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           bc_nmsg;
 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           br_nmsg;
 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           br_nmsg;
 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           br_nmsg;
 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           br_nmsg;
 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(struct netmsg *);
 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(struct netmsg *);
 398 static void     bridge_rtinstall_handler(struct netmsg *);
 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(struct netmsg *);
 462 static void     bridge_handoff(struct ifnet *, struct mbuf *);
 463
 464 static void     bridge_del_bif_handler(struct netmsg *);
 465 static void     bridge_add_bif_handler(struct netmsg *);
 466 static void     bridge_set_bifflags_handler(struct netmsg *);
 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.nm_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.nm_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(struct netmsg *nmsg)
 711 {
 712         struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
 713         struct bridge_softc *sc = lmsg->u.ms_resultp;
 714         struct ifnet *bifp = sc->sc_ifp;
 715         struct bridge_iflist *bif;
 716
 717         ifnet_serialize_all(bifp);
 718
 719         while ((bif = LIST_FIRST(&sc->sc_iflists[mycpuid])) != NULL)
 720                 bridge_delete_member(sc, bif, 0);
 721
 722         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL)
 723                 bridge_delete_span(sc, bif);
 724
 725         ifnet_deserialize_all(bifp);
 726
 727         lwkt_replymsg(lmsg, 0);
 728 }
 729
 730 /*
 731  * bridge_clone_destroy:
 732  *
 733  *      Destroy a bridge instance.
 734  */
 735 static void
 736 bridge_clone_destroy(struct ifnet *ifp)
 737 {
 738         struct bridge_softc *sc = ifp->if_softc;
 739         struct lwkt_msg *lmsg;
 740         struct netmsg nmsg;
 741
 742         ifnet_serialize_all(ifp);
 743
 744         bridge_stop(ifp);
 745         ifp->if_flags &= ~IFF_UP;
 746
 747         ifnet_deserialize_all(ifp);
 748
 749         netmsg_init(&nmsg, NULL, &curthread->td_msgport,
 750                     0, bridge_delete_dispatch);
 751         lmsg = &nmsg.nm_lmsg;
 752         lmsg->u.ms_resultp = sc;
 753         lwkt_domsg(BRIDGE_CFGPORT, lmsg, 0);
 754
 755         crit_enter();   /* XXX MP */
 756         LIST_REMOVE(sc, sc_list);
 757         crit_exit();
 758
 759         ether_ifdetach(ifp);
 760
 761         /* Tear down the routing table. */
 762         bridge_rtable_fini(sc);
 763
 764         /* Free per-cpu member iface lists */
 765         kfree(sc->sc_iflists, M_DEVBUF);
 766
 767         kfree(sc, M_DEVBUF);
 768 }
 769
 770 /*
 771  * bridge_ioctl:
 772  *
 773  *      Handle a control request from the operator.
 774  */
 775 static int
 776 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 777 {
 778         struct bridge_softc *sc = ifp->if_softc;
 779         struct bridge_control_arg args;
 780         struct ifdrv *ifd = (struct ifdrv *) data;
 781         const struct bridge_control *bc;
 782         int error = 0;
 783
 784         ASSERT_IFNET_SERIALIZED_ALL(ifp);
 785
 786         switch (cmd) {
 787         case SIOCADDMULTI:
 788         case SIOCDELMULTI:
 789                 break;
 790
 791         case SIOCGDRVSPEC:
 792         case SIOCSDRVSPEC:
 793                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 794                         error = EINVAL;
 795                         break;
 796                 }
 797                 bc = &bridge_control_table[ifd->ifd_cmd];
 798
 799                 if (cmd == SIOCGDRVSPEC &&
 800                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 801                         error = EINVAL;
 802                         break;
 803                 } else if (cmd == SIOCSDRVSPEC &&
 804                            (bc->bc_flags & BC_F_COPYOUT)) {
 805                         error = EINVAL;
 806                         break;
 807                 }
 808
 809                 if (bc->bc_flags & BC_F_SUSER) {
 810                         error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY);
 811                         if (error)
 812                                 break;
 813                 }
 814
 815                 if (ifd->ifd_len != bc->bc_argsize ||
 816                     ifd->ifd_len > sizeof(args.bca_u)) {
 817                         error = EINVAL;
 818                         break;
 819                 }
 820
 821                 memset(&args, 0, sizeof(args));
 822                 if (bc->bc_flags & BC_F_COPYIN) {
 823                         error = copyin(ifd->ifd_data, &args.bca_u,
 824                                        ifd->ifd_len);
 825                         if (error)
 826                                 break;
 827                 }
 828
 829                 error = bridge_control(sc, cmd, bc->bc_func, &args);
 830                 if (error) {
 831                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 832                         break;
 833                 }
 834
 835                 if (bc->bc_flags & BC_F_COPYOUT) {
 836                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 837                         if (args.bca_len != 0) {
 838                                 KKASSERT(args.bca_kptr != NULL);
 839                                 if (!error) {
 840                                         error = copyout(args.bca_kptr,
 841                                                 args.bca_uptr, args.bca_len);
 842                                 }
 843                                 kfree(args.bca_kptr, M_TEMP);
 844                         } else {
 845                                 KKASSERT(args.bca_kptr == NULL);
 846                         }
 847                 } else {
 848                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 849                 }
 850                 break;
 851
 852         case SIOCSIFFLAGS:
 853                 if (!(ifp->if_flags & IFF_UP) &&
 854                     (ifp->if_flags & IFF_RUNNING)) {
 855                         /*
 856                          * If interface is marked down and it is running,
 857                          * then stop it.
 858                          */
 859                         bridge_stop(ifp);
 860                 } else if ((ifp->if_flags & IFF_UP) &&
 861                     !(ifp->if_flags & IFF_RUNNING)) {
 862                         /*
 863                          * If interface is marked up and it is stopped, then
 864                          * start it.
 865                          */
 866                         ifp->if_init(sc);
 867                 }
 868                 break;
 869
 870         case SIOCSIFMTU:
 871                 /* Do not allow the MTU to be changed on the bridge */
 872                 error = EINVAL;
 873                 break;
 874
 875         default:
 876                 error = ether_ioctl(ifp, cmd, data);
 877                 break;
 878         }
 879         return (error);
 880 }
 881
 882 /*
 883  * bridge_mutecaps:
 884  *
 885  *      Clear or restore unwanted capabilities on the member interface
 886  */
 887 static void
 888 bridge_mutecaps(struct bridge_ifinfo *bif_info, struct ifnet *ifp, int mute)
 889 {
 890         struct ifreq ifr;
 891         int error;
 892
 893         if (ifp->if_ioctl == NULL)
 894                 return;
 895
 896         bzero(&ifr, sizeof(ifr));
 897         ifr.ifr_reqcap = ifp->if_capenable;
 898
 899         if (mute) {
 900                 /* mask off and save capabilities */
 901                 bif_info->bifi_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
 902                 if (bif_info->bifi_mutecap != 0)
 903                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
 904         } else {
 905                 /* restore muted capabilities */
 906                 ifr.ifr_reqcap |= bif_info->bifi_mutecap;
 907         }
 908
 909         if (bif_info->bifi_mutecap != 0) {
 910                 ifnet_serialize_all(ifp);
 911                 error = ifp->if_ioctl(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL);
 912                 ifnet_deserialize_all(ifp);
 913         }
 914 }
 915
 916 /*
 917  * bridge_lookup_member:
 918  *
 919  *      Lookup a bridge member interface.
 920  */
 921 static struct bridge_iflist *
 922 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 923 {
 924         struct bridge_iflist *bif;
 925
 926         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 927                 if (strcmp(bif->bif_ifp->if_xname, name) == 0)
 928                         return (bif);
 929         }
 930         return (NULL);
 931 }
 932
 933 /*
 934  * bridge_lookup_member_if:
 935  *
 936  *      Lookup a bridge member interface by ifnet*.
 937  */
 938 static struct bridge_iflist *
 939 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 940 {
 941         struct bridge_iflist *bif;
 942
 943         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 944                 if (bif->bif_ifp == member_ifp)
 945                         return (bif);
 946         }
 947         return (NULL);
 948 }
 949
 950 /*
 951  * bridge_lookup_member_ifinfo:
 952  *
 953  *      Lookup a bridge member interface by bridge_ifinfo.
 954  */
 955 static struct bridge_iflist *
 956 bridge_lookup_member_ifinfo(struct bridge_softc *sc,
 957                             struct bridge_ifinfo *bif_info)
 958 {
 959         struct bridge_iflist *bif;
 960
 961         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 962                 if (bif->bif_info == bif_info)
 963                         return (bif);
 964         }
 965         return (NULL);
 966 }
 967
 968 /*
 969  * bridge_delete_member:
 970  *
 971  *      Delete the specified member interface.
 972  */
 973 static void
 974 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 975     int gone)
 976 {
 977         struct ifnet *ifs = bif->bif_ifp;
 978         struct ifnet *bifp = sc->sc_ifp;
 979         struct bridge_ifinfo *bif_info = bif->bif_info;
 980         struct bridge_iflist_head saved_bifs;
 981
 982         ASSERT_IFNET_SERIALIZED_ALL(bifp);
 983         KKASSERT(bif_info != NULL);
 984
 985         ifs->if_bridge = NULL;
 986
 987         /*
 988          * Release bridge interface's serializer:
 989          * - To avoid possible dead lock.
 990          * - Various sync operation will block the current thread.
 991          */
 992         ifnet_deserialize_all(bifp);
 993
 994         if (!gone) {
 995                 switch (ifs->if_type) {
 996                 case IFT_ETHER:
 997                 case IFT_L2VLAN:
 998                         /*
 999                          * Take the interface out of promiscuous mode.
1000                          */
1001                         ifpromisc(ifs, 0);
1002                         bridge_mutecaps(bif_info, ifs, 0);
1003                         break;
1004
1005                 case IFT_GIF:
1006                         break;
1007
1008                 default:
1009                         panic("bridge_delete_member: impossible");
1010                         break;
1011                 }
1012         }
1013
1014         /*
1015          * Remove bifs from percpu linked list.
1016          *
1017          * Removed bifs are not freed immediately, instead,
1018          * they are saved in saved_bifs.  They will be freed
1019          * after we make sure that no one is accessing them,
1020          * i.e. after following netmsg_service_sync()
1021          */
1022         LIST_INIT(&saved_bifs);
1023         bridge_del_bif(sc, bif_info, &saved_bifs);
1024
1025         /*
1026          * Make sure that all protocol threads:
1027          * o  see 'ifs' if_bridge is changed
1028          * o  know that bif is removed from the percpu linked list
1029          */
1030         netmsg_service_sync();
1031
1032         /*
1033          * Free the removed bifs
1034          */
1035         KKASSERT(!LIST_EMPTY(&saved_bifs));
1036         while ((bif = LIST_FIRST(&saved_bifs)) != NULL) {
1037                 LIST_REMOVE(bif, bif_next);
1038                 kfree(bif, M_DEVBUF);
1039         }
1040
1041         /* See the comment in bridge_ioctl_stop() */
1042         bridge_rtmsg_sync(sc);
1043         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL | IFBF_FLUSHSYNC);
1044
1045         ifnet_serialize_all(bifp);
1046
1047         if (bifp->if_flags & IFF_RUNNING)
1048                 bstp_initialization(sc);
1049
1050         /*
1051          * Free the bif_info after bstp_initialization(), so that
1052          * bridge_softc.sc_root_port will not reference a dangling
1053          * pointer.
1054          */
1055         kfree(bif_info, M_DEVBUF);
1056 }
1057
1058 /*
1059  * bridge_delete_span:
1060  *
1061  *      Delete the specified span interface.
