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 1.54 2008/11/21 12:43:42 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 #include "opt_inet.h"
 192 #include "opt_inet6.h"
 193
 194 #include <sys/param.h>
 195 #include <sys/mbuf.h>
 196 #include <sys/malloc.h>
 197 #include <sys/protosw.h>
 198 #include <sys/systm.h>
 199 #include <sys/time.h>
 200 #include <sys/socket.h> /* for net/if.h */
 201 #include <sys/sockio.h>
 202 #include <sys/ctype.h>  /* string functions */
 203 #include <sys/kernel.h>
 204 #include <sys/random.h>
 205 #include <sys/sysctl.h>
 206 #include <sys/module.h>
 207 #include <sys/proc.h>
 208 #include <sys/lock.h>
 209 #include <sys/thread.h>
 210 #include <sys/thread2.h>
 211 #include <sys/mpipe.h>
 212
 213 #include <net/bpf.h>
 214 #include <net/if.h>
 215 #include <net/if_dl.h>
 216 #include <net/if_types.h>
 217 #include <net/if_var.h>
 218 #include <net/pfil.h>
 219 #include <net/ifq_var.h>
 220 #include <net/if_clone.h>
 221
 222 #include <netinet/in.h> /* for struct arpcom */
 223 #include <netinet/in_systm.h>
 224 #include <netinet/in_var.h>
 225 #include <netinet/ip.h>
 226 #include <netinet/ip_var.h>
 227 #ifdef INET6
 228 #include <netinet/ip6.h>
 229 #include <netinet6/ip6_var.h>
 230 #endif
 231 #include <netinet/if_ether.h> /* for struct arpcom */
 232 #include <net/bridge/if_bridgevar.h>
 233 #include <net/if_llc.h>
 234 #include <net/netmsg2.h>
 235
 236 #include <net/route.h>
 237 #include <sys/in_cksum.h>
 238
 239 /*
 240  * Size of the route hash table.  Must be a power of two.
 241  */
 242 #ifndef BRIDGE_RTHASH_SIZE
 243 #define BRIDGE_RTHASH_SIZE              1024
 244 #endif
 245
 246 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 247
 248 /*
 249  * Maximum number of addresses to cache.
 250  */
 251 #ifndef BRIDGE_RTABLE_MAX
 252 #define BRIDGE_RTABLE_MAX               100
 253 #endif
 254
 255 /*
 256  * Spanning tree defaults.
 257  */
 258 #define BSTP_DEFAULT_MAX_AGE            (20 * 256)
 259 #define BSTP_DEFAULT_HELLO_TIME         (2 * 256)
 260 #define BSTP_DEFAULT_FORWARD_DELAY      (15 * 256)
 261 #define BSTP_DEFAULT_HOLD_TIME          (1 * 256)
 262 #define BSTP_DEFAULT_BRIDGE_PRIORITY    0x8000
 263 #define BSTP_DEFAULT_PORT_PRIORITY      0x80
 264 #define BSTP_DEFAULT_PATH_COST          55
 265
 266 /*
 267  * Timeout (in seconds) for entries learned dynamically.
 268  */
 269 #ifndef BRIDGE_RTABLE_TIMEOUT
 270 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 271 #endif
 272
 273 /*
 274  * Number of seconds between walks of the route list.
 275  */
 276 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 277 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 278 #endif
 279
 280 /*
 281  * List of capabilities to mask on the member interface.
 282  */
 283 #define BRIDGE_IFCAPS_MASK              IFCAP_TXCSUM
 284
 285 typedef int     (*bridge_ctl_t)(struct bridge_softc *, void *);
 286
 287 struct netmsg_brctl {
 288         struct netmsg           bc_nmsg;
 289         bridge_ctl_t            bc_func;
 290         struct bridge_softc     *bc_sc;
 291         void                    *bc_arg;
 292 };
 293
 294 struct netmsg_brsaddr {
 295         struct netmsg           br_nmsg;
 296         struct bridge_softc     *br_softc;
 297         struct ifnet            *br_dst_if;
 298         struct bridge_rtinfo    *br_rtinfo;
 299         int                     br_setflags;
 300         uint8_t                 br_dst[ETHER_ADDR_LEN];
 301         uint8_t                 br_flags;
 302 };
 303
 304 struct netmsg_braddbif {
 305         struct netmsg           br_nmsg;
 306         struct bridge_softc     *br_softc;
 307         struct bridge_ifinfo    *br_bif_info;
 308         struct ifnet            *br_bif_ifp;
 309 };
 310
 311 struct netmsg_brdelbif {
 312         struct netmsg           br_nmsg;
 313         struct bridge_softc     *br_softc;
 314         struct bridge_ifinfo    *br_bif_info;
 315         struct bridge_iflist_head *br_bif_list;
 316 };
 317
 318 struct netmsg_brsflags {
 319         struct netmsg           br_nmsg;
 320         struct bridge_softc     *br_softc;
 321         struct bridge_ifinfo    *br_bif_info;
 322         uint32_t                br_bif_flags;
 323 };
 324
 325 eventhandler_tag        bridge_detach_cookie = NULL;
 326
 327 extern  struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
 328 extern  int (*bridge_output_p)(struct ifnet *, struct mbuf *);
 329 extern  void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
 330
 331 static int      bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 332
 333 static int      bridge_clone_create(struct if_clone *, int);
 334 static void     bridge_clone_destroy(struct ifnet *);
 335
 336 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
 337 static void     bridge_mutecaps(struct bridge_ifinfo *, struct ifnet *, int);
 338 static void     bridge_ifdetach(void *, struct ifnet *);
 339 static void     bridge_init(void *);
 340 static void     bridge_stop(struct ifnet *);
 341 static void     bridge_start(struct ifnet *);
 342 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 343 static int      bridge_output(struct ifnet *, struct mbuf *);
 344
 345 static void     bridge_forward(struct bridge_softc *, struct mbuf *m);
 346
 347 static void     bridge_timer_handler(struct netmsg *);
 348 static void     bridge_timer(void *);
 349
 350 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 351                     struct mbuf *, int);
 352 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 353
 354 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 355                     struct ifnet *, uint8_t);
 356 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
 357 static void     bridge_rtreap(struct bridge_softc *);
 358 static void     bridge_rttrim(struct bridge_softc *);
 359 static int      bridge_rtage_finddead(struct bridge_softc *);
 360 static void     bridge_rtage(struct bridge_softc *);
 361 static void     bridge_rtflush(struct bridge_softc *, int);
 362 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
 363 static int      bridge_rtsaddr(struct bridge_softc *, const uint8_t *,
 364                     struct ifnet *, uint8_t);
 365 static void     bridge_rtmsg_sync(struct bridge_softc *sc);
 366 static void     bridge_rtreap_handler(struct netmsg *);
 367 static void     bridge_rtinstall_handler(struct netmsg *);
 368 static int      bridge_rtinstall_oncpu(struct bridge_softc *, const uint8_t *,
 369                     struct ifnet *, int, uint8_t, struct bridge_rtinfo **);
 370
 371 static void     bridge_rtable_init(struct bridge_softc *);
 372 static void     bridge_rtable_fini(struct bridge_softc *);
 373
 374 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 375 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 376                     const uint8_t *);
 377 static void     bridge_rtnode_insert(struct bridge_softc *,
 378                     struct bridge_rtnode *);
 379 static void     bridge_rtnode_destroy(struct bridge_softc *,
 380                     struct bridge_rtnode *);
 381
 382 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 383                     const char *name);
 384 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 385                     struct ifnet *ifp);
 386 static struct bridge_iflist *bridge_lookup_member_ifinfo(struct bridge_softc *,
 387                     struct bridge_ifinfo *);
 388 static void     bridge_delete_member(struct bridge_softc *,
 389                     struct bridge_iflist *, int);
 390 static void     bridge_delete_span(struct bridge_softc *,
 391                     struct bridge_iflist *);
 392
 393 static int      bridge_control(struct bridge_softc *, u_long,
 394                                bridge_ctl_t, void *);
 395 static int      bridge_ioctl_init(struct bridge_softc *, void *);
 396 static int      bridge_ioctl_stop(struct bridge_softc *, void *);
 397 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 398 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 399 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 400 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 401 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 402 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 403 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 404 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 405 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 406 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 407 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 408 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 409 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 410 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 411 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 412 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 413 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 414 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 415 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 416 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 417 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 418 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 419 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 420 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 421 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 422 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 423                     int);
 424 static int      bridge_ip_checkbasic(struct mbuf **mp);
 425 #ifdef INET6
 426 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 427 #endif /* INET6 */
 428 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 429                     struct ether_header *, int, struct llc *);
 430 static void     bridge_enqueue_internal(struct ifnet *, struct mbuf *m,
 431                                         netisr_fn_t);
 432 static void     bridge_enqueue_handler(struct netmsg *);
 433 static void     bridge_pfil_enqueue_handler(struct netmsg *);
 434 static void     bridge_pfil_enqueue(struct ifnet *, struct mbuf *, int);
 435 static void     bridge_handoff(struct ifnet *, struct mbuf *);
 436
 437 static void     bridge_del_bif_handler(struct netmsg *);
 438 static void     bridge_add_bif_handler(struct netmsg *);
 439 static void     bridge_set_bifflags_handler(struct netmsg *);
 440 static void     bridge_del_bif(struct bridge_softc *, struct bridge_ifinfo *,
 441                     struct bridge_iflist_head *);
 442 static void     bridge_add_bif(struct bridge_softc *, struct bridge_ifinfo *,
 443                     struct ifnet *);
 444 static void     bridge_set_bifflags(struct bridge_softc *,
 445                     struct bridge_ifinfo *, uint32_t);
 446
 447 SYSCTL_DECL(_net_link);
 448 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 449
 450 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 451 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 452 static int pfil_member = 1; /* run pfil hooks on the member interface */
 453 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 454     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 455 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 456     &pfil_bridge, 0, "Packet filter on the bridge interface");
 457 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 458     &pfil_member, 0, "Packet filter on the member interface");
 459
 460 struct bridge_control_arg {
 461         union {
 462                 struct ifbreq ifbreq;
 463                 struct ifbifconf ifbifconf;
 464                 struct ifbareq ifbareq;
 465                 struct ifbaconf ifbaconf;
 466                 struct ifbrparam ifbrparam;
 467         } bca_u;
 468         int     bca_len;
 469         void    *bca_uptr;
 470         void    *bca_kptr;
 471 };
 472
 473 struct bridge_control {
 474         bridge_ctl_t    bc_func;
 475         int             bc_argsize;
 476         int             bc_flags;
 477 };
 478
 479 #define BC_F_COPYIN             0x01    /* copy arguments in */
 480 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 481 #define BC_F_SUSER              0x04    /* do super-user check */
 482
 483 const struct bridge_control bridge_control_table[] = {
 484         { bridge_ioctl_add,             sizeof(struct ifbreq),
 485           BC_F_COPYIN|BC_F_SUSER },
 486         { bridge_ioctl_del,             sizeof(struct ifbreq),
 487           BC_F_COPYIN|BC_F_SUSER },
 488
 489         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 490           BC_F_COPYIN|BC_F_COPYOUT },
 491         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 492           BC_F_COPYIN|BC_F_SUSER },
 493
 494         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 495           BC_F_COPYIN|BC_F_SUSER },
 496         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 497           BC_F_COPYOUT },
 498
 499         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 500           BC_F_COPYIN|BC_F_COPYOUT },
 501         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 502           BC_F_COPYIN|BC_F_COPYOUT },
 503
 504         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 505           BC_F_COPYIN|BC_F_SUSER },
 506
 507         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 508           BC_F_COPYIN|BC_F_SUSER },
 509         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 510           BC_F_COPYOUT },
 511
 512         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 513           BC_F_COPYIN|BC_F_SUSER },
 514
 515         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 516           BC_F_COPYIN|BC_F_SUSER },
 517
 518         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 519           BC_F_COPYOUT },
 520         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 521           BC_F_COPYIN|BC_F_SUSER },
 522
 523         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 524           BC_F_COPYOUT },
 525         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 526           BC_F_COPYIN|BC_F_SUSER },
 527
 528         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 529           BC_F_COPYOUT },
 530         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 531           BC_F_COPYIN|BC_F_SUSER },
 532
 533         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 534           BC_F_COPYOUT },
 535         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 536           BC_F_COPYIN|BC_F_SUSER },
 537
 538         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 539           BC_F_COPYIN|BC_F_SUSER },
 540
 541         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 542           BC_F_COPYIN|BC_F_SUSER },
 543
 544         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 545           BC_F_COPYIN|BC_F_SUSER },
 546         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 547           BC_F_COPYIN|BC_F_SUSER },
 548 };
 549 static const int bridge_control_table_size =
 550     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 551
 552 LIST_HEAD(, bridge_softc) bridge_list;
 553
 554 struct if_clone bridge_cloner = IF_CLONE_INITIALIZER("bridge",
 555                                 bridge_clone_create,
 556                                 bridge_clone_destroy, 0, IF_MAXUNIT);
 557
 558 static int
 559 bridge_modevent(module_t mod, int type, void *data)
 560 {
 561         switch (type) {
 562         case MOD_LOAD:
 563                 LIST_INIT(&bridge_list);
 564                 if_clone_attach(&bridge_cloner);
 565                 bridge_input_p = bridge_input;
 566                 bridge_output_p = bridge_output;
 567                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 568                     ifnet_detach_event, bridge_ifdetach, NULL,
 569                     EVENTHANDLER_PRI_ANY);
 570 #if notyet
 571                 bstp_linkstate_p = bstp_linkstate;
 572 #endif
 573                 break;
 574         case MOD_UNLOAD:
 575                 if (!LIST_EMPTY(&bridge_list))
 576                         return (EBUSY);
 577                 EVENTHANDLER_DEREGISTER(ifnet_detach_event,
 578                     bridge_detach_cookie);
 579                 if_clone_detach(&bridge_cloner);
 580                 bridge_input_p = NULL;
 581                 bridge_output_p = NULL;
 582 #if notyet
 583                 bstp_linkstate_p = NULL;
 584 #endif
 585                 break;
 586         default:
 587                 return (EOPNOTSUPP);
 588         }
 589         return (0);
 590 }
 591
 592 static moduledata_t bridge_mod = {
 593         "if_bridge",
 594         bridge_modevent,
 595         0
 596 };
 597
 598 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 599
 600
 601 /*
 602  * bridge_clone_create:
 603  *
 604  *      Create a new bridge instance.
