191  *
 192  * Percpu member interface list iteration with blocking operation:
 193  * Since one bridge could only delete one member interface at a time and
 194  * the deleted member interface is not freed after netmsg_service_sync(),
 195  * following way is used to make sure that even if the certain member
 196  * interface is ripped from the percpu list during the blocking operation,
 197  * the iteration still could keep going:
 198  *
 199  * LIST_FOREACH_MUTABLE(bif, sc->sc_iflists[mycpuid], bif_next, nbif) {
 200  *     blocking operation;
 201  *     blocking operation;
 202  *     ...
 203  *     ...
 204  *     if (nbif != NULL && !nbif->bif_onlist) {
 205  *         KKASSERT(bif->bif_onlist);
 206  *         nbif = LIST_NEXT(bif, bif_next);
 207  *     }
 208  * }
 209  *
 210  * As mentioned above only one member interface could be unlinked from the
 211  * percpu member interface list, so either bif or nbif may be not on the list,
 212  * but _not_ both.  To keep the list iteration, we don't care about bif, but
 213  * only nbif.  Since removed member interface will only be freed after we
 214  * finish our work, it is safe to access any field in an unlinked bif (here
 215  * bif_onlist).  If nbif is no longer on the list, then bif must be on the
 216  * list, so we change nbif to the next element of bif and keep going.
 217  */
 218
 219 #include "opt_inet.h"
 220 #include "opt_inet6.h"
 221
 222 #include <sys/param.h>
 223 #include <sys/mbuf.h>
 224 #include <sys/malloc.h>
 225 #include <sys/protosw.h>
 226 #include <sys/systm.h>
 227 #include <sys/time.h>
 228 #include <sys/socket.h> /* for net/if.h */
 229 #include <sys/sockio.h>
 230 #include <sys/ctype.h>  /* string functions */
 231 #include <sys/kernel.h>
 232 #include <sys/random.h>
 233 #include <sys/sysctl.h>
 234 #include <sys/module.h>
 235 #include <sys/proc.h>
 236 #include <sys/priv.h>
 237 #include <sys/lock.h>
 238 #include <sys/thread.h>
 239 #include <sys/thread2.h>
 240 #include <sys/mpipe.h>
 241
 242 #include <net/bpf.h>
 243 #include <net/if.h>
 244 #include <net/if_dl.h>
 245 #include <net/if_types.h>
 246 #include <net/if_var.h>
 247 #include <net/pfil.h>
 248 #include <net/ifq_var.h>
 249 #include <net/if_clone.h>
 250
 251 #include <netinet/in.h> /* for struct arpcom */
 252 #include <netinet/in_systm.h>
 253 #include <netinet/in_var.h>
 254 #include <netinet/ip.h>
 255 #include <netinet/ip_var.h>
 256 #ifdef INET6
 257 #include <netinet/ip6.h>
 258 #include <netinet6/ip6_var.h>
 259 #endif
 260 #include <netinet/if_ether.h> /* for struct arpcom */
 261 #include <net/bridge/if_bridgevar.h>
 262 #include <net/if_llc.h>
 263 #include <net/netmsg2.h>
 264
 265 #include <net/route.h>
 266 #include <sys/in_cksum.h>
 267
 268 /*
 269  * Size of the route hash table.  Must be a power of two.
 270  */
 271 #ifndef BRIDGE_RTHASH_SIZE
 272 #define BRIDGE_RTHASH_SIZE              1024
 273 #endif
 274
 275 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 276
 277 /*
 278  * Maximum number of addresses to cache.
 279  */
 280 #ifndef BRIDGE_RTABLE_MAX
 281 #define BRIDGE_RTABLE_MAX               100
 282 #endif
 283
 284 /*
 285  * Spanning tree defaults.
 286  */
 287 #define BSTP_DEFAULT_MAX_AGE            (20 * 256)
 288 #define BSTP_DEFAULT_HELLO_TIME         (2 * 256)
 289 #define BSTP_DEFAULT_FORWARD_DELAY      (15 * 256)
 290 #define BSTP_DEFAULT_HOLD_TIME          (1 * 256)
 291 #define BSTP_DEFAULT_BRIDGE_PRIORITY    0x8000
 292 #define BSTP_DEFAULT_PORT_PRIORITY      0x80
 293 #define BSTP_DEFAULT_PATH_COST          55
 294
 295 /*
 296  * Timeout (in seconds) for entries learned dynamically.
 297  */
 298 #ifndef BRIDGE_RTABLE_TIMEOUT
 299 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 300 #endif
 301
 302 /*
 303  * Number of seconds between walks of the route list.
 304  */
 305 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 306 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 307 #endif
 308
 309 /*
 310  * List of capabilities to mask on the member interface.
 311  */
 312 #define BRIDGE_IFCAPS_MASK              IFCAP_TXCSUM
 313
 314 typedef int     (*bridge_ctl_t)(struct bridge_softc *, void *);
 315
 316 struct netmsg_brctl {
 317         struct netmsg_base      base;
 318         bridge_ctl_t            bc_func;
 319         struct bridge_softc     *bc_sc;
 320         void                    *bc_arg;
 321 };
 322
 323 struct netmsg_brsaddr {
 324         struct netmsg_base      base;
 325         struct bridge_softc     *br_softc;
 326         struct ifnet            *br_dst_if;
 327         struct bridge_rtinfo    *br_rtinfo;
 328         int                     br_setflags;
 329         uint8_t                 br_dst[ETHER_ADDR_LEN];
 330         uint8_t                 br_flags;
 331 };
 332
 333 struct netmsg_braddbif {
 334         struct netmsg_base      base;
 335         struct bridge_softc     *br_softc;
 336         struct bridge_ifinfo    *br_bif_info;
 337         struct ifnet            *br_bif_ifp;
 338 };
 339
 340 struct netmsg_brdelbif {
 341         struct netmsg_base      base;
 342         struct bridge_softc     *br_softc;
 343         struct bridge_ifinfo    *br_bif_info;
 344         struct bridge_iflist_head *br_bif_list;
 345 };
 346
 347 struct netmsg_brsflags {
 348         struct netmsg_base      base;
 349         struct bridge_softc     *br_softc;
 350         struct bridge_ifinfo    *br_bif_info;
 351         uint32_t                br_bif_flags;
 352 };
 353
 354 eventhandler_tag        bridge_detach_cookie = NULL;
 355
 356 extern  struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
 357 extern  int (*bridge_output_p)(struct ifnet *, struct mbuf *);
 358 extern  void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
 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 void     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 void     bridge_stop(struct ifnet *);
 370 static void     bridge_start(struct ifnet *);
 371 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 372 static int      bridge_output(struct ifnet *, struct mbuf *);
 373
 374 static void     bridge_forward(struct bridge_softc *, struct mbuf *m);
 375
 376 static void     bridge_timer_handler(netmsg_t);
 377 static void     bridge_timer(void *);
 378
 379 static void     bridge_start_bcast(struct bridge_softc *, struct mbuf *);
 380 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 381                     struct mbuf *);
 382 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 383
 384 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 385                     struct ifnet *, uint8_t);
 386 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
 387 static void     bridge_rtreap(struct bridge_softc *);
 388 static void     bridge_rtreap_async(struct bridge_softc *);
 389 static void     bridge_rttrim(struct bridge_softc *);
 390 static int      bridge_rtage_finddead(struct bridge_softc *);
 391 static void     bridge_rtage(struct bridge_softc *);
 392 static void     bridge_rtflush(struct bridge_softc *, int);
 393 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
 394 static int      bridge_rtsaddr(struct bridge_softc *, const uint8_t *,
 395                     struct ifnet *, uint8_t);
 396 static void     bridge_rtmsg_sync(struct bridge_softc *sc);
 397 static void     bridge_rtreap_handler(netmsg_t);
 398 static void     bridge_rtinstall_handler(netmsg_t);
 399 static int      bridge_rtinstall_oncpu(struct bridge_softc *, const uint8_t *,
 400                     struct ifnet *, int, uint8_t, struct bridge_rtinfo **);
 401
 402 static void     bridge_rtable_init(struct bridge_softc *);
 403 static void     bridge_rtable_fini(struct bridge_softc *);
 404
 405 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 406 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 407                     const uint8_t *);
 408 static void     bridge_rtnode_insert(struct bridge_softc *,
 409                     struct bridge_rtnode *);
 410 static void     bridge_rtnode_destroy(struct bridge_softc *,
 411                     struct bridge_rtnode *);
 412
 413 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 414                     const char *name);
 415 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 416                     struct ifnet *ifp);
 417 static struct bridge_iflist *bridge_lookup_member_ifinfo(struct bridge_softc *,
 418                     struct bridge_ifinfo *);
 419 static void     bridge_delete_member(struct bridge_softc *,
 420                     struct bridge_iflist *, int);
 421 static void     bridge_delete_span(struct bridge_softc *,
 422                     struct bridge_iflist *);
 423
 424 static int      bridge_control(struct bridge_softc *, u_long,
 425                                bridge_ctl_t, void *);
 426 static int      bridge_ioctl_init(struct bridge_softc *, void *);
 427 static int      bridge_ioctl_stop(struct bridge_softc *, void *);
 428 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 429 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 430 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 431 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 432 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 433 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 434 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 435 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 436 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 437 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 438 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 439 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 440 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 441 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 442 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 443 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 444 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 445 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 446 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 447 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 448 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 449 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 450 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 451 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 452 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 453 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 454                     int);
 455 static int      bridge_ip_checkbasic(struct mbuf **mp);
 456 #ifdef INET6
 457 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 458 #endif /* INET6 */
 459 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 460                     struct ether_header *, int, struct llc *);
 461 static void     bridge_enqueue_handler(netmsg_t);
 462 static void     bridge_handoff(struct ifnet *, struct mbuf *);
 463
 464 static void     bridge_del_bif_handler(netmsg_t);
 465 static void     bridge_add_bif_handler(netmsg_t);
 466 static void     bridge_set_bifflags_handler(netmsg_t);
 467 static void     bridge_del_bif(struct bridge_softc *, struct bridge_ifinfo *,
 468                     struct bridge_iflist_head *);
 469 static void     bridge_add_bif(struct bridge_softc *, struct bridge_ifinfo *,
 470                     struct ifnet *);
 471 static void     bridge_set_bifflags(struct bridge_softc *,
 472                     struct bridge_ifinfo *, uint32_t);
 473
 474 SYSCTL_DECL(_net_link);
 475 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 476
 477 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 478 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 479 static int pfil_member = 1; /* run pfil hooks on the member interface */
 480 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 481     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 482 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 483     &pfil_bridge, 0, "Packet filter on the bridge interface");
 484 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 485     &pfil_member, 0, "Packet filter on the member interface");
 486
 487 struct bridge_control_arg {
 488         union {
 489                 struct ifbreq ifbreq;
 490                 struct ifbifconf ifbifconf;
 491                 struct ifbareq ifbareq;
 492                 struct ifbaconf ifbaconf;
 493                 struct ifbrparam ifbrparam;
 494         } bca_u;
 495         int     bca_len;
 496         void    *bca_uptr;
 497         void    *bca_kptr;
 498 };
 499
 500 struct bridge_control {
 501         bridge_ctl_t    bc_func;
 502         int             bc_argsize;
 503         int             bc_flags;
 504 };
 505
 506 #define BC_F_COPYIN             0x01    /* copy arguments in */
 507 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 508 #define BC_F_SUSER              0x04    /* do super-user check */
 509
 510 const struct bridge_control bridge_control_table[] = {
 511         { bridge_ioctl_add,             sizeof(struct ifbreq),
 512           BC_F_COPYIN|BC_F_SUSER },
 513         { bridge_ioctl_del,             sizeof(struct ifbreq),
 514           BC_F_COPYIN|BC_F_SUSER },
 515
 516         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 517           BC_F_COPYIN|BC_F_COPYOUT },
 518         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 519           BC_F_COPYIN|BC_F_SUSER },
 520
 521         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 522           BC_F_COPYIN|BC_F_SUSER },
 523         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 524           BC_F_COPYOUT },
 525
 526         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 527           BC_F_COPYIN|BC_F_COPYOUT },
 528         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 529           BC_F_COPYIN|BC_F_COPYOUT },
 530
 531         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 532           BC_F_COPYIN|BC_F_SUSER },
 533
 534         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 535           BC_F_COPYIN|BC_F_SUSER },
 536         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 537           BC_F_COPYOUT },
 538
 539         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 540           BC_F_COPYIN|BC_F_SUSER },
 541
 542         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 543           BC_F_COPYIN|BC_F_SUSER },
 544
 545         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 546           BC_F_COPYOUT },
 547         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 548           BC_F_COPYIN|BC_F_SUSER },
 549
 550         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 551           BC_F_COPYOUT },
 552         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 553           BC_F_COPYIN|BC_F_SUSER },
 554
 555         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 556           BC_F_COPYOUT },
 557         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 558           BC_F_COPYIN|BC_F_SUSER },
 559
 560         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 561           BC_F_COPYOUT },
 562         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 563           BC_F_COPYIN|BC_F_SUSER },
 564
 565         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 566           BC_F_COPYIN|BC_F_SUSER },
 567
 568         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 569           BC_F_COPYIN|BC_F_SUSER },
 570
 571         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 572           BC_F_COPYIN|BC_F_SUSER },
 573         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 574           BC_F_COPYIN|BC_F_SUSER },
 575 };
 576 static const int bridge_control_table_size =
 577     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 578
 579 LIST_HEAD(, bridge_softc) bridge_list;
 580
 581 struct if_clone bridge_cloner = IF_CLONE_INITIALIZER("bridge",
 582                                 bridge_clone_create,
 583                                 bridge_clone_destroy, 0, IF_MAXUNIT);
 584
 585 static int
 586 bridge_modevent(module_t mod, int type, void *data)
 587 {
 588         switch (type) {
 589         case MOD_LOAD:
 590                 LIST_INIT(&bridge_list);
 591                 if_clone_attach(&bridge_cloner);
 592                 bridge_input_p = bridge_input;
 593                 bridge_output_p = bridge_output;
 594                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 595                     ifnet_detach_event, bridge_ifdetach, NULL,
 596                     EVENTHANDLER_PRI_ANY);
 597 #if notyet
 598                 bstp_linkstate_p = bstp_linkstate;
 599 #endif
 600                 break;
 601         case MOD_UNLOAD:
 602                 if (!LIST_EMPTY(&bridge_list))
 603                         return (EBUSY);
 604                 EVENTHANDLER_DEREGISTER(ifnet_detach_event,
 605                     bridge_detach_cookie);
 606                 if_clone_detach(&bridge_cloner);
 607                 bridge_input_p = NULL;
 608                 bridge_output_p = NULL;
 609 #if notyet
 610                 bstp_linkstate_p = NULL;
 611 #endif
 612                 break;
 613         default:
 614                 return (EOPNOTSUPP);
 615         }
 616         return (0);
 617 }
 618
 619 static moduledata_t bridge_mod = {
 620         "if_bridge",
 621         bridge_modevent,
 622         0
 623 };
 624
 625 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 626
 627
 628 /*
 629  * bridge_clone_create:
 630  *
 631  *      Create a new bridge instance.