1062  */
1063 static void
1064 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1065 {
1066         KASSERT(bif->bif_ifp->if_bridge == NULL,
1067             ("%s: not a span interface", __func__));
1068
1069         LIST_REMOVE(bif, bif_next);
1070         kfree(bif, M_DEVBUF);
1071 }
1072
1073 static int
1074 bridge_ioctl_init(struct bridge_softc *sc, void *arg __unused)
1075 {
1076         struct ifnet *ifp = sc->sc_ifp;
1077
1078         if (ifp->if_flags & IFF_RUNNING)
1079                 return 0;
1080
1081         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1082             bridge_timer, sc);
1083
1084         ifp->if_flags |= IFF_RUNNING;
1085         bstp_initialization(sc);
1086         return 0;
1087 }
1088
1089 static int
1090 bridge_ioctl_stop(struct bridge_softc *sc, void *arg __unused)
1091 {
1092         struct ifnet *ifp = sc->sc_ifp;
1093         struct lwkt_msg *lmsg;
1094
1095         if ((ifp->if_flags & IFF_RUNNING) == 0)
1096                 return 0;
1097
1098         callout_stop(&sc->sc_brcallout);
1099
1100         crit_enter();
1101         lmsg = &sc->sc_brtimemsg.nm_lmsg;
1102         if ((lmsg->ms_flags & MSGF_DONE) == 0) {
1103                 /* Pending to be processed; drop it */
1104                 lwkt_dropmsg(lmsg);
1105         }
1106         crit_exit();
1107
1108         bstp_stop(sc);
1109
1110         ifp->if_flags &= ~IFF_RUNNING;
1111
1112         ifnet_deserialize_all(ifp);
1113
1114         /* Let everyone know that we are stopped */
1115         netmsg_service_sync();
1116
1117         /*
1118          * Sync ifnetX msgports in the order we forward rtnode
1119          * installation message.  This is used to make sure that
1120          * all rtnode installation messages sent by bridge_rtupdate()
1121          * during above netmsg_service_sync() are flushed.
1122          */
1123         bridge_rtmsg_sync(sc);
1124         bridge_rtflush(sc, IFBF_FLUSHDYN | IFBF_FLUSHSYNC);
1125
1126         ifnet_serialize_all(ifp);
1127         return 0;
1128 }
1129
1130 static int
1131 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1132 {
1133         struct ifbreq *req = arg;
1134         struct bridge_iflist *bif;
1135         struct bridge_ifinfo *bif_info;
1136         struct ifnet *ifs, *bifp;
1137         int error = 0;
1138
1139         bifp = sc->sc_ifp;
1140         ASSERT_IFNET_SERIALIZED_ALL(bifp);
1141
1142         ifs = ifunit(req->ifbr_ifsname);
1143         if (ifs == NULL)
1144                 return (ENOENT);
1145
1146         /* If it's in the span list, it can't be a member. */
1147         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1148                 if (ifs == bif->bif_ifp)
1149                         return (EBUSY);
1150
1151         /* Allow the first Ethernet member to define the MTU */
1152         if (ifs->if_type != IFT_GIF) {
1153                 if (LIST_EMPTY(&sc->sc_iflists[mycpuid])) {
1154                         bifp->if_mtu = ifs->if_mtu;
1155                 } else if (bifp->if_mtu != ifs->if_mtu) {
1156                         if_printf(bifp, "invalid MTU for %s\n", ifs->if_xname);
1157                         return (EINVAL);
1158                 }
1159         }
1160
1161         if (ifs->if_bridge == sc)
1162                 return (EEXIST);
1163
1164         if (ifs->if_bridge != NULL)
1165                 return (EBUSY);
1166
1167         bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO);
1168         bif_info->bifi_priority = BSTP_DEFAULT_PORT_PRIORITY;
1169         bif_info->bifi_path_cost = BSTP_DEFAULT_PATH_COST;
1170         bif_info->bifi_ifp = ifs;
1171
1172         /*
1173          * Release bridge interface's serializer:
1174          * - To avoid possible dead lock.
1175          * - Various sync operation will block the current thread.
1176          */
1177         ifnet_deserialize_all(bifp);
1178
1179         switch (ifs->if_type) {
1180         case IFT_ETHER:
1181         case IFT_L2VLAN:
1182                 /*
1183                  * Place the interface into promiscuous mode.
1184                  */
1185                 error = ifpromisc(ifs, 1);
1186                 if (error) {
1187                         ifnet_serialize_all(bifp);
1188                         goto out;
1189                 }
1190                 bridge_mutecaps(bif_info, ifs, 1);
1191                 break;
1192
1193         case IFT_GIF: /* :^) */
1194                 break;
1195
1196         default:
1197                 error = EINVAL;
1198                 ifnet_serialize_all(bifp);
1199                 goto out;
1200         }
1201
1202         /*
1203          * Add bifs to percpu linked lists
1204          */
1205         bridge_add_bif(sc, bif_info, ifs);
1206
1207         ifnet_serialize_all(bifp);
1208
1209         if (bifp->if_flags & IFF_RUNNING)
1210                 bstp_initialization(sc);
1211         else
1212                 bstp_stop(sc);
1213
1214         /*
1215          * Everything has been setup, so let the member interface
1216          * deliver packets to this bridge on its input/output path.
1217          */
1218         ifs->if_bridge = sc;
1219 out:
1220         if (error) {
1221                 if (bif_info != NULL)
1222                         kfree(bif_info, M_DEVBUF);
1223         }
1224         return (error);
1225 }
1226
1227 static int
1228 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1229 {
1230         struct ifbreq *req = arg;
1231         struct bridge_iflist *bif;
1232
1233         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1234         if (bif == NULL)
1235                 return (ENOENT);
1236
1237         bridge_delete_member(sc, bif, 0);
1238
1239         return (0);
1240 }
1241
1242 static int
1243 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1244 {
1245         struct ifbreq *req = arg;
1246         struct bridge_iflist *bif;
1247
1248         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1249         if (bif == NULL)
1250                 return (ENOENT);
1251
1252         req->ifbr_ifsflags = bif->bif_flags;
1253         req->ifbr_state = bif->bif_state;
1254         req->ifbr_priority = bif->bif_priority;
1255         req->ifbr_path_cost = bif->bif_path_cost;
1256         req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1257
1258         return (0);
1259 }
1260
1261 static int
1262 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1263 {
1264         struct ifbreq *req = arg;
1265         struct bridge_iflist *bif;
1266         struct ifnet *bifp = sc->sc_ifp;
1267
1268         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1269         if (bif == NULL)
1270                 return (ENOENT);
1271
1272         if (req->ifbr_ifsflags & IFBIF_SPAN) {
1273                 /* SPAN is readonly */
1274                 return (EINVAL);
1275         }
1276
1277         if (req->ifbr_ifsflags & IFBIF_STP) {
1278                 switch (bif->bif_ifp->if_type) {
1279                 case IFT_ETHER:
1280                         /* These can do spanning tree. */
1281                         break;
1282
1283                 default:
1284                         /* Nothing else can. */
1285                         return (EINVAL);
1286                 }
1287         }
1288
1289         ifnet_deserialize_all(bifp);
1290         bridge_set_bifflags(sc, bif->bif_info, req->ifbr_ifsflags);
1291         ifnet_serialize_all(bifp);
1292
1293         if (bifp->if_flags & IFF_RUNNING)
1294                 bstp_initialization(sc);
1295
1296         return (0);
1297 }
1298
1299 static int
1300 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1301 {
1302         struct ifbrparam *param = arg;
1303         struct ifnet *ifp = sc->sc_ifp;
1304
1305         sc->sc_brtmax = param->ifbrp_csize;
1306
1307         ifnet_deserialize_all(ifp);
1308         bridge_rttrim(sc);
1309         ifnet_serialize_all(ifp);
1310
1311         return (0);
1312 }
1313
1314 static int
1315 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1316 {
1317         struct ifbrparam *param = arg;
1318
1319         param->ifbrp_csize = sc->sc_brtmax;
1320
1321         return (0);
1322 }
1323
1324 static int
1325 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1326 {
1327         struct bridge_control_arg *bc_arg = arg;
1328         struct ifbifconf *bifc = arg;
1329         struct bridge_iflist *bif;
1330         struct ifbreq *breq;
1331         int count, len;
1332
1333         count = 0;
1334         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next)
1335                 count++;
1336         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1337                 count++;
1338
1339         if (bifc->ifbic_len == 0) {
1340                 bifc->ifbic_len = sizeof(*breq) * count;
1341                 return 0;
1342         } else if (count == 0 || bifc->ifbic_len < sizeof(*breq)) {
1343                 bifc->ifbic_len = 0;
1344                 return 0;
1345         }
1346
1347         len = min(bifc->ifbic_len, sizeof(*breq) * count);
1348         KKASSERT(len >= sizeof(*breq));
1349
1350         breq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1351         if (breq == NULL) {
1352                 bifc->ifbic_len = 0;
1353                 return ENOMEM;
1354         }
1355         bc_arg->bca_kptr = breq;
1356
1357         count = 0;
1358         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1359                 if (len < sizeof(*breq))
1360                         break;
1361
1362                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1363                         sizeof(breq->ifbr_ifsname));
1364                 breq->ifbr_ifsflags = bif->bif_flags;
1365                 breq->ifbr_state = bif->bif_state;
1366                 breq->ifbr_priority = bif->bif_priority;
1367                 breq->ifbr_path_cost = bif->bif_path_cost;
1368                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1369                 breq++;
1370                 count++;
1371                 len -= sizeof(*breq);
1372         }
1373         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1374                 if (len < sizeof(*breq))
1375                         break;
1376
1377                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1378                         sizeof(breq->ifbr_ifsname));
1379                 breq->ifbr_ifsflags = bif->bif_flags;
1380                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1381                 breq++;
1382                 count++;
1383                 len -= sizeof(*breq);
1384         }
1385
1386         bifc->ifbic_len = sizeof(*breq) * count;
1387         KKASSERT(bifc->ifbic_len > 0);
1388
1389         bc_arg->bca_len = bifc->ifbic_len;
1390         bc_arg->bca_uptr = bifc->ifbic_req;
1391         return 0;
1392 }
1393
1394 static int
1395 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1396 {
1397         struct bridge_control_arg *bc_arg = arg;
1398         struct ifbaconf *bac = arg;
1399         struct bridge_rtnode *brt;
1400         struct ifbareq *bareq;
1401         int count, len;
1402
1403         count = 0;
1404         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list)
1405                 count++;
1406
1407         if (bac->ifbac_len == 0) {
1408                 bac->ifbac_len = sizeof(*bareq) * count;
1409                 return 0;
1410         } else if (count == 0 || bac->ifbac_len < sizeof(*bareq)) {
1411                 bac->ifbac_len = 0;
1412                 return 0;
1413         }
1414
1415         len = min(bac->ifbac_len, sizeof(*bareq) * count);
1416         KKASSERT(len >= sizeof(*bareq));
1417
1418         bareq = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1419         if (bareq == NULL) {
1420                 bac->ifbac_len = 0;
1421                 return ENOMEM;
1422         }
1423         bc_arg->bca_kptr = bareq;
1424
1425         count = 0;
1426         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
1427                 struct bridge_rtinfo *bri = brt->brt_info;
1428                 unsigned long expire;
1429
1430                 if (len < sizeof(*bareq))
1431                         break;
1432
1433                 strlcpy(bareq->ifba_ifsname, bri->bri_ifp->if_xname,
1434                         sizeof(bareq->ifba_ifsname));
1435                 memcpy(bareq->ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1436                 expire = bri->bri_expire;
1437                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1438                     time_second < expire)
1439                         bareq->ifba_expire = expire - time_second;
1440                 else
1441                         bareq->ifba_expire = 0;
1442                 bareq->ifba_flags = bri->bri_flags;
1443                 bareq++;
1444                 count++;
1445                 len -= sizeof(*bareq);
1446         }
1447
1448         bac->ifbac_len = sizeof(*bareq) * count;
1449         KKASSERT(bac->ifbac_len > 0);
1450
1451         bc_arg->bca_len = bac->ifbac_len;
1452         bc_arg->bca_uptr = bac->ifbac_req;
1453         return 0;
1454 }
1455
1456 static int
1457 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1458 {