 605  */
 606 static int
 607 bridge_clone_create(struct if_clone *ifc, int unit)
 608 {
 609         struct bridge_softc *sc;
 610         struct ifnet *ifp;
 611         u_char eaddr[6];
 612         int cpu, rnd;
 613
 614         sc = kmalloc(sizeof(*sc), M_DEVBUF, M_WAITOK | M_ZERO);
 615         ifp = sc->sc_ifp = &sc->sc_if;
 616
 617         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 618         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 619         sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
 620         sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
 621         sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
 622         sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
 623         sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
 624
 625         /* Initialize our routing table. */
 626         bridge_rtable_init(sc);
 627
 628         callout_init(&sc->sc_brcallout);
 629         netmsg_init(&sc->sc_brtimemsg, &netisr_adone_rport,
 630                     MSGF_DROPABLE | MSGF_PRIORITY, bridge_timer_handler);
 631         sc->sc_brtimemsg.nm_lmsg.u.ms_resultp = sc;
 632
 633         callout_init(&sc->sc_bstpcallout);
 634         netmsg_init(&sc->sc_bstptimemsg, &netisr_adone_rport,
 635                     MSGF_DROPABLE | MSGF_PRIORITY, bstp_tick_handler);
 636         sc->sc_bstptimemsg.nm_lmsg.u.ms_resultp = sc;
 637
 638         /* Initialize per-cpu member iface lists */
 639         sc->sc_iflists = kmalloc(sizeof(*sc->sc_iflists) * ncpus,
 640                                  M_DEVBUF, M_WAITOK);
 641         for (cpu = 0; cpu < ncpus; ++cpu)
 642                 LIST_INIT(&sc->sc_iflists[cpu]);
 643
 644         LIST_INIT(&sc->sc_spanlist);
 645
 646         ifp->if_softc = sc;
 647         if_initname(ifp, ifc->ifc_name, unit);
 648         ifp->if_mtu = ETHERMTU;
 649         ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
 650         ifp->if_ioctl = bridge_ioctl;
 651         ifp->if_start = bridge_start;
 652         ifp->if_init = bridge_init;
 653         ifp->if_type = IFT_BRIDGE;
 654         ifq_set_maxlen(&ifp->if_snd, ifqmaxlen);
 655         ifp->if_snd.ifq_maxlen = ifqmaxlen;
 656         ifq_set_ready(&ifp->if_snd);
 657         ifp->if_hdrlen = ETHER_HDR_LEN;
 658
 659         /*
 660          * Generate a random ethernet address and use the private AC:DE:48
 661          * OUI code.
 662          */
 663         rnd = karc4random();
 664         bcopy(&rnd, &eaddr[0], 4); /* ETHER_ADDR_LEN == 6 */
 665         rnd = karc4random();
 666         bcopy(&rnd, &eaddr[2], 4); /* ETHER_ADDR_LEN == 6 */
 667
 668         eaddr[0] &= ~1; /* clear multicast bit */
 669         eaddr[0] |= 2;  /* set the LAA bit */
 670
 671         ether_ifattach(ifp, eaddr, NULL);
 672         /* Now undo some of the damage... */
 673         ifp->if_baudrate = 0;
 674         ifp->if_type = IFT_BRIDGE;
 675
 676         crit_enter();   /* XXX MP */
 677         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 678         crit_exit();
 679
 680         return (0);
 681 }
 682
 683 static void
 684 bridge_delete_dispatch(struct netmsg *nmsg)
 685 {
 686         struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
 687         struct bridge_softc *sc = lmsg->u.ms_resultp;
 688         struct ifnet *bifp = sc->sc_ifp;
 689         struct bridge_iflist *bif;
 690
 691         lwkt_serialize_enter(bifp->if_serializer);
 692
 693         while ((bif = LIST_FIRST(&sc->sc_iflists[mycpuid])) != NULL)
 694                 bridge_delete_member(sc, bif, 0);
 695
 696         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL)
 697                 bridge_delete_span(sc, bif);
 698
 699         lwkt_serialize_exit(bifp->if_serializer);
 700
 701         lwkt_replymsg(lmsg, 0);
 702 }
 703
 704 /*
 705  * bridge_clone_destroy:
 706  *
 707  *      Destroy a bridge instance.
 708  */
 709 static void
 710 bridge_clone_destroy(struct ifnet *ifp)
 711 {
 712         struct bridge_softc *sc = ifp->if_softc;
 713         struct lwkt_msg *lmsg;
 714         struct netmsg nmsg;
 715
 716         lwkt_serialize_enter(ifp->if_serializer);
 717
 718         bridge_stop(ifp);
 719         ifp->if_flags &= ~IFF_UP;
 720
 721         lwkt_serialize_exit(ifp->if_serializer);
 722
 723         netmsg_init(&nmsg, &curthread->td_msgport, 0, bridge_delete_dispatch);
 724         lmsg = &nmsg.nm_lmsg;
 725         lmsg->u.ms_resultp = sc;
 726         lwkt_domsg(BRIDGE_CFGPORT, lmsg, 0);
 727
 728         crit_enter();   /* XXX MP */
 729         LIST_REMOVE(sc, sc_list);
 730         crit_exit();
 731
 732         ether_ifdetach(ifp);
 733
 734         /* Tear down the routing table. */
 735         bridge_rtable_fini(sc);
 736
 737         /* Free per-cpu member iface lists */
 738         kfree(sc->sc_iflists, M_DEVBUF);
 739
 740         kfree(sc, M_DEVBUF);
 741 }
 742
 743 /*
 744  * bridge_ioctl:
 745  *
 746  *      Handle a control request from the operator.
 747  */
 748 static int
 749 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data, struct ucred *cr)
 750 {
 751         struct bridge_softc *sc = ifp->if_softc;
 752         struct bridge_control_arg args;
 753         struct ifdrv *ifd = (struct ifdrv *) data;
 754         const struct bridge_control *bc;
 755         int error = 0;
 756
 757         ASSERT_SERIALIZED(ifp->if_serializer);
 758
 759         switch (cmd) {
 760         case SIOCADDMULTI:
 761         case SIOCDELMULTI:
 762                 break;
 763
 764         case SIOCGDRVSPEC:
 765         case SIOCSDRVSPEC:
 766                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 767                         error = EINVAL;
 768                         break;
 769                 }
 770                 bc = &bridge_control_table[ifd->ifd_cmd];
 771
 772                 if (cmd == SIOCGDRVSPEC &&
 773                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 774                         error = EINVAL;
 775                         break;
 776                 } else if (cmd == SIOCSDRVSPEC &&
 777                            (bc->bc_flags & BC_F_COPYOUT)) {
 778                         error = EINVAL;
 779                         break;
 780                 }
 781
 782                 if (bc->bc_flags & BC_F_SUSER) {
 783                         error = suser_cred(cr, NULL_CRED_OKAY);
 784                         if (error)
 785                                 break;
 786                 }
 787
 788                 if (ifd->ifd_len != bc->bc_argsize ||
 789                     ifd->ifd_len > sizeof(args.bca_u)) {
 790                         error = EINVAL;
 791                         break;
 792                 }
 793
 794                 memset(&args, 0, sizeof(args));
 795                 if (bc->bc_flags & BC_F_COPYIN) {
 796                         error = copyin(ifd->ifd_data, &args.bca_u,
 797                                        ifd->ifd_len);
 798                         if (error)
 799                                 break;
 800                 }
 801
 802                 error = bridge_control(sc, cmd, bc->bc_func, &args);
 803                 if (error) {
 804                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 805                         break;
 806                 }
 807
 808                 if (bc->bc_flags & BC_F_COPYOUT) {
 809                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 810                         if (args.bca_len != 0) {
 811                                 KKASSERT(args.bca_kptr != NULL);
 812                                 if (!error) {
 813                                         error = copyout(args.bca_kptr,
 814                                                 args.bca_uptr, args.bca_len);
 815                                 }
 816                                 kfree(args.bca_kptr, M_TEMP);
 817                         } else {
 818                                 KKASSERT(args.bca_kptr == NULL);
 819                         }
 820                 } else {
 821                         KKASSERT(args.bca_len == 0 && args.bca_kptr == NULL);
 822                 }
 823                 break;
 824
 825         case SIOCSIFFLAGS:
 826                 if (!(ifp->if_flags & IFF_UP) &&
 827                     (ifp->if_flags & IFF_RUNNING)) {
 828                         /*
 829                          * If interface is marked down and it is running,
 830                          * then stop it.
 831                          */
 832                         bridge_stop(ifp);
 833                 } else if ((ifp->if_flags & IFF_UP) &&
 834                     !(ifp->if_flags & IFF_RUNNING)) {
 835                         /*
 836                          * If interface is marked up and it is stopped, then
 837                          * start it.
 838                          */
 839                         ifp->if_init(sc);
 840                 }
 841                 break;
 842
 843         case SIOCSIFMTU:
 844                 /* Do not allow the MTU to be changed on the bridge */
 845                 error = EINVAL;
 846                 break;
 847
 848         default:
 849                 error = ether_ioctl(ifp, cmd, data);
 850                 break;
 851         }
 852         return (error);
 853 }
 854
 855 /*
 856  * bridge_mutecaps:
 857  *
 858  *      Clear or restore unwanted capabilities on the member interface
 859  */
 860 static void
 861 bridge_mutecaps(struct bridge_ifinfo *bif_info, struct ifnet *ifp, int mute)
 862 {
 863         struct ifreq ifr;
 864         int error;
 865
 866         if (ifp->if_ioctl == NULL)
 867                 return;
 868
 869         bzero(&ifr, sizeof(ifr));
 870         ifr.ifr_reqcap = ifp->if_capenable;
 871
 872         if (mute) {
 873                 /* mask off and save capabilities */
 874                 bif_info->bifi_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
 875                 if (bif_info->bifi_mutecap != 0)
 876                         ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
 877         } else {
 878                 /* restore muted capabilities */
 879                 ifr.ifr_reqcap |= bif_info->bifi_mutecap;
 880         }
 881
 882         if (bif_info->bifi_mutecap != 0) {
 883                 lwkt_serialize_enter(ifp->if_serializer);
 884                 error = ifp->if_ioctl(ifp, SIOCSIFCAP, (caddr_t)&ifr, NULL);
 885                 lwkt_serialize_exit(ifp->if_serializer);
 886         }
 887 }
 888
 889 /*
 890  * bridge_lookup_member:
 891  *
 892  *      Lookup a bridge member interface.
 893  */
 894 static struct bridge_iflist *
 895 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 896 {
 897         struct bridge_iflist *bif;
 898
 899         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 900                 if (strcmp(bif->bif_ifp->if_xname, name) == 0)
 901                         return (bif);
 902         }
 903         return (NULL);
 904 }
 905
 906 /*
 907  * bridge_lookup_member_if:
 908  *
 909  *      Lookup a bridge member interface by ifnet*.
 910  */
 911 static struct bridge_iflist *
 912 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 913 {
 914         struct bridge_iflist *bif;
 915
 916         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 917                 if (bif->bif_ifp == member_ifp)
 918                         return (bif);
 919         }
 920         return (NULL);
 921 }
 922
 923 /*
 924  * bridge_lookup_member_ifinfo:
 925  *
 926  *      Lookup a bridge member interface by bridge_ifinfo.
 927  */
 928 static struct bridge_iflist *
 929 bridge_lookup_member_ifinfo(struct bridge_softc *sc,
 930                             struct bridge_ifinfo *bif_info)
 931 {
 932         struct bridge_iflist *bif;
 933
 934         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
 935                 if (bif->bif_info == bif_info)
 936                         return (bif);
 937         }
 938         return (NULL);
 939 }
 940
 941 /*
 942  * bridge_delete_member:
 943  *
 944  *      Delete the specified member interface.
 945  */
 946 static void
 947 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 948     int gone)
 949 {
 950         struct ifnet *ifs = bif->bif_ifp;
 951         struct ifnet *bifp = sc->sc_ifp;
 952         struct bridge_ifinfo *bif_info = bif->bif_info;
 953         struct bridge_iflist_head saved_bifs;
 954
 955         ASSERT_SERIALIZED(bifp->if_serializer);
 956         KKASSERT(bif_info != NULL);
 957
 958         ifs->if_bridge = NULL;
 959
 960         /*
 961          * Release bridge interface's serializer:
 962          * - To avoid possible dead lock.
 963          * - Various sync operation will block the current thread.
 964          */
 965         lwkt_serialize_exit(bifp->if_serializer);
 966
 967         if (!gone) {
 968                 switch (ifs->if_type) {
 969                 case IFT_ETHER:
 970                 case IFT_L2VLAN:
 971                         /*
 972                          * Take the interface out of promiscuous mode.
 973                          */
 974                         ifpromisc(ifs, 0);
 975                         bridge_mutecaps(bif_info, ifs, 0);
 976                         break;
 977
 978                 case IFT_GIF:
 979                         break;
 980
 981                 default:
 982                         panic("bridge_delete_member: impossible");
 983                         break;
 984                 }
 985         }
 986
 987         /*
 988          * Remove bifs from percpu linked list.
 989          *
 990          * Removed bifs are not freed immediately, instead,
 991          * they are saved in saved_bifs.  They will be freed
 992          * after we make sure that no one is accessing them,
 993          * i.e. after following netmsg_service_sync()
 994          */
 995         LIST_INIT(&saved_bifs);
 996         bridge_del_bif(sc, bif_info, &saved_bifs);
 997
 998         /*
 999          * Make sure that all protocol threads:
1000          * o  see 'ifs' if_bridge is changed
1001          * o  know that bif is removed from the percpu linked list
1002          */
1003         netmsg_service_sync();
1004
1005         /*
1006          * Free the removed bifs
1007          */
1008         KKASSERT(!LIST_EMPTY(&saved_bifs));
1009         while ((bif = LIST_FIRST(&saved_bifs)) != NULL) {
1010                 LIST_REMOVE(bif, bif_next);
1011                 kfree(bif, M_DEVBUF);
1012         }
1013
1014         /* See the comment in bridge_ioctl_stop() */
1015         bridge_rtmsg_sync(sc);
1016         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1017
1018         lwkt_serialize_enter(bifp->if_serializer);
1019
1020         if (bifp->if_flags & IFF_RUNNING)
1021                 bstp_initialization(sc);
1022
1023         /*
1024          * Free the bif_info after bstp_initialization(), so that
1025          * bridge_softc.sc_root_port will not reference a dangling
1026          * pointer.