 632  */
 633 static int
 634 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused)
 635 {
 636         struct bridge_softc *sc;
 637         struct ifnet *ifp;
 638         u_char eaddr[6];
 639         int cpu, rnd;
 640
 641         sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
 642         ifp = sc->sc_ifp = &sc->sc_if;
 643
 644         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 645         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 646         sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
 647         sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
 648         sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
 649         sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
 650         sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
 651
 652         /* Initialize our routing table. */
 653         bridge_rtable_init(sc);
 654
 655         callout_init(&sc->sc_brcallout);
 656         netmsg_init(&sc->sc_brtimemsg, NULL, &netisr_adone_rport,
 657                     MSGF_DROPABLE, bridge_timer_handler);
 658         sc->sc_brtimemsg.lmsg.u.ms_resultp = sc;
 659
 660         callout_init(&sc->sc_bstpcallout);
 661         netmsg_init(&sc->sc_bstptimemsg, NULL, &netisr_adone_rport,
 662                     MSGF_DROPABLE, bstp_tick_handler);
 663         sc->sc_bstptimemsg.lmsg.u.ms_resultp = sc;
 664
 665         /* Initialize per-cpu member iface lists */
 666         sc->sc_iflists = kmalloc(sizeof(*sc->sc_iflists) * ncpus,
 667                                  M_DEVBUF, M_WAITOK);
 668         for (cpu = 0; cpu < ncpus; ++cpu)
 669                 LIST_INIT(&sc->sc_iflists[cpu]);
 670
 671         LIST_INIT(&sc->sc_spanlist);
 672
 673         ifp->if_softc = sc;
 674         if_initname(ifp, ifc->ifc_name, unit);
 675         ifp->if_mtu = ETHERMTU;
 676         ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
 677         ifp->if_ioctl = bridge_ioctl;
 678         ifp->if_start = bridge_start;
 679         ifp->if_init = bridge_init;
 680         ifp->if_type = IFT_BRIDGE;
 681         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
 682         ifq_set_ready(&ifp->if_snd);
 683         ifp->if_hdrlen = ETHER_HDR_LEN;
 684
 685         /*
 686          * Generate a random ethernet address and use the private AC:DE:48
 687          * OUI code.
 688          */
 689         rnd = karc4random();
 690         bcopy(&rnd, &eaddr[0], 4); /* ETHER_ADDR_LEN == 6 */
 691         rnd = karc4random();
 692         bcopy(&rnd, &eaddr[2], 4); /* ETHER_ADDR_LEN == 6 */
 693
 694         eaddr[0] &= ~1; /* clear multicast bit */
 695         eaddr[0] |= 2;  /* set the LAA bit */
 696
 697         ether_ifattach(ifp, eaddr, NULL);
 698         /* Now undo some of the damage... */
 699         ifp->if_baudrate = 0;
 700         ifp->if_type = IFT_BRIDGE;
 701
 702         crit_enter();   /* XXX MP */
 703         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 704         crit_exit();
 705
 706         return (0);
 707 }
 708
 709 static void
 710 bridge_delete_dispatch(netmsg_t msg)
 711 {
 712         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
 713         struct ifnet *bifp = sc->sc_ifp;
 714         struct bridge_iflist *bif;
 715
 716         ifnet_serialize_all(bifp);
 717
 718         while ((bif = LIST_FIRST(&sc->sc_iflists[mycpuid])) != NULL)
 719                 bridge_delete_member(sc, bif, 0);
 720
 721         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL)
 722                 bridge_delete_span(sc, bif);
 723
 724         ifnet_deserialize_all(bifp);
 725
 726         lwkt_replymsg(&msg->lmsg, 0);
 727 }
 728
 729 /*
 730  * bridge_clone_destroy:
 731  *
 732  *      Destroy a bridge instance.
 733  */
 734 static void
 735 bridge_clone_destroy(struct ifnet *ifp)
 736 {
 737         struct bridge_softc *sc = ifp->if_softc;
 738         struct netmsg_base msg;
 739
 740         ifnet_serialize_all(ifp);
 741
 742         bridge_stop(ifp);
 743         ifp->if_flags &= ~IFF_UP;
 744
 745         ifnet_deserialize_all(ifp);
 746
 747         netmsg_init(&msg, NULL, &curthread->td_msgport,
 748                     0, bridge_delete_dispatch);
 749         msg.lmsg.u.ms_resultp = sc;
 750         lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0);
 751
 752         crit_enter();   /* XXX MP */
 753         LIST_REMOVE(sc, sc_list);
 754         crit_exit();
 755
 756         ether_ifdetach(ifp);
 757
 758         /* Tear down the routing table. */
 759         bridge_rtable_fini(sc);
 760
 761         /* Free per-cpu member iface lists */
 762         kfree(sc->sc_iflists, M_DEVBUF);
 763
 764         kfree(sc, M_DEVBUF);
 765 }
 766
 767 /*
 768  * bridge_ioctl:
 769  *
 770  *      Handle a control request from the operator.
 771  */
 772 static int
 773 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 774 {
 775         struct bridge_softc *sc = ifp->if_softc;
 776         struct bridge_control_arg args;
 777         struct ifdrv *ifd = (struct ifdrv *) data;
 778         const struct bridge_control *bc;
 779         int error = 0;
 780
 781         ASSERT_IFNET_SERIALIZED_ALL(ifp);
 782
 783         switch (cmd) {
 784         case SIOCADDMULTI:
 785         case SIOCDELMULTI:
 786                 break;
 787
 788         case SIOCGDRVSPEC:
 789         case SIOCSDRVSPEC:
 790                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 791                         error = EINVAL;
 792                         break;
 793                 }
 794                 bc = &bridge_control_table[ifd->ifd_cmd];
 795
 796                 if (cmd == SIOCGDRVSPEC &&
 797                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 798                         error = EINVAL;
 799                         break;
 800                 } else if (cmd == SIOCSDRVSPEC &&
 801                            (bc->bc_flags & BC_F_COPYOUT)) {
 802                         error = EINVAL;
 803                         break;
 804                 }
 805
 806                 if (bc->bc_flags & BC_F_SUSER) {
 807                         error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY);
 808                         if (error)
 809                                 break;
 810                 }
 811
 812                 if (ifd->ifd_len != bc->bc_argsize ||
 813                     ifd->ifd_len > sizeof(args.bca_u)) {
 814                         error = EINVAL;
 815                         break;
 816                 }
 817
 818                 memset(&args, 0, sizeof(args));
 819                 if (bc->bc_flags & BC_F_COPYIN) {
 820                         error = copyin(ifd->ifd_data, &args.bca_u,
 821                                        ifd->ifd_len);
 822                         if (error)
 823                                 break;
 824                 }
 825
 826                 error = bridge_control(sc, cmd, bc->bc_func, &args);
 827                 if (error) {
 828                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 829                         break;
 830                 }
 831
 832                 if (bc->bc_flags & BC_F_COPYOUT) {
 833                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 834                         if (args.bca_len != 0) {
 835                                 KKASSERT(args.bca_kptr != NULL);
 836                                 if (!error) {
 837                                         error = copyout(args.bca_kptr,
 838                                                 args.bca_uptr, args.bca_len);
 839                                 }
 840                                 kfree(args.bca_kptr, M_TEMP);
 841                         } else {
 842                                 KKASSERT(args.bca_kptr == NULL);
 843                         }
 844                 } else {
 845                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 846                 }
 847                 break;
 848
 849         case SIOCSIFFLAGS:
 850                 if (!(ifp->if_flags & IFF_UP) &&
 851                     (ifp->if_flags & IFF_RUNNING)) {
 852                         /*
 853                          * If interface is marked down and it is running,
 854                          * then stop it.
 855                          */
 856                         bridge_stop(ifp);
 857                 } else if ((ifp->if_flags & IFF_UP) &&
 858                     !(ifp->if_flags & IFF_RUNNING)) {
 859                         /*
 860                          * If interface is marked up and it is stopped, then
 861                          * start it.
 862                          */
 863                         ifp->if_init(sc);
 864                 }
 865                 break;
 866
 867         case SIOCSIFMTU:
 868                 /* Do not allow the MTU to be changed on the bridge */
 869                 error = EINVAL;
 870                 break;
 871
 872         default:
 873                 error = ether_ioctl(ifp, cmd, data);
 874                 break;
 875         }
 876         return (error);
 877 }
 878
 879 /*
 880  * bridge_mutecaps:
 881  *
 882  *      Clear or restore unwanted capabilities on the member interface
 883  */
 884 static void
 885 bridge_mutecaps(struct bridge_ifinfo *bif_info, struct ifnet *ifp, int mute)
 886 {
 887         struct ifreq ifr;
 888         int error;
 889
 890         if (ifp->if_ioctl == NULL)
 891                 return;
 892
 893         bzero(&ifr, sizeof(ifr));
 894         ifr.ifr_reqcap = ifp->if_capenable;
 895
 896         if (mute) {
 897                 /* mask off and save capabilities */
 898                 bif_info->bifi_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
 899                 if (bif_info->bifi_mutecap != 0)
 900                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
 901         } else {
 902                 /* restore muted capabilities */
 903                 ifr.ifr_reqcap |= bif_info->bifi_mutecap;
 904         }
 905
 906         if (bif_info->bifi_mutecap != 0) {
 907                 ifnet_serialize_all(ifp);
 908                 error = ifp->if_ioctl(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL);
 909                 ifnet_deserialize_all(ifp);
 910         }
 911 }
 912
 913 /*
 914  * bridge_lookup_member:
 915  *
 916  *      Lookup a bridge member interface.
 917  */
 918 static struct bridge_iflist *
 919 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 920 {
 921         struct bridge_iflist *bif;
 922
 923         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 924                 if (strcmp(bif->bif_ifp->if_xname, name) == 0)
 925                         return (bif);
 926         }
 927         return (NULL);
 928 }
 929
 930 /*
 931  * bridge_lookup_member_if:
 932  *
 933  *      Lookup a bridge member interface by ifnet*.
 934  */
 935 static struct bridge_iflist *
 936 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 937 {
 938         struct bridge_iflist *bif;
 939
 940         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 941                 if (bif->bif_ifp == member_ifp)
 942                         return (bif);
 943         }
 944         return (NULL);
 945 }
 946
 947 /*
 948  * bridge_lookup_member_ifinfo:
 949  *
 950  *      Lookup a bridge member interface by bridge_ifinfo.
 951  */
 952 static struct bridge_iflist *
 953 bridge_lookup_member_ifinfo(struct bridge_softc *sc,
 954                             struct bridge_ifinfo *bif_info)
 955 {
 956         struct bridge_iflist *bif;
 957
 958         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 959                 if (bif->bif_info == bif_info)
 960                         return (bif);
 961         }
 962         return (NULL);
 963 }
 964
 965 /*
 966  * bridge_delete_member:
 967  *
 968  *      Delete the specified member interface.
 969  */
 970 static void
 971 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 972     int gone)
 973 {
 974         struct ifnet *ifs = bif->bif_ifp;
 975         struct ifnet *bifp = sc->sc_ifp;
 976         struct bridge_ifinfo *bif_info = bif->bif_info;
 977         struct bridge_iflist_head saved_bifs;
 978
 979         ASSERT_IFNET_SERIALIZED_ALL(bifp);
 980         KKASSERT(bif_info != NULL);
 981
 982         ifs->if_bridge = NULL;
 983
 984         /*
 985          * Release bridge interface's serializer:
 986          * - To avoid possible dead lock.
 987          * - Various sync operation will block the current thread.
 988          */
 989         ifnet_deserialize_all(bifp);
 990
 991         if (!gone) {
 992                 switch (ifs->if_type) {
 993                 case IFT_ETHER:
 994                 case IFT_L2VLAN:
 995                         /*
 996                          * Take the interface out of promiscuous mode.
 997                          */
 998                         ifpromisc(ifs, 0);
 999                         bridge_mutecaps(bif_info, ifs, 0);
1000                         break;
1001
1002                 case IFT_GIF:
1003                         break;
1004
1005                 default:
1006                         panic("bridge_delete_member: impossible");
1007                         break;
1008                 }
1009         }
1010
1011         /*
1012          * Remove bifs from percpu linked list.