1459         struct ifbareq *req = arg;
1460         struct bridge_iflist *bif;
1461         struct ifnet *ifp = sc->sc_ifp;
1462         int error;
1463
1464         ASSERT_IFNET_SERIALIZED_ALL(ifp);
1465
1466         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1467         if (bif == NULL)
1468                 return (ENOENT);
1469
1470         ifnet_deserialize_all(ifp);
1471         error = bridge_rtsaddr(sc, req->ifba_dst, bif->bif_ifp,
1472                                req->ifba_flags);
1473         ifnet_serialize_all(ifp);
1474         return (error);
1475 }
1476
1477 static int
1478 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1479 {
1480         struct ifbrparam *param = arg;
1481
1482         sc->sc_brttimeout = param->ifbrp_ctime;
1483
1484         return (0);
1485 }
1486
1487 static int
1488 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1489 {
1490         struct ifbrparam *param = arg;
1491
1492         param->ifbrp_ctime = sc->sc_brttimeout;
1493
1494         return (0);
1495 }
1496
1497 static int
1498 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1499 {
1500         struct ifbareq *req = arg;
1501         struct ifnet *ifp = sc->sc_ifp;
1502         int error;
1503
1504         ifnet_deserialize_all(ifp);
1505         error = bridge_rtdaddr(sc, req->ifba_dst);
1506         ifnet_serialize_all(ifp);
1507         return error;
1508 }
1509
1510 static int
1511 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1512 {
1513         struct ifbreq *req = arg;
1514         struct ifnet *ifp = sc->sc_ifp;
1515
1516         ifnet_deserialize_all(ifp);
1517         bridge_rtflush(sc, req->ifbr_ifsflags | IFBF_FLUSHSYNC);
1518         ifnet_serialize_all(ifp);
1519
1520         return (0);
1521 }
1522
1523 static int
1524 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1525 {
1526         struct ifbrparam *param = arg;
1527
1528         param->ifbrp_prio = sc->sc_bridge_priority;
1529
1530         return (0);
1531 }
1532
1533 static int
1534 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1535 {
1536         struct ifbrparam *param = arg;
1537
1538         sc->sc_bridge_priority = param->ifbrp_prio;
1539
1540         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1541                 bstp_initialization(sc);
1542
1543         return (0);
1544 }
1545
1546 static int
1547 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1548 {
1549         struct ifbrparam *param = arg;
1550
1551         param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1552
1553         return (0);
1554 }
1555
1556 static int
1557 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1558 {
1559         struct ifbrparam *param = arg;
1560
1561         if (param->ifbrp_hellotime == 0)
1562                 return (EINVAL);
1563         sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1564
1565         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1566                 bstp_initialization(sc);
1567
1568         return (0);
1569 }
1570
1571 static int
1572 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1573 {
1574         struct ifbrparam *param = arg;
1575
1576         param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1577
1578         return (0);
1579 }
1580
1581 static int
1582 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1583 {
1584         struct ifbrparam *param = arg;
1585
1586         if (param->ifbrp_fwddelay == 0)
1587                 return (EINVAL);
1588         sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1589
1590         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1591                 bstp_initialization(sc);
1592
1593         return (0);
1594 }
1595
1596 static int
1597 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1598 {
1599         struct ifbrparam *param = arg;
1600
1601         param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1602
1603         return (0);
1604 }
1605
1606 static int
1607 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1608 {
1609         struct ifbrparam *param = arg;
1610
1611         if (param->ifbrp_maxage == 0)
1612                 return (EINVAL);
1613         sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1614
1615         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1616                 bstp_initialization(sc);
1617
1618         return (0);
1619 }
1620
1621 static int
1622 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1623 {
1624         struct ifbreq *req = arg;
1625         struct bridge_iflist *bif;
1626
1627         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1628         if (bif == NULL)
1629                 return (ENOENT);
1630
1631         bif->bif_priority = req->ifbr_priority;
1632
1633         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1634                 bstp_initialization(sc);
1635
1636         return (0);
1637 }
1638
1639 static int
1640 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1641 {
1642         struct ifbreq *req = arg;
1643         struct bridge_iflist *bif;
1644
1645         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1646         if (bif == NULL)
1647                 return (ENOENT);
1648
1649         bif->bif_path_cost = req->ifbr_path_cost;
1650
1651         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1652                 bstp_initialization(sc);
1653
1654         return (0);
1655 }
1656
1657 static int
1658 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1659 {
1660         struct ifbreq *req = arg;
1661         struct bridge_iflist *bif;
1662         struct ifnet *ifs;
1663
1664         ifs = ifunit(req->ifbr_ifsname);
1665         if (ifs == NULL)
1666                 return (ENOENT);
1667
1668         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1669                 if (ifs == bif->bif_ifp)
1670                         return (EBUSY);
1671
1672         if (ifs->if_bridge != NULL)
1673                 return (EBUSY);
1674
1675         switch (ifs->if_type) {
1676         case IFT_ETHER:
1677         case IFT_GIF:
1678         case IFT_L2VLAN:
1679                 break;
1680
1681         default:
1682                 return (EINVAL);
1683         }
1684
1685         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
1686         bif->bif_ifp = ifs;
1687         bif->bif_flags = IFBIF_SPAN;
1688         /* NOTE: span bif does not need bridge_ifinfo */
1689
1690         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1691
1692         sc->sc_span = 1;
1693
1694         return (0);
1695 }
1696
1697 static int
1698 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1699 {
1700         struct ifbreq *req = arg;
1701         struct bridge_iflist *bif;
1702         struct ifnet *ifs;
1703
1704         ifs = ifunit(req->ifbr_ifsname);
1705         if (ifs == NULL)
1706                 return (ENOENT);
1707
1708         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1709                 if (ifs == bif->bif_ifp)
1710                         break;
1711
1712         if (bif == NULL)
1713                 return (ENOENT);
1714
1715         bridge_delete_span(sc, bif);
1716
1717         if (LIST_EMPTY(&sc->sc_spanlist))
1718                 sc->sc_span = 0;
1719
1720         return (0);
1721 }
1722
1723 static void
1724 bridge_ifdetach_dispatch(struct netmsg *nmsg)
1725 {
1726         struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
1727         struct ifnet *ifp, *bifp;
1728         struct bridge_softc *sc;
1729         struct bridge_iflist *bif;
1730
1731         ifp = lmsg->u.ms_resultp;
1732         sc = ifp->if_bridge;
1733
1734         /* Check if the interface is a bridge member */
1735         if (sc != NULL) {
1736                 bifp = sc->sc_ifp;
1737
1738                 ifnet_serialize_all(bifp);
1739
1740                 bif = bridge_lookup_member_if(sc, ifp);
1741                 if (bif != NULL) {
1742                         bridge_delete_member(sc, bif, 1);
1743                 } else {
1744                         /* XXX Why bif will be NULL? */
1745                 }
1746
1747                 ifnet_deserialize_all(bifp);
1748                 goto reply;
1749         }
1750
1751         crit_enter();   /* XXX MP */
1752
1753         /* Check if the interface is a span port */
1754         LIST_FOREACH(sc, &bridge_list, sc_list) {
1755                 bifp = sc->sc_ifp;
1756
1757                 ifnet_serialize_all(bifp);
1758
1759                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1760                         if (ifp == bif->bif_ifp) {
1761                                 bridge_delete_span(sc, bif);
1762                                 break;
1763                         }
1764
1765                 ifnet_deserialize_all(bifp);
1766         }
1767
1768         crit_exit();
1769
1770 reply:
1771         lwkt_replymsg(lmsg, 0);
1772 }
1773
1774 /*
1775  * bridge_ifdetach:
1776  *
1777  *      Detach an interface from a bridge.  Called when a member
1778  *      interface is detaching.
1779  */
1780 static void
1781 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1782 {
1783         struct lwkt_msg *lmsg;
1784         struct netmsg nmsg;
1785
1786         netmsg_init(&nmsg, NULL, &curthread->td_msgport,
1787                     0, bridge_ifdetach_dispatch);
1788         lmsg = &nmsg.nm_lmsg;
1789         lmsg->u.ms_resultp = ifp;
1790
1791         lwkt_domsg(BRIDGE_CFGPORT, lmsg, 0);
1792 }
1793
1794 /*
1795  * bridge_init:
1796  *
1797  *      Initialize a bridge interface.
1798  */
1799 static void
1800 bridge_init(void *xsc)
1801 {
1802         bridge_control(xsc, SIOCSIFFLAGS, bridge_ioctl_init, NULL);
1803 }
1804
1805 /*
1806  * bridge_stop:
1807  *
1808  *      Stop the bridge interface.
1809  */
1810 static void
1811 bridge_stop(struct ifnet *ifp)
1812 {
1813         bridge_control(ifp->if_softc, SIOCSIFFLAGS, bridge_ioctl_stop, NULL);
1814 }
1815
1816 /*
1817  * bridge_enqueue:
1818  *
1819  *      Enqueue a packet on a bridge member interface.
1820  *
1821  */
1822 void
1823 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m)
1824 {
1825         struct netmsg_packet *nmp;
1826
1827         nmp = &m->m_hdr.mh_netmsg;
1828         netmsg_init(&nmp->nm_netmsg, NULL, &netisr_apanic_rport,
1829                     0, bridge_enqueue_handler);
1830         nmp->nm_packet = m;
1831         nmp->nm_netmsg.nm_lmsg.u.ms_resultp = dst_ifp;
1832
1833         lwkt_sendmsg(curnetport, &nmp->nm_netmsg.nm_lmsg);
1834 }
1835
1836 /*
1837  * bridge_output:
1838  *
1839  *      Send output from a bridge member interface.  This
1840  *      performs the bridging function for locally originated
1841  *      packets.
1842  *
1843  *      The mbuf has the Ethernet header already attached.  We must
1844  *      enqueue or free the mbuf before returning.
1845  */
1846 static int
1847 bridge_output(struct ifnet *ifp, struct mbuf *m)
1848 {
1849         struct bridge_softc *sc = ifp->if_bridge;
1850         struct ether_header *eh;
1851         struct ifnet *dst_if, *bifp;
1852
1853         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
1854
1855         /*
1856          * Make sure that we are still a member of a bridge interface.
1857          */
1858         if (sc == NULL) {
1859                 m_freem(m);
1860                 return (0);
1861         }
1862         bifp = sc->sc_ifp;
1863
1864         if (m->m_len < ETHER_HDR_LEN) {
1865                 m = m_pullup(m, ETHER_HDR_LEN);
1866                 if (m == NULL)
1867                         return (0);
1868         }
1869         eh = mtod(m, struct ether_header *);
1870
1871         /*
1872          * If bridge is down, but the original output interface is up,
1873          * go ahead and send out that interface.  Otherwise, the packet
1874          * is dropped below.
1875          */
1876         if ((bifp->if_flags & IFF_RUNNING) == 0) {
1877                 dst_if = ifp;
1878                 goto sendunicast;
1879         }
1880
1881         /*
1882          * If the packet is a multicast, or we don't know a better way to
1883          * get there, send to all interfaces.
1884          */
1885         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1886                 dst_if = NULL;
1887         else
1888                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1889         if (dst_if == NULL) {
1890                 struct bridge_iflist *bif, *nbif;
1891                 struct mbuf *mc;
1892                 int used = 0;
1893
1894                 if (sc->sc_span)
1895                         bridge_span(sc, m);
1896
1897                 LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid],
1898                                      bif_next, nbif) {
1899                         dst_if = bif->bif_ifp;
1900                         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1901                                 continue;
1902
1903                         /*
1904                          * If this is not the original output interface,
1905                          * and the interface is participating in spanning
1906                          * tree, make sure the port is in a state that
1907                          * allows forwarding.