1027          */
1028         kfree(bif_info, M_DEVBUF);
1029 }
1030
1031 /*
1032  * bridge_delete_span:
1033  *
1034  *      Delete the specified span interface.
1035  */
1036 static void
1037 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1038 {
1039         KASSERT(bif->bif_ifp->if_bridge == NULL,
1040             ("%s: not a span interface", __func__));
1041
1042         LIST_REMOVE(bif, bif_next);
1043         kfree(bif, M_DEVBUF);
1044 }
1045
1046 static int
1047 bridge_ioctl_init(struct bridge_softc *sc, void *arg __unused)
1048 {
1049         struct ifnet *ifp = sc->sc_ifp;
1050
1051         if (ifp->if_flags & IFF_RUNNING)
1052                 return 0;
1053
1054         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1055             bridge_timer, sc);
1056
1057         ifp->if_flags |= IFF_RUNNING;
1058         bstp_initialization(sc);
1059         return 0;
1060 }
1061
1062 static int
1063 bridge_ioctl_stop(struct bridge_softc *sc, void *arg __unused)
1064 {
1065         struct ifnet *ifp = sc->sc_ifp;
1066         struct lwkt_msg *lmsg;
1067
1068         if ((ifp->if_flags & IFF_RUNNING) == 0)
1069                 return 0;
1070
1071         callout_stop(&sc->sc_brcallout);
1072
1073         crit_enter();
1074         lmsg = &sc->sc_brtimemsg.nm_lmsg;
1075         if ((lmsg->ms_flags & MSGF_DONE) == 0) {
1076                 /* Pending to be processed; drop it */
1077                 lwkt_dropmsg(lmsg);
1078         }
1079         crit_exit();
1080
1081         bstp_stop(sc);
1082
1083         ifp->if_flags &= ~IFF_RUNNING;
1084
1085         lwkt_serialize_exit(ifp->if_serializer);
1086
1087         /* Let everyone know that we are stopped */
1088         netmsg_service_sync();
1089
1090         /*
1091          * Sync ifnetX msgports in the order we forward rtnode
1092          * installation message.  This is used to make sure that
1093          * all rtnode installation messages sent by bridge_rtupdate()
1094          * during above netmsg_service_sync() are flushed.
1095          */
1096         bridge_rtmsg_sync(sc);
1097         bridge_rtflush(sc, IFBF_FLUSHDYN);
1098
1099         lwkt_serialize_enter(ifp->if_serializer);
1100         return 0;
1101 }
1102
1103 static int
1104 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1105 {
1106         struct ifbreq *req = arg;
1107         struct bridge_iflist *bif;
1108         struct bridge_ifinfo *bif_info;
1109         struct ifnet *ifs, *bifp;
1110         int error = 0;
1111
1112         bifp = sc->sc_ifp;
1113         ASSERT_SERIALIZED(bifp->if_serializer);
1114
1115         ifs = ifunit(req->ifbr_ifsname);
1116         if (ifs == NULL)
1117                 return (ENOENT);
1118
1119         /* If it's in the span list, it can't be a member. */
1120         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1121                 if (ifs == bif->bif_ifp)
1122                         return (EBUSY);
1123
1124         /* Allow the first Ethernet member to define the MTU */
1125         if (ifs->if_type != IFT_GIF) {
1126                 if (LIST_EMPTY(&sc->sc_iflists[mycpuid])) {
1127                         bifp->if_mtu = ifs->if_mtu;
1128                 } else if (bifp->if_mtu != ifs->if_mtu) {
1129                         if_printf(bifp, "invalid MTU for %s\n", ifs->if_xname);
1130                         return (EINVAL);
1131                 }
1132         }
1133
1134         if (ifs->if_bridge == sc)
1135                 return (EEXIST);
1136
1137         if (ifs->if_bridge != NULL)
1138                 return (EBUSY);
1139
1140         bif_info = kmalloc(sizeof(*bif_info), M_DEVBUF, M_WAITOK | M_ZERO);
1141         bif_info->bifi_priority = BSTP_DEFAULT_PORT_PRIORITY;
1142         bif_info->bifi_path_cost = BSTP_DEFAULT_PATH_COST;
1143         bif_info->bifi_ifp = ifs;
1144
1145         /*
1146          * Release bridge interface's serializer:
1147          * - To avoid possible dead lock.
1148          * - Various sync operation will block the current thread.
1149          */
1150         lwkt_serialize_exit(bifp->if_serializer);
1151
1152         switch (ifs->if_type) {
1153         case IFT_ETHER:
1154         case IFT_L2VLAN:
1155                 /*
1156                  * Place the interface into promiscuous mode.
1157                  */
1158                 error = ifpromisc(ifs, 1);
1159                 if (error) {
1160                         lwkt_serialize_enter(bifp->if_serializer);
1161                         goto out;
1162                 }
1163                 bridge_mutecaps(bif_info, ifs, 1);
1164                 break;
1165
1166         case IFT_GIF: /* :^) */
1167                 break;
1168
1169         default:
1170                 error = EINVAL;
1171                 lwkt_serialize_enter(bifp->if_serializer);
1172                 goto out;
1173         }
1174
1175         /*
1176          * Add bifs to percpu linked lists
1177          */
1178         bridge_add_bif(sc, bif_info, ifs);
1179
1180         lwkt_serialize_enter(bifp->if_serializer);
1181
1182         if (bifp->if_flags & IFF_RUNNING)
1183                 bstp_initialization(sc);
1184         else
1185                 bstp_stop(sc);
1186
1187         /*
1188          * Everything has been setup, so let the member interface
1189          * deliver packets to this bridge on its input/output path.
1190          */
1191         ifs->if_bridge = sc;
1192 out:
1193         if (error) {
1194                 if (bif_info != NULL)
1195                         kfree(bif_info, M_DEVBUF);
1196         }
1197         return (error);
1198 }
1199
1200 static int
1201 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1202 {
1203         struct ifbreq *req = arg;
1204         struct bridge_iflist *bif;
1205
1206         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1207         if (bif == NULL)
1208                 return (ENOENT);
1209
1210         bridge_delete_member(sc, bif, 0);
1211
1212         return (0);
1213 }
1214
1215 static int
1216 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1217 {
1218         struct ifbreq *req = arg;
1219         struct bridge_iflist *bif;
1220
1221         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1222         if (bif == NULL)
1223                 return (ENOENT);
1224
1225         req->ifbr_ifsflags = bif->bif_flags;
1226         req->ifbr_state = bif->bif_state;
1227         req->ifbr_priority = bif->bif_priority;
1228         req->ifbr_path_cost = bif->bif_path_cost;
1229         req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1230
1231         return (0);
1232 }
1233
1234 static int
1235 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1236 {
1237         struct ifbreq *req = arg;
1238         struct bridge_iflist *bif;
1239         struct ifnet *bifp = sc->sc_ifp;
1240
1241         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1242         if (bif == NULL)
1243                 return (ENOENT);
1244
1245         if (req->ifbr_ifsflags & IFBIF_SPAN) {
1246                 /* SPAN is readonly */
1247                 return (EINVAL);
1248         }
1249
1250         if (req->ifbr_ifsflags & IFBIF_STP) {
1251                 switch (bif->bif_ifp->if_type) {
1252                 case IFT_ETHER:
1253                         /* These can do spanning tree. */
1254                         break;
1255
1256                 default:
1257                         /* Nothing else can. */
1258                         return (EINVAL);
1259                 }
1260         }
1261
1262         lwkt_serialize_exit(bifp->if_serializer);
1263         bridge_set_bifflags(sc, bif->bif_info, req->ifbr_ifsflags);
1264         lwkt_serialize_enter(bifp->if_serializer);
1265
1266         if (bifp->if_flags & IFF_RUNNING)
1267                 bstp_initialization(sc);
1268
1269         return (0);
1270 }
1271
1272 static int
1273 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1274 {
1275         struct ifbrparam *param = arg;
1276         struct ifnet *ifp = sc->sc_ifp;
1277
1278         sc->sc_brtmax = param->ifbrp_csize;
1279
1280         lwkt_serialize_exit(ifp->if_serializer);
1281         bridge_rttrim(sc);
1282         lwkt_serialize_enter(ifp->if_serializer);
1283
1284         return (0);
1285 }
1286
1287 static int
1288 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1289 {
1290         struct ifbrparam *param = arg;
1291
1292         param->ifbrp_csize = sc->sc_brtmax;
1293
1294         return (0);
1295 }
1296
1297 static int
1298 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1299 {
1300         struct bridge_control_arg *bc_arg = arg;
1301         struct ifbifconf *bifc = arg;
1302         struct bridge_iflist *bif;
1303         struct ifbreq *breq;
1304         int count, len;
1305
1306         count = 0;
1307         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next)
1308                 count++;
1309         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1310                 count++;
1311
1312         if (bifc->ifbic_len == 0) {
1313                 bifc->ifbic_len = sizeof(*breq) * count;
1314                 return 0;
1315         } else if (count == 0 || bifc->ifbic_len < sizeof(*breq)) {
1316                 bifc->ifbic_len = 0;
1317                 return 0;
1318         }
1319
1320         len = min(bifc->ifbic_len, sizeof(*breq) * count);
1321         KKASSERT(len >= sizeof(*breq));
1322
1323         breq = kmalloc(len, M_TEMP, M_INTWAIT | M_NULLOK | M_ZERO);
1324         if (breq == NULL) {
1325                 bifc->ifbic_len = 0;
1326                 return ENOMEM;
1327         }
1328         bc_arg->bca_kptr = breq;
1329
1330         count = 0;
1331         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1332                 if (len < sizeof(*breq))
1333                         break;
1334
1335                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1336                         sizeof(breq->ifbr_ifsname));
1337                 breq->ifbr_ifsflags = bif->bif_flags;
1338                 breq->ifbr_state = bif->bif_state;
1339                 breq->ifbr_priority = bif->bif_priority;
1340                 breq->ifbr_path_cost = bif->bif_path_cost;
1341                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1342                 breq++;
1343                 count++;
1344                 len -= sizeof(*breq);
1345         }
1346         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1347                 if (len < sizeof(*breq))
1348                         break;
1349
1350                 strlcpy(breq->ifbr_ifsname, bif->bif_ifp->if_xname,
1351                         sizeof(breq->ifbr_ifsname));
1352                 breq->ifbr_ifsflags = bif->bif_flags;
1353                 breq->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1354                 breq++;
1355                 count++;
1356                 len -= sizeof(*breq);
1357         }
1358
1359         bifc->ifbic_len = sizeof(*breq) * count;
1360         KKASSERT(bifc->ifbic_len > 0);
1361
1362         bc_arg->bca_len = bifc->ifbic_len;
1363         bc_arg->bca_uptr = bifc->ifbic_req;
1364         return 0;
1365 }
1366
1367 static int
1368 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1369 {
1370         struct bridge_control_arg *bc_arg = arg;
1371         struct ifbaconf *bac = arg;
1372         struct bridge_rtnode *brt;
1373         struct ifbareq *bareq;
1374         int count, len;
1375
1376         count = 0;
1377         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list)
1378                 count++;
1379
1380         if (bac->ifbac_len == 0) {
1381                 bac->ifbac_len = sizeof(*bareq) * count;
1382                 return 0;
1383         } else if (count == 0 || bac->ifbac_len < sizeof(*bareq)) {
1384                 bac->ifbac_len = 0;
1385                 return 0;
1386         }
1387
1388         len = min(bac->ifbac_len, sizeof(*bareq) * count);
1389         KKASSERT(len >= sizeof(*bareq));
1390
1391         bareq = kmalloc(len, M_TEMP, M_INTWAIT | M_NULLOK | M_ZERO);
1392         if (bareq == NULL) {
1393                 bac->ifbac_len = 0;
1394                 return ENOMEM;
1395         }
1396         bc_arg->bca_kptr = bareq;
1397
1398         count = 0;
1399         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
1400                 struct bridge_rtinfo *bri = brt->brt_info;
1401                 unsigned long expire;
1402
1403                 if (len < sizeof(*bareq))
1404                         break;
1405
1406                 strlcpy(bareq->ifba_ifsname, bri->bri_ifp->if_xname,
1407                         sizeof(bareq->ifba_ifsname));
1408                 memcpy(bareq->ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1409                 expire = bri->bri_expire;
1410                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1411                     time_second < expire)
1412                         bareq->ifba_expire = expire - time_second;
1413                 else
1414                         bareq->ifba_expire = 0;
1415                 bareq->ifba_flags = bri->bri_flags;
1416                 bareq++;
1417                 count++;
1418                 len -= sizeof(*bareq);
1419         }
1420
1421         bac->ifbac_len = sizeof(*bareq) * count;
1422         KKASSERT(bac->ifbac_len > 0);
1423
1424         bc_arg->bca_len = bac->ifbac_len;
1425         bc_arg->bca_uptr = bac->ifbac_req;
1426         return 0;
1427 }
1428
1429 static int
1430 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1431 {
1432         struct ifbareq *req = arg;
1433         struct bridge_iflist *bif;
1434         struct ifnet *ifp = sc->sc_ifp;
1435         int error;
1436
1437         ASSERT_SERIALIZED(ifp->if_serializer);
1438
1439         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1440         if (bif == NULL)
1441                 return (ENOENT);
1442
1443         lwkt_serialize_exit(ifp->if_serializer);
1444         error = bridge_rtsaddr(sc, req->ifba_dst, bif->bif_ifp,
1445                                req->ifba_flags);
1446         lwkt_serialize_enter(ifp->if_serializer);
1447         return (error);
1448 }
1449
1450 static int
1451 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1452 {
1453         struct ifbrparam *param = arg;
1454
1455         sc->sc_brttimeout = param->ifbrp_ctime;
1456
1457         return (0);
1458 }
1459
1460 static int
1461 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1462 {
1463         struct ifbrparam *param = arg;
1464
1465         param->ifbrp_ctime = sc->sc_brttimeout;
1466
1467         return (0);
1468 }
1469
1470 static int
1471 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1472 {
1473         struct ifbareq *req = arg;
1474         struct ifnet *ifp = sc->sc_ifp;
1475         int error;
1476
1477         lwkt_serialize_exit(ifp->if_serializer);
1478         error = bridge_rtdaddr(sc, req->ifba_dst);
1479         lwkt_serialize_enter(ifp->if_serializer);
1480         return error;
1481 }
1482
1483 static int
1484 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1485 {
1486         struct ifbreq *req = arg;
1487         struct ifnet *ifp = sc->sc_ifp;
1488
1489         lwkt_serialize_exit(ifp->if_serializer);
1490         bridge_rtflush(sc, req->ifbr_ifsflags);
1491         lwkt_serialize_enter(ifp->if_serializer);
1492
1493         return (0);
1494 }
1495
1496 static int
1497 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1498 {
1499         struct ifbrparam *param = arg;