1013          *
1014          * Removed bifs are not freed immediately, instead,
1015          * they are saved in saved_bifs.  They will be freed
1016          * after we make sure that no one is accessing them,
1017          * i.e. after following netmsg_service_sync()
1018          */
1019         LIST_INIT(&saved_bifs);
1020         bridge_del_bif(sc, bif_info, &saved_bifs);
1021
1022         /*
1023          * Make sure that all protocol threads:
1024          * o  see 'ifs' if_bridge is changed
1025          * o  know that bif is removed from the percpu linked list
1026          */
1027         netmsg_service_sync();
1028
1029         /*
1030          * Free the removed bifs
1031          */
1032         KKASSERT(!LIST_EMPTY(&saved_bifs));
1033         while ((bif = LIST_FIRST(&saved_bifs)) != NULL) {
1034                 LIST_REMOVE(bif, bif_next);
1035                 kfree(bif, M_DEVBUF);
1036         }
1037
1038         /* See the comment in bridge_ioctl_stop() */
1039         bridge_rtmsg_sync(sc);
1040         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL | IFBF_FLUSHSYNC);
1041
1042         ifnet_serialize_all(bifp);
1043
1044         if (bifp->if_flags & IFF_RUNNING)
1045                 bstp_initialization(sc);
1046
1047         /*
1048          * Free the bif_info after bstp_initialization(), so that
1049          * bridge_softc.sc_root_port will not reference a dangling
1050          * pointer.
1051          */
1052         kfree(bif_info, M_DEVBUF);
1053 }
1054
1055 /*
1056  * bridge_delete_span:
1057  *
1058  *      Delete the specified span interface.
1059  */
1060 static void
1061 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1062 {
1063         KASSERT(bif->bif_ifp->if_bridge == NULL,
1064             ("%s: not a span interface", __func__));
1065
1066         LIST_REMOVE(bif, bif_next);
1067         kfree(bif, M_DEVBUF);
1068 }
1069
1070 static int
1071 bridge_ioctl_init(struct bridge_softc *sc, void *arg __unused)
1072 {
1073         struct ifnet *ifp = sc->sc_ifp;
1074
1075         if (ifp->if_flags & IFF_RUNNING)
1076                 return 0;
1077
1078         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1079             bridge_timer, sc);
1080
1081         ifp->if_flags |= IFF_RUNNING;
1082         bstp_initialization(sc);
1083         return 0;
1084 }
1085
1086 static int
1087 bridge_ioctl_stop(struct bridge_softc *sc, void *arg __unused)
1088 {
1089         struct ifnet *ifp = sc->sc_ifp;
1090         struct lwkt_msg *lmsg;
1091
1092         if ((ifp->if_flags & IFF_RUNNING) == 0)
1093                 return 0;
1094
1095         callout_stop(&sc->sc_brcallout);
1096
1097         crit_enter();
1098         lmsg = &sc->sc_brtimemsg.lmsg;
1099         if ((lmsg->ms_flags & MSGF_DONE) == 0) {
1100                 /* Pending to be processed; drop it */
1101                 lwkt_dropmsg(lmsg);
1102         }
1103         crit_exit();
1104
1105         bstp_stop(sc);
1106
1107         ifp->if_flags &= ~IFF_RUNNING;
1108
1109         ifnet_deserialize_all(ifp);
1110
1111         /* Let everyone know that we are stopped */
1112         netmsg_service_sync();
1113
1114         /*
1115          * Sync ifnetX msgports in the order we forward rtnode
1116          * installation message.  This is used to make sure that
1117          * all rtnode installation messages sent by bridge_rtupdate()
1118          * during above netmsg_service_sync() are flushed.
1119          */
1120         bridge_rtmsg_sync(sc);
1121         bridge_rtflush(sc, IFBF_FLUSHDYN | IFBF_FLUSHSYNC);
1122
1123         ifnet_serialize_all(ifp);
1124         return 0;
1125 }
1126
1127 static int
1128 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1129 {
1130         struct ifbreq *req = arg;
1131         struct bridge_iflist *bif;
1132         struct bridge_ifinfo *bif_info;
1133         struct ifnet *ifs, *bifp;
1134         int error = 0;
1135
1136         bifp = sc->sc_ifp;
1137         ASSERT_IFNET_SERIALIZED_ALL(bifp);
1138
1139         ifs = ifunit(req->ifbr_ifsname);
1140         if (ifs == NULL)
1141                 return (ENOENT);
1142
1143         /* If it's in the span list, it can't be a member. */
1144         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1145                 if (ifs == bif->bif_ifp)
1146                         return (EBUSY);
1147
1148         /* Allow the first Ethernet member to define the MTU */
1149         if (ifs->if_type != IFT_GIF) {
1150                 if (LIST_EMPTY(&sc->sc_iflists[mycpuid])) {
1151                         bifp->if_mtu = ifs->if_mtu;
1152                 } else if (bifp->if_mtu != ifs->if_mtu) {
1153                         if_printf(bifp, "invalid MTU for %s\n", ifs->if_xname);
1154                         return (EINVAL);
1155                 }
1156         }
1157
1158         if (ifs->if_bridge == sc)
1159                 return (EEXIST);
1160
1161         if (ifs->if_bridge != NULL)
1162                 return (EBUSY);
1163
1164         bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO);
1165         bif_info->bifi_priority = BSTP_DEFAULT_PORT_PRIORITY;
1166         bif_info->bifi_path_cost = BSTP_DEFAULT_PATH_COST;
1167         bif_info->bifi_ifp = ifs;
1168
1169         /*
1170          * Release bridge interface's serializer:
1171          * - To avoid possible dead lock.
1172          * - Various sync operation will block the current thread.
1173          */
1174         ifnet_deserialize_all(bifp);
1175
1176         switch (ifs->if_type) {
1177         case IFT_ETHER:
1178         case IFT_L2VLAN:
1179                 /*
1180                  * Place the interface into promiscuous mode.
1181                  */
1182                 error = ifpromisc(ifs, 1);
1183                 if (error) {
1184                         ifnet_serialize_all(bifp);
1185                         goto out;
1186                 }
1187                 bridge_mutecaps(bif_info, ifs, 1);
1188                 break;
1189
1190         case IFT_GIF: /* :^) */
1191                 break;
1192
1193         default:
1194                 error = EINVAL;
1195                 ifnet_serialize_all(bifp);
1196                 goto out;
1197         }
1198
1199         /*
1200          * Add bifs to percpu linked lists
1201          */
1202         bridge_add_bif(sc, bif_info, ifs);
1203
1204         ifnet_serialize_all(bifp);
1205
1206         if (bifp->if_flags & IFF_RUNNING)
1207                 bstp_initialization(sc);
1208         else
1209                 bstp_stop(sc);
1210
1211         /*
1212          * Everything has been setup, so let the member interface
1213          * deliver packets to this bridge on its input/output path.
1214          */
1215         ifs->if_bridge = sc;
1216 out:
1217         if (error) {
1218                 if (bif_info != NULL)
1219                         kfree(bif_info, M_DEVBUF);
1220         }
1221         return (error);
1222 }
1223
1224 static int
1225 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1226 {
1227         struct ifbreq *req = arg;
1228         struct bridge_iflist *bif;
1229
1230         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1231         if (bif == NULL)
1232                 return (ENOENT);
1233
1234         bridge_delete_member(sc, bif, 0);
1235
1236         return (0);
1237 }
1238
1239 static int
1240 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1241 {
1242         struct ifbreq *req = arg;
1243         struct bridge_iflist *bif;
1244
1245         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1246         if (bif == NULL)
1247                 return (ENOENT);
1248
1249         req->ifbr_ifsflags = bif->bif_flags;
1250         req->ifbr_state = bif->bif_state;
1251         req->ifbr_priority = bif->bif_priority;
1252         req->ifbr_path_cost = bif->bif_path_cost;
1253         req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1254
1255         return (0);
1256 }
1257
1258 static int
1259 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1260 {
1261         struct ifbreq *req = arg;
1262         struct bridge_iflist *bif;
1263         struct ifnet *bifp = sc->sc_ifp;
1264
1265         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1266         if (bif == NULL)
1267                 return (ENOENT);
1268
1269         if (req->ifbr_ifsflags & IFBIF_SPAN) {
1270                 /* SPAN is readonly */
1271                 return (EINVAL);
1272         }
1273
1274         if (req->ifbr_ifsflags & IFBIF_STP) {
1275                 switch (bif->bif_ifp->if_type) {
1276                 case IFT_ETHER:
1277                         /* These can do spanning tree. */
1278                         break;
1279
1280                 default:
1281                         /* Nothing else can. */
1282                         return (EINVAL);
1283                 }
1284         }
1285
1286         ifnet_deserialize_all(bifp);
1287         bridge_set_bifflags(sc, bif->bif_info, req->ifbr_ifsflags);
1288         ifnet_serialize_all(bifp);
1289
1290         if (bifp->if_flags & IFF_RUNNING)
1291                 bstp_initialization(sc);
1292
1293         return (0);
1294 }
1295
1296 static int
1297 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1298 {
1299         struct ifbrparam *param = arg;
1300         struct ifnet *ifp = sc->sc_ifp;
1301
1302         sc->sc_brtmax = param->ifbrp_csize;
1303
1304         ifnet_deserialize_all(ifp);
1305         bridge_rttrim(sc);
1306         ifnet_serialize_all(ifp);
1307
1308         return (0);
1309 }
1310
1311 static int
1312 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1313 {
1314         struct ifbrparam *param = arg;
1315
1316         param->ifbrp_csize = sc->sc_brtmax;
1317
1318         return (0);
1319 }
1320
1321 static int
1322 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1323 {
1324         struct bridge_control_arg *bc_arg = arg;
1325         struct ifbifconf *bifc = arg;
1326         struct bridge_iflist *bif;
1327         struct ifbreq *breq;
1328         int count, len;
1329
1330         count = 0;
1331         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next)
1332                 count++;
1333         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1334                 count++;
1335
1336         if (bifc->ifbic_len == 0) {
1337                 bifc->ifbic_len = sizeof(*breq) * count;
1338                 return 0;
1339         } else if (count == 0 || bifc->ifbic_len < sizeof(*breq)) {
1340                 bifc->ifbic_len = 0;
1341                 return 0;
1342         }
1343
1344         len = min(bifc->ifbic_len, sizeof(*breq) * count);
1345         KKASSERT(len >= sizeof(*breq));
1346
1347         breq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1348         if (breq == NULL) {
1349                 bifc->ifbic_len = 0;
1350                 return ENOMEM;
1351         }
1352         bc_arg->bca_kptr = breq;
1353
1354         count = 0;
1355         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1356                 if (len < sizeof(*breq))
1357                         break;
1358
1359                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1360                         sizeof(breq->ifbr_ifsname));
1361                 breq->ifbr_ifsflags = bif->bif_flags;
1362                 breq->ifbr_state = bif->bif_state;
1363                 breq->ifbr_priority = bif->bif_priority;
1364                 breq->ifbr_path_cost = bif->bif_path_cost;
1365                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1366                 breq++;
1367                 count++;
1368                 len -= sizeof(*breq);
1369         }
1370         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1371                 if (len < sizeof(*breq))
1372                         break;
1373
1374                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1375                         sizeof(breq->ifbr_ifsname));
1376                 breq->ifbr_ifsflags = bif->bif_flags;
1377                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1378                 breq++;
1379                 count++;
1380                 len -= sizeof(*breq);
1381         }
1382
1383         bifc->ifbic_len = sizeof(*breq) * count;
1384         KKASSERT(bifc->ifbic_len > 0);
1385
1386         bc_arg->bca_len = bifc->ifbic_len;
1387         bc_arg->bca_uptr = bifc->ifbic_req;
1388         return 0;
1389 }
1390
1391 static int
1392 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1393 {
1394         struct bridge_control_arg *bc_arg = arg;
1395         struct ifbaconf *bac = arg;
1396         struct bridge_rtnode *brt;
1397         struct ifbareq *bareq;
1398         int count, len;
1399
1400         count = 0;
1401         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list)
1402                 count++;
1403
1404         if (bac->ifbac_len == 0) {
1405                 bac->ifbac_len = sizeof(*bareq) * count;
1406                 return 0;
1407         } else if (count == 0 || bac->ifbac_len < sizeof(*bareq)) {
1408                 bac->ifbac_len = 0;
1409                 return 0;
1410         }
1411
1412         len = min(bac->ifbac_len, sizeof(*bareq) * count);
1413         KKASSERT(len >= sizeof(*bareq));
1414
1415         bareq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1416         if (bareq == NULL) {
1417                 bac->ifbac_len = 0;
1418                 return ENOMEM;
1419         }
1420         bc_arg->bca_kptr = bareq;
1421
1422         count = 0;
1423         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
1424                 struct bridge_rtinfo *bri = brt->brt_info;
1425                 unsigned long expire;
1426
1427                 if (len < sizeof(*bareq))
1428                         break;
1429
1430                 strlcpy(bareq->ifba_ifsname, bri->bri_ifp->if_xname,
1431                         sizeof(bareq->ifba_ifsname));
1432                 memcpy(bareq->ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1433                 expire = bri->bri_expire;
1434                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1435                     time_second < expire)
1436                         bareq->ifba_expire = expire - time_second;
1437                 else
1438                         bareq->ifba_expire = 0;
1439                 bareq->ifba_flags = bri->bri_flags;
1440                 bareq++;
1441                 count++;
1442                 len -= sizeof(*bareq);
1443         }
1444
1445         bac->ifbac_len = sizeof(*bareq) * count;
1446         KKASSERT(bac->ifbac_len > 0);
1447
1448         bc_arg->bca_len = bac->ifbac_len;
1449         bc_arg->bca_uptr = bac->ifbac_req;
1450         return 0;
1451 }
1452
1453 static int
1454 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1455 {
1456         struct ifbareq *req = arg;
1457         struct bridge_iflist *bif;
1458         struct ifnet *ifp = sc->sc_ifp;
1459         int error;
1460
1461         ASSERT_IFNET_SERIALIZED_ALL(ifp);
1462
1463         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1464         if (bif == NULL)
1465                 return (ENOENT);
1466
1467         ifnet_deserialize_all(ifp);
1468         error = bridge_rtsaddr(sc, req->ifba_dst, bif->bif_ifp,
1469                                req->ifba_flags);
1470         ifnet_serialize_all(ifp);
1471         return (error);
1472 }
1473
1474 static int
1475 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1476 {
1477         struct ifbrparam *param = arg;
1478
1479         sc->sc_brttimeout = param->ifbrp_ctime;
1480
1481         return (0);
1482 }
1483
1484 static int
1485 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1486 {
1487         struct ifbrparam *param = arg;
1488
1489         param->ifbrp_ctime = sc->sc_brttimeout;
1490
1491         return (0);
1492 }
1493
1494 static int
1495 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1496 {