1908                          */
1909                         if (dst_if != ifp &&
1910                             (bif->bif_flags & IFBIF_STP) != 0) {
1911                                 switch (bif->bif_state) {
1912                                 case BSTP_IFSTATE_BLOCKING:
1913                                 case BSTP_IFSTATE_LISTENING:
1914                                 case BSTP_IFSTATE_DISABLED:
1915                                         continue;
1916                                 }
1917                         }
1918
1919                         if (LIST_NEXT(bif, bif_next) == NULL) {
1920                                 used = 1;
1921                                 mc = m;
1922                         } else {
1923                                 mc = m_copypacket(m, MB_DONTWAIT);
1924                                 if (mc == NULL) {
1925                                         bifp->if_oerrors++;
1926                                         continue;
1927                                 }
1928                         }
1929                         bridge_handoff(dst_if, mc);
1930
1931                         if (nbif != NULL && !nbif->bif_onlist) {
1932                                 KKASSERT(bif->bif_onlist);
1933                                 nbif = LIST_NEXT(bif, bif_next);
1934                         }
1935                 }
1936                 if (used == 0)
1937                         m_freem(m);
1938                 return (0);
1939         }
1940
1941 sendunicast:
1942         /*
1943          * XXX Spanning tree consideration here?
1944          */
1945         if (sc->sc_span)
1946                 bridge_span(sc, m);
1947         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1948                 m_freem(m);
1949         else
1950                 bridge_handoff(dst_if, m);
1951         return (0);
1952 }
1953
1954 /*
1955  * bridge_start:
1956  *
1957  *      Start output on a bridge.
1958  *
1959  */
1960 static void
1961 bridge_start(struct ifnet *ifp)
1962 {
1963         struct bridge_softc *sc = ifp->if_softc;
1964
1965         ASSERT_IFNET_SERIALIZED_TX(ifp);
1966
1967         ifp->if_flags |= IFF_OACTIVE;
1968         for (;;) {
1969                 struct ifnet *dst_if = NULL;
1970                 struct ether_header *eh;
1971                 struct mbuf *m;
1972
1973                 m = ifq_dequeue(&ifp->if_snd, NULL);
1974                 if (m == NULL)
1975                         break;
1976
1977                 if (m->m_len < sizeof(*eh)) {
1978                         m = m_pullup(m, sizeof(*eh));
1979                         if (m == NULL) {
1980                                 ifp->if_oerrors++;
1981                                 continue;
1982                         }
1983                 }
1984                 eh = mtod(m, struct ether_header *);
1985
1986                 BPF_MTAP(ifp, m);
1987                 ifp->if_opackets++;
1988
1989                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0)
1990                         dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1991
1992                 if (dst_if == NULL)
1993                         bridge_start_bcast(sc, m);
1994                 else
1995                         bridge_enqueue(dst_if, m);
1996         }
1997         ifp->if_flags &= ~IFF_OACTIVE;
1998 }
1999
2000 /*
2001  * bridge_forward:
2002  *
2003  *      The forwarding function of the bridge.
2004  */
2005 static void
2006 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
2007 {
2008         struct bridge_iflist *bif;
2009         struct ifnet *src_if, *dst_if, *ifp;
2010         struct ether_header *eh;
2011
2012         src_if = m->m_pkthdr.rcvif;
2013         ifp = sc->sc_ifp;
2014
2015         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2016
2017         ifp->if_ipackets++;
2018         ifp->if_ibytes += m->m_pkthdr.len;
2019
2020         /*
2021          * Look up the bridge_iflist.
2022          */
2023         bif = bridge_lookup_member_if(sc, src_if);
2024         if (bif == NULL) {
2025                 /* Interface is not a bridge member (anymore?) */
2026                 m_freem(m);
2027                 return;
2028         }
2029
2030         if (bif->bif_flags & IFBIF_STP) {
2031                 switch (bif->bif_state) {
2032                 case BSTP_IFSTATE_BLOCKING:
2033                 case BSTP_IFSTATE_LISTENING:
2034                 case BSTP_IFSTATE_DISABLED:
2035                         m_freem(m);
2036                         return;
2037                 }
2038         }
2039
2040         eh = mtod(m, struct ether_header *);
2041
2042         /*
2043          * If the interface is learning, and the source
2044          * address is valid and not multicast, record
2045          * the address.
2046          */
2047         if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
2048             ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
2049             (eh->ether_shost[0] == 0 &&
2050              eh->ether_shost[1] == 0 &&
2051              eh->ether_shost[2] == 0 &&
2052              eh->ether_shost[3] == 0 &&
2053              eh->ether_shost[4] == 0 &&
2054              eh->ether_shost[5] == 0) == 0)
2055                 bridge_rtupdate(sc, eh->ether_shost, src_if, IFBAF_DYNAMIC);
2056
2057         if ((bif->bif_flags & IFBIF_STP) != 0 &&
2058             bif->bif_state == BSTP_IFSTATE_LEARNING) {
2059                 m_freem(m);
2060                 return;
2061         }
2062
2063         /*
2064          * At this point, the port either doesn't participate
2065          * in spanning tree or it is in the forwarding state.
2066          */
2067
2068         /*
2069          * If the packet is unicast, destined for someone on
2070          * "this" side of the bridge, drop it.
2071          */
2072         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2073                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2074                 if (src_if == dst_if) {
2075                         m_freem(m);
2076                         return;
2077                 }
2078         } else {
2079                 /* ...forward it to all interfaces. */
2080                 ifp->if_imcasts++;
2081                 dst_if = NULL;
2082         }
2083
2084         if (dst_if == NULL) {
2085                 bridge_broadcast(sc, src_if, m);
2086                 return;
2087         }
2088
2089         /*
2090          * At this point, we're dealing with a unicast frame
2091          * going to a different interface.
2092          */
2093         if ((dst_if->if_flags & IFF_RUNNING) == 0) {
2094                 m_freem(m);
2095                 return;
2096         }
2097         bif = bridge_lookup_member_if(sc, dst_if);
2098         if (bif == NULL) {
2099                 /* Not a member of the bridge (anymore?) */
2100                 m_freem(m);
2101                 return;
2102         }
2103
2104         if (bif->bif_flags & IFBIF_STP) {
2105                 switch (bif->bif_state) {
2106                 case BSTP_IFSTATE_DISABLED:
2107                 case BSTP_IFSTATE_BLOCKING:
2108                         m_freem(m);
2109                         return;
2110                 }
2111         }
2112
2113         if (inet_pfil_hook.ph_hashooks > 0
2114 #ifdef INET6
2115             || inet6_pfil_hook.ph_hashooks > 0
2116 #endif
2117             ) {
2118                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2119                         return;
2120                 if (m == NULL)
2121                         return;
2122
2123                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2124                         return;
2125                 if (m == NULL)
2126                         return;
2127         }
2128         bridge_handoff(dst_if, m);
2129 }
2130
2131 /*
2132  * bridge_input:
2133  *
2134  *      Receive input from a member interface.  Queue the packet for
2135  *      bridging if it is not for us.
2136  */
2137 static struct mbuf *
2138 bridge_input(struct ifnet *ifp, struct mbuf *m)
2139 {
2140         struct bridge_softc *sc = ifp->if_bridge;
2141         struct bridge_iflist *bif;
2142         struct ifnet *bifp, *new_ifp;
2143         struct ether_header *eh;
2144         struct mbuf *mc, *mc2;
2145
2146         ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
2147
2148         /*
2149          * Make sure that we are still a member of a bridge interface.
2150          */
2151         if (sc == NULL)
2152                 return m;
2153
2154         new_ifp = NULL;
2155         bifp = sc->sc_ifp;
2156
2157         if ((bifp->if_flags & IFF_RUNNING) == 0)
2158                 goto out;
2159
2160         /*
2161          * Implement support for bridge monitoring.  If this flag has been
2162          * set on this interface, discard the packet once we push it through
2163          * the bpf(4) machinery, but before we do, increment various counters
2164          * associated with this bridge.
2165          */
2166         if (bifp->if_flags & IFF_MONITOR) {
2167                 /* Change input interface to this bridge */
2168                 m->m_pkthdr.rcvif = bifp;
2169
2170                 BPF_MTAP(bifp, m);
2171
2172                 /* Update bridge's ifnet statistics */
2173                 bifp->if_ipackets++;
2174                 bifp->if_ibytes += m->m_pkthdr.len;
2175                 if (m->m_flags & (M_MCAST | M_BCAST))
2176                         bifp->if_imcasts++;
2177
2178                 m_freem(m);
2179                 m = NULL;
2180                 goto out;
2181         }
2182
2183         eh = mtod(m, struct ether_header *);
2184
2185         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
2186                 /*
2187                  * If the packet is for us, set the packets source as the
2188                  * bridge, and return the packet back to ifnet.if_input for
2189                  * local processing.
2190                  */
2191                 KASSERT(bifp->if_bridge == NULL,
2192                         ("loop created in bridge_input"));
2193                 new_ifp = bifp;
2194                 goto out;
2195         }
2196
2197         /*
2198          * Tap all packets arriving on the bridge, no matter if
2199          * they are local destinations or not.  In is in.
2200          */
2201         BPF_MTAP(bifp, m);
2202
2203         bif = bridge_lookup_member_if(sc, ifp);
2204         if (bif == NULL)
2205                 goto out;
2206
2207         if (sc->sc_span)
2208                 bridge_span(sc, m);
2209
2210         if (m->m_flags & (M_BCAST | M_MCAST)) {
2211                 /* Tap off 802.1D packets; they do not get forwarded. */
2212                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2213                     ETHER_ADDR_LEN) == 0) {
2214                         ifnet_serialize_all(bifp);
2215                         bstp_input(sc, bif, m);
2216                         ifnet_deserialize_all(bifp);
2217
2218                         /* m is freed by bstp_input */
2219                         m = NULL;
2220                         goto out;
2221                 }
2222
2223                 if (bif->bif_flags & IFBIF_STP) {
2224                         switch (bif->bif_state) {
2225                         case BSTP_IFSTATE_BLOCKING:
2226                         case BSTP_IFSTATE_LISTENING:
2227                         case BSTP_IFSTATE_DISABLED:
2228                                 goto out;
2229                         }
2230                 }
2231
2232                 /*
2233                  * Make a deep copy of the packet and enqueue the copy
2234                  * for bridge processing; return the original packet for
2235                  * local processing.
2236                  */
2237                 mc = m_dup(m, MB_DONTWAIT);
2238                 if (mc == NULL)
2239                         goto out;
2240
2241                 bridge_forward(sc, mc);
2242
2243                 /*
2244                  * Reinject the mbuf as arriving on the bridge so we have a
2245                  * chance at claiming multicast packets. We can not loop back
2246                  * here from ether_input as a bridge is never a member of a
2247                  * bridge.
2248                  */
2249                 KASSERT(bifp->if_bridge == NULL,
2250                         ("loop created in bridge_input"));
2251                 mc2 = m_dup(m, MB_DONTWAIT);
2252 #ifdef notyet
2253                 if (mc2 != NULL) {
2254                         /* Keep the layer3 header aligned */
2255                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2256                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2257                 }
2258 #endif
2259                 if (mc2 != NULL) {
2260                         /*
2261                          * Don't tap to bpf(4) again; we have
2262                          * already done the tapping.
2263                          */
2264                         ether_reinput_oncpu(bifp, mc2, 0);
2265                 }
2266
2267                 /* Return the original packet for local processing. */
2268                 goto out;
2269         }
2270
2271         if (bif->bif_flags & IFBIF_STP) {
2272                 switch (bif->bif_state) {
2273                 case BSTP_IFSTATE_BLOCKING:
2274                 case BSTP_IFSTATE_LISTENING:
2275                 case BSTP_IFSTATE_DISABLED:
2276                         goto out;
2277                 }
2278         }
2279
2280         /*
2281          * Unicast.  Make sure it's not for us.
2282          *
2283          * This loop is MPSAFE; the only blocking operation (bridge_rtupdate)
2284          * is followed by breaking out of the loop.