1500
1501         param->ifbrp_prio = sc->sc_bridge_priority;
1502
1503         return (0);
1504 }
1505
1506 static int
1507 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1508 {
1509         struct ifbrparam *param = arg;
1510
1511         sc->sc_bridge_priority = param->ifbrp_prio;
1512
1513         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1514                 bstp_initialization(sc);
1515
1516         return (0);
1517 }
1518
1519 static int
1520 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1521 {
1522         struct ifbrparam *param = arg;
1523
1524         param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
1525
1526         return (0);
1527 }
1528
1529 static int
1530 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1531 {
1532         struct ifbrparam *param = arg;
1533
1534         if (param->ifbrp_hellotime == 0)
1535                 return (EINVAL);
1536         sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
1537
1538         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1539                 bstp_initialization(sc);
1540
1541         return (0);
1542 }
1543
1544 static int
1545 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1546 {
1547         struct ifbrparam *param = arg;
1548
1549         param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
1550
1551         return (0);
1552 }
1553
1554 static int
1555 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1556 {
1557         struct ifbrparam *param = arg;
1558
1559         if (param->ifbrp_fwddelay == 0)
1560                 return (EINVAL);
1561         sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1562
1563         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1564                 bstp_initialization(sc);
1565
1566         return (0);
1567 }
1568
1569 static int
1570 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1571 {
1572         struct ifbrparam *param = arg;
1573
1574         param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
1575
1576         return (0);
1577 }
1578
1579 static int
1580 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1581 {
1582         struct ifbrparam *param = arg;
1583
1584         if (param->ifbrp_maxage == 0)
1585                 return (EINVAL);
1586         sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
1587
1588         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1589                 bstp_initialization(sc);
1590
1591         return (0);
1592 }
1593
1594 static int
1595 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1596 {
1597         struct ifbreq *req = arg;
1598         struct bridge_iflist *bif;
1599
1600         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1601         if (bif == NULL)
1602                 return (ENOENT);
1603
1604         bif->bif_priority = req->ifbr_priority;
1605
1606         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1607                 bstp_initialization(sc);
1608
1609         return (0);
1610 }
1611
1612 static int
1613 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1614 {
1615         struct ifbreq *req = arg;
1616         struct bridge_iflist *bif;
1617
1618         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1619         if (bif == NULL)
1620                 return (ENOENT);
1621
1622         bif->bif_path_cost = req->ifbr_path_cost;
1623
1624         if (sc->sc_ifp->if_flags & IFF_RUNNING)
1625                 bstp_initialization(sc);
1626
1627         return (0);
1628 }
1629
1630 static int
1631 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1632 {
1633         struct ifbreq *req = arg;
1634         struct bridge_iflist *bif;
1635         struct ifnet *ifs;
1636
1637         ifs = ifunit(req->ifbr_ifsname);
1638         if (ifs == NULL)
1639                 return (ENOENT);
1640
1641         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1642                 if (ifs == bif->bif_ifp)
1643                         return (EBUSY);
1644
1645         if (ifs->if_bridge != NULL)
1646                 return (EBUSY);
1647
1648         switch (ifs->if_type) {
1649         case IFT_ETHER:
1650         case IFT_GIF:
1651         case IFT_L2VLAN:
1652                 break;
1653
1654         default:
1655                 return (EINVAL);
1656         }
1657
1658         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
1659         bif->bif_ifp = ifs;
1660         bif->bif_flags = IFBIF_SPAN;
1661         /* NOTE: span bif does not need bridge_ifinfo */
1662
1663         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1664
1665         return (0);
1666 }
1667
1668 static int
1669 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1670 {
1671         struct ifbreq *req = arg;
1672         struct bridge_iflist *bif;
1673         struct ifnet *ifs;
1674
1675         ifs = ifunit(req->ifbr_ifsname);
1676         if (ifs == NULL)
1677                 return (ENOENT);
1678
1679         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1680                 if (ifs == bif->bif_ifp)
1681                         break;
1682
1683         if (bif == NULL)
1684                 return (ENOENT);
1685
1686         bridge_delete_span(sc, bif);
1687
1688         return (0);
1689 }
1690
1691 static void
1692 bridge_ifdetach_dispatch(struct netmsg *nmsg)
1693 {
1694         struct lwkt_msg *lmsg = &nmsg->nm_lmsg;
1695         struct ifnet *ifp, *bifp;
1696         struct bridge_softc *sc;
1697         struct bridge_iflist *bif;
1698
1699         ifp = lmsg->u.ms_resultp;
1700         sc = ifp->if_bridge;
1701
1702         /* Check if the interface is a bridge member */
1703         if (sc != NULL) {
1704                 bifp = sc->sc_ifp;
1705
1706                 lwkt_serialize_enter(bifp->if_serializer);
1707
1708                 bif = bridge_lookup_member_if(sc, ifp);
1709                 if (bif != NULL) {
1710                         bridge_delete_member(sc, bif, 1);
1711                 } else {
1712                         /* XXX Why bif will be NULL? */
1713                 }
1714
1715                 lwkt_serialize_exit(bifp->if_serializer);
1716                 goto reply;
1717         }
1718
1719         crit_enter();   /* XXX MP */
1720
1721         /* Check if the interface is a span port */
1722         LIST_FOREACH(sc, &bridge_list, sc_list) {
1723                 bifp = sc->sc_ifp;
1724
1725                 lwkt_serialize_enter(bifp->if_serializer);
1726
1727                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1728                         if (ifp == bif->bif_ifp) {
1729                                 bridge_delete_span(sc, bif);
1730                                 break;
1731                         }
1732
1733                 lwkt_serialize_exit(bifp->if_serializer);
1734         }
1735
1736         crit_exit();
1737
1738 reply:
1739         lwkt_replymsg(lmsg, 0);
1740 }
1741
1742 /*
1743  * bridge_ifdetach:
1744  *
1745  *      Detach an interface from a bridge.  Called when a member
1746  *      interface is detaching.
1747  */
1748 static void
1749 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1750 {
1751         struct lwkt_msg *lmsg;
1752         struct netmsg nmsg;
1753
1754         netmsg_init(&nmsg, &curthread->td_msgport, 0, bridge_ifdetach_dispatch);
1755         lmsg = &nmsg.nm_lmsg;
1756         lmsg->u.ms_resultp = ifp;
1757
1758         lwkt_domsg(BRIDGE_CFGPORT, lmsg, 0);
1759 }
1760
1761 /*
1762  * bridge_init:
1763  *
1764  *      Initialize a bridge interface.
1765  */
1766 static void
1767 bridge_init(void *xsc)
1768 {
1769         bridge_control(xsc, SIOCSIFFLAGS, bridge_ioctl_init, NULL);
1770 }
1771
1772 /*
1773  * bridge_stop:
1774  *
1775  *      Stop the bridge interface.
1776  */
1777 static void
1778 bridge_stop(struct ifnet *ifp)
1779 {
1780         bridge_control(ifp->if_softc, SIOCSIFFLAGS, bridge_ioctl_stop, NULL);
1781 }
1782
1783 static void
1784 bridge_enqueue_internal(struct ifnet *dst_ifp, struct mbuf *m,
1785                         netisr_fn_t handler)
1786 {
1787         struct netmsg_packet *nmp;
1788         lwkt_port_t port;
1789         int cpu = mycpu->gd_cpuid;
1790
1791         nmp = &m->m_hdr.mh_netmsg;
1792         netmsg_init(&nmp->nm_netmsg, &netisr_apanic_rport, 0, handler);
1793         nmp->nm_packet = m;
1794         nmp->nm_netmsg.nm_lmsg.u.ms_resultp = dst_ifp;
1795
1796         port = cpu_portfn(cpu);
1797         lwkt_sendmsg(port, &nmp->nm_netmsg.nm_lmsg);
1798 }
1799
1800 static void
1801 bridge_pfil_enqueue(struct ifnet *dst_ifp, struct mbuf *m,
1802                     int runfilt)
1803 {
1804         netisr_fn_t handler;
1805
1806         if (runfilt && (inet_pfil_hook.ph_hashooks > 0
1807 #ifdef INET6
1808             || inet6_pfil_hook.ph_hashooks > 0
1809 #endif
1810             )) {
1811                 handler = bridge_pfil_enqueue_handler;
1812         } else {
1813                 handler = bridge_enqueue_handler;
1814         }
1815         bridge_enqueue_internal(dst_ifp, m, handler);
1816 }
1817
1818 /*
1819  * bridge_enqueue:
1820  *
1821  *      Enqueue a packet on a bridge member interface.
1822  *
1823  */
1824 void
1825 bridge_enqueue(struct ifnet *dst_ifp, struct mbuf *m)
1826 {
1827         bridge_enqueue_internal(dst_ifp, m, bridge_enqueue_handler);
1828 }
1829
1830 /*
1831  * bridge_output:
1832  *
1833  *      Send output from a bridge member interface.  This
1834  *      performs the bridging function for locally originated
1835  *      packets.
1836  *
1837  *      The mbuf has the Ethernet header already attached.  We must
1838  *      enqueue or free the mbuf before returning.
1839  */
1840 static int
1841 bridge_output(struct ifnet *ifp, struct mbuf *m)
1842 {
1843         struct bridge_softc *sc = ifp->if_bridge;
1844         struct ether_header *eh;
1845         struct ifnet *dst_if;
1846
1847         ASSERT_NOT_SERIALIZED(ifp->if_serializer);
1848
1849         /*
1850          * Make sure that we are still a member of a bridge interface.
1851          */
1852         if (sc == NULL) {
1853                 m_freem(m);
1854                 return (0);
1855         }
1856
1857         if (m->m_len < ETHER_HDR_LEN) {
1858                 m = m_pullup(m, ETHER_HDR_LEN);
1859                 if (m == NULL)
1860                         return (0);
1861         }
1862
1863         /* Serialize our bridge interface. */
1864         lwkt_serialize_enter(sc->sc_ifp->if_serializer);
1865
1866         eh = mtod(m, struct ether_header *);
1867
1868         /*
1869          * If bridge is down, but the original output interface is up,
1870          * go ahead and send out that interface.  Otherwise, the packet
1871          * is dropped below.
1872          */
1873         if ((sc->sc_ifp->if_flags & IFF_RUNNING) == 0) {
1874                 dst_if = ifp;
1875                 goto sendunicast;
1876         }
1877
1878         /*
1879          * If the packet is a multicast, or we don't know a better way to
1880          * get there, send to all interfaces.
1881          */
1882         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1883                 dst_if = NULL;
1884         else
1885                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1886         if (dst_if == NULL) {
1887                 struct bridge_iflist *bif;
1888                 struct mbuf *mc;
1889                 int used = 0;
1890
1891                 bridge_span(sc, m);
1892
1893                 /*
1894                  * Following loop is MPSAFE; nothing is blocking
1895                  * in the loop body.
1896                  */
1897                 LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
1898                         dst_if = bif->bif_ifp;
1899                         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1900                                 continue;
1901
1902                         /*
1903                          * If this is not the original output interface,
1904                          * and the interface is participating in spanning
1905                          * tree, make sure the port is in a state that
1906                          * allows forwarding.
1907                          */
1908                         if (dst_if != ifp &&
1909                             (bif->bif_flags & IFBIF_STP) != 0) {
1910                                 switch (bif->bif_state) {
1911                                 case BSTP_IFSTATE_BLOCKING:
1912                                 case BSTP_IFSTATE_LISTENING:
1913                                 case BSTP_IFSTATE_DISABLED:
1914                                         continue;
1915                                 }
1916                         }
1917
1918                         if (LIST_NEXT(bif, bif_next) == NULL) {
1919                                 used = 1;
1920                                 mc = m;
1921                         } else {
1922                                 mc = m_copypacket(m, MB_DONTWAIT);
1923                                 if (mc == NULL) {
1924                                         sc->sc_ifp->if_oerrors++;
1925                                         continue;
1926                                 }
1927                         }
1928                         bridge_enqueue(dst_if, mc);
1929                 }
1930                 if (used == 0)
1931                         m_freem(m);
1932                 lwkt_serialize_exit(sc->sc_ifp->if_serializer);
1933                 return (0);
1934         }
1935
1936 sendunicast:
1937         /*
1938          * XXX Spanning tree consideration here?
1939          */
1940
1941         bridge_span(sc, m);
1942         lwkt_serialize_exit(sc->sc_ifp->if_serializer);
1943         if ((dst_if->if_flags & IFF_RUNNING) == 0)
1944                 m_freem(m);
1945         else
1946                 bridge_enqueue(dst_if, m);
1947         return (0);
1948 }
1949
1950 /*
1951  * bridge_start:
1952  *
1953  *      Start output on a bridge.
1954  *
1955  */
1956 static void
1957 bridge_start(struct ifnet *ifp)
1958 {
1959         struct bridge_softc *sc = ifp->if_softc;
1960
1961         ASSERT_SERIALIZED(ifp->if_serializer);
1962
1963         ifp->if_flags |= IFF_OACTIVE;
1964         for (;;) {
1965                 struct ifnet *dst_if = NULL;
1966                 struct ether_header *eh;
1967                 struct mbuf *m;
1968
1969                 m = ifq_dequeue(&ifp->if_snd, NULL);
1970                 if (m == NULL)
1971                         break;
1972
1973                 if (m->m_len < sizeof(*eh)) {
1974                         m = m_pullup(m, sizeof(*eh));
1975                         if (m == NULL) {
1976                                 ifp->if_oerrors++;
1977                                 continue;
1978                         }
1979                 }
1980                 eh = mtod(m, struct ether_header *);
1981
1982                 BPF_MTAP(ifp, m);
1983                 ifp->if_opackets++;
1984
1985                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0)
1986                         dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1987
1988                 if (dst_if == NULL)
1989                         bridge_broadcast(sc, ifp, m, 0);
1990                 else
1991                         bridge_enqueue(dst_if, m);
1992         }
1993         ifp->if_flags &= ~IFF_OACTIVE;
1994 }
1995
1996 /*
1997  * bridge_forward:
1998  *
1999  *      The forwarding function of the bridge.