1497         struct ifbareq *req = arg;
1498         struct ifnet *ifp = sc->sc_ifp;
1499         int error;
1500
1501         ifnet_deserialize_all(ifp);
1502         error = bridge_rtdaddr(sc, req->ifba_dst);
1503         ifnet_serialize_all(ifp);
1504         return error;
1505 }
1506
1507 static int
1508 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1509 {
1510         struct ifbreq *req = arg;
1511         struct ifnet *ifp = sc->sc_ifp;
1512
1513         ifnet_deserialize_all(ifp);
1514         bridge_rtflush(sc, req->ifbr_ifsflags | IFBF_FLUSHSYNC);
1515         ifnet_serialize_all(ifp);
1516
1517         return (0);
1518 }
1519
1520 static int
1521 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1522 {
1523         struct ifbrparam *param = arg;
1524
1525         param->ifbrp_prio = sc->sc_bridge_priority;
1526
1527         return (0);
1528 }
1529
1530 static int
1531 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1532 {
1533         struct ifbrparam *param = arg;
1534
1535         sc->sc_bridge_priority = param->ifbrp_prio;
1536
1537         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1538                 bstp_initialization(sc);
1539
1540         return (0);
1541 }
1542
1543 static int
1544 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1545 {
1546         struct ifbrparam *param = arg;
1547
1548         param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1549
1550         return (0);
1551 }
1552
1553 static int
1554 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1555 {
1556         struct ifbrparam *param = arg;
1557
1558         if (param->ifbrp_hellotime == 0)
1559                 return (EINVAL);
1560         sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1561
1562         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1563                 bstp_initialization(sc);
1564
1565         return (0);
1566 }
1567
1568 static int
1569 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1570 {
1571         struct ifbrparam *param = arg;
1572
1573         param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1574
1575         return (0);
1576 }
1577
1578 static int
1579 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1580 {
1581         struct ifbrparam *param = arg;
1582
1583         if (param->ifbrp_fwddelay == 0)
1584                 return (EINVAL);
1585         sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1586
1587         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1588                 bstp_initialization(sc);
1589
1590         return (0);
1591 }
1592
1593 static int
1594 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1595 {
1596         struct ifbrparam *param = arg;
1597
1598         param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1599
1600         return (0);
1601 }
1602
1603 static int
1604 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1605 {
1606         struct ifbrparam *param = arg;
1607
1608         if (param->ifbrp_maxage == 0)
1609                 return (EINVAL);
1610         sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1611
1612         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1613                 bstp_initialization(sc);
1614
1615         return (0);
1616 }
1617
1618 static int
1619 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1620 {
1621         struct ifbreq *req = arg;
1622         struct bridge_iflist *bif;
1623
1624         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1625         if (bif == NULL)
1626                 return (ENOENT);
1627
1628         bif->bif_priority = req->ifbr_priority;
1629
1630         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1631                 bstp_initialization(sc);
1632
1633         return (0);
1634 }
1635
1636 static int
1637 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1638 {
1639         struct ifbreq *req = arg;
1640         struct bridge_iflist *bif;
1641
1642         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1643         if (bif == NULL)
1644                 return (ENOENT);
1645
1646         bif->bif_path_cost = req->ifbr_path_cost;
1647
1648         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1649                 bstp_initialization(sc);
1650
1651         return (0);
1652 }
1653
1654 static int
1655 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1656 {
1657         struct ifbreq *req = arg;
1658         struct bridge_iflist *bif;
1659         struct ifnet *ifs;
1660
1661         ifs = ifunit(req->ifbr_ifsname);
1662         if (ifs == NULL)
1663                 return (ENOENT);
1664
1665         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1666                 if (ifs == bif->bif_ifp)
1667                         return (EBUSY);
1668
1669         if (ifs->if_bridge != NULL)
1670                 return (EBUSY);
1671
1672         switch (ifs->if_type) {
1673         case IFT_ETHER:
1674         case IFT_GIF:
1675         case IFT_L2VLAN:
1676                 break;
1677
1678         default:
1679                 return (EINVAL);
1680         }
1681
1682         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
1683         bif->bif_ifp = ifs;
1684         bif->bif_flags = IFBIF_SPAN;
1685         /* NOTE: span bif does not need bridge_ifinfo */
1686
1687         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1688
1689         sc->sc_span = 1;
1690
1691         return (0);
1692 }
1693
1694 static int
1695 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1696 {
1697         struct ifbreq *req = arg;
1698         struct bridge_iflist *bif;
1699         struct ifnet *ifs;
1700
1701         ifs = ifunit(req->ifbr_ifsname);
1702         if (ifs == NULL)
1703                 return (ENOENT);
1704
1705         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1706                 if (ifs == bif->bif_ifp)
1707                         break;
1708
1709         if (bif == NULL)
1710                 return (ENOENT);
1711
1712         bridge_delete_span(sc, bif);
1713
1714         if (LIST_EMPTY(&sc->sc_spanlist))
1715                 sc->sc_span = 0;
1716
1717         return (0);
1718 }
1719
1720 static void
1721 bridge_ifdetach_dispatch(netmsg_t msg)
1722 {
1723         struct ifnet *ifp, *bifp;
1724         struct bridge_softc *sc;
1725         struct bridge_iflist *bif;
1726
1727         ifp = msg->lmsg.u.ms_resultp;
1728         sc = ifp->if_bridge;
1729
1730         /* Check if the interface is a bridge member */
1731         if (sc != NULL) {
1732                 bifp = sc->sc_ifp;
1733
1734                 ifnet_serialize_all(bifp);
1735
1736                 bif = bridge_lookup_member_if(sc, ifp);
1737                 if (bif != NULL) {
1738                         bridge_delete_member(sc, bif, 1);
1739                 } else {
1740                         /* XXX Why bif will be NULL? */
1741                 }
1742
1743                 ifnet_deserialize_all(bifp);
1744                 goto reply;
1745         }
1746
1747         crit_enter();   /* XXX MP */
1748
1749         /* Check if the interface is a span port */
1750         LIST_FOREACH(sc, &bridge_list, sc_list) {
1751                 bifp = sc->sc_ifp;
1752
1753                 ifnet_serialize_all(bifp);
1754
1755                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1756                         if (ifp == bif->bif_ifp) {
1757                                 bridge_delete_span(sc, bif);
1758                                 break;
1759                         }
1760
1761                 ifnet_deserialize_all(bifp);
1762         }
1763
1764         crit_exit();
1765
1766 reply:
1767         lwkt_replymsg(&msg->lmsg, 0);
1768 }
1769
1770 /*
1771  * bridge_ifdetach:
1772  *
1773  *      Detach an interface from a bridge.  Called when a member
1774  *      interface is detaching.
1775  */
1776 static void
1777 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1778 {
1779         struct netmsg_base msg;
1780
1781         netmsg_init(&msg, NULL, &curthread->td_msgport,
1782                     0, bridge_ifdetach_dispatch);
1783         msg.lmsg.u.ms_resultp = ifp;
1784
1785         lwkt_domsg(BRIDGE_CFGPORT, &msg.lmsg, 0);
1786 }
1787
1788 /*
1789  * bridge_init:
1790  *
1791  *      Initialize a bridge interface.
1792  */
1793 static void
1794 bridge_init(void *xsc)
1795 {
1796         bridge_control(xsc, SIOCSIFFLAGS, bridge_ioctl_init, NULL);
1797 }
1798
1799 /*
1800  * bridge_stop:
1801  *
1802  *      Stop the bridge interface.
1803  */
1804 static void
1805 bridge_stop(struct ifnet *ifp)
1806 {
1807         bridge_control(ifp->if_softc, SIOCSIFFLAGS, bridge_ioctl_stop, NULL);
1808 }
1809
1810 /*
1811  * bridge_enqueue:
1812  *
1813  *      Enqueue a packet on a bridge member interface.
1814  *
1815  */
1816 void
1817 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m)
1818 {
1819         struct netmsg_packet *nmp;
1820
1821         nmp = &m->m_hdr.mh_netmsg;
1822         netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
1823                     0, bridge_enqueue_handler);
1824         nmp->nm_packet = m;
1825         nmp->base.lmsg.u.ms_resultp = dst_ifp;
1826
1827         lwkt_sendmsg(ifnet_portfn(mycpu->gd_cpuid), &nmp->base.lmsg);
1828 }
1829
1830 /*
1831  * bridge_output:
1832  *
1833  *      Send output from a bridge member interface.  This
1834  *      performs the bridging function for locally originated
1835  *      packets.
1836  *
1837  *      The mbuf has the Ethernet header already attached.  We must
1838  *      enqueue or free the mbuf before returning.
1839  */
1840 static int
1841 bridge_output(struct ifnet *ifp, struct mbuf *m)
1842 {
1843         struct bridge_softc *sc = ifp->if_bridge;
1844         struct ether_header *eh;
1845         struct ifnet *dst_if, *bifp;
1846
1847         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
1848
1849         /*
1850          * Make sure that we are still a member of a bridge interface.
1851          */
1852         if (sc == NULL) {
1853                 m_freem(m);
1854                 return (0);
1855         }
1856         bifp = sc->sc_ifp;
1857
1858         if (m->m_len < ETHER_HDR_LEN) {
1859                 m = m_pullup(m, ETHER_HDR_LEN);
1860                 if (m == NULL)
1861                         return (0);
1862         }
1863         eh = mtod(m, struct ether_header *);
1864
1865         /*
1866          * If bridge is down, but the original output interface is up,
1867          * go ahead and send out that interface.  Otherwise, the packet
1868          * is dropped below.
1869          */
1870         if ((bifp->if_flags & IFF_RUNNING) == 0) {
1871                 dst_if = ifp;
1872                 goto sendunicast;
1873         }
1874
1875         /*
1876          * If the packet is a multicast, or we don't know a better way to
1877          * get there, send to all interfaces.
1878          */
1879         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1880                 dst_if = NULL;
1881         else
1882                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1883         if (dst_if == NULL) {
1884                 struct bridge_iflist *bif, *nbif;
1885                 struct mbuf *mc;
1886                 int used = 0;
1887
1888                 if (sc->sc_span)
1889                         bridge_span(sc, m);
1890
1891                 LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid],
1892                                      bif_next, nbif) {
1893                         dst_if = bif->bif_ifp;
1894                         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1895                                 continue;
1896
1897                         /*
1898                          * If this is not the original output interface,
1899                          * and the interface is participating in spanning
1900                          * tree, make sure the port is in a state that
1901                          * allows forwarding.
1902                          */
1903                         if (dst_if != ifp &&
1904                             (bif->bif_flags & IFBIF_STP) != 0) {
1905                                 switch (bif->bif_state) {
1906                                 case BSTP_IFSTATE_BLOCKING:
1907                                 case BSTP_IFSTATE_LISTENING:
1908                                 case BSTP_IFSTATE_DISABLED:
1909                                         continue;
1910                                 }
1911                         }
1912
1913                         if (LIST_NEXT(bif, bif_next) == NULL) {
1914                                 used = 1;
1915                                 mc = m;
1916                         } else {
1917                                 mc = m_copypacket(m, MB_DONTWAIT);
1918                                 if (mc == NULL) {
1919                                         bifp->if_oerrors++;
1920                                         continue;
1921                                 }
1922                         }
1923                         bridge_handoff(dst_if, mc);
1924
1925                         if (nbif != NULL && !nbif->bif_onlist) {
1926                                 KKASSERT(bif->bif_onlist);
1927                                 nbif = LIST_NEXT(bif, bif_next);
1928                         }
1929                 }
1930                 if (used == 0)
1931                         m_freem(m);
1932                 return (0);
1933         }
1934
1935 sendunicast:
1936         /*
1937          * XXX Spanning tree consideration here?
1938          */
1939         if (sc->sc_span)
1940                 bridge_span(sc, m);
1941         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1942                 m_freem(m);
1943         else
1944                 bridge_handoff(dst_if, m);
1945         return (0);
1946 }
1947
1948 /*
1949  * bridge_start:
1950  *
1951  *      Start output on a bridge.
1952  *
1953  */
1954 static void
1955 bridge_start(struct ifnet *ifp)
1956 {
1957         struct bridge_softc *sc = ifp->if_softc;
1958
1959         ASSERT_IFNET_SERIALIZED_TX(ifp);
1960
1961         ifp->if_flags |= IFF_OACTIVE;
1962         for (;;) {
1963                 struct ifnet *dst_if = NULL;
1964                 struct ether_header *eh;
1965                 struct mbuf *m;
1966
1967                 m = ifq_dequeue(&ifp->if_snd, NULL);
1968                 if (m == NULL)
1969                         break;
1970
1971                 if (m->m_len < sizeof(*eh)) {
1972                         m = m_pullup(m, sizeof(*eh));
1973                         if (m == NULL) {
1974                                 ifp->if_oerrors++;
1975                                 continue;
1976                         }
1977                 }
1978                 eh = mtod(m, struct ether_header *);
1979
1980                 BPF_MTAP(ifp, m);
1981                 ifp->if_opackets++;
1982
1983                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0)
1984                         dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1985
1986                 if (dst_if == NULL)
1987                         bridge_start_bcast(sc, m);
1988                 else
1989                         bridge_enqueue(dst_if, m);
1990         }
1991         ifp->if_flags &= ~IFF_OACTIVE;
1992 }
1993
1994 /*
1995  * bridge_forward:
1996  *
1997  *      The forwarding function of the bridge.
1998  */
1999 static void
2000 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
2001 {
2002         struct bridge_iflist *bif;
2003         struct ifnet *src_if, *dst_if, *ifp;
2004         struct ether_header *eh;
2005
2006         src_if = m->m_pkthdr.rcvif;
2007         ifp = sc->sc_ifp;
2008
2009         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2010
2011         ifp->if_ipackets++;
2012         ifp->if_ibytes += m->m_pkthdr.len;
2013
2014         /*
2015          * Look up the bridge_iflist.