2285          */
2286         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2287                 if (bif->bif_ifp->if_type != IFT_ETHER)
2288                         continue;
2289
2290                 /* It is destined for us. */
2291                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
2292                     ETHER_ADDR_LEN) == 0) {
2293                         if (bif->bif_ifp != ifp) {
2294                                 /* XXX loop prevention */
2295                                 m->m_flags |= M_ETHER_BRIDGED;
2296                                 new_ifp = bif->bif_ifp;
2297                         }
2298                         if (bif->bif_flags & IFBIF_LEARNING) {
2299                                 bridge_rtupdate(sc, eh->ether_shost,
2300                                                 ifp, IFBAF_DYNAMIC);
2301                         }
2302                         goto out;
2303                 }
2304
2305                 /* We just received a packet that we sent out. */
2306                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
2307                     ETHER_ADDR_LEN) == 0) {
2308                         m_freem(m);
2309                         m = NULL;
2310                         goto out;
2311                 }
2312         }
2313
2314         /* Perform the bridge forwarding function. */
2315         bridge_forward(sc, m);
2316         m = NULL;
2317 out:
2318         if (new_ifp != NULL) {
2319                 ether_reinput_oncpu(new_ifp, m, 1);
2320                 m = NULL;
2321         }
2322         return (m);
2323 }
2324
2325 /*
2326  * bridge_start_bcast:
2327  *
2328  *      Broadcast the packet sent from bridge to all member
2329  *      interfaces.
2330  *      This is a simplified version of bridge_broadcast(), however,
2331  *      this function expects caller to hold bridge's serializer.
2332  */
2333 static void
2334 bridge_start_bcast(struct bridge_softc *sc, struct mbuf *m)
2335 {
2336         struct bridge_iflist *bif;
2337         struct mbuf *mc;
2338         struct ifnet *dst_if, *bifp;
2339         int used = 0;
2340
2341         bifp = sc->sc_ifp;
2342         ASSERT_IFNET_SERIALIZED_ALL(bifp);
2343
2344         /*
2345          * Following loop is MPSAFE; nothing is blocking
2346          * in the loop body.
2347          */
2348         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2349                 dst_if = bif->bif_ifp;
2350
2351                 if (bif->bif_flags & IFBIF_STP) {
2352                         switch (bif->bif_state) {
2353                         case BSTP_IFSTATE_BLOCKING:
2354                         case BSTP_IFSTATE_DISABLED:
2355                                 continue;
2356                         }
2357                 }
2358
2359                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2360                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2361                         continue;
2362
2363                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2364                         continue;
2365
2366                 if (LIST_NEXT(bif, bif_next) == NULL) {
2367                         mc = m;
2368                         used = 1;
2369                 } else {
2370                         mc = m_copypacket(m, MB_DONTWAIT);
2371                         if (mc == NULL) {
2372                                 bifp->if_oerrors++;
2373                                 continue;
2374                         }
2375                 }
2376                 bridge_enqueue(dst_if, mc);
2377         }
2378         if (used == 0)
2379                 m_freem(m);
2380 }
2381
2382 /*
2383  * bridge_broadcast:
2384  *
2385  *      Send a frame to all interfaces that are members of
2386  *      the bridge, except for the one on which the packet
2387  *      arrived.
2388  */
2389 static void
2390 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2391     struct mbuf *m)
2392 {
2393         struct bridge_iflist *bif, *nbif;
2394         struct mbuf *mc;
2395         struct ifnet *dst_if, *bifp;
2396         int used = 0;
2397
2398         bifp = sc->sc_ifp;
2399         ASSERT_IFNET_NOT_SERIALIZED_ALL(bifp);
2400
2401         if (inet_pfil_hook.ph_hashooks > 0
2402 #ifdef INET6
2403             || inet6_pfil_hook.ph_hashooks > 0
2404 #endif
2405             ) {
2406                 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0)
2407                         return;
2408                 if (m == NULL)
2409                         return;
2410
2411                 /* Filter on the bridge interface before broadcasting */
2412                 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0)
2413                         return;
2414                 if (m == NULL)
2415                         return;
2416         }
2417
2418         LIST_FOREACH_MUTABLE(bif, &sc->sc_iflists[mycpuid], bif_next, nbif) {
2419                 dst_if = bif->bif_ifp;
2420                 if (dst_if == src_if)
2421                         continue;
2422
2423                 if (bif->bif_flags & IFBIF_STP) {
2424                         switch (bif->bif_state) {
2425                         case BSTP_IFSTATE_BLOCKING:
2426                         case BSTP_IFSTATE_DISABLED:
2427                                 continue;
2428                         }
2429                 }
2430
2431                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2432                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2433                         continue;
2434
2435                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2436                         continue;
2437
2438                 if (LIST_NEXT(bif, bif_next) == NULL) {
2439                         mc = m;
2440                         used = 1;
2441                 } else {
2442                         mc = m_copypacket(m, MB_DONTWAIT);
2443                         if (mc == NULL) {
2444                                 sc->sc_ifp->if_oerrors++;
2445                                 continue;
2446                         }
2447                 }
2448
2449                 /*
2450                  * Filter on the output interface.  Pass a NULL bridge
2451                  * interface pointer so we do not redundantly filter on
2452                  * the bridge for each interface we broadcast on.
2453                  */
2454                 if (inet_pfil_hook.ph_hashooks > 0
2455 #ifdef INET6
2456                     || inet6_pfil_hook.ph_hashooks > 0
2457 #endif
2458                     ) {
2459                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2460                                 continue;
2461                         if (mc == NULL)
2462                                 continue;
2463                 }
2464                 bridge_handoff(dst_if, mc);
2465
2466                 if (nbif != NULL && !nbif->bif_onlist) {
2467                         KKASSERT(bif->bif_onlist);
2468                         nbif = LIST_NEXT(bif, bif_next);
2469                 }
2470         }
2471         if (used == 0)
2472                 m_freem(m);
2473 }
2474
2475 /*
2476  * bridge_span:
2477  *
2478  *      Duplicate a packet out one or more interfaces that are in span mode,
2479  *      the original mbuf is unmodified.
2480  */
2481 static void
2482 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2483 {
2484         struct bridge_iflist *bif;
2485         struct ifnet *dst_if, *bifp;
2486         struct mbuf *mc;
2487
2488         bifp = sc->sc_ifp;
2489         ifnet_serialize_all(bifp);
2490
2491         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2492                 dst_if = bif->bif_ifp;
2493
2494                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2495                         continue;
2496
2497                 mc = m_copypacket(m, MB_DONTWAIT);
2498                 if (mc == NULL) {
2499                         sc->sc_ifp->if_oerrors++;
2500                         continue;
2501                 }
2502                 bridge_enqueue(dst_if, mc);
2503         }
2504
2505         ifnet_deserialize_all(bifp);
2506 }
2507
2508 static void
2509 bridge_rtmsg_sync_handler(struct netmsg *nmsg)
2510 {
2511         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
2512 }
2513
2514 static void
2515 bridge_rtmsg_sync(struct bridge_softc *sc)
2516 {
2517         struct netmsg nmsg;
2518
2519         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2520
2521         netmsg_init(&nmsg, NULL, &curthread->td_msgport,
2522                     0, bridge_rtmsg_sync_handler);
2523         ifnet_domsg(&nmsg.nm_lmsg, 0);
2524 }
2525
2526 static __inline void
2527 bridge_rtinfo_update(struct bridge_rtinfo *bri, struct ifnet *dst_if,
2528                      int setflags, uint8_t flags, uint32_t timeo)
2529 {
2530         if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2531             bri->bri_ifp != dst_if)
2532                 bri->bri_ifp = dst_if;
2533         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2534             bri->bri_expire != time_second + timeo)
2535                 bri->bri_expire = time_second + timeo;
2536         if (setflags)
2537                 bri->bri_flags = flags;
2538 }
2539
2540 static int
2541 bridge_rtinstall_oncpu(struct bridge_softc *sc, const uint8_t *dst,
2542                        struct ifnet *dst_if, int setflags, uint8_t flags,
2543                        struct bridge_rtinfo **bri0)
2544 {
2545         struct bridge_rtnode *brt;
2546         struct bridge_rtinfo *bri;
2547
2548         if (mycpuid == 0) {
2549                 brt = bridge_rtnode_lookup(sc, dst);
2550                 if (brt != NULL) {
2551                         /*
2552                          * rtnode for 'dst' already exists.  We inform the
2553                          * caller about this by leaving bri0 as NULL.  The
2554                          * caller will terminate the intallation upon getting
2555                          * NULL bri0.  However, we still need to update the
2556                          * rtinfo.
2557                          */
2558                         KKASSERT(*bri0 == NULL);
2559
2560                         /* Update rtinfo */
2561                         bridge_rtinfo_update(brt->brt_info, dst_if, setflags,
2562                                              flags, sc->sc_brttimeout);
2563                         return 0;
2564                 }
2565
2566                 /*
2567                  * We only need to check brtcnt on CPU0, since if limit
2568                  * is to be exceeded, ENOSPC is returned.  Caller knows
2569                  * this and will terminate the installation.
2570                  */
2571                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2572                         return ENOSPC;
2573
2574                 KKASSERT(*bri0 == NULL);
2575                 bri = kmalloc(sizeof(struct bridge_rtinfo), M_DEVBUF,
2576                                   M_WAITOK | M_ZERO);
2577                 *bri0 = bri;
2578
2579                 /* Setup rtinfo */
2580                 bri->bri_flags = IFBAF_DYNAMIC;
2581                 bridge_rtinfo_update(bri, dst_if, setflags, flags,
2582                                      sc->sc_brttimeout);
2583         } else {
2584                 bri = *bri0;
2585                 KKASSERT(bri != NULL);
2586         }
2587
2588         brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF,
2589                       M_WAITOK | M_ZERO);
2590         memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2591         brt->brt_info = bri;
2592
2593         bridge_rtnode_insert(sc, brt);
2594         return 0;
2595 }
2596
2597 static void
2598 bridge_rtinstall_handler(struct netmsg *nmsg)
2599 {
2600         struct netmsg_brsaddr *brmsg = (struct netmsg_brsaddr *)nmsg;
2601         int error;
2602
2603         error = bridge_rtinstall_oncpu(brmsg->br_softc,
2604                                        brmsg->br_dst, brmsg->br_dst_if,
2605                                        brmsg->br_setflags, brmsg->br_flags,
2606                                        &brmsg->br_rtinfo);
2607         if (error) {
2608                 KKASSERT(mycpuid == 0 && brmsg->br_rtinfo == NULL);
2609                 lwkt_replymsg(&nmsg->nm_lmsg, error);
2610                 return;
2611         } else if (brmsg->br_rtinfo == NULL) {
2612                 /* rtnode already exists for 'dst' */
2613                 KKASSERT(mycpuid == 0);
2614                 lwkt_replymsg(&nmsg->nm_lmsg, 0);
2615                 return;
2616         }
2617         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
2618 }
2619
2620 /*
2621  * bridge_rtupdate:
2622  *
2623  *      Add/Update a bridge routing entry.
2624  */
2625 static int
2626 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2627                 struct ifnet *dst_if, uint8_t flags)
2628 {
2629         struct bridge_rtnode *brt;
2630
2631         /*
2632          * A route for this destination might already exist.  If so,
2633          * update it, otherwise create a new one.