2000  */
2001 static void
2002 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
2003 {
2004         struct bridge_iflist *bif;
2005         struct ifnet *src_if, *dst_if, *ifp;
2006         struct ether_header *eh;
2007
2008         src_if = m->m_pkthdr.rcvif;
2009         ifp = sc->sc_ifp;
2010
2011         ASSERT_SERIALIZED(ifp->if_serializer);
2012
2013         ifp->if_ipackets++;
2014         ifp->if_ibytes += m->m_pkthdr.len;
2015
2016         /*
2017          * Look up the bridge_iflist.
2018          */
2019         bif = bridge_lookup_member_if(sc, src_if);
2020         if (bif == NULL) {
2021                 /* Interface is not a bridge member (anymore?) */
2022                 m_freem(m);
2023                 return;
2024         }
2025
2026         if (bif->bif_flags & IFBIF_STP) {
2027                 switch (bif->bif_state) {
2028                 case BSTP_IFSTATE_BLOCKING:
2029                 case BSTP_IFSTATE_LISTENING:
2030                 case BSTP_IFSTATE_DISABLED:
2031                         m_freem(m);
2032                         return;
2033                 }
2034         }
2035
2036         eh = mtod(m, struct ether_header *);
2037
2038         /*
2039          * If the interface is learning, and the source
2040          * address is valid and not multicast, record
2041          * the address.
2042          */
2043         if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
2044             ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
2045             (eh->ether_shost[0] == 0 &&
2046              eh->ether_shost[1] == 0 &&
2047              eh->ether_shost[2] == 0 &&
2048              eh->ether_shost[3] == 0 &&
2049              eh->ether_shost[4] == 0 &&
2050              eh->ether_shost[5] == 0) == 0)
2051                 bridge_rtupdate(sc, eh->ether_shost, src_if, IFBAF_DYNAMIC);
2052
2053         if ((bif->bif_flags & IFBIF_STP) != 0 &&
2054             bif->bif_state == BSTP_IFSTATE_LEARNING) {
2055                 m_freem(m);
2056                 return;
2057         }
2058
2059         /*
2060          * At this point, the port either doesn't participate
2061          * in spanning tree or it is in the forwarding state.
2062          */
2063
2064         /*
2065          * If the packet is unicast, destined for someone on
2066          * "this" side of the bridge, drop it.
2067          */
2068         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2069                 dst_if = bridge_rtlookup(sc, eh->ether_dhost);
2070                 if (src_if == dst_if) {
2071                         m_freem(m);
2072                         return;
2073                 }
2074         } else {
2075                 /* ...forward it to all interfaces. */
2076                 sc->sc_ifp->if_imcasts++;
2077                 dst_if = NULL;
2078         }
2079
2080         if (dst_if == NULL) {
2081                 bridge_broadcast(sc, src_if, m, 1);
2082                 return;
2083         }
2084
2085         /*
2086          * At this point, we're dealing with a unicast frame
2087          * going to a different interface.
2088          */
2089         if ((dst_if->if_flags & IFF_RUNNING) == 0) {
2090                 m_freem(m);
2091                 return;
2092         }
2093         bif = bridge_lookup_member_if(sc, dst_if);
2094         if (bif == NULL) {
2095                 /* Not a member of the bridge (anymore?) */
2096                 m_freem(m);
2097                 return;
2098         }
2099
2100         if (bif->bif_flags & IFBIF_STP) {
2101                 switch (bif->bif_state) {
2102                 case BSTP_IFSTATE_DISABLED:
2103                 case BSTP_IFSTATE_BLOCKING:
2104                         m_freem(m);
2105                         return;
2106                 }
2107         }
2108
2109         lwkt_serialize_exit(ifp->if_serializer);
2110
2111         /* run the packet filter */
2112         if (inet_pfil_hook.ph_hashooks > 0
2113 #ifdef INET6
2114             || inet6_pfil_hook.ph_hashooks > 0
2115 #endif
2116             ) {
2117                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2118                         goto done;
2119                 if (m == NULL)
2120                         goto done;
2121
2122                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2123                         goto done;
2124                 if (m == NULL)
2125                         goto done;
2126         }
2127         bridge_handoff(dst_if, m);
2128
2129         /*
2130          * ifp's serializer was held on entry and is expected to be held
2131          * on return.
2132          */
2133 done:
2134         lwkt_serialize_enter(ifp->if_serializer);
2135 }
2136
2137 /*
2138  * bridge_input:
2139  *
2140  *      Receive input from a member interface.  Queue the packet for
2141  *      bridging if it is not for us.
2142  */
2143 static struct mbuf *
2144 bridge_input(struct ifnet *ifp, struct mbuf *m)
2145 {
2146         struct bridge_softc *sc = ifp->if_bridge;
2147         struct bridge_iflist *bif;
2148         struct ifnet *bifp, *new_ifp;
2149         struct ether_header *eh;
2150         struct mbuf *mc, *mc2;
2151
2152         /*
2153          * Make sure that we are still a member of a bridge interface.
2154          */
2155         if (sc == NULL)
2156                 return m;
2157
2158         new_ifp = NULL;
2159         bifp = sc->sc_ifp;
2160
2161         lwkt_serialize_enter(bifp->if_serializer);
2162
2163         if ((bifp->if_flags & IFF_RUNNING) == 0)
2164                 goto out;
2165
2166         /*
2167          * Implement support for bridge monitoring.  If this flag has been
2168          * set on this interface, discard the packet once we push it through
2169          * the bpf(4) machinery, but before we do, increment various counters
2170          * associated with this bridge.
2171          */
2172         if (bifp->if_flags & IFF_MONITOR) {
2173                 /* Change input interface to this bridge */
2174                 m->m_pkthdr.rcvif = bifp;
2175
2176                 BPF_MTAP(bifp, m);
2177
2178                 /* Update bridge's ifnet statistics */
2179                 bifp->if_ipackets++;
2180                 bifp->if_ibytes += m->m_pkthdr.len;
2181                 if (m->m_flags & (M_MCAST | M_BCAST))
2182                         bifp->if_imcasts++;
2183
2184                 m_freem(m);
2185                 m = NULL;
2186                 goto out;
2187         }
2188
2189         eh = mtod(m, struct ether_header *);
2190
2191         m->m_flags &= ~M_PROTO1; /* XXX Hack - loop prevention */
2192
2193         if (memcmp(eh->ether_dhost, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
2194                 /*
2195                  * If the packet is for us, set the packets source as the
2196                  * bridge, and return the packet back to ifnet.if_input for
2197                  * local processing.
2198                  */
2199                 KASSERT(bifp->if_bridge == NULL,
2200                         ("loop created in bridge_input"));
2201                 new_ifp = bifp;
2202                 goto out;
2203         }
2204
2205         /*
2206          * Tap all packets arriving on the bridge, no matter if
2207          * they are local destinations or not.  In is in.
2208          */
2209         BPF_MTAP(bifp, m);
2210
2211         bif = bridge_lookup_member_if(sc, ifp);
2212         if (bif == NULL)
2213                 goto out;
2214
2215         bridge_span(sc, m);
2216
2217         if (m->m_flags & (M_BCAST | M_MCAST)) {
2218                 /* Tap off 802.1D packets; they do not get forwarded. */
2219                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2220                     ETHER_ADDR_LEN) == 0) {
2221                         m = bstp_input(sc, bif, m);
2222                         KASSERT(m == NULL,
2223                                 ("attempt to deliver 802.1D packet\n"));
2224                         goto out;
2225                 }
2226
2227                 if (bif->bif_flags & IFBIF_STP) {
2228                         switch (bif->bif_state) {
2229                         case BSTP_IFSTATE_BLOCKING:
2230                         case BSTP_IFSTATE_LISTENING:
2231                         case BSTP_IFSTATE_DISABLED:
2232                                 goto out;
2233                         }
2234                 }
2235
2236                 /*
2237                  * Make a deep copy of the packet and enqueue the copy
2238                  * for bridge processing; return the original packet for
2239                  * local processing.
2240                  */
2241                 mc = m_dup(m, MB_DONTWAIT);
2242                 if (mc == NULL)
2243                         goto out;
2244
2245                 bridge_forward(sc, mc);
2246
2247                 /*
2248                  * Reinject the mbuf as arriving on the bridge so we have a
2249                  * chance at claiming multicast packets. We can not loop back
2250                  * here from ether_input as a bridge is never a member of a
2251                  * bridge.
2252                  */
2253                 KASSERT(bifp->if_bridge == NULL,
2254                         ("loop created in bridge_input"));
2255                 mc2 = m_dup(m, MB_DONTWAIT);
2256 #ifdef notyet
2257                 if (mc2 != NULL) {
2258                         /* Keep the layer3 header aligned */
2259                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2260                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2261                 }
2262 #endif
2263                 if (mc2 != NULL) {
2264                         mc2->m_pkthdr.rcvif = bifp;
2265                         bifp->if_ipackets++;
2266                         bifp->if_input(bifp, mc2);
2267                 }
2268
2269                 /* Return the original packet for local processing. */
2270                 goto out;
2271         }
2272
2273         if (bif->bif_flags & IFBIF_STP) {
2274                 switch (bif->bif_state) {
2275                 case BSTP_IFSTATE_BLOCKING:
2276                 case BSTP_IFSTATE_LISTENING:
2277                 case BSTP_IFSTATE_DISABLED:
2278                         goto out;
2279                 }
2280         }
2281
2282         /*
2283          * Unicast.  Make sure it's not for us.
2284          *
2285          * This loop is MPSAFE; the only blocking operation (bridge_rtupdate)
2286          * is followed by breaking out of the loop.
2287          */
2288         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2289                 if (bif->bif_ifp->if_type != IFT_ETHER)
2290                         continue;
2291
2292                 /* It is destined for us. */
2293                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
2294                     ETHER_ADDR_LEN) == 0) {
2295                         if (bif->bif_ifp != ifp) {
2296                                 /* XXX loop prevention */
2297                                 m->m_flags |= M_PROTO1;
2298                                 new_ifp = bif->bif_ifp;
2299                         }
2300                         if (bif->bif_flags & IFBIF_LEARNING) {
2301                                 bridge_rtupdate(sc, eh->ether_shost,
2302                                                 ifp, IFBAF_DYNAMIC);
2303                         }
2304                         goto out;
2305                 }
2306
2307                 /* We just received a packet that we sent out. */
2308                 if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
2309                     ETHER_ADDR_LEN) == 0) {
2310                         m_freem(m);
2311                         m = NULL;
2312                         goto out;
2313                 }
2314         }
2315
2316         /* Perform the bridge forwarding function. */
2317         bridge_forward(sc, m);
2318         m = NULL;
2319 out:
2320         lwkt_serialize_exit(bifp->if_serializer);
2321
2322         if (new_ifp != NULL) {
2323                 lwkt_serialize_enter(new_ifp->if_serializer);
2324
2325                 m->m_pkthdr.rcvif = new_ifp;
2326                 new_ifp->if_ipackets++;
2327                 new_ifp->if_input(new_ifp, m);
2328                 m = NULL;
2329
2330                 lwkt_serialize_exit(new_ifp->if_serializer);
2331         }
2332         return (m);
2333 }
2334
2335 /*
2336  * bridge_broadcast:
2337  *
2338  *      Send a frame to all interfaces that are members of
2339  *      the bridge, except for the one on which the packet
2340  *      arrived.
2341  */
2342 static void
2343 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2344     struct mbuf *m, int runfilt)
2345 {
2346         struct bridge_iflist *bif;
2347         struct mbuf *mc;
2348         struct ifnet *dst_if, *bifp;
2349         int used = 0;
2350
2351         bifp = sc->sc_ifp;
2352
2353         ASSERT_SERIALIZED(bifp->if_serializer);
2354
2355         /* run the packet filter */
2356         if (runfilt && (inet_pfil_hook.ph_hashooks > 0
2357 #ifdef INET6
2358             || inet6_pfil_hook.ph_hashooks > 0
2359 #endif
2360             )) {
2361                 lwkt_serialize_exit(bifp->if_serializer);
2362
2363                 /* Filter on the bridge interface before broadcasting */
2364
2365                 if (bridge_pfil(&m, bifp, src_if, PFIL_IN) != 0)
2366                         goto filt;
2367                 if (m == NULL)
2368                         goto filt;
2369
2370                 if (bridge_pfil(&m, bifp, NULL, PFIL_OUT) != 0)
2371                         m = NULL;
2372 filt:
2373                 lwkt_serialize_enter(bifp->if_serializer);
2374                 if (m == NULL)
2375                         return;
2376         }
2377
2378         /*
2379          * Following loop is MPSAFE; nothing is blocking
2380          * in the loop body.
2381          */
2382         LIST_FOREACH(bif, &sc->sc_iflists[mycpuid], bif_next) {
2383                 dst_if = bif->bif_ifp;
2384                 if (dst_if == src_if)
2385                         continue;
2386
2387                 if (bif->bif_flags & IFBIF_STP) {
2388                         switch (bif->bif_state) {
2389                         case BSTP_IFSTATE_BLOCKING:
2390                         case BSTP_IFSTATE_DISABLED:
2391                                 continue;
2392                         }
2393                 }
2394
2395                 if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2396                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2397                         continue;
2398
2399                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2400                         continue;
2401
2402                 if (LIST_NEXT(bif, bif_next) == NULL) {
2403                         mc = m;
2404                         used = 1;
2405                 } else {
2406                         mc = m_copypacket(m, MB_DONTWAIT);
2407                         if (mc == NULL) {
2408                                 sc->sc_ifp->if_oerrors++;
2409                                 continue;
2410                         }
2411                 }
2412                 bridge_pfil_enqueue(dst_if, mc, runfilt);
2413         }
2414         if (used == 0)
2415                 m_freem(m);
2416 }
2417
2418 /*
2419  * bridge_span:
2420  *
2421  *      Duplicate a packet out one or more interfaces that are in span mode,
2422  *      the original mbuf is unmodified.