2016          */
2017         bif = bridge_lookup_member_if(sc, src_if);
2018         if (bif == NULL) {
2019                 /* Interface is not a bridge member (anymore?) */
2020                 m_freem(m);
2021                 return;
2022         }
2023
2024         if (bif->bif_flags & IFBIF_STP) {
2025                 switch (bif->bif_state) {
2026                 case BSTP_IFSTATE_BLOCKING:
2027                 case BSTP_IFSTATE_LISTENING:
2028                 case BSTP_IFSTATE_DISABLED:
2029                         m_freem(m);
2030                         return;
2031                 }
2032         }
2033
2034         eh = mtod(m, struct ether_header *);
2035
2036         /*
2037          * If the interface is learning, and the source
2038          * address is valid and not multicast, record
2039          * the address.
2040          */
2041         if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
2042             ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
2043             (eh->ether_shost[0] == 0 &&
2044              eh->ether_shost[1] == 0 &&
2045              eh->ether_shost[2] == 0 &&
2046              eh->ether_shost[3] == 0 &&
2047              eh->ether_shost[4] == 0 &&
2048              eh->ether_shost[5] == 0) == 0)
2049                 bridge_rtupdate(sc, eh->ether_shost, src_if, IFBAF_DYNAMIC);
2050
2051         if ((bif->bif_flags & IFBIF_STP) != 0 &&
2052             bif->bif_state == BSTP_IFSTATE_LEARNING) {
2053                 m_freem(m);
2054                 return;
2055         }
2056
2057         /*
2058          * At this point, the port either doesn't participate
2059          * in spanning tree or it is in the forwarding state.
2060          */
2061
2062         /*
2063          * If the packet is unicast, destined for someone on
2064          * "this" side of the bridge, drop it.
2065          */
2066         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2067                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2068                 if (src_if == dst_if) {
2069                         m_freem(m);
2070                         return;
2071                 }
2072         } else {
2073                 /* ...forward it to all interfaces. */
2074                 ifp->if_imcasts++;
2075                 dst_if = NULL;
2076         }
2077
2078         if (dst_if == NULL) {
2079                 bridge_broadcast(sc, src_if, m);
2080                 return;
2081         }
2082
2083         /*
2084          * At this point, we're dealing with a unicast frame
2085          * going to a different interface.
2086          */
2087         if ((dst_if->if_flags & IFF_RUNNING) == 0) {
2088                 m_freem(m);
2089                 return;
2090         }
2091         bif = bridge_lookup_member_if(sc, dst_if);
2092         if (bif == NULL) {
2093                 /* Not a member of the bridge (anymore?) */
2094                 m_freem(m);
2095                 return;
2096         }
2097
2098         if (bif->bif_flags & IFBIF_STP) {
2099                 switch (bif->bif_state) {
2100                 case BSTP_IFSTATE_DISABLED:
2101                 case BSTP_IFSTATE_BLOCKING:
2102                         m_freem(m);
2103                         return;
2104                 }
2105         }
2106
2107         if (inet_pfil_hook.ph_hashooks > 0
2108 #ifdef INET6
2109             || inet6_pfil_hook.ph_hashooks > 0
2110 #endif
2111             ) {
2112                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2113                         return;
2114                 if (m == NULL)
2115                         return;
2116
2117                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2118                         return;
2119                 if (m == NULL)
2120                         return;
2121         }
2122         bridge_handoff(dst_if, m);
2123 }
2124
2125 /*
2126  * bridge_input:
2127  *
2128  *      Receive input from a member interface.  Queue the packet for
2129  *      bridging if it is not for us.
2130  */
2131 static struct mbuf *
2132 bridge_input(struct ifnet *ifp, struct mbuf *m)
2133 {
2134         struct bridge_softc *sc = ifp->if_bridge;
2135         struct bridge_iflist *bif;
2136         struct ifnet *bifp, *new_ifp;
2137         struct ether_header *eh;
2138         struct mbuf *mc, *mc2;
2139
2140         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2141
2142         /*
2143          * Make sure that we are still a member of a bridge interface.
2144          */
2145         if (sc == NULL)
2146                 return m;
2147
2148         new_ifp = NULL;
2149         bifp = sc->sc_ifp;
2150
2151         if ((bifp->if_flags & IFF_RUNNING) == 0)
2152                 goto out;
2153
2154         /*
2155          * Implement support for bridge monitoring.  If this flag has been
2156          * set on this interface, discard the packet once we push it through
2157          * the bpf(4) machinery, but before we do, increment various counters
2158          * associated with this bridge.
2159          */
2160         if (bifp->if_flags & IFF_MONITOR) {
2161                 /* Change input interface to this bridge */
2162                 m->m_pkthdr.rcvif = bifp;
2163
2164                 BPF_MTAP(bifp, m);
2165
2166                 /* Update bridge's ifnet statistics */
2167                 bifp->if_ipackets++;
2168                 bifp->if_ibytes += m->m_pkthdr.len;
2169                 if (m->m_flags & (M_MCAST | M_BCAST))
2170                         bifp->if_imcasts++;
2171
2172                 m_freem(m);
2173                 m = NULL;
2174                 goto out;
2175         }
2176
2177         eh = mtod(m, struct ether_header *);
2178
2179         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
2180                 /*
2181                  * If the packet is for us, set the packets source as the
2182                  * bridge, and return the packet back to ifnet.if_input for
2183                  * local processing.
2184                  */
2185                 KASSERT(bifp->if_bridge == NULL,
2186                         ("loop created in bridge_input"));
2187                 new_ifp = bifp;
2188                 goto out;
2189         }
2190
2191         /*
2192          * Tap all packets arriving on the bridge, no matter if
2193          * they are local destinations or not.  In is in.
2194          */
2195         BPF_MTAP(bifp, m);
2196
2197         bif = bridge_lookup_member_if(sc, ifp);
2198         if (bif == NULL)
2199                 goto out;
2200
2201         if (sc->sc_span)
2202                 bridge_span(sc, m);
2203
2204         if (m->m_flags & (M_BCAST | M_MCAST)) {
2205                 /* Tap off 802.1D packets; they do not get forwarded. */
2206                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2207                     ETHER_ADDR_LEN) == 0) {
2208                         ifnet_serialize_all(bifp);
2209                         bstp_input(sc, bif, m);
2210                         ifnet_deserialize_all(bifp);
2211
2212                         /* m is freed by bstp_input */
2213                         m = NULL;
2214                         goto out;
2215                 }
2216
2217                 if (bif->bif_flags & IFBIF_STP) {
2218                         switch (bif->bif_state) {
2219                         case BSTP_IFSTATE_BLOCKING:
2220                         case BSTP_IFSTATE_LISTENING:
2221                         case BSTP_IFSTATE_DISABLED:
2222                                 goto out;
2223                         }
2224                 }
2225
2226                 /*
2227                  * Make a deep copy of the packet and enqueue the copy
2228                  * for bridge processing; return the original packet for
2229                  * local processing.
2230                  */
2231                 mc = m_dup(m, MB_DONTWAIT);
2232                 if (mc == NULL)
2233                         goto out;
2234
2235                 bridge_forward(sc, mc);
2236
2237                 /*
2238                  * Reinject the mbuf as arriving on the bridge so we have a
2239                  * chance at claiming multicast packets. We can not loop back
2240                  * here from ether_input as a bridge is never a member of a
2241                  * bridge.
2242                  */
2243                 KASSERT(bifp->if_bridge == NULL,
2244                         ("loop created in bridge_input"));
2245                 mc2 = m_dup(m, MB_DONTWAIT);
2246 #ifdef notyet
2247                 if (mc2 != NULL) {
2248                         /* Keep the layer3 header aligned */
2249                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2250                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2251                 }
2252 #endif
2253                 if (mc2 != NULL) {
2254                         /*
2255                          * Don't tap to bpf(4) again; we have
2256                          * already done the tapping.
2257                          */
2258                         ether_reinput_oncpu(bifp, mc2, 0);
2259                 }
2260
2261                 /* Return the original packet for local processing. */
2262                 goto out;
2263         }
2264
2265         if (bif->bif_flags & IFBIF_STP) {
2266                 switch (bif->bif_state) {
2267                 case BSTP_IFSTATE_BLOCKING:
2268                 case BSTP_IFSTATE_LISTENING:
2269                 case BSTP_IFSTATE_DISABLED:
2270                         goto out;
2271                 }
2272         }
2273
2274         /*
2275          * Unicast.  Make sure it's not for us.
2276          *
2277          * This loop is MPSAFE; the only blocking operation (bridge_rtupdate)
2278          * is followed by breaking out of the loop.
2279          */
2280         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2281                 if (bif->bif_ifp->if_type != IFT_ETHER)
2282                         continue;
2283
2284                 /* It is destined for us. */
2285                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
2286                     ETHER_ADDR_LEN) == 0) {
2287                         if (bif->bif_ifp != ifp) {
2288                                 /* XXX loop prevention */
2289                                 m->m_flags |= M_ETHER_BRIDGED;
2290                                 new_ifp = bif->bif_ifp;
2291                         }
2292                         if (bif->bif_flags & IFBIF_LEARNING) {
2293                                 bridge_rtupdate(sc, eh->ether_shost,
2294                                                 ifp, IFBAF_DYNAMIC);
2295                         }
2296                         goto out;
2297                 }
2298
2299                 /* We just received a packet that we sent out. */
2300                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
2301                     ETHER_ADDR_LEN) == 0) {
2302                         m_freem(m);
2303                         m = NULL;
2304                         goto out;
2305                 }
2306         }
2307
2308         /* Perform the bridge forwarding function. */
2309         bridge_forward(sc, m);
2310         m = NULL;
2311 out:
2312         if (new_ifp != NULL) {
2313                 ether_reinput_oncpu(new_ifp, m, 1);
2314                 m = NULL;
2315         }
2316         return (m);
2317 }
2318
2319 /*
2320  * bridge_start_bcast:
2321  *
2322  *      Broadcast the packet sent from bridge to all member
2323  *      interfaces.
2324  *      This is a simplified version of bridge_broadcast(), however,
2325  *      this function expects caller to hold bridge's serializer.
2326  */
2327 static void
2328 bridge_start_bcast(struct bridge_softc *sc, struct mbuf *m)
2329 {
2330         struct bridge_iflist *bif;
2331         struct mbuf *mc;
2332         struct ifnet *dst_if, *bifp;
2333         int used = 0;
2334
2335         bifp = sc->sc_ifp;
2336         ASSERT_IFNET_SERIALIZED_ALL(bifp);
2337
2338         /*
2339          * Following loop is MPSAFE; nothing is blocking
2340          * in the loop body.
2341          */
2342         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2343                 dst_if = bif->bif_ifp;
2344
2345                 if (bif->bif_flags & IFBIF_STP) {
2346                         switch (bif->bif_state) {
2347                         case BSTP_IFSTATE_BLOCKING:
2348                         case BSTP_IFSTATE_DISABLED:
2349                                 continue;
2350                         }
2351                 }
2352
2353                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2354                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2355                         continue;
2356
2357                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2358                         continue;
2359
2360                 if (LIST_NEXT(bif, bif_next) == NULL) {
2361                         mc = m;
2362                         used = 1;
2363                 } else {
2364                         mc = m_copypacket(m, MB_DONTWAIT);
2365                         if (mc == NULL) {
2366                                 bifp->if_oerrors++;
2367                                 continue;
2368                         }
2369                 }
2370                 bridge_enqueue(dst_if, mc);
2371         }
2372         if (used == 0)
2373                 m_freem(m);
2374 }
2375
2376 /*
2377  * bridge_broadcast:
2378  *
2379  *      Send a frame to all interfaces that are members of
2380  *      the bridge, except for the one on which the packet
2381  *      arrived.
2382  */
2383 static void
2384 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2385     struct mbuf *m)
2386 {
2387         struct bridge_iflist *bif, *nbif;
2388         struct mbuf *mc;
2389         struct ifnet *dst_if, *bifp;
2390         int used = 0;
2391
2392         bifp = sc->sc_ifp;
2393         ASSERT_IFNET_NOT_SERIALIZED_ALL(bifp);
2394
2395         if (inet_pfil_hook.ph_hashooks > 0
2396 #ifdef INET6
2397             || inet6_pfil_hook.ph_hashooks > 0
2398 #endif
2399             ) {
2400                 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0)
2401                         return;
2402                 if (m == NULL)
2403                         return;
2404
2405                 /* Filter on the bridge interface before broadcasting */
2406                 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0)
2407                         return;
2408                 if (m == NULL)
2409                         return;
2410         }
2411
2412         LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
2413                 dst_if = bif->bif_ifp;
2414                 if (dst_if == src_if)
2415                         continue;
2416
2417                 if (bif->bif_flags & IFBIF_STP) {
2418                         switch (bif->bif_state) {
2419                         case BSTP_IFSTATE_BLOCKING:
2420                         case BSTP_IFSTATE_DISABLED:
2421                                 continue;
2422                         }
2423                 }
2424
2425                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2426                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2427                         continue;
2428
2429                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2430                         continue;
2431
2432                 if (LIST_NEXT(bif, bif_next) == NULL) {
2433                         mc = m;
2434                         used = 1;
2435                 } else {
2436                         mc = m_copypacket(m, MB_DONTWAIT);
2437                         if (mc == NULL) {
2438                                 sc->sc_ifp->if_oerrors++;
2439                                 continue;
2440                         }
2441                 }
2442
2443                 /*
2444                  * Filter on the output interface.  Pass a NULL bridge
2445                  * interface pointer so we do not redundantly filter on
2446                  * the bridge for each interface we broadcast on.