2634          */
2635         if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
2636                 struct netmsg_brsaddr *brmsg;
2637
2638                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2639                         return ENOSPC;
2640
2641                 brmsg = kmalloc(sizeof(*brmsg), M_LWKTMSG, M_WAITOK | M_NULLOK);
2642                 if (brmsg == NULL)
2643                         return ENOMEM;
2644
2645                 netmsg_init(&brmsg->br_nmsg, NULL, &netisr_afree_rport,
2646                             0, bridge_rtinstall_handler);
2647                 memcpy(brmsg->br_dst, dst, ETHER_ADDR_LEN);
2648                 brmsg->br_dst_if = dst_if;
2649                 brmsg->br_flags = flags;
2650                 brmsg->br_setflags = 0;
2651                 brmsg->br_softc = sc;
2652                 brmsg->br_rtinfo = NULL;
2653
2654                 ifnet_sendmsg(&brmsg->br_nmsg.nm_lmsg, 0);
2655                 return 0;
2656         }
2657         bridge_rtinfo_update(brt->brt_info, dst_if, 0, flags,
2658                              sc->sc_brttimeout);
2659         return 0;
2660 }
2661
2662 static int
2663 bridge_rtsaddr(struct bridge_softc *sc, const uint8_t *dst,
2664                struct ifnet *dst_if, uint8_t flags)
2665 {
2666         struct netmsg_brsaddr brmsg;
2667
2668         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2669
2670         netmsg_init(&brmsg.br_nmsg, NULL, &curthread->td_msgport,
2671                     0, bridge_rtinstall_handler);
2672         memcpy(brmsg.br_dst, dst, ETHER_ADDR_LEN);
2673         brmsg.br_dst_if = dst_if;
2674         brmsg.br_flags = flags;
2675         brmsg.br_setflags = 1;
2676         brmsg.br_softc = sc;
2677         brmsg.br_rtinfo = NULL;
2678
2679         return ifnet_domsg(&brmsg.br_nmsg.nm_lmsg, 0);
2680 }
2681
2682 /*
2683  * bridge_rtlookup:
2684  *
2685  *      Lookup the destination interface for an address.
2686  */
2687 static struct ifnet *
2688 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2689 {
2690         struct bridge_rtnode *brt;
2691
2692         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2693                 return NULL;
2694         return brt->brt_info->bri_ifp;
2695 }
2696
2697 static void
2698 bridge_rtreap_handler(struct netmsg *nmsg)
2699 {
2700         struct bridge_softc *sc = nmsg->nm_lmsg.u.ms_resultp;
2701         struct bridge_rtnode *brt, *nbrt;
2702
2703         LIST_FOREACH_MUTABLE(brt, &sc->sc_rtlists[mycpuid], brt_list, nbrt) {
2704                 if (brt->brt_info->bri_dead)
2705                         bridge_rtnode_destroy(sc, brt);
2706         }
2707         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
2708 }
2709
2710 static void
2711 bridge_rtreap(struct bridge_softc *sc)
2712 {
2713         struct netmsg nmsg;
2714
2715         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2716
2717         netmsg_init(&nmsg, NULL, &curthread->td_msgport,
2718                     0, bridge_rtreap_handler);
2719         nmsg.nm_lmsg.u.ms_resultp = sc;
2720
2721         ifnet_domsg(&nmsg.nm_lmsg, 0);
2722 }
2723
2724 static void
2725 bridge_rtreap_async(struct bridge_softc *sc)
2726 {
2727         struct netmsg *nmsg;
2728
2729         nmsg = kmalloc(sizeof(*nmsg), M_LWKTMSG, M_WAITOK);
2730
2731         netmsg_init(nmsg, NULL, &netisr_afree_rport,
2732                     0, bridge_rtreap_handler);
2733         nmsg->nm_lmsg.u.ms_resultp = sc;
2734
2735         ifnet_sendmsg(&nmsg->nm_lmsg, 0);
2736 }
2737
2738 /*
2739  * bridge_rttrim:
2740  *
2741  *      Trim the routine table so that we have a number
2742  *      of routing entries less than or equal to the
2743  *      maximum number.
2744  */
2745 static void
2746 bridge_rttrim(struct bridge_softc *sc)
2747 {
2748         struct bridge_rtnode *brt;
2749         int dead;
2750
2751         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2752
2753         /* Make sure we actually need to do this. */
2754         if (sc->sc_brtcnt <= sc->sc_brtmax)
2755                 return;
2756
2757         /*
2758          * Find out how many rtnodes are dead
2759          */
2760         dead = bridge_rtage_finddead(sc);
2761         KKASSERT(dead <= sc->sc_brtcnt);
2762
2763         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2764                 /* Enough dead rtnodes are found */
2765                 bridge_rtreap(sc);
2766                 return;
2767         }
2768
2769         /*
2770          * Kill some dynamic rtnodes to meet the brtmax
2771          */
2772         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2773                 struct bridge_rtinfo *bri = brt->brt_info;
2774
2775                 if (bri->bri_dead) {
2776                         /*
2777                          * We have counted this rtnode in
2778                          * bridge_rtage_finddead()
2779                          */
2780                         continue;
2781                 }
2782
2783                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2784                         bri->bri_dead = 1;
2785                         ++dead;
2786                         KKASSERT(dead <= sc->sc_brtcnt);
2787
2788                         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2789                                 /* Enough rtnodes are collected */
2790                                 break;
2791                         }
2792                 }
2793         }
2794         if (dead)
2795                 bridge_rtreap(sc);
2796 }
2797
2798 /*
2799  * bridge_timer:
2800  *
2801  *      Aging timer for the bridge.
2802  */
2803 static void
2804 bridge_timer(void *arg)
2805 {
2806         struct bridge_softc *sc = arg;
2807         struct lwkt_msg *lmsg;
2808
2809         KKASSERT(mycpuid == BRIDGE_CFGCPU);
2810
2811         crit_enter();
2812
2813         if (callout_pending(&sc->sc_brcallout) ||
2814             !callout_active(&sc->sc_brcallout)) {
2815                 crit_exit();
2816                 return;
2817         }
2818         callout_deactivate(&sc->sc_brcallout);
2819
2820         lmsg = &sc->sc_brtimemsg.nm_lmsg;
2821         KKASSERT(lmsg->ms_flags & MSGF_DONE);
2822         lwkt_sendmsg(BRIDGE_CFGPORT, lmsg);
2823
2824         crit_exit();
2825 }
2826
2827 static void
2828 bridge_timer_handler(struct netmsg *nmsg)
2829 {
2830         struct bridge_softc *sc = nmsg->nm_lmsg.u.ms_resultp;
2831
2832         KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
2833
2834         crit_enter();
2835         /* Reply ASAP */
2836         lwkt_replymsg(&nmsg->nm_lmsg, 0);
2837         crit_exit();
2838
2839         bridge_rtage(sc);
2840         if (sc->sc_ifp->if_flags & IFF_RUNNING) {
2841                 callout_reset(&sc->sc_brcallout,
2842                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2843         }
2844 }
2845
2846 static int
2847 bridge_rtage_finddead(struct bridge_softc *sc)
2848 {
2849         struct bridge_rtnode *brt;
2850         int dead = 0;
2851
2852         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2853                 struct bridge_rtinfo *bri = brt->brt_info;
2854
2855                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2856                     time_second >= bri->bri_expire) {
2857                         bri->bri_dead = 1;
2858                         ++dead;
2859                         KKASSERT(dead <= sc->sc_brtcnt);
2860                 }
2861         }
2862         return dead;
2863 }
2864
2865 /*
2866  * bridge_rtage:
2867  *
2868  *      Perform an aging cycle.
2869  */
2870 static void
2871 bridge_rtage(struct bridge_softc *sc)
2872 {
2873         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2874
2875         if (bridge_rtage_finddead(sc))
2876                 bridge_rtreap(sc);
2877 }
2878
2879 /*
2880  * bridge_rtflush:
2881  *
2882  *      Remove all dynamic addresses from the bridge.
2883  */
2884 static void
2885 bridge_rtflush(struct bridge_softc *sc, int bf)
2886 {
2887         struct bridge_rtnode *brt;
2888         int reap;
2889
2890         reap = 0;
2891         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2892                 struct bridge_rtinfo *bri = brt->brt_info;
2893
2894                 if ((bf & IFBF_FLUSHALL) ||
2895                     (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2896                         bri->bri_dead = 1;
2897                         reap = 1;
2898                 }
2899         }
2900         if (reap) {
2901                 if (bf & IFBF_FLUSHSYNC)
2902                         bridge_rtreap(sc);
2903                 else
2904                         bridge_rtreap_async(sc);
2905         }
2906 }
2907
2908 /*
2909  * bridge_rtdaddr:
2910  *
2911  *      Remove an address from the table.
2912  */
2913 static int
2914 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2915 {
2916         struct bridge_rtnode *brt;
2917
2918         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
2919
2920         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2921                 return (ENOENT);
2922
2923         /* TODO: add a cheaper delete operation */
2924         brt->brt_info->bri_dead = 1;
2925         bridge_rtreap(sc);
2926         return (0);
2927 }
2928
2929 /*
2930  * bridge_rtdelete:
2931  *
2932  *      Delete routes to a speicifc member interface.
2933  */
2934 void
2935 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int bf)
2936 {
2937         struct bridge_rtnode *brt;
2938         int reap;
2939
2940         reap = 0;
2941         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2942                 struct bridge_rtinfo *bri = brt->brt_info;
2943
2944                 if (bri->bri_ifp == ifp &&
2945                     ((bf & IFBF_FLUSHALL) ||
2946                      (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
2947                         bri->bri_dead = 1;
2948                         reap = 1;
2949                 }
2950         }
2951         if (reap) {
2952                 if (bf & IFBF_FLUSHSYNC)
2953                         bridge_rtreap(sc);
2954                 else
2955                         bridge_rtreap_async(sc);
2956         }
2957 }
2958
2959 /*
2960  * bridge_rtable_init:
2961  *
2962  *      Initialize the route table for this bridge.
2963  */
2964 static void
2965 bridge_rtable_init(struct bridge_softc *sc)
2966 {
2967         int cpu;
2968
2969         /*
2970          * Initialize per-cpu hash tables
2971          */
2972         sc->sc_rthashs = kmalloc(sizeof(*sc->sc_rthashs) * ncpus,
2973                                  M_DEVBUF, M_WAITOK);
2974         for (cpu = 0; cpu < ncpus; ++cpu) {
2975                 int i;
2976
2977                 sc->sc_rthashs[cpu] =
2978                 kmalloc(sizeof(struct bridge_rtnode_head) * BRIDGE_RTHASH_SIZE,
2979                         M_DEVBUF, M_WAITOK);
2980
2981                 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2982                         LIST_INIT(&sc->sc_rthashs[cpu][i]);
2983         }
2984         sc->sc_rthash_key = karc4random();
2985
2986         /*
2987          * Initialize per-cpu lists
2988          */
2989         sc->sc_rtlists = kmalloc(sizeof(struct bridge_rtnode_head) * ncpus,
2990                                  M_DEVBUF, M_WAITOK);
2991         for (cpu = 0; cpu < ncpus; ++cpu)
2992                 LIST_INIT(&sc->sc_rtlists[cpu]);
2993 }
2994
2995 /*
2996  * bridge_rtable_fini:
2997  *
2998  *      Deconstruct the route table for this bridge.
2999  */
3000 static void
3001 bridge_rtable_fini(struct bridge_softc *sc)
3002 {
3003         int cpu;
3004
3005         /*
3006          * Free per-cpu hash tables
3007          */
3008         for (cpu = 0; cpu < ncpus; ++cpu)
3009                 kfree(sc->sc_rthashs[cpu], M_DEVBUF);
3010         kfree(sc->sc_rthashs, M_DEVBUF);
3011
3012         /*
3013          * Free per-cpu lists
3014          */
3015         kfree(sc->sc_rtlists, M_DEVBUF);
3016 }
3017
3018 /*
3019  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3020  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3021  */
3022 #define mix(a, b, c)                                                    \
3023 do {                                                                    \
3024         a -= b; a -= c; a ^= (c >> 13);                                 \
3025         b -= c; b -= a; b ^= (a << 8);                                  \
3026         c -= a; c -= b; c ^= (b >> 13);                                 \
3027         a -= b; a -= c; a ^= (c >> 12);                                 \
3028         b -= c; b -= a; b ^= (a << 16);                                 \
3029         c -= a; c -= b; c ^= (b >> 5);                                  \
3030         a -= b; a -= c; a ^= (c >> 3);                                  \
3031         b -= c; b -= a; b ^= (a << 10);                                 \
3032         c -= a; c -= b; c ^= (b >> 15);                                 \
3033 } while (/*CONSTCOND*/0)
3034
3035 static __inline uint32_t
3036 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3037 {
3038         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3039
3040         b += addr[5] << 8;
3041         b += addr[4];
3042         a += addr[3] << 24;
3043         a += addr[2] << 16;
3044         a += addr[1] << 8;
3045         a += addr[0];
3046
3047         mix(a, b, c);
3048
3049         return (c & BRIDGE_RTHASH_MASK);
3050 }
3051
3052 #undef mix
3053
3054 static int
3055 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3056 {
3057         int i, d;
3058
3059         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3060                 d = ((int)a[i]) - ((int)b[i]);
3061         }
3062
3063         return (d);
3064 }
3065
3066 /*
3067  * bridge_rtnode_lookup:
3068  *
3069  *      Look up a bridge route node for the specified destination.