2423  */
2424 static void
2425 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2426 {
2427         struct bridge_iflist *bif;
2428         struct ifnet *dst_if;
2429         struct mbuf *mc;
2430
2431         if (LIST_EMPTY(&sc->sc_spanlist))
2432                 return;
2433
2434         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2435                 dst_if = bif->bif_ifp;
2436
2437                 if ((dst_if->if_flags & IFF_RUNNING) == 0)
2438                         continue;
2439
2440                 mc = m_copypacket(m, MB_DONTWAIT);
2441                 if (mc == NULL) {
2442                         sc->sc_ifp->if_oerrors++;
2443                         continue;
2444                 }
2445
2446                 bridge_enqueue(dst_if, mc);
2447         }
2448 }
2449
2450 static void
2451 bridge_rtmsg_sync_handler(struct netmsg *nmsg)
2452 {
2453         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
2454 }
2455
2456 static void
2457 bridge_rtmsg_sync(struct bridge_softc *sc)
2458 {
2459         struct netmsg nmsg;
2460
2461         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2462
2463         netmsg_init(&nmsg, &curthread->td_msgport, 0,
2464                     bridge_rtmsg_sync_handler);
2465         ifnet_domsg(&nmsg.nm_lmsg, 0);
2466 }
2467
2468 static __inline void
2469 bridge_rtinfo_update(struct bridge_rtinfo *bri, struct ifnet *dst_if,
2470                      int setflags, uint8_t flags, uint32_t timeo)
2471 {
2472         if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2473             bri->bri_ifp != dst_if)
2474                 bri->bri_ifp = dst_if;
2475         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2476             bri->bri_expire != time_second + timeo)
2477                 bri->bri_expire = time_second + timeo;
2478         if (setflags)
2479                 bri->bri_flags = flags;
2480 }
2481
2482 static int
2483 bridge_rtinstall_oncpu(struct bridge_softc *sc, const uint8_t *dst,
2484                        struct ifnet *dst_if, int setflags, uint8_t flags,
2485                        struct bridge_rtinfo **bri0)
2486 {
2487         struct bridge_rtnode *brt;
2488         struct bridge_rtinfo *bri;
2489
2490         if (mycpuid == 0) {
2491                 brt = bridge_rtnode_lookup(sc, dst);
2492                 if (brt != NULL) {
2493                         /*
2494                          * rtnode for 'dst' already exists.  We inform the
2495                          * caller about this by leaving bri0 as NULL.  The
2496                          * caller will terminate the intallation upon getting
2497                          * NULL bri0.  However, we still need to update the
2498                          * rtinfo.
2499                          */
2500                         KKASSERT(*bri0 == NULL);
2501
2502                         /* Update rtinfo */
2503                         bridge_rtinfo_update(brt->brt_info, dst_if, setflags,
2504                                              flags, sc->sc_brttimeout);
2505                         return 0;
2506                 }
2507
2508                 /*
2509                  * We only need to check brtcnt on CPU0, since if limit
2510                  * is to be exceeded, ENOSPC is returned.  Caller knows
2511                  * this and will terminate the installation.
2512                  */
2513                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2514                         return ENOSPC;
2515
2516                 KKASSERT(*bri0 == NULL);
2517                 bri = kmalloc(sizeof(struct bridge_rtinfo), M_DEVBUF,
2518                                   M_WAITOK | M_ZERO);
2519                 *bri0 = bri;
2520
2521                 /* Setup rtinfo */
2522                 bri->bri_flags = IFBAF_DYNAMIC;
2523                 bridge_rtinfo_update(bri, dst_if, setflags, flags,
2524                                      sc->sc_brttimeout);
2525         } else {
2526                 bri = *bri0;
2527                 KKASSERT(bri != NULL);
2528         }
2529
2530         brt = kmalloc(sizeof(struct bridge_rtnode), M_DEVBUF,
2531                       M_WAITOK | M_ZERO);
2532         memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2533         brt->brt_info = bri;
2534
2535         bridge_rtnode_insert(sc, brt);
2536         return 0;
2537 }
2538
2539 static void
2540 bridge_rtinstall_handler(struct netmsg *nmsg)
2541 {
2542         struct netmsg_brsaddr *brmsg = (struct netmsg_brsaddr *)nmsg;
2543         int error;
2544
2545         error = bridge_rtinstall_oncpu(brmsg->br_softc,
2546                                        brmsg->br_dst, brmsg->br_dst_if,
2547                                        brmsg->br_setflags, brmsg->br_flags,
2548                                        &brmsg->br_rtinfo);
2549         if (error) {
2550                 KKASSERT(mycpuid == 0 && brmsg->br_rtinfo == NULL);
2551                 lwkt_replymsg(&nmsg->nm_lmsg, error);
2552                 return;
2553         } else if (brmsg->br_rtinfo == NULL) {
2554                 /* rtnode already exists for 'dst' */
2555                 KKASSERT(mycpuid == 0);
2556                 lwkt_replymsg(&nmsg->nm_lmsg, 0);
2557                 return;
2558         }
2559         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
2560 }
2561
2562 /*
2563  * bridge_rtupdate:
2564  *
2565  *      Add/Update a bridge routing entry.
2566  */
2567 static int
2568 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2569                 struct ifnet *dst_if, uint8_t flags)
2570 {
2571         struct bridge_rtnode *brt;
2572
2573         /*
2574          * A route for this destination might already exist.  If so,
2575          * update it, otherwise create a new one.
2576          */
2577         if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
2578                 struct netmsg_brsaddr *brmsg;
2579
2580                 if (sc->sc_brtcnt >= sc->sc_brtmax)
2581                         return ENOSPC;
2582
2583                 brmsg = kmalloc(sizeof(*brmsg), M_LWKTMSG, M_WAITOK | M_NULLOK);
2584                 if (brmsg == NULL)
2585                         return ENOMEM;
2586
2587                 netmsg_init(&brmsg->br_nmsg, &netisr_afree_rport, 0,
2588                             bridge_rtinstall_handler);
2589                 memcpy(brmsg->br_dst, dst, ETHER_ADDR_LEN);
2590                 brmsg->br_dst_if = dst_if;
2591                 brmsg->br_flags = flags;
2592                 brmsg->br_setflags = 0;
2593                 brmsg->br_softc = sc;
2594                 brmsg->br_rtinfo = NULL;
2595
2596                 ifnet_sendmsg(&brmsg->br_nmsg.nm_lmsg, 0);
2597                 return 0;
2598         }
2599         bridge_rtinfo_update(brt->brt_info, dst_if, 0, flags,
2600                              sc->sc_brttimeout);
2601         return 0;
2602 }
2603
2604 static int
2605 bridge_rtsaddr(struct bridge_softc *sc, const uint8_t *dst,
2606                struct ifnet *dst_if, uint8_t flags)
2607 {
2608         struct netmsg_brsaddr brmsg;
2609
2610         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2611
2612         netmsg_init(&brmsg.br_nmsg, &curthread->td_msgport, MSGF_PRIORITY,
2613                     bridge_rtinstall_handler);
2614         memcpy(brmsg.br_dst, dst, ETHER_ADDR_LEN);
2615         brmsg.br_dst_if = dst_if;
2616         brmsg.br_flags = flags;
2617         brmsg.br_setflags = 1;
2618         brmsg.br_softc = sc;
2619         brmsg.br_rtinfo = NULL;
2620
2621         return ifnet_domsg(&brmsg.br_nmsg.nm_lmsg, 0);
2622 }
2623
2624 /*
2625  * bridge_rtlookup:
2626  *
2627  *      Lookup the destination interface for an address.
2628  */
2629 static struct ifnet *
2630 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2631 {
2632         struct bridge_rtnode *brt;
2633
2634         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2635                 return NULL;
2636         return brt->brt_info->bri_ifp;
2637 }
2638
2639 static void
2640 bridge_rtreap_handler(struct netmsg *nmsg)
2641 {
2642         struct bridge_softc *sc = nmsg->nm_lmsg.u.ms_resultp;
2643         struct bridge_rtnode *brt, *nbrt;
2644
2645         LIST_FOREACH_MUTABLE(brt, &sc->sc_rtlists[mycpuid], brt_list, nbrt) {
2646                 if (brt->brt_info->bri_dead)
2647                         bridge_rtnode_destroy(sc, brt);
2648         }
2649         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
2650 }
2651
2652 static void
2653 bridge_rtreap(struct bridge_softc *sc)
2654 {
2655         struct netmsg nmsg;
2656
2657         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2658
2659         netmsg_init(&nmsg, &curthread->td_msgport, 0, bridge_rtreap_handler);
2660         nmsg.nm_lmsg.u.ms_resultp = sc;
2661
2662         ifnet_domsg(&nmsg.nm_lmsg, 0);
2663 }
2664
2665 /*
2666  * bridge_rttrim:
2667  *
2668  *      Trim the routine table so that we have a number
2669  *      of routing entries less than or equal to the
2670  *      maximum number.
2671  */
2672 static void
2673 bridge_rttrim(struct bridge_softc *sc)
2674 {
2675         struct bridge_rtnode *brt;
2676         int dead;
2677
2678         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2679
2680         /* Make sure we actually need to do this. */
2681         if (sc->sc_brtcnt <= sc->sc_brtmax)
2682                 return;
2683
2684         /*
2685          * Find out how many rtnodes are dead
2686          */
2687         dead = bridge_rtage_finddead(sc);
2688         KKASSERT(dead <= sc->sc_brtcnt);
2689
2690         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2691                 /* Enough dead rtnodes are found */
2692                 bridge_rtreap(sc);
2693                 return;
2694         }
2695
2696         /*
2697          * Kill some dynamic rtnodes to meet the brtmax
2698          */
2699         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2700                 struct bridge_rtinfo *bri = brt->brt_info;
2701
2702                 if (bri->bri_dead) {
2703                         /*
2704                          * We have counted this rtnode in
2705                          * bridge_rtage_finddead()
2706                          */
2707                         continue;
2708                 }
2709
2710                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2711                         bri->bri_dead = 1;
2712                         ++dead;
2713                         KKASSERT(dead <= sc->sc_brtcnt);
2714
2715                         if (sc->sc_brtcnt - dead <= sc->sc_brtmax) {
2716                                 /* Enough rtnodes are collected */
2717                                 break;
2718                         }
2719                 }
2720         }
2721         if (dead)
2722                 bridge_rtreap(sc);
2723 }
2724
2725 /*
2726  * bridge_timer:
2727  *
2728  *      Aging timer for the bridge.
2729  */
2730 static void
2731 bridge_timer(void *arg)
2732 {
2733         struct bridge_softc *sc = arg;
2734         struct lwkt_msg *lmsg;
2735
2736         KKASSERT(mycpuid == BRIDGE_CFGCPU);
2737
2738         crit_enter();
2739
2740         if (callout_pending(&sc->sc_brcallout) ||
2741             !callout_active(&sc->sc_brcallout)) {
2742                 crit_exit();
2743                 return;
2744         }
2745         callout_deactivate(&sc->sc_brcallout);
2746
2747         lmsg = &sc->sc_brtimemsg.nm_lmsg;
2748         KKASSERT(lmsg->ms_flags & MSGF_DONE);
2749         lwkt_sendmsg(BRIDGE_CFGPORT, lmsg);
2750
2751         crit_exit();
2752 }
2753
2754 static void
2755 bridge_timer_handler(struct netmsg *nmsg)
2756 {
2757         struct bridge_softc *sc = nmsg->nm_lmsg.u.ms_resultp;
2758
2759         KKASSERT(&curthread->td_msgport == BRIDGE_CFGPORT);
2760
2761         crit_enter();
2762         /* Reply ASAP */
2763         lwkt_replymsg(&nmsg->nm_lmsg, 0);
2764         crit_exit();
2765
2766         bridge_rtage(sc);
2767         if (sc->sc_ifp->if_flags & IFF_RUNNING) {
2768                 callout_reset(&sc->sc_brcallout,
2769                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2770         }
2771 }
2772
2773 static int
2774 bridge_rtage_finddead(struct bridge_softc *sc)
2775 {
2776         struct bridge_rtnode *brt;
2777         int dead = 0;
2778
2779         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2780                 struct bridge_rtinfo *bri = brt->brt_info;
2781
2782                 if ((bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2783                     time_second >= bri->bri_expire) {
2784                         bri->bri_dead = 1;
2785                         ++dead;
2786                         KKASSERT(dead <= sc->sc_brtcnt);
2787                 }
2788         }
2789         return dead;
2790 }
2791
2792 /*
2793  * bridge_rtage:
2794  *
2795  *      Perform an aging cycle.
2796  */
2797 static void
2798 bridge_rtage(struct bridge_softc *sc)
2799 {
2800         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2801
2802         if (bridge_rtage_finddead(sc))
2803                 bridge_rtreap(sc);
2804 }
2805
2806 /*
2807  * bridge_rtflush:
2808  *
2809  *      Remove all dynamic addresses from the bridge.
2810  */
2811 static void
2812 bridge_rtflush(struct bridge_softc *sc, int full)
2813 {
2814         struct bridge_rtnode *brt;
2815         int reap;
2816
2817         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2818
2819         reap = 0;
2820         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2821                 struct bridge_rtinfo *bri = brt->brt_info;
2822
2823                 if (full ||
2824                     (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2825                         bri->bri_dead = 1;
2826                         reap = 1;
2827                 }
2828         }
2829         if (reap)
2830                 bridge_rtreap(sc);
2831 }
2832
2833 /*
2834  * bridge_rtdaddr:
2835  *
2836  *      Remove an address from the table.
2837  */
2838 static int
2839 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2840 {
2841         struct bridge_rtnode *brt;
2842
2843         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2844
2845         if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2846                 return (ENOENT);
2847
2848         /* TODO: add a cheaper delete operation */
2849         brt->brt_info->bri_dead = 1;
2850         bridge_rtreap(sc);
2851         return (0);
2852 }
2853
2854 /*
2855  * bridge_rtdelete:
2856  *
2857  *      Delete routes to a speicifc member interface.