2447                  */
2448                 if (inet_pfil_hook.ph_hashooks > 0
2449 #ifdef INET6
2450                     || inet6_pfil_hook.ph_hashooks > 0
2451 #endif
2452                     ) {
2453                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2454                                 continue;
2455                         if (mc == NULL)
2456                                 continue;
2457                 }
2458                 bridge_handoff(dst_if, mc);
2459
2460                 if (nbif != NULL && !nbif->bif_onlist) {
2461                         KKASSERT(bif->bif_onlist);
2462                         nbif = LIST_NEXT(bif, bif_next);
2463                 }
2464         }
2465         if (used == 0)
2466                 m_freem(m);
2467 }
2468
2469 /*
2470  * bridge_span:
2471  *
2472  *      Duplicate a packet out one or more interfaces that are in span mode,
2473  *      the original mbuf is unmodified.
2474  */
2475 static void
2476 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2477 {
2478         struct bridge_iflist *bif;
2479         struct ifnet *dst_if, *bifp;
2480         struct mbuf *mc;
2481
2482         bifp = sc->sc_ifp;
2483         ifnet_serialize_all(bifp);
2484
2485         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2486                 dst_if = bif->bif_ifp;
2487
2488                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2489                         continue;
2490
2491                 mc = m_copypacket(m, MB_DONTWAIT);
2492                 if (mc == NULL) {
2493                         sc->sc_ifp->if_oerrors++;
2494                         continue;
2495                 }
2496                 bridge_enqueue(dst_if, mc);
2497         }
2498
2499         ifnet_deserialize_all(bifp);
2500 }
2501
2502 static void
2503 bridge_rtmsg_sync_handler(netmsg_t msg)
2504 {
2505         ifnet_forwardmsg(&msg->lmsg, mycpuid + 1);
2506 }
2507
2508 static void
2509 bridge_rtmsg_sync(struct bridge_softc *sc)
2510 {
2511         struct netmsg_base msg;
2512
2513         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2514
2515         netmsg_init(&msg, NULL, &curthread->td_msgport,
2516                     0, bridge_rtmsg_sync_handler);
2517         ifnet_domsg(&msg.lmsg, 0);
2518 }
2519
2520 static __inline void
2521 bridge_rtinfo_update(struct bridge_rtinfo *bri, struct ifnet *dst_if,
2522                      int setflags, uint8_t flags, uint32_t timeo)
2523 {
2524         if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2525             bri->bri_ifp != dst_if)
2526                 bri->bri_ifp = dst_if;
2527         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2528             bri->bri_expire != time_second + timeo)
2529                 bri->bri_expire = time_second + timeo;
2530         if (setflags)
2531                 bri->bri_flags = flags;
2532 }
2533
2534 static int
2535 bridge_rtinstall_oncpu(struct bridge_softc *sc, const uint8_t *dst,
2536                        struct ifnet *dst_if, int setflags, uint8_t flags,
2537                        struct bridge_rtinfo **bri0)
2538 {
2539         struct bridge_rtnode *brt;
2540         struct bridge_rtinfo *bri;
2541
2542         if (mycpuid == 0) {
2543                 brt = bridge_rtnode_lookup(sc, dst);
2544                 if (brt != NULL) {
2545                         /*
2546                          * rtnode for 'dst' already exists.  We inform the
2547                          * caller about this by leaving bri0 as NULL.  The
2548                          * caller will terminate the intallation upon getting
2549                          * NULL bri0.  However, we still need to update the
2550                          * rtinfo.
2551                          */
2552                         KKASSERT(*bri0 == NULL);
2553
2554                         /* Update rtinfo */
2555                         bridge_rtinfo_update(brt->brt_info, dst_if, setflags,
2556                                              flags, sc->sc_brttimeout);
2557                         return 0;
2558                 }
2559
2560                 /*
2561                  * We only need to check brtcnt on CPU0, since if limit
2562                  * is to be exceeded, ENOSPC is returned.  Caller knows
2563                  * this and will terminate the installation.
2564                  */
2565                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2566                         return ENOSPC;
2567
2568                 KKASSERT(*bri0 == NULL);
2569                 bri = kmalloc(sizeof(struct bridge_rtinfo), M_DEVBUF,
2570                                   M_WAITOK | M_ZERO);
2571                 *bri0 = bri;
2572
2573                 /* Setup rtinfo */
2574                 bri->bri_flags = IFBAF_DYNAMIC;
2575                 bridge_rtinfo_update(bri, dst_if, setflags, flags,
2576                                      sc->sc_brttimeout);
2577         } else {
2578                 bri = *bri0;
2579                 KKASSERT(bri != NULL);
2580         }
2581
2582         brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF,
2583                       M_WAITOK | M_ZERO);
2584         memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2585         brt->brt_info = bri;
2586
2587         bridge_rtnode_insert(sc, brt);
2588         return 0;
2589 }
2590
2591 static void
2592 bridge_rtinstall_handler(netmsg_t msg)
2593 {
2594         struct netmsg_brsaddr *brmsg = (struct netmsg_brsaddr *)msg;
2595         int error;
2596
2597         error = bridge_rtinstall_oncpu(brmsg->br_softc,
2598                                        brmsg->br_dst, brmsg->br_dst_if,
2599                                        brmsg->br_setflags, brmsg->br_flags,
2600                                        &brmsg->br_rtinfo);
2601         if (error) {
2602                 KKASSERT(mycpuid == 0 && brmsg->br_rtinfo == NULL);
2603                 lwkt_replymsg(&brmsg->base.lmsg, error);
2604                 return;
2605         } else if (brmsg->br_rtinfo == NULL) {
2606                 /* rtnode already exists for 'dst' */
2607                 KKASSERT(mycpuid == 0);
2608                 lwkt_replymsg(&brmsg->base.lmsg, 0);
2609                 return;
2610         }
2611         ifnet_forwardmsg(&brmsg->base.lmsg, mycpuid + 1);
2612 }
2613
2614 /*
2615  * bridge_rtupdate:
2616  *
2617  *      Add/Update a bridge routing entry.
2618  */
2619 static int
2620 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2621                 struct ifnet *dst_if, uint8_t flags)
2622 {
2623         struct bridge_rtnode *brt;
2624
2625         /*
2626          * A route for this destination might already exist.  If so,
2627          * update it, otherwise create a new one.
2628          */
2629         if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
2630                 struct netmsg_brsaddr *brmsg;
2631
2632                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2633                         return ENOSPC;
2634
2635                 brmsg = kmalloc(sizeof(*brmsg), M_LWKTMSG, M_WAITOK | M_NULLOK);
2636                 if (brmsg == NULL)
2637                         return ENOMEM;
2638
2639                 netmsg_init(&brmsg->base, NULL, &netisr_afree_rport,
2640                             0, bridge_rtinstall_handler);
2641                 memcpy(brmsg->br_dst, dst, ETHER_ADDR_LEN);
2642                 brmsg->br_dst_if = dst_if;
2643                 brmsg->br_flags = flags;
2644                 brmsg->br_setflags = 0;
2645                 brmsg->br_softc = sc;
2646                 brmsg->br_rtinfo = NULL;
2647
2648                 ifnet_sendmsg(&brmsg->base.lmsg, 0);
2649                 return 0;
2650         }
2651         bridge_rtinfo_update(brt->brt_info, dst_if, 0, flags,
2652                              sc->sc_brttimeout);
2653         return 0;
2654 }
2655
2656 static int
2657 bridge_rtsaddr(struct bridge_softc *sc, const uint8_t *dst,
2658                struct ifnet *dst_if, uint8_t flags)
2659 {
2660         struct netmsg_brsaddr brmsg;
2661
2662         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2663
2664         netmsg_init(&brmsg.base, NULL, &curthread->td_msgport,
2665                     0, bridge_rtinstall_handler);
2666         memcpy(brmsg.br_dst, dst, ETHER_ADDR_LEN);
2667         brmsg.br_dst_if = dst_if;
2668         brmsg.br_flags = flags;
2669         brmsg.br_setflags = 1;
2670         brmsg.br_softc = sc;
2671         brmsg.br_rtinfo = NULL;
2672
2673         return ifnet_domsg(&brmsg.base.lmsg, 0);
2674 }
2675
2676 /*
2677  * bridge_rtlookup:
2678  *
2679  *      Lookup the destination interface for an address.
2680  */
2681 static struct ifnet *
2682 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2683 {
2684         struct bridge_rtnode *brt;
2685
2686         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2687                 return NULL;
2688         return brt->brt_info->bri_ifp;
2689 }
2690
2691 static void
2692 bridge_rtreap_handler(netmsg_t msg)
2693 {
2694         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
2695         struct bridge_rtnode *brt, *nbrt;
2696
2697         LIST_FOREACH_MUTABLE(brt, &sc->sc_rtlists[mycpuid], brt_list, nbrt) {
2698                 if (brt->brt_info->bri_dead)
2699                         bridge_rtnode_destroy(sc, brt);
2700         }
2701         ifnet_forwardmsg(&msg->lmsg, mycpuid + 1);
2702 }
2703
2704 static void
2705 bridge_rtreap(struct bridge_softc *sc)
2706 {
2707         struct netmsg_base msg;
2708
2709         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2710
2711         netmsg_init(&msg, NULL, &curthread->td_msgport,
2712                     0, bridge_rtreap_handler);
2713         msg.lmsg.u.ms_resultp = sc;
2714
2715         ifnet_domsg(&msg.lmsg, 0);
2716 }
2717
2718 static void
2719 bridge_rtreap_async(struct bridge_softc *sc)
2720 {
2721         struct netmsg_base *msg;
2722
2723         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_WAITOK);
2724
2725         netmsg_init(msg, NULL, &netisr_afree_rport,
2726                     0, bridge_rtreap_handler);
2727         msg->lmsg.u.ms_resultp = sc;
2728
2729         ifnet_sendmsg(&msg->lmsg, 0);
2730 }
2731
2732 /*
2733  * bridge_rttrim:
2734  *
2735  *      Trim the routine table so that we have a number
2736  *      of routing entries less than or equal to the
2737  *      maximum number.
2738  */
2739 static void
2740 bridge_rttrim(struct bridge_softc *sc)
2741 {
2742         struct bridge_rtnode *brt;
2743         int dead;
2744
2745         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2746
2747         /* Make sure we actually need to do this. */
2748         if (sc->sc_brtcnt <= sc->sc_brtmax)
2749                 return;
2750
2751         /*
2752          * Find out how many rtnodes are dead
2753          */
2754         dead = bridge_rtage_finddead(sc);
2755         KKASSERT(dead <= sc->sc_brtcnt);
2756
2757         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2758                 /* Enough dead rtnodes are found */
2759                 bridge_rtreap(sc);
2760                 return;
2761         }
2762
2763         /*
2764          * Kill some dynamic rtnodes to meet the brtmax
2765          */
2766         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2767                 struct bridge_rtinfo *bri = brt->brt_info;
2768
2769                 if (bri->bri_dead) {
2770                         /*
2771                          * We have counted this rtnode in
2772                          * bridge_rtage_finddead()
2773                          */
2774                         continue;
2775                 }
2776
2777                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2778                         bri->bri_dead = 1;
2779                         ++dead;
2780                         KKASSERT(dead <= sc->sc_brtcnt);
2781
2782                         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2783                                 /* Enough rtnodes are collected */
2784                                 break;
2785                         }
2786                 }
2787         }
2788         if (dead)
2789                 bridge_rtreap(sc);
2790 }
2791
2792 /*
2793  * bridge_timer:
2794  *
2795  *      Aging timer for the bridge.
2796  */
2797 static void
2798 bridge_timer(void *arg)
2799 {
2800         struct bridge_softc *sc = arg;
2801         struct netmsg_base *msg;
2802
2803         KKASSERT(mycpuid == BRIDGE_CFGCPU);
2804
2805         crit_enter();
2806
2807         if (callout_pending(&sc->sc_brcallout) ||
2808             !callout_active(&sc->sc_brcallout)) {
2809                 crit_exit();
2810                 return;
2811         }
2812         callout_deactivate(&sc->sc_brcallout);
2813
2814         msg = &sc->sc_brtimemsg;
2815         KKASSERT(msg->lmsg.ms_flags & MSGF_DONE);
2816         lwkt_sendmsg(BRIDGE_CFGPORT, &msg->lmsg);
2817
2818         crit_exit();
2819 }
2820
2821 static void
2822 bridge_timer_handler(netmsg_t msg)
2823 {
2824         struct bridge_softc *sc = msg->lmsg.u.ms_resultp;
2825
2826         KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
2827
2828         crit_enter();
2829         /* Reply ASAP */
2830         lwkt_replymsg(&msg->lmsg, 0);
2831         crit_exit();
2832
2833         bridge_rtage(sc);
2834         if (sc->sc_ifp->if_flags & IFF_RUNNING) {
2835                 callout_reset(&sc->sc_brcallout,
2836                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2837         }
2838 }
2839
2840 static int
2841 bridge_rtage_finddead(struct bridge_softc *sc)
2842 {
2843         struct bridge_rtnode *brt;
2844         int dead = 0;
2845
2846         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2847                 struct bridge_rtinfo *bri = brt->brt_info;
2848
2849                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2850                     time_second >= bri->bri_expire) {
2851                         bri->bri_dead = 1;
2852                         ++dead;
2853                         KKASSERT(dead <= sc->sc_brtcnt);
2854                 }
2855         }
2856         return dead;
2857 }
2858
2859 /*
2860  * bridge_rtage:
2861  *
2862  *      Perform an aging cycle.
2863  */
2864 static void
2865 bridge_rtage(struct bridge_softc *sc)
2866 {
2867         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2868
2869         if (bridge_rtage_finddead(sc))
2870                 bridge_rtreap(sc);
2871 }
2872
2873 /*
2874  * bridge_rtflush:
2875  *
2876  *      Remove all dynamic addresses from the bridge.