3070  */
3071 static struct bridge_rtnode *
3072 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
3073 {
3074         struct bridge_rtnode *brt;
3075         uint32_t hash;
3076         int dir;
3077
3078         hash = bridge_rthash(sc, addr);
3079         LIST_FOREACH(brt, &sc->sc_rthashs[mycpuid][hash], brt_hash) {
3080                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3081                 if (dir == 0)
3082                         return (brt);
3083                 if (dir > 0)
3084                         return (NULL);
3085         }
3086
3087         return (NULL);
3088 }
3089
3090 /*
3091  * bridge_rtnode_insert:
3092  *
3093  *      Insert the specified bridge node into the route table.
3094  *      Caller has to make sure that rtnode does not exist.
3095  */
3096 static void
3097 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3098 {
3099         struct bridge_rtnode *lbrt;
3100         uint32_t hash;
3101         int dir;
3102
3103         hash = bridge_rthash(sc, brt->brt_addr);
3104
3105         lbrt = LIST_FIRST(&sc->sc_rthashs[mycpuid][hash]);
3106         if (lbrt == NULL) {
3107                 LIST_INSERT_HEAD(&sc->sc_rthashs[mycpuid][hash], brt, brt_hash);
3108                 goto out;
3109         }
3110
3111         do {
3112                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3113                 KASSERT(dir != 0, ("rtnode already exist\n"));
3114
3115                 if (dir > 0) {
3116                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3117                         goto out;
3118                 }
3119                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
3120                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3121                         goto out;
3122                 }
3123                 lbrt = LIST_NEXT(lbrt, brt_hash);
3124         } while (lbrt != NULL);
3125
3126         panic("no suitable position found for rtnode\n");
3127 out:
3128         LIST_INSERT_HEAD(&sc->sc_rtlists[mycpuid], brt, brt_list);
3129         if (mycpuid == 0) {
3130                 /*
3131                  * Update the brtcnt.
3132                  * We only need to do it once and we do it on CPU0.
3133                  */
3134                 sc->sc_brtcnt++;
3135         }
3136 }
3137
3138 /*
3139  * bridge_rtnode_destroy:
3140  *
3141  *      Destroy a bridge rtnode.
3142  */
3143 static void
3144 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3145 {
3146         LIST_REMOVE(brt, brt_hash);
3147         LIST_REMOVE(brt, brt_list);
3148
3149         if (mycpuid + 1 == ncpus) {
3150                 /* Free rtinfo associated with rtnode on the last cpu */
3151                 kfree(brt->brt_info, M_DEVBUF);
3152         }
3153         kfree(brt, M_DEVBUF);
3154
3155         if (mycpuid == 0) {
3156                 /* Update brtcnt only on CPU0 */
3157                 sc->sc_brtcnt--;
3158         }
3159 }
3160
3161 static __inline int
3162 bridge_post_pfil(struct mbuf *m)
3163 {
3164         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED)
3165                 return EOPNOTSUPP;
3166
3167         /* Not yet */
3168         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED)
3169                 return EOPNOTSUPP;
3170
3171         return 0;
3172 }
3173
3174 /*
3175  * Send bridge packets through pfil if they are one of the types pfil can deal
3176  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3177  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3178  * that interface.
3179  */
3180 static int
3181 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3182 {
3183         int snap, error, i, hlen;
3184         struct ether_header *eh1, eh2;
3185         struct ip *ip;
3186         struct llc llc1;
3187         u_int16_t ether_type;
3188
3189         snap = 0;
3190         error = -1;     /* Default error if not error == 0 */
3191
3192         if (pfil_bridge == 0 && pfil_member == 0)
3193                 return (0); /* filtering is disabled */
3194
3195         i = min((*mp)->m_pkthdr.len, max_protohdr);
3196         if ((*mp)->m_len < i) {
3197                 *mp = m_pullup(*mp, i);
3198                 if (*mp == NULL) {
3199                         kprintf("%s: m_pullup failed\n", __func__);
3200                         return (-1);
3201                 }
3202         }
3203
3204         eh1 = mtod(*mp, struct ether_header *);
3205         ether_type = ntohs(eh1->ether_type);
3206
3207         /*
3208          * Check for SNAP/LLC.
3209          */
3210         if (ether_type < ETHERMTU) {
3211                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3212
3213                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3214                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3215                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3216                     llc2->llc_control == LLC_UI) {
3217                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3218                         snap = 1;
3219                 }
3220         }
3221
3222         /*
3223          * If we're trying to filter bridge traffic, don't look at anything
3224          * other than IP and ARP traffic.  If the filter doesn't understand
3225          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3226          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3227          * but of course we don't have an AppleTalk filter to begin with.
3228          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3229          * ARP traffic.)
3230          */
3231         switch (ether_type) {
3232         case ETHERTYPE_ARP:
3233         case ETHERTYPE_REVARP:
3234                 return (0); /* Automatically pass */
3235
3236         case ETHERTYPE_IP:
3237 #ifdef INET6
3238         case ETHERTYPE_IPV6:
3239 #endif /* INET6 */
3240                 break;
3241
3242         default:
3243                 /*
3244                  * Check to see if the user wants to pass non-ip
3245                  * packets, these will not be checked by pfil(9)
3246                  * and passed unconditionally so the default is to drop.
3247                  */
3248                 if (pfil_onlyip)
3249                         goto bad;
3250         }
3251
3252         /* Strip off the Ethernet header and keep a copy. */
3253         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3254         m_adj(*mp, ETHER_HDR_LEN);
3255
3256         /* Strip off snap header, if present */
3257         if (snap) {
3258                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3259                 m_adj(*mp, sizeof(struct llc));
3260         }
3261
3262         /*
3263          * Check the IP header for alignment and errors
3264          */
3265         if (dir == PFIL_IN) {
3266                 switch (ether_type) {
3267                 case ETHERTYPE_IP:
3268                         error = bridge_ip_checkbasic(mp);
3269                         break;
3270 #ifdef INET6
3271                 case ETHERTYPE_IPV6:
3272                         error = bridge_ip6_checkbasic(mp);
3273                         break;
3274 #endif /* INET6 */
3275                 default:
3276                         error = 0;
3277                 }
3278                 if (error)
3279                         goto bad;
3280         }
3281
3282         error = 0;
3283
3284         /*
3285          * Run the packet through pfil
3286          */
3287         switch (ether_type) {
3288         case ETHERTYPE_IP:
3289                 /*
3290                  * before calling the firewall, swap fields the same as
3291                  * IP does. here we assume the header is contiguous
3292                  */
3293                 ip = mtod(*mp, struct ip *);
3294
3295                 ip->ip_len = ntohs(ip->ip_len);
3296                 ip->ip_off = ntohs(ip->ip_off);
3297
3298                 /*
3299                  * Run pfil on the member interface and the bridge, both can
3300                  * be skipped by clearing pfil_member or pfil_bridge.
3301                  *
3302                  * Keep the order:
3303                  *   in_if -> bridge_if -> out_if
3304                  */
3305                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) {
3306                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3307                         if (*mp == NULL || error != 0) /* filter may consume */
3308                                 break;
3309                         error = bridge_post_pfil(*mp);
3310                         if (error)
3311                                 break;
3312                 }
3313
3314                 if (pfil_member && ifp != NULL) {
3315                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp, dir);
3316                         if (*mp == NULL || error != 0) /* filter may consume */
3317                                 break;
3318                         error = bridge_post_pfil(*mp);
3319                         if (error)
3320                                 break;
3321                 }
3322
3323                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) {
3324                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3325                         if (*mp == NULL || error != 0) /* filter may consume */
3326                                 break;
3327                         error = bridge_post_pfil(*mp);
3328                         if (error)
3329                                 break;
3330                 }
3331
3332                 /* check if we need to fragment the packet */
3333                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3334                         i = (*mp)->m_pkthdr.len;
3335                         if (i > ifp->if_mtu) {
3336                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3337                                             &llc1);
3338                                 return (error);
3339                         }
3340                 }
3341
3342                 /* Recalculate the ip checksum and restore byte ordering */
3343                 ip = mtod(*mp, struct ip *);
3344                 hlen = ip->ip_hl << 2;
3345                 if (hlen < sizeof(struct ip))
3346                         goto bad;
3347                 if (hlen > (*mp)->m_len) {
3348                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3349                                 goto bad;
3350                         ip = mtod(*mp, struct ip *);
3351                         if (ip == NULL)
3352                                 goto bad;
3353                 }
3354                 ip->ip_len = htons(ip->ip_len);
3355                 ip->ip_off = htons(ip->ip_off);
3356                 ip->ip_sum = 0;
3357                 if (hlen == sizeof(struct ip))
3358                         ip->ip_sum = in_cksum_hdr(ip);
3359                 else
3360                         ip->ip_sum = in_cksum(*mp, hlen);
3361
3362                 break;
3363 #ifdef INET6
3364         case ETHERTYPE_IPV6:
3365                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3366                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3367                                         dir);
3368
3369                 if (*mp == NULL || error != 0) /* filter may consume */
3370                         break;
3371
3372                 if (pfil_member && ifp != NULL)
3373                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3374                                         dir);
3375
3376                 if (*mp == NULL || error != 0) /* filter may consume */
3377                         break;
3378
3379                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3380                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3381                                         dir);
3382                 break;
3383 #endif
3384         default:
3385                 error = 0;
3386                 break;
3387         }
3388
3389         if (*mp == NULL)
3390                 return (error);
3391         if (error != 0)
3392                 goto bad;
3393
3394         error = -1;
3395
3396         /*
3397          * Finally, put everything back the way it was and return
3398          */
3399         if (snap) {
3400                 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT);
3401                 if (*mp == NULL)
3402                         return (error);
3403                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3404         }
3405
3406         M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT);
3407         if (*mp == NULL)
3408                 return (error);
3409         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3410
3411         return (0);
3412
3413 bad:
3414         m_freem(*mp);
3415         *mp = NULL;
3416         return (error);
3417 }
3418
3419 /*
3420  * Perform basic checks on header size since
3421  * pfil assumes ip_input has already processed
3422  * it for it.  Cut-and-pasted from ip_input.c.
3423  * Given how simple the IPv6 version is,
3424  * does the IPv4 version really need to be
3425  * this complicated?
3426  *
3427  * XXX Should we update ipstat here, or not?
3428  * XXX Right now we update ipstat but not
3429  * XXX csum_counter.