2858  */
2859 void
2860 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2861 {
2862         struct bridge_rtnode *brt;
2863         int reap;
2864
2865         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
2866
2867         reap = 0;
2868         LIST_FOREACH(brt, &sc->sc_rtlists[mycpuid], brt_list) {
2869                 struct bridge_rtinfo *bri = brt->brt_info;
2870
2871                 if (bri->bri_ifp == ifp &&
2872                     (full ||
2873                      (bri->bri_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)) {
2874                         bri->bri_dead = 1;
2875                         reap = 1;
2876                 }
2877         }
2878         if (reap)
2879                 bridge_rtreap(sc);
2880 }
2881
2882 /*
2883  * bridge_rtable_init:
2884  *
2885  *      Initialize the route table for this bridge.
2886  */
2887 static void
2888 bridge_rtable_init(struct bridge_softc *sc)
2889 {
2890         int cpu;
2891
2892         /*
2893          * Initialize per-cpu hash tables
2894          */
2895         sc->sc_rthashs = kmalloc(sizeof(*sc->sc_rthashs) * ncpus,
2896                                  M_DEVBUF, M_WAITOK);
2897         for (cpu = 0; cpu < ncpus; ++cpu) {
2898                 int i;
2899
2900                 sc->sc_rthashs[cpu] =
2901                 kmalloc(sizeof(struct bridge_rtnode_head) * BRIDGE_RTHASH_SIZE,
2902                         M_DEVBUF, M_WAITOK);
2903
2904                 for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2905                         LIST_INIT(&sc->sc_rthashs[cpu][i]);
2906         }
2907         sc->sc_rthash_key = karc4random();
2908
2909         /*
2910          * Initialize per-cpu lists
2911          */
2912         sc->sc_rtlists = kmalloc(sizeof(struct bridge_rtnode_head) * ncpus,
2913                                  M_DEVBUF, M_WAITOK);
2914         for (cpu = 0; cpu < ncpus; ++cpu)
2915                 LIST_INIT(&sc->sc_rtlists[cpu]);
2916 }
2917
2918 /*
2919  * bridge_rtable_fini:
2920  *
2921  *      Deconstruct the route table for this bridge.
2922  */
2923 static void
2924 bridge_rtable_fini(struct bridge_softc *sc)
2925 {
2926         int cpu;
2927
2928         /*
2929          * Free per-cpu hash tables
2930          */
2931         for (cpu = 0; cpu < ncpus; ++cpu)
2932                 kfree(sc->sc_rthashs[cpu], M_DEVBUF);
2933         kfree(sc->sc_rthashs, M_DEVBUF);
2934
2935         /*
2936          * Free per-cpu lists
2937          */
2938         kfree(sc->sc_rtlists, M_DEVBUF);
2939 }
2940
2941 /*
2942  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2943  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2944  */
2945 #define mix(a, b, c)                                                    \
2946 do {                                                                    \
2947         a -= b; a -= c; a ^= (c >> 13);                                 \
2948         b -= c; b -= a; b ^= (a << 8);                                  \
2949         c -= a; c -= b; c ^= (b >> 13);                                 \
2950         a -= b; a -= c; a ^= (c >> 12);                                 \
2951         b -= c; b -= a; b ^= (a << 16);                                 \
2952         c -= a; c -= b; c ^= (b >> 5);                                  \
2953         a -= b; a -= c; a ^= (c >> 3);                                  \
2954         b -= c; b -= a; b ^= (a << 10);                                 \
2955         c -= a; c -= b; c ^= (b >> 15);                                 \
2956 } while (/*CONSTCOND*/0)
2957
2958 static __inline uint32_t
2959 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2960 {
2961         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2962
2963         b += addr[5] << 8;
2964         b += addr[4];
2965         a += addr[3] << 24;
2966         a += addr[2] << 16;
2967         a += addr[1] << 8;
2968         a += addr[0];
2969
2970         mix(a, b, c);
2971
2972         return (c & BRIDGE_RTHASH_MASK);
2973 }
2974
2975 #undef mix
2976
2977 static int
2978 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2979 {
2980         int i, d;
2981
2982         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2983                 d = ((int)a[i]) - ((int)b[i]);
2984         }
2985
2986         return (d);
2987 }
2988
2989 /*
2990  * bridge_rtnode_lookup:
2991  *
2992  *      Look up a bridge route node for the specified destination.
2993  */
2994 static struct bridge_rtnode *
2995 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2996 {
2997         struct bridge_rtnode *brt;
2998         uint32_t hash;
2999         int dir;
3000
3001         hash = bridge_rthash(sc, addr);
3002         LIST_FOREACH(brt, &sc->sc_rthashs[mycpuid][hash], brt_hash) {
3003                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3004                 if (dir == 0)
3005                         return (brt);
3006                 if (dir > 0)
3007                         return (NULL);
3008         }
3009
3010         return (NULL);
3011 }
3012
3013 /*
3014  * bridge_rtnode_insert:
3015  *
3016  *      Insert the specified bridge node into the route table.
3017  *      Caller has to make sure that rtnode does not exist.
3018  */
3019 static void
3020 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3021 {
3022         struct bridge_rtnode *lbrt;
3023         uint32_t hash;
3024         int dir;
3025
3026         hash = bridge_rthash(sc, brt->brt_addr);
3027
3028         lbrt = LIST_FIRST(&sc->sc_rthashs[mycpuid][hash]);
3029         if (lbrt == NULL) {
3030                 LIST_INSERT_HEAD(&sc->sc_rthashs[mycpuid][hash], brt, brt_hash);
3031                 goto out;
3032         }
3033
3034         do {
3035                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3036                 KASSERT(dir != 0, ("rtnode already exist\n"));
3037
3038                 if (dir > 0) {
3039                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3040                         goto out;
3041                 }
3042                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
3043                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3044                         goto out;
3045                 }
3046                 lbrt = LIST_NEXT(lbrt, brt_hash);
3047         } while (lbrt != NULL);
3048
3049         panic("no suitable position found for rtnode\n");
3050 out:
3051         LIST_INSERT_HEAD(&sc->sc_rtlists[mycpuid], brt, brt_list);
3052         if (mycpuid == 0) {
3053                 /*
3054                  * Update the brtcnt.
3055                  * We only need to do it once and we do it on CPU0.
3056                  */
3057                 sc->sc_brtcnt++;
3058         }
3059 }
3060
3061 /*
3062  * bridge_rtnode_destroy:
3063  *
3064  *      Destroy a bridge rtnode.
3065  */
3066 static void
3067 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3068 {
3069         LIST_REMOVE(brt, brt_hash);
3070         LIST_REMOVE(brt, brt_list);
3071
3072         if (mycpuid + 1 == ncpus) {
3073                 /* Free rtinfo associated with rtnode on the last cpu */
3074                 kfree(brt->brt_info, M_DEVBUF);
3075         }
3076         kfree(brt, M_DEVBUF);
3077
3078         if (mycpuid == 0) {
3079                 /* Update brtcnt only on CPU0 */
3080                 sc->sc_brtcnt--;
3081         }
3082 }
3083
3084 static __inline int
3085 bridge_post_pfil(struct mbuf *m)
3086 {
3087         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED)
3088                 return EOPNOTSUPP;
3089
3090         /* Not yet */
3091         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED)
3092                 return EOPNOTSUPP;
3093
3094         return 0;
3095 }
3096
3097 /*
3098  * Send bridge packets through pfil if they are one of the types pfil can deal
3099  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3100  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3101  * that interface.
3102  */
3103 static int
3104 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3105 {
3106         int snap, error, i, hlen;
3107         struct ether_header *eh1, eh2;
3108         struct ip *ip;
3109         struct llc llc1;
3110         u_int16_t ether_type;
3111
3112         snap = 0;
3113         error = -1;     /* Default error if not error == 0 */
3114
3115         if (pfil_bridge == 0 && pfil_member == 0)
3116                 return (0); /* filtering is disabled */
3117
3118         i = min((*mp)->m_pkthdr.len, max_protohdr);
3119         if ((*mp)->m_len < i) {
3120                 *mp = m_pullup(*mp, i);
3121                 if (*mp == NULL) {
3122                         kprintf("%s: m_pullup failed\n", __func__);
3123                         return (-1);
3124                 }
3125         }
3126
3127         eh1 = mtod(*mp, struct ether_header *);
3128         ether_type = ntohs(eh1->ether_type);
3129
3130         /*
3131          * Check for SNAP/LLC.
3132          */
3133         if (ether_type < ETHERMTU) {
3134                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3135
3136                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3137                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3138                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3139                     llc2->llc_control == LLC_UI) {
3140                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3141                         snap = 1;
3142                 }
3143         }
3144
3145         /*
3146          * If we're trying to filter bridge traffic, don't look at anything
3147          * other than IP and ARP traffic.  If the filter doesn't understand
3148          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3149          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3150          * but of course we don't have an AppleTalk filter to begin with.
3151          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3152          * ARP traffic.)
3153          */
3154         switch (ether_type) {
3155         case ETHERTYPE_ARP:
3156         case ETHERTYPE_REVARP:
3157                 return (0); /* Automatically pass */
3158
3159         case ETHERTYPE_IP:
3160 #ifdef INET6
3161         case ETHERTYPE_IPV6:
3162 #endif /* INET6 */
3163                 break;
3164
3165         default:
3166                 /*
3167                  * Check to see if the user wants to pass non-ip
3168                  * packets, these will not be checked by pfil(9)
3169                  * and passed unconditionally so the default is to drop.
3170                  */
3171                 if (pfil_onlyip)
3172                         goto bad;
3173         }
3174
3175         /* Strip off the Ethernet header and keep a copy. */
3176         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3177         m_adj(*mp, ETHER_HDR_LEN);
3178
3179         /* Strip off snap header, if present */
3180         if (snap) {
3181                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3182                 m_adj(*mp, sizeof(struct llc));
3183         }
3184
3185         /*
3186          * Check the IP header for alignment and errors
3187          */
3188         if (dir == PFIL_IN) {
3189                 switch (ether_type) {
3190                 case ETHERTYPE_IP:
3191                         error = bridge_ip_checkbasic(mp);
3192                         break;
3193 #ifdef INET6
3194                 case ETHERTYPE_IPV6:
3195                         error = bridge_ip6_checkbasic(mp);
3196                         break;
3197 #endif /* INET6 */
3198                 default:
3199                         error = 0;
3200                 }
3201                 if (error)
3202                         goto bad;
3203         }
3204
3205         error = 0;
3206
3207         /*
3208          * Run the packet through pfil
3209          */
3210         switch (ether_type) {
3211         case ETHERTYPE_IP:
3212                 /*
3213                  * before calling the firewall, swap fields the same as
3214                  * IP does. here we assume the header is contiguous
3215                  */
3216                 ip = mtod(*mp, struct ip *);
3217
3218                 ip->ip_len = ntohs(ip->ip_len);
3219                 ip->ip_off = ntohs(ip->ip_off);
3220
3221                 /*
3222                  * Run pfil on the member interface and the bridge, both can
3223                  * be skipped by clearing pfil_member or pfil_bridge.
3224                  *
3225                  * Keep the order:
3226                  *   in_if -> bridge_if -> out_if
3227                  */
3228                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL) {
3229                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3230                         if (*mp == NULL || error != 0) /* filter may consume */
3231                                 break;
3232                         error = bridge_post_pfil(*mp);
3233                         if (error)
3234                                 break;
3235                 }
3236
3237                 if (pfil_member && ifp != NULL) {
3238                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp, dir);
3239                         if (*mp == NULL || error != 0) /* filter may consume */
3240                                 break;
3241                         error = bridge_post_pfil(*mp);
3242                         if (error)
3243                                 break;
3244                 }
3245
3246                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL) {
3247                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp, dir);
3248                         if (*mp == NULL || error != 0) /* filter may consume */
3249                                 break;
3250                         error = bridge_post_pfil(*mp);
3251                         if (error)
3252                                 break;
3253                 }
3254
3255                 /* check if we need to fragment the packet */
3256                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3257                         i = (*mp)->m_pkthdr.len;
3258                         if (i > ifp->if_mtu) {
3259                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3260                                             &llc1);
3261                                 return (error);
3262                         }
3263                 }
3264
3265                 /* Recalculate the ip checksum and restore byte ordering */
3266                 ip = mtod(*mp, struct ip *);
3267                 hlen = ip->ip_hl << 2;
3268                 if (hlen < sizeof(struct ip))
3269                         goto bad;
3270                 if (hlen > (*mp)->m_len) {
3271                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3272                                 goto bad;
3273                         ip = mtod(*mp, struct ip *);
3274                         if (ip == NULL)
3275                                 goto bad;
3276                 }
3277                 ip->ip_len = htons(ip->ip_len);
3278                 ip->ip_off = htons(ip->ip_off);
3279                 ip->ip_sum = 0;
3280                 if (hlen == sizeof(struct ip))
3281                         ip->ip_sum = in_cksum_hdr(ip);
3282                 else
3283                         ip->ip_sum = in_cksum(*mp, hlen);
3284
3285                 break;
3286 #ifdef INET6
3287         case ETHERTYPE_IPV6:
3288                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3289                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3290                                         dir);
3291
3292                 if (*mp == NULL || error != 0) /* filter may consume */
3293                         break;
3294
3295                 if (pfil_member && ifp != NULL)
3296                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3297                                         dir);
3298
3299                 if (*mp == NULL || error != 0) /* filter may consume */
3300                         break;
3301
3302                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3303                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3304                                         dir);
3305                 break;
3306 #endif
3307         default:
3308                 error = 0;
3309                 break;
3310         }
3311
3312         if (*mp == NULL)
3313                 return (error);
3314         if (error != 0)
3315                 goto bad;
3316
3317         error = -1;
3318
3319         /*
3320          * Finally, put everything back the way it was and return
3321          */
3322         if (snap) {
3323                 M_PREPEND(*mp, sizeof(struct llc), MB_DONTWAIT);
3324                 if (*mp == NULL)
3325                         return (error);
3326                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3327         }
3328
3329         M_PREPEND(*mp, ETHER_HDR_LEN, MB_DONTWAIT);
3330         if (*mp == NULL)
3331                 return (error);
3332         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3333
3334         return (0);
3335
3336 bad:
3337         m_freem(*mp);
3338         *mp = NULL;
3339         return (error);
3340 }
3341
3342 /*
3343  * Perform basic checks on header size since
3344  * pfil assumes ip_input has already processed
3345  * it for it.  Cut-and-pasted from ip_input.c.
3346  * Given how simple the IPv6 version is,
3347  * does the IPv4 version really need to be
3348  * this complicated?