2877  */
2878 static void
2879 bridge_rtflush(struct bridge_softc *sc, int bf)
2880 {
2881         struct bridge_rtnode *brt;
2882         int reap;
2883
2884         reap = 0;
2885         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2886                 struct bridge_rtinfo *bri = brt->brt_info;
2887
2888                 if ((bf & IFBF_FLUSHALL) ||
2889                     (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2890                         bri->bri_dead = 1;
2891                         reap = 1;
2892                 }
2893         }
2894         if (reap) {
2895                 if (bf & IFBF_FLUSHSYNC)
2896                         bridge_rtreap(sc);
2897                 else
2898                         bridge_rtreap_async(sc);
2899         }
2900 }
2901
2902 /*
2903  * bridge_rtdaddr:
2904  *
2905  *      Remove an address from the table.
2906  */
2907 static int
2908 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2909 {
2910         struct bridge_rtnode *brt;
2911
2912         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2913
2914         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2915                 return (ENOENT);
2916
2917         /* TODO: add a cheaper delete operation */
2918         brt->brt_info->bri_dead = 1;
2919         bridge_rtreap(sc);
2920         return (0);
2921 }
2922
2923 /*
2924  * bridge_rtdelete:
2925  *
2926  *      Delete routes to a speicifc member interface.
2927  */
2928 void
2929 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int bf)
2930 {
2931         struct bridge_rtnode *brt;
2932         int reap;
2933
2934         reap = 0;
2935         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2936                 struct bridge_rtinfo *bri = brt->brt_info;
2937
2938                 if (bri->bri_ifp == ifp &&
2939                     ((bf & IFBF_FLUSHALL) ||
2940                      (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
2941                         bri->bri_dead = 1;
2942                         reap = 1;
2943                 }
2944         }
2945         if (reap) {
2946                 if (bf & IFBF_FLUSHSYNC)
2947                         bridge_rtreap(sc);
2948                 else
2949                         bridge_rtreap_async(sc);
2950         }
2951 }
2952
2953 /*
2954  * bridge_rtable_init:
2955  *
2956  *      Initialize the route table for this bridge.
2957  */
2958 static void
2959 bridge_rtable_init(struct bridge_softc *sc)
2960 {
2961         int cpu;
2962
2963         /*
2964          * Initialize per-cpu hash tables
2965          */
2966         sc->sc_rthashs = kmalloc(sizeof(*sc->sc_rthashs) * ncpus,
2967                                  M_DEVBUF, M_WAITOK);
2968         for (cpu = 0; cpu < ncpus; ++cpu) {
2969                 int i;
2970
2971                 sc->sc_rthashs[cpu] =
2972                 kmalloc(sizeof(struct bridge_rtnode_head) * BRIDGE_RTHASH_SIZE,
2973                         M_DEVBUF, M_WAITOK);
2974
2975                 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2976                         LIST_INIT(&sc->sc_rthashs[cpu][i]);
2977         }
2978         sc->sc_rthash_key = karc4random();
2979
2980         /*
2981          * Initialize per-cpu lists
2982          */
2983         sc->sc_rtlists = kmalloc(sizeof(struct bridge_rtnode_head) * ncpus,
2984                                  M_DEVBUF, M_WAITOK);
2985         for (cpu = 0; cpu < ncpus; ++cpu)
2986                 LIST_INIT(&sc->sc_rtlists[cpu]);
2987 }
2988
2989 /*
2990  * bridge_rtable_fini:
2991  *
2992  *      Deconstruct the route table for this bridge.
2993  */
2994 static void
2995 bridge_rtable_fini(struct bridge_softc *sc)
2996 {
2997         int cpu;
2998
2999         /*
3000          * Free per-cpu hash tables
3001          */
3002         for (cpu = 0; cpu < ncpus; ++cpu)
3003                 kfree(sc->sc_rthashs[cpu], M_DEVBUF);
3004         kfree(sc->sc_rthashs, M_DEVBUF);
3005
3006         /*
3007          * Free per-cpu lists
3008          */
3009         kfree(sc->sc_rtlists, M_DEVBUF);
3010 }
3011
3012 /*
3013  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3014  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3015  */
3016 #define mix(a, b, c)                                                    \
3017 do {                                                                    \
3018         a -= b; a -= c; a ^= (c >> 13);                                 \
3019         b -= c; b -= a; b ^= (a << 8);                                  \
3020         c -= a; c -= b; c ^= (b >> 13);                                 \
3021         a -= b; a -= c; a ^= (c >> 12);                                 \
3022         b -= c; b -= a; b ^= (a << 16);                                 \
3023         c -= a; c -= b; c ^= (b >> 5);                                  \
3024         a -= b; a -= c; a ^= (c >> 3);                                  \
3025         b -= c; b -= a; b ^= (a << 10);                                 \
3026         c -= a; c -= b; c ^= (b >> 15);                                 \
3027 } while (/*CONSTCOND*/0)
3028
3029 static __inline uint32_t
3030 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3031 {
3032         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3033
3034         b += addr[5] << 8;
3035         b += addr[4];
3036         a += addr[3] << 24;
3037         a += addr[2] << 16;
3038         a += addr[1] << 8;
3039         a += addr[0];
3040
3041         mix(a, b, c);
3042
3043         return (c & BRIDGE_RTHASH_MASK);
3044 }
3045
3046 #undef mix
3047
3048 static int
3049 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3050 {
3051         int i, d;
3052
3053         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3054                 d = ((int)a[i]) - ((int)b[i]);
3055         }
3056
3057         return (d);
3058 }
3059
3060 /*
3061  * bridge_rtnode_lookup:
3062  *
3063  *      Look up a bridge route node for the specified destination.
3064  */
3065 static struct bridge_rtnode *
3066 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
3067 {
3068         struct bridge_rtnode *brt;
3069         uint32_t hash;
3070         int dir;
3071
3072         hash = bridge_rthash(sc, addr);
3073         LIST_FOREACH(brt, &sc->sc_rthashs[mycpuid][hash], brt_hash) {
3074                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3075                 if (dir == 0)
3076                         return (brt);
3077                 if (dir > 0)
3078                         return (NULL);
3079         }
3080
3081         return (NULL);
3082 }
3083
3084 /*
3085  * bridge_rtnode_insert:
3086  *
3087  *      Insert the specified bridge node into the route table.
3088  *      Caller has to make sure that rtnode does not exist.
3089  */
3090 static void
3091 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3092 {
3093         struct bridge_rtnode *lbrt;
3094         uint32_t hash;
3095         int dir;
3096
3097         hash = bridge_rthash(sc, brt->brt_addr);
3098
3099         lbrt = LIST_FIRST(&sc->sc_rthashs[mycpuid][hash]);
3100         if (lbrt == NULL) {
3101                 LIST_INSERT_HEAD(&sc->sc_rthashs[mycpuid][hash], brt, brt_hash);
3102                 goto out;
3103         }
3104
3105         do {
3106                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3107                 KASSERT(dir != 0, ("rtnode already exist\n"));
3108
3109                 if (dir > 0) {
3110                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3111                         goto out;
3112                 }
3113                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
3114                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3115                         goto out;
3116                 }
3117                 lbrt = LIST_NEXT(lbrt, brt_hash);
3118         } while (lbrt != NULL);
3119
3120         panic("no suitable position found for rtnode\n");
3121 out:
3122         LIST_INSERT_HEAD(&sc->sc_rtlists[mycpuid], brt, brt_list);
3123         if (mycpuid == 0) {
3124                 /*
3125                  * Update the brtcnt.
3126                  * We only need to do it once and we do it on CPU0.
3127                  */
3128                 sc->sc_brtcnt++;
3129         }
3130 }
3131
3132 /*
3133  * bridge_rtnode_destroy:
3134  *
3135  *      Destroy a bridge rtnode.
3136  */
3137 static void
3138 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3139 {
3140         LIST_REMOVE(brt, brt_hash);
3141         LIST_REMOVE(brt, brt_list);
3142
3143         if (mycpuid + 1 == ncpus) {
3144                 /* Free rtinfo associated with rtnode on the last cpu */
3145                 kfree(brt->brt_info, M_DEVBUF);
3146         }
3147         kfree(brt, M_DEVBUF);
3148
3149         if (mycpuid == 0) {
3150                 /* Update brtcnt only on CPU0 */
3151                 sc->sc_brtcnt--;
3152         }
3153 }
3154
3155 static __inline int
3156 bridge_post_pfil(struct mbuf *m)
3157 {
3158         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED)
3159                 return EOPNOTSUPP;
3160
3161         /* Not yet */
3162         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED)
3163                 return EOPNOTSUPP;
3164
3165         return 0;
3166 }
3167
3168 /*
3169  * Send bridge packets through pfil if they are one of the types pfil can deal
3170  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3171  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3172  * that interface.
3173  */
3174 static int
3175 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3176 {
3177         int snap, error, i, hlen;
3178         struct ether_header *eh1, eh2;
3179         struct ip *ip;
3180         struct llc llc1;
3181         u_int16_t ether_type;
3182
3183         snap = 0;
3184         error = -1;     /* Default error if not error == 0 */
3185
3186         if (pfil_bridge == 0 && pfil_member == 0)
3187                 return (0); /* filtering is disabled */
3188
3189         i = min((*mp)->m_pkthdr.len, max_protohdr);
3190         if ((*mp)->m_len < i) {
3191                 *mp = m_pullup(*mp, i);
3192                 if (*mp == NULL) {
3193                         kprintf("%s: m_pullup failed\n", __func__);
3194                         return (-1);
3195                 }
3196         }
3197
3198         eh1 = mtod(*mp, struct ether_header *);
3199         ether_type = ntohs(eh1->ether_type);
3200
3201         /*
3202          * Check for SNAP/LLC.
3203          */
3204         if (ether_type < ETHERMTU) {
3205                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3206
3207                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3208                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3209                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3210                     llc2->llc_control == LLC_UI) {
3211                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3212                         snap = 1;
3213                 }
3214         }
3215
3216         /*
3217          * If we're trying to filter bridge traffic, don't look at anything
3218          * other than IP and ARP traffic.  If the filter doesn't understand
3219          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3220          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3221          * but of course we don't have an AppleTalk filter to begin with.
3222          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3223          * ARP traffic.)
3224          */
3225         switch (ether_type) {
3226         case ETHERTYPE_ARP:
3227         case ETHERTYPE_REVARP:
3228                 return (0); /* Automatically pass */
3229
3230         case ETHERTYPE_IP:
3231 #ifdef INET6
3232         case ETHERTYPE_IPV6:
3233 #endif /* INET6 */
3234                 break;
3235
3236         default:
3237                 /*
3238                  * Check to see if the user wants to pass non-ip
3239                  * packets, these will not be checked by pfil(9)
3240                  * and passed unconditionally so the default is to drop.
3241                  */
3242                 if (pfil_onlyip)
3243                         goto bad;
3244         }
3245
3246         /* Strip off the Ethernet header and keep a copy. */
3247         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3248         m_adj(*mp, ETHER_HDR_LEN);
3249
3250         /* Strip off snap header, if present */
3251         if (snap) {
3252                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3253                 m_adj(*mp, sizeof(struct llc));
3254         }
3255
3256         /*
3257          * Check the IP header for alignment and errors
3258          */
3259         if (dir == PFIL_IN) {
3260                 switch (ether_type) {
3261                 case ETHERTYPE_IP:
3262                         error = bridge_ip_checkbasic(mp);
3263                         break;
3264 #ifdef INET6
3265                 case ETHERTYPE_IPV6:
3266                         error = bridge_ip6_checkbasic(mp);
3267                         break;
3268 #endif /* INET6 */
3269                 default:
3270                         error = 0;
3271                 }
3272                 if (error)
3273                         goto bad;
3274         }
3275
3276         error = 0;
3277
3278         /*
3279          * Run the packet through pfil
3280          */
3281         switch (ether_type) {
3282         case ETHERTYPE_IP:
3283                 /*
3284                  * before calling the firewall, swap fields the same as
3285                  * IP does. here we assume the header is contiguous
3286                  */
3287                 ip = mtod(*mp, struct ip *);
3288
3289                 ip->ip_len = ntohs(ip->ip_len);
3290                 ip->ip_off = ntohs(ip->ip_off);
3291
3292                 /*
3293                  * Run pfil on the member interface and the bridge, both can
3294                  * be skipped by clearing pfil_member or pfil_bridge.
3295                  *
3296                  * Keep the order:
3297                  *   in_if -> bridge_if -> out_if
3298                  */
3299                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) {
3300                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3301                         if (*mp == NULL || error != 0) /* filter may consume */
3302                                 break;
3303                         error = bridge_post_pfil(*mp);
3304                         if (error)
3305                                 break;
3306                 }
3307
3308                 if (pfil_member && ifp != NULL) {
3309                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp, dir);
3310                         if (*mp == NULL || error != 0) /* filter may consume */
3311                                 break;
3312                         error = bridge_post_pfil(*mp);
3313                         if (error)
3314                                 break;
3315                 }
3316
3317                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) {
3318                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3319                         if (*mp == NULL || error != 0) /* filter may consume */
3320                                 break;
3321                         error = bridge_post_pfil(*mp);
3322                         if (error)
3323                                 break;
3324                 }
3325
3326                 /* check if we need to fragment the packet */
3327                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3328                         i = (*mp)->m_pkthdr.len;
3329                         if (i > ifp->if_mtu) {
3330                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3331                                             &llc1);
3332                                 return (error);
3333                         }
3334                 }
3335
3336                 /* Recalculate the ip checksum and restore byte ordering */
3337                 ip = mtod(*mp, struct ip *);
3338                 hlen = ip->ip_hl << 2;
3339                 if (hlen < sizeof(struct ip))
3340                         goto bad;
3341                 if (hlen > (*mp)->m_len) {
3342                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3343                                 goto bad;
3344                         ip = mtod(*mp, struct ip *);
3345                         if (ip == NULL)
3346                                 goto bad;
3347                 }
3348                 ip->ip_len = htons(ip->ip_len);
3349                 ip->ip_off = htons(ip->ip_off);
3350                 ip->ip_sum = 0;
3351                 if (hlen == sizeof(struct ip))
3352                         ip->ip_sum = in_cksum_hdr(ip);
3353                 else
3354                         ip->ip_sum = in_cksum(*mp, hlen);
3355
3356                 break;
3357 #ifdef INET6
3358         case ETHERTYPE_IPV6:
3359                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3360                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3361                                         dir);
3362
3363                 if (*mp == NULL || error != 0) /* filter may consume */
3364                         break;
3365
3366                 if (pfil_member && ifp != NULL)
3367                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3368                                         dir);
3369
3370                 if (*mp == NULL || error != 0) /* filter may consume */
3371                         break;
3372
3373                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3374                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3375                                         dir);
3376                 break;
3377 #endif
3378         default:
3379                 error = 0;
3380                 break;
3381         }
3382
3383         if (*mp == NULL)
3384                 return (error);
3385         if (error != 0)
3386                 goto bad;
3387
3388         error = -1;
3389
3390         /*
3391          * Finally, put everything back the way it was and return
3392          */
3393         if (snap) {
3394                 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT);
3395                 if (*mp == NULL)
3396                         return (error);
3397                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3398         }
3399
3400         M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT);
3401         if (*mp == NULL)
3402                 return (error);
3403         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3404
3405         return (0);
3406
3407 bad:
3408         m_freem(*mp);
3409         *mp = NULL;
3410         return (error);
3411 }
3412
3413 /*
3414  * Perform basic checks on header size since
3415  * pfil assumes ip_input has already processed
3416  * it for it.  Cut-and-pasted from ip_input.c.