3430  */
3431 static int
3432 bridge_ip_checkbasic(struct mbuf **mp)
3433 {
3434         struct mbuf *m = *mp;
3435         struct ip *ip;
3436         int len, hlen;
3437         u_short sum;
3438
3439         if (*mp == NULL)
3440                 return (-1);
3441 #if notyet
3442         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3443                 if ((m = m_copyup(m, sizeof(struct ip),
3444                         (max_linkhdr + 3) & ~3)) == NULL) {
3445                         /* XXXJRT new stat, please */
3446                         ipstat.ips_toosmall++;
3447                         goto bad;
3448                 }
3449         } else
3450 #endif
3451 #ifndef __predict_false
3452 #define __predict_false(x) x
3453 #endif
3454          if (__predict_false(m->m_len < sizeof (struct ip))) {
3455                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3456                         ipstat.ips_toosmall++;
3457                         goto bad;
3458                 }
3459         }
3460         ip = mtod(m, struct ip *);
3461         if (ip == NULL) goto bad;
3462
3463         if (ip->ip_v != IPVERSION) {
3464                 ipstat.ips_badvers++;
3465                 goto bad;
3466         }
3467         hlen = ip->ip_hl << 2;
3468         if (hlen < sizeof(struct ip)) { /* minimum header length */
3469                 ipstat.ips_badhlen++;
3470                 goto bad;
3471         }
3472         if (hlen > m->m_len) {
3473                 if ((m = m_pullup(m, hlen)) == 0) {
3474                         ipstat.ips_badhlen++;
3475                         goto bad;
3476                 }
3477                 ip = mtod(m, struct ip *);
3478                 if (ip == NULL) goto bad;
3479         }
3480
3481         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3482                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3483         } else {
3484                 if (hlen == sizeof(struct ip)) {
3485                         sum = in_cksum_hdr(ip);
3486                 } else {
3487                         sum = in_cksum(m, hlen);
3488                 }
3489         }
3490         if (sum) {
3491                 ipstat.ips_badsum++;
3492                 goto bad;
3493         }
3494
3495         /* Retrieve the packet length. */
3496         len = ntohs(ip->ip_len);
3497
3498         /*
3499          * Check for additional length bogosity
3500          */
3501         if (len < hlen) {
3502                 ipstat.ips_badlen++;
3503                 goto bad;
3504         }
3505
3506         /*
3507          * Check that the amount of data in the buffers
3508          * is as at least much as the IP header would have us expect.
3509          * Drop packet if shorter than we expect.
3510          */
3511         if (m->m_pkthdr.len < len) {
3512                 ipstat.ips_tooshort++;
3513                 goto bad;
3514         }
3515
3516         /* Checks out, proceed */
3517         *mp = m;
3518         return (0);
3519
3520 bad:
3521         *mp = m;
3522         return (-1);
3523 }
3524
3525 #ifdef INET6
3526 /*
3527  * Same as above, but for IPv6.
3528  * Cut-and-pasted from ip6_input.c.
3529  * XXX Should we update ip6stat, or not?
3530  */
3531 static int
3532 bridge_ip6_checkbasic(struct mbuf **mp)
3533 {
3534         struct mbuf *m = *mp;
3535         struct ip6_hdr *ip6;
3536
3537         /*
3538          * If the IPv6 header is not aligned, slurp it up into a new
3539          * mbuf with space for link headers, in the event we forward
3540          * it.  Otherwise, if it is aligned, make sure the entire base
3541          * IPv6 header is in the first mbuf of the chain.
3542          */
3543 #if notyet
3544         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3545                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3546                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3547                             (max_linkhdr + 3) & ~3)) == NULL) {
3548                         /* XXXJRT new stat, please */
3549                         ip6stat.ip6s_toosmall++;
3550                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3551                         goto bad;
3552                 }
3553         } else
3554 #endif
3555         if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3556                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3557                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3558                         ip6stat.ip6s_toosmall++;
3559                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3560                         goto bad;
3561                 }
3562         }
3563
3564         ip6 = mtod(m, struct ip6_hdr *);
3565
3566         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3567                 ip6stat.ip6s_badvers++;
3568                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3569                 goto bad;
3570         }
3571
3572         /* Checks out, proceed */
3573         *mp = m;
3574         return (0);
3575
3576 bad:
3577         *mp = m;
3578         return (-1);
3579 }
3580 #endif /* INET6 */
3581
3582 /*
3583  * bridge_fragment:
3584  *
3585  *      Return a fragmented mbuf chain.
3586  */
3587 static int
3588 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3589     int snap, struct llc *llc)
3590 {
3591         struct mbuf *m0;
3592         struct ip *ip;
3593         int error = -1;
3594
3595         if (m->m_len < sizeof(struct ip) &&
3596             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3597                 goto out;
3598         ip = mtod(m, struct ip *);
3599
3600         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3601                     CSUM_DELAY_IP);
3602         if (error)
3603                 goto out;
3604
3605         /* walk the chain and re-add the Ethernet header */
3606         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3607                 if (error == 0) {
3608                         if (snap) {
3609                                 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT);
3610                                 if (m0 == NULL) {
3611                                         error = ENOBUFS;
3612                                         continue;
3613                                 }
3614                                 bcopy(llc, mtod(m0, caddr_t),
3615                                     sizeof(struct llc));
3616                         }
3617                         M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT);
3618                         if (m0 == NULL) {
3619                                 error = ENOBUFS;
3620                                 continue;
3621                         }
3622                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3623                 } else
3624                         m_freem(m);
3625         }
3626
3627         if (error == 0)
3628                 ipstat.ips_fragmented++;
3629
3630         return (error);
3631
3632 out:
3633         if (m != NULL)
3634                 m_freem(m);
3635         return (error);
3636 }
3637
3638 static void
3639 bridge_enqueue_handler(struct netmsg *nmsg)
3640 {
3641         struct netmsg_packet *nmp;
3642         struct ifnet *dst_ifp;
3643         struct mbuf *m;
3644
3645         nmp = (struct netmsg_packet *)nmsg;
3646         m = nmp->nm_packet;
3647         dst_ifp = nmp->nm_netmsg.nm_lmsg.u.ms_resultp;
3648
3649         bridge_handoff(dst_ifp, m);
3650 }
3651
3652 static void
3653 bridge_handoff(struct ifnet *dst_ifp, struct mbuf *m)
3654 {
3655         struct mbuf *m0;
3656
3657         /* We may be sending a fragment so traverse the mbuf */
3658         for (; m; m = m0) {
3659                 struct altq_pktattr pktattr;
3660
3661                 m0 = m->m_nextpkt;
3662                 m->m_nextpkt = NULL;
3663
3664                 if (ifq_is_enabled(&dst_ifp->if_snd))
3665                         altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
3666
3667                 ifq_dispatch(dst_ifp, m, &pktattr);
3668         }
3669 }
3670
3671 static void
3672 bridge_control_dispatch(struct netmsg *nmsg)
3673 {
3674         struct netmsg_brctl *bc_msg = (struct netmsg_brctl *)nmsg;
3675         struct ifnet *bifp = bc_msg->bc_sc->sc_ifp;
3676         int error;
3677
3678         ifnet_serialize_all(bifp);
3679         error = bc_msg->bc_func(bc_msg->bc_sc, bc_msg->bc_arg);
3680         ifnet_deserialize_all(bifp);
3681
3682         lwkt_replymsg(&nmsg->nm_lmsg, error);
3683 }
3684
3685 static int
3686 bridge_control(struct bridge_softc *sc, u_long cmd,
3687                bridge_ctl_t bc_func, void *bc_arg)
3688 {
3689         struct ifnet *bifp = sc->sc_ifp;
3690         struct netmsg_brctl bc_msg;
3691         struct netmsg *nmsg;
3692         int error;
3693
3694         ASSERT_IFNET_SERIALIZED_ALL(bifp);
3695
3696         bzero(&bc_msg, sizeof(bc_msg));
3697         nmsg = &bc_msg.bc_nmsg;
3698
3699         netmsg_init(nmsg, NULL, &curthread->td_msgport,
3700                     0, bridge_control_dispatch);
3701         bc_msg.bc_func = bc_func;
3702         bc_msg.bc_sc = sc;
3703         bc_msg.bc_arg = bc_arg;
3704
3705         ifnet_deserialize_all(bifp);
3706         error = lwkt_domsg(BRIDGE_CFGPORT, &nmsg->nm_lmsg, 0);
3707         ifnet_serialize_all(bifp);
3708         return error;
3709 }
3710
3711 static void
3712 bridge_add_bif_handler(struct netmsg *nmsg)
3713 {
3714         struct netmsg_braddbif *amsg = (struct netmsg_braddbif *)nmsg;
3715         struct bridge_softc *sc;
3716         struct bridge_iflist *bif;
3717
3718         sc = amsg->br_softc;
3719
3720         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
3721         bif->bif_ifp = amsg->br_bif_ifp;
3722         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
3723         bif->bif_onlist = 1;
3724         bif->bif_info = amsg->br_bif_info;
3725
3726         LIST_INSERT_HEAD(&sc->sc_iflists[mycpuid], bif, bif_next);
3727
3728         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
3729 }
3730
3731 static void
3732 bridge_add_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3733                struct ifnet *ifp)
3734 {
3735         struct netmsg_braddbif amsg;
3736
3737         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3738
3739         netmsg_init(&amsg.br_nmsg, NULL, &curthread->td_msgport,
3740                     0, bridge_add_bif_handler);
3741         amsg.br_softc = sc;
3742         amsg.br_bif_info = bif_info;
3743         amsg.br_bif_ifp = ifp;
3744
3745         ifnet_domsg(&amsg.br_nmsg.nm_lmsg, 0);
3746 }
3747
3748 static void
3749 bridge_del_bif_handler(struct netmsg *nmsg)
3750 {
3751         struct netmsg_brdelbif *dmsg = (struct netmsg_brdelbif *)nmsg;
3752         struct bridge_softc *sc;
3753         struct bridge_iflist *bif;
3754
3755         sc = dmsg->br_softc;
3756
3757         /*
3758          * Locate the bif associated with the br_bif_info
3759          * on the current CPU
3760          */
3761         bif = bridge_lookup_member_ifinfo(sc, dmsg->br_bif_info);
3762         KKASSERT(bif != NULL && bif->bif_onlist);
3763
3764         /* Remove the bif from the current CPU's iflist */
3765         bif->bif_onlist = 0;
3766         LIST_REMOVE(bif, bif_next);
3767
3768         /* Save the removed bif for later freeing */
3769         LIST_INSERT_HEAD(dmsg->br_bif_list, bif, bif_next);
3770
3771         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
3772 }
3773
3774 static void
3775 bridge_del_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3776                struct bridge_iflist_head *saved_bifs)
3777 {
3778         struct netmsg_brdelbif dmsg;
3779
3780         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3781
3782         netmsg_init(&dmsg.br_nmsg, NULL, &curthread->td_msgport,
3783                     0, bridge_del_bif_handler);
3784         dmsg.br_softc = sc;
3785         dmsg.br_bif_info = bif_info;
3786         dmsg.br_bif_list = saved_bifs;
3787
3788         ifnet_domsg(&dmsg.br_nmsg.nm_lmsg, 0);
3789 }
3790
3791 static void
3792 bridge_set_bifflags_handler(struct netmsg *nmsg)
3793 {
3794         struct netmsg_brsflags *smsg = (struct netmsg_brsflags *)nmsg;
3795         struct bridge_softc *sc;
3796         struct bridge_iflist *bif;
3797
3798         sc = smsg->br_softc;
3799
3800         /*
3801          * Locate the bif associated with the br_bif_info
3802          * on the current CPU
3803          */
3804         bif = bridge_lookup_member_ifinfo(sc, smsg->br_bif_info);
3805         KKASSERT(bif != NULL && bif->bif_onlist);
3806
3807         bif->bif_flags = smsg->br_bif_flags;
3808
3809         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
3810 }
3811
3812 static void
3813 bridge_set_bifflags(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3814                     uint32_t bif_flags)
3815 {
3816         struct netmsg_brsflags smsg;
3817
3818         ASSERT_IFNET_NOT_SERIALIZED_ALL(sc->sc_ifp);
3819
3820         netmsg_init(&smsg.br_nmsg, NULL, &curthread->td_msgport,
3821                     0, bridge_set_bifflags_handler);
3822         smsg.br_softc = sc;
3823         smsg.br_bif_info = bif_info;
3824         smsg.br_bif_flags = bif_flags;
3825
3826         ifnet_domsg(&smsg.br_nmsg.nm_lmsg, 0);
3827 }