3349  *
3350  * XXX Should we update ipstat here, or not?
3351  * XXX Right now we update ipstat but not
3352  * XXX csum_counter.
3353  */
3354 static int
3355 bridge_ip_checkbasic(struct mbuf **mp)
3356 {
3357         struct mbuf *m = *mp;
3358         struct ip *ip;
3359         int len, hlen;
3360         u_short sum;
3361
3362         if (*mp == NULL)
3363                 return (-1);
3364 #if notyet
3365         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3366                 if ((m = m_copyup(m, sizeof(struct ip),
3367                         (max_linkhdr + 3) & ~3)) == NULL) {
3368                         /* XXXJRT new stat, please */
3369                         ipstat.ips_toosmall++;
3370                         goto bad;
3371                 }
3372         } else
3373 #endif
3374 #ifndef __predict_false
3375 #define __predict_false(x) x
3376 #endif
3377          if (__predict_false(m->m_len < sizeof (struct ip))) {
3378                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3379                         ipstat.ips_toosmall++;
3380                         goto bad;
3381                 }
3382         }
3383         ip = mtod(m, struct ip *);
3384         if (ip == NULL) goto bad;
3385
3386         if (ip->ip_v != IPVERSION) {
3387                 ipstat.ips_badvers++;
3388                 goto bad;
3389         }
3390         hlen = ip->ip_hl << 2;
3391         if (hlen < sizeof(struct ip)) { /* minimum header length */
3392                 ipstat.ips_badhlen++;
3393                 goto bad;
3394         }
3395         if (hlen > m->m_len) {
3396                 if ((m = m_pullup(m, hlen)) == 0) {
3397                         ipstat.ips_badhlen++;
3398                         goto bad;
3399                 }
3400                 ip = mtod(m, struct ip *);
3401                 if (ip == NULL) goto bad;
3402         }
3403
3404         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3405                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3406         } else {
3407                 if (hlen == sizeof(struct ip)) {
3408                         sum = in_cksum_hdr(ip);
3409                 } else {
3410                         sum = in_cksum(m, hlen);
3411                 }
3412         }
3413         if (sum) {
3414                 ipstat.ips_badsum++;
3415                 goto bad;
3416         }
3417
3418         /* Retrieve the packet length. */
3419         len = ntohs(ip->ip_len);
3420
3421         /*
3422          * Check for additional length bogosity
3423          */
3424         if (len < hlen) {
3425                 ipstat.ips_badlen++;
3426                 goto bad;
3427         }
3428
3429         /*
3430          * Check that the amount of data in the buffers
3431          * is as at least much as the IP header would have us expect.
3432          * Drop packet if shorter than we expect.
3433          */
3434         if (m->m_pkthdr.len < len) {
3435                 ipstat.ips_tooshort++;
3436                 goto bad;
3437         }
3438
3439         /* Checks out, proceed */
3440         *mp = m;
3441         return (0);
3442
3443 bad:
3444         *mp = m;
3445         return (-1);
3446 }
3447
3448 #ifdef INET6
3449 /*
3450  * Same as above, but for IPv6.
3451  * Cut-and-pasted from ip6_input.c.
3452  * XXX Should we update ip6stat, or not?
3453  */
3454 static int
3455 bridge_ip6_checkbasic(struct mbuf **mp)
3456 {
3457         struct mbuf *m = *mp;
3458         struct ip6_hdr *ip6;
3459
3460         /*
3461          * If the IPv6 header is not aligned, slurp it up into a new
3462          * mbuf with space for link headers, in the event we forward
3463          * it.  Otherwise, if it is aligned, make sure the entire base
3464          * IPv6 header is in the first mbuf of the chain.
3465          */
3466 #if notyet
3467         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3468                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3469                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3470                             (max_linkhdr + 3) & ~3)) == NULL) {
3471                         /* XXXJRT new stat, please */
3472                         ip6stat.ip6s_toosmall++;
3473                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3474                         goto bad;
3475                 }
3476         } else
3477 #endif
3478         if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3479                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3480                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3481                         ip6stat.ip6s_toosmall++;
3482                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3483                         goto bad;
3484                 }
3485         }
3486
3487         ip6 = mtod(m, struct ip6_hdr *);
3488
3489         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3490                 ip6stat.ip6s_badvers++;
3491                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3492                 goto bad;
3493         }
3494
3495         /* Checks out, proceed */
3496         *mp = m;
3497         return (0);
3498
3499 bad:
3500         *mp = m;
3501         return (-1);
3502 }
3503 #endif /* INET6 */
3504
3505 /*
3506  * bridge_fragment:
3507  *
3508  *      Return a fragmented mbuf chain.
3509  */
3510 static int
3511 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3512     int snap, struct llc *llc)
3513 {
3514         struct mbuf *m0;
3515         struct ip *ip;
3516         int error = -1;
3517
3518         if (m->m_len < sizeof(struct ip) &&
3519             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3520                 goto out;
3521         ip = mtod(m, struct ip *);
3522
3523         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3524                     CSUM_DELAY_IP);
3525         if (error)
3526                 goto out;
3527
3528         /* walk the chain and re-add the Ethernet header */
3529         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3530                 if (error == 0) {
3531                         if (snap) {
3532                                 M_PREPEND(m0, sizeof(struct llc), MB_DONTWAIT);
3533                                 if (m0 == NULL) {
3534                                         error = ENOBUFS;
3535                                         continue;
3536                                 }
3537                                 bcopy(llc, mtod(m0, caddr_t),
3538                                     sizeof(struct llc));
3539                         }
3540                         M_PREPEND(m0, ETHER_HDR_LEN, MB_DONTWAIT);
3541                         if (m0 == NULL) {
3542                                 error = ENOBUFS;
3543                                 continue;
3544                         }
3545                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3546                 } else
3547                         m_freem(m);
3548         }
3549
3550         if (error == 0)
3551                 ipstat.ips_fragmented++;
3552
3553         return (error);
3554
3555 out:
3556         if (m != NULL)
3557                 m_freem(m);
3558         return (error);
3559 }
3560
3561 static void
3562 bridge_enqueue_handler(struct netmsg *nmsg)
3563 {
3564         struct netmsg_packet *nmp;
3565         struct ifnet *dst_ifp;
3566         struct mbuf *m;
3567
3568         nmp = (struct netmsg_packet *)nmsg;
3569         m = nmp->nm_packet;
3570         dst_ifp = nmp->nm_netmsg.nm_lmsg.u.ms_resultp;
3571
3572         bridge_handoff(dst_ifp, m);
3573 }
3574
3575 static void
3576 bridge_pfil_enqueue_handler(struct netmsg *nmsg)
3577 {
3578         struct netmsg_packet *nmp;
3579         struct ifnet *dst_ifp;
3580         struct mbuf *m;
3581
3582         nmp = (struct netmsg_packet *)nmsg;
3583         m = nmp->nm_packet;
3584         dst_ifp = nmp->nm_netmsg.nm_lmsg.u.ms_resultp;
3585
3586         /*
3587          * Filter on the output interface. Pass a NULL bridge interface
3588          * pointer so we do not redundantly filter on the bridge for
3589          * each interface we broadcast on.
3590          */
3591         if (inet_pfil_hook.ph_hashooks > 0
3592 #ifdef INET6
3593             || inet6_pfil_hook.ph_hashooks > 0
3594 #endif
3595             ) {
3596                 if (bridge_pfil(&m, NULL, dst_ifp, PFIL_OUT) != 0)
3597                         return;
3598                 if (m == NULL)
3599                         return;
3600         }
3601         bridge_handoff(dst_ifp, m);
3602 }
3603
3604 static void
3605 bridge_handoff(struct ifnet *dst_ifp, struct mbuf *m)
3606 {
3607         struct mbuf *m0;
3608
3609         lwkt_serialize_enter(dst_ifp->if_serializer);
3610
3611         /* We may be sending a fragment so traverse the mbuf */
3612         for (; m; m = m0) {
3613                 struct altq_pktattr pktattr;
3614
3615                 m0 = m->m_nextpkt;
3616                 m->m_nextpkt = NULL;
3617
3618                 if (ifq_is_enabled(&dst_ifp->if_snd))
3619                         altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
3620
3621                 ifq_handoff(dst_ifp, m, &pktattr);
3622         }
3623
3624         lwkt_serialize_exit(dst_ifp->if_serializer);
3625 }
3626
3627 static void
3628 bridge_control_dispatch(struct netmsg *nmsg)
3629 {
3630         struct netmsg_brctl *bc_msg = (struct netmsg_brctl *)nmsg;
3631         struct ifnet *bifp = bc_msg->bc_sc->sc_ifp;
3632         int error;
3633
3634         lwkt_serialize_enter(bifp->if_serializer);
3635         error = bc_msg->bc_func(bc_msg->bc_sc, bc_msg->bc_arg);
3636         lwkt_serialize_exit(bifp->if_serializer);
3637
3638         lwkt_replymsg(&nmsg->nm_lmsg, error);
3639 }
3640
3641 static int
3642 bridge_control(struct bridge_softc *sc, u_long cmd,
3643                bridge_ctl_t bc_func, void *bc_arg)
3644 {
3645         struct ifnet *bifp = sc->sc_ifp;
3646         struct netmsg_brctl bc_msg;
3647         struct netmsg *nmsg;
3648         int error;
3649
3650         ASSERT_SERIALIZED(bifp->if_serializer);
3651
3652         bzero(&bc_msg, sizeof(bc_msg));
3653         nmsg = &bc_msg.bc_nmsg;
3654
3655         netmsg_init(nmsg, &curthread->td_msgport, 0, bridge_control_dispatch);
3656         bc_msg.bc_func = bc_func;
3657         bc_msg.bc_sc = sc;
3658         bc_msg.bc_arg = bc_arg;
3659
3660         lwkt_serialize_exit(bifp->if_serializer);
3661         error = lwkt_domsg(BRIDGE_CFGPORT, &nmsg->nm_lmsg, 0);
3662         lwkt_serialize_enter(bifp->if_serializer);
3663         return error;
3664 }
3665
3666 static void
3667 bridge_add_bif_handler(struct netmsg *nmsg)
3668 {
3669         struct netmsg_braddbif *amsg = (struct netmsg_braddbif *)nmsg;
3670         struct bridge_softc *sc;
3671         struct bridge_iflist *bif;
3672
3673         sc = amsg->br_softc;
3674
3675         bif = kmalloc(sizeof(*bif), M_DEVBUF, M_WAITOK | M_ZERO);
3676         bif->bif_ifp = amsg->br_bif_ifp;
3677         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
3678         bif->bif_onlist = 1;
3679         bif->bif_info = amsg->br_bif_info;
3680
3681         LIST_INSERT_HEAD(&sc->sc_iflists[mycpuid], bif, bif_next);
3682
3683         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
3684 }
3685
3686 static void
3687 bridge_add_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3688                struct ifnet *ifp)
3689 {
3690         struct netmsg_braddbif amsg;
3691
3692         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
3693
3694         netmsg_init(&amsg.br_nmsg, &curthread->td_msgport, 0,
3695                     bridge_add_bif_handler);
3696         amsg.br_softc = sc;
3697         amsg.br_bif_info = bif_info;
3698         amsg.br_bif_ifp = ifp;
3699
3700         ifnet_domsg(&amsg.br_nmsg.nm_lmsg, 0);
3701 }
3702
3703 static void
3704 bridge_del_bif_handler(struct netmsg *nmsg)
3705 {
3706         struct netmsg_brdelbif *dmsg = (struct netmsg_brdelbif *)nmsg;
3707         struct bridge_softc *sc;
3708         struct bridge_iflist *bif;
3709
3710         sc = dmsg->br_softc;
3711
3712         /*
3713          * Locate the bif associated with the br_bif_info
3714          * on the current CPU
3715          */
3716         bif = bridge_lookup_member_ifinfo(sc, dmsg->br_bif_info);
3717         KKASSERT(bif != NULL && bif->bif_onlist);
3718
3719         /* Remove the bif from the current CPU's iflist */
3720         bif->bif_onlist = 0;
3721         LIST_REMOVE(bif, bif_next);
3722
3723         /* Save the removed bif for later freeing */
3724         LIST_INSERT_HEAD(dmsg->br_bif_list, bif, bif_next);
3725
3726         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
3727 }
3728
3729 static void
3730 bridge_del_bif(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3731                struct bridge_iflist_head *saved_bifs)
3732 {
3733         struct netmsg_brdelbif dmsg;
3734
3735         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
3736
3737         netmsg_init(&dmsg.br_nmsg, &curthread->td_msgport, 0,
3738                     bridge_del_bif_handler);
3739         dmsg.br_softc = sc;
3740         dmsg.br_bif_info = bif_info;
3741         dmsg.br_bif_list = saved_bifs;
3742
3743         ifnet_domsg(&dmsg.br_nmsg.nm_lmsg, 0);
3744 }
3745
3746 static void
3747 bridge_set_bifflags_handler(struct netmsg *nmsg)
3748 {
3749         struct netmsg_brsflags *smsg = (struct netmsg_brsflags *)nmsg;
3750         struct bridge_softc *sc;
3751         struct bridge_iflist *bif;
3752
3753         sc = smsg->br_softc;
3754
3755         /*
3756          * Locate the bif associated with the br_bif_info
3757          * on the current CPU
3758          */
3759         bif = bridge_lookup_member_ifinfo(sc, smsg->br_bif_info);
3760         KKASSERT(bif != NULL && bif->bif_onlist);
3761
3762         bif->bif_flags = smsg->br_bif_flags;
3763
3764         ifnet_forwardmsg(&nmsg->nm_lmsg, mycpuid + 1);
3765 }
3766
3767 static void
3768 bridge_set_bifflags(struct bridge_softc *sc, struct bridge_ifinfo *bif_info,
3769                     uint32_t bif_flags)
3770 {
3771         struct netmsg_brsflags smsg;
3772
3773         ASSERT_NOT_SERIALIZED(sc->sc_ifp->if_serializer);
3774
3775         netmsg_init(&smsg.br_nmsg, &curthread->td_msgport, 0,
3776                     bridge_set_bifflags_handler);
3777         smsg.br_softc = sc;
3778         smsg.br_bif_info = bif_info;
3779         smsg.br_bif_flags = bif_flags;
3780
3781         ifnet_domsg(&smsg.br_nmsg.nm_lmsg, 0);
3782 }