3417  * Given how simple the IPv6 version is,
3418  * does the IPv4 version really need to be
3419  * this complicated?
3420  *
3421  * XXX Should we update ipstat here, or not?
3422  * XXX Right now we update ipstat but not
3423  * XXX csum_counter.
3424  */
3425 static int
3426 bridge_ip_checkbasic(struct mbuf **mp)
3427 {
3428         struct mbuf *m = *mp;
3429         struct ip *ip;
3430         int len, hlen;
3431         u_short sum;
3432
3433         if (*mp == NULL)
3434                 return (-1);
3435 #if notyet
3436         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3437                 if ((m = m_copyup(m, sizeof(struct ip),
3438                         (max_linkhdr + 3) & ~3)) == NULL) {
3439                         /* XXXJRT new stat, please */
3440                         ipstat.ips_toosmall++;
3441                         goto bad;
3442                 }
3443         } else
3444 #endif
3445 #ifndef __predict_false
3446 #define __predict_false(x) x
3447 #endif
3448          if (__predict_false(m->m_len < sizeof (struct ip))) {
3449                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3450                         ipstat.ips_toosmall++;
3451                         goto bad;
3452                 }
3453         }
3454         ip = mtod(m, struct ip *);
3455         if (ip == NULL) goto bad;
3456
3457         if (ip->ip_v != IPVERSION) {
3458                 ipstat.ips_badvers++;
3459                 goto bad;
3460         }
3461         hlen = ip->ip_hl << 2;
3462         if (hlen < sizeof(struct ip)) { /* minimum header length */
3463                 ipstat.ips_badhlen++;
3464                 goto bad;
3465         }
3466         if (hlen > m->m_len) {
3467                 if ((m = m_pullup(m, hlen)) == 0) {
3468                         ipstat.ips_badhlen++;
3469                         goto bad;
3470                 }
3471                 ip = mtod(m, struct ip *);
3472                 if (ip == NULL) goto bad;
3473         }
3474
3475         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3476                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3477         } else {
3478                 if (hlen == sizeof(struct ip)) {
3479                         sum = in_cksum_hdr(ip);
3480                 } else {
3481                         sum = in_cksum(m, hlen);
3482                 }
3483         }
3484         if (sum) {
3485                 ipstat.ips_badsum++;
3486                 goto bad;
3487         }
3488
3489         /* Retrieve the packet length. */
3490         len = ntohs(ip->ip_len);
3491
3492         /*
3493          * Check for additional length bogosity
3494          */
3495         if (len < hlen) {
3496                 ipstat.ips_badlen++;
3497                 goto bad;
3498         }
3499
3500         /*
3501          * Check that the amount of data in the buffers
3502          * is as at least much as the IP header would have us expect.
3503          * Drop packet if shorter than we expect.
3504          */
3505         if (m->m_pkthdr.len < len) {
3506                 ipstat.ips_tooshort++;
3507                 goto bad;
3508         }
3509
3510         /* Checks out, proceed */
3511         *mp = m;
3512         return (0);
3513
3514 bad:
3515         *mp = m;
3516         return (-1);
3517 }
3518
3519 #ifdef INET6
3520 /*
3521  * Same as above, but for IPv6.
3522  * Cut-and-pasted from ip6_input.c.
3523  * XXX Should we update ip6stat, or not?
3524  */
3525 static int
3526 bridge_ip6_checkbasic(struct mbuf **mp)
3527 {
3528         struct mbuf *m = *mp;
3529         struct ip6_hdr *ip6;
3530
3531         /*
3532          * If the IPv6 header is not aligned, slurp it up into a new
3533          * mbuf with space for link headers, in the event we forward
3534          * it.  Otherwise, if it is aligned, make sure the entire base
3535          * IPv6 header is in the first mbuf of the chain.
3536          */
3537 #if notyet
3538         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3539                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3540                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3541                             (max_linkhdr + 3) & ~3)) == NULL) {
3542                         /* XXXJRT new stat, please */
3543                         ip6stat.ip6s_toosmall++;
3544                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3545                         goto bad;
3546                 }
3547         } else
3548 #endif
3549         if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3550                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3551                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3552                         ip6stat.ip6s_toosmall++;
3553                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3554                         goto bad;
3555                 }
3556         }
3557
3558         ip6 = mtod(m, struct ip6_hdr *);
3559
3560         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3561                 ip6stat.ip6s_badvers++;
3562                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3563                 goto bad;
3564         }
3565
3566         /* Checks out, proceed */
3567         *mp = m;
3568         return (0);
3569
3570 bad:
3571         *mp = m;
3572         return (-1);
3573 }
3574 #endif /* INET6 */
3575
3576 /*
3577  * bridge_fragment:
3578  *
3579  *      Return a fragmented mbuf chain.
3580  */
3581 static int
3582 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3583     int snap, struct llc *llc)
3584 {
3585         struct mbuf *m0;
3586         struct ip *ip;
3587         int error = -1;
3588
3589         if (m->m_len < sizeof(struct ip) &&
3590             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3591                 goto out;
3592         ip = mtod(m, struct ip *);
3593
3594         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3595                     CSUM_DELAY_IP);
3596         if (error)
3597                 goto out;
3598
3599         /* walk the chain and re-add the Ethernet header */
3600         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3601                 if (error == 0) {
3602                         if (snap) {
3603                                 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT);
3604                                 if (m0 == NULL) {
3605                                         error = ENOBUFS;
3606                                         continue;
3607                                 }
3608                                 bcopy(llc, mtod(m0, caddr_t),
3609                                     sizeof(struct llc));
3610                         }
3611                         M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT);
3612                         if (m0 == NULL) {
3613                                 error = ENOBUFS;
3614                                 continue;
3615                         }
3616                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3617                 } else
3618                         m_freem(m);
3619         }
3620
3621         if (error == 0)
3622                 ipstat.ips_fragmented++;
3623
3624         return (error);
3625
3626 out:
3627         if (m != NULL)
3628                 m_freem(m);
3629         return (error);
3630 }
3631
3632 static void
3633 bridge_enqueue_handler(netmsg_t msg)
3634 {
3635         struct netmsg_packet *nmp;
3636         struct ifnet *dst_ifp;
3637         struct mbuf *m;
3638
3639         nmp = &msg->packet;
3640         m = nmp->nm_packet;
3641         dst_ifp = nmp->base.lmsg.u.ms_resultp;
3642
3643         bridge_handoff(dst_ifp, m);
3644 }
3645
3646 static void
3647 bridge_handoff(struct ifnet *dst_ifp, struct mbuf *m)
3648 {
3649         struct mbuf *m0;
3650
3651         /* We may be sending a fragment so traverse the mbuf */
3652         for (; m; m = m0) {
3653                 struct altq_pktattr pktattr;
3654
3655                 m0 = m->m_nextpkt;
3656                 m->m_nextpkt = NULL;
3657
3658                 if (ifq_is_enabled(&dst_ifp->if_snd))
3659                         altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
3660
3661                 ifq_dispatch(dst_ifp, m, &pktattr);
3662         }
3663 }
3664
3665 static void
3666 bridge_control_dispatch(netmsg_t msg)
3667 {
3668         struct netmsg_brctl *bc_msg = (struct netmsg_brctl *)msg;
3669         struct ifnet *bifp = bc_msg->bc_sc->sc_ifp;
3670         int error;
3671
3672         ifnet_serialize_all(bifp);
3673         error = bc_msg->bc_func(bc_msg->bc_sc, bc_msg->bc_arg);
3674         ifnet_deserialize_all(bifp);
3675
3676         lwkt_replymsg(&bc_msg->base.lmsg, error);
3677 }
3678
3679 static int
3680 bridge_control(struct bridge_softc *sc, u_long cmd,
3681                bridge_ctl_t bc_func, void *bc_arg)
3682 {
3683         struct ifnet *bifp = sc->sc_ifp;
3684         struct netmsg_brctl bc_msg;
3685         int error;
3686
3687         ASSERT_IFNET_SERIALIZED_ALL(bifp);
3688
3689         bzero(&bc_msg, sizeof(bc_msg));
3690
3691         netmsg_init(&bc_msg.base, NULL, &curthread->td_msgport,
3692                     0, bridge_control_dispatch);
3693         bc_msg.bc_func = bc_func;
3694         bc_msg.bc_sc = sc;
3695         bc_msg.bc_arg = bc_arg;
3696
3697         ifnet_deserialize_all(bifp);
3698         error = lwkt_domsg(BRIDGE_CFGPORT, &bc_msg.base.lmsg, 0);
3699         ifnet_serialize_all(bifp);
3700         return error;
3701 }
3702
3703 static void
3704 bridge_add_bif_handler(netmsg_t msg)
3705 {
3706         struct netmsg_braddbif *amsg = (struct netmsg_braddbif *)msg;
3707         struct bridge_softc *sc;
3708         struct bridge_iflist *bif;
3709
3710         sc = amsg->br_softc;
3711
3712         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
3713         bif->bif_ifp = amsg->br_bif_ifp;
3714         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
3715         bif->bif_onlist = 1;
3716         bif->bif_info = amsg->br_bif_info;
3717
3718         LIST_INSERT_HEAD(&sc->sc_iflists[mycpuid], bif, bif_next);
3719
3720         ifnet_forwardmsg(&amsg->base.lmsg, mycpuid + 1);
3721 }
3722
3723 static void
3724 bridge_add_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3725                struct ifnet *ifp)
3726 {
3727         struct netmsg_braddbif amsg;
3728
3729         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3730
3731         netmsg_init(&amsg.base, NULL, &curthread->td_msgport,
3732                     0, bridge_add_bif_handler);
3733         amsg.br_softc = sc;
3734         amsg.br_bif_info = bif_info;
3735         amsg.br_bif_ifp = ifp;
3736
3737         ifnet_domsg(&amsg.base.lmsg, 0);
3738 }
3739
3740 static void
3741 bridge_del_bif_handler(netmsg_t msg)
3742 {
3743         struct netmsg_brdelbif *dmsg = (struct netmsg_brdelbif *)msg;
3744         struct bridge_softc *sc;
3745         struct bridge_iflist *bif;
3746
3747         sc = dmsg->br_softc;
3748
3749         /*
3750          * Locate the bif associated with the br_bif_info
3751          * on the current CPU
3752          */
3753         bif = bridge_lookup_member_ifinfo(sc, dmsg->br_bif_info);
3754         KKASSERT(bif != NULL && bif->bif_onlist);
3755
3756         /* Remove the bif from the current CPU's iflist */
3757         bif->bif_onlist = 0;
3758         LIST_REMOVE(bif, bif_next);
3759
3760         /* Save the removed bif for later freeing */
3761         LIST_INSERT_HEAD(dmsg->br_bif_list, bif, bif_next);
3762
3763         ifnet_forwardmsg(&dmsg->base.lmsg, mycpuid + 1);
3764 }
3765
3766 static void
3767 bridge_del_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3768                struct bridge_iflist_head *saved_bifs)
3769 {
3770         struct netmsg_brdelbif dmsg;
3771
3772         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3773
3774         netmsg_init(&dmsg.base, NULL, &curthread->td_msgport,
3775                     0, bridge_del_bif_handler);
3776         dmsg.br_softc = sc;
3777         dmsg.br_bif_info = bif_info;
3778         dmsg.br_bif_list = saved_bifs;
3779
3780         ifnet_domsg(&dmsg.base.lmsg, 0);
3781 }
3782
3783 static void
3784 bridge_set_bifflags_handler(netmsg_t msg)
3785 {
3786         struct netmsg_brsflags *smsg = (struct netmsg_brsflags *)msg;
3787         struct bridge_softc *sc;
3788         struct bridge_iflist *bif;
3789
3790         sc = smsg->br_softc;
3791
3792         /*
3793          * Locate the bif associated with the br_bif_info
3794          * on the current CPU
3795          */
3796         bif = bridge_lookup_member_ifinfo(sc, smsg->br_bif_info);
3797         KKASSERT(bif != NULL && bif->bif_onlist);
3798
3799         bif->bif_flags = smsg->br_bif_flags;
3800
3801         ifnet_forwardmsg(&smsg->base.lmsg, mycpuid + 1);
3802 }
3803
3804 static void
3805 bridge_set_bifflags(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3806                     uint32_t bif_flags)
3807 {
3808         struct netmsg_brsflags smsg;
3809
3810         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3811
3812         netmsg_init(&smsg.base, NULL, &curthread->td_msgport,
3813                     0, bridge_set_bifflags_handler);
3814         smsg.br_softc = sc;
3815         smsg.br_bif_info = bif_info;
3816         smsg.br_bif_flags = bif_flags;
3817
3818         ifnet_domsg(&smsg.base.lmsg, 0);
3819 }